Stephen Marlett

Summer Institute of Linguistics

and University of North Dakota
Fall 2001 Edition
This is a working, pre-publication draft. Please do not quote.
Please do not duplicate without written permission.
Revisions (including addition of new exercises and corrections) are made yearly.
Please inquire about the latest version. Earlier versions are no longer availabl
e.
[email protected]
Copyright 2001 by Stephen A. Marlett
A N
I N T R O D U C T I O N
T O
PHONOLOGICAL ANALYSIS
Table of Contents
Preface ........................................................................
............................................................ iii
Chapter 1 Introduction..........................................................
......................................................
1
Section 1
Morphological Rules
Chapter 2 Word Structure ......................................................
....................................................
7
Chapter 3 Suppletive Allomorphy.................................................
............................................. 12
Chapter 4 Multiple Function Formatives .........................................
.......................................... 29
Chapter 5 Morphologically Triggered Rules .....................................
........................................ 35
Chapter 6 Features and Natural Classes .........................................
............................................ 40
Chapter 7 Reduplication ........................................................
..................................................... 44
Chapter 8 Word or Affix?........................................................
................................................... 46
Summary and Review Questions for Section 1......................................
.............................................. 51
Section 2
Phonological Rules: Assimilation
Chapter 9 Voicing Assimilation .................................................
................................................ 55
Chapter 10 Choosing Underlying Forms............................................
.......................................... 61
Chapter 11 Place Assimilation (Nasals) .........................................
.............................................. 64
Chapter 12 Features in the Lexicon..............................................
................................................ 71
Chapter 13 Feature Spreading ...................................................
................................................... 74
Chapter 14 Constraining Rule Application........................................
........................................... 79
Chapter 15 Palatalization and Labialization.....................................
............................................ 83
Chapter 16 Nasalization.........................................................
....................................................... 88
Chapter 17 Manner Assimilation..................................................
................................................ 90
Chapter 18 Vowel Changes .......................................................
................................................... 93
Chapter 19 Place Assimilation (Non-nasals) .....................................
........................................... 99
Chapter 20 Dissimilation and the Obligatory Contour Principle ..................

............................... 104
Chapter 21 Miscellaneous........................................................
..................................................... 107
Summary and Review Questions for Section 2......................................
.............................................. 109
Section 3
Phonological Rules: Some Practical Procedures
Chapter 22 Contrastive Features.................................................
.................................................. 113
Chapter 23 Noncontrastive Features..............................................
............................................... 121
Chapter 24 Suspicious Pairs ....................................................
..................................................... 134
Summary and Review Questions for Section 3 .....................................
.............................................. 136
Section 4
Phonological Rules: Structural Issues
Chapter 25 Underspecification ..................................................
................................................... 139
Chapter 26 Edge Phenomena.......................................................
................................................. 145
Chapter 27 Syllable Structure...................................................
.................................................... 151
Chapter 28 Syllable Structure Constraints.......................................
............................................. 156
Chapter 29 Linking Features to the Syllable ....................................
............................................ 160
Chapter 30 Stress-conditioned Processes ........................................
............................................. 168
Chapter 31 Epenthesis ..........................................................
........................................................ 170
Chapter 32 Deletion.............................................................
......................................................... 176
Chapter 33 Underlying Forms ....................................................
.................................................. 182
Chapter 34 Rule Ordering........................................................
..................................................... 188
Summary and Review Questions for Section 4......................................
..............................................197
Section 5
Phonological Rules: Suprasegmental Properties
Chapter 35 Stress ..............................................................
...........................................................200
Chapter 36 Introduction to Pitch................................................
...................................................204
Chapter 37 Intonation...........................................................
.........................................................206
Chapter 38 Tone ................................................................
.........................................................212
Chapter 39 Tone Rules...........................................................
.......................................................214
Summary and Review Questions for Section 5......................................
..............................................222
Appendices
Appendix A Features ............................................................
....................................................224
Appendix B Orthography Design ..................................................
...........................................231
Appendix C Phonology Write-ups ................................................
...........................................247
Appendix D Symbols Tables ......................................................
..............................................251
Appendix E Language Index and Source Information .............................

...............................253
Appendix F Open-ended Exercises ...............................................
...........................................261
Appendix G Glossary.............................................................
...................................................282
Appendix H Topic Index..........................................................
.................................................287
iii
Preface
Phonology is a broad topic of study and currently comprises many theories, each
of which requires (at least)
a course and a book for adequate treatment. This book is an introduc
tion to phonology in general, and a
very brief introduction to the ideas addressed by various of these theories, inc
luding generative phonology,
lexical phonology, underspecification theory, autosegmental phonology, feature t
heory, phonemics, and CV
phonology. We believe that each of these has contributed in a signifi
cant way to our understanding of
language, regardless of how the theories themselves may fare in the future.
We expect that some users of this book may not find these theories inherently in
teresting. Nevertheless,
someone who wishes to use linguistic theory for practical problems (such as lang
uage learning, orthography
development, literacy programs) will benefit from learning more about h
ow languages work. Thus we
encourage all to jump in and try to master each topic.
The presentation of material in this book is different from many othe
r books. Most importantly for
those who are studying phonology in order to do field work, it does
not present theory in the same order
that one might apply theory. We know that not everyone approaches pro
blems in the same way or in the
same order. The book does not begin with the same kind of facts that usually fir
st confront a linguist in a
field situation. Nevertheless, after one has completed reading the book, one sho
uld have a clear idea of how
to work in such situations.
One of the major reasons for presenting material in the order chosen
is that it allows us to present
phonological detail in small steps. Many introductory books begin with
phonetics and then lead the user
through the morass of detail to graphically simpler levels of representation. Bu
t since many students are not
well-trained in phonetics at the time when they take their first phon
ology courseor at least could use
some reviewwe have opted to teach simple but not necessarily phonetic processes e
arly in the course and
postpone discussions of phonetic detail until after the general princip
les of phonological description and
analysis are understood. Of course, this means that careful attention
must be paid to helping the students
apply the principles that are taught.
The various interactive exercises included in the book refer to the e
nd of each chapter where a
suggested answer is provided (called feedback). The suggested answer is not meant
to stifle all alternative
solutions, but is intended as an additional teaching aid to the user.
This edition of the book incorporates a section called Postscript for
Teachers, which provides

additional background and explanation that instructors may find helpful.

We thank the help of phonology teachers, teaching assistants, and stud
ents at SIL North Dakota and
others for their input over the years. These include Gayle Aasen, Anita Bickford
, Steve Clark, Mark Datson,
Margie Doty, Roy Eberhardt, Mary Huttar, Andreas Joswig, Mike Maxwell,
Jim Meyer, Steve Parker,
Steve Quakenbush, Jim Robertson, Amy Schondelmeyer, Doug Trick, Cathy Marlett, a
nd Stephen Walker.
All are absolved of responsibility for deficiencies that remain.
Corrections and suggestions for further improvements are welcome and ma
y be directed to: Steve
Marlett, PO Box 8987, Catalina, Arizona, 85738-0987; or e-mail: steve_marlett@s
il.org.
1
Chapter 1 - Introduction
Human beings have an extraordinary and unique communicative ability. Wi
th a limited set of sounds, a
speaker of any language can express an infinite number of sentences. This abilit
y is extremely complex; in
reality, we are still unable to characterize its nature with any degree of preci
sion.
This book is an introduction to one facet of human communication: phonology, the
study of sounds and
how they are organized and used in natural languages. Before jumping
into the study of phonology,
however, it is helpful to see a bit of the bigger picture in order to know how p
honology fits in.
Syntax
Syntax has to do with the positioning of words. There are many ways in which one
could give an explicit
account of the syntax of a language, and it is not our purpose here
to recommend one over another. A
commonly used notation is the rewrite rule. For example, suppose we were trying
to describe a language
which had noun phrases (some longer, some shorter) such as the following:
(1) work
the work
interesting work
the interesting work
the very interesting work
We might propose that a noun phrase is composed of at least a noun
(N), or a noun preceded by a
determiner (D) the, or a noun preceded by an adjective phrase (AP), or a noun pr
eceded by a determiner and
an adjective phrase. An adjective phrase consists of at least an adjective (A),
or an adjective preceded by a
degree adverb (Deg) such as very. The rewrite rules to describe these facts woul
d be something like:
(2) NP N
NP D N
NP AP N
NP D AP N
AP A
AP Deg A
We might even propose ways to combine these rules into one rule in order to show
that they are related in
some way. The parentheses in the rules shown in (3) indicate optional elements i
n the phrase.
(3) NP (D) (AP) N

AP (Deg) A
The point is that we are being explicit about the position and organization of w
ords in this language. Notice
also that we are going from the top down, from the bigger units to the smaller uni
ts, from the phrases to
the individual words.
The rewrite rules we have given above are part of the grammar (or s
yntax) of English. But we also
need to tell what the actual words of English are. This is done through the lexi
con. The lexicon is (at least)
a list of words of the language, such as the following:
(4) N work
N flower
N dog
D the
D a
A interesting
A pretty
A cute
Deg very
2
The lexicon also contains lots of information about these words in so
me fashion or another, including
pronunciation, meaning and usage. And the lexicon may actually be quite differe
nt in appearance or form
than the kind of rules that were just presented. Those details are irrelevant he
re.
If we combine all of the rules we have given so far, we can appropriately genera
te phrases such as:
(5)
NP
D
AP
N [Produced by the rule NP (D) (A
P) N]
Deg
A
[Produced by the rule AP (Deg) A]
the
very
cute
dog
Morphology
Morphology has to do with the structure of words. And just as with the positioni
ng of words in a language,
we need to be explicit about the order in which the different parts of a word go
together. For example, it is
part of your knowledge of English that the plural suffix -s must fol
low the stem of the word, rather than
precede it.
(6) dog-s
cat-s
tree-s
stone-s
Each functional piece of a word is called a formative or, (almost) alternatively
, a morpheme.
a
The word
dog has one morpheme, the word dogs has two morphemes (dog-s), and the word edit
ors has three (edit-ors). In the next chapter we consider ways in which the structure of words may be
described.
Phonology
Phonology is the study of the organization of sounds in language. Our
study of phonology looks at two
major aspects. One aspect that we consider is the inventory of sounds
that a language has. For example,

English has sounds which do not occur in French, and vice versa. If
one is studying a language that has
never been analyzed or written down before, this is an important area of study.
A second aspect we consider is the set of rules which specify exactl
y how each sound is pronounced
and how sounds affect and are affected by the sounds around them. Understanding
this part of language is
also crucial for the design of writing systems for languages. It is also importa
nt for learning the language.
Every language has internal structure and organization, regardless of the social
position of its speakers, but
until one unlocks the secrets, it may remain mysterious and seem difficult.
Phonetics
Phonetics deals with the physical aspects of the sounds of languages, especially
how sounds are articulated
and perceived, but not how they are organized. A person trained in p
honetics is able to transcribe words
from virtually any language. This transcription is most often the basi
s on which phonological analysis is
done, although acoustic studies may also play an important role.
A couple of brief examples may help clarify the distinction between p
honetics and phonology. A
phonetic transcription of a language may include sounds that are similar to the
t, d, th, and r of English, as
well as other sounds. As a result of analysis, however, we may discover that the
re is in fact only one t-like
sound phonologically in the language. The other sounds that we hear are variatio
ns of this one t-like sound.
Consequently, it is likely that the alphabet of the language will include a sing
le symbol to represent these
sounds instead of four or five symbols.
a
These terms are not exactly synonymous, but enough so for our purposes here.
3
Consider also the words vain and vanity in English, or the pair sane and sanity.
The first word in each
pair has the phonetic vowel (or diphthong) [cj] and the second has the phonetic
vowel []. If we were to
write the words scientifically, then these are the symbols we should use, one migh
t say. But these vowels,
be what they may be phonetically, are typically referred to in traditional studi
es of English as long a and
short a, and often written as the letter a with and without a macron over the lett
ers. This transcription is
not phoneticwhat does the letter a have to do with the phonetic transcription [c
j]? However, the names
long a and short a are important in that they point to a systematic co
rrespondence between phonetic
sequences which is fundamental to understanding the sound patterns of
English. This correspondence is
captured in the imperfect English spelling system by the use of the letter a for
both the long and the short
version, with an additional something added for the long version (the
use of the silent e in sane and the
combination with i in vain). Successful literacy programs in other lan
guages also depend on similar
knowledge of how the sound systems in those languages workof their phonologies. O
ur understanding of
these systems may begin with phonetics, but it does not end there.

Key Concepts
syntax rewrite rule morphology
formative / morpheme phonology phonetics
Postscript for Teachers
The formalisms adopted for syntactic and morphological generalizations a
re not a major issue for our
purposes. We are primarily interested in clarifying the distinction between and
interrelatedness of the major
components of formal structure. However, this chapter does introduce th
e rewrite rule notation, which is
used extensively in the book. It also introduces the idea of developing explicit
rules as a means to describe
recurrent patterns found in language. It should be noted that we are
embedding phrases within phrases.
Therefore, it is important that the noun phrase be described as containing an op
tional adjective phrase, not
an optional adjective.
4
1.1 Try it for yourself with English
Use the following rewrite rule and lexicon, in addition to the rules
and lexicon found in the chapter, and
determine at least eight good English phrases that may be described by them. (Ch
eck your answers with the
feedback section at the end of this chapter.)
1
PP P NP
P to
P on
P for
1.2 Try it for yourself with Seri
Examine the following rewrite rules and lexicon. Determine which of the phrases
provided are possible and
which are not possible according to the grammar of Seri. (Check your answers wit
h the feedback section at
the end of this chapter.)
2
PP NP P
NP N (D)
P ano (in)
P iti (on)
N xepe (sea)
N hasaj (basket)
D com (the)
To evaluate:
(a) ano xepe
(b) iti hasaj com
(c) xepe com
(d) com xepe
(e) xepe com ano
(f) iti com hasaj
5
Feedback for Chapter 1
1.1
1
Some possible answers: on the dog, on a dog, on the flower, to a flower, to wo
rk, for the flower, for work,
for the dog, to the pretty flower, to the very pretty flower, on the cute dog.
1.2 Seri
2
(a) ano xepe (ungrammatical: P must follow NP)
(b) iti hasaj com (ungrammatical: P must follow NP)

(c) xepe com (grammatical)

(d) com xepe (ungrammatical: D must follow N)
(e) xepe com ano (grammatical)
(f) iti com hasaj (ungrammatical: P must follow NP, D must follow N)
6
Section 1
Morphological Rules
In this section we show that it is important to account for every p
iece of every word in the grammar of a
language. We introduce morphological rulesrules having to do with the structure o
f words and the shapes
of morphemes in words. We also introduce some basic notions such as features.
It should be noted that most of the topics in this section do not pertain to pho
nology per se, and so one
may choose to begin with Section 2 instead if one has already had a
n introduction to morphology
previously. Nevertheless, certain concepts that are helpful later are i
ntroduced in this section. These
concepts may be reviewed by examining the "key concepts" box at the end of each
section.
7
Chapter 2 - Word Structure
Some words consist of only one formative (or morpheme), such as table. Others co
nsist of several, such as
disappearances (dis-appear-ance-s). If a language had only one morpheme
in each word (and some
languages may be almost like this), there would not be much to say about word st
ructure in that language.
But most languages do have combinations of morphemes in words, so rules specifyi
ng how the formatives
may be combined are necessary. (Note: These rules do not tell us how
we should make words and
sentences; they tell us how we do make words and sentences. Dont conf
use these with the rules you
remember from high school which told you not to use dangling participles, etc.)
Just as there are many approaches to syntax, so there are many appro
aches to morphology, and there
are many issues of importance. But any approach must specify the orde
r in which formatives appear. We
adopt here a fairly simplistic approach. For simple cases like dog-s, we write a
rule such as the following:
(7) N N - PLURAL
(A noun may consist of a noun and a plural morpheme.)
The lexicon of English must then also include information about how the Plural m
orpheme is realized
(the most common form being -s, of course). (This is discussed more in chapter 3
.)
For more complicated words, such as disappearances, we assume the following addi
tional rules:
(8) a. V NEGATIVE - Verb
stem
cf. dis-appear
b. N V - NOMINALIZER
a
cf. appear-ance
The rules in (7) and (8) are somewhat different in productivity. Most nouns can
be pluralized, but most
verbs cannot occur with the prefix dis- and most verbs cannot occur
with the suffix -ance. But rule (8a)
creates verbs from which rule (8b) can form nouns, from which rule (7) can form

plural nouns. Given rules

such as these, we see that words such as disappearances have a nested structure li
ke (9a) rather than the
flat structure of (9b).
(9) a.
N
N
Plural
V
Nominalizer
Neg.
V
stem
dis
appear
ance
s
b.
N
dis appear ance
s
Nevertheless, as long as we know the exact order in which the morphemes occur, w
e have the information
we need for most of our present purposes.
If one were looking only at simple English words, one would not understand well
how other languages
construct words. For example, one can take off each of the formatives from the
word disappearances and
find another word: disappearance, appearances, appearance, disappear, app
ear. This is not always the
case in other languages.
b
Consider the following example from Seri:
c
a
A Nominalizer is a morpheme that changes some category of word (such as Verb, f
or example appear) into a Noun
(appear-ance).
b
Of course, it is not always possible in English either. We have ch
osen simple examples here to illustrate. But
8
(10) mayomazt s/he didnt tattoo you
This word is parsed as follows:
(11) ma - yo - m - azt
stem meaning tattoo
Negative prefix
Past tense prefix
Second person singular direct object agreement prefix
If we wanted to express a first or second person subject (I, we, or
you), another prefix would occur
between the direct object prefix and the tense prefix:
(12) mahyomazt I didn t tattoo you
The point here is that Seri is not like English. If we were to tak
e off the prefix ma, the resulting string
would not be a possible word of Seri. If we were to take the stem azt by itself,
or the stem plus any one of
the prefixes shown, the results would be nonsense in Seri. Nevertheless, all of
this is not our real concern
here. All that we want to know for our purposes is the exact order
of morphemes, and a word structure
rule (also known as a word formation rule), such as the following, suffices:
(13) Verb Direct Object

Subject
Tense
(Negative)
Verb


stem
Agreement
d
Agreement

This rewrite rule expresses the fullest expansion of the verb shown in our data.
It does not indicate that the
absence of overt subject agreement means that the subject is third person (altho
ugh that is important).
e
Practical Procedures
Word structure is discovered by making morpheme cuts. One looks for the (largest
) string of letters which
regularly corresponds to a certain meaning. The nominalizing suffix
an
ce appears repeatedly in the
following list of words, and everything before it is different. Therefore we can
make a morpheme cut before
this string of sounds.
f
(14) appearance
defiance
utterance
conveyance
guidance
An important aid in morphological and phonological analysis is the par
adigm. A paradigm involves
columns and rows of lexical material, where each column and row has
some common element, such as a
stem or a suffix. In the paradigm below, the first column is the form which the
verb has in the present tense
if the subject is first person, the second column is the form which
the verb has in the present tense if the
subject is third person singular, the third column is the verb in the past tense
, and the fourth column is the

nevertheless, the difference between English word structure and that of

many other languages is important to
recognize. All verb stems in Seri must occur with some affix; they a
re said to be bound stems. Most verb stems in
English are not bound stems. The morpheme demonstr
in English, however, must o
ccur with a suffix: demonstr
ate,
demonstr
able.
c
The Seri words are italicized here since they are given in a non
technical orth
ography.
d
Many languages include an affix (prefix or suffix) in the word to
indicate the subject or direct object. These are
called agreement affixes in this book.
e
Some linguists refer to this absence of phonological material as a ze
ro morpheme. We hesitate to do this in many
cases, although nothing important hinges on this matter in this course.
f
The suffix
ance has a grammatical meaning as a Nominalizer (it makes something i
nto a noun). Some formatives
do not actually have a meaning, but this problem is not of interest to us here.
9
verb in the present participle (the ing form). Each row is a different verb. (The
English data are presented
here with phonetic symbols.)
g
Any time a morpheme appears in more than one combination, a paradigm (however sm

all or large) can

and should be constructed to help in the analysis.
The next step is to identify, as much as possible, the phonological material in
each column or row that
contributes the meaning of that column or row. Looking across the first row above,
for example, we can
see that the morpheme fake is represented consistently through the paradigm by t
he string [fejk]. Looking
down the first column, we can see that there is no phonological material associa
ted with the present tense,
first person. The phonological material associated with the second column is an
[s], that of the third column
[t], and that of the fourth [ip]. We then make (tentative) morpheme cuts as show
n below:
Rule of Thumb: Going down a column which relates to an affix, try to keep as muc
h identical or similar
phonological material with the affix as possible. Going across a row which
relates to a single stem, try to
keep as much identical or similar phonological material with the stem as possibl
e. If there is discrepancy or
doubt, draw a circle around the sound(s) for which you have questions and keep y
our options open.
We can then make note of the results by putting the individual morph
emes at the top of the columns
and at the beginning of the rows. (We have written the suffixes with a hyphen be
fore them.)
h
(17) (no suffix) s t ip
Icjk
dip
nik
The combination of the morpheme {Icjk} and the morpheme {s} gives the
form [Icjks], as desired.
Likewise, the other morpheme combinations give the correct surface forms (the fo
rms which we are trying
to account for). Sometimes the facts are more complicated, as we will
see in later chapters. Nevertheless,
the techniques for cutting morphemes are basically the same.
g
The symbols used in this book are, for the most part, drawn from
the set used by the International Phonetics
Association.
h
One can only make hypotheses at this time since we dont know much
for sure yet; perhaps one column has two
suffixes rather than only one, or other facts will be brought to bear on the dec
isions, as we will see later. If there are
two affixes, one needs to set up a column for the combination, if the affixes co
occur.
(15) Present (1
st
) Present (3
rd
) Past Present Participle
fake
Icjk Icjks Icjkt Icjkip
dip
dip dips dipt dipip
nick

nik niks nikt nikip

(16) Present (1
st
) Present (3
rd
) Past Present Participle
fake
Icjk Icjk s Icjk t Icjk ip
dip
dip dip s dip t dip ip
nick
nik nik s nik t nik ip
10
Key Concepts
word structure rule paradigm morpheme cut
Postscript for Teachers
When several inflectional affixes appear in the same word, such as in
the Seri verbs in this chapter,
approaches to morphology differ as to how they should be handled. Some approache
s add one formative at
a time, whereas others add them all at once, as in rule (13). This difference in
approach is irrelevant to the
issue here. All that we care about is that the various formatives be put in the
right order.
2.1 Try it for yourself with Walmatjari
The data for this exercise are in Appendix F at the back of this b
ook. These data contain a number of
complications; nevertheless, you should be able to determine the word structure
rule for nouns (which are
inflected for "case") in this language, even by looking at only one noun.
3
2.2 Try if for yourself with Manam
Give a word structure rule for the following data and a list of the morphemes. (
Check the feedback section
at the end of this chapter.)
4
1. tomogu my father 6. dcbugu my hand
2. tomop your (sg.) father 7. motogu my eye
3. tomo his/her father 8. wosogu my breath
4. tomodi their father 9. moIopomip your (pl.) voices
5. tomomip your (pl.) father
2.3 Now try it again with Selepet
Give word structure rules for the following data and a list of the morphemes. (M
orpheme breaks have been
given for you.)
5
1. oto
nc my elder brother 5. borot
Iip
gc your daughters
2. borot
nc my daughter 6. tcbc
gc your bow
3. borot
Iip
nc my daughters 7. cmct
nc my house
4. oto
gc your elder brother 8. cmct
gc your house
2.4 Try it once more with Manam
Give a word structure rule (showing positions for the subject and object agreeme
nt affixes) and a list of the
morphemes for the following data.
6
1. u
piIc I spoke 5. u
noku I jumped
2. di
piIc they spoke 6. u
tc
di I saw them
3. u
Ioko I went 7. di
tc
o they saw me
4. di
Ioko they went 8. u
bogo
di I brought them
11

Feedback for Chapter 2

2.1 Walmatjari
3
Word Structure Rule: Noun Noun
stem

Case
2.2 Manam
4
Word Structure Rule: Noun Noun
stem

Possessor
Morphemes:
father tomo breath woso
eye moto voice moIopo hand dcbu
FIRST PERSON SINGULAR POSSESSOR
gu
SECOND PERSON SINGULAR POSSESSOR
p
SECOND PERSON PLURAL POSSESSOR
mip
THIRD PERSON PLURAL POSSESSOR
di
(The absence of overt possessor indicates third person singular.)
2.3 Selepet
5
Word Structure Rules:
Noun Noun
stem

Possessor Noun Noun
stem

Plural
Possessor
(Some might suggest combining these using parentheses around Plural.)
Morphemes:
oto elder brother tcbc bow
cmct house borot daughter
nc FIRST PERSON SINGULAR POSSESSOR Iip PLURAL
gc SECOND PERSON SINGULAR POSSESSOR
2.4 Manam
6
Word Structure Rule: Verb Subject

Verb
stem
Direct Object


Agreement
Agreement
Morphemes:
piIc speak noku jump bogo bring
tc see Ioko go
u
FIRST PERSON SINGULAR SUBJECT
di
THIRD PERSON PLURAL SUBJECT

o FIRST PERSON SINGULAR DIRECT OBJECT

di THIRD PERSON PLURAL DIRECT OBJECT
(Actually, the appearance of the morpheme di twice in this list is artificial. Giv
en the word structure rule,
it would be clear that the string di is a prefix in one use and a suffix in the ot
her, so di could be glossed
simply as THIRD PERSON PLURAL. The use of a hyphen to indicate where the morphem
e break occurs is a
convention, not a formalism.)
12
Chapter 3
Suppletive Allomorphy
3.1 Word classes
Many languages of the world have their lexicons divided into classes
of one sort or another. People who
have studied or learned Romance languages know about masculine and femini

ne words. One fact of a

Romance language such as Spanish happens to be that the word for the which occurs
before plural nouns
has two alternating shapes: los and las. Sometimes one is correct to use and som
etimes the other. One has to
memorize that the word casas houses takes the form las and the word hogares homes ta
kes los. Its just
an arbitrary fact of Spanish. A noun belongs to either one class or to the other
.
It is also very common to find that word classes such as nouns or verbs have sub
classes with respect to
morphology. For example, one group of nounsespecially body part nounsmigh
t require an affix to
indicate the possessor. Another group of nouns might never allow such
an affix to occur on them. The
lexicon of the language must include this information since we want t
o reflect the speakers knowledge
about the words. A sample set of morphemes is given for Seri, where +Poss means
that the word must occur
with a possessive prefix,
Poss that it never does, and +/Poss that it may or may
not.
a
(18) Morphemes
hast stone Poss
tom money Poss
caac nephew/niece +Poss
lit head +Poss
With this information, we understand why some words exist and some words do not
exist in Seri, as shown
below (nonexistent words are given with an asterisk before them).
(19) Words
hast stone *hihast (my stone)
tom money hitom my money
*caac nephew/niece hicaac my nephew/niece
*lit head hilit my head
The word hihast does not exist because the word hast is a Poss noun and cannot t
ake a possessive prefix.
The words caac and lit do not exist because the roots in question are +Poss noun
s and cannot occur without
a possessive prefix.
This is one type of morphological subclassification of roots; the roots are divi
ded into groups based on
whether or not they may co
occur with other morphemes. The particular type of cl
assification illustrated in
the examples above is very common: body part words and kinship terms often are t
he only words that may
or must have possessive affixes in a language. One often sees the terms alienabl
e class for words like stone,
and inalienable class for words like head and mother, since the latter do not oc
cur without the mention of
the possessor in many languages. These classes of words do not line up exactly f
rom language to language,
however.
Other times we find that we must posit classes of words to account
for other facts. For example,
consider the following illustrative words from Seri:
(20)
camiz shirt
hi
lit my head hi
camiz my shirt
mi
lit your head mi
camiz your shirt
i
lit his/her head
i
camiz his/her shirt

(21) hi
xiiha my older brother hi
quipaz my grandchild

ma
xiiha your older brother ma
quipaz your grandchild
a
xiiha his/her older brother
a
quipaz his grandchild
a
If the language uses the lack of an overt affix to indicate third person, the di
fference between +Poss and
Poss nouns
cannot be determined simply by examining whether or not the noun occurs without
a person affix.
13
Notice that there are two shapes each for the second person and third person pos
sessive prefixes. With one
class of words the prefixes mi
and i
occur, and with the other class of words
the prefixes ma
and a
occur.
When a formative, such as third person possessor, has two shapes, we say that it
has two allomorphs.
Allomorphs must be handled one way or another in the description of a language.
If they are handled in
the lexicon, they are called suppletive allomorphs.
b
If they are handled by the phonological component of
the language, they are phonological allomorphs. The latter are a major emphasis
of phonological theory and
we look at them in the second section of this book. The difference between the t
wo is fairly simple, and will
be developed in the next few chapters. Suppletive allomorphy is essentially allo
morphy which is a property
of one morpheme, and for that reason is in the lexicon. Phonological
allomorphy is something that is a
property of the sound pattern of the language and is not limited to one morpheme
.
The lexicon must include some way to handle suppletive allomorphs. We might thin
k of a lexicon as a
set of rewrite rules that map from a semantic label, such as DOG, to the languag
e particular form, as shown
below.
c
(22) (English) DOG dog (Spanish) DOG perro
(French) DOG chien (Seri) DOG haxz
The realizations of affixes could also be indicated by such rules.
d
(23) (English) Plural
s
(Seri) 1st Singular Possessor hi

(Seri) 2nd Singular Direct Object ma

When a morpheme has suppletive allomorphs, both of these must appear
in the rule. The Seri rules for
second and third person possessor are given below.
(24) 2
nd
POSSESSOR ma
with kinship nouns
mi
elsewhere
3
rd
POSSESSOR a
with kinship nouns
i
elsewhere
The allomorphs mi
and i
were chosen as the elsewhere cases since they have the w
idest distribution in
the language.
e
In the Seri examples above, the word classes worked out fairly neatly: kinship n

ouns worked one way,

and body part nouns another. This is not always the case, however. O
ften there are simply two or more
arbitrary classes of roots, and some may be quite small. Consider the following
English plurals:
(25) Singular Plural
dog dog
s
cat cat
s
cow cow
s
deer deer
sheep sheep
ox ox
en
The noun ox is one of the very few nouns in English which still pluralize with t
he Old English plural suffix

en (children and brethren are more complicated examples). The lexical entry for
PLURAL in English must
mention this small class of nouns, and also the class of nouns which have no suf
fix for plural. The lexical
entries for the nouns of English must tell the class to which the noun belongs (
if it is not in the elsewhere
class, at least).
b
Morris Halle (1989) On Abstract Morphemes and Their Treatment (paper
presented to the Arizona Phonology
Conference), refers to morphemes with suppletive allomorphs as abstract morpheme
s.
c
These are sometimes called spell
out rules.
d
Although there is no linguistic reason to include hyphens with affixes in these
rules, we will do so simply to remind
the reader that a given morpheme is an affix. But it is really not important.
e
There are more non
kinship possessed nouns than kinship nouns. In additio
n, the allomorphs mi
and i
occur with
nominalized verbs as well as simple nouns. It is not always so clear what the els
ewhere case is.
14
(26) PLURAL
en with class A Nouns:

with class B cow

s elsewhere deer [Class B]
ox [Class A]
sheep [Class B]
table
etc.
(You have probably thought of other words which dont use the
s suffi
x. We discuss some of them in a
later chapter.)
It is appropriate that the
s allomorph is chosen as the elsewhere case in Engli
sh. This allomorph occurs
with by far the largest class of words (which class is in fact so large as never
to be enumerated exhaustively
as Classes A and B could be). It also is the suffix that people us
e when they have not learned the special
lists of words which take the other allomorphs, as in the childrens speech form f
oots, rather than feet.
Of course, such arbitrary classes are harder to remember, and therefore more uns
table over time, than
classes which are based on something tangible (such as classes compose

d of body part nouns). For this

reason, they tend to get smaller with passing generations, as illustra
ted by the gradual replacement of
brethren with brothers.
An illustration from Kikuyu
The following data are from Kikuyu which, like other Bantu languages, has prefix
es on both singular and
plural nouns. Take a quick look at the data and make some morpheme cuts to see t
hat the noun stems follow
the morphemes that indicate singular and plural.
(27) Singular Plural
1.
teacher murutoni orutoni
2.
girl muirctu oirctu
3.
woman mutumio otumio
4.
buyer muguri oguri
5.
root muri miri
6.
tree muti miti
7.
lion muroodi miroodi
8.
mattress muuto miuto
9.
chair gcti cti
10. yam gikoo ikoo
11. tray gitoruru itoruru
We can therefore propose the following word structure rule:
(28) N Singular/Plural N
stem
You will also notice that the prefixes arent the same for all nouns. The singular
prefix is sometimes mu

and sometimes g
(we think). The plural prefix is sometimes o
, somet
imes mi
, and sometimes null. So
how many classes of nouns are there? For the singular prefix, there are
two classes. For the plural prefix,
there are three classes. There is more than one way to describe these results, b
ut we will just present one.
Lets try to get by with just three classes. The plural prefix would have the foll
owing lexical entry:
(29) PLURAL o

with Class A nouns
mi
with Class B nouns


with Class C nouns
By this analysis, tumio is a Class A noun stem, ti is a Class B noun stem, and k
oa is a Class C noun stem.
This information about these noun stems must be included in the lexicon so that
one will know how to form
the actual singular and plural words.
15
Writing up the results
A quasi
formal write
up of an analysis is appropriate for some audienc
es, but not all. Here, as
elsewhere, we are going to suggest prose descriptions that might be m
ore appropriate in many cases,
making the results available to a wider set of audiences. The following paragrap
h says the same thing as the
formal statement (29) above, accounting only for the data that we hav
e at hand. A proper write
up would
probably include more examples.
Nouns in Kikuyu typically have two forms: singular and plural. A singular prefix
appears on each
singular noun, and a plural prefix appears on each plural noun. There are two si
ngular prefixes: mu

and g
; and there are three plural prefixes: o
, mi
, and null. Each noun in Kik
uyu belong to one of

several arbitrary or partially arbitrary groups, or classes, depending on which of

these prefixes the
noun occurs with. Class A nouns (which may all be human beings?) take the plural
prefix o
. Class
B nouns (which include animals and some plants?) take the plural prefix mi
. Cla
ss C nouns (which
include inanimate objects and some plants?) take the null plural prefix. The sin
gular prefix for Class
A and Class B nouns is mu
(as in mu
ti tree). It is g
for class C nouns (as in
g
cti chair).
Singular Plural
Class A muguri oguri buyer
Class B muti miti tree
Class C gcti cti chair
Basic Practice with Kikuyu
How would you complete the following lexical entry, making reference to
just the classes shown above?
7
Check your answer before continuing.
(30) Singular mu

g

3.2 Syntactic Conditioning
Arbitrary word classes are a last ditch measure. If there is suppletive allomorp
hy, one should always check
out the possibility of some conditioning for the allomorphy which doesnt need to
be memorized separately
from the other information about the word. In this chapter and those
that follow, we look at cases where
suppletive allomorphy is determined by something other than arbitrary word class
es.
The distribution of suppletive allomorphs may be sensitive to syntactic factors. W
hen this is true, the
language learner does not have to memorize lists of words.
The Seri first person singular subject prefix (for I ) has two suppl
etive allomorphs, hp
and h
. The
distribution of these allomorphs depends on whether the clause is supe
rficially transitive or intransitive
(whether there is an overt direct object in the clause or not).
f
(31) hp
yo
m
afp I didnt arrive h
yo
m
aho I didnt see it
hp
yo
m
atax I didnt go h
yo
m
pii I didnt taste it
hp
yo
m
panzx I didnt run h
yo
m
azt I didnt tattoo him/her
hp
yo
m
p
azt I wasnt tattooed ma
h
yo
m
azt I didnt tattoo you
The rule for this morpheme does not need to refer to arbitrary class
es of verbs. Rather, it can refer to
something which is a basic characteristic of verbsor more accurately, p
erhaps, of the clauses in which
they occurthat we find in many languages, their transitivity.
(32) FIRST PERSON SINGULAR SUBJECT
h

in transitive clauses
hp
in intransitive clauses
f
These words would actually also have an i at the beginning
said in isolation.
16
It is important to recognize the advantages of this
over one which simply posits
arbitrary classes. An arbitrary class solution would
or every single verb in the language
which allomorph of the first person singular subject

of them if they were

kind of solution
have to specify f
prefix it may oc

cur with. This would imply that

speakers have to learn that fact as a random fact for every verb which they use.
The solution given above,
however, takes advantage of the fact that we already know from other sources whe
ther a verb is transitive
or not; the speaker also knows this (although perhaps not consciously)
and can use this information to
decide on the correct allomorph of this prefix. The arbitrary class s
olution would add a great deal of
complexity to the grammar of Seri. The syntactic conditioning solution,
on the other hand, is extremely
simple, although it is outside of the word itself. We choose the solution which
adds the least complexity to
the description of the language.
3.3 Phonological Conditioning
It is not uncommon for suppletive allomorphy to be sensitive to the
phonological shape of an adjacent
morpheme rather than to some morpheme class (arbitrary or otherwise). When this
is true, the lexicon does
not have to list classes of morphemes since a simple inspection of the sounds ma
kes the choice of allomorph
clear.
A simple example from English illustrates this. The indefinite article
has two shapes: a and an. The
first allomorph occurs before consonant
initial words and the second before vowe
l
initial words: a tree, an
orange.
g
Speakers of English dont memorize which word occurs with a and which with an and
then guess
at words they dont know. They figure out the rule, use it, and apply
it to new words.
The same thing
happens in Korean where the article has the shape i after words which end in con
sonants, and ka after words
which end in vowels: salam i (man Article), ai ka (boy Article).
The passive morpheme in Seri has two suppletive allomorphs, and the c
hoice depends primarily on
whether the next morpheme begins with a consonant or with a vowel.
h
(33) t
ah
cazni was s/he bitten? t
p
azt was s/he tattooed?
t
ah
tiip was s/he pinched? t
p
ahit was it eaten?
t
ah
fain was it tied up? t
p
am was it swallowed?
t
p
ezi was s/he defeated?
t
p
emen was it winnowed?
The lexical entry for the passive morpheme would look something like the followi
ng:
(34) PASSIVE ah
before consonants
p

before vowels
Somewhat more formally, this rule can be written with the following notation, wh
ere the slash (/) is short
for in the environment and the placement of the dash ( __ ) indicates whether the
item precedes or follows
the environment on the right of the slash:
(35) X Y / __ Z
X is rewritten as Y if X precedes Z.
(36) R S / T __
R is rewritten as S if R follows T.
The Seri rule would be the following, where C represents any consonant and V repre
sents any vowel:
(37) PASSIVE ah
/ ___ C

p

/ ___ V

g
It is the sounds that count, not the spelling. Notice that the allomorph a occu
rs before unicorn since the noun begins
with a glide and not a vowel phonetically: [ junikoJn].
h
It also matters in Seri whether the next morpheme is a prefix or a root, but we
ignore this complication here.
There are two important cross
linguistic observations which we mention he
re but do not elaborate on. First, roots do
not exhibit phonologically conditioned suppletive allomorphy, unlike affixes. Se
cond, morphemes are sensitive to the
shape of morphemes closer to the root, but not to morphemes farther out.
17
From this rule we know that the shape of the passive morpheme in Seri depends on
whether the morpheme
following it begins with a consonant or a vowel.
The conditioning for the suppletive allomorphy could be something more
specific than consonant vs.
vowel, of course. For example, consider the object nominalizer in Seri, which tu
rns a verb stem like eat into
a noun which means what (someone) eats/ate. It has three shapes of relevance to
us here: null, y
, and o
.
Some examples of the roots with which these allomorphs occur are given below.
(38) Null
ip straighten,
ii hear,
iic plant,

iip carry on head,
ixim fear
y
aazi carry,
aat cook in ashes,

aatjc spread around,
eefel stumble on
o
quesejc gnaw,
afmoj carry firewood,
pii taste,

oocta look at,
oi delouse
This allomorphy depends on the kind of vowel that the verb begins wi
th.
If the verb begins with a close
front vowel (i or ii), then it takes the null allomorph. If it begins with a lon
g open vowel (aa or ee), then it
takes the allomorph y. Otherwise, before all other vowels and before all conson
antsthis is the elsewhere
casethe allomorph o occurs.
In Walmatjari, the emphatic morpheme has two shapes: [
n i| following vowels, a
nd [
pon i] following
consonants. Similarly, the morpheme for then has two shapes:
[
Io] f
ollowing vowels, and [
poIo]
following consonants.
i
Also in Walmatjari, certain suffixes have allomorphs which are accounted for par
tially by the number
of syllables in the stem to which they attach: one allomorph occurs
only when the host word has exactly
two syllables (the minimum), while the other occurs with host words which are lo
nger.
Basic Practice
Write the following rules in formal notation and check your answers before conti
nuing.
(a) The infinitive prefix has the shape a
before consonants and n
before vowe
ls.
8
(b) The negative suffix has the shape
t after vowels and
on after consonants.
9
(c) The dative suffix has the shape
wu after vowels, and
ku after consonants
(Walmatjari, Australia).

10
3.4 Morphological Conditioning
The shape of a morpheme is sometimes sensitive to the presence of ot
her morphemes in the word,
regardless of their shape. (Syntactic conditioning had to do with some
factor outside of the word form
itself.)
A simple example from English is found with some basic adjectives. Yo
u probably know that many
adjectives occur in three forms:
basic, comparative, and superlative,
such as blue, bluer, bluest, and tall,
taller, tallest. The suffixes
er and
est are added to the adjective
stem.
The adjective good has three
suppletive stems, however; compare good, bett
er, and b
est.
In Spanish, some verbs have suppletive stems that depend on some othe
r morphological fact of the
word. For example, the regular stem for know is sab
(cf. saber, sabemos, sabes,
saba, sabiendo), but the
stem sep
occurs when the verb is in the present subjunctive (cf. sep
a, sepamos, sepas). The regular stem
for do is hac

(cf. hace, hacemos, haces, haca, haciendo), but the st
em hag
occurs in the present
subjunctive (cf. haga, hagamos, hagas). One could view these cases as ones where
there is a special stem
(sep
, hag
) that occurs when the verb is inflected for present subju
nctive, and regular stems (sab
, hac
)
otherwise.
j
i
There are a couple of things that are a bit odd about this set of facts. The fi
rst is that both morphemes have the same
extra syllable [po]. The second is that in each case one allomorph contains the ot
her. One might want to look at these
facts more carefully.
j
These particular verbs happen to have other stems as well, and other contexts w
here these stems may be used.
18
In Seri the first person singular object agreement prefix on the verb (for me) h
as two different shapes:
hpo
in commands and him
otherwise.
k
(39) hpo
sanj Carry me! him
ihasanj to carry me
hpo
moocta Dont look at me! him
xosanj s/he carried me!
Since the shape of the imperative morpheme is irrelevant, the lexical
entry for the object agreement
prefix would look something like this:
(40) 1 SINGULAR OBJECT hpo
/ <Imperative>
him
/ elsewhere
We use the angle brackets here to indicate that if the Imperative morpheme is fo
und somewhere in the
word, the allomorph hpo
must be chosen.
Another prefix in Seri has three shapes, each sensitive to the presence of other
morphemes in the word
(dont worry about the technical names which we give to morphemes):
(41) SUBJECT NOMINALIZER i
/ <Negative>
ha
/ <Passive>
c
/ elsewhere
Again, it is not the shape of the other morpheme that counts, but the mere prese

nce of it somewhere in the

word. (It may be adjacent, but it may not be.)
To see how this works in the bigger picture, consider the following
partial grammar and lexicon of
Seri. (The word formation rule given here is simplified slightly.)
(42) Word Structure Rule:
N NOMINALIZER
(NEGATIVE)
(PASSIVE)
Verb
stem
Lexicon: PASSIVE (see rule (37), p. 16)
NEGATIVE m

SUBJECT NOMINALIZER (see rule (41), p. 18)
Verb stems: ahit eat, atax go
First, we will derive the word for who goes, a subject nominalized for
m. We start with the word
structure rule and choose only the obligatory parts of the rule since we dont wan
t a negative or a passive
form this time.
(43) NOMINALIZER
Verb
stem
We insert the morphemes, starting with the stem. The form of the Sub
ject Nominalizer that must be
chosen, according to rule (41), is c
because the verb is not negative or passiv
e.
(44) (i) NOMINALIZER
atax
(ii) c
atax
We will now derive the word for who does not go, which is a negative
subject nominalized form.
Again we start with the word structure rule, but this time we choose the Negativ
e option as well.
(45) NOMINALIZER
NEGATIVE
Verb
stem
We then insert the morphemes, again starting with the stem.
(46) (i) NOMINALIZER
NEGATIVE
atax
(ii) NOMINALIZER
m
atax
(iii) i
m
atax
k

The Seri words are italicized here since they are given in a non
t
echnical orthography. The prefix hpo
would be
preceded by an i if the word on which it occurs were pronounced in isolation.
19
Now consider a slightly more complicated example, the negative subject nominaliz
ed form what is not
eaten. The steps are the same, but one morpheme (PASSIVE) has phonolog
ically
conditioned allomorphs
and another (NOMINALIZER) has morphologically
conditioned allomorphs.
(47) Nominalizer
Negative
Passive
Verb
stem
(48) (i) NOMINALIZER
NEGATIVE
PASSIVE
ahit
(ii) NOMINALIZER
NEGATIVE
p
ahit
(iii) NOMINALIZER
m
p
ahit
(iv) i
m
p
ahit
In step (i) the root is inserted. In step (ii) the Passive allomorph is chosen;
according to rule (37) the allo

morph p
must be used since the passive morpheme precedes a vowel. In step (iii)
the Negative morpheme
is inserted. In step (iv) the Nominalizer allomorph is chosen; according to rule
(41), the allomorph i
must
be used since the Nominalizer is in a verb which also contains the
Negative morpheme. This is why the

order of the subparts of rule (41) is important: to provide an economical and ac

curate account of the facts.
Key Concepts
morphological subclassification allomorphs suppletive allomorphs
non
arbitrary word classes alienable class inalienable class
arbitrary word classes C, V "slash
dash" notation
suppletive allomorphy conditioned by syntactic, phonological, and morphological
factors
Postscript for Teachers
The concept of suppletion is one that figures in a minor way in some books on mo
rphology and phonology,
but it usually is not developed very much, and often in a rather un
satisfying way. In addition, the term
partial suppletion is also introduced, although that term seems oxymoronic and unh
elpful. How does that
designation affect the formal analysis? Either a morpheme has one unde
rlying form which is changed by
rules of one sort or another, or it has more than one underlying (suppletive) fo
rms which replace each other.
In the case of the latter, these suppletive forms may be totally unlike each oth
er, or they may be somewhat
similar. The term partial suppletion has been used to refer to the sit
uation where they are somewhat
similar.
This chapter introduces the first kind of conditioning for suppletive allomorphy
that we present in this
book. Most books do not discuss facts such as these at all, perhaps
because phonologists usually ignore
suppletive allomorphy, or because of lack of familiarity with such fac
ts. Many descriptions of suppletive
allomorphy consider only the kind of suppletive allomorphy which relies
on arbitrary word classes. We
believe that this is a serious mistake.
The question may arise as to whether a given pair of allomorphs are
suppletive allomorphs or not.
(Students learn about non
suppletive allomorphy in the next major secti
on of the book.) One cannot say
whether two allomorphs are suppletive or not based on the facts alone, but only
as one sees how the facts
are treated within an analysis. One analyst may treat two allomorphs as if they
were suppletive, and another
analyst may analyze them as non
suppletive. The real question is whether the all
omorphs should be treated
as suppletive or non
suppletive. This can only be answered by comparing the comp
eting analyses.
When there are several morphemes which are being looked up in the lexicon, one m
ight wonder which
one of them should be looked up first, especially if the shape of one may depend
on the shape of another. In
general, it seems that one starts with the root and works outwards. Prefixes may
depend on the shape of the
following morpheme, and suffixes may depend on the shape of the preceding morphe
me.
20
3.1 Try it for yourself with Seri
Examine the following data and quickly make some tentative morpheme cuts. The si
ngular form precedes
the plural form.
1. cait, caitj black jewfish 11. camopxa, camopxaj moth (sp.)
2. car, caroj sheep 12. cacmama, cacmamatoj stink bug (sp.)

3. cay, caytaj horse 13. cascarera, cascareraj ladder

4. sarapi, sarapitaj blanket 14. siimenil, siimeniloj Cooper s Hawk
5. canaa, canaataj boat 15. santar, santaroj soldier
6. sh, shtaj owl (species) 16. jone, jonetaj Passiflora
7. taaz, taazoj handkerchief palmeri (plant)
8. xtasi, xtasitoj estuary 17. cahcosa, cahcosaj young deer
9. sir, siroj saddle 18. nosi, nosilc Mourning Dove
10. otye, otyej bottle
Which of the following word structure rules is correct?
11
N N
stem
Pl
or
N Pl
N
stem
List the allomorphs of the plural suffix that you see.
12
Do you see any way to predict to tell which noun takes one allomorph instead of
another?
13
Propose a lexical entry for the plural morpheme that is appropriate for these fa
cts.
14
3.2 Try it for yourself with Huajuapan Mixtec
Examine the following data (given in a broad phonetic transcription).
You will note that there are two
allomorphs for the pronoun meaning our. (The nouns also sometimes change a bit,
but you can ignore that
fact here.)
our ... your (pl.)
1. no mo no mo so no mo ndo soap
2. to mi to mi so to mi ndo feather
3. ti no ti no so ti no ndo dog
4. yki yki o yki ndo niece
5. kotji kotji so kotji ndo cotton
6. jito jito o jito ndo grandmother
7. c1c c1c so c1c ndo door
8. oo o o oo ndo tongue
9. tiji tij o tiji ndo stomach
10. ndo1o ndo1 o ndo1o ndo hand
11. tjiko tjiko o tjiko ndo chest
Complete the following lexical insertion rule for the pronoun our assuming that
there is no way to predict
which noun takes which allomorph.
15
21
1 PLURAL POSSESSOR o /
so /
Now quickly make a list of each group of nouns and examine the noun
s in each group. Are there any
unifying characteristics that would suggest that these are not just ar
bitrary lists of nouns? Recall that
arbitrary classes are something the language learner has to memorize and so alte
rnatives to arbitrary classes
should always be explored.
16
Propose a revised lexical entry for this pronoun that takes this information int
o consideration.
17
After checking your answer, propose a prose write
up of these facts.
18

3.3 Try it for yourself with Madija

The prefix for third person subject has two shapes: i
and null, as shown below
.
l
Give the lexical entry for
third person subject. Write your analysis up in quasi
formal format and also in
prose.
19
1. i
ncbohcroni he did not leave her
2. do i
noboki she did not give it to them
3. bobo poic i
hidzo he plucked the tick from him
4. dzoho i
nohonohoio he carried her toward there
5. kopi i
ninchoro she vomited it
6.
modii he lives
7. howo howo
nohoii he shouted and shouted
8. wchc wchc
nohoio she sways and sways
9.
wotidzc s/he is/was happy
3.4 Now try it for yourself with Seri
Look over the following data quickly and provide the word structure r
ule for infinitives.
20
(Morpheme
breaks have been provided for you. And again, see the appropriate end
notes at the end of this chapter for
answers to these questions after you have jotted down your own answers.)
1. iha
nifz to kick (something) 7. iha
nip to buffet (something)
2. iha
tis to point at (something) 8. iha
tiixp to squeeze (something)
3. iha
poin to close (something) 9. iha
pii to taste (something)
4. ica
panzx to run 10. ica
xapz to congeal
5. ica
xaplc to tremble 11. ica
poct to be full
6. ica
poozi to be pear
shaped 12. ica
neepni to be stooped
Make a list of the roots according to which allomorph they take. Do
you think that you need to posit
arbitrary classes in order to account for them.
21
Why is the best solution not one in which simply two arbitrary classes of verbs
are posited?
22
l
This is a slight oversimplification.
22
3.5 Try it for yourself with Seri
Examine the following data and make the morpheme cuts. The Negative prefix is m

.
Positive Negative
Imperative Imperative
1. hoocta cmoocta look
2. hexl cmexl take
3. hoonl cmoonl stir
4. haat cmaat cook in ashes
5. haanj cmaanj poison arrowheads
Give the word structure rule for imperative forms.
23
Now give the lexical entry for the imperative prefix, making reference
to the presence of the Negative
prefix for one allomorph. The allomorph which occurs in non
negative v
erbs should be the elsewhere
case. (Although from these data it may look like the answer has some
thing to do with consonants and

vowels, it does not. Unfortunately, the data needed to show this present other c
omplications that you do not
want to see yet.)
24
The following data complicate the picture somewhat. How would you integrate them
into your analysis (in
particular, the lexical entry for the imperative prefix)? In this case
, the fact that all of these verb stems
begins with the vowel a (not double aa) is significant.
25
It is a coincidence that the Imperative prefix has the
same shape (c
) in this context as in negative verbs.
Positive Negative
Imperative Imperative
6. catax cmatax go
7. caitom cmaitom speak
8. camjc cmamjc bring (something)
9. catni cmatni touch (something)
10. casquim cmasquim paddle
11. caxpx cmaxpx get mad at (someone)
Now write up all of these facts in prose.
26
3.6 Try it for yourself with Tzotzil
Examine the following data and give the word structure rule(s) necessary to desc
ribe these facts. (The data
are given in a practical orthography in which the j represents a vel
ar fricative and y a palatal
glide/semivowel. Dont forget to take care of the words for wives, as compared wit
h wife.)
27
his/her/their... my/our...
1. stot jtot father
2. stzek jtzek skirt
3. svex jvex trousers
4. sbi jbi name
5. yajnil kajnil wife
6. yajnilik kajnilik wives
7. yixlel kixlel younger sister
8. yalak kalak chicken
23
The Plural suffix has only one shape:
ik. You should have noticed that there
are two allomorphs for each
of the prefixes, however. Complete the following lexical insertion rule
s (using slash/dash notation) and
discuss why you do not appeal to arbitrary word classes.
28
3
rd
POSSESSOR 1
st
POSSESSOR
Now give a prose account of the facts, both the morphological structure and the
allomorphy.
29
3.7 Try it for yourself with Ngemba
In Ngemba, as in other Bantu languages, nouns carry prefixes known as
noun class markers. The noun
class markers for Class III singular nouns have two shapes in this l
anguage, including null, as explicitly
shown. The null form occurs before the consonants n, m, and p these

are the nasal consonants. The

allomorph a
occurs before a variety of consonants, so it should be viewed as th
e elsewhere case. Give an
account of their distribution by completing the following lexical entry.
30
NOUN CLASS MARKER CLASS III SINGULAR /
o /
1. ndo house 6. odi1 place
2. pvu dog 7. oku foot
3. mbp forest 8. oIo1 village
4.
pki basket 9. oti tree
5. pgob skin
3.8 Try it for yourself with Tairora
Give the word structure rule for the following words from Tairora. (T
he data are limited, so morpheme
breaks have been given for you.)
31
Give the lexical entry for the two indirect object prefixes.
32
1
st
person indirect object 2
nd
/3
rd
singular indirect object
1. ti
mi o
mi give
2. ti
ti o
ti say
3. h
oorcc
oorcc call
4. h
umi?cc
umi1cc show
5. h
ori
ori hit
3.9 Try it yourself with Coatecas Altas Zapotec
The first person morpheme in Coatecas Altas Zapotec has three shapes.
Provide an account of the
distribution of these allomorphs by giving an explicit and adequate lexical entr
y for this morpheme. In this
case, you will have to draw on the concepts of more than one section.
33
1. jcnn
o my neck 6. n,o
n my ear
2. jik
o my head 7. mbcI
do my fish
3. ti:tj
o my back 8. boj
do my shawl
4. zoIo
n my eye 9. ji
do my flower
24
5. jo
n my hand 10. zo
do my bean
3.10 Try it for yourself with Mengen
The transitive suffix has two allomorphs. What does the lexical entry for this suf
fix look like?
34
Note: the
vowels in this language are either front (i, e) or back (a, u, o).
Infinitive Transitive
1. omo omoc rub
2. bo boc send
3. kovu kovuc chop
4. kiou kiouc call
5. too tooc walk
6. motc motco like
7. ossi ossio lose
8. soi soio sew
9. toc toco climb

3.11 Try it for yourself with Chol

Analyze the following data and provide a lexicon for the prefixes, as
suming the allomorphs of each are
suppletive.
35
my ... your ... his/her ... Unpossessed
1. kpusikoI opusikoI ipusikoI pusikoI heart
2. kot ot owot ot ijot ot ot ot house
3. hkobo1 okobo1 ikobo1 kobo1 name
4. hko1 oko1 iko1 ko1 grandmother
5. kijim owijim ijijim ijim corn
6. kbuhk obuhk ibuhk buhk blouse
7. kpom opom ipom pom forehead
8. hkchIob okchIob ikchIob kchIob shoulder
9. hkutj okutj ikutj kutj load
10. kmut omut imut mut chicken
11. kotsom owotsom ijotsom otsom salt
12. kti1 oti1 iti1 ti1 lips
25
3.12 Try it for yourself with Madija
Make some quick morpheme cuts on the following data, which are given
in a very broad phonetic
transcription (some detail has been omitted). You may ignore the change from e t
o a in the word for hair.
my ... your ... his ... her ...
1. okonc tikonc konc kononi hair
2. ototi titoti toti totini head
3. opono tipono pono pononi face
4. onotsopc tinotsopc notsopc notsoponi saliva
5. otonc titonc tonc tononi bone
6. owcnc tcnc cnc cncni nose
7. owcbcno tcbcno cbcno cbcnoni tongue
8. owoti woti wotini liver
9. oworibo woribo woriboni ear
10. owipo tipo ipo iponi lower lip
Give the word structure rule for possessed nouns. You may refer to the suffix
n
i as the Feminine suffix
(used only in third person). The third person possessor prefix is null.
36
There are two allomorphs for the first person possessor prefix and tw
o allomorphs for the second person
possessor prefix. Assume they are suppletive allomorphs for now (they
arent really, but you dont know
that and it doesnt make any difference at this point). Give the lexical entry for
each of them.
37
Feedback for Chapter 3
Basic Practice with Kikuyu
7
The allomorph mu
occurs with Class A and B nouns, and the allomorph g
occurs w
ith Class C nouns.
Basic Practice
8
Infinitive o
/ __ C
n
/ __ V
9
Negative
t / V __

on / C __
10


wu / V __

ku / C __
(The allomorphs of the Dative are similar enough to suggest that another analysi
s might be possible, as we
will see later, especially if some other morphemes have similar allomorphy, as i
s the case.)
3.1 Seri
11
The first word structure rule is correct, of course. The plural morpheme follow
s is a suffix.
12
The allomorphs of the plural suffix are: j, oj, taj, toj, lc.
13
There doesnt seem to be any way to predict which noun takes which allomorph.
26
Dative

14
PLURAL
j with Class A

oj with Class B

taj with Class C

toj with Class D

lc with Class E
The labels for these arbitrary classes dont have any significance. We could have
used Class 1, etc., or
Declension 150, etc., or whatever. This rule does not give any indication that o
ne of these allomorphs is the
elsewhere case. That would be determined, if possible, by looking at more facts.
3.2 Huajuapan Mixtec
15
o with Class A, so with Class B.
16
The nouns that take the allomorph o are body part nouns and kinship nouns, wher
eas the other nouns are
not (the only possible exception is feather; it would be interesting to know how
a group of birds would say
our feathers!).
17
1 Plural Possessor o with inalienable nouns
so elsewhere
This rule assumes that the class of alienable nouns (like door) is much larger t
han the class of inalienable
nouns. (After all, there are only so many body part names and names for relative
s.)
18
There are two pronouns for our (first person plural possessor): o and so. The p
ronoun o is used with
inalienable nounsthose which require the expression of some possessor, such as bo
dy parts and
kinship terms. Example: o oour tongues, and jito o our grandmother(s). The pron
oun sois used
with all other nouns. Example: c1c so our doors.
3.3 Madija
19
3
rd
SUBJECT i
in transitive clauses

in intransitive clauses
A third person subject in Madija is indicated on the verb by two different prefi

xes. One of these is null,

and is used on verbs in intransitive clauses. The other is i
, and is used on ve
rbs in transitive clauses.
3.4 Seri
20
The word structure rule would be:
Infinitive INFINITIVE PREFIX
Verb
stem
21
No. The verbs that take the allomorph iha
are all transitive verbs. Those that
take the allomorph ica
are
all intransitive.
22
Arbitrary classes require extra, but unnecessary, information to be included in
the lexicon (the class
labeling). The classes are not arbitrary, however, and the language learner does
not have to memorize these
facts.
3.5 Seri
23
Word Structure Rule: Imperative Verb IMPERATIVE
(NEGATIVE)
Verb
stem
24
IMPERATIVE c / <Negative>
h / elsewhere
25
IMPERATIVE c / <Negative>
c / __ a
h / elsewhere
27

The order of the first two parts of this rule is not crucial (if the verb is neg
ative, then there is always a
consonant before the root). The elsewhere part should be last, of course.
26
An imperative in Seri is formed by adding an imperative prefix to a verb stem.
If the imperative is
negative, the negative morpheme m
appears between the imperative prefix and the
verb stem.
Examples: catax go!, cmatax dont go!.
The imperative prefix has two basic shapes: c
and h
. The shape c
is used, fi
rst of all, whenever a verb
is negative. The examples cmatax dont go! and cmoocta dont look at it! illustrate thi
s. When the
verb is not negative and the morpheme after the imperative prefix begins with th
e vowel a (as opposed
to any other vowel, consonant, or even long aa), the imperative prefix is c
, as
in catax go!. Otherwise,
when the verb is not negative, the imperative prefix is h
, as in haat cook it in
the ashes!, and hoocta
look at it!.
3.6 Tzotzil
27
Noun Possessor
Noun
stem

(Plural)
You may have presented the rule above as two rules, one with Plural and one withou
t Plural.
28

3
rd
POSSESSOR s
/ __ C 1
st
POSSESSOR j
/ __ C
y
/ __ V
k
/ __ V
There is no reason to set up arbitrary word classes because the distribution of
these suppletive allomorphs
can be adequately described by simply referring to the phonological shape of the
noun stems.
29
One kind of noun in Tzotzil, a possessed noun, is formed by adding a prefix in
dicating the possessor
to a noun stem. A plural suffix may occur. (Not enough information is given her
e to know exactly how
it works.) Examples: yajnil his wife, kajnil my wife, yajnilik their wives (maybe al
so his wives?).
The prefix for third person possessor has two basic shapes; s
occurs when foll
owed by a consonant,
and j
when followed by a vowel. Examples: stot his/her/their father, yixlel his/h
er younger sister.
The prefix for first person possessor has two basic shapes; j
occurs when foll
owed by a consonant, and
k
when followed by a vowel. Examples: jtot my/our father, kixlel my/our younger
sister.
3.7 Ngemba
30
NOUN CLASS MARKER CLASS III SINGULAR / __ nasal consonants
o / elsewhere
3.8 Tairora
31
Verb Indirect Object Prefix
Verb
stem
32
1
st
INDIRECT OBJECT ti
/ __ C 2
nd
/3
rd
INDIRECT OBJECT
o
/ __ C
h
/ __ V
/ __ V
3.9 Coatecas Altas Zapotec
32
1
st
SINGULAR POSSESSOR
n / V __ with body part (or inalienable) nouns

o / C __ with body part (or inalienable) nouns

do / elsewhere
3.10 Mengen
34
TRANSITIVE c / back vowels ___
o / front vowels ___
28

3.11 Chol
35
Each of the prefixes has two suppletive allomorphs.

1
st
SG POSSESSOR h / __ k
sounds
k / elsewhere
2
nd
SG POSSESSOR ow / __ V
o / __ C
3
rd
SG POSSESSOR ij / __ V
i / __ C
3.12 Madija
36
Noun Possessor
Noun
stem

FEMININE
37
1
st
POSSESSOR o / __ C 2
nd
POSSESSOR
ti / __ C
ow / __ V t / __ V
29
Chapter 4
Multiple Function Formatives
Formatives (morphemes) do not always carry a single meaning. They may i
ndicate two (or more)
meanings, such as tense and aspect, person and number, or tense and pe
rson. We may refer to them as
multiple function formatives (although this is not a commonly used term).
a
You are well aware that I in
English indicates first person singular, and we first person plural. These morph
emes therefore indicate both
person and number.
In Manam there are two sets of subject agreement prefixes, one for realis mood (
present and past), the
other for irrealis (future):
(49) Realis Irrealis
1 singular u m
2 singular 1u go
3 singular i po
1 plural exclusive 1i go
1 plural inclusive to to
2 plural 1o 1omo
3 plural di do
Since the difference between these two sets is related to the mood o
f the verb, we do not consider the
formatives {u
} and {m
}, etc. to be allomorphs; rather, they are simply two dif
ferent morphemes. They are
formatives that have two functions: to indicate person/number and to indicate mo
od.
In Greek there are case suffixes that occur on nouns. But a suffix such as
os act
ually indicates three
things: case (nominative), number (singular), and word class (masculine)
. The suffix cannot be further
dissected so as to reveal a formative for each of these functions. The rule migh
t look something as follows:
(50) +Masculine
os

Plural
+Nominative
In Seri there are several words for the. The choice between them largely has to
do with the position and
number of the noun: cop if it is erect, com if it is prone, quij if it is squat
or seated, coi if it is plural, etc. We
also view these as separate morphemes with more than one function, not allomorph
s of a single morpheme.
Examples like the Greek case suffixes are not uncommon; the morphology
is built around these
multiple function formatives. In some languages, however, the multiple function
formative is unusual since
the typical formatives in the paradigm do not conflate multiple functions.
Some multiple function formatives have been given the special name of portmantea
u (a term coined
by Charles Hockett). A portmanteau conflates two or more morphemes whi
ch otherwise occur
independently in the language. For example, was in English could be considered
a portmanteau; it conflates
the verb be and Past tense (which is elsewhere manifested as
d). The word am c
onflates three things: the
verb be, Present tense, and first person singular subject. The word w
orse conflates two morphemes:
the
adjective bad and the comparative (which is usually
er). The morpheme du in Fre
nch is a portmanteau that
conflates de of and le the (masc. sg.). The verb form s I know in Spani
h is a portmanteau that
replaces what one might otherwise expect based on the regular stem and regular m
orphology (sab
plus
o).
The verb go to (somewhere) in Seri has two shapes: one if the subje
ct is singular and another if the
subject is plural. Since the plural stem displaces the regular affixes that norm
ally indicate agreement with a
plural subject in this language, the plural stem in this case may be considered
a portmanteau.
(51) i
t
yaai did s/he go to it?
i
t
oziit did they go to it?
a
One recent book (Andrew Radford, Martin Atkinson, David Britain, Hara
ld Clahsen, and Andrew Spencer, 1999,
Linguistics: an introduction; Cambridge, New York, Melbourne: University
of Cambridge Press) uses the term
cumulation for situations like this; a morpheme cumulates various properties (e.g. c
ase and number).
30
We will write the lexical entry for a portmanteau morpheme as follows, with the
understanding that it
displaces the regular morphemes.
(52) GO TO, PLURAL oziit
The singular form will have the following lexical entry.
(53) GO TO yaai
There are two question particles in Madija: ki and ko. The word ki is used if th
e subject of the verb is
feminine, and the word ko is used if the subject of the verb is masculine.
b
In this sense, there is a kind of
agreement going on between the Question particle morpheme and some other element
in the clause. (These
are not portmanteaux since there arent any other morphemes that these forms displ

ace.)
(54) toko ki Did she go?
toko ko Did he go?
The lexical entry for this morpheme would be:
(55) QUESTION PARTICLE ko QUESTION PARTICLE ki
[
Feminine]
[+Feminine]
Key Concept
multiple function formative portmanteau
4.1 Now try it for yourself with French
In French there is a word for to [a], and a word for the (masc. sg.) [lo]. These
two morphemes are displaced
by a portmanteau, [o], when they co
occur. (This does not happen when
[a] to precedes the feminine
singular word for the [la].) Give the lexical entry for the portmanteau.
38
Try stating this in prose as well.
b
This is a slight oversimplification. The conditioning factor is actually the ge
nder of the absolutive of the clause (the
subject of an intransitive clause, the direct object of a transitive clause).
31
4.2 Try it for yourself with Swahili
Give the word structure rule and the lexicon (affixes and stems). (This problem
draws on topics presented in
various of the preceding chapters.)
39
Three of the morphemes have two suppletive allomorphs each.
c
There
is also one portmanteau.
1. ninopcndo I love 17. ninoono I see
2. unopcndo you (sg.) love 18. unoono you (sg.) see
3. onopcndo s/he loves 19. onoono s/he sees
4. tunopcndo we love 20. tunoono we see
5. mnopcndo you (pl.) love 21. mnoono you (pl.) see
6. wonopcndo they love 22. wonoono they see
7. ninompcndo I love him/her 23. ninomwono I see him/her
8. ninomIcto I carry him/her 24. ninomwombo I ask him/her
9. nitopcndo I will love 25. nitoono I will see
10. niIipcndo I loved 26. niIiono I saw
11. sitopcndo I wont love 27. sikupcndo I didnt love
12. hutopcndo you (sg.) wont love 28. hukupcndo you didnt love
13. hotopcndo s/he wont love 29. hokupcndo s/he didnt love
14. hotutopcndo we wont love 30. howokupcndo they didnt love
15. homtopcndo you (pl.) wont love
16. howotopcndo they wont love
c
At least you should treat them as suppletive for the purposes of this exercise.
32
4.3 Try it for yourself again with Swahili
40
The verb form which these data introduce is called the infinitive (so

called because it does not show any
reference to time or tense). The form which is glossed for (someone) is often call
ed the applicative (who
knows why!). Another term you might use (for other forms) is inversive (opposite
). And there is something
else that you should analyze here as a formative that you can just call Final Vo

wel.
Account for all morphemes by providing:
(1)
the word structure rul
e for infinitives, (2) a lexicon for
morphemes which do not vary, (3) rules for the morpheme(s) displaying
variation.
(Assume they are
suppletive.)
Some traditional vowel labels might be helpful to you.
HIGH i
u
MID e
o
LOW
a
For the morpheme(s) showing variation, explain in prose of sufficient
length why you analyze it/them the
way you do (conditioned by syntactic factors, conditioned by phonologic
al factors, conditioned by
morphological factors, or simply completely unpredictable and therefore requirin
g arbitrary classes).
1. kupcndo to love 17. kuIitjo to hide
2. kupoto to get 18. kuIitjuo to reveal
3. kuonzo to start 19. kupikio to cook for (someone)
4. kuIcto to carry 20. kuIipio to pay for (someone)
5. kuIo to eat 21. kubcbco to carry for (someone)
6. kuscmo to speak 22. kuimbio to sing for (someone)
7. kucndo to go 23. kuIungio to close for (someone)
8. kuIipo to pay 24. kupotio to get for (someone)
9. kupiko to cook 25. kucndco to go for (someone)
10. kubcbo to carry 26. kuombco to pray for (someone)
11. kuondiko to write 27. kujonco to sew for (someone)
12. kujono to sew
13. kuIungo to close Extra challenge:
14. kuIunguo to open 28. kuIunguIio to open for (someone)
15. kuIumo to weave
16. kuIumuo to loosen, untie 29. kuIumuIio to loosen for (someone)
33
Feedback for Chapter 4
4.1 French
38
TO, THE (masc. sg.) o
When the preposition o is expected in front of the masculine singular article Io, th
e expected
sequence o Io does not occur. Instead, a single word o appears.
4.2 Swahili
39
Word Structure Rule: (NEGATIVE)
Subject Agr.
Tense
(Object Agr.)
Verb
stem
Lexicon:
1s Subject ni PRESENT no pcndo love
2s Subject u FUTURE to ono see
3s Subject o PAST (see below) Icto carry
1p Subject tu ombo ask
2p Subject m
3p Subject wo
PAST ku / <NEGATIVE> 3s OBJECT mw / __ V
Ii / elsewhere
m / __ C
NEGATIVE h / __ V NEGATIVE, 1s SUBJECT si
ho / __ C
The final vowel of the verb is actually a suffix, but this is not apparent from
the data given.
4.3 Swahili
40

Word structure rule for infinitives:

V
infinitive
Infinitive Prefix
Root
(Inversive)
(Applicative)
Final Vowel
The order of the Inversive and Applicative morphemes is only known from the exam
ples given in (28
29).
Lexicon for morphemes which do not vary:

pcnd love ku
Infinitive Prefix

pot get
u Inversive*

I eat
o Final Vowel

Ium weave

etc.
* An alternative analysis to the one presented here would posit two suppletive a
llomorphs of the Inversive:

uI / __ <Applicative> and
u / elsewhere.
Rules for the morpheme displaying variation:
APPLICATIVE Ii / V ___
c / root with mid vowel
i / (elsewhere)
34

The Applicative suffix has three forms (assuming the l to be part of it rather t
han part of the Inversive
suffix). It has the shape
Ii when it follows a vowel which in the data shown is
only the Inversive suffix.
The analysis given is that the conditioning is phonological. It is also possible
to say that it has the shape
only after the Inversive suffix, in which case the conditioning is morphological
.
Otherwise, the Applicative suffix is
e following roots/syllables which have a m
id vowel, and
i elsewhere
(i.e., following roots/syllables with high vowels or low vowelsnot a natural clas
s, and therefore
appropriately treated as the elsewhere case). This is phonological conditioning
and it adequately handles all
of the cases shown. Therefore it is not necessary to look for other conditions;
it is certainly not a matter of
arbitrary classes.
While it certainly may not be true that the Inversive suffix may co
occur with j
ust any verb of Swahili, there
is no evidence given in the data of any restrictions. The facts discussed above
would not be affected by any
such restrictions.
35
Chapter 5
Morphologically Triggered Rules
Consider the following data from English:
(56) Singular Plural
leaf leaves
knife knives
wife wives
half halves
loaf loaves
Notice that the final consonant of the noun changes between the singular and the
plural. It would seem
unfortunate to have to list each of these nouns in the lexicon with two roots, a
s follows:
(57) WIFE wive / <Plural>
wife / elsewhere

The forms are just too similar to treat in this way. An alternative would be to
choose one of them as more
basic than the other and derive one allomorph from the other. Let us
assume that the singular form is the
basic form, and the plural form of these nouns is created by adding the suffix

s plus a tweak on the conso

nant. We refer to this tweak as a consonant mutation rule. The rule might be state
d as follows (ignoring
the English spelling convention using the silent e):
(58) F/V Rule: Change the last f of a root to v if the root is of the class leaf
, loaf, etc.
We have to indicate somewhere that nouns like leaf are subject to the F/V Rule w
hen they are pluralized;
this rule does not apply to nouns such as puffs, coughs, cliffs, chiefs, waifs,
and fifes. Since this rule is re

stricted to a small subset of nouns only under very restricted circumstances, we
refer to it as a minor rule.
The realization of the plural morpheme in English should therefore look somethin
g like the following:
(59) PLURAL
en with class A (ox, etc.)

with class B (deer, etc.)

s plus F/V Rule elsewhere
We can illustrate the steps we have taken so far with a derivation. A derivatio
n shows the steps from
the input to the output.
a
(60) Word Structure Rule Noun
PLURAL
Lexical Insertion leaf
s
F/V Rule leave
s
Final Output leave
s
A similar kind of rule would be posited to account for the alternati
ons seen in mouse/mice, and
goose/geese. A rule that changes a vowel in this way is often referred to as an
ablaut rule. These rules for
English are also minor rules, as are the ones that change the vowels in some ver
bs sto form the Past tense
(sing/sang, sit/sat).
b
(61) Minor Rule (Ablaut): Change ou to i if the word is louse, mouse, etc.
(62) Minor Rule (Ablaut): Change oo to ee if the word is goose.
(63) Minor Rule (Ablaut): Change i to a if the word is sing, sit, etc.
The zero allomorph of PLURAL in English therefore also triggers a min
or rule for some nouns. For
nouns like deer and moose, the zero allomorph occurs, but since these
roots do not appear in the list of
words which undergo Ablaut, there is no change between singular and plural.
a
This derivation is a bit informal since it still utilizes English spelling conv
entions. Lexical insertion means going to
the lexicon and choosing a Noun, the appropriate plural suffix, etc., and combin
ing them.
b
Remember that for the sake of convenience we are using the English spelling her
e rather than a technical alphabet.
36
(64) PLURAL
en with class A (ox, etc.)

plus Ablaut with class B (deer, moose,goose, louse, etc.)
c

s plus F/V Rule elsewhere

Not all morphologically triggered rules are minor rules; it depends if

they specify a special class of
morphemes in the lexicon. For example, in Seri the passive allomorph
{p
}, which we saw earlier, also
triggers a rule of ablaut on every morpheme that immediately follows it. It does
not refer to any ad hoc class
of morphemes, unlike the English examples above. Among other things, t
he rule changes an o (long or
short) to short a. (Morphologically triggered rules are often rather odd.) This
is illustrated by the following
derivation of the passive verb tpanl.
(65) Word Structure Rule Tense
PASSIVE
Stem
Lexical Insertion t
p
oonl
Ablaut Rule tpanl
Final Output tpanl was it stirred?
A rather unusual case of a morphologically triggered rule in Seri (one of many i
n that language) is one
that happens in a number of verbs when they form the plural. Consider the follow
ing data:
(66) Singular Subject Plural Subject (with suffixes)

sic
scooc
xam grind

zip
zcoop
xam kiss

fiz
fcooz
xam tie knot

ziim
zcoom
t enjoy

ficj
fcoocl wrap oneself with

pii
pcoo
yo taste

sii
scoo
yo smell
In each of these verbs, the vowel of the stem is changed to coo wh
en the stem for a plural subject is
formed. The minor rule would look something as follows, and it would apply only
in the formation of Plural
stems of verbs that are marked to use it (which all have a close front vowel):
(67) Minor Rule: Vowel of stem coo
Underlying forms
In the preceding section we chose one allomorph as basic, and derived
another allomorph from it by
applying a rule to it. This is a very important concept in generativ
e phonology since all allomorphs other
than suppletive allomorphs are derived in this way. It is important to know how
and why one form is chosen
as basic instead of another one. In this chapter we discuss this issue.
In the case of leaf and leaves, we had an alternation between f and v that we ne
eded to account for. We
saw that we could not simply let the Mutation rule apply to any noun ending in f
because of words such as
chief/chiefs (no change). But what if we chose leave as the basic form and deriv
ed leaf from it by doing the
Mutation rule in the opposite direction? That also will not work across the boar
d since there are words such
as cove/coves and wave/waves where we have v in both forms. These fa
cts establish that we have three
groups of words: words whose stems end in f in singular and plural, words whose
stems end in v in singular
and plural, and words that have an alternation between f and v.
But we still have not shown any reason for choosing f as the input to the minor
rule. Why not have a
rule like (68)?
(68) Minor Rule: Change v to f.
One piece of evidence in favor of having the rule change f to v rather than the
other way around is that the

form leaf is used as the base for the adjective leafy. It also appears in the co
mpound leaf
like.
Also, evidence from language acquisition indicates that the rule is le
arned the other way around, that
the basic, or underlying form of the stem in leaves is leaf. Children say leafs
until they learn that this word
undergoes a funny rule. In another hundred years or so, the rule may disappear a
ltogether and the plural of
leaf will be leafs in standard English. (Or it just may persist, as it has!)
c
The exact subclasses of words (some very small) would need to be listed.
37
The underlying form of a morpheme is the form from which all of the
non
suppletive variants
(allomorphs) may be derived. In other cases, the reasons for choosing one form a
s the underlying form are
clearer. This was the case with the Seri plurals which we just saw in chapter 5.
A few forms are repeated
below.
(69) Singular Plural
tepol tepl
oj blacktailed jackrabbit
mojpe mojp
toj saguaro cactus
A minor rule deleting the vowel in the plural is possible; it would
be impossible to formulate a rule
inserting a vowel in the singular because we wouldnt know which vowel
to insert. Therefore the
underlying form of the stem for saguaro cactus is mojpe rather than mojp,
and the underling form for
blacktailed jackrabbit is tepol rather than tepl. Other reasons for making the choic
e of underlying form
are given in later chapters. This issue of choosing one underlying fo
rm of a morpheme from which to
account for allomorphs is one of central importance in phonology and we will spe
nd more time on it.
Key Concepts
minor rules consonant mutation rules ablaut rules
derivation underlying (or basic) form
5.1 Try it for yourself with Seri
Consider the following data. You may recall from chapter 3 that Seri words plura
lize with various suffixes.
Two of these suffixes are represented in the data below:
toj and
oj. But besid
es that fact, the noun stems
sometimes show the presence vs. absence of a vowel in the syllable after the str
essed vowel (which may be
long or short). (Note: the change from qu to c in the word for her child is just a
spelling convention of Seri
similar to that of Spanish.)
Singular Plural
1. sepol sepl
oj shrub (species)
2. cascamma cascamma
toj stink bug (species)
3. tepol tepl
oj blacktailed jackrabbit
4. iquet ict
oj her child
5. icahme icahm
toj abandoned camp
6. xtasi xtasi
toj estuary
7. mojpe mojp
toj saguaro cactus
Is the vowel that appears in the singular stems the same vowel or a variety of v
owels?
41
Which of the following rules is correct?

42
(a) Delete the first vowel in the syllable after the stressed syllable.
(b) Delete the first vowel in the syllable after the stressed syllable in the Pl
ural.
(c) Delete the vowel in the syllable after the stressed syllable in t
he Plural if
the noun belongs to Class Q.
38
5.2 Try it for yourself again with Seri
Provide an account for the following data: word structure rule and lexicon. Ass
ume that the prefix for you
(subject) is in
, and that the prefix for interrogative past is t
. The data ar
e not entirely representative of the
language as a whole; some important data are omitted in order to kee
p the problem manageable at this
stage. The hypothesis that you will probably come up with will not be en
tirely unlike what is true for the
language, however.
43
did you .... it? (transitive) did you ...? (no direct object mentionable)
1. int ko:o into ko:o chew
2. int kokx into kokx chop
3. in tnoix into noix dream
4. in tsop into sop slap
5. in tjip into jip kiss
6. in timoy in tmoy butcher
7. in tinx in tnx yell at
8. in tip in tp straighten
9. in ti:p in tp carry on head
10. in ti:k in tk sow
11. in to: in to: grab
12. in to:om in to:om beg
13. in toix in toix fetch water
14. in tomx
in tomx hunt
15. in tosokim in tosokim comb (hair)
39
Feedback for Chapter 5
5.1 Seri
41
A variety of vowels. This means that the vowel which appears in the singular st
em must be part of the
underlying form since it is not predictable.
42
The third rule, a minor rule which applies only to certain nouns, is correct. T
he other rules apply too
generally and would delete vowels where they shouldnt.
5.2 Seri
43
The word structure rule would be: Subject Agreement Tense (Detransitivizer)
Root
Lexicon: 2
nd
person (singular) subject: n

past interrogative: t

ko:o chew, inx yell at, o: grab, ...
Detransitivizer (causing the omission of the direct object from the clause)
o / __ consonants
/ __ low vowels
plus Ablaut / __ high vowels

Ablaut rule: Change the first vowel of the root to the vowel (the short low equ
ivalent of the high vowel
that it started out as).
40
Chapter 6
Features and Natural Classes
There are other words that display consonant mutations in the plural besides f/v
, although the change is not
shown in the spelling system of English. You will have to listen carefully to he
ar the difference; the square
brackets enclose phonetic transcriptions (transcriptions that include exact deta
ils of pronunciation using a
standard international alphabet):
a
(70) house hou[s]e houses hou[z]es
bath ba[] baths ba[]s
path pa[] paths pa[]s
Again, this is a small group of words (compare these with myth/myths,
death/deaths, douse/douses,
noose/nooses, grouse/grouses.)
There is a similarity between the alternations found here and the f/v alternatio
n seen earlier. The sounds
that occur in the singular forms are pronounced without vibration of
the vocal cords they are voiceless;
those which occur in the plural are all pronounced with vibration of the vocal c
ords they are voiced.
b
(71) Voiceless Voiced
I
v
s
z

It woul be simple (an easy) to generalize the minor rule we evelo

pe earlier to hanle these
alternations as a group if we refer to the voicing of the souns.
(72) Minor Rule: Change the final consonant from voiceless to voice if the wor
is of the
class leaf, house, path, etc.
By referring to the voicing of these souns, we are picking out a single feature
, namely voice, which
they share, an changing its value. This step has a big avantage over a rule th
at simply lists the souns out
(such as change f to v, s to z, etc.). It enables us to state the rule simply. If
the consonants in question were
v, k, an s, an the change was to make them f, t, an z (respectively), a simpl
e rule woul not be possible.
Therefore, we can look at the kin of rule we come up with an evaluate how iff
icult these changes are for
the language learner to acquire. The use of features is part of a s
ystem for measuring the simplicity of a
escription; simplicity is consiere to be a favorable attribute. The use of sy
mbols may be simpler to use in
informal situations, but the use of symbols such as f oes not permit
the straightforwar comparison an
evaluation of competing analyses.
Another fact about the consonant mutations illustrate above shoul not go unnot
ice. The consonants
which unergo the minor rule are all fricatives, souns in which there is just e
nough obstruction in the oral
cavity to cause the air stream to prouce friction upon passing. Souns which fu
nction alike cross-linguis-

tically are calle natural classes. They are single out by specifying
one or few features. For example,
fricatives is one natural class an voiceless fricatives is another natural clas
s (a subset of the former). The
more features that are specifie, the smaller the natural class that
is being esignate. To see this more
clearly, examine the following noun-verb pairs in English. They show a
n alternation which is not always
represente in the spelling system.
a

The fact that both [] and [] are spelle th in the English spelling system probab
ly makes it more ifficult for some
English speakers to even realize that they are in fact saying two ifferent soun
s here. The ifference between [] and
[] is clear in ether an either.
b
The vocal cors (or fols) are membranes that are foun in the lar
ynx an which are manipulate to moify the
airstream. If these are relaxe an separate, a voiceless soun is prouce, li
ke h or f. But if they are pulle together
so that they vibrate when the air passes, a voice soun is prouce
, like a or v. There are other possible configurations of the vocal cors also. See Appenix A.
41
(73) Noun Verb
house house [s], [z]
avice avise [s], [z]
breath breathe [O], []
mouth mouth [O], []
bath bathe [O], []
strife strive [I], [v]
belief believe [I], [v]
grief grieve [I], [v]
thief thieve [I], [v]
You shoul recognize that the feature [voice] is again consistently ch
anging in these pairs. The
voiceless fricative in the noun is replace by its voice counterpart
in the verb. If the noun is consiere
basic (simply an assumption at this time), then one might propose a rule such as
the following that operates
in the erivation of the verb form:
(74) Make the final consonant of the stem [+voice].
But again, it is not just any consonant that participates in this change. There
are no noun-verb pairs in
English such as stick*stig, or pipe*pibe. An it is not just a haphaza
r set of consonants, but a natural
class. The following chart inclues the consonants of English
an th
e feature specifications for two
important features. (The efinitions of these features are not important at this
point; but see Appenix A.)
(75)
p b t 
k g tj  j I v O s z
sonorant
continuant + + + + + + + +
m n I + j w h
sonorant + + + + + +
continuant + + + +
Of these two features, neither one by itself selects exactly the set of consonan
ts that participate in the

noun-verb alternations. The feature [+continuant] comes very close. The

combination [-sonorant,
+continuant] comes even closer; it specifies the natural class of fricatives. If w
e really wante to specify a
smaller set, fricatives that are pronounce in the front of the mouth,
we
woul have to a another
feature. This illustrates the point that we mae above that the fewer the featur
es specifie, the larger the set,
while the more the features specifie, the smaller the set.
Key Concepts
features [voice] natural class
evaluation of simplicity
42
6.1 Try it for yourself (features an natural classes)
Assume the following inventory of souns:
p t k b  g I s j v i o u
Which of the above souns woul be inclue in the natural classes that each of
the following features or
feature combinations efine:
44
1. [+voice]
2. [+labial] (souns mae with the lips as active or passive articulators)
3. [voice], [+labial]
6.2 Try it for yourself with English
The consonants which may appear in the same syllable following the diphthongs [o
w] and [5j] in English
seem to be restricted. Consider the following data:
Consonants which appear: t, d, tj, d, O, , s, z, n, I, +
Examples:
route, uoit, loud, tabloid, crouch, gouge, mouth (n.), mou
th (v.), house, voice, rouse, noise,
town, groin, cowl, boil, hour. (Not all appear after [5j].)
Consonant seuences which appear: nd, st
Examples: mound, joust, hoist
Consonants which do not appear: p, b, k, g, I, v, h, j, , m, j, w
Consonant seuences which do not appear:
c
pk, mp, sp
Using the features shown on the following charts, what feature(s) define class o
f consonants which appear
after [aw], apparently a natural class, as closely as possible?
45
(See Appendix A for definitions of the
features.)
Based on this natural class, what consonants unexpectedly do not occur after the
se diphthongs?
46
p b t d k g tj d j I v O s z
Labial + + + +
Coronal + + + + + + + + + +
Dorsal + +
m n I + j w h
Labial + +
Coronal + + +
Dorsal + +
6.3 Try it for yourself with Misantla Totonac
Assume the following (partial) inventory of segments and features:
I ts tj I s j
sonorant +
continuant + + +
strident + + + + +

lateral + +
c

The word oink is exceptional in that it does have the seuence [5jpk].
43
There are at least two reasons in this language to group the sounds
I, s, and j together as a group,
distinguishing the group from the other sounds in this chart. What fe
ature or features define this natural
class? (Use more than one feature only if necessary.)
47

Feedback for Chapter 6

6.1 Features and Natural Classes
44
[+voice]:
b d g v i o u
[+labial]:
p b I v u
[
voice], [+labial]:
p I
6.2 English
45
The feature [+coronal] defines a class that includes all of the correct consona
nts, plus two.
46
The sounds [j] and [] are included in the class, but do not occur (in the same s
yllable) after these
diphthongs. (Interestingly enough, this is probably due to another factor. They
do not occur after any long
or tense vowel in English, with one exception, leash (and unleash), besides loa
nwords such as microfiche,
uiche, and hashish.)
6.3 Misantla Totonac
46
The feature [+continuant] defines a class that includes the correct sounds. (Thi
s feature would also include
vowels and glides, so in reality another feature such as [
sonorant] is also nee
ded.)
44
Chapter 7
Reduplication
In the preceding chapters we have seen cases where formatives are realized by a
string of sounds, and other
cases where they are realized by changing sounds (ablaut and mutation). In this
chapter we see an example
of a uite different kind of formative.
Consider the following data from Madija:
(76) Verb Noun
kosi whip kokosi whip
hipo eat hihipo food
bo weave bobo shuttle
do pound dodo pestle
dzobi dance dzodzobi dance
koro throw kokoro hook, net
If we try to isolate some common consonants and/or vowels for the morpheme
which changes a verb
into a noun in this language, we find that we are unable to do so.
Rather, the nominalization of a verb is
indicated by the reduplication of the first syllable of the word. For
example, you take the verb for whip
(kosi), take the first syllable of this word (ko) and put it in front of the ver
b, and you have the noun whip.
The word structure rule and lexicon for Madija would include the following:

(77) N

NOMINALIZER
Verb
stem
NOMINALIZER

Syllable Reduplication
a
kosi whip
hipo eat
bo weave
etc.
Reduplication is not always an entire syllable. It may be the first
consonant, the first consonant and
vowel, or something else. Reduplication is found in many languages aro
und the world, with all kinds of
variations on the theme. Discussion of the many various interesting co
mplications which reduplication
presents is not relevant here, however.
b
The point is that reduplication is another type of realization of a
morpheme that is rule governed.
Key Concept
reduplication
a
The fact that it is the first syllable that is reduplicated need not be indicat
ed in this rule since the word structure rule
indicates that the nominalizing morpheme is a prefix.
b
Relevant work to consult is: E. I. Broselow and J. McCarthy (1984) A Theory of
Internal Reduplication, Linguistic
Review 3; A. Marantz (1982) Re Reduplication, Linguistic Inuiry 13:435
82.
45
7.1 Try it yourself with Tohono O'odham
Give the word structure and the lexicon for the following data, inclu
ding the lexical entry for the plural
morpheme (which may be written in simple, concise prose).
48
Singular Plural
1. jud um jujud um bear
2. komkcud kokomkcud desert tortoise
3. gogs gogogs dog
4. kui kukui mesuite tree
5. hodoi hohodoi rock
6. koIhoi kokoIhoi fence
7. piIso pipiIso blanket
Feedback for Chapter 7
7.1 Tohono O'odham
48
Word Structure Rule Noun (PLURAL)
Noun
stem
The stems for the nouns are identical to the singular forms.
The plural morpheme is a reduplicative prefix that consists of the first consona
nt and vowel of the stem.
46
Chapter 8
Word or Affix?
We have been discussing the realization of morphemes within words in the past fe
w chapters. But how does

one know that a given morpheme is to be accounted for by a syntactic rule or a w

ord structure rule? You
may have thought that this is a simple matter, but that is because
you have been relying on the spelling
system of English.
Especially problematic are unstressed morphemes which follow the noun. We assume
that morphemes
which are part of words are properly accounted for by word structure rules, and
that morphemes which are
part of phrases are accounted for by syntactic rules.
Case suffixes in Greek and Latin, like the plural suffix
s of English, are part
of the nouns. Therefore
they should be described by word structure rules.
Morphemes like the English prepositions in and to precede noun phrases
and are part of the
prepositional phrases in which they occur. Their placement should be described b
y syntactic rules.
Morphemes like the s in the Queen of Englans hat, attach to noun phr
ases; they are not
grammatically part of the preceing wor. (However, s oes interact wit
h the preceing wor
phonologically as if it were an affix. Notice that the wors juges an juges ar
e phonologically ientical.)
The position of such morphemes, known as phrasal affixes, is escribe by syntac
tic rules.
We posit the following test to ecie if an unstresse morpheme is to be ac
counte for by the syntax
(like in or s ) or by the morphology (like -s).
(78) If an inepenent wor, or a phrase, can appear between two morphemes, the
morphemes
in question are not part of the same morphological wor.
a
This test works positively for s, as in the Queens hat vs. the Queen [of Englan]s
hat; an for in time vs.
in goo time. It oes not give a positive result for the plural morpheme: ays,
*ay [gone by]s.
Other examples
The wors the an a(n) are typically unstresse an
pronounce with
the wor which follows them. Are
they prefixes, as in the-stuent an a-lesson, or are they separate wors, as in
the stuent an a lesson? The
intervention test given above establishes the latter: the careful stuent, a sim
ple lesson. This oes not mean
that there is no phonological interaction or epenency between the morphemes in
question. We know, for
example, that the inefinite article is an when the next morpheme begins with a
vowel an a when it begins
with a consonant. Likewise, the efinite article is th[i] when the next morpheme
begins with a vowel, an
th[e] when the next morpheme begins with a consonant.
Consier the morpheme za in Maija. It follows nouns, as in expressi
ons such as toniniza in her
nest (possible wor breaks purposely omitte). Now is za a Locative suffix which
is part of the wor her
nest, or a Locative morpheme which is part of the noun phrase in her nest (either a
s a phrasal affix or a
postposition)?
b
Note that za follows the ajective onii other in the expression tonini
onii za (her.nest

other in) in her other nest (wor breaks purposely inclue now). This
is inicative that it shoul be
analyze as part of a phrase rather than part of a wor. Therefore the position
of the morpheme za shoul
be accounte for by a syntactic rule rather than a wor structure rule.
In Maija there is also a morpheme eni which means plural. One might easily think
that it is just a
suffix like -s in English from expressions such as maihaeni the (Ma
ija) people an tamineeni the
chiefs (possible wor breaks purposely omitte). The fact that eni fol
lows the relative clause in the
following expression inicates that it is part of the phrase an not
part of any wor: bani [tokehenahari]
eni (animal who.became plural) the ones who became animals.
a
b

Exceptions to this criterion are presente by some cases of noun incorporation.

A postposition is the counterpart to preposition which is foun in

many languages. Rather than preceing its
complement, it follows the complement.
47
This topic is not simple an we are purposely omitting iscussion of
morphemes which have been
labele clitics in the literature since these issues woul lea us far astray of o
ur central concern.
c
Key Concepts
part of wor vs. part of phrase postposition
Postscript for Teachers
The issue of the istinction between wors an affixes is not trivial. We have n
ot aresse all of the issues
here, obviously. We have ignore the phenomenon of noun incorporation, for examp
le. We feel that in the
past unue attention has been given to the question of whether a morpheme may co
nstitute an utterance in
an of itself. This criterion may provie a goo test for what may be an inepen
ent utterance (naturally),
but not for what may be a wor. Some language analyses are unuly complicate by
trying to account for
certain unstresse morphemes through the morphological component rather
than through the syntactic
component. The criterion we are trying to use oes not say whether a
particular wor is stresse or
unstresse, or whether it may lean on or interact with another wor phonological
ly.
8.1 Try it for yourself with Isthmus Zapotec
Discuss whether the morpheme ko is part of a wor or part of a phr
ase. (No wor breaks have been
given.)
49
1. ngi u man
2. ngi uko that man
3. ngi uro?ko that big man
4. Icju rabbit
5. Icjuko that rabbit
6. Icjuwi niko that little rabbit
8.2 Try it for yourself with Houailou
Discuss whether the morphemes rrc (that) an no (subject case) are parts of wor
s or parts of phrases an

show your interpretation of the facts by writing the ata again accoring to you
r analysis. (No wor breaks
are given.)
50
1. bwc1 woman
2. bc ori ol
3. bwc1rrc that woman
4. bwc1bco ri rrc that ol woman
5. kom1 man (not subject)
6. nokom1 man (subject)
7.
cyo s other, some
8. nocyo skom1 (the) other man (subject)
c

See Arnol M. Zwicky (1985) Clitics an Particles, Language 61:283-305; an Arn
ol M. Zwicky an Geoffrey K.
Pullum (1983) Cliticization vs. Inflection: English n t, Language 59:502-13.
48
8.3 Try it once more with Damana
Discuss whether the morphemes [go| (Ergative case), [mbo] (Locative case), an [
nc] (Genitive case) are
parts of wors or parts of phrases. (The names of these cases are irrelevant to
the problem. No wor breaks
have been given.)
51
1.
umogogo lion
2. onogo s/he
3. rogo I
4. suigo bir
5. suibonjigo white bir
6. ipgunombo on the roa
7. ipgunozukwcgombo on the goo roa
8. mcnonc the woman s
9. mcnotjukkwcgonc the big woman s
49
8.4 Try it for yourself with Corongo Quechua
Discuss whether the morphemes which are not nouns (like oven) or not ajectives
(like re) are affixes or
wors. Your iscussion shoul inclue the wors like this, that, one, an two.
52
Write the Corongo Quechua
wor or phrase meaning in those enormous vehicles accoring to your a
nalysis. (The ata are given in a
non-phonetic transcription.)
1. urnu oven
2. urnuta oven (when irect object)
3. urnuman into the oven
4. urnumanxa into the oven (topicalize)
5. puka re
6. utsu hot pepper
7. pukautsuta re hot pepper (when irect object)
8. ka:rru vehicle
9. ka:rruta vehicle (when irect object)
10. ishkeyka:rruta two vehicles (when irect object)
11. ukka:rrullata just one vehicle (when irect object)
12. akshu potato
13. akshukuna potatoes
14. ishkeyakshu two potatoes
15. akshukunaxa potatoes (topicalize)
16. ishkeysurkullapik from just two furrows

17.
18.
18.
18.
19.
20.
21.
22.
23.

24.

keykajunman in this box

keylugar this place
lugar place
tseylugar that place
tseylugarman at that place
tseyna:nikunataxa those roas (topicalize, irect object)
tseyatusaxka:rrukunachow in those enormous vehicles
atusax enormous
atun large
ukatunkorona one large crown


Dont try to relate this wor to the wor for enormous.
50
Feeback for Chapter 8
8.1 Isthmus Zapotec
49
Since the morpheme [ko] may be separate from the nouns for man an rabbit by th
e wors big an little,
we consier [ko] to be part of the noun phrase.
8.2 Houailou
50
Since the morphemes [rrc| an [no| may be separate from the nouns by other wor
s, we shoul consier
[rrc| an [no| to be parts of the noun phrase. The phrases woul be transcribe
as [bwc1 rrc|,
[bwc1 bco ri rrc|, [no kom1|, [no cyo s kom1|, etc.
8.3 Damana
51
Since the morphemes [go], [mbu], an [nc] may be separate from the nouns by othe
r wors, we shoul
consier these morphemes to be parts of a phrase. (They are phonologically epen
ent in some way, as a
matter of fact, but their istribution is accounte for by the syntax, not by wo
r structure rules.) The phrases
woul be transcribe as [umogo go|, [ro go], [sui bonji go|, [ipguno mbo|,
[ipguno zukwcgo mbo|, [mcno tjukkwcga nc|, etc.
8.4 Corongo Quechua
51
There are no cases where the following morphemes are separate from the nouns by
other wors:
ta (Direct Object), man (into, at), xa (Topicalizer), kuna (Plural), lla just,
pik from, an chow in.
This oes not mean that these morphemes are suffixes, although one or more of th
em may be. We simply o
not have irect evience from these ata (if any) in favor of analyzing them as
separate wors.
On the other han, the numbers an emonstratives may be separate from the noun
s by an ajective.
Base on this evience, it is arguable that the emonstratives an numbers are s
eparate wors.
The wor meaning in those enormous vehicles coul be written in one of three way
s, epening on how one
ecies about the post-nominal elements:
tsey atusax ka:rru kuna chow
(those enormous vehicle Plural in)
or
tsey atusax ka:rrukuna chow
(those enormous vehicles in)
or

tsey atusax ka:rrukunachow (those enormous vehicles/Locative)

At this point, we simply o not know what is the correct (or best) analysis.
51
Summary an Review Questions for Section 1
In these chapters we have consiere various kins of rules which are neee to
account for the istribution
an realization of morphemes within wors.
Wor structure rules escribe the orer of morphemes within wors. (Un
er the appropriate analysis,
they may also escribe the internal, layere structure of wors).
The lexicon contains items which can be inserte into the appropriate positions
in the structure. Most
lexical items have a single unerlying form, an where they o not, explicit rul
es must give the suppletive
allomorphs an their istribution. The istribution may be relate to
syntactic, morphological, an/or
phonological factors, an/or arbitrary wor classes. (Recall that the latter are
the last resort.)
Sometimes a morpheme is realize by effecting a change on an existing
form: the mutation of a
consonant, the ablaut of a vowel, or reuplication of some material from the wor
.
A complete escription of a language must therefore inclue, among oth
er things: (1) wor structure
rules to account for the way morphemes are combine into wors, (2)
a lexicon to give an inventory of
morphemes, incluing rules for suppletive allomorphs, an (3) rules to account f
or changes in morphemes
which are triggere by other morphemes, if there are any such rules.
If you have mastere the material in this section, you shoul be able to examine
ata an
a. ientify formatives
b. write an unerstan rewrite rules
c. construct paraigms
. write an unerstan wor structure rules that escribe the orer of morphe
mes within
wors
e. provie an unerstan lexicons of stems an affixes
f. provie an unerstan lexical entry rules for suppletive allomorphs; these
rules make
reference to phonological, syntactic, morphological, or other factors (incluing
arbitrary
classes)
g. unerstan an use the abbreviations C an V
h. explain an illustrate the notions multiple function formative an portmant
eau
i. explain an illustrate the concept of morphologically triggere rule
j. ientify common wor class istinctions such as inalienable vs. alienable n
ouns,
transitive vs. intransitive verbs
k. be able to use the feature [voice] to istinguish souns
l. explain the iea of natural class
m. explain the iea of unerlying form
n. unerstan the iea of reuplication
o. be able to argue that a morpheme is an affix or a separate wor
For Further Reaing:
Aronoff, Mark. 1976. Wor formation in generative grammar. Linguistic Inquiry Mo
nograph 1. Cambrige,
Mass: MIT Press.
Carstairs, Anrew. 1988. On some implications of phonologically conitio

ne suppletion, Yearbook of

Morphology 1.
Dressler, Wolfgang. 1985. Morphonology: The ynamics of erivation. Ann Arbor, M
ich., Karoma.
52
Hammon, Michael an Michael Noonan. 1988. Theoretical morphology: appro
aches in moern
linguistics. New York, Lonon, Acaemic Press.
Matthews, P. H. 1974. Morphology. Cambrige, Lonon, New York: Cambrige Univers
ity Press.
McCarthy, John. 1981. A prosoic theory of nonconcatenative morphology.
Linguistic Inquiry 12:373418.
a
Nia, Eugene. 1949. Morphology: the escriptive analysis of wors. Ann
Arbor: University of Michigan
Press.
Review Questions
The following questions are to help you review the material in the p
receing section.
53
The numbers in
brackets refer to the most relevant page number which eals with the topic.
1. ____________ is the part of linguistic theory that eals with the positioning
of wors. [1]
2. The list of wors an meaningful parts of wors is calle the _________ of th
e language. [1]
3. ____________ is the part of linguistic theory that eals with the structure o
f wors. [2]
4. The wor ogs is compose of two functional pieces calle __________. [2]
5 NP (Art) (AP) N is an example of a ___________ _______. [1]
6. The phonology of a language eals with the organization of the _______ of tha
t language. [2]
7. Wor structure rules inicate the orer of ______________ in wors. [2]
8. A collection of wors which share morphemes an which are organize
in rows an columns is
calle a _______________. [8]
9. When a morpheme has more than one shape, it is sai to have more
than one
__________________. [13]
10. A rule such as
3 POSS a- with kinship nouns, i- elsewhere
is inclue in the lexicon of a language to hanle __________ _________. [13]
11. -en an -s are ____________ ____________ of the morpheme for PLURAL in Engli
sh. [13]
12. The lexical entry of a morpheme may refer to various facts, inclu
ing arbitrary classes,
phonological information, morphological information, an syntactic information.
Ientify which of
the following is of which type:
(a)
X Y / ___ consonants ___________ [16]
Z / ___ vowels
(b)
X Y / ___ transitive clauses ___________ [15]
Z / ___ intransitive clauses
(c)
W X / ___ Class I ___________ [14]
Y / ___ Class II
Z / ___ Class III
()
X Y / ___ PASSIVE ___________ [18]
Z / ___ elsewhere
13. An affix which simultaneously inicates number, gener, an case is

an example of a
____________ ___________ formative. [29]
a

There is a kin of morphology which we have not consiere, which

is important for unerstaning
languages such as Arabic an Hebrew. This is iscusse in McCarthy 1981 an much
later work.
53
14. A rule which is cause to apply by the presence of a morpheme i
s a _________________
___________________ ___________________. An example is the rule changing
f
to v in the
plural of certain wors (e.g. leaf / leaves). [35]
15. An explicit emonstration of the application of rules, starting from the inp
ut an ening with the
output, is calle a ___________________. [35]
16. The basic form of a morpheme, which serves as the input to the
rules of the phonology, is the
_____________________ form. [36]
17. Some rules affect one characteristic of a soun. For example, the
change from f to v in certain
nouns of English when they are pluralize is a change in one _______
________, namely
_________________. [40]
18. Groups of souns which function alike cross-linguistically because they shar
e certain features are
calle _________________ ___________. [40]
19. (T or F) The larger the natural class, the fewer the features are neee to
specify it. [40]
20. (T or F) One of the reasons features are use in phonological analysis is t
hat by using features we
are able to capture generalizations where informal notation using segme
nts often fails to o so.
[40]
21. Different forms of the same morpheme are calle __________________. [13]
22. A special kin of multiple function formative that isplaces two or more oth
er formatives is calle
a _______________________. [29]
Feeback on Review Questions
53
1. syntax 2. lexicon 3. morphology 4. morphemes 5. rewrite rule
6. souns 7. morphemes 8. paraigm 9. allomorph
10. suppletive allomorphs 11. suppletive allomorphs
12. (a) phonological (b) syntactic (c) arbitrary classes () morphological
13. multiple function formative 14. morphologically triggere rule
15. erivation 16. unerlying 17. feature, voice 18. natural classes
19. T 20. T 21. allomorphs 22. portmanteau
54
Section 2
Phonological Rules:
Assimilation
In this section we begin to look at phonology proper. We assume that morphologic
al rules have inserte all
of the morphemes in a form that might be thought of (informally) as a string of
souns, as shown below.
Syntactic rules yiel: Noun Verb Noun
Lexical insertion an morphological rules yiel:
VAL SELL-3SG SHOE-PLURAL

vI sIz juz
At this point, phonological ruls apply. Phonological ruls account for
th xact pronunciation of th
morphms sinc thr is vidnc that som phonological information ab
out morphms is not stord in
th lxicon.
W hav as a goal that any kind of prdictabl information about th
phonological shap of a
morphm should b xtractd from its lxical rprsntation and statd in th p
honology. For xampl, th
fact that th first consonant of Val is voicd and not voiclss is
not prdictabl from anything ls
English has an f as wll as a v. Thrfor this fact must b includd in th lx
ical rprsntation of Val. But
th fact that th l is voicd and not voiclss is prdictabl (within English)
sinc th sonorants of English
ar all voicd; th fact that it is voicd should not b includd in th lxical
rprsntation.
55
Chaptr 9
Voicing Assimilation
In th prvious sction on morphological ruls w xamind primarily a
llomorphs which wr quit
diffrnt from ach othr (although that is not ncssarily th cas
with suppltiv allomorphs). W now
look at allomorphs which tnd to b vry similar, and w s that t
hy ar proprly tratd in a fashion
which is diffrnt from that of suppltiv allomorphs.
Considr again English plurals, but this tim w look at th prcis
pronunciation of th final
consonants (bracktd bcaus of th phontic transcription).
(79) caps [ps] cabs [bz]
fats [ts] fads [dz]
books [ks] rags [gz]
graphs [fs] gravs [vz]
myths [s] times [mz]
pins [nz]
pills [lz]
We see here that PLURAL has another allomorph which we have been ignoring up unt
il now: [z]. Shall we
simply add this allomorph to our list of suppletive allomorphs for PL
URAL?
This would not seem
appropriate for at least three reasons. First, the allomorphs [z] and [s] are ve
ry similar phonologically; they
differ only in the feature voice. We refer to them as phonological allomorphs. S
uppletive allomorphs are
generally more distinct. Second, the feature voice by which they differ is also
the feature that distinguishes
the consonants which precede them. [s] follows voiceless consonants, an
d [z] follows voiced consonants.
Third, there are other morphemes in English which show a similar set
of allomorphs, although they have
nothing to do with plural.
(80) Third person singular, present tense
sips [ps] grabs [bz]
lets [ts] rids [dz]
looks [ks] drags [gz]
laughs [fs] raves [vz]
rhymes [mz]
grins [nz]
sells [lz]

(81) Possessive
cap's [ps] cab's [bz]
cat's [ts] fad's [dz]
Jack's [ks] Greg's [gz]
leaf's [fs] Marv's [vz]
Keith's [s] Jim's [mz]
Ann's [nz]
Bill's [lz]
If the allomorphs of 3SING, POSS, and PLURAL are all treated as supp
letive, the description of English
would have to contain rules for their realization which are embarrassi
ngly similar. If something has to be
said in exactly the same way twice or three times in the grammar/phonology of a
language, we start to look
for an alternative solution which avoids the repetition.
To account for the distribution of such allomorphs, the strategy in g
enerative phonology has been to
propose a single underlying form and provide phonological rules that a
djust this form according to the
context in which it occurs, as necessary. In the case of Plural, one of the unde
rlying forms which we have
given earlier is
s. But now we need to make this underlying form explicit and n
ot continue to use English
spelling conventions (which mask the [s]/[z] distinction). We need to
choose one of these allomorphs (or
perhaps something else) as the underlying form. For now, we will choo
se [z] (reasons will come later).
Therefore the lexicon includes the following rule:
56
(82) PLURAL
en with class A

plus Ablaut with class B

z plus F/V Rule elsewhere
Using this rule and other items in the lexicon of English, we put t
ogether words such as the following,
putting curly braces around underlying forms:
(83) Underlying Forms
tags tacks
{tgz} {tkz}
Underlying forms of words contain all of the morphemes in the words
juxtaposed, and each of the
morphemes appears in its underlying form.
If we were to pronounce the morphemes {tg} and {z} together, the result would be
just like it needs
to be phonetically: [tgz]. But of course, the pronunciation of tacks i
s not [tkz]; it is [tks]. So a
phonological rule is necessary.
As we have seen, the s allomorph occurs after voiceless consonants. This fact i
s important. Evidently
the z is becoming more like the preceding consonant; it is assimilating to it. T
he vocal cords, which are not
vibrating during the pronunciation of the k, are not reactivated durin
g the pronunciation of the next
consonant, so the plural suffix is pronounced voiceless, as an s.
We can make the following generalization for English:
a
(84) A consonant assimilates in voicing to an immediately preceding consonant.
This is a rule of voicing assimilation. As a phonological rule, it refers to fea
tures of sounds, rather than to
particular morphemes. And we have stated it as generally as we can; it
refers to consonants generally rather than to a specific class of

consonants such as fricatives or sibilants. A first guess at a rule

may
be incorrect, but it is generally preferable to start out with too general
a rule than too specific a rule.
To show how the rule works in the phonology of English, we set up
a derivation, as shown below.
When a rule applies, the output of the rule is indicated. A rule may apply vacuo
usly; the output may be the
same as the input. When it does not apply because the conditions are not met for
it to apply, then a line is
inserted to show this.
b
(85) Underlying Form {tg
z} {tk
z} {tk}
Voicing Assimilation tg
z tks
Phonetic Form [tgz] [tks] [tk]
As one can see, the combination of appropriate underlying form and ap
propriate phonological rule
yields the correct phonetic results. We therefore conclude that the ph
onology of English includes a rule
which is something along the lines of (84).
c
Consonants may become voiced in the context of vowels, especially when
between vowels. In
Japanese, for example, consonants are voiced in compounds such as the following:
a
This rule makes incorrect predictions as stated here. See the discussion in cha
pter 14.
b
Note: a phonological rule must always be attempted on all forms. A
phonological rule is meant to have a general
application. If the rule is formulated in too broad a fashion and wo
uld make incorrect claims, it must be made more
specific in an appropriate way or else reformulated entirely.
c
The rule is somewhat more complicated, however, since the suffix
st
er does not show voicing allomorphs at all:
mobster (not mob[z]ter), gangster, roadster. Similarly, the suffix
ship does no
t show voicing allomorphs: friendship,
readership.
Rule of Thumb: When writing a rule, start out
with as simple a formulation as possible. Add
extra features only as necessary to prevent it
from making incorrect claims.
57
(86) iro + komi [irogomi] colored paper
color paper
jo + sokIro [jozokIro] blossoms at night
night cherry
mizI + scmc
[mizIzcmc] water torture
water torture
jomo + tcro [jomodcro] mountain temple
mountain temple
jI + tooII [jIdooII] boiled tofu
hot water tofu
Vowels can also assimilate to neighboring consonants in voicing. In Tataltepec C
hatino, for example,
an unstressed vowel becomes voiceless when it occurs between voiceless
consonants. The word /ki'su/
avocado is pronounced [ki'su], with a voiceless i.

Similarly, in Mokilese, the vowels i and u tend to be pronounced voi

celess when they occur in
unstressed syllables between voiceless consonants: [ki'sa] we two, [sup'wo] pile
of firewood.
Key Concepts
phonological allomorphs phonological rule derivation
assimilation of a sound to its context
9.1 Try it for yourself with Lithuanian
Examine the following data. (Morpheme cuts have been made for you.)
1 Sg. Present 1 Sg. Past Freuentative
1. iIjip
u iIjib
dovo
u get in
2. djirb
u djirb
dovo
u work
3. suk
u sug
dovo
u turn
4. vjc
u vjc
dovo
u transport
5. injcj
u injc
dovo
u bring in
You will have noted that some verb stems end in voiceless consonants in the firs
t column, but none of them
do in the second column. We want to set up underlying forms for these verbs
and provide a phonological
rule that will account for the allomorphs. The rule should be true for all of th
e data, not just for some verbs.
Which of the following solutions will work, using the verb turn to illustrate? W
hy?
54
(a) The underlying form of the verb is {sug} and the rule is: A co
nsonant becomes voiceless
when it precedes a vowel.
(b) The underlying form of the verb is {suk} and the rule is: A consonant assim
ilates in voicing to
an immediately following consonant.
Fill in all of the boxes below, as shown in (85) in this chapter,
to complete the derivation of the words
shown.
55
58
TURN
1 SG TURN
FREQUENTATIVE
1 SG
Underlying Form
Phonological Rule:
(name)
Surface
[suku] [sugdovou]
9.2 Try it for yourself with Yatzachi el Bajo Zapotec
You may assume that the underlying forms of the nouns are the forms
that appear without a prefix. The
prefix {
}, which is a voiceless retroflex sibilant, is added when the
noun is possessed (it turns the noun
from a noun which cannot co
occur with a possessor to one which must co
occur wi
th a possessor).
Unposssessed Possessed
1. bej pej handkerchief
2. tos tos cup
3. do to cord, string
4. koj koj pig
5. gonn konn ox
6. sij sij chair
7. zo so bean
8. jib jib goat
9. in jin work
10. il jil tortilla griddle
Which of the following rules is the best one to account for the allomorphy displ

ayed by the stems?

56
(a) b changes to p, d to t, g to k, z to s, and = to 6 when they follow .
(b) A consonant becomes voiceless when it follows .
(c) A consonant becomes voiceless when it follows a fricative.
(d) A consonant assimilates in voicing to an immediately preceding consonant.
Provide a derivation, as shown in this chapter, for the two forms of
bean. Use the label POSS for th
prfix.
57
59
9.3 Try it for yourslf with Copainal Zoqu
Two suffixs hav bn undrlind in th following data. On is th incompltiv s
uffix, and on indicats
plurality of th subjct (first and scond prson only).
1. muspo s/h can
2. wanbo s/h sings
3. tj:kpo s/h dos it
4. n:mbo s/h says
5. minbo s/h coms
6. pojohpo thy run
7. tj:kjohpo thy do it
8. mindomu you (pl.) cam
9. ndz:ktomumis you (pl.) did it
You ar abl to formulat a phonological rul to account for this allomorphy v
n without dciding on th
undrlying forms of ths suffixs. That is, whthr it is undrlyingly {po} or
undrlyingly {bo} maks no
diffrnc at this point.
Provid a phonological rul that accounts for th allomorphy of th morphms wh
ich ar undrlind.
58
9.4 Walmatjari
Examin th Walmatjari data st in Appndix F. Undr what conditions ar consona
nts voicd or dvoicd?
What rul is ndd to account for voicing altrnations?
59
60
Fdback for Chaptr 9
9.1 Lithuanian
54
Solution (b). Th first solution dos not work for all of th data. Not all of
th stm
final consonants
bcom voiclss whn thy prcd a vowl. Solution (b) also maks som sns i
n that it is an
assimilation rul (unlik th first). Th consonants ar bcoming mor lik ach
othr (voiclss bcoms
voicd in th contxt of voicd). This is not tru of Solution (a).
55
TURN
1 SG TURN
FREQUENTATIVE
1 SG
Undrlying Form
{suk
u} {suk
dovo
u}
Phonological Rul:
Voicing Assimilation








sugdovou
Surfac
[suku] [sugdovou]
9.2 Yatzachi l Bajo Zapotc

56
Th bst rul is (d). Rul (a) is a list and not a rul. Ruls (b) and (c) ar
still too spcific and dont
mntion anything about th fact that th first consonant is voiclss, vn thou
gh this is what is rlvant to
th rul. Rul (d) follows th rul of thumb givn in this chaptr in liminatin
g as much dtail as possibl
and kping only what is absolutly ncssary to show what is going on and to ma
k corrct statmnt
about th data.
57
ban POSS
ban
Undrlying Form {zo} {
zo}
Voicing Assimilation so
Surfac Form [zo] [so]
9.3 Copainal Zoqu
55
Th rul would b th sam as w hav sn alrady: A consonant assimilats in
voicing to an immdiatly
prcding consonant.
9.4 Walmatjari
59
A consonant assimilats in voicing to an immdiatly prcding nasal consonant.
61
Chaptr 10
Choosing Undrlying Forms
Whn a morphm has two (or mor) allomorphs which ar rlatd by ph
onological rul, on of thm is
usually chosn as th undrlying form. (It is also possibl that th
undrlying form is not quivalnt to
ithr of th surfac forms.) Th qustion which must thn b dcidd is which o
f th allomorphs is th un

drlying on? Th answr is simpl to articulat but not always asy to prform
: choos as undrlying form
th form from which th surfac allomorphs can b most asily drivd.
In this chaptr w giv som prliminary instructions as to how undrlying forms
ar chosn. But first
w displ som common misconcptions. Th undrlying form is:
not ncssarily th most common form of th morphm
not ncssarily th morphm as it occurs in isolation
(if it dos in fact occur in isolation at all)
not ncssarily any on of th surfac allomorphs
not ncssarily a complt syllabl.
Now for som positiv advic: if on allomorph has a fatur which cannot b xp
laind by its contxt,
that fatur is probably part of th undrlying form of th morphm.
For xampl, th fact that th plural suffix in English is a fricativ, rathr t
han a stop, is not xplainabl
from th contxt. Similarly, it is not xplainabl from th contxt t
hat it is pronouncd with th tongu
rathr than with th lips (it is s rathr than f ). So th undrlying form of th
 plural suffix must includ such
information.
So how dos on choos btwn th voicd
z and th voiclss
s (
or som othr shap) as th
undrlying form of th plural suffix? Th allomorph
s always follows voiclss
sounds, and th allomorph

z always follows voicd sounds. Eithr way it looks lik th contxt could b r
sponsibl for th voicing
(or lack of). In this cas, othr facts nd to b takn into considration, as
w will s in latr chaptrs.

At this point w also wish to introduc an important division btwn

sounds which shows up
rpatdly in phonology. Sounds ar classifid as bing ithr obstrun
ts or sonorants. Obstrunts ar
thos sounds which ar producd with nough obstruction in th vocal tract to si
gnificantly incras th air
prssur in th mouth. Mor air prssur is rquird to produc such
sounds, which includ th fricativs
(such as f or s), and stops (such as p and t). Nasal stops ar no
t obstrunts bcaus th opn nasal cavity
allows air to scap through th nos, kping air prssur lowr. Sonorants, on
th othr hand, involv lss
obstruction and ar producd with a rsonanc in th vocal tract. Th
y includ nasals, liquids (such as l),
and vowl
lik sounds (such as j or i or a). A sound is classifid
as ithr [+sonorant] or [
sonorant].
(Crtain sounds, such as glottal stop, hav bn difficult to classify
. In som languags thy appar to
function as sonorants, and in othrs as obstrunts.) Sonorants ar usually voic
d (all languags hav voicd
vowls; no languag has only voiclss nasals). Obstrunts ar mor commonly voi
clss. Of cours, many
languags hav voicd stops or fricativs, but if thy do, thy also
usually hav voiclss stops and
fricativs.
Ky Concpts
choosing undrlying form sonorant vs. obstrunt [sonorant] (th fatur)
62
10.1 Try it for yourslf with Pngo
Th following stms hav two allomorphs ach. On nds in a voicd consonant and
th othr in a voiclss
consonant.
Grund 2 sg. Imp. 3 sg. Past
1. tu:b
di tu:b
o tu:p
ton blow
2. tog
di tog
o tok
ton stp on
3. ro:g
di ro:k
o ro:k
ton offr worship
4. hi:b
di hi:p
o hi:p
ton swp
Complt th following obsrvations as to th distribution of ths allomorphs.
60
(a) Bfor voiclss consonants (lik th suffix
tan), th allomorph which n
ds in a ______
consonant always occurs.
(b) Bfor voicd consonants (lik th suffix
di) , th allomorph which nds i
n a ______
consonant always occurs.
(c) Bfor vowls, _________________________________.
Of ths thr contxts (a
c), which prmits th undrlying form of th stms to
show throughis th last
likly to b causing som chang to th stm?
61
What is th phonological rul which accounts for th voicing altrnations?
62
Giv drivations for all thr forms of th word for swp.
63
Provid a pros account of what is going on hr as it might b wr
ittn for a grammar of th languag
intndd for taching Pngo to forignrs who ar not linguistically traind, bu
t who do know about roots
and suffixs and voiclss/voicd consonants.
64

63
Fdback for Chaptr 10
10.1 Pngo
60
(a) voiclss; (b) voicd; (c) both kinds of allomorph occur.
61
Th undrlying form is sn in th pr
vocalic contxt. In th othr contxts,
th voicing of th stm
final
consonant is bing affctd by th following consonant. Whn th undrlying form
prcds a vowl, th
voicing of its final consonant is not affctd.
62
Voicing Assimilation: A consonant assimilats in voicing to an immdiatly foll
owing consonant.
63
UF {hi:p
di} {hi:p
o} {hi:p
ton}
Voicing Assimilation hi:bdi
hi:pton
SF [hi:bdi] [hi:po] [hi:pton]
64
A vrb root in Pngo may nd in ithr a voiclss or voicd consonant. This is
sn by looking at th
imprativ form, which has th suffix
a: hi:pa swp! and tu:ba blow!. Whn a cons
onant
initial
suffix is addd, th root
final consonant changs to agr in voicing with that
suffix
initial consonant:
hi:bdi swp (grund), and tu:ptan blow (3 sg. Past).
64
Chaptr 11
Plac Assimilation (Nasals)
In th prvious chaptr w saw instancs of whr th valu of th fatur [voic
] of on sound affctd that
fatur in anothr sound. In this chaptr w xamin cass whr svral fatur
s act as a group. Ths ar
th major plac faturs.
Som sounds (such as [p] and [f]) ar pronouncd with th lips; thy
ar said to b labial. Som
consonants (such as [t], [s], and [j]) ar pronouncd with th front of th tong
u; thy ar said to hav th
fatur coronal (as in crown of th tongu). Yt othr sounds (such as [k] and [g]
) ar pronouncd with
th back of th tongu; thy ar said to hav th fatur dorsal. Ths thr la
bls ar th important plac
labls in phonology. (Finr distinctions ar built on thm; for xampl, th dif
frnc btwn bilabial and
labiodntal is a subdivision undr labial. S Appndix A. Phonological faturs
ar diffrnt from phontic
labls in that th formr forc a coars gnralization first (labial, coronal,
dorsal) and thn allow for dtails
to b addd as ncssary.)
With that background, considr th following forms from Sri (in broad phontic
transcription):
(87) imii who dosnt har it
impii who dosnt tast it
imit who dosnt stand up
insii who dosnt smll it
intis who dosnt point at it
ipjmx who dosnt div
ipjoo who dosnt own it
ipkoo who dosnt look for it
iyop who dosnt dig it

Th ngativ prfix has fiv allomorphs hr: [m], [n], [p], [p], and
[]. Sinc ths ar similar
phonologically to ach othr, w suspct that this is not a cas of suppltiv a
llomorphy. Th distribution of
ths allomorphs is rgular: [m] (a bilabial nasal) occurs bfor bilabial conso
nants and bfor vowls; [n]
(a coronal nasal) occurs bfor coronal consonants, [p ] (a palatal nasal) occur
s bfor palatals, [p] (a vlar
nasal) occurs bfor vlar consonants, and [] (a uvular nasal) occurs bfor uvul
ar consonants.
Sinc th allomorphs which occur bfor consonants hav a prdictabl
plac of articulation (it is
always th sam as th following consonant), this information can b
xtractd and statd as a rul in th
phonology of Sri:
(88) Plac Assimilation: A nasal consonant assimilats to th plac of articula
tion of an
immdiatly following consonant.
W hav givn this rul in broad gnral stroks. Othr data may sho
w that it should b narrowd
down som (and thy do, but thy ar not rlvant hr). Ruls such
as this ar vry common, howvr.
Espcially in th cas of th pronunciation of nasal consonants, th tongu and
lips ar movd into position
in anticipation of th nxt consonant.
What is th undrlying form of th ngativ prfix in Sri? Th only allomorph
whos shap appars
not to b dtrmind by its contxt is th allomorph which appars b
for vowls: [m]. Vowls do not
usually hav a major influnc on th plac of articulation of an adjacnt
consonant and thrfor usually
rprsnt th most nutral contxt for cass lik this. Sinc th lab
ial fatur of this allomorph is not
prdictabl, it must b spcifid in th lxicon. Thrfor w choos
{m} as th undrlying form of th
Ngativ prfix.
(89) Undrlying {imii} {imtii} {imkoo}
Plac Assimilation intii ipkoo
Phontic Rprsnt. [imii] [intii] [ipkoo]
65
Faturs
An important claim of gnrativ phonology (among othr thoris) is t
hat th basic building blocks of
phonology ar not th consonants and vowls (such as p and ), but th faturs.
A symbol such as p is an
abbrviation for a configuration of faturs, including som spcificati
on that th faturs [labial] and
[
voic] ar prsnt. If th fatur spcification for voic is changd, th rs
ulting sound will b diffrnt (it
will b a b rathr than a p). Mor dtails about faturs ar prs
ntd in nsuing chaptrs and also in
Appndix A.
Ky Concpts
major plac faturs [Labial] [Coronal]
[Dorsal]
Rlvant Phontic Information
Th following information, grouping sounds by point of articulation and showing
othr faturs, will b
hlpful in doing th xrciss for this chaptr. Th grouping of th sounds may
not b prcis in th

alvopalatal and palatal rgions.

Sounds without major plac faturs: glottal stop and [h].
11.1 Try it for yourslf with Wantoat
Th suffixs for my and our hav two allomorphs ach. Choos an undrlying form
for ach suffix and giv
rasons for your choic.
65
my our
1. kotok
po kotok
pin hand
2. kpi
no kpi
nin foot
3. jot
no jot
nin hous
Giv th phonological rul ncssary to account for th allomorph. You
should mak it as gnral as
possibl, making prdictions for data which you do not hav, in th absnc of c
ountrxampls.
66
[+voic]
j
w
[+sonorant]
[+voic]
m n
j
p
[+sonorant]
[+voic]
b d
g [sonorant]
[voic]
p t
k [sonorant]
[voic]
j
[sonorant]
[+voic]

[sonorant]
66
11.2 Try it for yourslf with English
Th ngativ prfix in English has two allomorphs in th following da
ta. Which on is th rsult of th
nvironmnt in which it occurs and which on is bst chosn as th undrlying fo
rm?
67
1. [in]dibl
2. [in]alinabl
3. [im]possibl
4. [im]partial
5. [in]flxibl
6. [in]variabl
7. [in]comptnt
8. [in]combustibl
9. [in]sincr
10. [in]tstat
Giv th phonological rul ndd to account for th allomorphy, in as gnral f
orm as possibl. Th rul
cannot b as gnral as th Plac Assimilation rul shown in this chaptr bcaus
 th nasal dos not bcom
vlar in th last two xampls.
68
11.3 Try it for yourslf with Min Nan Chins
Th word for vry has thr allomorphs in th following data. Choos an undrlyi
ng form and giv rasons
for your choic.
69
1. tsim p vry whit
2. tsim boi vry ugly
3. tsin to vry dry

4. tsin sin vry nw

5. tsin ho vry good
6. tsin o vry black
7. tsip koo vry thick
Giv th phonological rul ndd to account for th allomorphy.
70
67
11.4 Try it for yourslf with Sirra Popoluca
Th prfix for my has svral allomorphs. Choos an undrlying form and giv ra
sons for your choic.
71
1. omptkuj my broom
2. ompiju my hn
3. omm:m my buttrfly
4. onh:p my cup
5. opkowoh my hors
6. ojjmkuj my fan
7. onho:jo my husband
8. opw:tjo:mo my wif
9. onsuup my cooking pot
10. ojjopun my soap
Giv th phonological rul ncssary to account for th allomorphy.
72
11.5 Try it for yourslf with Javans
Suggst an undrlying form for th vrbalizr morphm and a rason for your cho
ic.
73
Noun Vrbalizd
1. bubor porridg mbubor to mak porridg
2. doprp story ndoprp to tll a story
3. guntp scissors pguntp to mak scissors
4. otjor pickl potjor to pickl
Giv th rul ndd to account for th surfac forms.
74
11.6 Try it for yourslf with Nabak
Giv a word structur rul for th data blow.
75
Suggst undrlying forms for th suffixs and rasons for
choosing ach.
76
my his
1. ibm ibmop sistr
2. bibm bibmop fathr
3. bon bopop pig
4. mkon mkopop hous
5. dodn dodnop ldr brothr
6. todn todnop parnts
in
law
7. sogp sogpop grandmothr
8. sgp sgpop skin
Giv a singl phonological rul to account for all th allomorphs.
77
68
11.7 Try it for yourslf with Nuyoo Mixtc
Assum th following undrlying forms for th data blow. Two ruls a
r ndd, on for ach typ of
chang. Giv th two ruls. Giv a drivation of th word for srvd to
show th application of th ruls.
(Th morphm {n} indicats compltd action.)
78
1. {n tso1o} [ndso1o] hard

2. {n kiku} [pgiku] srvd

3. {n kini} [pgini] wantd
4. {n so1o} [nzo1o] did, mad
5. {n jiko} [piko] sold
6. {n tjiso} [pdiso] hav put on
7. {n xo1o} [pyo1o] passd
11.8 Try it for yourslf with Pastaza Quchua
Propos a word structur rul to account for th words blow (including th word
[wasinda] at his hous,
which has thr morphms), a list of undrlying forms, and th phonological rul
s ncssary. (Rcall that
nasal stops ar sonorants and that oral stops and fricativs ar obst
runts; obstrunts ar [
sonorant]. This
will b important to know sinc obviously plac assimilation is not happning v
rywhr in ths data.)
79
to... at.. in... of ...
wosi wosimo wosito wosipi wosipo (th) hous
wosip wosipmo wosindo wosimbi wosimbo his hous
11.9 Try it for yourslf with Copainal Zoqu
Propos phonological ruls to account for th allomorphy shown blow (two ruls
ar ndd).
80
Th stms
of ths nouns all bgin with stops. Nouns which bgin with othr sounds ar pur
posly omittd. Th ruls
you writ should b as simpl as rasonably possibl.
Unpossssd my ...
1. pomo mbomo clothing
2. potso mbotso youngr sibling
3. totoh ndotoh fathr
4. t:k nd:k hous
5. tuwi nduwi dog
6. koju pgoju hors
7. komo pgomo cornfild
8. kos pgos oldr sistr
69
11.10 Try it for yourslf with North Publa Nahuatl
Rlvant morphm braks hav bn givn for you in th following dat
a. Assign a gloss to ach
morphm. Assum that th simpl prsnt tns contains th full stm of th vr
b and that othr tnss ar
formd by adding a suffix and dlting th last vowl of th stm (a kind of mor
phologically triggrd rul,
as dscribd in chaptr 5). Giv th undrlying form of th vrb for walk and th
 noun for hous; and th
phonological rul ncssary to account for thir allomorphs.
81
1. kolon outsid
2. kolompo th plac outsid
3. notjon my hous
4. tjompo th plac whr th hous is
5. nmi h walks
6. nntok h is walking
7. totjpono h swps
8. totjpontok h is swping
9. totjpopko h had swpt
10. motmo h baths
11. motpko h had bathd

Fdback for Chaptr 11

11.1 Wantoat
65
Th suffixs both bgin with a coronal nasal in th most nutral contxt, follo
wing a vowl. Thrfor th
bst choics for undrlying forms ar {
no} and {
nin}.
66
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly p
rcding consonant.
11.2 English
67
Th allomorph [im] occurs only bfor bilabial consonants and is probably du t
o th nvironmnt in
which it occurs. th allomorph [in] occurs in a varity of contxts, including b
for vowls and bfor
labio
dntal consonants, and is th bst choic for undrlying form.
68
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly fo
llowing bilabial
consonant.
11.3 Min Nan Chins
69
Th allomorph [tsin] occurs bfor vowls and bfor [h]; ths sounds ar lss
likly to affct th plac
of articulation of th nasal consonant. Thrfor [tsin] is th bst choic for
th undrlying form of th word
vry.
70
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly f
ollowing consonant. [h]
has no plac faturs, so thr is no assimilation bfor this consonant.
70

11.4 Sirra Popoluca

71
Th allomorph [on] is th only on which occurs in a contxt, bfor [h], which
dos not hav plac
faturs to affct th nasal consonant.
72
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly f
ollowing consonant.
11.5 Javans
73
Th bst choic for undrlying form is [p], which appars in th most nutral c
ontxt (prcding a
vowl).
74
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly f
ollowing consonant.
11.6 Nabak
75
Word Structur Rul: N N
stm

Possssor
76
Undrlying forms: my {
n}, his {
pop}. Ths ar th allomorphs which appar
aftr vowls, th most
nutral contxt so far as plac assimilation is concrnd.

77
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly p
rcding consonant.
11.7 Nuyoo Mixtc
78
Phonological Ruls:
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly fo
llowing consonant.
Voic Assimilation: A consonant assimilats in voicing to an immdiatly prcd
ing consonant.
Undrlying Form {n
kiku}
Plac Assimilation pkiku
Voic Assimilation pgiku
Surfac Form
pgiku
11.8 Pastaza Quchua
79
Word structur rul: N N
stm

(Possssor)
(Cas)
Morphms: wasi hous,
mo to,
to at,
pi in,
po of,
p his.
Phonological ruls:
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly fol
lowing obstrunt.
Voic Assimilation: A consonant assimilats in voicing to an immdiatly prcd
ing consonant.
11.9 Copainal Zoqu
80
Th ruls would b:
Voicing Assimilation: A consonant assimilats in voicing to an immdiatly prc
ding consonant.
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly fo
llowing consonant.
11.10 North Publa Nahuatl
81
Undrlying forms: hous {tjon}, walk {nmi}.
Plac Assimilation: A nasal consonant assimilats in plac to an immdiatly fo
llowing consonant.
71
Chaptr 12
Faturs in th Lxicon
Rcall again th Sri data prsntd in chaptr 11.
(90) imii who dosnt har it
impii who dosnt tast it
imit who dosnt stand up
insii who dosnt smll it
intis who dosnt point at it
ipjmx who dosnt div
ipjoo who dosnt own it
ipkoo who dosnt look for it
iyop who dosnt dig it
W saw that th various allomorphs of th ngativ prfix could b drivd from
th undrlying form {m}.
As a mattr of fact, all of th occurrncs of th sounds [p] and [p] and [] in t
h languag can b drivd.
Thr ar no occurrncs of ths sounds which ar unprdictabl.
a
If thr ar no unprdictabl
occurrncs of ths sounds, thn thr is no rason to hav [p] or [p] or [] in
any undrlying forms.
W say thn that whras w must posit faturs that would yild m in th lxico
n of Sri, w do not

posit a combination of faturs which would yild [p] or [p] or [] in th lxicon
of Sri.
b
If a rul or st of
ruls compltly account for th occurrncs of a fatur, that fatur is not c
ontrastiv in th languag in
qustion. In common parlanc, [m] is a phonm of Sri, but [p], [p] and [] ar n
ot.
A sound which is th rsult of a combination of faturs including on or mor n
on
contrastiv faturs
is rfrrd to as an allophon. W might say that [p] and [p] and [] ar allopho
ns of th phonm /m/ in
Sri. (Diagonals ar oftn usd around a symbol whn w hav phonms
in mind rathr than phontic
ntitis.)
In contrast to Sri, th languag Nabak has a combination of faturs
in undrlying rprsntations
which will rsult in /p/: /p/ is a phonm of Nabak. In solving th
Nabak problm from th prvious
chaptr, you saw that th undrlying form of th third prson possssiv suffix
was {
pop}. If w posit /p/
in undrlying forms, w ar saying that it cannot b dscribd as a rul
govrn
d variant of anothr sound
in all cass. W couldn t gt by with positing {
non} as th undrly
ing form for this morphm in Nabak
bcaus w wouldn t b abl to tll th conditions undr which th sound n bcam
 [p].
In latr chaptrs w will look at othr ways in which th contrastiv
 (or distinctiv) faturs of a
languag ar found.
Ky Concpts
contrastiv faturs and combinations of faturs allophon
phonm
a

Wll, almost. Th nam for a crtain spcis of duck [ kto:pk] unprdictably ha
s th vlar nasal. It is unprdictably
vlar bcaus othrwis in th languag (i) n dos not assimilat at
all, and (ii) m dos not assimilat in strssd
syllabls. This word is simply xtraordinarily xcptional.
b
Or in othr trms, [m] is a phonm of Sri, but [p], [p], and [N] ar not. W
follow th position outlind in Diana
Archangli and Douglas Pullyblank (in prparation) Th Contnt and Structur of
Phonological Rprsntations, who
point out that w rally must think of invntoris of faturs and t
hir combinations, rathr than invntoris of
phonms.
72
Postscript for Tachrs
Th notion of phonm has bn in th vryday vocabulary of linguists for many d
cads now and yt
continus to b problmatic in dfinition and plac in phonological th
ory. This book dos not tak th
viw of som vrsions of classical phonmics in which crtain strong cl
aims ar mad about th
rlationship btwn phontic rprsntations and phonmic rprsntations hold.
For that rason, som of
th procdurs for doing phonmic analysis ar lss rigidly applid in this book t
han in th past.

12 Try it for yourslf

Using th data providd in th problms, dcid whthr th sounds in
qustion should b considrd as
bing phonms of th languag or not. To do this, dcid whthr th rul that
was formulatd for th data
gnrats all of th instancs of th sound in qustion. If it dos, thn on sh
ould hypothsiz (until thr is
strong vidnc to th contrary) that th sound in qustion is not a phonm of
that languag.
12.1 Lithuanian (chaptr 9): Is [] a phonm of Lithuanian?
82
12.2 Javans (chaptr 11): Is [p] a phonm of Javans?
83
12.3 Nuyoo Mixtc (chaptr 11): Ar [n], [p], [p] and [z] phonms of Nuyoo Mi
xtc?
84
12.4 Pastaza Quchua (chaptr 11): Ar [d] and [b] phonms of Pastaza Quchua
?
85
12.5 North Publa Nahuatl (chaptr 11): Is [p] a phonm of North Publa Nahua
tl?
86
73
12.1 Lithuanian
82
Ys, [] must b a phonm sinc only som occurrncs of this sound ar gnratd
by th voic
assimilation rul. Undrlying forms includ this sound.
12.2 Javans
83
Ys, [p] must b a phonm of Javans sinc only som occurrncs of this sound
ar gnratd by th
plac assimilation rul. Undrlying forms includ this sound.
12.3 Nuyoo Mixtc
84
[n] must b a phonm bcaus it is positd in th undrlying forms, and is nd
d thr. Th sounds [p ],
[p] and [z] ar not phonms bcaus thy ar not ndd in undrlying forms (so
far as w know). Th
ruls of plac assimilation and voic assimilation account for thm.
12.4 Pastaza Quchua
85
[d] and [b] ar not phonms of Pastaza Quchua sinc thy ar not ndd in und
rlying forms; thy ar
accountd for by th rul of voic assimilation.
12.5 North Publa Nahuatl
86
[p] is not a phonm of Publa Nahuatl sinc it is not ndd in undrlying form
s and is gnratd ntirly
by th rul of plac assimilation.
74
Chaptr 13
Fatur Sprading
In chaptr 9 w saw that on sound was affcting anothr sound in a
particular fatur. Various attmpts
hav bn mad in th history of linguistics at formalizing ruls tha
t captur th ssnc of what is
happning. A common informal way to show an assimilation rul simply uss th ph
ontic symbols thm

slvs, as shown blow. Th following rul says that p bcoms b whn it follows

m.
(91) p b / m __
Thr ar two major problms with this informal notation. First, on
cannot asily xprss
gnralizations. What would on do if p bcam b, t bcam d, and k bcam g wh
n thy followd m, n, z,
tc.? A laundry list is not th sam as a simpl gnralization. Scond, this ki
nd of notation dos not allow
us to valuat th rul s simplicity. A simpl ordinary rul is just as asy to
writ in this notation as a vry
odd unlikly rul such as th following:
(92) z p / r __
W want our thory of phonology to hlp us valuat th ruls that w writ.
In standard gnrativ phonology of th 1960 s and 1970 s, a common notation for
ruls usd faturs,
such as [voic] in th following way.
a
(93) C [+voic] /
C
___
[+voic]
This rul claims that any consonant bcoms voicd aftr a voicd con
sonant. If th rul should b mor
spcific, mor faturs would b includd; if it should b mor gnral, fwr f
aturs would b includd.
For xampl, if only p and nvr t wr voicd in this nvironmnt, w could wri
t th rul as shown b

low.
(94)
C [+voic] /
C
___
[+labial]
[+voic]
This typ of notation is suprior to th us of phontic symbols in
that odd ruls usually rquir many
faturs whras common ruls rquir fw faturs. Th lattr ar formally simp
lr.
b
But this kind of notation still has crtain dficincis. Th following rul is
not any mor complicatd
in trms of faturs than (93), but it is crtainly not a typical rul.
(95) C [+voic] /
C
___
[+labial]
Nw formalism has attmptd to solv this problm. In currnt thory, assimilati
on is sn as on or mor
faturs sprading to anothr sound. To undrstand how it works, howv
r, som background on th
architctur of th rprsntation of sounds is ncssary.
W assum that sounds ar composd of faturs. On way to think of
thm is as individual
componnts of a sound which ar linkd to som anchor. For xampl, th sounds p
, b, and m might b
thought of as consonant positions (C) with th faturs [voic], [labial], and [nas
al].
c
a

Th most influntial work of th priod was Noam Chomsky and Morris Hall (1968
) Th Sound Pattrn of English,
Evanston, Nw York, and London, Harpr and Row (rprintd in 1991 by MIT Prss).
b
This didn t always turn out to b th cas, howvr. Somtims vry
common and natural ruls rquird mor
faturs than uncommon ruls. This has ld to som rvisions in th thory.
c

W ignor hr th important issu of whthr a fatur may b com

pltly absnt rathr than b spcifid with a
ngativ valu. W also ignor hr th important issu of how th faturs ar
organizd. S Appndix A.
75
(96) C p
[+labial] [voic] [nasal]
C b
[+labial] [+voic] [nasal]
C m
[+labial] [+voic] [+nasal]
With this in mind, thrfor, w might rprsnt th undrlying form of tacks as
follows, omitting for
th sak of prsntation all thos faturs which ar not rlvant to th rul o
f Voicing Assimilation:
(97)
t

k
z
|
|
|
|
C
V
C
C
|
|
|
|
[voic] [+voic] [voic] [+voic]
A rul of Voicing Assimilation sprads th fatur [voic] from th k
to th following sound. This
sprading can b indicatd by a dottd lin, as blow. Onc [voic] sprads to th
 z, th fatur [+voic]
that is alrady thr is cancld or droppd, and this is indicatd by th "x" o
n th lin.
d
(98)
t

k
z
|
|
|
|
C
V
C
C
|
|
|
x
[voic] [+voic] [voic] [+voic]
Th rul of Voicing Assimilation can b somwhat formalizd as:
This rul has a simpl intrprtation: th valu of th
fatur [voic] which th first consonant hasnot th
solid linis associatd with th scond consonantnot th dottd lin. (Not that
w hav statd th rul
as sprading [voic]whatvr valu it may havrathr than just sprading [voic]; t
his maks th rul
a bit mor gnral.)
Th rul of Plac Assimilation in Sri which was discussd in chaptr 11 might b
 viwd as sprading
th plac of articulation faturs of a consonant to th prcding nasal consona
nt. Rathr than mntion ach
particular plac fatur (labial, coronal, back, tc.), w can stat t
h rul quit simply as shown blow,
whr Plac is a catgory labl (in itslf it is not a fatur):
d

This cancllation of an incompatibl fatur whn anothr fatur is

addd on th sam C or V position) is
commonly considrd to happn by convntion.
(99) Voicing Assimilation
C
C
[voic]
76
Th first consonant in this rul must b spcifid as bing [+nasal], sinc only
nasal consonants assimilat
in this way to following consonants in Sri (as in many languags).
And again, th rul has a simpl

intrprtation: th first consonant (which may or may not hav a spcifid
plac of articulation bfor th
application of th rul) bcoms associatd with th Plac faturs of th follo
wing consonant.
Ky Concpts
fatur sprading Plac
13 Try it for yourslf with formalization
Complt th formalization of th assimilation ruls for th following problms
by drawing th dottd lin
that shows th chang.
13.1 Pngo (chaptr 10): A consonant assimilats in voicing to an immdiatly
following
consonant.
C
C
[voic]
13.2 Nabak (chaptr 11) : A nasal consonant assimilats in plac to an immdia
tly prcding
consonant.
C
C
Plac [+nasal]
13.3 Nuyoo Mixtc (chaptr 11): A nasal consonant assimilats in plac to an i
mmdiatly
following consonant.
13.4 Pastaza Quchua (chaptr 11): A nasal consonant assimilats in plac to an
immdiatly
following obstrunt.
(100)
C
C
[+nasal] Plac
C
C
[+nasal]
Plac
C
C
[+nasal]
Plac [
sonorant]
77
13.5 Pastaza Quchua (chaptr 11): A consonant assimilats in voicing to an im
mdiatly
prcding consonant.
13.6 Walmatjari (chaptr 9): A consonant assimilats in voicing to an immdiat
ly prcding nasal
consonant.
C
C
[voic]
C
C
[+nasal] [voic]
78
Fdback for Chaptr 13
13.1 Pngo: A consonant assimilats in voicing to an immdiatly following con
sonant.
C
C
[voic]
13.2 Nabak (chaptr 11) : A nasal consonant assimilats in plac to an immdia
tly prcding consonant.
C
C
Plac [+nasal]
13.3 Nuyoo Mixtc (chaptr 11): A nasal consonant assimilats in plac to an i
mmdiatly following
consonant.
13.4 Pastaza Quchua (chaptr 11): A nasal consonant assimilats in plac to an
immdiatly following
obstrunt.
13.5 Pastaza Quchua (chaptr 11): A consonant assimilats in voicing to an im

mdiatly prcding
consonant.
13.6 Walmatjari (chaptr 9): A consonant assimilats in voicing to an immdiat
ly prcding nasal
consonant.
C
C
[+nasal]
Plac
C
C
[+nasal]
Plac [
sonorant]
C
C
[voic]
C
C
[+nasal] [voic]
79
Chaptr 14
Constraining Rul Application
Whn a rul such as that of Voicing Assimilation (99) is proposd in
th phonology of a languag, th
xpctation is that it will apply vrywhr that it can apply, xcp
t as spcifically stipulatd (by adding
conditions to its contxt) or as gnrally constraind.
Th most intrsting ruls ar thos which hav a wid application, a
nd so th attmpt is to avoid
making th ruls any mor complicatd than thy hav to b. If a rul such as Vo
icing Assimilation (99) is a
rul of English, on must know why it dosn t also apply to words s
uch as flagship (its not pronouncd
with a [] sound instad of th [j] sound, and also words such as goldsmith and pi
gsty (thy ar pronouncd
with [s] and not [z]). Voicing is not sprading to a following conso
nant. Ths facts suggst that it is
ncssary to mak th rul mor spcific in som way, and of cours
th dilmma will b to dcid how.
Lt us assum that th rul should b fixd by adding th constraint (wll ignor
th problm of notation at
this point) both consonants must b in th sam syllabl. In othr words, voicing
assimilation in English
taks plac only whn th two consonants ar tautosyllabic (in th sam syllabl
).
This amndmnt dos not tak car of words such as slip and shrik,
howvr. Ths words contain
voiclss consonants followd by voicd consonants: th l and r do not bcom vo
iclss, vn though thy
ar in th sam syllabl with th prcding voiclss consonant.
Thr ar various ways in which Voicing Assimilation (99) could b mad vn mor
 complicatd so
that it will not apply to words such as slip and strip, but still
allow it to corrctly produc words such as
tacks. For xampl, on might mak th rul apply only if th two consonants ar
at th nd of a syllabl (in
slip th consonants ar at th bginning of th syllabl), or at th nd of a wo
rd. Or on might propos that
th dvoicing taks plac only if th scond consonant is a sibilant (that is, a
n s
lik sound). But thr ar
othr ways to constrain th rul which ar of mor intrst cross
lin
guistically sinc thy hav to do with
th natur of th sounds involvd. Situations lik th on w ar ncountring h
r happn in languag aftr
languag, with ruls that ar vry diffrnt.
On obsrvation is that th voicing assimilation rul in qustion applis
to z, an obstrunt, whras it

dos not apply to l and r, which ar sonorants. Th distinction btwn sonorant
s and obstrunts shows up
in many languags and could b th rlvant fact. If so, rul (99)
would b rvisd to spcify that only
obstrunts undrgo th rul.
(101) Voicing Assimilation (rvisd)
C
C
[voic] [
sonorant]
if th consonants ar tautosyllabic
Whil this kind of rvision works, it is not th most intrsting wa
y to fix th problm of a rul that
applis in th wrong placs. Somtims a fatur dos hav to b addd to th ru
l, but othr tims thr is a
bttr solution. W not in this cas that th sound to which w s
 Voicing Assimilation applying in
English is z, a sound which is a phonm of English. Th rsult of th applicati
on of Voicing Assimilation
is s, also a phonm of English. But if Voicing Assimilation wr to
apply to slip, th rsult would b a
voiclss l, which is not a phonm of English. If Voicing Assimilation wr to
apply, a novl combination
of faturs for English (on which dos not occur in undrlying forms
) would rsult. If w stipulat that
Voicing Assimilation in English is a structur prsrving rul, this w
ould corrctly prvnt it from
applying to th word slip.
a
Th output of a structur
prsrving rul is similar to th kinds o
f structurs
that occur as input to th rul; for xampl, such a rul dos not
produc sounds which do not occur in
a

Elaborat and intrsting attmpts hav bn mad to mak structur

prsrvation rlat to othr facts in th
phonology, spcially in th thory of Lxical Phonology. For an intro
duction to that thory (within th gnrativ
tradition), including problms that aris, s Michal Knstowicz (1994)
Phonology in Gnrativ Grammar,
Cambridg, Mass, Blackwll. Th ida of structur prsrvation in particular is
discussd first in Paul Kiparsky (1982)
From Cyclic Phonology to Lxical Phonology, H. van dr Hulst and N.
Smith (ds.), Th Structur of Phonological
Rprsntations, Parts I and II, Dordrcht, Foris; and Paul Kiparsky (1982) Lxi
cal Morphology and Phonology, I.
S.
Yang (d.), Linguistics in th Morning Calm, Soul, Hanshin.
80
undrlying forms. If Voicing Assimilation in English is stipulatd to b a struc
tur prsrving rul, th rul
may apply whr th conditions ar right xcpt that it is blockd f
rom applying in thos spcific cass
whr it would produc somthing novl (such as a voiclss l or r).
To argu that a rul is structur
prsrving, on must hav dirct vidnc such as this whr its application is
actually blockd.
W might also not that rul (99) applis in English only in drivd
nvironmnts, such as whn
morphms com togthr. In th cas of th word slip, th s and l occur nxt
to ach othr in th undr

lying form; rul (99) dos not apply to thm. But in a word such as tacks, t
h morphms tack and
z ar

juxtaposd by mans of a word formation rul, crating a drivd situation to wh

ich (99) can apply. If w
stipulat that (99) is a drivd nvironmnt only rul, th facts would b account
d for corrctly without
complicating th rul in any spcific way.
b
All this is mant to show that not vry fact about how a rul applis to a form
is ncssarily ncodd
dirctly into th formulation of th rul. A constraint such as that
a rul b structur prsrving or apply
only in drivd nvironmnts might b addd to it (or somhow rsult from th th
ory), gratly rstricting it
without dirctly complicating it. Linguists ar constantly looking for ways to g
nraliz, to rlat facts, and
to find xplanations that go byond th nitty
gritty dtails of a particular cas
.
Ky Concpts
constraints on ruls structur
prsrving rul drivd nvironmnt
14.1 Try it for yourslf with English
How do words such as flt, smlt (compar with fl, smll) affct th way that
you viw th formulation of
Voicing Assimilation. Ar th consonants lt in th sam syllabl? Ar
thy in a drivd nvironmnt?
Would th application of Voicing Assimilation b blockd by stipulating that it
is structur prsrving?
87
14.2 Try it for yourslf with Amrican English
Th two suffixs with th form
O in English, unlik th suffix s {

z}, caus rgrssiv voicing
assimilation in som words; voicing sprads to th lft. (Ignor th vowl altr
nations in ths data.) Th
rul might b formulatd as: A consonant assimilats in voicing to a consonant t
hat immdiatly follows it.
Discuss two distinct ways in which th voicing assimilation rul could b formul
atd or constraind so that
it will proprly chang v to f (as in twlfth) but not dvoic n (as in ninth).
Which way is to b prfrrd and
why?
88
b

Th notion of a structur
prsrving rul or a rul that applis only in a dri

vd nvironmnt has som affinity with
morphophonmic rul. Th lattr trm, introducd in structuralist phonology
, is still commonly usd in an informal
way to rfr to ruls which do not gnrat allophons. Th trm is
oftn usd in opposition to allophonic ruls
and/or phontic ruls. Ruls of th lattr sort, which ar not structur prsrvin
g and which do apply in nondrivd
nvironmnts, ar discussd in dtail in chaptr 23.
81
1. fojv fiv fif
O fifth
2. siks six siks
O sixth
3. scvn svn scvn
O svnth
c
4. jt ight jt
O ighth
5. nojn nin nojn
O ninth
6. tcn tn tcn
O tnth
7. twclv twlv twclf
O twlfth
8. tw\ni twnty tw\ni
O twntith

9. wojd wid wit
O width


10. b+5d broad b+ct
O bradth
12. w5+m warm w5+m
O warmth
f
11. l5p long
g
lcpk
O lngth
12. str5p strong strcpk
O strngth
14.3 Try it for yourslf with Amrican English
In ths data you s that thr ar thr allomorphs of th suffix. Dos
th undrlying form of th suffix
bgin with a voicd or a voiclss sound? How do you know?
89
Assum that th last consonant of ight dlts whn it prcds this
suffix. Also, you may assum that a
rul changs a coronal stop into a flap whn it occurs following a
continuant sonorant.
(Not: this givs
vidnc from English that th r of English is [+continuant] and that nasals ar [

continuant].)
Two voicing assimilation ruls ar ndd. On is rgrssiv (sprading to th l
ft), and on is progrssiv
(sprading to th right). What ar th ruls?
90
Can you kp thm simpl? Ar thy structur prsrving or
not or cant w tll?
91
(Ths ruls do not apply only to tautosyllabic consonantsanothr ch
allng for
th corrct formulation of Voicing Assimilation in English!)
1. Io+ four fo+
ri forty
2. Iojv fiv IiI
ti fifty
3. siks six siks
ti sixty
4. scvn svn scvn
di svnty
5. cjt ight cj
ri ighty
6. nojn nin nojn
di ninty
c

For many spakrs with this dialct, thr is an intrusiv voiclss

stop btwn th n and th 6 in word such as
this: [scvntO]. For this xrcis, considr this dialct (prhaps artificial
at this point) that dos not hav this intrusiv
stop.
d
Th allomorph
O occurs aftr vowls.

Th dvoicing of th root
final consonant is tru of this dialct of English; i
t is apparntly not tru of all dialcts.
f
Thr may b an intrusiv voiclss bilabial stop in this word: [w5+mpO]. S
not c.
g
Assum that ths roots hav an undrlying final g which dlts in word
final
position and that th k which appars
bfor
6 is not an intrusiv consonant (s footnot 2).
82
14.4 Now try it on som othr languags
Dcid whthr th ruls for th following problms ar structur prsrving rul
s or not. Th way to do

this is to show that th rul producs a sound which dos not appar in undrlyi
ng forms. If it dos produc
such a sound, thn th rul is not structur prsrving. It dos not produc suc
h a sound, and w think that it
should in a givn nvironmnt, thn th rul is structur prsrving.
If th rul dos not produc such a
sound but w dont know what it dos in th crucial cass, thn w simply dont know
whthr its structur
prsrving or not.
If th data ar vry limitd, you wont hav much to go on for your dcisions. But
do th bst you can.
14.4.1 Nuyoo Mixtc (chaptr 11); Voicing Assimilation and Plac Assimilation
92
Assum that th
symbols which appar in th undrlying forms ar rprsntativ of all of th ph
onms of th
languag.
14.4.2 Copainal Zoqu (chaptr 11); Voicing Assimilation and Plac Assimilation
93
Assum that
th symbols which appar in th Unpossssd column ar rprsntativ of all of
th phonms of
th languag.
Fdback for Chaptr 14
14.1 English
87
Th consonants ar in th sam syllabl (thy ar tautosyllabic). Thrfor that
condition is mt. Thy ar
also in a drivd nvironmnt, sinc th words ar a combination of a vrb plus
a past tns suffix
t. Th
application of Voicing Assimilation would not b blockd by stipulating that it
is structur prsrving sinc
th output d is a phonm of English. Th output is consistnt with structur pr
srvation. It appars that
th rul simply has to b formulatd to b fairly spcific in its contxt.
14.2 Amrican English
87
On way to account for th facts would b to complicat th rul by stipulating
that only obstrunts undrgo
th rul (including v, d, and g and xcluding n and m). An altrnativ is to sti
pulat that th rul is structur
prsrving. This would prvnt th rul from applying to n and m sinc thr ar
no voiclss sonorant
phonms in English. This is to b prfrrd in principl bcaus it maks rcou
rs to a gnral constraint
rathr than a fatur
spcific constraint.
14.3 Amrican English
89
Th suffix must bgin with a voiclss sound. Othrwis, w would hav no way t
o account for th
dvoicing of th v to f in th drivation of fifty.
90
Rgrssiv Voicing Assimilation: A consonant assimilats in voicing to a conso
nant that immdiatly
follows it. (This accounts for th dvoicing of fiv in th word fifty.)
Progrssiv Voicing Assimilation: A consonant assimilats in voicing to a co
nsonant that immdiatly
prcds it. (This accounts for th voicing of
ty in words such as ninty.)
91

To kp ths ruls simpl, w must first rquir Rgrssiv Assimilation to b

attmptd bfor
Progrssiv Assimilation. Othrwis undrlying fiv
ty would yild fivdy insta
d of th corrct fifty. Also,
w must stipulat that Rgrssiv Assimilation is structur prsrving, bcaus
it dos not apply in th
drivation of ninty; th output contains a voicd n rathr than a voiclss n.
14.4 Othr languags
14.4.1 Nuyoo Mixtc
92
Nithr rul is structur prsrving sinc thy both produc sounds that do not
occur in undrlying forms.
14.4.2 Copainal Zoqu
92
Nithr rul is structur prsrving sinc thy both produc sounds that do not
occur in undrlying forms.
83
Chaptr 15
Palatalization and Labialization
In chaptr 11 w saw cass whr a consonant assimilatd to a nighboring conson
ant in its plac faturs.
In this chaptr w look at cass whr consonants assimilat to nigh
boring vowls with rspct to plac
faturs.
Consonants somtims ar pronouncd with th tongu body drawn toward
th hard palat.
a
For
xampl, a back consonant, such as k, is oftn pronouncd slightly forward bfor
a front vowl. Th k of
ky is pronouncd farthr forward than th k of Kool
Aid, for xampl.
In som languags th assimilation of a consonant to a vowl rsults
in a transitional j off
glid
btwn th consonant and th vowl. In Tataltpc Chatino, a k has an audibl j
off
glid whn it prcds
an . For xampl, th word /k/ had is pronouncd [k cc|.
Both of th prcding kinds of procsss hav bn rfrrd to as pa
latalization (sinc thr is an
attraction towards th hard palat), although th rsulting sound may
not ncssarily b a palatalizd
consonant in th trminology usd in a phontics class. High front vowls and gl
ids ar spcially common
nvironmnts for palatalization.
Palatalization in this broad sns may rsult in an vn mor notica
bl chang in th sound. For
xampl, in Dakota Sioux, a /k/ bcoms [tj| whn it follows an /i/
(and also prcds a vowl). For
xampl, th /k/ of th morphm {kox} mak bcoms [tj] whn th pr
fix {i
} prcds it, as in th
word [itjoyc] instrumnt (th morphm {kox} is th part [tjoy] in this word).
English has many xampls of palatalization which originatd with Latin, such as
whr k bcoms s
bfor i. Compar lastic [k] with lasticity [s], and analogu [g] with analogy
[d]. S also pairs such as
Egypt [t] and Egyptian [j], Grc [s] and Grcian [j]. Othr xampl
s of historical palatalization in
English ar: rgn[t] / rgn[s]y, na[t]iv / na[j]ion.
But thr ar also many mor instancs of palatalization in English that ar goi
ng on today. Considr
th commonly hard pronunciations of th words issu (with [j] in Am
rican English, [s] in British

English) and matur (with [tj] in Amrican English gnrally). Th t

of th root of captiv palatalizs to
[tj] in th word captur. Th d of th word fraud palatalizs to [d] in th word
fraudulnt. Th s of miss is
typically palatalizd in Amrican English in a phras such as I ll mi
ss you, spcially in casual spch.
Similarly, th t of lt (and othr words) is pronouncd as [tj] in phrass such
as If I lt you rad this.... And
thr is a distinct palatalization of t and d bfor r in many dial
cts of English, such that childrn oftn
bgin writing words lik truck with ch bcaus thy har th distinctiv palatal
ization.
15.1 Try it for yourslf with Yaqui
What kind of altrnation is sn in th following data? What is th likly condi
tioning factor for this?
94
(102) tcput fla
tcputj
im flas
15.2 Try it for yourslf with Karuk
What kind of altrnation is sn in th following data? What is th likly condi
tioning factor for it?
95
Basd
on ths vry limitd data, is th procss structur
prsrving?
(103) s/h... I...
1u
skok ni
jkok jump
1u
suprih ni
juprih masur
1u
si:tvo ni
ji:tvo stal
a

Th hard palat is th bony part of th uppr sid of th mouth; th soft palat
 is bhind it. If you mov your tongu
along th uppr sid of th mouth, you can fl th plac whr th hard palat
stops.
84
Palatalization is an ara that has bn challnging to formaliz proprly. For s
om vry hlpful currnt
proposals, s Clmnts and Hum 1995.
b
Labialization
Roundd vowls such as u also somtims influnc consonants around th
m, causing thm to bcom
roundd. This is known as labialization. Back consonants ar spcially suscpti
bl.
In som cass, th lips ar simply roundd bcaus th following vowl has lip r
ounding, as in English
tour (th t is pronouncd with roundd lips). In othr cass, th co
nsonants may b pronouncd with a w
off
glid bfor th round vowl, although this situation is apparntly lss com
mon than simpl rounding of
th lips.
Ky Concpts
palatalization (in th broad sns) labialization
15.3 Try it for yourslf with Ukrainian
Giv th undrlying form of ach suffix and giv th undrlying form
of ach stm. Giv th ncssary
phonological rul.
96
Indicat whthr th rul appars to b structur
prsrving or not, basd on t
hs vry

limitd data.
Nom. Sg. Gn. Sg. Loc. Sg.
1. t ilo t ilo t il i body
2. kolso kolso kols i whl
3. ozro ozro ozr i lak
4. slovo slovo slov i word
15.4 Try it for yourslf with Pars
Giv th undrlying form of ach prfix and giv th undrlying form
of ach stm. Giv th ncssary
phonological rul.
97
Indicat whthr th rul appars to b structur
prsrving or not.
[Som phontic
dtail has bn omittd.]
your ... my ...
1. hik ono nokono arm
2. him o nomo cloth
3. him li nomli skin
4. hih oni nohoni hous
b

G. N. Clmnts and Elizabth V. Hum (1995) Th intrnal organizatio

n of spch sounds. In John A. Goldsmith,
d., Th handbook of phonological thory, Blackwll, Cambridg, Mass. and Oxford
, pp. 245
306.
85
15.5 Try it for yourslf with Sri
Considr th morphms blow which hav allomorphs that ar diffrnt
only in th fatur of [+round].
Giv th phonological rul that is appropriat in its simplst form.
How can this rul b prvntd from
applying to s and t?
98
did / was ? s/h will s/h was going to
1. toop soop ko1o soop to mak baskt
2. tiim siim ko1o siim to slp
3. tit sit ko1o sit to aris
4. tis sis ko1o sis to b raw
5. itoon soon ko1o soon to carry (thm)
6. itkooj skooj ko1o skooj to rob (him/hr)
7. tooix sooix ko1o sooix to dizzy
8. itomx
siimx ko1o siimx to hunt (it)
9. tk sk ko1o sk to sow (unspc. objct)
10. itok
siik ko1o siik to kill (it/him/hr)
11. tiipy siipy ko1o siipy to lay gg
12. itiisy
siisy ko1o siisy to hid (it/him/hr)
13. kotikpon who works (morphm brak indicatd)
14. kkotikpon who works lik him/hr
15. ksotikpan s/h will work lik him/hr
86
15.6 Try it for yourslf with Modrn Grk
Th words ar givn as thy appar whn followd by a vowl
initial word. Somti
ms a word has mor
than on possibl pronunciation. Account for th allomorphy; two optional ruls
will b ndd. Prsnt th
rsults in th form of a discussion, with argumntation and vidnc. B concis
but also clar. (Th ruls
do hav som complications.) An astrisk mans that th form is not corrct. It
is not possibl to includ all
of th words that on cant say in Grk, but th ons includd ar important as a

sampl of data that may

b rlvant.
99
1. kIistiko ~ kIisti ko I was shut
2. IcrOiko ~ IcrOi ko I was brought
3. pc toxtiko ~ pc toxti ko I was thrown
4. trc IoOiko ~ trc IoOi ko I bcam insan
5. ro tiOiko ~ ro tiOi ko I was askd
6. simbo Oo ~ *si mbo Oo ~ *s imbo Oo I lik
7. sindi ro I support
8. ipos x cOiko ~ ipos x cOi ko
I promisd
9. po Ocno ~ *po Ocno I suffr
10. po Ii ~ po I i ~ *po I i much
11.
Iiyo ~
I iyo ~ * I i yo
littl
12. ni si ~ n i si ~ *n i si island
13.
skoni ~
skon i dust
14. poro Iipo ~ poro I ipo I omit
15.
k irii ~ * k ir ii gntlmn
16. kriI to ~ *kr i I to I shall hid myslf
17. cpi strcIo ~ cpi strcIo
I giv or com back
18. o nixtik c ~ o n ixtik c ~ o n ixti k c ~ *o n ixt i k c it was opnd
19. nix isu kopik c ~ n ix isu kopik c ~ n ix i su
kopi k c
~ nix i su kopi k c Your fingrnail is cut.
87

Fdback for Chaptr 15

15.1 Yaqui
94
Th altrnation is on of palatalization, with th [t] bcoming [tj] bfor [i].
15.2 Karuk
94
Th altrnation is on of palatalization, with [s] bcoming [j] aftr [i]. Th p
rocss is not structur

prsrving sinc [j] dos not appar in undrlying forms.
15.3 Ukrainian
95
Th undrlying forms of th affixs ar: {o}, {o}, and {i}.
Th undrlying forms of th stms ar: {t il}, {kols}, {ozr}, and {slov}, so f
ar as w can tll.
Rul: A consonant is palatalizd (has a j
offglid) prcding i.
Th rul is not structur
prsrving sinc som of th outputs do not appar in
undrlying forms.
15.4 Pars
97
Th undrlying forms of th prfixs ar: {hi} and {no}.
Th undrlying forms of th stms ar {kono}, {mo}, {mli}, and {honi}.
Rul: A consonant is palatalizd (has a j
offglid) following i.
Th rul is not structur
prsrving.
15.5 Sri
97
Th labialization ruls sprads th fatur [+round] to an immdiatly following
consonant.
Th rul is structur prsrving: it dos not sprad th fatur [+round] to s a
nd t bcaus th only roundd
consonants in undrlying forms ar th back consonants, as is vidnt from thir
usag in roots. If on did
not stipulat this, on would hav to complicat th rul and spcify that it sp
rads th fatur [+round]

only to Dorsal consonants.

15.6 Modrn Grk
99
Two ruls ar ndd:
Palatalization: Th consonants n and l palataliz bfor i xcpt whn thy fol
low a consonant. Othr
consonants do not palataliz in this position. This rul is optional.
Dvoicing of vowls: High vowls (i and u) dvoic whn thy occur btwn two
voiclss sounds and
ar not strssd. This rul is optional.
88
Chaptr 16
Nasalization
Th fatur [+nasal] charactrizs som consonants, such as m, in that thy ar
producd with air scaping
through th nos. Vowls can also b producd with th fatur [nasal
]. For xampl, many spakrs of
English slightly nasaliz th vowl of a word such as can, and som do it mor t
han othrs. This procss of
nasalization, whr a sound assimilats to a nighboring sound with r
spct to th fatur nasal, is vry
common. In som languags it is mor noticabl than in othrs. It m
ight b formalizd as follows
(dpnding on whthr th nasal consonant prcds or follows th vowl):
(104) Nasalization
C
V
V
C
[+nasal] [+nasal]
In som languags, th nasality of a vowl affcts nighboring consona
nts. Considr th following data
from Waorani:
(105) o
bo I s
o
ki
mo I shall s
o
do sh ss
o
ki
no sh will s
Th first prson singular suffix has two allomorphs, [bo] and [mo]; and
th third prson singular
fminin suffix has two allomorphs: [do ] and [no ]. Assuming that th undrlying
forms of ths suffixs
hav oral consonants, w can posit a rul nasalizing consonants whn thy follow
nasalizd vowls.
a
(106) Consonants bcom [+nasal] following [+nasal] vowls.
V
C
[+nasal]
In Copainal Zoqu, a nasal consonant mrgs with a non
consonantal sono
rant that follows it to
produc a nasalizd vrsion of that sonorant. Th first prson singula
r possssiv prfix (my) is a nasal
consonant (s th xrcis in chaptr 11).
b
(107) Unpossssd my ...
jomo jomo wif
woko woko baskt
hojoh hojoh husband
In Kikongo, th prsnc of a nasal consonant in a stm affcts th
nasality of coronal consonants in
suffixs. Nasality is sprading to th suffix. Thus th morphm for prfctiv
activ has two allomorphs:
[
idi] and [
ini]; and th prfctiv passiv has two allomorphs: [
ulu] and [
u
nu].
(108) Stms without nasals:

n
suk
idi I washd
n
suk
ulu I was washd
m
bud
idi I hit
a

W ar uncrtain about th analysis. Th data providd in William R

. Mrrifild t al., ds. (1987) Laboratory
manual for morphology and syntax, Dallas, SIL, contain xampls with [t] followi
ng nasalizd vowls.
b
Th mrgr, or coalscnc, of two sounds is mor complicatd than th simpl s
prading of a fatur from on to
th othr. A formalism for this is not prsntd hr.
89
Stms with nasals:
tu
kun
ini w plantd
tu
nik
ini w ground
masangu ma
nik
unu th maiz was ground
Ky Concpts
[nasal] nasalization
16.1 Try it for yourslf with Lalana Chinantc
Assuming that th undrlying form of th first prson plural inclusiv
suffix is {
ro}, giv th rul(s)
ndd to account for th allomorph [
no ]. Giv th rul(s) in pros and in for
mal notation.
100
1. zo:?ro w (incl.) practic 5. ?ju:??nno w (incl.) ti him
2. kud?nno w (incl.) snd 6. ?u:nno w (incl.) pour on him
3. ?mi:ro w (incl.) mnd 7. zu?nno w (incl.) fold ovr
4. ?i:?ro w (incl.) rciv
Fdback for Chaptr 16
16.1 Lalana Chinantc
100
Nasalization of Consonants: A consonant (prhaps only coronals, prhaps only t
h coronal sonorant)
bcoms nasal following a nasal consonant.
Nasalization of Vowls: A vowl bcoms [+nasal] following a [+nasal] consonant
.
Prhaps ths can b combind into on rul: A sgmnt bcoms [+nasal] followi
ng a [+nasal] consonant.
Assuming this, th rul could b formalizd as follows (with a dottd lin going
from th fatur [+nas] to
th X). This rul would hav to b apply mor than onc itrativly in words li
k w pour on him.
C
X
(whr X = C or V)
|
[+nas]
90
Chaptr 17
Mannr Assimilation
Consonants may tak on faturs of nighboring sounds. On common chan
g is for th fatur
[continuant] to sprad to contiguous sounds. Sounds that ar [+continuant] inclu
d thos in which th air
stram is not compltly blockd at th primary constriction in th v
ocal tract (Chomsky & Hall
1968:317). Vowls, of cours, ar [+continuant], as ar fricativs. Sou
nds that ar [
continuant] includ
thos in which th air stram is prvntd from scaping through th oral cavity
, which includs oral stops,

affricats, and nasal stops, as wll as l (by this somwhat controvrsial dfini
tion), bcaus th air stram is
blockd at th primary strictur.
A common xampl of [+continuant] sprading to a consonant is whn a stop bcom
s a fricativ ithr
bfor or aftr a vowl or btwn vowls. For xampl, in Salasaca Quichua, th
Locativ suffix {
bi} has
th allomorph [i] whn it follows a vowl
final morphm, as in [wasii]
(hous
Loc). And th Plural
suffix {
guna} has th allomorph [
yuna] whn it follows a vowl
final morphm,
as in [tajtayuna] (fathr

Pl).
In Spanish, som obstrunts ar pronouncd as fricativs aftr vowls
and flaps, .g. th d in nada
nothing and ardr to burn; but as stops whn thy follow nasals or a l
atral, as in andar to go and
bald buckt.
In som languags, th combination of [+continuant] and [coronal] rsul
ts in an r
lik continuant
instad of a fricativ. In Amrican English, for xampl, th consonants t and
d ar both pronouncd as a
flap whn thy occur btwn vowls: city, Hidi.
Ky Concpt
[continuant]
a

Strictly spaking, this may b limitd to th non
latral fricativs.

(109) [+continuant] [
continuant]
Vowls, Glids
Fricativs
a
Oral Stops, Glottal Stop, Affricats
Flaps, Trills Nasal Stops, Latral Stop
91
17.1 Try it for yourslf with Ejagham
Giv th rul(s) ncssary to account for th altrnations, using th
faturs in th tabl givn in this
chaptr. You should b abl to do this in pros and also in formal
notation. You will nd to considr
carfully about th undrlying form of th root. Discuss how you cam up with yo
ur dcision. Assum that
th undrlying form of th prfix for I is {n
}. Provid a drivation for th w
ords I brok and I dancd.
101
I ... h ...
1. mbon obon brok
2. mbon oon dancd
3. mbc obc scapd
4. mbc oc plantd
5. ndi odi at
6. ndi ori crid
7. ndop odop trid
8. ndop orop touchd
92
Fdback for Chaptr 17
17.1 Ejagham
101
Rul 1: Th fatur [
continuant] sprads from a consonant to an immdiatly fo
llowing consonant.

(Fricativs and flaps ar changd into stops.)

C
C (dottd lin ndd from th fatur to th scond consonant)
|
[
cont]
Rul 2: A nasal consonant assimilats in plac to an immdiatly following cons
onant.
C C (dottd lin ndd from th Plac nod to th first consonant)
|
Plac
Th undrlying form of th stm is mor clarly rvald in th scond column, w
hn it follows a vowl. In
that column, w s two diffrnt forms for th vrbs brak and danc, for xamp
l, whras in th first
column th diffrncs btwn ths two vrbs ar obscurd (by th sprading of
th fatur [
cont], as a
mattr of fact).
Drivations: (th UF of th prfix is assumd to b n
)
I brok I dancd I crid
UF n
bon n
on n
ri
Mannr Assim. nbon nbon ndi
Plac Assim. mbon mbon ndi
SF mbon mbon ndi
Th ruls apply in all thr of ths words vn though thy only actually mak
a chang in crtain
instancs.
93
Chaptr 18
Vowl Changs
Vowls somtims tak on faturs of nighboring consonants. W hav a
lrady discussd how thy can
bcom nasalizd contiguous to nasal consonants.
Somtims vowls also ar changd to b highr or lowr, mor front or mor back
, dpnding on th
point of articulation of a contiguous consonant. In Sri, for xampl,
an i is backd and lowrd
considrably if it prcds a uvular fricativ (a sound which is producd in
th back of th throat without
raising th back of th tongu). Thrfor a word such as {jiiy} is pronouncd [
jIIy].
A vowl can also assimilat to a vowl in anothr syllabl. This typ
 of assimilation has bn calld
vowl harmony. For xampl, in Hungarian, [u] (back roundd) bcoms [
y] (front roundd) whn th
prcding syllabl has a front vowl, but not whn th prcding syllabl has a
back vowl.
a
(Th vowl a
in ths data rprsnts a vowl that is farthr back than th mor cntral vow
l usually rprsntd by th
symbol a in this txt.)
(110) 1 plural Subjunctiv
ioo
jupk scold
kop
jupk obtain
tud
jupk know
tj
ypk put
vid
jypk carry
Ikyd
jypk rclin
Th rul might b statd tntativly as:
b
(111) Backnss Harmony: A vowl assimilats in backnss to th vowl in th imm
diatly

prcding syllabl.
V
C
0
V
[back]
This rul is givn in a fairly gnral form, and prdicts that vowl
s othr than u will also assimilat in
backnss. If this is not tru, thn th rul must b modifid. But othr Hungari
an data do show that o also
bcoms [
back] in th sam nvironmnt.
(112) 3 singular Subjunctiv
roo
jon scold
kop
jon obtain
tud
jon know
fkyd
jon rclin
Th rul also prdicts that th back vowl a will also b frontd w
hn th prcding syllabl contains a
front vowl. And it is.
a

Facts from situations lik this in a languag that has cntral vowls ar usual
ly similar, suggsting that vowls should
b classifid phonologically as bing ithr [+back] or [
back], whr so
calld
cntral vowls ar [+back]. That is,
th IPA labls front, cntral, and backa thr
way distinction, ar rplacd
by a binary distinction in th
phonology which is [
back] and [+back]. Th phonological fatur [back] thrfor
 must b clarly distinguishd from
th phontic labl back.
b
Th C subscript 0 is a way to indicat that th numbr of consonants is irrlv
ant.
94
(113) 1 singular Subjunctiv
roo
jok scold
kop
jok obtain
tj
k put
vij
k carry
kr
jk ask
Now for th wrinkl. Somtims th o also bcoms [
round] as wll a
s [
back]. Instad of just
bcoming [o] (front roundd) as th Backnss Harmony rul prdicts, it
bcoms [] (low front
unroundd), Compar th data blow with thos in (113)prsumably th undrlying f
orm of th suffix is

jon.
(114) 3 singular Subjunctiv
td
jn put
vid
jn carry
If w assum that Backnss Harmony rul (111) taks car of th fronting, all w
nd to do now is hav a
rul that in addition to that rul maks o into an unroundd vowl in th right
contxt, and low.
Th Backnss Harmony rul applis to all vowls (so far as w know at this point
), but th Roundnss
Harmony rul applis only to [
high] vowls (namly o) and only whn
th prcding syllabl contains a
front vowl. (Not that th squncs a...u, a...o, ...y, and i...y ar found
abov, with no rounding harmony
having takn plac.)

Th combination of th application of ths two ruls, Backnss Harmony and Roun
dnss Harmony, to
o almost yilds th corrct rsult. Thy should giv th vowl [],
but th rsult is actually []. So th
situation is slightly mor complicatd.
c
(115) Roundnss Harmony: Nonhigh vowls assimilat in rounding to a front vowl
in th
immdiatly prcding syllabl.
Sinc Backnss Harmony and Roundnss Harmony affct diffrnt substs of vowls
and caus distinct
changs (on fronts, on unrounds), w considr thm as two sparat and indpn
dnt phonological ruls
of Hungarian.
Vowl Faturs
Th discussion abov uss crtain faturs to rfr to vowls. As mntiond, th
fatur [back] is commonly
usd to distinguish btwn front vowls on th on hand, and cntral and back v
owls on th othr.
(116)
[
back] [+back]
Front vowls Cntral & Back Vowls
Eg.: i c y Eg.: u o a u I
Th phonological faturs [high] and [low] hav bn usd to distinguish clos,
nar
clos, nar
opn,
and opn vowls.
d
c

This discussion is all quit tntativ in th absnc of a complt st of fact
s. Th vowl systm of Hungarian would
nd to b lookd at in mor dtail.
d
Th trm clos is standard trminology, but th trm nar
clos is not.
Nar
clos
vowls lik [i] ar calld
mid
cntralizd clos vowls in th IPA systm. S Handbook to th Intrnational
Phontic Association: a guid
to th us of th Intrnational Phontic Alphabt (1999) Cambridg Univrsity Pr
ss, Cambridg.
95
(117)
[high] [low]
Clos and
nar
clos vowls
+
i y u u
i
u
Clos
mid,
mid, and
opn
mid vowls

 o r o
o
c
:
Nar
opn
and opn vowls

+ a u
Th fatur [advancd tongu root], or [ATR] is commonly usd to distinguish clo
s from nar
clos,
and clos
mid from opn
mid vowls. (S Appndix A for mor discussion.)


(118)
clos, clos
mid
i u
c o
[+ATR]
nar
clos, opn
mid, opn
i o
c
[ATR]
Som languags xhibit vowl harmony with rspct to th fatur [ATR]. This can
b sn in Yoruba
whr th nominalizing prfix taks on th sam valu of ATR as th vowl of th
nxt syllabl. (Ignor th
altrnation in rounding.)
(119) Noun Vrb
dc dc hunt
cio io think
ci ci fabricat
tu tu shoot
oku ku di
Ky Concpts
vowls assimilating to consonants vowl harmony [back]
[high] [low] [ATR]


It is also notd that th IPA has som spcial diacritics for marking [ATR]
, although th distinct lttrs prsntd
hr ar oftn usd in actual practic.
96
18.1 Try it for yourslf with Lamba
Th Applid suffix has two allomorphs. What ar thy?
102
Past Passiv Applid Rciprocal
1. tjito tjitwo tjitiIo tjitono do
2. tuIo tuIwo tuIiIo tuIono dig
3. tjcto tjctwo tjctcIo tjctono spy
4. sopko sopkwo sopkcIo sopkono pay tax
5. poto potwo potiIo potono scold
6. tjcso tjcswo tjcscIo tjcsono cut
7. koso koswo koscIo kosono b strong
What vowls occur in th syllabls prcding ach of ths allomorphs?
103
Which of ths allomorphs has th most rstrictd (and thrfor most xplainabl
) distribution?
104
Which allomorph is th bst choic for th undrlying form?
105
Complt th following rul.
106
V
C
0
V
[
low] [
high] [
low]
97
18.2 Try it for yourslf with Turkish
Considr only th vowl altrnations (a chart of usful faturs is p
rovidd blow th data). Two fairly
simpl ruls ar ndd to account for th following data (plural suffix and var
ious cas suffixs). What ar
thy? (You will not b abl to dtrmin th xact undrlying form of th suffix
s at this point, but you will
discovr som facts about thm.)

107
You might start this invstigation by listing th kinds of altrnations that occ
ur in th suffixs and th kinds
that do not. Provid drivations for th Gnitiv forms (singular and
plural) of tooth and slav; includ a
lin for th rul that palatalizs a latral consonant bfor a front vowl. Ass
um that th undrlying form
of th Plural is {Ior} and of th Gnitiv is {in}.
108
tooth hand ros villag daughtr stalk slav nd
Sg. Nom. dij c jy coj kmz sop kuI son
Acc. diji ci jyy cojy kmzm sopm kuIu sonu
Dat. cc cojc sopo sono
Gn. dijin cin jyyn cojyn kmzmn sopmn kuIun sonun
Pl. Nom. dijcr ccr jycr cojcr kmzIor sopIor kuIIor sonIor
Acc. ccri cojcri sopIorm sonIorm
Dat. ccrc cojcrc sopIoro sonIoro
Gn. dijcrin ccrin jycrin cojcrin kmzIormn sopIormn kuIIormn
ormn
[
round]
[+round]
[
back] [+back]
[
back] [+ back]
[+high]
i
m
y
u
[
high]
c
o
o
o
18.3 Try it for yourslf with Turkish again
Considr th following data. Can you account for all of ths data? If not, tll
what is th problm.
109
com laugh s tak run
to gc jy gor oI kos
I gcim gorym oImm kosum
I am ing gcijorum jyyjorum goryjorum oImjorum kosujorum
h is ing gcijor goryjor oImjor kosujor
98
Fdback for Chaptr 18
18.1. Lamba
102
Th allomorphs ar [
ilo] and [
lo].
103
Th allomorph [
ilo] follows syllabls with [i], [u] and [o]; th allomorph [

lo] follows syllabls with
[] and [o].
104
Th allomorph [
lo] is th most rstrictd. It only follows syllabls with clo
s
mid vowls, whras
[
ilo] occurs following syllabls with clos vowls and also nar
opn vowls.
105
Th allomorph [
ilo], bcaus w could not xplain why it follows syllabls wit
h opn vowls.
106
Add a dottd lin from th [
high] fatur of th first V to th scond V.
18.2 Turkish
107
Th plural suffix only shows th altrnation btwn c and o. Th othr suffix
s show altrnations
btwn front roundd, front unroundd, back roundd, and back unroundd. Non o
f th altrnations is
btwn high and non
high vowls.

sonI

108
Drivations:
UF kuI
in kuI
Ior
in dij
in dij
Ior
in
Roundnss Sprad kuIun kuIIorin dijin dijIorin
Backnss Sprad kuIun kuIIormn dijin dijIcrin
Latral Palatalization dijcrin
SF kuIun kuIIormn dijin dijcrin
18.3 Turkish
109
Th suffix {im} is no problm sinc it dos xactly what w xpct. Th first vo
wl of th suffix
{
ijor} also changs as xpctd, but th scond vowl changs nithr for backn
ss nor for roundnss.
Sinc it is not a high vowl, w dont xpct it to chang for roundnss, but w d
o xpct it to chang for
backnss. It is a problm if w xpct our rul of Backnss Sprading to apply t
o all suffixs, as w saw
happning in th noun data. (Th solution to this situation has bn proposd as
following: som suffixs
ar spcifid undrlyingly as bing [back], but som suffixs ar not spcifid f
or this fatur. Th fatur
sprads only to thos suffixs that ar not undrlyingly spcifid.)
99
Chaptr 19
Plac Assimilation (Non
nasals)
Consonants somtims assimilat partially or wholly to nighboring conso
nants. Considr th following
data from Sri.
(120) i
s
pii s/h will tast it i
j
join s/h will fnc it in
i
s
tis s/h will point at it i
s
mis s/h will rsmbl it
i
s
koo s/h will look for it i
s
noix s/h will grab it
i
s
sii s/h will smll it i
s
xost s/h will stp on it
Th futur prfix has two allomorphs in ths data: [s] and [j]. Th
allomorph [j] occurs only bfor [j].
Th morphm obviously has an undrlying form containing an s and it
assimilats to th point of
articulation of a following alvopalatal consonant. A rul such as th following
can b writtn:
(121) s bcoms j bfor j
If this rul can b statd mor gnrally and still b accurat, thi
s would b dsirabl. W might hav
thought that prhaps w could hav th rul b mor gnral in th following way
:
(122) s bcoms idntical to a following fricativ
But (122) is inaccurat bcaus it claims that s should bcom x bf
or x, but it dos not. Rul (121) is
corrct vn though it is vry spcific.
Th following data show anothr xampl of assimilation.
(123) i
[m]
atay who didnt go
po
[m]
atay if s/h dosnt go
t
[m]
atay if s/h didnt go
s
[m]
atay who will not go
k
[w]
atay dont go!
Th ngativ morphm shows up as [m] in ths forms, xcpt whn it
follows a k. Aftr k, a vlar
consonant, it bcoms a vlar consonant, a nasalizd w. Th rul (informally sta
td) is:
(124) m bcoms a nasalizd labiovlar approximant aftr a vlar consonant
Th Plac faturs of th k ar sprading to th nasal consonant, but it rmains
labial vn though it is also
vlar.

Th fatur [latral] may also sprad to adjacnt consonants (somtims vn if
a vowl is intrvning,
an important fact that w will ignor hr). In Tralfn Flmish, th suffix {n}
assimilats to a prcding
latral; undrlying {sp:l
n} bcoms [spc:II] (th final consonant is also syll
abic). This happns whn th
ngativ prfix in
of English is addd to a stm such as lgibl illgibl. It
has bn proposd that th
fatur [latral] is dpndnt on th Coronal nod (s Appndix A). Thrfor i
f th Plac faturs sprad,
th fatur [latral] will also automatically sprad.
Oftn an assimilation is vry slight and only noticd if th words h
av bn dscribd with a narrow
phontic transcription. For xampl, th j of Sri is typically rtroflxd, but
whn it prcds t, it loss its
rtroflxion. Also, th s which is usually alvolar in Sri, is mad dntal whn
it follows th (usually) dntal
t.
But th assimilation may also b vry drastic, vn whr a consonant
changs to bcom totally
idntical to an adjacnt consonant which is vry diffrnt from it. C
onsidr th following data from
Hungarian:
100
(125) 3 sg. Subjunctiv
roo
jon scold
kop
jon obtain
tud
jon know
vzj
jn lad
vs
sn prish
huuz
zon pull
Th j bcoms an s following s, an j following j, and a z following z. Th gnr
alization appars to b
that j assimilats totally to a prcding coronal continuant. A complt assimil
ation of this sort is dscribd
formally by th us of th Root nod which dominats all of th oth
r faturs. (Th rul blow spcifis
that th first consonant must b a coronal continuant and that th s
cond consonant must b a sonorant;
ths faturs may or may not b all that ar ncssary to corrctly limit th r
ul.)
In many languags, including Madija, /w/ is pronouncd as a voicd bilabial fric
ativ, [|, bfor front
vowls and [w] bfor back vowls. (In Madija, /w/ dos not occur b
for /o/. Th front vowl is
ssntially fronting th back vlar approximant.
(127) [w]opo monky (spcis) ha[]i trail
o[w]o tr []ni rivr
[w]opi many []i dig
ho[w]o shout o[]i xtinguishd
Vowls may also assimilat to an adjacnt consonant, as sn in chaptr 18.
Ky Concpt
plac assimilation
(126)
C
C
|
|
Root
Root
|
|
[+cont] Plac
Plac [+son]
|
Coronal

101
19.1 Try it for yourslf with Xavant
What is th undrlying form of th prfix for scond prson possssor
?
110
Giv th phonological rul
ncssary to account for th allomorphy, using formal notation.
111
your
1. du 1oddu stomach
2. 1ro 1oj1ro child
3. hi1roti 1ojhi1roti kn
4. t 1ott y(s)
5. 1wo 1oj1wo tooth
6. jc:rc 1ojjc:rc hair
7. po:ro 1ojpo:ro foot
8. bo:bo 1ojbo:bo fathr
19.2 Try it for yourslf with Cairo Arabic
What is th undrlying form of th prfix which appars in th scond column?
112
Giv th rul ncssary
to account for th allomorphs.
113
th...
1. nuus innuus popl
2. kursi ilkursi chair
3. durs iddurs lsson
4. buub ilbuub door
5. sitt issitt woman
6. gb ilgb pockt
7. loon illoon color
8. ju11u ijju11u apartmnt
9. nimiu innimiu grad
10. suti issuti lin
11. juntu ijjuntu bag
12. muduiiis ilmuduiiis tachr
13. qism ilqism sction
14. wuugib ilwuugib assignmnt
102
19.3 Try it for yourslf with Slayars
Assum that th word for six blow nds in a nasal consonant undrly
ingly; it is irrlvant whthr you
know what nasal consonant that is. Giv th rul(s) ndd to account for th d
ata. Provid a drivation for
th phrass six gardns and six bananas.
114
1. onnom pokc six spars
2. onnon tou six prsons
3. onnon +upo six kinds
4. onnop joron six horss
5. onnop koko six gardns
6. onnoI Ioko six bananas
Fdback for Chaptr 19
19.1 Xavant
110
Th undrlying form is {?oj}.
111
Th phonological rul is an assimilation rul: j assimilats totally to a follo
wing coronal consonant. Th

rul almost crtainly nds to b rstrictd to apply only to j, hnc th fatu
rs [+son], Dorsal. Not: a
dottd lin nds to b drawn btwn th top lft C and th Root nod on th righ
t, to indicat th total
assimilation in vry fatur.
C
C
|
|
Root
Root
/ |
|
[+son] Plac
Plac
|
|
Dorsal Coronal
19.2 Cairo Arabic
112
Th undrlying form is {iI}.
113
Th phonological rul is an assimilation rul: /I/ assimilats totally to a fo
llowing coronal consonant.
Th following rul is formulatd xplicitly to only apply whn th first consona
nt is /I/ may b adquat in
trms of faturs, or it may nd to dlimit th first consonant in som way No
t: a dottd lin nds to b
drawn btwn th top lft C and th Root nod on th right, to indicat th assim
ilation.
C
C
|
|
Root
Root
|
|
Plac Plac
|
|
Coronal Coronal
|
[+latral]
103

19.3 Slayars
114
Th phonological rul is an assimilation rul: a nasal assimilats to th plac
 of a following consonant.
If that consonant is a latral, th nasal bcoms a latral. This should follow
automatically if th fatur
[latral] is dpndnt on th Coronal nod, as mntiond in th txt. This is a
cas whr corrct fatur
gomtry should giv th corrct rsult, making it unncssary to hav a sparat
 rul for assimilation to /l/
or additional faturs. W also assum that whn th nasal bcoms [+latral],
it must bcom [
nasal].
UF onnon koko onnon Ioko
Assim. onnop koko onnoI Ioko
SF onnop koko onnoI Ioko
C
C
|
|
[+nasal] Plac
Not: a dottd lin nds to b drawn btwn th top lft C and th Plac nod on
th right, to indicat
th assimilation.
104

Chaptr 20
Dissimilation and th Obligatory
Contour Principl
Th phonological procsss which w hav considrd in th past svra
l chaptrs hav all bn
assimilatory. Nvrthlss, xampls of dissimilation ar also found in languag
s, although thy ar much
lss common than assimilatory ons. Dissimilation is th procss by which two si
milar sounds bcom lss
lik ach othr.
A possibl xampl of dissimilation from Sri is th chang from /I/
to /j/ whn a suffix with /I/
immdiatly follows.
a
(128) ko:pI + iI ko:pj + iI suffocat
knopI + iI knopj + iI sink
Kuman also has a rul of latral dissimilation. Whn a suffix containing l is a
ddd to a stm that nds
in l, th first l bcoms [r]. Thus, th root {joI} changs to [jor] whn it is
followd by th suffix {aI}.
In North Publa Nahuatl, a k dissimilats to th vlar fricativ [x]
whn it prcds anothr k. Thus
okikakk thy hard it coms out as [okikaxk]. Similarly, th affricat
t j dissimilats to j whn it
prcds tj. Thus th morphm n:tj bcoms [nc:j| bfor th root {tjontio} r
sid.
Dissimilation has gnrally bn formalizd in rcnt yars as th loss of a cr
tain fatur spcification
which is triggrd by th prsnc of an idntical fatur in th nvironmnt. H
rs how th rasoning has
gon. First, considr th mattr of a long k sound. How would this
b rprsntd formally? Is it two
consonant positions that ach indpndntly has th faturs of k (as
in 129a), or is it two consonant
positions that shar th faturs of k (as in 129b)?
(129)
This may sm lik just a littl diffrnc, but th issu is quit
important. Dos a languag contain
both kinds of structurs? Do thy contrast? If so, how dos a spak
r know th diffrnc? Aftr
considrabl dbat during th past coupl of dcads, th gnral ans
wr has bn quit clarly that th
structur in (129b) is th corrct on, or at last th common on,
morphm
intrnally. Th nam of th
univrsal principl is th Obligatory Contour Principl, or OCP, for s
hort. This nam cam from th
original contxt in which this was proposd, which had to do with tons.
b
If a thr syllabl word had th
tons High
Low
High pronouncd on thm, th tons would b just thos: H
L
H.
But if a thr syllabl
word had high ton on ach syllabl, th OCP says that it is actual
ly on high ton that is shard by th
thr syllabls. Th ton pattrn has to hav a contour it cant b H
H
H, that is,
without a contour.
So what dos this hav to do with dissimilation? Considr again th
Nahuatl facts mntiond abov.
Th root is {kok} and a suffix is addd that bgins with /k/. So somthing that
looks just lik (129a) is
cratd by th word formation ruls of th languag. Nvrthlss, this is still
a violation of th Obligatory

Contour Principl. What do languags do? Actually, thr

inds of rpair stratgis. On of
a

ar

various k

This rul is vry dp in th phonology and is not immdiatly obvious.
b
John A. Goldsmith (1979) Autosgmntal phonology. [1976 Ph.D. Dissrta
tion, MIT.] Nw York: Garland Prss.
Th OCP was statd as: Adjacnt idntical lmnts ar prohibitd.
a. [Faturs of k] [Faturs of k]
|
|
C
C
b.
[Faturs of k]
C
C
105
ths is simply to mrg th faturs, and mak thm into (129b) formally. Anoth
r stratgy is to liminat
on of th offnding faturs. And this sms to b what is going on with cass
of dissimilation: a fatur is
dlinkd bcaus of prssur from th OCP.
Th loss of a fatur may thn b compnsatd for by th insrtion
of som dfault fatur.
For
xampl, in th North Publa Nahuatl cas th fatur [continuant] on th first k
is droppd in th prsnc
of th fatur [continuant] of an affix k. Th assumption is thn tha
t som rul supplis th fatur
[+continuant] in this languag in thos cass whr th consonant is not othrwi
s spcifid for this fatur.
Ky Concpt
dissimilation
20.1 Try it for yourslf with Tzltal
Th prfix in th data blow indicats third prson; th suffix, which has two a
llomorphs, indicats
inanimat possssor (compar th first word with [spok] his cloth). Dscrib wh
at appars to b going on
hr.
115
What fatur is bing dlinkd?
Root its
1. pok
spokuI cloth
2. wits switsuI hill
3. Icw sIcwuI fat
4. Ium sIumiI ground
5. poj spojiI mdicin
6. si1 si1uI firwood
20.2 Try it for yourslf with Lalao Chinantc
Th words [go] and [ko] indicat Rcnt Past. Th word [go] (mid to
n) is a scond prson dirctional
imprativ. Th words [go ] and [ko ] (high ton) ar usd to indic
at Optativ mood. Choos an
undrlying form for ach of ths thr morphms and dscrib what is going on.
116
What fatur is bing
dlinkd?
1. go hmc: to h did work
2. ko do1 hmc: to thy did work
3. ko do1 to:j thy fll
4. go to:j h fll
5. go hmc: to go do work!
6. go bo go strik it!

7. go hmc: to may h do work!

8. ko do1 hmc: to may thy do work!
106
Fdback for Chaptr 20
20.1 Tzltal
115
Th suffix is probably undrlyingly {
uI}; this is th allomorph that occurs aft
r front vowls and th low
vowl /o/. It dissimilats to [
iI] whn it follows a round vowl. Th fatur
[+round] is bing dlinkd.
W must also assum that th loss of [+round] also ntails th loss of [+back],
ithr by convntion or
stipulation. This is an intrsting topic that w do not xplor hr.
20.2 Lalao Chinantc
116
Undrlying forms: Rcnt Past {ko},
Imprativ {go},
Optativ {ko}
Notic that th Imprativ morphm dos not vary in voicing (it always bgins w
ith [g]), but that th
Optativ morphm varis btwn [k]and [g]. It sms rasonabl to posit diff
rnt undrlying forms
for thm: th voicd on stays voicd, th voiclss on (lik th Rcnt Past)
varis in voicing.
Dissimilation: A voiclss consonant (ths ar all vlars) bcoms voicd if t
h nxt syllabl bgins with
a voiclss consonant.
Th fatur in qustion is [voic] (which sms odd, actually; usually it is th m
arkd valu of a fatur
that is dlinkd).
107
Chaptr 21
Miscllanous
Som phonological procsss which ar oftn found in languags ar hard to class
ify by labls currntly in
us. Th phontic motivation for som of ths ruls is not asy to
xplain (vn if w know what it is).
Nvrthlss, it is good to b awar of som phonological ruls that ar lss w
ll
undrstood, vn though
thy ar not all common. (Th list blow is growing informally with th drafts o
f this book, as tim prmits.
Latr chaptrs in this book dal with th common procsss of insrtion and dl
tion.)
Aspiration of s
In som dialcts of Spanish, thr is a tndncy to pronounc syllabl
final s a
s [h]. In Capanahua, any of
th thr sibilants of th languag (alvolar, alvopalatal, and rtroflxd alv
opalatal) tnds to b rplacd
by [h] or dltd in fast spch whn it prcds th flap r.
(130) bo1kijri1bi bo1kihri1bi bo1kiri1bi tomorrow, too
hisisri1bi hisihri1bi hisiri1bi ant also
morori1bi morohri1bi morori1bi poison also
Altrnation btwn l and r
Liquids (l and r sounds) altrnat in som languags. In Koran, th
liquid is ralizd as somthing mor
lik an [l] whn it is syllabl
final; othrwis it is mor lik [r].
(131) po[I] foot u[r]i w
to[I] ston mo[r]u woodn floor
i[I]kop svn
In Madija ths sounds ar variants of a singl phonm. Th liquid
is ralizd as [l] whn it is
simultanously prcdd and followd by i; othrwis it is ralizd as [r].

(132) po[r]c pircd hi[l]i sing

o[r]i paddl
wi[l]idi grandfathr
o[r]o tr (spcis)
dzc[r]o grass
c[r]ibo our ars
hi[r]cc b without
Labialization bfor unroundd vowls
In Capanahua, th labial consonants may b labializd whn thy prcd a high b
ack unroundd vowl.
a
(133) kop[]m alligator m[] m1i bark
p[] m1i laf 1om[]mn capybara
Variation in vowls
In many languags, vowls hav a rang of phontic variation that is conditiond
by many factors. Considr
th following statmnts rgarding th variation in th thr vowls (a, i, u) o
f Inga (Lvinsohn 1976, p. 26;
th symbols in th quot hav bn adaptd to IPA):
a

Loos 1967 spcifis that th consonants which may b labializd ar p and m; ap
parntly b and w ar not labializd
by this rul. It is also said that th rul is a rgular fatur with som spak
rs, but absnt with othrs or usd spora

dically (p. 183). Th procss is dscribd using th acoustic fatur [flat], an
d is givn in a formulation that would
suggst it is a kind of dissimilation. Th consonant bcoms [+flat] bfor a [

flat] vowl (with othr faturs spcifid
in th rul).
108
/o/: has a variant [] which varis frly with [o] word
initial, and in syllabls
closd by a nasal.
/i/:
has a variant [] which varis frly with mid
low [c] non
fina
lly following /d, n, l, r, j/
and
prcding /p/. [] varis with [i] non
finally following /s/. [] occu
rs finally, xcpt following non

contiguous [i].
/u/:
has various allophons, including [o], [o], [5], and [u]. Th
distributional statmnts ar quit
complx, and th ruls accounting for thm would also b complx.
109
Summary and Rviw Qustions for Sction 2
An xamination of morphms oftn shows that a singl morphm has mor than on
shap (allomorph),
and that ths shaps ar similar phontically. Th standard practic within gn
rativ phonology has bn
to posit a singl undrlying form for such morphms and to account
for th diffrnt shaps by gnral
phonological ruls. Th most common phonological ruls ar assimilatory, in whic
h a fatur (or group of
faturs) of on sound sprads to anothr sound. (Som othr vry common ruls r
lat to syllabl structur
and ar tratd in a latr chaptr.)
Undrlying forms ar chosn and phonological ruls ar proposd in ord
r to account for th facts in
th most straightforward fashion using th smallst amount of languag
spcific
complications. Th guiding
principl of gnrativ phonology was minimum storag, maximum computation.
W hav only minimally introducd th rol of th Obligatory Contour
Principl in phonological

thory. Som of th rfrncs blow contain a much mor dtaild xplanation.
For Furthr Rading:
Andrson, Stphn. 1974. Th organization of phonology. Nw York, London, Acadm
ic Prss.
Chomsky, Noam and Morris Hall. 1968. Th sound pattrn of English. N
w York, Harpr and Row.
(Rprintd in 1991 by MIT Prss.)
Clmnts, Gorg N. 1985. Th gomtry of phonological faturs, Phonology Yarb
ook 2:225
52.
Clmnts, G. N. and Elizabth V. Hum. 1995. Th intrnal organization
of spch sounds. In John A.
Goldsmith, d., Th handbook of phonological thory, Blackwll: Cambridg, Mass.
, pp. 245
306.
Durand, Jacqus. 1990. Gnrativ and non
linar phonology. Longman: Nw York an
d London.
Goldsmith, John A. 1979. Autosgmntal phonology. [1976 Ph.D. Dissrtation, MIT.
] Garland Prss: Nw
York.
Hulst, Harry van dr, and Jron van d Wijr. 1995. Vowl harmony. In John A.
Goldsmith, Th hand

book of phonological thory, pp. 495
534. Blackwll: Cambridg, Mass.
Knstowicz, Michal. 1993. Phonology in gnrativ grammar. Blackwll: Cambridg
, Mass.
Mohanan, K. P. 1986. Lxical phonology, Dordrcht, Ridl.
Oddn, David (1986). On th rol of th Obligatory Contour Principl
in phonological thory. Languag
62:35383.
Oddn, David (1988). Anti antigmination and th OCP. Linguistic Inquiry 19:45175
.
Sagy, Elizabth. 1986. Th rprsntation of faturs and rlations in
non
linar phonology, Ph.D.
Dissrtation, Massachustts Institut of Tchnology.
Yip, Moira (1988). Th obligatory contour principl and phonological ruls: a los
s of idntity. Linguistic
Inquiry 19:65100.
Rviw Qustions
Th following qustions ar to hlp you rviw th matrial in th prcding sc
tion.
117
1. Allomorphs which ar similar phonologically ar usually bst dscrib
d as bing drivd from a
singl undrlying form and th application of a _______________ rul.
2. Sounds oftn bcom lik, or _____________ to, nighboring sounds.
3. Th chang from z to s in th English plural is an xampl of _____________ a
ssimilation.
4. Sounds (such as s) which ar producd with a significant obstructio
n in th mouth, causing an
incras in air prssur, ar calld __________________.
110
5. Sounds (such as l) which ar producd with lss obstruction in th mouth and
mor opportunity for
rsonanc in th vocal tract ar calld _________________.
6. Nam th thr major plac faturs: ___________, __________, ____________.
7. Nasals oftn assimilat in ____________ faturs to an adjacnt consonant.
8. Combinations of distinctiv (unprdictabl) faturs in a languag ar common
ly rfrrd to as th
______________ of th languag; th inclusion of a non
contrastiv fat
ur producs an
_____________.
9. Assimilation may b bst formalizd as th _________________ of on

fatur (or group of

faturs) from on unit to anothr.
10. Som ruls apply only in ____________ nvironmnts, such as whn t
wo morphms com
togthr.
11. Th chang from t to tj bfor i is an xampl of _______________.
12. Th rounding of t in th word tool is an xampl of _______________.
13. Th fatur which is sprad to th vowl in th word can [kon] i
s th fatur
___________________.
14. Th assimilation of a vowl of on syllabl to a vowl in anoth
r syllabl is calld vowl
_______________.
15. Th chang of on sound to bcom lss lik a nighboring sound
is calld
____________________.
16. What typ of procss is illustratd by th following data?
{nitos} [nitjos] ________________________
17. Allomorphs must b listd in th lxicon
(a) always
(b) whn
thy cannot b prdictd by
phonological ruls (c) whn thy ar accountd for by phonological ruls.
18. Circl th sgmnts that ar [+sonorant]. o
k
r
z
w
n
s
I
19. (T or F) Fatur sprading is a vry common typ of phonological rul.
111
20. Th following data show th nd for a rul that assimilats th vowl of on
 syllabl to bcom
idntical to th vowl of th prcding syllabl. What might b on rason for w
hy this rul dos
not chang th scond vowl of ritu to b icy to b i ?
Tarahumara
ritu to b icy rituku ic
rm to mak tortillas rmk tortillas
potji to grow ars of corn potjiki ar of corn
opotjo to b drssd opotjoko garmnt
21. Giv th valus of th faturs [high] and [low] of mid vowls:
________________________________________.
If you hav mastrd th matrial in this sction, you should b abl to xamin
data and
(a) distinguish suppltiv allomorphy from non
suppltiv allomorphy
(b) rcogniz phonological allomorphs which ar rlatd by phonological assimila
tory procsss
(c) dscrib xampls of voic assimilation, plac assimilation, mannr
assimilation, and
nasalization in pros and with formal notation
(d) dscrib xampls of palatalization, labialization, and vowl harmony in pro
s
() xplain th importanc writing gnral ruls
(f) xplain how th application of gnral ruls ruls may b rstric
td if th rul is structur

prsrving or if it applis only in drivd nvironmnts
(g) follow widly
accptd principls for choosing undrlying forms
Fdback for Rviw Qustions
117
1. phonological 2. assimilat 3. voicing 4. obstrunts
5. sonorants 6. labial, coronal, dorsal 7. plac
8. phonms, allophon 9. sprading 10. drivd 11. palatalization
12. labialization 13. [nasal] 14. harmony
15. dissimilation 16. palatalization17. (b) 18. o r w n I 19. T

20. It is a rul that applis only in a drivd nvironmnt. 21. [
high] [
low
]
112
Sction 3
Phonological Ruls:
Som Practical Procdurs
This sction has th subtitl "Som Practical Procdurs" bcaus it o
utlins som stps which ar oftn
takn in th initial stags of work on an unwrittn languag. W ar
prsnting ths stps now, aftr
xtnsiv discussion of assimilatory procsss, sinc it is important t
o know what kinds of procsss on
should b looking for whn on is doing such procdurs.
Th stps illustratd in this sction ar important for clarifying wha
t faturs and combinations of
faturs ar distinctiv in a languag, and hnc which faturs will
b found in undrlying forms of that
languag. Ths stps, commonly rfrrd to as phonmic analysis, hav bn cons
idrd crucial for undr

standing nough of th phonology of a prviously unwrittn languag to b abl t
o dvlop a writing systm
for it.
113
Chaptr 22
Contrastiv Faturs
W hav sn on way to find out whthr a fatur (or combination
of faturs) nds to b prsnt in
undrlying rprsntations of a languag. If w writ a rul (to account for som
 allomorphy, for xampl)
which can account for all of th occurrncs of a fatur/combination
in th languag, that
fatur/combination will not appar in th lxical rprsntation of an
y morphm. (Th combination is
thrfor not a phonm, although it is an allophon of a phonm.)
A good xampl of this was sn in chaptr
19 whr w saw that [w] in Sri was an
allomorph of th ngativ morphm which is
basically [m]. Onc w wrot a rul accounting
for this altrnation, w would chck around and s if that rul (al
ong with anothr on, prhaps) would
account for all of th occurrncs of [w] in th languag. And as a mattr of fa
ct, thy do in Sri. Thr is
no phonm [w]. All occurrncs of [w] follow a vlar consonant; and
thr ar no occurrncs of [m]
following a vlar consonant.
a
Not all allophons ar discovrd through th xamination of allomorphi
c altrnation, although such
xamination may b vry important. Many allophons ar found by looking at th d
istribution of sounds.
This is usually a job for th analyst. Nativ spakrs may b compl
tly unawar of phontic variation
which to an outsidr is quit obvious.
Th most obvious vidnc that a fatur/combination must appar in a
lxical rprsntation of a
morphm is that you simply cannot xplain it in any othr way. For
xampl, considr th sound h in
English. It is a sound which dos not occur just anywhr; in fact,
it only occurs at th bginning of
syllabls: har, hold, rhat, bhold. On might wondr whthr it must appar i
n th lxical rprsntation

of th morphm HEAR. It must, sinc w must distinguish btwn ar

and har, and btwn old and
hold, and thr is no obvious othr way to distinguish ths words phonologicall
y.
b
Thrfor, w say that
/h/ is a phonm of English. That is, th faturs or combination of faturs th
at ar ncssary to rprsnt
[h] (or distinguish it from othr sounds) must b prsnt undrlyingly in Englis
h.
Words such as old and hold ar calld minimal pairs. Th only diffr
nc btwn thm can b
narrowd down to th diffrnc btwn th prsnc and absnc of h;
th diffrnc is minimal but is
significant. (W will assum that minimal pairs must b idntical word
s xcpt for th sgmnts in
qustion.)
If th simpl rplacmnt of phon x by phon y in a word rsults
in a chang of maning, th two
phons ar said to contrast in idntical nvironmnts (CIE).
c
Considr th two phons [J] and [I] of
English. If w tak th uttranc [Jip] and rplac th [J] with [I]
, th rsulting uttranc [Iip] mans
somthing diffrnt. Rip and lip ar a minimal pair. This is takn as initial v
idnc that th two phons [J]
and [I] ar rlatd to distinct lmnts in th lxical rprsntations, probabl
y th phonms /J/ and /I/. W
us th following modl to summariz ths obsrvations.
a

This is slightly simplifid; th rul must actually spcify that th vlar cons
onant must b in th sam syllabl as th
m, although th xampls ndd to show this ar xtrmly rar. A coupl of dc
ads ago, thr was som optionality
to th rul in qustion, and so som words tndd to fluctuat btwn having an
[m] and having a [w].
b
Th prsnc of a glottal stop bfor initial vowls in som dialc
ts of English in crtain contxts (spcially
uttranc
initial position) clouds th issu a bit.
c
Procdurs for discovring phonms ar laboratd in dtail in txtb
ooks such as Knnth L. Pik (1947)
Phonmics: A Tchniqu for Rducing Languags to Writing, Ann Arbor, U
nivrsity of Michigan Prss. Som
assumptions mad in work of that priod ar not mad today. Our prs
ntation blow includs som warnings about
possibl rrors that analysts somtims mak.
Major Point: If a givn fatur may b accountd for
ntirly by a rul, it should not appar in any
undrlying form.
114
(134) [J] and [I] contrast in idntical nvironmnts
or
[J] : [I] CIE (at last word
initially)
.g. __ip [Jip], [Iip]
__ik
[Jik], [Iik]
__ok
[Jk], [Ik]
Probabl implication: [J] and [I] blong to sparat phonms.

Lxical rprsntations (assuming othr sounds also to b phonmic): {Jip}, {Ii

p}, {Jik}, {Iik},
{Jk}, {Ik}.
W tak this vidnc as bing only initial vidnc for contrast sin
c th facts of th languag as a
whol must b takn into account. For xampl, thr ar words such as [pcro] bu
t (with a flap) and [pcro]
dog (with a trill) in Spanish which might b (and hav bn) takn
as vidnc for two phonms in this
languag. Nvrthlss, th contrast is found only intrvocalically. An altrnat
iv analysis is that th trilld
r is th phontic ralization of /rr/, sinc /r/ can occur at th nd of a sylla
bl and also at th bginning of a
syllabl in this languag.
d
Th minimal pair shown abov argus for a diffrnc in th undrlying forms
of ths words, but that diffrnc may not b th diffrnc which shows up pho
ntically.
22.1 Try it for yourslf with Karuk
Using a format similar to that shown abov, prsnt th vidnc that
vowl lngth is contrastiv in this
languag.
118
1. tos to build fnc 4. xos thn, but
2. pi:p to say 5. to:s plant (spcis)
3. xo:s almost 6. pip
to sting
22.2 Now try it for yourslf with Sri
Using a format similar to that shown abov, prsnt th vidnc that th diffr
nc btwn [s] and [j] is
contrastiv in Sri.
119
1. nos milkwd pod 4. ko:jx string up (nom.)
2. ko:sx unfold (nom.) 5. noj bush (spcis)
3. ji:m wd (spcis) 6. si:m will slp
If two sounds do not contrast in idntical nvironmnts, on must bgin to look
to s whthr on of
thm occurs in an nvironmnt which might affct its pronunciation. Th
at is th point of looking for
minimal pairs, whn possiblth nvironmnt is bing controlld compltl
y. If th nvironmnt is
clarly not affcting th pronunciation, thn w may say that th two
sounds contrast in noninfluncing
nvironmnts (CNE) although th nvironmnt is not idntical.

For xampl, th following vidnc
would b sufficint to stablish that [m] and [n] blong to diffrnt phonms i
n English, vn though thr
ar no minimal pairs in ths data.
d

S Jams W. Harris (1983) Syllabl Structur and Strss in Spanish:

A Nonlinar Analysis, Cambridg, London,
MIT Prss.

Idntical nvironmnts ar of cours simply on spcial kind of noninfluncing
nvironmnt. (In othr prsntations
of phonmic analysis, this kind of nvironmnt is rfrrd to as analogous nviron
mnt.)
115
(135) [mt] mat [muwn] moon

[nb]
nab [nuwt] nwt
Th vidnc would b summarizd as follows:
(136) [m] and [n] contrast in noninfluncing nvironmnts.
or
[m] : [n] CNE (at last word
initially)
.g. bfor th vowl []: [mt] [nb]
bfor th vowl [u]: [muwn] [nuwt]
Probabl implication: [m] and [n] blong to sparat phonms (/m/ and /n/ proba
bly).
Lxical rprsntations (assuming othr sounds also to b phonmic): {mt}, {nb},
{muwn},
{nuwt}.
If w wr to discovr minimal pairs latr, such as mat and gnat, our conclusion
s would b confirmd. But
such minimal pairs ar not ncssary or vn always possibl. (Rathr,
thy ar mor lik icing on th
cak.)
Again, vidnc from contrast in noninfluncing nvironmnts must b ca
rfully considrd from th
prspctiv of th rst of th languag. For xampl, in English on
might tak words such as [ mijn]
mission and [ vin] vision and claim that English has th phonms /j/
and //. But this would b an
unwarrantd conclusion. Thr is no dfct in th importanc of th pair, howv
r. What such a pair shows
is that in English thr is somthing diffrnt about th undrlying
form of mission from th undrlying
form of vision byond th diffrnc btwn m and v. But it dos n
ot tll us what this diffrnc is. It
might b that th diffrnc is btwn /sj/ and /zj/, or btwn /j
/ and /zj/, or btwn /s/ and /z/, or
somthing ls. On must b carful about th conclusion on draws.
For anothr xampl, in Sri it was obsrvd that many words apparnt
ly nding in /k/ could b
pronouncd with ithr [k] or somthing lik a voiclss rlas [k] in uttran
c
final position: [?ok],
[?ok] wood. Thn on word was found which was always pronouncd with [k] (th vo
iclss rlas)
and did not altrnat with [k] (th voicd rlas): [?ojoo k] anklbon. Shoul
d this b takn as clar
vidnc that w hav two phonms, /k/ and /k/?
No, this would b
 a dramatic stp basd on such
littl vidnc. Th mystry was solvd latr: th word anklbon nds in a vla
r fricativ /x/ which sounds
just lik labializd aspiration in this position.
f
Th two words ar phonmically /?ok/ and /?ojokx/.
Th contrast btwn th two words xists, but th diffrnc is not localizd i
n th aspiration itslf, but in
th prsnc of anothr sound.
22.3 Try it for yourslf with Kuskokwim Eskimo
Prsnt th vidnc that [k] and [q] contrast in this languag.
120
1. nuIiq wif 4. nuuq buttocks
2. kouk thundr 5. qiIuk bark
3. koIiqoq papr 6. uu:ok chk
f
Th prsnc of th /x/ might b suggstd by indirct vidnc havin
g to do with crtain allomorphy. Dirct

vidnc was found whn furthr probing producd th uttranc [?o jokoxojo] an
klbon, my y!, which includs
a suffix [
ojo] and an infix [o]. Not th [x] which appars bfor th suffix [
o
jo].
116
22.4 Try it for yourslf with Vnda
Prsnt th vidnc that alvolar [n] and dntal [n ] contrast in Vnda.
121
1. non u at your plac 6. cnc h
2. I in o tooth 7. nono childhood
3. mun c mattr 8. zit onu fiv
4. mon o four 9. zino now
5. ononi s (pl.)! 10. kouno thr
11. n uri buffalo
Ky Concpts
distributional vidnc vs. allomorphic vidnc contrast in idntical nvironm
nts
contrast in noninfluncing nvironmnts minimal pair
For ach of th following xrciss, giv th rsults in a mannr similar to tha
t shown abov.
22.5 Try it for yourslf with Damana
Which fricativs ar contrastiv in Damana (just compar ach fricativ
with ach othr fricativ at this
point)? Giv th vidnc, bing sur to considr ach possibl pair (six).
122
1. sigi bothr 7. ojcgo b satd
2. oscgo cut off 8. o to at
3. jo hair 9. jigi tomorrow
4. omojo to sing 10. sino arrow
5. u
narrow 11. zu
good
6. zino
papr 12. jomojo
widow
22.6 Try it for yourslf with Guanano
Which stops ar contrastiv in Guanano? (Just compar th stops at th sam poin
t of articulation with ach
othrnin pairs.) Giv th vidnc.
123
1. duoho I sll 7. bo:ho I swim
2. po:ro toy 8. tuoho I hav strngth
3. po:ho I play 9. tuoho I rturn hom
4. koho I fl (pain) 10. ko?oko nar
5. po:ro stomach 11. ko?ogo gt drunk!
6. koho I tak a fish out of a trap
117
22.7 Try it for yourslf with Sirra Miwok
Prsnt th vidnc that alvolar [t] contrasts with dntal [t ].
124
1. t o:tji? ldr brothr 5. hijo:t o? astrn
2. how:otu? bads 6. tojijmu? jay
3. ?utu: Gt out of th way! 7. ?cno:t soon
4. not u:wok on far sid 8. ?otkimc? two days
22.8 Try it for yourslf with Agusan Manobo
Compar [p] with [b], [t] with [d], and [k] with [g]. Prsnt vidnc of contra
st in word
initial position,
intrvocalic position, and word
final position.
125
1. pitu1 svn 11. bogo1 lung
2. tobo1 fat 12. bu1uk pic
3. boto1 child 13. bitu1 hol
4. topo1 saltd mat 14. hipog othr sid
5. boko1 jaw 15. 1oncd float

6. kondip goat 16. dogho chst

7. bogo mbr 17. ?usob again
8. ?usip to ask 18. poIi1 wound
9. bognct wd 19. bu1op foolish
10. bodo1 pass 20. gomit us
22.9 Try it for yourslf with Eastrn Quichua
Prsnt vidnc that thr is contrast btwn th following pairs of sounds:
[t] and [ts], [t] and [r], [ts]
and [r], [k] and [g]; all of th nasals (all possibl pairs).
126
If thr is no vidnc of contrast, say so.
1. koru far 11. tsoko rough
2. roku thick 12. moti gourd bowl
3. indi sun 13. turi woman s brothr
4. mopgo pot 14. poni man s sistr
5. kuri gold 15. popo woman s sistr
6. tikto trap 16. rosu snow
7. rumi ston 17. kipgi twistd
8. tosin nst 18. tsinzu thin child
9. nopgi bird trap 19. tomio rain
10. powi y / fac 20. xuktu hol
118
22.10 Try it for yourslf with Tojolabal
Prsnt vidnc that thr is contrast btwn th following pairs of sounds: [
k] and [k]; [p] and [p] (th
lattr is actually an implosiv rathr than an jctiv); [t] and [t]; [t] and
[tj]; [ts] and [tj].
127
1. ki sim my bard 10. tso moI prtty
2. po koI corncob 11. tjo ko chop it down
3. sok whit 12. 1ok rd
4. tji tom pig 13. tjo tot plant (spcis)
5. tjo moI rottn 14. ko nip squash flowr
6. po tot plant (spcis) 15. no hot long
7. ti non upsid down 16. nu pon marrid
8. po kon stting 17. 1i not sd
9. 1i poj armadillo 18. tsokop thrad
Fdback for Chaptr 22
22.1 Karuk
118
[V] : [V:] CIE
.g. t__s [tos] to build fnc, [to:s] plant (spcis)
p__p [pip] to sting, [pi:p] to say
Probabl implication: long vowls and short vowls ar distinguishd ithr by b
ing sparat phonms or
by bing vowl clustrs.
22.2 Sri
118
[s] : [j] CIE finally ([nos] milkwd pod, [noj] bush (spcis))
[s] : [j] CIE prconsonantally ([ko:sx] unfold, [ko:jx] string up)
[s] : [j] CIE initially ([si:m] will slp, [ji:m] wd (spcis))
Probabl implication: [s] and [j] blong to sparat phonms.
22.3 Kuskokwim Eskimo
120
.g. [kouk] thundr vs. [nuuq] buttocks
[k] : [q] CNE
Probabl implication: [k] and [q] blong to sparat phonms.
22.4 Vnda
121
[n] : [n ] CNE
([nunu] childhood and [mun u] four)

Probabl implication: [n] and [n] blong to sparat phonms.

119

22.5 Damana
122
[s] : [j] CIE .g. [sigi] bothr, [jigi] tomorrow
[z] : [] CIE .g. [zu] good, [u] narrow
[s] : [z] CIE .g. [sino] arrow, [zino] papr
[j] : [] CIE .g. [jo] hair, [o] to at
[j] : [z] CNE .g. word
initially, bfor vowls [jo] hair, [zu] good
[s] : [] CNE .g. word
initially, bfor vowls [sigi] bothr, [o] to at
Probabl implication: [s], [j], [z] and [] blong to sparat phonms.
22.6 Guanano
123
[p] : [p] CIE .g. [po:io] toy, [po:io] stomach
[p] : [b] CIE .g. [po:ho] I play, [bo:ho] I swim
[p] : [b] CNE .g. [po:io] stomach, [bo:ho] I swim
[t]
: [t]
CIE .g. [tuoho] I hav strngth, [tuoho] I rturn hom
[t]
: [d] CIE
.g. [tuoho] I hav strngth, [duoho] I sll
[t] : [d] CIE .g. [tuoho] I rturn hom, [duoho] I sll
[k] : [k] CIE .g. [koho] I fl (pain), [koho] I tak fish out of a trap
.g. [ko?oko] nar, [ko?ogo] gt drunk!
[k] : [g]
CIE
[k] :
[g] No dirct vidnc in ths limitd data.
Probabl implication: voiclss unaspiratd, voiclss aspiratd, and voicd sto
ps blong to sparat
phonms.
22.7 Sirra Miwok
124
[t] : [t ] CNE
[t o:tji?] ldr brothr, [tojijmu1] jay)
Probabl implication: [t] and [t ] blong to sparat phonms.
22.8 Agusan Manobo
125
[p] : [b] CIE initially (svn, hol)
[p] : [b] CIE intrvocalically (saltd mat, fat)
[p] : [b] CNE finally (ask, again)
[t] : [d] C initially (fat, chst)
[t] : [d] CIE intrvocalically (child, pass)
[t] : [d] CNE finally (wd, float)
[k] : [g] CNE initially (goat, us)
[k] : [g] CIE intrvocalically (jaw, lung)
[k] : [g] CNE finally (pic, othr sid)
Probabl implication: [p], [b], [t], [d], [k], and [g] blong to sparat phonm
s.
120

22.9 Eastrn Quichua

126
[t] : [ts] CNE
[tosin] nst, [tsoko] rough
[t] : [r] CNE
[tosin] nst, [roku] thick
[ts] : [r] CNE
[tsoko] rough, [roku] thick
[k] : [g] no good vidnc sinc [g| only occurs aftr [p] (voicd contxt), wh
ras [k] dos not
[m] : [n] CNE
[tomio] ston, [poni] man s sistr
[m] : [p] CNE
[mopgo] pot, [powi] y / fac
[m] : [p] no good vidnc sinc [p| only occurs bfor [g] (vlar contxt) whr
as [m] dos not

[n] : [p] CNE [nopgi] bird trap, [powi] y / fac

[n] : [p] no good vidnc sinc [p| only occurs bfor [g] (vlar contxt) wh
ras [n]dos not
[p] : [p] no good vidnc sinc [p| only occurs bfor [g] (vlar contxt) wh
ras [p] dos not
Probabl implications: [t], [ts], and [r] blong to sparat phonms. Thr is
vidnc for thr nasal
phonms: [m], [n], and [p]. [p] may wll not b distinct from ths sinc it d
os not appar to contrast
with thm. [k] and [g] ar not shown to b in contrast by th data providd.
22.10 Tojolobal
127
[k] : [k] CNE
[sok] my bard, [1ok] rd
Probabl implication: [k] and [k] blong to sparat phonms.
[p] : [p] CNE [po kon] stting, [po koI] corncob
[nu pon] marrid, [1i poj] armadillo
[ko nip] squash flowr, [tso kop] thrad
Probabl implication: [p] and [p] blong to sparat phonms. (Evidnc in in
itial, mdial, and final
positions.)
[t] : [t] CNE [po tot] plant (spcis), [no hot] long
Probabl implication: [t] and [t] blong to sparat phonms.
[t] : [tj]
CNE [ti non] upsid down, [tji tom] pig
Probabl implication: [t] and [tj] blong to sparat phonms.
[ts] : [tj]
CIE [tso moI] prtty, [tjo moI] rottn
Probabl implication: [ts] and [tj] blong to sparat phonms.
Not: Th vidnc has still not bn prsntd that [t] and [ts] blong to dis
tinct phonms. Do you
xpct to find such vidnc?
121
Chaptr 23
Noncontrastiv Faturs
If a crtain sound x can nvr b found to occur in th sam or a
noninfluncing nvironmnt as sound y
occurs in, thn it may b that th two sounds ar variant pronunciations (bcaus
 of contxt)allophons
of th sam phonm. For xampl, in Tswana on finds syllabls of th following
typs:
(137) r
V
I
V
ru
kuru tortois
ri
bariso hrd

Io IoIcmc tongu

Ic IoIcmc tongu

Io xoboIo had
Th phon [I] occurs bfor [c, o, o], and th phon [r] nvr dos. Th phon [
r] occurs bfor [u, i] and
th phon [I] nvr dos. Th phons [r] and [I] ar said to occur
in complmntary distribution, or in
mutually xclusiv nvironmnts. Whr on occurs, th othr dos not occur.
a
Sinc thy nvr occur in
th sam nvironmnt, a statmnt of contrast in idntical nvironmnt is not po
ssibl for ths phons in
Tswana. Rathr, w can mak th following statmnts:
(138) [r] dos not contrast with [I]. Thy occur in diffrnt nvironmnts.
At this point w will simply assum that [r] is somhow mor basic
than [I] in Tswana and stat th
following:

b
(139) [r] : [I] CD
/r/ is pronouncd as [I] whn it immdiatly prcds nonhigh vowls
or
/r/ has th allophon [I] whn it immdiatly prcds nonhigh vowls
or
r I / ___ nonhigh vowls
What is happning hr is that th pronunciation of th sound /r/ is
affctd by its nvironmnt in
Tswana.
A word such as [IoIcmc] would b writtn phonmically as /rorcmc/.
A mthod which may hlp in finding out whthr two sounds occur in complmntary
distribution or
not is th us of y
diagrams, as shown blow (whr th squar brackt symbol "[
" mans th dg of th
word).
c
a
b

For anothr xampl, s th bginning of chaptr 22.

Th dcision rgarding which sound is mor basic is not basd on f

rquncy but rathr on rstrictivnss of
distribution and phonological simplicity (th last rstrictd sound or th mor
simpl sound is considrd mor basic
usually, it sms). That is, on dosn t count up th numbr of words that conta
in on allophon or anothr; on looks
at whr th sounds occur and considrs how many kinds of nvironmnts th allop
hons occur in.
Evn so, it is not always asy, and it may in som cass not vn b possibl or
ncssary to dcid on a basic phon.
Also, it should b notd that in traditional phonmic dscriptions, a
phonm such as /r/ was said to hav two
allophons: [r] and also [I]. Th important phonological fact sms to
b that th sound is a non
nasal coronal
sonorant, not whthr it is a latral or non
latral, a flap or not a flap.
c
Th symbol # is also commonly sn usd for this purpos, spcially in structu
ralist and arlir gnrativ work.
122
(140)
r
I

u u kuru tortois
o i
boriso hrd
[ o IoIcmc tongu
o c
IoIcmc tongu
o o
xoboIo had
Somtims it is ncssary to plac mor of th nvironmnt on ach s
id of th diagram; th influncing
nvironmnt may not ncssarily b immdiatly adjacnt. On should ch
ck at last as far as th nxt
syllabl.
Now considr th situation whr th substitution of on phon for anothr in so
m contxt is possibl
but th substitution rsults in no chang of maning. Th following d
ata from Spanish illustrat this
situation. Th phon [z] may b substitutd for th phon [s] in cr
tain nvironmnts, such as in fastr
spch styls:

(141) [mismo] sam [dcsdc] sinc

[mizmo] sam [dczdc] sinc
[sopo] frog [koso] hous
[zopo] (not possibl) [kozo] (not possibl)
Th phons [s] and [z] do not contrast in idntical nvironmnts sinc
 no chang in maning vr
rsults from th substitution of on for th othr in Spanish. Thrf
or, w hav no basis for considring
thm to both b phonms. Sinc th [z] occurs in a rstrictd nvir
onmnt, w can considr it to b a
variant of /s/ which occurs undr crtain conditions. This is somtim
s rfrrd to as conditiond fr
variation (FV)that is, two allophons altrnat with ach othr in th
sam contxt undr crtain con

ditions.
(142) [s]:[z] FV bfor voicd consonants (fast spch phnomnon)
or
/s/ is pronouncd as [z] whn it prcds a voicd consonant in fast spch.
Somtims th factors which affct th substitution of on phon for
anothr ar not wll
known. It
may hav to do with how tird th spakr is, how fast s/h is sp
aking, tc. In such cass, th
gnralization for fr variation might b statd as in th following xampl:
(143) /c/ is optionally pronouncd as [c] (in any contxt).
This is somtims rfrrd to as unconditiond fr variation, although th trmin
ology is not th bst.
For xampl, in various dialcts of Nahuatl (spokn in Mxico) crtain short vow
ls ar said to frly
vary btwn tns and lax variants.
d
A word such as /kowit/ tr is pronouncd as ithr [kowit] or
[kowit]. No dscription of Nahuatl (to our knowldg) has pinnd down
any nvironmnt for this
variation.
Illustrativ Sctions
In th following sctions, w show how xamination of th distribution
of sounds (as opposd to
xamination of allomorphic altrnation) rvals that phonological procss
s such as palatalization,
nasalization, tc. may b discovrd in a languag.
d

S, for xampl, Harold and Mary Ky (1953) Th Phonms of Sirra Nahuat, Int
rnational Journal of Amrican
Linguistics 19:53
6.
123
Voicing Assimilation
S th discussion abov rgarding th voicing of s in Spanish.
23.1 Try it for yourslf with North Publa Nahuatl
Considr th phons [j] and []. Is thr vidnc for contrast? If not, stat th
distributional facts and mak
a hypothsis about ths sounds by proposing a phonological rul.
128
A coupl of words that do not contain
ithr of ths sounds ar includd to hlp you mak your rul mor accurat.
(144) 1. [ojon] now 4. [mitsmoko] h givs to you
2. [iwok]
dry 5. [ijojo]
tar
3. [ncjtik]
blu 6. [onikmot]
I knw it

23.2 Try it for yourslf with Walmatjari

Examin th Walmatjari data in Appndix F. In chaptr 9 you proposd
a rul of voicing assimilation
(voicing consonants aftr nasals). This might lad you to qustion wh
thr both voicd and voiclss
consonants ar ndd undrlyingly. Ar thy in fact in complmntary distributi
on? If so, what othr rul
of voicing assimilation is rquird?
129
Plac Assimilation
In North Publa Nahuatl, a nasal is always vlar prcding vlar stops: [jopkik
] nw. This is a common
typ of assimilation. If a nasal is always vlar in this position, and hnc th
r is no contrast btwn [p]
and [n], and also no contrast btwn [p] and [m], th qustion thn ariss as t
o what [p] is phonmically.
Is it phonmically /m/, or /n/? In som thoris, this was a srious problm, a
nd a dcision had to b mad.
Usually th most similar sound (in this cas /n/) was chosn as th phonm, or
th most common. Thus a
word such as [jopkik] would b /jonkik/ phonmically. But undr othr assumpti
ons, mad xplicit in
chaptr 25, [p] is simply a nasal consonant which is nithr /n/ or /m/, but do
s not contrast with ithr.

English also has plac assimilation of nasals. Thus th nasal in word
s such as pump, finch, pant, and
think ar homorganic with th following consonant. Considr how this fact may b
rlvant for th analysis
of [p]. This sound occurs in a word lik think, of cours. And on also finds pa
irs of words such as [Oin]
thin, and [Oip] thing. Words such as ths, which ar minimal pairs,
hav bn adducd as vidnc of a
phonmic contrast btwn [n] and [p]. But this dirct conclusion has bn chall
ngd. Th minimal pairs
show that somthing is diffrnt, but not xactly what. First of all,
on should not that [p] nvr occurs
syllabl
initially in English. Thr ar no words lik *[pip]. And yt
syllabl
initial position is th plac
whr on xpcts to find most phonmic contrasts. Now considr th following da
ta:
(145) Nasal plus voiclss consonant:
Imp lamp Iintj finch pnt pant sipk sink
Nasal plus voicd consonant:
(non) bind bing
snd sand (non)
Thr ar no words or syllabls that nd in a nasal
consonant clustr whr th
consonant is a voicd labial
or vlar. This fact has bn usd to argu that th phonm [p] dos not xist i
n English, and that a word
lik [sip] sing is corrctly analyzd as bing {sing} phonologically.
Th nasal assimilats (as it must
anyway), and th g dlts by a phonological rul.
f


In som cass, thr may b altrnation vidnc that will provid vidnc in
favor of a particular analysis, but thr
ar also cass whr thr is no altrnation vidnc.
f

S th chaptr on dltion latr in th book. Th contxt for dl
tion of th g in English has to rfr to th word
124
Considr th sounds [] and [w] in th following data from Trio (Sipal
iwini Rivr dialct). Ths
sounds show som fluctuation, with [] or [w] occurring bfor front vo
wls, but only [w] (a back
consonant) occurring bfor cntral and back vowls. Thrfor, [w] and
[] ar allophons of on
phonm. (This variation in th pronunciation of /w/ is actually quit common cr
oss
linguistically.)
(146) woc I am ci ~ wci sun
wmi cassava cc ~ wcwc wood
iwonmoc ~ wiwonmoc
I tach him
icmc ~ wiwcmc
slippry
Palatalization and Labialization
In Brazilian Portugus, a t bcoms [tj] bfor i, and a d bcoms [d] in th sa
m nvironmnt.
(147) gotu cat mhtji
dath
gotjiju small cat ondi
whr
tjipu typ diIcic tji
diffrnt
dcdiju fingr (dim.) dcdu
fingr
Thrfor, a word such as [tjipu] is phonmically transcribd /tipu/.
It is oftn claimd that English has a phonm //. This is hard at th nd of th
 word roug and lug.
But curiously nough, th phonm dos not occur word
initially (although popl
may point to th nam
Zsa Zsa Gabor). Stting asid such words (roug is a loanword from Frnch, as ar
 many words in English,
of cours), lt us look at word
mdial xampls of [] and dcid whthr thr is
rally good vidnc that
this is a phonm of English.
Th sound [] primarily occurs intrvocalically in English: occasion, pl
asur, closur, trasur,
sizur, division, incision. Simply th fact that thr ar no good xampls of
contrast btwn [] and [z]
should indicat to us that somthing is going on. W thn may notic
that svral of ths words hav
rlatd words with [z] in thm:
plas, clos, siz, divisibility,
incisor. And that othr words in English
hav ndings which bgin with [j], a prim candidat for inducing pal
atalization:
rbllion, tnur. With
ths facts in mind, w could rasonably propos that [] should b an
alyzd as /zj/. Th word incision
[insin] would b analyzd /insizjn/.
g
23.3 Try it for yourslf with Chimalapa Zoqu
Focus on th affricats [t s] and [t j], and th fricativs [s] and [j|. Prsnt
th vidnc to stablish which
ar ndd in undrlying forms. Giv th rul ndd (informally) to
driv th othrs.
130
B sur to look
byond th immdiatly continguous sounds.

boundary, and this is complicatd by words such as singr. Not th

diffrnc btwn fingr (with g pronouncd)
and singr (without g). Thr hav bn thoris of phonology whr such contras

ts ar takn as dirct vidnc that

[pg] and [p] contrast at som dpr lvl, dspit th facts shown abov and d
spit th rlationship btwn words
lik long (without g) and longr (with g). It has also bn pointd
out that [p] occurs morphm
mdially, prhaps
uniquly, in th word dinghy.
g
This analysis dos not carry ovr to th fw words lik roug and lug. Ths r
main xcptional.
125
1. mctso1op two 7. jis
mat
2. mctjhi1
th two of thm 8. 1ujpi
alligator
3. tItspo
it is drying 9. 1uji1ip
a littl bit
4. tItji
dry 10. 1ijnitum straight
5. tI1tjkin castor plant 11. tjin
pin tr
6. kospo
h is scolding him
Nasalization
In um Mixtc, as wll as som othr Mixtc languags, a t has th pr
onunciation [t
n
] (a t with a nasal
rlas) whn it prcds a nasalizd vowl.
(148) [t
n
u ?u ] word
Sinc [t
n
] occurs only in this nvironmnt, it dos not appar in lxical rprsntations
of Mixtc words; it
is an allophon of t. Th lxical rprsntation of th word word contains a
simpl t.
h
Similarly, in Busa
thr is a nasal transition btwn crtain consonants and nasalizd vowls, as
in /do bo/ [d
n
o
m
bo] lif; and
slight prnasalization prcds glids that bgin syllabls with nasaliz
d vowls, as in /ji / [
p
ji ] cold, wt,
and /wc / [
p
wc ] yar.
In Sri, /m/ bcoms a nasalizd w aftr /k/, and th nasalization t
hn sprads phontically to th
vowls which follow. A word such as /kmooi/ dont mak it! is phontically [kw o o
i ].
Yoruba has oral vowls and nasalizd vowls. Nasalization sprads from nasalizd
vowls to sonorant
consonants which ar in th sam syllabl.
(149) JI [J I ] to walk
ju [j u ] to dispns
hu [h u ] to wav
wi [w i ] to lnd
23.4 Try it for yourslf with Busa
Considr th sonorant consonants in th following data. Considr what
phonms ar ndd. (Ton has

bn omittd.) Th symbol [r] rprsnts a flap, but it usually may b rplacd
by a latral flap. Similarly,
th [n] in intrvocalic position may b rplacd by a nasal flap.
131
h

Som languags hav bn dscribd as having morphms or words which ar chara
ctrizd as bing ithr nasal or
oral. Th nasal fatur affcts a prdictabl rang of consonants and vowls. Fo
r xampl, in Mixtc languags, it has
bn claimd that th fatur [nasal] attachs to th right sid of a word, link
ing up with as many sounds lftward as
ar compatibl with it. Th compatibl sounds ar sonorants: vowls, smivowls
(lik j and w) and nasal consonants.
Th sounds which ar not compatibl with th fatur [nasal] ar obst
runts: stops and fricativs. Whn a vowl is
linkd with [nasal], it is pronouncd nasalizd; othrwis it is oral.
Whn j is linkd with [nasal], it typically is [j ]
(which sounds almost lik [j]); othrwis it typically sounds lik []. Whn w is
linkd with [nasal], it is pronouncd
[m]; whn not linkd with [nasal], it usually varis btwn [w] and []. Yt anot
hr consonant, usually thought of as
n, is pronouncd [n] whn it is linkd with [nasal] and pronouncd a
s [n
d
] whn it is not. Th fatur [nasal] is
distinctiv (phonmic) in this languag family; it works diffrntly from othr
languags in that it is not a proprty of
individual sounds. For mor discussion, s Stphn A. Marltt (1992)
Nasalization in Mixtc Languags,
Intrnational Journal of Amrican Linguistics 58:425
35.
126
(150) Icrc aris ni b5 gust
Iou rain oru
blood
mo fll oro strtch
n5 wif Icrc fall
Ii tr nu nu not having born
Ic mouth gbc dog
bIc at on5 thy
pIc choos
mo no good
mIc snak mc snak
Mannr Assimilation
In Spanish, th voicd stops b, d, and g tak on th fatur [continuant] of a p
rcding vowl or flap if not
immdiatly followd by a tru consonant (not a glid); thy thrfor
bcom fricativs. Thrfor th
stops [b], [d], and [g] ar in complmntary distribution with [], [], and [y].
i
(151) pud [pwcc] is abl anda [ondo] walks
ard [oic] burns
rd [ic] nt habla [obIo] spaks
cab [koc] fits samba [sombo] samba
lago [Ioyo] lak lngua [Icpgwo] tongu
agua [oywo| watr digno [digno] worthy
(152)

|
[+continuant] [+voic]
In Doyayo, howvr, th

V
|

fatur [continuant] is takn on by th voic

d stops b and g (but not d,

curiously nough, which dos occur bfor z) whn thy occur bfor a consonant.
Prsumably th fatur
[continuant] still coms from th vowl, but th contxt of bfor a consonant is
crucial to th procss in
this languag.
j
i

Spirantization (i.., fricativization) following /I/ happns with /b/ but not w
ith /d/, howvr. If /I/ is a [continuant],
this diffrntial bhavior is unxpctd, but if /I/ is [+continuant], it is als
o not xpctd. Compar /koIbo/ [koIo]
bald and /koIdo/ [koIdo] broth.
j
Spirantization of /b/ dos not happn prcding /d/, anothr curious
fact. Ths anomalis mak it difficult to
formaliz th rul.
127
(153) Stop allophons
bo: [bo:] danc a dirg
gob [gob] b man
pobi [pobi] uncl
dog [dog] gt, mt, tast
jogi [jogi] rat with bushy tail
Fricativ allophons
hobz [hoz] prish quickly
kibI [kiI] roll on th ground
gubr [gur] pour out
dogd [doyd] alrady got
Icgrj5 [Icyrj5] spidr wb
zogI [zoyI] knl, crawl
r5gz [r5yz] just cluckd, smoothd
23.5 Try it for yourslf with Oaxaca Chontal
Considr th labial sounds [p], [b], [I], and []; th coronal sounds [t], [d], an
d []; and th vlar sounds
[k], [g], and [y|. Which ar phonms? Giv th vidnc. Giv th ruls for th
othr sounds.
132
1. mojyI? tomorrow 9. toyuj
I grow
2. Iiio his grandmothr 10. Iupguj
sh grows fat
3. wouj h carris 12. cntco
livr
4. konduj h lavs 13. toyo
thick
5. poyuj sh washs 14. go1
hron
6. opondo1 lam 15. Iooj
thy saw
7. popuj h livs 16. bomo1
tn
8. owo:to girl
Vowl changs
Vowls may assimilat to th plac of articulation of adjacnt consona
nts. For xampl, in Corongo
Quchua, which has only th vowls /i/, /u/, and /o/ in nativ words
, th high vowls bcom mid whn
thy follow a word
initial uvular consonant (which has th fatur [
high] sinc
th tongu is not raisd), or
whn thy prcd a uvular consonant.
(154) ycru wood purup
fallow
ycjwo nst purip
walks
yoto lak
uro blow
yoskip h givs it all umo had
otoy fox

Th high vowls ar also optionally lowrd whn thy ar othrwis s
imply adjacnt to a uvular
consonant.
(155) isyup ~ isyop nin
jonoyi ~ jonoyc companion
royup ~ royop
it cuts
128
23.6 Try it for yourslf with Tlahuitoltpc Mix
Considr th vowls [o] and [o] (its front countrpart), ignoring th
lngth distinctions. Thy do not
contrast. with ach othr. Giv th rul (informally).
133
1. mo:pj giv 8. 1oh cough
2. koj
rabbit 9. koh
wav
3. pokj
sin 10. mo:hkh corn
4. kohj
wov 11. ho:n
bird
5. mo:nj
mony 12. po:ph
whit
6. to:kj
sold 13. jo:ts
cloud
7. xo:kh
dampn 14. to:kh
sll
Miscllanous
In Sri consonants ar lngthnd considrably if thy follow a strssd vowl a
nd prcd a vowl. Thus a
word such as / ?osotox/ [ ?os:o:tox] stons has a rathr long s. Similarly, th
vowl in th syllabl with
this lngthnd consonant is also lngthnd if it is followd by a consonant. S
o th scond a in th word
/ 1osotox/ is also rathr long. This prdictabl consonant lngth and
vowl lngth is not analyzd as
contrastiv in Sri, dspit th xistnc of contrastiv vowl lngth in th st
rssd syllabl.
23.7 Try it for yourslf with Daga
Considr [t], [s], and [r] in th following data. Two of ths sound
s contrast with ach othr and two do
not. Discuss th facts, prsnt th vidnc, and giv th rul (informallyit wont
b a rul with obvious
motivation). Writ th word whispr phonmically.
134
1. osi
grunt 10. onct
w should go
2. scnoo
shout 11. wogot
holiday
3. urosc
hol 12. otu
littl
4. sinoo
drum 13. topcn
hit
5. simuro whispr 14. toc
old
6. usc
thr 15. tuion
I kill
7. siuron salt 16. worio
h taks
8. worop
I tak 17. wotop
I opn
9. ncsip
stps 18. mcrip
vomit
129
23.8 Try it for yourslf with Ashninca Campa
Considr th vowl pairs [i, i], and [c, c]. Th fatur distinguishi
ng clos from nar
clos vowls, and
clos
mid from opn
mid vowls, is not distinctiv. Giv th rul informally.
135
1. ctini
armadillo 7. ijohi
his bill
2. ipgoni
rain 8. jicndi
dragonfly
3. nojco
I at 9. oorintsi
roastd mat
4. pitiro
cockroach 10. pokitc
cook it!
5. topctso
vin 11. nombokotcmbi
I ll show you
6. nihondo
far away 12. iiri
his nos
Ky Concpts
complmntary distribution (mutually xclusiv) fr variation (conditiond and u

nconditiond)
23.9 Try it for yourslf with Sirra Nahuat
Th sounds [w] and [w ] (voicd and voiclss countrparts) do not contrast in t
his languag, nor th sounds
[j] and [j ] contrast. What is th distribution of th voicd vs. voiclss soun
ds. Choos an undrlying form
and giv a phonological rul.
136
1. kitji:wok h did it 9. kjw potj moss
2. kjw ji:pkch sawyr 10. woIo:s
h will com
3. jo:s
h will go 11. jo:Ii:k
slowly
4. jiwit
laf 12. kc:jo
frog
5. kow kckc:j woodpckr 13. mokcj tio
on visits
6. pcpctsi:w tok fastnd 14. tchwo n
w
7. kowmch
trs 15. nomo:j
my hand
8. nc:toto:to:w ti:Ii:s bgging
23.10 Try it for yourslf with Italian
Th sounds [n] and [p] do not contrast with ach othr. Which has th most rstr
ictd distribution? Giv th
rul (statd as gnrally as possibl) to driv th most rstrictd on from th
 othr.
137
1. ncro black 5. Iupgo mud
2. bjupko whit 6. Iipgwu tongu, languag
3. upgju (fingr)nail 7. Iinc nd
4. ondu wav 8. duntsu danc
130
23.11 Try it for yourslf with Damana
Th sounds [p], [n], and [j] do not contrast with ach othr. Giv th rul or r
uls ndd to driv thm
from on undrlying nasal /n/.
138
1. ji what? 5. opgo it is
2. jikumo gg 6. nojcpgo I will go
3. jipgo nothing 7. kon wood
4. nopko s/h is 8. supkIno nos
23.12 Try it for yourslf with Ttlcingo Nahuatl
Th phons [w], [I], [w ], and [] ar in complmntary distribution. (
Chck it out.) Assum that th
phonm is basically /w/, and account for th allophons by two pros ruls, on
for th fronting (making
th w into a fricativ, ignoring th fact that it is bilabial or labiodntal) an
d on for th dvoicing.
139
1. iIIitI fast 6. wotsinko tomorrow
2. iwo and 7. iccitsi biggish
3. cjok long 8. tcusii hungry
4. chIcji big 9. tchw w
5. ijoIw o ystrday 10. mitsIiko h taks you
23.13 Try it for yourslf with Tairora
Th sounds [b] and [] do not contrast with ach othr; nor do th sounds [r] and
[I]. Giv th ruls ndd
(informally).
140
1. burouko I wnt 6. buIo I am going!
2. biIo
h gos 7. bio h wnt
3. binou
w wnt 8. bircro I will go
4. irio
listn! 5. huro uorcro I might go tomorrow
131
23.14 Try it for yourslf with Cashinahua

Considr th following sts of sounds:

[p, b, |; [t, ts, tj, d, r|; [t
s, s, j, h|. Prsnt th vidnc of
contrast. Giv th ruls ndd to driv any allophons.
141
1. poko bamboo 17. m:su swolln hand
2. toko livr 18. m:ju dark, black
3. koko baskt (typ) 19. b:tu spottd fac
4. boko fish 20. b:tsu to b squzd
5. duku to rst 21. b:su fac
6. tsoko to kill 22. kutjo arrow (typ)
7. tjoko bad 23. kujo to hit
8. topo floor 24. noko to chw
9. tomo panut 25. too washboard
10. b:ru y 26. iso bird
11. isi unnding 27. is: had painting
12. isu spidr monky 28. bori sun
13. bo frind 29. nou popl
14. dor: plant (spcis) 30. s:pi waving dsign
15. jono fruit (spcis) 31. dosi all
16. h:p: palm (spcis) 32. moko rat
Fdback for Chaptr 23
23.1 North Publa Nahuatl
128
Thr is no vidnc of contrast. Th sounds ar in complmntary distribution.
Th voiclss fricativ
occurs intrvocalically and also bfor a voiclss consonant. Th voicd fricat
iv only occurs bfor
voicd consonants.
Th rul might b statd as gnrally as: Voicing sprads lftward to a fricati
v. Or it might b: Voicing
sprads lftward from a glid. W s that oral stops and affricats do not bco
m voicd bfor nasal
consonants; thrfor th rul cannot b gnralizd furthr (i.., th rul can
not b: Voicing sprads
lftward from a consonant).
23.2 Walmatjari
129
Voicd and voiclss consonants (th voiclss ons ar all stops) ar in compl
mntary distribution.
Anothr rul is ndd: consonants ar voicd whn thy occur btwn vowls.
23.3 Chimalapa Zoqu
130
[ts] : [s] CNE
[tItspo] it is drying vs. [kospo] h is scolding him (not th
 bst vidnc)
[tj] : [j] CNE
[tjin] pin tr vs. [jis] mat
[ts] dos not contrast with [tj], and [s] dos not contrast with [j|.
132

Rul: Coronal fricativs and affricats bcom alvopalatal whn followd in th

 sam syllabl or th nxt
syllabl by /i/.
23.4 Busa
131
Thr ar two contrastiv sonorant consonants in ths data. On is labial and o
n is coronal. Th labial
on, m, happns to only occur bfor nasalizd vowls. Th coronal sonorant has
thr forms: [I], [r], and

[n]. Of ths, only [I] occurs aftr a consonant. Othrwis, th phons [I] and
[r] occur only bfor oral
vowls, and [n] only bfor nasalizd vowls. Othrwis, th phon [I] occurs on
ly in word
initial onsts,
and [r] in word
mdial onsts. (Ths facts arnt vry complicatd, but thy ar
not trivial ithr.) In th
sourc, /n/ was chosn to rprsnt th phonm. Howvr, [I] has a widr distri
bution. That fact, plus th
fact that nasality is a contrastiv fatur on th vowl, suggsts that prhaps
[I] is mor basic.
23.5 Oaxaca Chontal
132
Th phonms ar: p, b, I; t, d; k, g.
[p] : [b]
CNE [popuj] h livs vs. [bomo1] tn
[p] : []
CNE [opondo1] lam vs. [Iiio] his grandmothr
[p] : [I]
CNE [popuj] h livs vs. [Iooj] thy saw
[b] : []
no good vidnc
[t] : [d]
CNE [cntco] livr vs. [konduj] h lavs
[t] : []
CNE [owo:to] girl vs. [Iooj] thy saw
[k] : [g]
CNE [konduj] h lavs vs. [go1] hron
[k] : [y] no good vidnc
Th fatur [+continuant] sprads rightward to voicd obstrunts. (Not that na
sal stops ar [
continuant]
and that fricativs, vowls and glids ar [+continuant].)
23.6 Tlahuitoltpc Mix
133
/o/ is frontd to [o] whn it is followd somwhr in th sam word (or syllabl
, actually) by th front
glid /j/ (or palatalization).
23.7 Daga
134
[t] and [s] contrast with [r|, as in [wotop] I talk and [ncsip] stps vs. [worop
] I opn and [mcrip] vomit;
but [t] dos not contrast with [s]. [t] nvr occurs prcding a front vowl, bu
t [s] only occurs prcding
front vowls. /t/ bcoms [s] bfor front vowls. Th word whispr is phonmica
lly /timuro/.
23.8 Campa
135
Clos and clos
mid vowls bcom nar
clos and opn
mid, rspctivly, whn th
y prcd nasals.
23.9 Sirra Nahuat
136
Th voiclss glids only occur bfor voiclss consonants. Th voicd glids o
ccur btwn vowls, aftr
h, bfor voicd consonants, word
finally, word
initially, tc.
Th rul is: Voicing sprads lftward to a glid.
23.10 Italian
137
Th vlar nasal has th most rstrictd distribution.
133

Th rul: A nasal assimilats to th plac of articulation of an immdiatly fo

llowing consonant.
23.11 Damana
138
A coronal nasal bcoms palatal bfor /i/. (This rul accounts for th occurrn
cs of th sound [j].

A nasal assimilats to th plac of articulation of an immdiatly following con

sonant. (This rul accounts
for th occurrncs of th sound [p].)
23.12 Ttlcingo Nahuatl
139
/w/ bcoms a bilabial fricativ bfor front vowls.
A continuant consonant assimilats in voicing to an immdiatly prcding conson
ant. (Not that all of th
forms of /w/ assimilat in voicing, but /k/ dos not assimilat in voicing to a
prcding nasal.)
23.13 Tairora
140
/b/ is a fricativ whn it occurs intrvocalically.
/r/ is [I] whn it prcds a round vowl.
23.14 Cashinahua
141
[p] : [b] CIE
[puku] bamboo vs. [buku] fish
[b] : [] No contrast
[p] : [] CIE
[tupu] floor vs. [tuu] washboard
[t] : [d] CIE
[tuku] livr vs. [duku] to rst
[d] : [r] No contrast
[t] : [r] CIE
[b:tu] spottd fac vs. [b:iu] y
[t] : [ts] CIE
[b:tu] spottd fac vs. [b:tsu] to b squzd
[t] : [tj] CIE
[tuku] livr vs. [tjuku] bad
[ts] : [tj] CIE [tsuku] to kill vs. [tjuku] bad
[ts] : [s] CIE [b:tsu] to b squzd vs. [b:su] fac
[ts] : [j] CNE [tsuku] to kill vs. [ jono] fruit (spcis)
[ts] : [h] CNE [tsuku] to kill vs. [ h:p:] palm (spcis)
[s] : [j] CIE
[m:su] swolln hand vs. [m:ju] dark, black
[h] : [s] CNE [h:p:] palm (spcis) vs. [s:pi] waving dsign
[h] : [j] CNE [h:p:] palm (spcis) vs. [junu] fruit (spcis)
Th fatur [+continuant] sprads rightward to voicd obstrunts. (In this langu
ag, th [+continuant]
vrsion of /d/ is [r].
134
Chaptr 24
Suspicious Pairs
Not vry pair of sounds in a languag must b invstigatd in th
mannr dscribd in th prcding
chaptrs. Crtainly in no languag is th phonm /b/ somtims pronouncd [s].
Thrfor w would not
vn bothr worrying about whthr th sounds [b] and [s] in a langu
ag ar phonmically distinct. W
simply assum that thy ar. Th rason for this is that th nvironmnt that a
phonm occurs in can affct
it only slightly (although probably mor than you might hav gussd).
Thrfor th analyst rstricts
attntion to phontically similar sounds, which ar calld th suspicious pairs.
It dos not tak too much xprinc to b abl to guss what sound
s must b invstigatd. Ths
sounds usually only diffr in on, two, or thr faturs. Th dgr to which s
ounds might b considrd
suspicious pairs can b shown by comparing th following sounds. Blow
ach sound w list crtain
faturs, th faturs that distinguish on or mor of th sounds from th othr
sounds. For xampl, sinc
th sounds listd hr all hav a common plac of articulation (thy ar bilabia
ls), w do not list any plac
faturs.
(156) p
b

w
m
sonorant

+
+

continuant

+
+




voic

+
+
+
+
nasal


+




W now tak ach pair of sounds and not how many faturs sparat thm.
(157)
On fatur
Two faturs
Thr or mor
p, b
p,
p, w
b,
b, m
p, m
, w
b, w
, m
w, m
Th pairs in th first column ar thos which w want to chck
first; w might call thm th primary
suspicious pairs. Th sounds in th scond column ar also worth
king out, although it is a littl lss
likly that thy ar allophons. W
might call thm th scondary
suspicious pairs. Th sounds in th
third column ar not out of th
qustion as suspicious pairs. In any
cas, if th analyst has difficulty in
finding contrast btwn two sounds
which shar som faturs, s/h should
b awar that thy may b allophons
of th sam phonm.
a
Ky Concpt
suspicious pairs
Ruls of thumb: Having diffrnt articulators is a vry significant
fact for obstrunts. Thrfor, although [p] and [t] ar dis

tinguishd only by thir point of articulation, this diffrnc
virtually guarants that thy ar not allophons of th sam
phonm. Having diffrnt articulators is not a vry significant fact
for nasals. A nasal oftn taks on th plac of articulation of
adjacnt consonant, spcially on that follows it.
Laryngal consonants (such as [h] and glottal stop) also mrit
spcial considration, and comparison with consonants such as [s]
and [t], rspctivly, is appropriat.
135
24.1 Try it for yourslf with Suspicious Pairs
For th groups of sounds listd blow, dcid which ar primary
cious pairs using th faturs
providd.
142
Group A:
s
z
t
b
m___
continuant
voic
Coronal
Labial
nasal
Group B:
s
I
v
m___
voic
Labial
Coronal
continuant
nasal
Fdback for Chaptr 24
24.1 Suspicious Pairs
142

out
chc

an

suspi

Group A: s,z
Group B: I,v

s,t
v,b

t,d
b,m

z,d

b,m

This mthod of dtrmining suspicious pairs dpnds, of cours, on w

hich faturs you choos to utiliz.
Nvrthlss, w fl it is mor important to know why two sounds should b comp
ard than to rly mchanically on a
chart of suspicious pairs.
136
Summary and Rviw Qustions for Sction 3
Just as an xamination of allomorphy rvals that procsss of assimilation ar
oprativ in a languag, so
an xamination of th distribution of sounds rvals th sam.
Th analyst provids vidnc that two sounds ar not rlatd by a rul of assim
ilation by showing that
th sounds contrast in non
influncing nvironmnts. If two phonological
ly similar sounds (suspicious
pairs) cannot b found to contrast in this way, w assum that thy
ar allophons and look for
nvironmnts which may account for thir distribution.
If you hav mastrd th matrial in this sction, you should b abl to xamin
data and
(a) rcogniz th opration of phonological ruls basd on distributional fact
s
(b) provid vidnc that crtain sounds contrast with ach othr
(c) provid vidnc that crtain sounds do not contrast with ach othr
(d) propos phonological ruls to dscrib allophonic variation
() rcogniz sounds which ar likly to b analyzd as allophons of a singl
phonm
(f) xplain th rol that contrastiv faturs hav in th sound systm of a l
anguag
(g) xplain th rol that non
contrastiv faturs hav in th sound systm of
a languag
For Furthr Rading:
Andrson, Stphn R. 1974. Th organization of phonology. Nw York, London, Acad
mic Prss.
Hyman, Larry M. 1975. Phonology: thory and analysis. Nw York, London, Sydny:
Holt, Rinhart and
Winston.
Knstowicz, Michal and Charls Kissbrth. 1979. Gnrativ phonology:
dscription and thory. Nw
York, London, Acadmic Prss.
Pik, Knnth L. 1947. Phonmics: A tchniqu for rducing languag to writing.
Ann Arbor, Univrsity of
Michigan Prss.
Rviw Qustions
Th following qustions ar to hlp you rviw th matrial in th prcding sc
tion.
143
1. (T or F) Allophons somtims appar in th undrlying form of a morphm.
2. (T or F) On cannot assum that two similar sounds contrast in a languag un
til on finds thm in
a minimal pair.
3. (T or F)
A minimal pair such as [noko] dog and [nogo] tabl pr
ovs that /k/ and /g/ ar
phonms.
4. (T or F) Whn thr is fr variation btwn two sgmnts, an optional rul

can b writtn.
5. (T or F) Contrast in non
influncing nvironmnt is hlpful for dmonstratin
g that two sounds ar
distinct phonms.
6. (T or F) Whn two similar sounds ar in complmntary distribution, on xp
cts to writ a rul
to account for th facts.
7. (T or F) Phontic diffrncs that mak a diffrnc in maning ar account
d for by phonological
ruls.
8. (T or F)
Th dmonstration of contrast btwn two sounds stabli
shs thm as sparat
allophons of a singl phonm.
137
9. (T or F)
In th cas of complmntary distribution, two or mor
sounds occur in mutually
xclusiv nvironmnts.
10. (T or F) If you cannot prov that th mmbrs of a suspicious pair contrast
, you should look for
complmntary distribution.
11. Two sounds ar _________________ of on ___________________ if thy
occur in
complmntary distribution.
12. If two phontically similar sounds occur in mutually xclusiv nvi
ronmnts, thy ar
_____________________.
13. A(n) _____________________ has th potntial to crat a diffrnc in mani
ng.
14. [p] and [p] ar __________________ in English.
Fdback for Rviw Qustions
143
1. F 2. F
3. F (This is vidnc that thr is somthing diffrnt btwn th two words
undrlyingly, but it dosn t
ncssarily man that th diffrnc is locatd in a voicing contrast btwn k
and g.)
4. T 5. T 6. T
7. F 8. F 9. T 10. T
11. allophons, phonm 12. allophons 13. phonm
14. allophons
138
Sction 4
Phonological Ruls:
Structural Issus
In this sction w look at a varity of issus, many of which rlat to faturs
in a largr structur, such as a
syllabl, a word, or an uttranc.
Sounds ar organizd into groups whn words ar formd. Th syllabl pattrns of
a languag tnd to
b prsrvd whn affixs ar addd to words. Thus thr ar phonolog
ical ruls that insrt or dlt
matrial in ordr to kp ths pattrns intact. Ths phonological ruls mop up a
ftr morphological ruls
hav don thir thing.
In this sction w also look a bit mor closly at how phonological ruls intra
ct. What happns whn
two or mor ruls might apply to th sam undrlying form? Do thy
apply in ordr or all at onc?
Qustions such as ths constitut an important part of linguistic rsarch ovr

th past thirty yars.

W also look at som issus that rlat to th natur of undrlying forms.
139
Chaptr 25
Undrspcification
Up to this point w hav prsntd undrlying forms of morphms as strings of p
honms. For xampl,
w hav givn th undrlying form of th most common plural suffix i
n English as {z}. W hav also
clarifid that symbols such as z ar abbrviations for combinations of
faturs. Thrfor, w might mor
corrctly giv th undrlying form of this morphm as th combination
of th faturs [
sonorant],
[+voic], [Coronal], [+continuant].
But this viw still may b unncssarily complicatd. It has bn suggstd that
in som cass crtain
faturs may b absnt in undrlying forms. That is, crtain charactr
istics of th morphm may b
undrspcifid. As an xampl, considr th following data from Sri.
a
(158) [m]
potis if you point at it
[n]
ttis did you point at it?
[j]
jotis you pointd at it
[]
yotis you pointd at it!
Th subjct prfix for scond prson singular is a nasal, it always happns to o
ccur in front of a consonant,
and it always has th sam point of articulation as that consonant. So thr is
no obvious way to choos on
of ths forms as undrlying. It would b arbitary (from ths data alon) to ch
oos m as basic, arbitary to
choos n as basic, and so forth.
An altrnativ to an arbitrary choic is to claim that th morphm is simply a
nasal consonant which is
not spcifid for plac of articulation. Th nasal will thn assimilat to th f
ollowing consonant.
In Salasaca Quichua, syllabl
final consonants ar xtrmly rstrictd.
If it is a nasal, its plac of
articulation is ithr (a) that of th following obstrunt, or (b) vlar.
(160) mundupgu had (nasal prcds vlar)
tsopkono to grind by hand (nasal prcds vlar)
jonto wood (nasal prcds alvolar)
timbuno to boil (nasal prcds labial)
guIup thundr (nasal is uttranc
final)
If th syllabl
final consonant is a non
stridnt obstrunt (i.. not
s, j, z, tj, tc.), its point of
articulation is always vlar, th xact pronunciation of which varis by contxt
: [x] word finally, [y] bfor
voicd fricativs, [k] bfor voiclss stops, tc.
(161) kuIux storag bnch (word
final position)
tuyo slipknot (prcding voicd fricativ)
waktsux bird (spcis) (prcding voiclss stop)
Thrfor, it appars that Plac faturs at th nd of syllabls in
this languag ar not prsnt if th
consonant is a nasal or a nonstridnt obstrunt. If th nasal dos n
ot obtain Plac faturs from th
following consonant, it is assignd th faturs of a vlar. And a n
onstridnt obstrunt in syllabl
final
position not only has no Plac faturs, it also has no Mannr faturs. It is a
ssignd th Plac faturs of a
vlar. It rcivs th faturs [voic] and [continuant] from th cont
xt, whn possibl, and is othrwis

assignd th faturs [
voic] and [+continuant]. Th undrlying form o

f th word for storag bnch is
a

Ths forms ar thos which follow vowl
final words. If a consonant
final word
prcdd thm, thy would bgin
with an i.
(159)
C
C
[+nasal] Plac
140
thrfor somthing lik {kuIuC}; th pronunciation of th final conson
ant is drivd from gnral
phonological pattrns of th languag.
It is possibl for othr faturs to b absnt in undrlying forms. Whnvr th
r is a lack of contrast
btwn sounds, on should considr th ida that a fatur or st o
f faturs is not spcifid rathr than
mak an arbitrary choic.
Thr is anothr way in which undrspcification is usd in phonology. Suppos,
for xampl, that you
had a common fiv vowl systm, as shown blow.
(162) o c i o u
high

+
+
low +




back +

+ +
round


+ +
Not that if a vowl is [+high], it is [
low]. And if a vowl is [+low], it is [

high]. In addition, th valus of
th fatur [round] ar prdictabl from othr faturs. If a vowl i
s [+back] and [
low], it is [+round] in
this languag. In fact, if as much rdundant information as possibl is omittd
from th spcifications of th
vowls, w can nd up with th rprsntations shown blow. And ach
vowl is still rprsntationally
distinct.
(163) o c i o u
high + +
low +
back + + +
round
Unlss othrwis spcifid, a vowl is [
high]. Unlss othrwis spcif
id, a vowl is [
low]. Unlss
othrwis spcifid, a vowl is [
back]. Th ruls might b xprssd as:
b
(164) [
] [
high] [
] [
low] [
] [
back]
[+back] [+round] [
] [
round]
[
low ]
This approach allows a considrabl simplification of undrlying forms in a lang
uag. For xampl, a
morphm such as pcb might hav th following shap (simplifying th formalism s
omwhat):
c
W know that th first consonant is a p (or I)
d
bcaus th fatur [voic] is not spcifid and th plac
fatur [+labial] is spcifid. On th othr hand, w know that th scond conso
nant is b (or v) bcaus th
faturs [+voic] and [+labial] ar both spcifid. Th vowl will surfac as /
/ bcaus of th ruls givn
abov.


It is possibl (but probably not common) that an undrspcifid consonant or vow
l contrasts with all
of th othr phonms of th languag. For xampl, it has bn clai
md that Sri has a numbr of roots
b

In ordr to mak ths rdundancy ruls giv th corrct rsults, crtain ons
must b applid bfor crtain othrs.
c
W ar assuming a rdundancy rul [
] [
voic] for obstrunts.
d
If th languag had I as wll as p, an xtra fatur would b ncssary for on
or th othr of ths sounds.

S Diana Archangli (1984) Undrspcification in Yawlmani phonology
and morphology, Ph.D. Dissrtation,
MIT; Diana Archangli (1988) Aspcts of undrspcification thory, Phono
logy 5; Douglas Pullyblank (1988)
Vocalic undrspcification in Yoruba, Linguistic Inquiry 19:233
270; and Poovi A
baglo and Diana Archangli (1989))
Languag
particular undrspcification: Gngb // and Yoruba /i/, Linguistic In
quiry 20:457
80.
(165)
C
V
C
[+voic]
[Labial] [Labial]
141
which bgin with a consonant which has no faturs of its own, but
which is distinct from all othr
consonants in th languag. Whn this consonant follows a consonant pr
fix, it assimilats ntirly to th
prcding consonant.
(166) t
toy is it hard?
k
koy that which is hard
s
soy it will b hard
im
moy it is hard
As discussd in chaptr 19, this is dscribd formally by a sprading of th Roo
t nod, which dominats all
of th faturs of a sound.
(167) C
C
|
Root
Whn it follows a vowl, it cannot assimilat and is not pronouncd.
f
(168) po
oy if it is hard
yo
oy it was hard
Th undrlying form of th root for b hard is thrfor th followi
ng (whr lttrs ar still usd as
abbrviations for fatur combinations):
(169) C V C
o y
W hav sn in this chaptr that if w allow undrlying forms to b undrspc
ifid for som faturs,
thn w ar abl to avoid making arbitrary choics in many cass. In this last 
xampl from Sri, w dont
hav much of an option at all.
Ky Concpts
undrspcification of faturs in undrlying forms
25.1 Try it for yourslf with Hausa

Propos undrlying forms for th roots and suffixs, and phonological ruls. B
sur not to propos xplicit
undrlying forms whr undrspcifid forms ar appropriat.
144
fathr mothr
1. 1ubopko 1uwokko your...
2. 1ubojji 1uwojji his...
3. 1ubonto 1uwotto hr...
4. 1ubommu 1uwommu our...
5. 1ubopku 1uwokku your (pl.) ...
6. 1ubonsu 1uwossu thir...
f

Stphn A. Marltt and Josph P. Stmbrgr (1983) Empty consonants in Sri, Li

nguistic Inquiry 14:617
39.
142
25.2 Try it for yourslf with Tarahumara
Propos an undrlying form for th nominalizing suffix in th following data, an
d a phonological rul.
145
1. mitjiru to mak shavings 6. rcmckc tortillas
2. mitjiruku shavings 7. potji to grow ars of corn
3. ritu to b icy 8. potjiki an ar of corn
4. rituku ic 9. opotjo to b drssd
5. rcmc to mak tortillas 10. opotjoko garmnt
25.3 Try it for yourslf with Walmatjari
Th following data apparntly show th full rang of allomorphy for th Irralis
suffix and th full rang of
facts for vrb stm
final consonants in this languag. Discuss th fac
ts and propos an analysis. (Th
diacritic
indicats a lamino
alvolar consonant. Th diacritic
indi
cats an apical post
alvolar
consonant.) For th purposs of this xrcis, you may omit considration of th
 allomorph which occurs
with th vrb chop. Also, rcall that voicing of th stop is dtrmind by th c
ontxt in which it occurs; it is
not distinctiv.
146
Stm Irralis
1. won d o won d o gt
2. kori kori stand
3 pod or pod oro chop
4. mon mon d o spak
5. kop kopgo carry
6. t on t ondo go
7. pod oron pod orondo continu to chop
8. kopon kopondo continu to carry
143
25.4 Try it for yourslf with Agutaynn
Propos an undrlying form and a rul or ruls to account for th allomorphy of
th scond prson pronoun
which follows th vrb.
147
1. mijip po you will blow
2. monIit to you will borrow
3. timiIik ko you will look
4. tumo1Iib bo you will pass by
5. tinIid do you will ntr
6. mobo1Iog go you will fall
7. moIitim mo you will grow hungry

8. moiIin no you will bcom full

9. mogIibip po you will attnd th funral
10. mongiIiI Io you will trmbl
11. mogpodir ro you will mak a fnc
12. mogbosis so you will us your voic
13. mogbaIaj jo you will mak a hous
14. mogpaoraw wo you will gt in th shad
15. mogsisi o you will scal (fish)
16. mogpoIipi o you will humbl yourslf
17. mogpotoko o you will inform
18. mogboIito o you will tll th nws
144
Fdback for Chaptr 25
25.1 Hausa
144
your {ko}, his {ji}, hr {to}, our {mu}, your (pl.) {ku}, thir {su}.
fathr {1uboN) (whr N indicats an undrspcifid nasal consonant)
mothr {1uwoC} (whr C indicats a fully undrspcifid consonant)
Rul: Th undrspcifid nasal consonant assimilats in Plac to an immdiatly
following consonant.
Rul: Th undrspcifid consonant assimilats in all faturs to an immdiatly
following consonant.
25.2 Tarahumara
145
Th suffix is {
kV} (whr V indicats a fully undrspcifid vowl). Th undr
spcifid vowl
assimilats in all faturs to th vowl of th prcding syllabl.
25.3 Walmatjari
146
First of all, thr ar probably two suppltiv allomorphs: null (aftr vowls
), and somthing ls aftr
consonants. Th undrlying form of th ovrt allomorph is indtrminat sinc it
always agrs in plac of
articulation with th prcding consonant. (It is dltd aftr a liquid.) Unls
s thr is othr vidnc of an
undrlying Plac fatur for this suffix, it would sm bst to simply posit an
undrlying form of {
Co}
but th C would hav th fatur [
sonorant] at last, but no Plac nod. A rul
sprads Plac to th right to
giv this consonant a Plac fatur in surfac rprsntation. Voicing would spr
ad to th right (by th rul
alrady sn).
25.4 Agutaynn
147
Th pronoun is {Co}. Th consonant has no faturs at all. Th Root nod of th
prcding consonant
sprads to it. If thr is no consonant immdiatly prcding it (such as whn t
h vrb nds in a vowl), no
faturs can sprad and th mpty consonant is not ralizd phontically at all. T
his xampl parallls th
Sri cas discussd in th txt xactly.
145
Chaptr 26
Edg Phnomna
Thr ar a numbr of phnomna that can b obsrvd happning at th dg of so
m unit of structur such
as th word, or syllabl, or uttranc. Bfor considring ths phno
mna, w nd to clarify what w
man by ths labls.
For som ruls of a languag, th word unit which is rlvant is th

 word producd by th word

structur ruls. For othrs, th word unit may includ a word produc
d by th word structur ruls plus
som smallr words which occur nxt to it. Th lattr has somtims bn r
frrd to as th phonological
word. W do not distinguish btwn ths hr, but us th symbols ]
w
(nd of word) and
w
[ (bginning of
word) to rfr to ithr.
a
For our purposs, an uttranc is any string of words which occurs b
twn two pauss (howvr
short). A simpl sntnc might b said so slowly that it is actuall
y composd of many uttrancs by this
dfinition. W us th symbols
]

(Grk uppr cas phi following squar brackt) to rfr to nd of

uttranc.
A syllabl is a small unit of structur which w discuss in mor dtail in th f
ollowing chaptr, as w
assum that you hav som ida of what a syllabl is. W us th convntio
n

[ (Greek lower cae igma

preceding quare bracket) to refer to the beginning of a yllable.
b
We now conider variou rule below which refer to the end or the beginning of o
ne of thee unit.
Final Poition
In variou language it ha been oberved that cononant tend to bec
ome voicele at the end of a unit,
epecially the yllable or utterance.
c
Although thi devoicing i common and natural, it i not due to
the
preading of ome feature from another ound.
It i common in Spanih, for example, for the r-ound to be voicele (and trill
ed) utterance-finally.
(170) flo[r] ~ flo[r ] flower
It can be hown that the relevant unit i the utterance and not the
word (or yllable) by noting that the
devoicing never happen when the word flor i ued in the middle of a entence.
The rule i therefore:
(171)
r ]

[-voice]
In German, obtruent are devoiced in yllable-final poition. Thu the
underlying g and d in the
following word become k and t in many intance.
(172) togo day
jogon to hunt
tok day jokdon hunting (pl.)
jokt hunting
kindij childih
kint child
a

The ymbol # (or a pair, ##) were commonplace in earlier generative

phonological work to ymbolize word

boundary.
b
The ymbol $ and . (period) are ometime een in the literature to ymbolize
yllable boundary.
c
In earlier generative phonological work, explicit reference to the y
llable wa avoided on principled ground.
Therefore, rather than indicate that omething happened at the end of
a yllable, the rule wa formulated uch that it
applied before a cononant or at the end of a word. Thi kind of c
omplicated environment wa expreed with the
brace ({}) notation, which wa ued elewhere in the theory to combine two unrel
ated environment. Direct reference
to the edge of the yllable i now no longer avoided.
146
In Sierra Nahuat, w and j are voicele in thi poition. Therefore a word uch
a /ipotiw/ it price ha
a voicele w when it occur before paue. It i alo not uncommon for vowel to
become voicele, or have
a voicele releae, in utterance-final poition.
Alo in Spanih, in highly emphatic peech, an r i trilled if it occ
ur at the end of the yllable.
d
(Otherwie it i a flap in that poition.)
(173) ma[r]te
~ ma[r]te Tueday
In Seri a p, t, or k i optionally pronounced three different way
at the end of an utterance: either
unreleaed, or with a voicele naal releae, or apirated. Following
a cononant and utterance-final, a k
may be glottalized intead.
In Chimalapa Zoque, a vowel i lengthened if it occur at the end of a treed
yllable before another
yllable.
(174)
kitpo bend,
ki:tI? bend
(root: kit)
A cononant may be tripped of certain feature in yllable-final or word-final
poition. For example, a
top uch a t may loe it coronal place feature and be pronounced
a glottal top, a in certain dialect
and peech tyle of Englih where hit i pronounced [hi1]. In ome
dialect of Nahuatl, a yllable-final
glide (j or w), which i [+continuant], how up a h, having lot ditinctive p
lace feature. In Spanih, in
certain dialect, an  i changed to h in yllable-final poition.
(175) de[]pue[] de[h]pue[h]
after
In many language, a (phonetic) glottal top i inerted between a vowel and pau
e. In Sierra Nahuat,
thi happen only after a vowel. Thu a word uch a look! ha two pronunciation
: [jkito1] at the end of
an utterance, and [jkito] if another word follow.
Initial Poition
In Englih the voicele top are apirated at the beginning of a word and alo
at the beginning of treed
yllable.
(176) [p]each
[p]ollute
ca[p]itulate
The r of Spanih i trilled if it occur at the beginning of a word, whether or no
t the word i inide a
larger utterance or not.

(177) [r]eto challenge

In Sierra Nahuat, a w optionally ha a velar onglide in word-initial
poition (making it ound like
[g]), epecially before back vowel.
Key Concept
procee which apply at edge of unit yllable word
utterance
d

We return to thee fact below. The rule i more general than thi.
147
26.1 Try it for yourelf with Korean
Complete the following rule to accounting for the ditribution of [r] and [I]. W
rite the word for road and
water phonemically, making ome choice about the repreentation of the phoneme.
148
The liquid i [I] at the end of a _______ and [r] at the beginning of a _______.
1. rubi ruby 6. muI water
2. kiri
road 7. poI
big
3. orom peron 8. ouI Seoul
4. irumi name 9. iIkop even
5. rodio radio 10. oImono how much
11. iboIzo barber
26.2 Try it for yourelf with Totonac
Conider the voicele vowel in the following word which are preented a they
would be pronounced in
a lit. I the lack of voicing in vowel predictable?
Give a rule
and dicu what other kind of data you
would like to ee.
149
1. kuku
uncle 4. jumpi porcupine
2. miki
now 5. t\pu
bean
3. n\p\p\ white 6. t\p\
he pile it up
26.3 Try it for yourelf with Seri
Doe [p] contrat with [m]? Doe it contrat with [n]? Write the word for did t
hey arrive? phonemically.
150
Give an explicit proe rule accounting for the ditribution of [p]. (The locatio
n of tre i relevant; the plurality of the ubject of the verb i not. The word are given a they occur utte
rance-finally.)
1.
tojkop did they arrive? 8.
kmomi1o it i cooked
2.
kojkomi1o they are arriving 9. i to:n i /he carrying them?
3.
toikon wa it hard? 10.
ko:ni1o /he i carrying them
4.
koikoni1o it i hard 11.
ti:m i /he leeping?
5.
toitop are there five? 12.
ki:mi1o /he i leeping
6.
koitomi1o there are five 13. to tikpon i /he working?
7.
tmom i it cooked? 14. ko tikponi1o /he i working
148
26.4 Try it for yourelf with North Puebla Nahuatl
Word ending in onorant cononant in North Puebla Nahuatl have a variant pronu
nciation when they are
in utterance-final poition. Give the rule. Similarly, word ending in top hav
e a variant pronunciation in
thi poition. Give that rule alo.
e
151
Medial Poition Final Poition

1. kc:mon kc:mon ~ kc:mon when

2. onikmot onikmot ~ onikmot I knew it
3. omokcp omokcp ~ omokcp he returned
4. tjipowok tjipowok ~ tjipowok clean
5. nokoI nokoI ~ nokoI my houe
6. niko:n niko:n ~ niko:n here
26.5 Try it for yourelf with North Puebla Nahuatl
Conider the phone [I], [I] (voiced lateral with a voicele releae),
[] (voicele lateral), [w] and [w ].
What are the phoneme and what are the rule?
152
1. potik ~ poItik wet 8. to:no:Ii day
2. to:joIi houefly 9. wc:ji big
3. wo:top blackberry 10. ko:Ii houe
4. ncw ojo ~ ncIw ojo root 11. iwok dry
5. noko ~ nokoI my houe 12. o:w po twice
6. mo:w i ~ mo:Iw i very 13. tjipowok clean
7. onikikow ~ onikiIkow
I forgot
26.6 Try it for yourelf with Sierra Popoluca
Conider the naal with voicele off-glide. I the off-glide predict
able?
Give a rule and dicu what
kind of data i lacking to make the rule precie.
153
1. jo:mo woman 5. mok corn
2. to:mm very 6. to:jj nake
3. ji:wi chili 7. wo:ji little girl
4. honn bird 8. nIt armadillo
e

Affricate and k apparently do not undergo the econd rule. Ignore thi fact.
149
26.7 Try it for yourelf with Pennylvania German
Focu on the ound [n] (repreenting a voiced uvular fricative and [
r]). They do not contrat with each
other, o both are not needed in underlying form. [r| occur in yllable onet
. Where doe [n] occur?
154
1. rot red 7. jtro treet
2. jrojo write 8. morik market
3. onjt auage 9. rcyoro rain
4. mon mare 10. nxpon neighbor
5. tin door 11. krumpin potato
6. tiro door 12. powroj farm
150
Feedback for Chapter 26
26.1 Korean
148
Syllable; yllable. /kiri/ and /mur/; or /kiIi/ and /muI/.
26.2 Totonac
149
Vowel are devoiced in final poition.
What i unclear i whether it i the end of the utterance or the end of the word
which i relevant. We would
need to ee thee word inide of a phrae to adequately formulate the rule. (In
actuality, a in o many
cae, the environment i utterance final.)
26.3 Seri
150
[p] contrat with [n], but not with [m]. Morpheme with [p] have allomorph wi

th [m]. The phonemic

form of did they arrive? i /tojkam/.
Rule: /m/ become velar when it occur at the end of an untreed utterance-fi
nal yllable (or word-final
yllableone cant tell from thee data, but the former i correct).
26.4 North Puebla Nahuatl
151
A top i optionally apirated utterance-finally.
A onorant cononant i optionally devoiced utterance-finally.
26.5 North Puebla Nahuatl
152
There are two phoneme: /I/ and /w/.
/I/ i either wholly or partially devoiced in yllable-final poition.
/w/ i voicele in yllable-final poition; it i alo voicele following a vo
icele cononant.
26.6 Sierra Popoluca
153
Naal have voicele off-glide in final poition.
What i unclear from thee data i whether it i the end of the utterance or the
end of the word which i
relevant. We would need to ee thee word inide of a phrae. (According to the
ource of the data, the
correct environment i utterance-final.)
26.7 Pennylvania German
154
/r/ i [n] when it occur in the coda (or, alternatively, in the rhyme) of a y
llable.
151
Chapter 27 - Syllable Structure
In the previou chapter we looked at ome phenomena which make reference to the
end or beginning of a
yllable. Mot linguit conider that a word mut be parable into yl
lable, and language impoe
retriction which have to do with poible yllable. One way in whi
ch thi ha been conceptualized
recently i in term of a maximal yllable template, where the templa
te pecifie the larget yllable
allowed.
a
For example, Madija ha the following maximal yllable template:
(178) [ C V ]
The larget yllable in thi language conit of a cononant and a
vowel, although of coure maller
yllable alo occur (coniting of a vowel).
b
We can pare the following word of Madija uing thi
template:
(179)

|
|
C V C V C V C V V C V V
|
| |
| |
|
|
|
|
| |
|
adne
n o p i p i n i
her tomach o h i c
There are no maximal yllable template maller than (178) in any lan
guage. In other word, every
language ha CV yllable in it inventory. But there are ome yllab
le template which are larger, of
coure. In Corongo Quechua, the maximal yllable template i:
(180) [ C V C ]
We can pare the following word uing thi template:
(181)

C V C C V C V C V C C V
| | | | |
| | | | |
| |
m o n t o k u n
it i upended o k j u
potato
Note that thi template i able to pare yllable type CV and VC a well a CVC
. An Englih word uch a
inpect could not be a word of Quechua ince it cannot be pared by the template
given above.
c
The following maximal yllable template i propoed for Ithmu Zapotec:
(182) [ C C V V ]
We can pare the following word uing thi template:
(183)

C C V C V C V V
| | |
| | | | |
j n c z o
correct g i c1
d
flower

[gjc1]

For example, ee Junko It (1986) Syllable theory in proodic phonology, Dierta
tion, Univerity of Maachuett,
Amhert.
b
In ome language (e.g. Japanee), loanword from other language are retructu
red according to the native yllable
template. In other, loanword are incorporated into the language with a yllabl
e tructure very much like the language
from which they were taken. For example, Madija peaker ue certain
word from Spanih which would not be
poible word in their language ince they violate the yllable template.
c
Loanword in Quechua do violate native yllable tructure contraint, however.
d
See Marlett and Pickett (1987) The yllable tructure and apect mor
phology of Ithmu Zapotec, International
Journal of American Linguitic 53:398-422, for argument that the [j]
of word uch a [gjc1] i to be analyzed a
hown here. The vowel of thi word i checked by a glottal top; the
glottal feature i not a cononant in the
language, but a vowel feature.
152

C V C V V C C V V
| |
| | | | | | |
ickne from embarr
n o n u i
fine  t u i
[twi]
ament
And in Seri, the maximal yllable template (excluding certain pecial cae) i
apparently the following:
e
(184) [ C C C V V V C C C ]
(185) Example:
ko:txk grahopper
tmoi:x i it circular?
kjyok who chop it off
Of coure, we are not giving a complete inventory of poible maximal
yllable template, but imply
giving ome example of what they look like.

Syllabification
We have ued the term pare above, and it need ome explanation. By ome algori
thm, a tring of ound
i canned uing a yllable template, and the tring i organized int
o yllable. There i a coniderable
amount of dicuion in the literature a to when thi i done, and
how it i done. We ignore thee iue
here, except for one. We aume, with other, that language alway organize the
tring CV into a yllable.
A equence uch a CVCV i claimed to be univerally yllabified a [CV] [CV], n
ever a [CVC] [V].
f
Internal Structure of Syllable
Some linguit have viewed the yllable a a flat tructure, a illutrated belo
w:
That i, the cononant and vowel which occur in a yllable are im
ply linked directly to the yllable
node.
g
A more common view i that the yllable ha ome kind of internal 
tructure, although there are
difference of viewpoint in thi matter. Some linguit view the ylla
ble a having two major part: onet
(O) and rhyme (R), a hown below for three word from Spanih.
h
(187)

O R
O R
O R O R
R
R
|
|
|
 c i 
p o  t o
p u c 
ix
pata
then
e
For extenive dicuion, ee Stephen A. Marlett (1988) The yllable 
tructure of Seri, International Journal of
American Linguitic 54:245-78.
f
Junko It (1986) Syllable theory in proodic phonology, propoe the following Un
iveral Core Syllable Condition: if
a cononant precede a vowel, it i in the ame yllable a that vowel.
g
George N. Clement and Samuel Jay Keyer (1983) CV phonology: a gener
ative theory of the yllable, London,
Cambridge, Ma., MIT Pre.
h
Jame Harri (1983) Syllable tructure and tre aignment in Spanih: a nonl
inear analyi, Cambridge, London,
MIT Pre.
(186)

t e l  t a r
153
Phonological rule may make reference to the unit rhyme. For example,
the rule in Spanih trilling r
mentioned in chapter 26 actually applie to any r in a rhyme, not jut to r whic
h occur at the end of the
yllable.
i
Another view of the internal tructure of the yllable which i now
quite commonly adopted, i that

there are three major contituent: Onet, Nucleu, and Coda, with the
latter two unit uually grouped
under a common node (Rhyme).
j
The nucleu i alway the part which i perceived a the yllable peak.
(188)

O
N

R
Co

f l a
t
Since there i till coniderable debate about thee different view, we do not
develop thi topic more here.
Ditribution of yllable
In ome language a yllable i never without an onet, o a condition uch a t
he following i neceary:
(189) Retriction: Onet obligatory.
All language, however, appear to have onet in mot yllable, altho
ugh they may never have complex
onet. (All language have nuclei, of coure.)
In ome language, treed yllable are pecially licened to contain
more cononant (or vowel)
than other yllable.
k
Special Syllable
In ome language yllable which appear at the edge of a word are pecial in on
e of two way. Firt, they
may be without onet, in pite of a general retriction requiring on
et in the language. Thi eem to be
handled typically by tipulating the retriction, a in (189) above.
Second,
they may include ound in the onet (if word-initial) or c
oda (if word-final) which are not
generally allowed there otherwie, or they allow for more ound than are genera
lly allowed.
l
Englih ha
ome pecial yllable. For example, there are yllable of the type
illutrated by the word aked, which
have a yllable coda [kt] that doe not occur word-medially nor in word coni
ting of only one morpheme
(note that the pat tene uffix occur in aked). (There are no nou
n like rakt in Englih, nor are they
poible.) Thee pecial yllable alway end in a coronal cononant in Englih.
m
Given thi fact, it i not
i

An example would be the r in perpicaz; the firt yllable

. See Harri (1983) Syllable tructure and tre
aignment in Spanih for a careful treatment of yllable in Spanih.
t he obtain are not thoe traditionally
held for Spanih. The argumentation i clear and careful and
ading. A major point to remember i that the
deciion about the internal tructure of the yllable cannot
without careful examination of fact other than
jut the tring of cononant and vowel.
j
Kenneth L. Pike and Eunice V. Pike (1947) Immediate contituent of

i

per

The reul
worth re
be made

Mazateco

yllable, International Journal of

American Linguitic 13:78-91, propoed a view of the yllable along thee line
.
k
For example, thi i true of Seri. See Stephen A. Marlett (1989) The yllable 
tructure of Seri.
l
Thee cononant are commonly called extrametrical. The theoretical point i th
at word are ometime compoed of
more than imply well-formed yllable. The practical point i that th
e characterization of the maximal yllable
template i thrown off by looking at word edge.
m
See Elizabeth Selkirk (1984) On the major cla feature and yllabl
e theory, in Mark Aronoff and Richard T.
Oehrle, ed. Language ound tructure, MIT Pre, Cambridge, Ma,; and John A.
Goldmith (1990) Autoegmental
and metrical phonology, Bail Blackwell, Oxford, Cambridge, Ma.
In more current term, the edge ound may be extrametrical, and may be exempt f
rom the condition impoed by the
154
poible or appropriate to imply look at a lit of monoyllabic word
 in order to determine what the
maximal yllable template i. Conider the following et of word:
(190) pog
pok
poktop
(no word like *pokpot)
We can pare all of thee word with the template [CVC] if we analyze the word 
pok a coniting of the
yllable pok plu the extra ound  which thi language allow at the beginning of
word. If the template
i [CVC], we can explain why word uch a pokpot do not occur. On the other ha
nd, if the template were
expanded to [CCVC] jut to account for the word pok, then we would
have no account for the fact that
thi kind of yllable cannot occur word-internally but may only occur at the beg
inning of a word.
Key Concept
(maximal) yllable template parability of a tring pecial yllable
internal contituent of yllable (onet, rhyme, nucleu, coda)
27.1 Try it for yourelf with North Puebla Nahuatl
What i the maximal yllable template for the following data? Show how each word
may be pared with it
by grouping the ound into yllable.
155
1. poItik wet 4. ojon now
2. onikmot I knew it 5. ijwok dry
3. okikokkc they heard it 6. ncjtik blue
27.2 Try it for yourelf with Marinahua
Thi language ha many obviou CV yllable and V yllable:
1. tono blue 5. o i to ee
2. jimo fih 6. jowo pig
3. Ii moquito 7. ojo dre
4. Ii i rubber 8. no nu duck
27.2.A. There are no word like [koto], [koto|, or [ko]. Explain how a [CV] m
aximal yllable template
account for all of thee fact, both atteted yllable and unatteted yllable
, with the aumption that there
are no pecial yllable.

156
27.2.B. But there are, in fact, word like [tmpm] waterfall, [tmjpi]
throat, and [niko i ] weat.
Firt,
how that thee cannot be handled by a [CV] template, with or withou
t pecial yllable (in the retricted
ene dicued in the chapter). Next, dicu the problem that arie
if the maximal template i imply
extended to be either [CVC] or [CCV]. What fact are unaccounted for?
157

baic yllable template. See It (1986) Syllable theory in proodic phonology for
dicuion.
155
27.3 Try it for yourelf with thee claim
The following data are baed on a preliminary, decriptive write-up of a languag
e in Ethiopia. Evaluate the
claim and provide an alternative account uing the notion of maximal yllable
and pecial yllable.
158
The language ha even kind of yllable (propoed yllable break hown by do
t):
CV do.ro chicken CVC Iom cow, dc.to happine
V o.o fih CVCC mcI anwer
VC oI mouth, on.bc.o lion VCC ird turmeric
CCVCC krcmt winter
The pattern VC and VCC only occur a ingle word or word-initially.
(Not tated explicitly, but apparently true: CC equence in coda only occur wo
rd-finally.)
Feedback for Chapter 27
27.1 North Puebla Nahuatl
155
Maximal yllable template: [CVC]
[poI] [tik|, [o] [nik] [mot], [o] [ki] [kok] [kc], [o] [jon], [ij] [wok], [ncj]
[tik].
27.2.A Marinahua
156
A maximal template of [CV] can account for the [V] yllable and for thoe that
are [CV]. A word uch
a Ii i can be analyzed a being [CV] plu a [V] yllable. If there are no pec
ial yllable allowed wordinitially or finally, then yllable uch a [ko| and [ko] cannot be generated
, and their abence i
explained.
27.2.B Marinahua
157
Thee word cannot be decribed by the [CV] template becaue niko i , for exam
ple, ha two
cononant intervocalically.
With repect to a [CCV] template, the abence of word like [koto] mut be acco
unted for. If the template
were [CCV], we would expect uch word to occur. Pike and Scott imply oberve
that the yllable type
CCV never occur phonological-word initial (p. 197).
With repect to a [CVC] template, the abence of word like [koto] and [ko] mu
t be accounted for ince
uch a template would lead one to expect them. One might propoe that a coda i
poible only in yllable

which are in a trong poition. (Elewhere, Pike and Scott claim that the firt y
llable of the phonological
word i the nuclear yllable, o thi i plauible.) The analyi mut be formaliz
ed to diallow a cloed
yllable in monoyllabic word.
27.3 Data from Ethiopia
158
A template [CVC] account for all of the data, with the following extra tateme
nt:
1. Onet i obligatory except word-intially.
2. Special yllable occur at word edge. Specifically, an extra cononant may
occur at either
edge of the word, giving CC onet and CC coda at word edge.
156
Chapter 28 - Syllable Structure Contraint
In many language there are contraint on the type of cononant th
at may occur in the coda of the
yllable, or that may cooccur in the onet or coda. For example, Englih yllabl
e may begin with fl but not
with fn. And h can occur in the onet of a yllable in Englih, but not in the c
oda. Furthermore it cannot cooccur with any other cononant in an onet ince there are not word
like [phin]. Such fact may be decribed by contraint uch a the following:
a
(191) Contraint: Onet mut be parable by the following maximal onet
template:
[
C
C
C
]
onet
[+cont]
[+on]
[Coronal]
(192) Contraint: h cannot occur in a coda.
(193) Contraint: h doe not co-occur with any other cononant in an onet.
Feature are ueful in uch contraint. The feature [onorant] properl
y define the natural cla of
cononant in the lat poition of (191). Thi condition allow the f
irt group of example below, but
diallow the econd group.
(194) print twitj kjut IIut
*ptint *pitj *gzut *Ijut
In thi way Englih reflect a common, but not univeral, trait of l
anguage to require ound which are
higher in onority (they are more reonant) to be cloer to the yllable nucleu
than ound which are lower
in onority. (A general onority cale i: top < fricative < liquid < naal
< glide < vowel.)
b
In mot cae, only vowel-like ound may occur in the Nucleu poiti
on of a yllable. But Englih
doe allow naal to be in thi poition, a in word uch a didnt. Sound which
commonly occur in Onet
or Coda poition in language are the top, fricative, naal, etc.. It i al
o not uncommon to find the high
vowel-like ound, uch a [i] and [u], in Onet or Coda poition.
c
In uch cae, they are uually
trancribed a j and w, repectively; but note that the difference i not one of
feature really, but of poition
in the yllable.
d

Nucleu, onet, or coda?

The matter of whether a ound i to be taken a being in the yllable nucleu or
the yllable onet or coda i
not decidable purely on phonetic ground. There are too many factor that inter
fere with our perception of
yllable, including knowledge of other language and deep-rooted prejud
ice. The deciion about how to
analyze a yllable i more difficult, however, when one of the vowel-like ound
i one that could be taken
a a glide, ince we know that there are language in which glide do function a
 onet (a in Englih yet) or
coda (no good example in Englih).
e
a

For one of many treatment of uch fact, ee Michael Hammond (1999
) The phonology of Englih: a proodic
optimality-theoretic approach, Oxford and New York, Oxford Univerity Pre.
b
See Elizabeth Selkirk (1984) On the major cla feature and yllable theory fo
r one attempt to make onority cale
more precie.
c
In the work of Kenneth L. Pike (1947) Phonemic, the iue of whether a high vo
wel could function in a non-nuclear
poition wa dicued under the rubric of interpretation. Sound which
can occur in either nuclear or nonnuclear
poition are called ambivalent ound.
d
In earlier generative work, they were ditinguihed by the feature yllabic. In
recent year, given the rie of yllable
theory, thi feature i generally avoided.
e
The analyi of word uch a toy uually poit a diphthong which occur in th
e yllable nucleu. Similar analye
are given for word uch a die [doj] and hay [hcj]. Thee are not analyzed a i
ntance of a glide in the coda.
157
Argument from phonological rule
Conider the following word from Spanih and Seri, which ound virtually the a
me except for the onet
(they both rhyme with Englih igh).
(195) Spanih:
[oj] there i
Seri:
[koj] bruh
Wherea one might think that the [j] in the Seri example hould be analyzed a a
yllable coda (ince Seri
doe allow cloed yllable), thi analyi would preent problem. Evi
dence from the phonology clearly
point to thi [j] a being a vowel in the yllable nucleu. For ex
ample, there are two allomorph of the
declarative morpheme: [?o] occur after vowel, and [i?o] occur after cononant
.
(196) [ po:?o] it i a wild pig cf. po:
[yo po:?o] it i a ea lion cf. yo po:
[ ktomi?o] it i a man cf. ktom
The fact that the allomorph [?o] rather than [i?o] occur after [koj

] bruh how that the [j] ound i

really functioning a a vowel.
(197) [ koj?o] (the word i not *[ koji?o])
If thi word actually ended in a cononant, we would not be able to eaily handl
e thee fact. Informed by
thee fact, the phonological repreentation of thi word i therefore
/koi/, a CCVV monoyllabic word.
(The word i written cai in the practical pelling ytem.) Thi kind of evide
nce i upported by other fact
in the language, and we can actually propoe the following contraint:
f
(198) Contraint: [j] cannot occur in a coda.
On the other hand, [j] doe occur at the beginning of a root in Se
ri. , and it i (almot) alway to be
taken a a yllable onet, not part of the yllable nucleu. For example, the ro
ot {jo:} have, own might be
analyzed (a priori) a either being {io:} or {jo:}. Phonological evide
nce can be brought to bear on thi
iue.
Prefixe uch a the dependent irreali in Seri have the ha
pe [po-] before cononant (compare
[ tponjy] when /he ran and [po ponjy] if /he run, and [p-] before vowel (com
pare [ ti:m] when /he
lept and [ pi:m] if /he leep). The allomorph [po] occur before t
he root in quetion:
[ipo jo:| if /he
own it, etablihing that thi i a cononant-initial root.
Another fact come from the familiar rule of naal place aimilation. The negat
ive prefix {m} alway
aimilate to the point of articulation of a following cononant in
Seri, (when not preceded by a treed
vowel) and it doe not change before a vowel. Since it change to [j] in word
like [itkoj jo:] doen t he
have it?, we have additional evidence that the [j] i a cononant in
thi word. In fact, the following
contraint general hold in the language:
g
(199) Contraint: [i] cannot precede another vowel (in the ame yllable).
Thi contraint correctly rule out morpheme like {iot}, but allow one like {
ooi}. The contraint on [j|
correctly diallow morpheme like {poj}, but allow one like {jo}.
Argument from pecial rule
The evidence that we have looked at o far ha been fairly traightforward phono
logical evidence, and we
have looked at cae where the [j] occur at the edge of a morpheme
. But what kind of direct evidence
f

To my knowledge, there i only one exception to thi contraint in Seri: the lo

anword [kooj] hore, which i identical
phonetically but not phonologically to the word [kooi] one who make
(it). The phonetic trancription given are
informed by our undertanding of the phonology.
g
I know of only a couple verb that may be exception.
158
might we find for the analyi of a word uch a [ 1oit] blood. The contraint o
n [j] that we have poited
above ought to rule out the analyi CVCC, o the word i not [ 1oj
t], although it ound jut like that.
Direct evidence in upport of thi come from an intereting rule of t
he language which break up

cononant cluter after a treed vowel under pecial condition. One

of thee condition i to ay
omething like Stone, my eye!. The word tone, which i [ 1ot] come out a [ 1o
ot ojo]. If there i no
cononant cluter, no a i be inerted: [ ktom] man, [ ktom ojo] Man, my eye!
If blood i CVCC, clearly
we expect a to be inerted; if it i CVVC, then we do not. The fact upport th
e latter analyi, jut a our
contraint on [j] predicted: [ 1oit ojo] Blood, my eye!
Argument from uppletive allomorphy
More evidence that upport our contraint prohibiting cluter of i followed by
another vowel i baed on
the uppletive allomorphy of the paive morpheme. A hown in chapter 3, the pa
ive morpheme ha two
uppletive form: {p} before vowel and {o1} before cononant. Since the paiv
e form of the verb own
contain the econd allomorph, a in [to1 jo:] i it owned?, we have
clear evidence that the verb own i
{jo:}; it i a CVV root.
Finally, the tre fact of the language upport the analyi that thi root
begin with a cononant [j]
and not a vowel [i]. Stre regularly fall on the firt vowel of the root. If t
he word were kiaa, other thing
being equal we would expect it to be pronounced with tre on the [i]. But the
vowel [o] i clearly more
prominent perceptually and i the locu of the phonetic correlate of tre in
the language.
h
Phonological and morphological evidence of the ort hown above i ver
y important for making
deciion about whether a ound i to be yllabified a a cononant or a vowel.
It hould be conidered more
relevant than evidence from poible type of yllable.
Key Concept
condition on yllable function of high vowel
28.1 Try it for yourelf with Marinahua
We will aume the [CVC] template propoed in chapter 34. The following kind of
cloed yllable are the
only one found in thi language: [ni], [ti] [ij], [m]. The coda may not be [
t], [t], [k], [h], or [r], for
example. Propoe a contraint, uing feature, to deal with thi.
159
28.2 Try it for yourelf with thee data
The language from Ethiopia referred to in the exercie at the end of the precedi
ng chapter ha onet cluter
which are limited to the following: br, gr, kr, and bI. Dicu the
e in term of the onority cale and
propoe a formal contraint.
160
The ame language ha coda which include the following:
t nt nd nz Ik
b
I
Id I
Ij Im
rk rd rj
Dicu thee in term of onority.
161
h

The fact are the ame for verb which dont have long vowel, uch a {-jokx} ha
ve ibling.
159

Feedback for chapter 28

28.1 Marinahua
159
If
C ]

then [+cont ]
[Coronal]
28.2 Data from Ethiopia
160
Thee equence follow the onority cale, but they are very retrictive. The f
ollowing contraint could
be propoed (written omewhat informally):
If

[ C
C

Then
Stop Liquid
161
The coda equence are much more liberal than the onet, ome of them violatin
g the onority cale in
that the yllable i not becoming le onorant at the edge. In one cae, the o
nority doe not change
(fricative-fricative), and in one cae it become more onorant (for example, t
op-fricative).
b top - fricative (violation)
I fricative - fricative (violation)
t fricative - top
nt nd naal - top
nz naal - fricative
Ik Id liquid - top
I Ij liquid - fricative
Im liquid - naal
rk rd liquid - top
rj liquid - fricative
160
Chapter 29 - Linking Feature to the Syllable
Up until now we have ignored complication with repect to how featur
e may link up to the yllable
tructure. For our dicuion we will aume that there i a level of tructure
which intervene between the
yllable and the bundle of feature (till informally repreented by letter),
often called the timing keleton
or the timing tier.
a
(200)

On
R
Nu Co
|
X X X X
timing skeleton / tier
|
|
|
|
s i p k
This level of structure has been claimed to be relevant for various reasons, but
we examine only one here.
There are pairs of sounds which occur together in a language and whi
ch function together as a unit in
certain ways. We refer to these as contour segments.
Given the concept of the skeletal tier, phonetic seuences such as [n
d] might be analyzed as
prenasalized stops (single consonants with two parts). If [nd] is anal

yzed as a prenasalized stop, a word

such as [ndop] might be represented as:
b
(201) X X X
|
|
n d o p
But the phonetic seuence [nd] might very well be linked to two X s
lots in the skeletal tier of a given
language (if the language has a syllable template which allows for it
), in which case a word like [ndap]
would be represented as follows:
(202) X X X X
|
|
|
|
n d o p
Various facts may be brought to bear on the decision as to how a given phonetic
seuence such as [nd]
or [tj] should be analyzed. Some of the possible arguments that have been or cou
ld be used are presented
below.
Argument based on the syllable template
One argument is based on whether the language has clear cases of consonant clust
ers in a single syllable. If
it does not, then one presumes that the phonetic seuence is linked
to a single X slot (why not?); if the
language does have clear cases of consonant clusters, then perhaps the
seuence [nd] is linked to two X
slots.
Seri has a word [tji1| (the name of a fish). The seuence [tj] might be analyzed
as a consonant cluster
(as shown), or as a single contour segment (an affricate). As a matt
er of fact, Seri does have consonant
a
For extensive discussion of the skeletal tier, see Goldsmith (1990) Autosegment
al & Metrical Phonology.
b
This has been proposed by Clements and Keyser (1983) CV phonology, among others
, and we adopt it here for the
sake of explication. But Goldsmith (1990) Autosegmental and metrical ph
onology gives some arguments why
affricates should not be analyzed this way. Similarly, Lombardi (1990)
The nonlinear organization of affricates,
Natural Language and Linguistic Theory 8:375
425 1990, considers affricates to h
ave a single set of features and not
contour segments. Affricates are distinguished from stops as [+delayed
release] in Chomsky and Halle (1968) The
sound pattern of English.
161
clusters inside of single syllables, as illustrated by the word [pso:k| hunger.
There is no argument from the
syllable structure to analyze [tj] as an affricate in Seri (and other reasons fo
r not analyzing it so.)
Quiegolani Zapotec has a word [gitj] grinding stone. Again, the seuence [tj] mi
ght be analyzed as a
consonant cluster, or as a single contour segment. In syllable
final position, Q
uiegolani Zapotec only has a
few clear cases of consonant clusters, and these are clusters like [It] (beginni
ng with a sonorant). This fact
provides an argument for treating the seuence [tj] in this language as an affri
cate.

c
Argument based on syllable stucture constraints
English has the word [htj], and one might wonder with the [tj] is a cons
onant cluster analogous to the
cluster [ts] that we observe in hats (compare hat), or whether it is a contour s
egment which occupies only a
single consonant position in the syllable. One argument against the cluster anal
ysis for [tj] in English is the
fact that Stop
Fricative clusters in syllable codas in English are always and on
ly Stop
s, as in lapse [Ips]
and tax [tks]. There are no syllables in English like [Ipj] and [tkI],
and therefore no independent
evidence for a cluster [tj] in a syllable coda.
Argument based on lack of independent existence of both parts
A second argument for making a decision of this sort might be based on the obser
vation that if a seuence
such as [nd] is analyzed as a consonant cluster, then one should exp
ect that the sound [d] (or [t], as they
might be allophones) occurs independently. For example, the word [t jot
o] flat in Spanish should be
analyzed as containing an affricate, not a consonant cluster. There is no phonem
e /j/ in Spanish (nor is [j]
an allophone of any other phoneme in this position in Spanish).
d
It is not possible for [tj] to be a cluster
unless it corresponds to two independently
attested phonemes.
e
This argument does not provide any help for the Seri, Quiegolani Zapotec, and En
glish facts discussed
above, however, since [t] and [j] both occur independently as well in these lan
guages.
Argument based on interaction with rules
An argument may be based on how the sounds in uestion interact with phonologica
l rules. Do they act as
a single consonant or as a cluster?
For example, a casual speech r
ule in Chimalapa Zoue voices stops
when they precede a voiced consonant: ho1tmuspo he can make twine is
pronounced ho1dmuspo in
casual speech. Fricatives do not undergo this rule: 1isjonpo he is watching it f
all is not pronounced with a
[z] in casual speech. The fact that the word tsctsmuspo he knows how
to carve is pronounced
tscdzmuspo in casual speech is entirely expected if [ts] is an affric
ate in this language, but not if it is a
stop+fricative cluster.
c
Affricates are represented with single symbols in some traditions; for example,
c
wedge (c ) is used for [tj].
d
The sound [j] occurs phonetically in some dialects of Spanish; for
these dialects, this argument would have to be
very carefully presented.
e
Those two phonemes might be /t/ and /s/ rather than /t/ and /j/, but one would
have to have an explanation for the
phonetic representation [j] regardless.
162
Alternative analyses reviewed
Phonetic seuences which might be taken as occupying either one X slot or two in

clude the following

(where t represents any consonant by example, although in some cases the two par
ts must be homorganic):
We should mention here that long vowels and consonants have been anal
yzed in recent work as single
segments (actually feature complexes, of course) linked to two X slots.
f
These decisions about how the features link up to the timing tier ar
e only indirectly related to claims
about the phonemic status of the sounds. If [d] is taken as an affricate, then it
still must be decided if [d]
is a phoneme distinct from [tj] or [d], for example. And if [d] is taken as a clu
ster [d], then it still must
be decided if [] is a phoneme distinct from [j] or [z] or [j], for example.
29.1 Try it for yourself with North Puebla Nahuatl
From a previous exercise you have found evidence for a maximal syllab
le template [CVC] for this
language. Do the following data provide evidence for expanding that te
mplate to something larger?
162
Consider the affricates (or seuences) and aspirated consonants (or se
uences) separately from the long
vowels.
1. omokwcph he returned it 4. tIokotI man
2. ko:Ii house 5. mo:wi XGT[
3. tjipowokh clean
29.2 Try it for yourself with Corongo Quechua
In chapter 27 you were told that Corongo Quechua had a maximal syllable template
[CVC]. Do words such
as the following reuire a change in that template?
163
1. i:jop his sheep
2. no:nc: my road
3. pu:kop he blows
Observation: there are no syllables such as [no:p] in this language.
Practical Procedures
We outline here some basic steps for figuring out the syllable struct
ure of a language. First, since all
languages have CV syllables, we attempt to parse all words with a CV
template. Consider the following
data from Tlapanec:
f
Early generative phonology work used the feature long; but given the skeletal t
ier, this feature has been avoided.
(203) phonetic transcription X XX
tj or ti or t' t' tj or ti
a
tw or tu or t
t
tw or tu
th or t t th
t1 or t
t
t1
ts t s ts
tj t j tj
tI t I tI
nd or
n
d
n
d or n
d
nd
(204)
X
X
X X

o
t
163
(205) gogi happy omo net
gido sown iji corn
All of these words are easily parsed as having two CV syllables. We would unders
tand from these facts that
Tlapanec has syllables which do not have onsets, at least word
initially. If th
e maximal syllable is CV, this
is uite normal.
(206) Tentative maximal syllable template: [CV]
We might then look further to see if there is anything special to be said about
the distribution of these
syllable types. An obvious place to start would be to check whether
the V syllable can occur somewhere
other than word initially. If it cannot (as appears to be the case in Tlapanec),
then we state a restriction such
as follows:
(207) Restriction: Onset obligatory except word
initially.
Now consider the following additional words from Tlapanec:
(208) ogu1 my fire mo1si watery c1dc blood
What do we do with the glottal stop? Perhaps we now have examples of syllables
with codas in Tlapanec,
where the final consonant can only be the glottal stop. Another possi
bility is that the vowels in this
language come either plain or with laryngeal modification (perceived as
a glottal stop). In the latter case,
there would be no CVC syllable. At this point, we do not have evidence to decide
between these analyses.
The following words have nonsyllabic high vowels (transcribed as glides, as we h
ear them):
(209) wobo soft ijo water ko1wi it is clean
Do we consider these words to have word
medial syllables without onsets? If so,
they would be (partially)
analyzed as:
(210) uobo (u
o
bo) iio (i
i
o) ko1ui
(ko1
u
i)
If we did this, however, we would run afoul of our proposed restrict
ion on such syllables (namely, that
syllables without onsets occur only word
initially). The facts do not
seem to warrant abandoning this
constraint since another possibility is open to us. We could consider the nonsyl
labic high vowels as linked
to an onset position in the syllable, in which case they could be (partially) an
alyzed as:
(211) wobo
ijo
ko1wi
We are able to maintain our simple syllable structure and proposed generalizatio
n.
The following words contain some possible consonant clusters:
(212) jtotso blanket mi1tsi fat
kutji thick nctho broken
okwo ant mbodi1 my (female) turkey
The seuence [jt] is not a problem. There is nothing we can do with the cluster
except admit a new syllable
type, CCV, although we do not know whether the syllable is a basic syllable or a
special syllable. (Does it
occur only word
initially? If so, it may be a special syllable.)
More difficult are the other consonant seuences in these data. Perhap
s the [tso ] syllable is another
instance of the CCV pattern, or perhaps we really have the affricate [t s], and
therefore it is the syllable type
CV. We do not have good evidence to decide between these alternatives yet; we wi

ll assume the CV analy

sis.
(213)

C C V C
V
| |
|
|
|
j t o t
o
164
Similarly, the equence [th] could be two cononant (t plu h) or a ingle con
onant (apirated t), and the
equence [kw] could be two cononant (k plu w) or one (k). (Note
that we do not conider the
poibility the equence [kwo ] i two yllable ([ku] plu [o ]) inc
e we have already propoed without
problem that non-initial yllable cannot be of the V type.) The equence [mb]
could be two cononant or
a prenaalized top. We do not have good evidence to decide between
thee alternative yet, and o will
aume the CV analyi.
(214) C V C V C V C V V C V
m i1 t

i n c t o o k o
Finally, look at the equence [tj] and [d]. A with the previou cae, thee mi
ght be analyzed a two
cononant or a ingle cononant (affricate). For thee cae, however, we ha
ve ome reaon to decide in
favor of the affricate analyi. If we were to analyze [d] a a cluter, we would
be claiming that Tlapanec
probably ha a phoneme //. Since [] doe not occur anywhere in the la
nguage other than following [d],
thi would be a trange claim to make. Therefore, we hould analyze
[d] a an affricate. If we have a
voiced affricate, we would expect that we would alo have a voicele
one, and therefore we might then
analyze [tj] a [t j] (depite the fact that [j] occur independently in the lan
guage, unlike []).
After tentative analye are made with repect to uch fact, the dec
iion and evidence upporting
them hould be written down. Evidence baed on other apect of the
phonology of the language may
become available a further tudy continue. Such evidence may force y
ou to revaluate previouly made
analytical deciion.
Key Concept
timing keleton / tier feature linked in more complex way to keletal tier
contour egment
29.3 Try it for yourelf with Seri
What i the maximal yllable template needed for the following data? I the
re evidence for treating either
[t] or [tj] a an affricate?
Show how the firt five word (in t
he firt column) hould be pared by thi
template.
164
1. toni pelican 7. tyopt i it extinguihed?
2. 1opot Papago 8. joty tiny thorn
3. tji1 fih (pecie) 9. 1oox baket
4. nop mountain lion 10. jm unet
5. kop glowworm 11. tok pumice
6. yIoIo fih (pecie) 12. txomoxo fih (pecie)

165
29.4 Try it for yourelf with Tetelcingo Nahuatl
What i the maximal yllable template needed for the following data? Show how th
e firt five word (in the
firt column) hould be pared by thi template.
165
1. timitkw Ill eat you 10. kitjiwo he make
2. vitti thorn 11. tunoI hadow
3. ittik cold 12. ok another
4. oIojojto orange grove 13. tIokctomo it bite
5. ojokti he i not here 14. itjkotI cotton
6. kipitkwo he pinche him 15. iwo and
7. kotjtcko he leeping 16. umi two
8. o only 17. puktIi moke
9. tjuko he crie 18. mo let
10. kitjiwo he make it 19. tIiintIi fire
11. itIopctj
hi bed
29.5 Try it for yourelf with other data
On page 155 you aw data from Ethiopia that motivated a [CVC] templa
te with the poibility of pecial
yllable (extra C) at word edge. Now conider the following additional data:
1. tjcr kind 6. totj under
2. mctjoI to be able 7. doro ear
3. hidi he goe 8. Iid child
4. kitjbiIoIctj he i itting 9. Iintj clue
5. tcromdi he walk 10. kintjc ground wheat
Evaluate the following propoal for the analyi of the phonetic equence [tj
] and [d]. Note: the ound
[j] occur with [t] before it, but [] never occur without [d] before it.
1. [tj] hould be analyzed a a equence of cononant becaue [t] and [j] bot
h exit in the language and
there i no reaon to poit an affricate a a phoneme. A word uch a [mctjoI] i
 CVC.CVC and word
uch a [totj] and [tjcr] how CVC plu pecial cae of C at the edge of the wo
rd. An utterance uch
a [kitjbiIoIctj] he i itting cant be analyzed properly until we know where th
e word break are.
166
2. [d] hould be analyzed a an affricate. Thi phonetic equence may be proper
ly yllabified a a ingle
cononant.
167
166
29.6 Try it for yourelf with Quioquitani Zapotec
What i the maximal yllable template needed for the following data?
What are the contraint on coda
cluter? Show how the firt five word (in the firt column) hould be pared
by thi template.
168
1. miIcntj moquito 13. Ion bed
2. winj mall 14. kpoj broom
3. ktonj Quioquitani 15. ni water
4. kwcj cheek 16. mto1 dove
5. tjc1 pitcher 17. mbon ad
6. tit egg 18. kit hair
7. jug nail 19. mbidz un
8. jutj onion 20. pc what
9. nIcpkw crooked 21. tu one
10. tipk hummingbird 22. job brain
11. jpid nout 23. pIog leaf

12.

topk corn mut 24.

kon bench

Feedback for Chapter 29

29.1 North Puebla Nahuatl
162
With repect to the cononant, thee word can all be pared with the [CVC] te
mplate if certain
equence are taken a apirated cononant and affricate. The apirated conon
ant have already been
een to be allophone of the top phoneme, o thi i appropriate. The long vow
el do preent a problem
for thi template. Long vowel occur in cloed yllable, o the long vowel cann
ot be imply occupying
both poition in the rhyme of the yllable.
29.2 Corongo Quechua
163
Thee word can till be pared with the [CVC] template if long vowel are anal
yzed a occupying both
poition in the rhyme of the yllable. Thi would account for why long vowel c
annot occur in yllable
with a final cononant.
29.3 Seri
164
Maximal yllable template: [CCVCC]
to ni
1o pot
tji1
nop
kop
There i no evidence for treating [t] or [tj] a an affricate ince cononant c
luter are perfectly fine in thi
language.
29.4 Tetelcingo Nahuatl
165
Maximal yllable template: [CVC]
ti mit

k
vit

t

i it

tik
o Io joj t

o o jok t

i
167

29.5 Other data from Ethiopia

166
Thi i all true. However, two word are till problematic. One i [kintjc].
It ha three cononant
word-medially, and one of them cannot be properly analyzed. It might be a compou
nd word, or an
expreion [kinDtjc], which would therefore be two word and not problematic. B
ut a word uch a
[Iintj] clue i very problematic for the equence analyi, which would have to
claim it i CVCCC. One i
only allowed to have one extra cononant at a word edge.
It alo eem to be the cae that [tj] occur together very often. Thi alo ugg

et that we really have an

affricate here.
167
While it i true that [d] may be yllabified a an affricate (ingle C), none of
the example how
concluively that thi mut be the cae. A word uch a [hidi] could be analyzed
a CVC.CV. Mot of the
other intance could be analyzed a C of the yllable template plu an extra e
gment at the edge of the
word.
The one problem i the word [tcromdi] he walk, which how a problematic word-i
nternal CCC unle
we adopt the affricate analyi. We would have to be convinced that thi i a i
ngle word, however, and
there i no evidence of thi yet. What if it i [tcrom di]?
The mot important evidence o far that i very important for the analyi of [d]
i the fact that there i no
[] without [d]. Thi would argue forcefully that we have an affricate and not a 
equence of cononant.
29.6 Quioquitani Zapotec
168
Maximal yllable template: [CCVCC]
Contraint: Coda cluter mut begin with a naal.
(If [j] i analyzed a a cluter, then thi contraint i not valid. The eviden
ce preented here i not
overwhelming.)
mi Icnt
j
win
kton
kc
t
j
c1
168
Chapter 30 - Stre-conditioned Procee
There are many procee in language that are affected by the preenc
e or abence of tre. (Stre i
dicued in more detail in chapter 35.) In ome cae, a rule applie only if 
tre i preent; in other cae,
a rule applie only if tre i abent.
In Englih, for example, t and d become a flap intervocalically, a
in latter and ladder, but not if the
econd vowel i treed (where it i apirated), a in fraternal.
In many language, including Englih, the vowel that are found in un
treed yllable have a le
ditinctive character and are often trancribed a chwa.
(215)
tot[o]l to t[]lity
hum[o]n hu m[]nity
On the other hand, treed vowel are often longer than untreed vowel. For
example, in Tagalog
the vowel of a non-final treed yllable i lengthened, but vowel length i no
t contrative in the language.
(216)
bohoj [ bo:hoj] houe
mog IoIobo [mog Io:Iobo]
will wah (clothe)
pom buto
[pom bu:to] intrument for making hole
In Seri, naal aimilation only applie in untreed yllable (the
verb below both end in m
underlyingly).
(217)
i:jko[p] ko1o they will arrive (untreed yllable)
mo[m] ko1o it will be cooked (treed yllable)
In Spanih, certain vowel alternate with diphthong when treed.
(218) [xu yor] to play [u]

[ xwcyo] I play [wc]

[pcn or] to think [c]
[ pjcno] I think [jc]
Key Concept
effect of tre on phonological rule
30.1 Try it for yourelf with Karuk
Conider the vowel [o:], [o], and [o]. Aume that tre i part o
f the underlying form of each of thee
word and that it i relevant to the ditribution of thee ound. What ound a
re contrative and which one
i an allophone of one of the other?
169
Give the phonological rule (in proe).
170
Write the word for bread
phonemically.
171
1. 1otro:x arm 6. 1otj no:t rat
2. tototj mama 7.
1oxok two
3. oro bread 8.
1oor wet
4. tjo no:k moquito 9. tonuk jo:nor hovel
5. no:piI beetle 10. o:o blanket
169
30.2 Try it for yourelf with Seri
Examine long cononant and long vowel in the data below and propoe
an analyi of the fact hown.
(Some Seri data preented elewhere doe not contain thi level of de
tail and o hould not be ued in
conjunction with thi exercie.) You may aume that vowel length in treed y
llable i ditinctive (ince
it i). But do not aume that jut becaue you can t predict everything you can
not predict anything. Make
morpheme cut when thi i called for.
172
Provide trancription of all of the following data which (a)
eliminate predictable length, and (b) how clearly what your undertanding of th
e phonological tructure of
thee word i.
173
1. in:o:I hi/her finger 7. i t:o:m did /he harvet it?
2. inI hi/her finger 8. kto:m who harvet it
3. 1op:o:tx ocean wave 9. kok:o:t what i bitter
4. k:om what i alive 10. tok:o:t i it bitter?
5. tkom i it alive? 11.
ko:p:o:I black widow pider
6.
montox ghot hrimp 12. optim enormou baket
Feedback for Chapter 30
30.1 Karuk
169
[o:] and [o] both occur in untreed yllable, and therefore contrat. [o] oc
cur only in treed
yllable, and [o] occur only in untreed yllable; they are likely to be al
lophone.
170
The vowel /o/ become [o] when it i untreed.
171
Bread i (probably) / oro/.
30.2 Seri
172
Long cononant are explained in the following two way. Firt, in ome cae t

hey are imply the reult

of two identical cononant (one a prefix and one a root-intial cononant) occur
ring next to each other
becaue of a word-formation rule putting them together. Second, in the remaining
cae they occur when a
CVC:V. Long
imple onet follow a treed open yllable. Thu CVCV become
vowel in untreed
yllable occur are the reult of thi ame lengthening proce: the vowel foll
owing uch a lengthened
cononant i alo lengthened. (There are, in fact, ome additional retriction
on the lengthening rule, but
thi i eentially correct.)
173
The word can be trancribed a hown below. Note that predictable cononant an
d vowel length i
omitted and that long cononant reulting from morphological juxtapoition are
written a geminate
cononant. Ditinctive vowel length (in the treed yllable) i till writte
n a uch.
1. inoI hi/her finger 7. i tto:m did /he harvet it?
2. inI hi/her finger 8. kto:m who harvet it
3. 1opotx ocean wave 9. kokot what i bitter
4. kkom what i alive 10. tokot i it bitter?
5. tkom i it alive? 11
ko:poI black widow pider .
6.
montox ghot hrimp 12. optim enormou baket
170
Chapter 31 - Epenthei
In thi chapter we look at ituation in which a phonological rule inert a ou
nd. Thee rule of inertion
are called epenthei. In ome cae, the rule are motivated by yllable-truct
ure conideration.
A yllable in Spanih cannot begin with an  followed by a cononant. In the Spa
nih word yugolavo
Yugolavian the  i part of the econd yllable, a hown below.
(219)

y u go
l a
v o
If the prefix yugo doe not occur, the  of the root lavo Slav cannot be properly
incorporated into a
yllable. In uch a cae, an e i inerted to allow the contruction of a proper
yllable for the  to attach to.
(220)


l a
v o
e 
l a
v o
In Seri a yllable cannot begin with a cononant cluter whoe firt
member i a onorant (which
include glottal top in thi language). When uch a cluter arie through word
-formation procee, an i i
inerted.
(221) 1 + jo + m + o1o i1jomo1o I didn t ee it
I
did not ee
In Madija, a j i inerted when an i precede a ditinct vowel, a
w i inerted when an o precede a
ditinct vowel, and otherwie a glottal top i inerted if a yllable would hav
e no onet.
a
(222) tio
you [tijo] ooo
other (mac.) [1owo1o]
onii other (fem.) [1oni1i ]oni
name [1oni]
poo
manioc [po1o]
Since glottal top and j do not occur otherwie in thi language, they do not ap
pear in underlying form and

are not phoneme of the language.

In Berber, a j i inerted between vowel a one way of enuring that every word
-internal yllable tart
with a cononant.
(223) Verb Verb + ox (for u)
tcu tcu-jox he made a bed
tcttu tcttu-jox he forgot
turi turi-jox he wrote
ini ini-jox he aid
a

There are ome vowel combination which do not have glottal inertio
n between them, at leat in ome word. An
example i /kaikai/ [kaikai] parrot.
171
31.1 Try it for yourelf with Ithmu Zapotec
All word-internal yllable begin with a cononant in Ithmu Zapotec. Aume th
at the Progreive prefix
i underlyingly {ko-} and tem uch a dre oneelf begin with a vowel. Accou
nt for the fact in the third
column of fact.
174
Syllable in Ithmu Zapotec may not begin with two cononant except
when the econd cononant i a
glide (and ome other pecial cae, not hown here). Aume that the Habitual p
refix i underlyingly {r-}
and that the Potential prefix i underlyingly {g-}. Note the allomorphy; account
for the fact.
175
Habitual Potential Progreive
1. roku goku kojoku dre oneelf
2. ro1dc go1dc kojo1dc receive gift
3. rc1 gc1 kojc1 drink
4. rc do gc do kojc do come
5. riniti giniti koniti lot
6. rirc girc korc leave
7. ridubi gidubi kodubi worn out
8. riIo giIo koIo puh
9. rjoko gjoko kojoko enjoy
Tranitional ound
Not all cae of epenthei are motivated by yllable tructure, however. Some a
re due to the tranition from
one point of articulation to another. Englih word like tene and fe
nce may have the equence /n/
underlyingly, and phonetically the equence [n] in ome dialect, but in other
dialect they are pronounced
with an intruive tranitional t (a part of the tranition from a voiced naal t
op to a voicele fricative)
and therefore are phonetically very cloe to or even inditinguihable from [nt
].
In Seri a tranitional vowel i inerted between nonround vowel and
round cononant that
immediately follow. Thi inertion i due to an anticipation of the round conon
ant.
(224) kik
[kiu k] who kill it itok [itoo k] did /he kill it
Formalization
One way to formalize epenthei rule i to ue an explicit rule that ay repla
ce null with omething in a
certain environment.
(225) X / Y __ Z

For example, the rule inerting the e in Spanih might be formalized a:
(226) c / __  C
But a Harri ha oberved,
b
thi formalization ignore the fact that the tatement of the enviro
nment
duplicate what one ha to ay anyway about the yllable tructure an
d contraint on the ditribution of
ound within yllable.
Alternative account have been propoed. Following It,
c
we might claim that tring of ound are
yllabified according to the template of Spanih yllable. Since Spani
h contain a contraint blocking
b

Harri (1983) Syllable tructure and tre aignment in Spanih.

c
It (1986) Syllable theory in proodic phonology.
172
yllable from beginning with C, the morpheme lavo i pared a (wi
th the  unincorporated into any
yllable at thi tage):
(227)

 I o v o
Cononant which have not been yllabified are tray cononant. Some
tray cononant in Spanih are
handled by the following rule:
(228) Inert e to the left of a tray .
d
(229)

 I o v o
c 
I o v o
Key Concept
epenthei tray cononant
d

Thi rule could be implified in variou way, although we do not go into thi
matter here.
Rule of Thumb:
The poibility of epenthetic vowel and cononant m
ake morpheme cutting more
difficult many time. When looking at data, you might do well to circle egment
which might be analyzed
a being epenthetic in order to keep track of them. You are more li
kely to find epenthetic vowel when a
morpheme-final cononant precede a morpheme-initial cononant. Therefore,
it i helpful to make early
hypothee about the underlying hape of morpheme. Then, ak yourelf if you c
an ee any motivation for
the putative inertion baed on the yllable tructure of the language. In the n
ext chapter, we ee that there
are other way that the language may ue to bring thing into line.
173
31.2 Try it for yourelf with Latin
Aume that the maximal yllable template for Latin i [CCVX] (where
X mean C or V) and that long
vowel occupy the lat two poition in the yllable.
The following noun are ome of thoe in Latin which have tem endi

ng in r. (You wouldnt know

that if we didnt tell you.) Give the underlying form for all the morpheme, incl
uding the uffixe marking
nominative and genitive, and the rule() needed to account for the urface form
.
176
Nominative Genitive
1. Iibcr Iibcri free
2. micr micri wretched
3. ogcr ogri field
4. initcr initri left
5. Iibcr Iibri book
31.3 Try it for yourelf with Serbo-Croatian
Account for the allomorph in the data below. Aume that the maximal yllable t
emplate for thi language
i [CCVCC] and that the final cononant of a maximal yllable cannot be a onora
nt. Give derivation for
the form of the word for good.
177
Maculine Feminine Neuter Plural
1. mIod mIodo mIodo mIodi young
2. put puto puto puti empty
3. zcIcn zcIcno zcIcno zcIcni green
4. tjct tjcto tjcto tjcti frequent
5. bogot bogoto bogoto bogoti rich
6. untjon untjono untjono untjoni unny
7. ropov ropovo ropovo ropovi rough
8. dobor dobro dobro dobri good
9. joon jono jono joni clear
10. Icdon Icdno Icdno Icdni frozen
11. iton itno itno itni tiny
12. ojtor ojtro ojtro ojtri harp
13. mokor mokro mokro mokri wet
174
31.4 Try it for yourelf with Southeatern Puebla Nahuatl
The following word are cited in the abolutive form. Account for the allomorphy o
f the abolutive uffix.
Aume that the maximal yllable in thi language i [CVC].
178
1.
popot drinking traw 6.
itjpotjti girl
2.
tokot man 7.
cti blood
3.
koyot coyote 8.
tcnkot goat
iwot woman 9.
4.
pokti moke
5.
tckiti egg
31.5 Try it for yourelf with American Englih
Give the alternation that are exhibited; the underlying form for four, even,
twelve, warm, and the uffix
marking ordinal; and the informal rule() needed to account for the urface for
m.
179
Cardinal Ordinal
1. Io+ Io+O four
2. ik ikO ix
3. cvon cvontO even
4. tcn tcntO ten
5. twcIv twcIIO twelve
Adjective Noun
6. w+m w+mpO warm
31.6 Try it for yourelf with Czech
Give the neceary rule and how how they account for the data. How

do they force you to conider the

equence [tj] and [t]? The Dative uffix i ometime u and ometim
e i; do not try to predict thi
alternation. Aume that the underlying form of the word for brain contain a z.
180
Accuative Dative Intrumental
1. dcn dpi dncm day
2. kmcn kmcpi kmcncm tem
3. bit bitu bitcm apartment
4. kIi:tj kIi:tji kIi:tjcm key
5. dvor dvoru dvorcm courtyard
6. potok potoku potokcm tream
7. koj koji kojcm baket
8. mozck moku mokcm brain
9. hrnct
hrnti hrntcm pot
10. rok roku rokcm year
175
Feedback for Chapter 31
31.1 Ithmu Zapotec
174
A j i inerted between when a word-internal yllable i lacking an onet.
175
An i i inerted to help yllabify a tray cononant.
31.2 Latin
176
Underlying form: {Ii:bcr} free, {micr} wretched, {ogr} field, {initr} left
, {Iibr} book.
Suffixe: Nominative (no uffix), Genitive: {i:}
Stray Epenthei (inerting an e to the left of a tray cononant) i motivated
by yllable tructure
conideration. The yllable template doe not permit final cluter. In the ca
e of {initr}, we might
propoe (a other have) that we proceed directionally (from right to left in La
tin) and inert vowel a
needed. Therefore, an c i inerted to the left of r, but it inertion enable
the t to not be a tray cononant,
blocking further epenthei.
31.3 Serbo-Croatian
177
Stray Epenthei (inerting an a to the left of a tray cononant) i motivated
by yllable tructure
conideration ince the coda Cononant-Sonorant i not permitted.
Derivation:
Underlying dobr dobr-o dobr-o dobr-i
(yllabification) (dob) r (dob) (ro) (dob) (ro) (dob) (ri)
Stray Epenthei
dobor

Surface dobor dobro dobro dobri

31.4 Southeatern Puebla Nahuatl
178
Stray Epenthei, inerting an i to the right of a tray cononant, i motivate
d by yllable-tructure
conideration. Syllable-final cononant cluter are not permitted.
31.5 Englih
179
A rule devoice a cononant before voicele cononant, a een in the form tw
cII-O.
A rule inert a top between a naal and a fricative. The epenthetic top agree
 in place of articulation with
the naal.

31.6 Czech
180
Stray Epenthei applie, inerting e to the left of a tray cononant. There a
re no yllable-final cluter.
The equence [tj] and [t] mut be analyzed a affricate.
The underlying form of brain mut be {mozk}, with devoicing of the z, either yl
lable-finally, or before a
voicele cononant.
The cononant n become palatalized before i.
176
Chapter 32 - Deletion
A cononant or a vowel can be completely lot under certain condition
. For example, in Spanih, a
cononant of the root i lot in the word preceding a cononant-initial uffix 
ince it cannot be incorporated
into any yllable. Conider the root {ckuIp} culpt, and {diting} ditingui
h when they are combined
with the uffixe {-i-r}, {-tor} (agentive nominalizer), and {-to} (adjectivaliz
er).
(230)

c
kuI pir
to culpt c kuI p tor culptor
di tin gir
to ditinguih di tin g to
ditinct
The theme vowel which occur in each of the three conjugation clae
f Spanih verb i deleted
when it occur before a vowel-initial uffix, uch a the uffix {o}
for firt peron ingular in preent
tene.
a
(231) we ... I ...
root + theme + uffix root + theme + uffix
obI + o + mo
obI + o + o peak
kom + c + mo
kom + c + o eat
bib + i + mo
bib + i + o live
Formalization
A common way to formalize deletion rule i with the ue of the null ymbol, rep
lacing a certain ound with
null in a certain context.
(232) X
/ Y __ Z
The rule deleting vowel in Spanih could be formalized a:
b
(233) V
/ ___ V
It ha been uggeted by Harri that an explicit rule deleting conon
ant in Spanih i not neceary,
however.
c
The aumption i that if a ound i not incorporated into any 
yllable, it i deleted by
convention.
d
Deletion rule often erve the function of enforcing common yllable tructure p
attern, although thi i
not alway the cae.
Coalecence
There are cae in which two ound merge into one, reulting in one ound that
ha ome feature of both
of the original ound.

In Copainala Zoque a coronal cononant and a j coalece to form a palatal or pal

atalized cononant.
a

Vowel deletion i blocked when the vowel in quetion carrie tre, a in the
vootro form (abli,for example).
b
Thi formalization doe not mention the morpheme boundary on the a
umption that thi i a rule which can only
apply in a derived environment.
c
Harri (1983) Syllable tructure and tre aignment in Spanih.
d
Thi ha been called the Stray Eraure Convention. The propoal i
that ome deletion hould be built into the
theory rather than be done by language-pecific rule. Thi reult in imple la
nguage-pecific phonological tatement
(but more complicated and intereting theorie).
177
(234) Root + Plural
wiht + joh wiht oh walk
mc1t + joh

mc1tjoh look for

oh + joh
ohjoh

cook
In Ithmu Zapotec an /j/ and a following onorant cononant coalece to become
a voicele verion
of the onorant: therefore /jI/ become voicele l, /jn/ become voicele n, e
tc.
In Corongo Quechua, certain equence of vowel and glide merge to bec
ome long vowel: /oj/
become [c:] and /ow/ become [o:].
(235) {ojwo-jko:-n}
[c:wc:kop] he i going
{owmi}

[o:mi]
ye
In ome cae of coalecence it might be poible to think of coale
cence a aimilation of one
egment to another followed by deletion of one egment. For example,
if a vowel plu naal equence i
replaced by a naalized vowel phonetically, one might argue that there are two r
ule (vowel aimilate in
naality, naal cononant delete) rather than one rule (vowel-naal co
alecence). Variou fact mut be
brought into the dicuion of thi iue which we cannot go into here.
Key Concept
deletion coalecence
32.1 Try it for yourelf with Maori
Give the neceary rule and how how they account for the data. What i the uf
fix for Paive? (Propoe a
ingle underlying form.) Show how your analyi account for example 1 and 8.
181
Active Paive
1. owhi owhitio embrace
2. hopu hopukio catch
3. oru orumio follow
4. tohu tohupio point out
5. mou mourio carry
6. wcro wcrohio tab
7. potu potuo trike, kill
8. kitc kitco ee, find
32.2 Try it for yourelf with Walmatjari
Examine the Walmatjari data in Appendix F. Propoe a imple rule to account for
the lo of cononant in

the form 4-7.

182
178
32.3 Try it for yourelf with French
Account for the allomorphy in the adjective.
183
... friend ... hore
1. oI oIami oIjovaI
only
2. poti potitami potijovaI
mall
3. gno gnozami gnojovaI
heavy
32.4 Try it for yourelf with Sre
Give the underlying form for each morpheme, and the phonological rule(
) needed. (V: indicate a long
vowel with falling pitch.)
184
Stem Cauative
1. du:h tondu:h fall
2. Iik toIik come out
3. po tonpo nure
4. mu:1 tomu:1 decend
5. tjhct tontjhct die
6. duh tonduh be hot
7. potj topotj goodbye (greeting)
8. guh tonguh wake up
9. h:r tonh:r flow
10. koh tonkoh remember
179
32.5 Try it for yourelf with Koine Greek
Give the underlying form for each morpheme, and the phonological rule() needed.
185
Nom. Sg. Gen. Sg. Dat. Pl. Acc. Pl.
1. IoiIop
IoiIopo
IoiIopi IoiIopo torm
2. ork orko
orki orko fleh
3. cIpi
cIpido
cIpii cIpido hope
4. orni
ornito
ornii ornito bird
5. orop
orobo
oropi orobo Arab
6. IIok
IIogo
IIoki IIogo flame
7. ornik
orniko
orniki orniko hen
8. kori
korito
korii korito grace
9. ocro
ocri ocro air
10. orcno
orci orcno male
11. nyk
nykto
nyki nykto night
32.6 Try it for yourelf with Aheninca Campa
Give the underlying form for each morpheme, and the phonological rule() needed.
186
Noun my ... your ...
1. mopi nomopini pimopini rock
2. opko noopkoni piopkoni ugar cane
3. topki notopkini pitopkini ant (pecie)
4. ipki nipkini pipkini peanut
5. ono nononi pononi black dye
6. opko nopkoni popkoni root (pecie)
180
Feedback for Chapter 32
32.1 Maori
181
Cononant Deletion: A word-final cononant i deleted. Or, alternatively (from
the data hown),
obtruent are not poible in the coda of the yllable, and tray cononant ar

e deleted.
Vowel Deletion: A vowel delete after a vowel (in derived environment only).
Active Paive
Underlying uwhit uwhit-iu
C-Deletion uwhi
V-Deletion
Surface uwhi uwhitiu
Active Paive
Underlying kitc kitc-iu
C-Deletion
V-Deletion kitcu
Surface kitc kitcu
The proper undertanding of thee fact ha been dicued in the literature. Th
ee data are taken from a
paper which argue that the cononant deletion analyi outlined above i not th
e correct analyi
ynchronically. It alo propoed that the allomorph [-iu] and [-u] of the Pai
ve uffix are uppletive, but it
doe not argue for thi analyi over the vowel deletion analyi propoed above
.
32.2 Walmatjari
182
A coronal cononant i deleted after a coronal liquid. Doral cononant (e.g.
k) are not deleted ee the
Dative uffix. Deletion doe not take place after obtruent or naal. We dont k
now what would happen
after glide ince thee do not occur in thi poition in the language. The feat
ure neceary to pecify
liquid are [+on], [-na]; if glide were to be excluded, we could add the feature
[+con].
32.3 French
183
From thee data it appear that obtruent cannot appear in yllable-final poi
tion. Stray Deletion
applie.
32.4 Sre
184
The unpredictable allomorph of the cauative prefix i {ton}. The naal delete
before a onorant
cononant. It may be neceary to retrict the deletion of the cononant to appl
y only to onorant, or naal
only. The data given do not warrant thi retriction, however.
Cononant Deletion: C
C
[+on]

32.5 Koine Greek

185
The voicing of a cononant and apiration of a cononant i not predictable bef
ore a vowel and mut be
included in the underlying form. Cononant are deapirated and devoiced before
a cononant. If thi rule i
pecified a a tructure preerving rule, the following i ufficient; if not, w
e mut pecify that both
cononant mut be obtruent, or mut be in different morpheme.
Rule: Laryngeal Aimilation: The laryngeal feature of a cononant pread to
a preceding
cononant.
181

Coronal top (oral and naal) are deleted before a cononant.

Rule: Coronal Stop Deletion: A coronal top delete before another cononant.
32.6 Aheninca Campa
186
The underlying form of the noun tem would be the form which appear in the f
irt column. The
affixe are: {no} (firt peron ingular poeor), {pi} (econd peron ingular
poeor), and {ni}
(perhap a uffix that change an alienable noun into one that can be poeed)
.
The rule needed i: V V

Both of the prefixe loe their vowel when they precede a vowel-initial root.
182
Chapter 33 - Underlying Form
The baic criterion for poiting underlying form i implicity. We po
it a the underlying form that form
from which the variou allomorph can mot eaily and naturally be derived by ph
onological rule uch a
aimilation, deletion, etc. There i no principle which ay that the
underlying form mut be one of the
urface allomorph, although it often i. It i a good idea to coni
der a range of poibilitie firt. For the
pat tene uffix of Englih, thee poibilitie include the following:
(236) -d (a in believed)
-t (a in walked)
-Id (a in heated)
-It (a compoite of the previou, not atteted in any urface form)
We alo conider how a decription poiting each of thee might go.
If the underlying form i {d}, then we mut account for the fact that an I i in
erted in word uch a
[potId]
a
potted and that the d i devoiced in word uch a [Icjkt] faked.
If the underlying form i {t}, then we mut account for the fact tha
t the t become voiced in word
uch a [obd] obbed, and that the t i voiced and an I inerted in form uch
a [notId] knotted.
If the underlying form i {Id}, then we mut account for the fact th
at the I i lot in word uch a
[obd] obbed, and that the I i lot and the d devoiced in word uch a [dipt]
dipped.
If the underlying form i {It}, then we mut account for the fact that the [I] i
lot in word uch a
[dipt] dipped, the I lot and the t voiced in word uch a [obd] obbed, and t
he t voiced in word uch a
[nodId] nodded and [notId] knotted.
We will begin by howing how each of thee poibilitie might be invetigated,
beginning with the lat
one and working backward. The clearet way to do thi i to put do
wn clearly what the propoed
underlying form i for a particular urface form, and then ee what rule are ne
eded to get from one to the
other. Each olution poit underlying form and phonological rule whi
ch apply to thee form, with the
intermediate repreentation (the output of each rule) hown. Every rule mut be
attempted for every form.
One cannot imply not apply a rule becaue one doe not want to. Th

e rule mut be formulated with

preciion o that it applie in jut the right context and not in the wrong con
text.
Before beginning thi demontration, we alo point out that we will b
e propoing two phonological
rule: one dealing with the alternation in voicing that we ee, and
one dealing with the preence (or
abence) of the vowel I. Since thee are related to eparate phonolog
ical procee, two rule are needed.
We do not write rule uch a the following:
(237) t Id / ...
The rule i formally too complex: it add a egment and it change the feature
of another egment at
the ame time. Furthermore, it conflate two procee which can be hown to be
independently operating
procee in the language.
We will now proceed with the comparion of the different poible ol
ution, tarting with one that
may appear to be the leat likely. (We ignore here the fact that in thee oluti
on, the rule mut apply in a
certain order; thi topic i taken up in chapter 34.)
Solution 1: Underlying {V}
(238) Deletion: I i deleted if it doe not follow a coronal top.
a

Thi form actually ha a flap intead of [t] in American Englih.

Thi fact i ignored here. The ame i true of the
word knotted mentioned below.
183
(239) Voicing Aimilation: A cononant aimilate in voicing to an immediatel
y preceding
ound.
knotted obbed dipped laced rowed
Underlying not - It
dip - It Icj - It Jow - It
ob - It
obt
Deletion
dipt Icjt Jowt
obd
Voicing nodId
dibd Icjzd Jowd
Surface *nodId
obd
*dibd *Icjzd Jowd
The Deletion rule in thi olution i complicated; notice that it ha to
include a negative environment
(not X). It applie in an environment which doe not conit of a natural cla.
b
The Voicing rule in thi olution caue problem in that it applie
incorrectly in a number of place
(even though ometime it may be debated how it i actually to be applied). Thi
olution mut be rejected
for thi reaon.
Solution 2: Underlying {F}
(240) Deletion (ame a above).
(241) Voicing Aimilation (ame a above except that we tipulate that it appli
e only in
derived environment, ee chapter 14)
Underlying not - Id ob - Id dip - Id Icj -Id Jow - Id
Deletion
obd
dipd Icjd Jowd
V. Aim. notId
dipt Icjt Jowd
obd
obd
Surface notId
dipt Icjt Jowd
Thi olution ha the ame problem a the previou one with repect to the compl
exity of the deletion
rule. It would alo have the ame problem with repect to voicing we
re it not for the poibility of
tipulating that it applie only in derived environment (a move that

would not have aved the previou

olution). Thi tipulation keep it from applying morpheme-internally t
o word uch a dipt dipped
(otherwie we would have derived either dibt or dibd, depending on how the rule
wa applied.
Solution 3: Underlying {V}
(242) Inertion: I i inerted between two coronal top.
(243) Voicing Aimilation (ame a in Solution 2 above)
Underlying not - t
ob - t
dip - t Icj - t Jow - t
Inertion notIt

obd
Voicing notId
dipt Icjt Jowd
obd
Surface notId
dipt Icjt Jowd
The inertion rule in thi olution i an improvement over the deleti
on rule in the previou olution;
the environment i imple. The rule of Voicing Aimilation applie wi
thout any problem alo if we
tipulate that it applie only in derived environment.
Solution 4: Underlying {F}
(244) Inertion: (ame a above).
(245) Voicing Aimilation: (ame a in Solution 2 above).
b

When rule are looked at from a hitorical perpective, negative environment a

re evaluated differently.
184
Underlying not - d
dip - d Icj - d Jow - d
ob - d
Inertion notId

obd
Voicing notId
dipt Icjt Jowd
Surface notId
obd
dipt Icjt Jowd
Thi olution alo work fine. Which lead u to the problem of deci
ding between Solution 3 and
Solution 4. The rule are the ame; the underlying form are differen
t; both olution work. Maybe the
voicing of the uffix i irrelevant; it would be underlying unpecified for voic
ing. People have alo pointed
out that we mut be able to ditinguih between the word burnt and burned. Whi
le there are perhap other
way to kin thi cat, the mot uual olution ha been to claim th
at the word burnt ha the irregular pat
tene uffix {-t} and that the word burned ha the regular pat tene uffix {-d
}. One mut till claim that
the form burnt doe not undergo the ame rule of Voicing Aimilation that regul
ar pat tene form do.
Phonological rule uch a Inertion and Voicing Aimilation are writt
en to account for morpheme
alternation. But they are meant to be general rule; therefore the analyt mut
invetigate and explain any
data which appear to contradict them. The analyt will alo attempt to make them
a imple and general a
poible. For example, the third peron preent uffix how alternation wh
ich are imilar to thoe of the
third peron pat: [], [z], [Iz]. We can immediately ee that our rule of Voicing
Aimilation will account for the voicelene of the plural uffix in word uch a /nik/ nick.
(246) Underlying nik - z
Devoicing nik
Surface nik
Likewie our rule of Inertion (with minor adjutmentthe environment i
two coronal cononant that
have the ame value of certain feature) will account for the inertion of the I

in word like race [+cjIz].

c
Key Concept
underlying form to account for the fact phonological rule which apply general
ly
imple environment
33.1 Try it for yourelf with Baque
Lit all allomorph. Give the underlying form for the noun tem and the uffix
e. Give the rule neceary
to account for the allomorph. (Hint: There are three rule, one of which change
 e to i when it precede a
vowel.)
187
Abolutive Abolutive Ergative Genitive
Singular Plural Proximate Indefinite Plural
1. giono gionok gionck gioncn man
2. tjokuro tjokurok tjokurck tjokurcn dog
3. oIoo oIook oIook oIocn daughter
4. pootio pootiok pootck pooticn wall
c

See Moira Yip (1988) The Obligatory Contour Principle and phonological rule: a
lo of identity Linguitic Inquiry
19:65-100 for more dicuion.
185
33.2 Try it for yourelf with Seri
The econd and third column are verb inflected for third peron ubject and th
ird peron object (when the
verb i tranitive). The tranitive verb have an extra morpheme (a prefix i) in
thee column. However, not
all cae of i in thi poition are thi morpheme. Note that the third column ha
 more intance of i than the
econd column.
Identify all morpheme. Give underlying form and the rule neceary.
Several rule are needed. Give
derivation for the variou form of the verb be painful, be cold, and look for
. Proe rule are OK.
188
Hint: Onet retriction in Seri: *

[
C
C
[+on]
(That i, no yllable onet cluter in Seri begin with a onorant cononant.)
Nominalized Unrealized Proximal
1.
kw k po mk i m:k be lukewarm
2.
kw i
ipo mi
i m:i reemble
3.
ki
pi
mi be raw
4.
ko:p
po:p
mo:p be cold
5.
ko:kto i po:kto i mo:kto look at
6.
kti
ipo ti
in ti point at
7.
kpi: ipo pi: im pi: tate
8.
ko:i i po:i i mo:i make
9.
ko:n i po:n i mo:n carry (pl.item)
10.
kko: ipo ko: ip ko: look for
11.
ki: i pi: i mi: hear
12.
kpoxim po poxim im poxim be overcooked
13.
kxiji po xiji ip xiji be painful
186
Feedback for Chapter 33

33.1 Baque
187
Two root have no allomorph: gion man and tjokur dog.
Two root have allomorph: oIo, oIoo daughter; pooti, pootc wall. (Chooe oIoo a U
F to
ditinguih it from cononant-final root like gion. Chooe {pootc} a UF, a p
er hint.)
Three uffixe have no allomorph: o Ab. Sg., ok Ab. Pl. Prox., cn Gen. Pl. C
hooe thee a UF.
One uffix ha allomorph: k, ck Erg. Indef. Chooe k a UF to ditinguih it fr
om vowel-initial
uffixe like ok.
Deletion: o delete before another vowel. (Other vowel do not delete. Alo, th
e rule may apply only in a
derived environment, or the double vowel ymbol indicate long vowel.)
o V
oIoo + cn oIocn
oIoo + o oIoo

Inertion: An e i inerted before a tray cononant. (Or, more decriptively,

an e i inerted between two
cononant, auming that [tj] i an affricate.)
gion + k gionck
Raiing: An c i raied to i before a vowel.
33.2 Seri
188
Nominalizer: k
Unreal: po
Proximal: m (from thee data; other data ugget there i more to the tory)
Tranitive: i
Stem: a in firt column without the initial k, but with m intead of naalize
d w.
Naal Velarization: m become [+back] and [+cont] after a [+back] cononant.
Naal Place Aimilation: Naal aimilate in place to an immediately followin
g cononant. (Actually, the
rule i not quite thi general; it only applie to m in Seri, urpriingly, and
only in untreed yllable.)
Deletion: V V (applie only in derived environment)

Epenthei: An i i inerted before a tray cononant.

187

Crucial Derivation:
Underlying poo:p mxiji imko: kmk
Naal Aimilation pxiji ipko:

Naal Velarization
kw k
Deletion po:p

Epenthei ipxiji

Surface po:p ipxiji

ipko:
kw k
188
Chapter 34 - Rule Ordering
In the previou chapter we aw that two phonological rule were needed to accoun
t for the allomorphy of
the pat tene uffix in Englih. Whenever two or more rule are involved, we ne
ed to conider, and perhap
tipulate, how the rule hould apply. Ideally, a theory of phonology
would tell u exactly how the rule

hould apply in mot cae, thereby implifying the phonological decri

ption of a language. Coniderable
effort ha been made to develop uch a theory of language, but to d
ate the mytery of rule application
remain largely unolved. Neverthele, coniderable progre ha been made in
finding out what theorie
of language are not adequate.
Conider firt of all the rule of Inertion and Devoicing from the previou cha
pter.
a
(247) Inertion: I i inerted between two coronal top.
(248) Voicing Aimilation: A cononant aimilate in voicing to an immediately
preceding
ound.
One might uppoe that phonological rule uch a thee apply imultaneouly to
any underlying form that
meet their tructural decription. Let u ee how thi hypothei might fare wit
h the Englih fact.
(249) Underlying form for knotted not - d
Would Inertion apply? (ye)
Would Voicing Aimilation apply? (ye)
Apply both rule at once: * notIt (incorrect)
Since the imultaneou application hypothei produce incorrect reult, and i
nce it i not obviou how it
might be fixed up to produce the correct reult, we will not conider it furthe
r.
b
Rather than take the reader through other hypothee of rule application
that have been propoed, we
will explain the hypothei for rule application which ha (with ome variation
) been widely ued within
generative phonology for over twenty-five year.
It i widely aumed that mot phonological rule appear in an ordere
d lit and apply once in the
derivation of a word.
c
All rule generally apply to all morpheme that meet their tructural decript
ion. The
linguit mut determine and pecify the order in which they apply; there i no c
ompletely reliable method
for predicting ahead of time the order in which two rule will apply.
If we aume that Inertion appear before Devoicing in the ordered l
it of rule of Englih, then we
can derive the correct reult in every cae. The firt rule applie
to the underlying form and give the
intermediate output hown beide it name. The next rule applie to the output o
f the firt rule and give the
output hown beide it name. The next rule applie imilarly, and o
on until all of the rule of the
phonology have been applied. If a rule doe not find anything to which it can ap
ply in the tring, thi lack of
application i hown by a line, a we howed in the lat chapter. (The next rule
in the derivation applie to
the lat output hown).
(250) Underlying not - d dip - d
Inertion notId
Voicing Aimilation dipt
Surface notId dipt
If we were to aume the oppoite order, incorrect reult (marked with an ater
ik) are ometime derived.

The dicuion in thi chapter till ignore the fact that the intervocalic t i
n word like potted i flapped in American
Englih. The introduction of a third rule (Flapping) to the dicuion
would make it more intereting, but alo more
complicated.
b
Such a hypothei i not to be lightly dimied, however.
c
Cyclic rule apply more than once, once on each cycle. Thi topic cannot be cov
ered in thi coure.
189
(251) Underlying not - d dip - d
Voicing Aimilation nott dipt
Inertion notIt
Surface * notIt dipt
The order in which Inertion and Voicing Aimilation apply i importa
nt. Crucial ordering uch a
thi are uually indicated in the lit of rule by mean of a curve
d line connecting the two rule. Thi i
hown in the following figure where rule W i crucially ordered befor
e rule X, and rule X i crucially
ordered before Inertion. Inertion and Rule Y are both crucially ordered before
Voicing Aimilation, but
they are not crucially ordered with repect to each other. Finally, Voicing Ai
milation i crucially ordered
before rule Z.
(252) Rule W
Rule X
Inertion
Rule Y
Voicing Aimilation
Rule Z
When one rule create input for another rule, it i aid that the firt
rule feed the econd rule. When
one rule rob another rule of input, it i aid that the firt rule bleed the 
econd rule.
When we claim that two rule mut apply in a particular order, we need to how t
hat thi i o. To do
thi, one contruct a derivation which how, a it did above, that one
order give the correct reult and
the oppoite order doe not. Thi i done for each pair of rule. The bet way t
o do thi i to find a form in
which only a given pair of rule will both apply and ee if they mut be ordered
.
When we think about a olution for a et of data, we mut contantl
y think about three part to the
olution:
(1) the underlying form,
(2) the phonological rule, and
(3) the way in which the rule apply.
d
Each of thee repreent a ditinct hypothei. We might have the right underlyi
ng form and the right
rule, but the wrong order of application. Or we might have the right rule, and
the right ordering, but the
wrong underlying form. Or everything might be right except that one phonologica
l rule i overly general or
overly retricted. Phonological analyi require the contant juggling of thee

factor to come up with the

bet olution. In general, however, rule ordering i preferred to complication o
f rule.
Try it for yourelf with Totonac
Conider the following two rule. Examine the data and determine the order in wh
ich they mut apply.
Give derivation to how the correct order and to demontrate that the oppoite
order i not correct.
189
Devoicing: Short non-laryngealized vowel are devoiced utterance-finally.
Data: /kuku/ uncle, pronounced a [kuku ] utterance-finally and [kuku] utteranc
e-medially;
h-Deletion: h i deleted utterance-finally.
Data: /kukuh/ and, pronounced a [kuku] utterance-finally and [kukuh] utteranc
e-medially,
a in [kukuh ku1] it i till and.
d

One mut alo conider the poible contraint dicued in chapter 14.
190
Effect of rule interaction
When two or more rule apply in a derivation, the effect of their i
nteraction ometime produce urface
form that look like exception to one rule or the other. Becaue of
the way in which the rule apply, the
true fact of the language may be more difficult to dicover. For example, after
looking at many word in
Hueyapan Nahuatl, one would eaily conclude that [k] and [y] are in
complementary ditribution. The
following rule of Spirantization (fricative-formation) would be formulated:
(253) Spirantization: A velar top become a fricative intervocalically.
e
But now we come acro the following data:
(254) [nikoo] I buy it [niyoo] I hell it
It look like [k] and [y] are contrating in identical environment here, and th
at we have a minimal pair. It
look like we have an exception to our rule of Spirantization. But actually it w
ould be wrong to conclude
from thee data that we have two phoneme. A cloer inpection reveal that more
i going on. Firt of all,
note that the word are morphologically complex; they have more than
one morpheme. In fact, a quick
comparion with other word how that the morpheme are the following:
f
(255) FIRST PERSON SUBJECT
niTHIRD PERSON OBJECT
kbuy koo
hell oo
The underlying form of the two word then mut be {nikkoo} I buy it and {nikoo} I 
hell it. In the
firt word the underlying cluter i reduced to a ingle k in the urface by the
following degemination rule
(given informally) which reduce a cluter of geminate cononant to a ingle co
nonant:
(256) Degemination:
k k

In the econd word the prevocalic k change to [y]. The correct derivation for
thee word mut have the

rule apply in a certain order.

(257) I buy it I hell it
Underlying ni - k - koo ni - k - oo
Spirantization niyoo
Degemination nikoo
Surface nikoo niyoo
If thee rule apply in the oppoite order, where Degemination would feed Spiran
tization, the correct form
are not derived.
(258) I buy it I hell it
Underlying ni - k - koo ni - k - oo
Degemination nikoo
Spirantization niyoo niyoo
Surface * niyoo niyoo
Thi example how very clearly the importance of our point above tha
t exception to rule may only be
apparent and not real. Rule interaction often account for what appear to be exc
eption.
e

The fricative i voiced a well, o it could be that the feature [voice] i wha
t i preading. Once we have one feature
pread, the other one can be gotten for free later on in thi language.
f
The i of the prefix ni- i actually epenthetic.
191
Key Concept
rule ordering feeding order bleeding order
Tip about Rule Application in Traditional Generative Phonology
1. Rule apply in a fixed order (to be dicovered for each language).
2. All rule are put in a lit of rule, with crucial ordering indicated. (Rule
 which are unordered in
relation to other rule are often put lat in the lit.)
3. A derivation how how all of the rule apply (or fail to apply) to a given w
ord.
4. The firt rule of the lit applie directly to the underlying form; each ub
equent rule applie to the
output of the lat rule which applied.
5. Every rule mut be attempted on every word. All rule mut be hown in each d
erivation, whether they
actually apply or not.
6. If a rule doe not apply to a word, thi i indicated by a line in the deriva
tion.
7. Crucial rule ordering mut be demontrated by one correct and one fale deri
vation for the ame
word (mot eaily done when only two rule actually apply in the derivation).
8. An incorrect output i indicated by an aterik.
9. All morpheme are included in the underlying form (at the top of the derivati
on).
g
g

In ome verion of Lexical Phonology, morpheme are added in cycle. We ignore

thi poibility here.
192
34.1 Try it for yourelf with Serbo-Croatian
The firt thirteen form below were included in an exercie in chapter 31. You 
hould review the data and
the olution for that part. Then propoe an account for the data in example 1422. Dicu the order of the

rule and how derivation to upport what you ay.

190
Maculine Feminine Neuter Plural
1. mIod mIodo mIodo mIodi young
2. put puto puto puti empty
3. zcIcn zcIcno zcIcno zcIcni green
4. tjct tjcto tjcto tjcti frequent
5. bogot bogoto bogoto bogoti rich
6. untjon untjono untjono untjoni unny
7. ropov ropovo ropovo ropovi rough
8. dobor dobro dobro dobri good
9. joon jono jono joni clear
10. Icdon Icdno Icdno Icdni frozen
11. iton itno itno itni tiny
12. ojtor ojtro ojtro ojtri harp
13. mokor mokro mokro mokri wet
14. dcbco dcbcIo dcbcIo dcbcIi fat
15. poutoo poutoIo poutoIo poutoIi tired
16. bco bcIo bcIo bcIi white
17. mio miIo miIo miIi dear
18. tco tcIo tcIo tcIi whole
19. okrugoo okrugIo okrugIo okrugIi round
20. oboo obIo obIo obIi plump
21. nogoo nogIo nogIo nogIi abrupt
22. podoo podIo podIo podIi bae
193
34.2 Try it for yourelf with Lamba
Two rule are needed for the following data. One of them i the following:
Vowel Harmony: [+high] vowel become [-high] when the preceding yllable ha a
[-high, -low]
vowel.
Thi Harmony rule take underlying {-iIo} and give the allomorph [-cI
o]. (Notice that the allomorph
[-iIo] ha a wider ditribution ince it occur after more kind of vowel.) Wha
t i the other rule?
191
Show that the two rule may apply in either order or that they mut apply in onl
y one of the two poible
order.
192
Dicu the interaction of thee rule uing the term feeding and/or bleeding.
193
Pat
Paive
Applied
Reciprocal
1. tjito
tjitwo
tjitiIo
tjitono
do
2. tuIo
tuIwo
tuIiIo
tuIono
dig
3. tjcto
tjctwo
tjctcIo
tjctono
py
opkwo
opkcIo
opkono
4. opko
pay tax
cold
5. poto
potwo
potiIo
potono
6. Iio
Iiwo
IijiIo
Iiono
hide
7. tjco
tjcwo
tjccIo
tjcono
cut
8. koo
kowo
kocIo
koono
be trong
9. Ioo
Iowo
IojiIo
Ioono
wound
10. moo
mowo
mojiIo
moono
plater
194
34.3 Try it for yourelf with Ocotepec Mixtec
Only one example of each type of root i given below. Several rule are needed.
They are:
Naal Place Aimilation: A naal aimilate in place to an immediately follow
ing cononant.
(nkoko pgoko)

Degemination: Identical cononant which are adjacent degeminate.

({nncnda} [ncndo])
Voicing: A cononant aimilate in voicing to an immediately preceding conona
nt.
({no1o} [nzo1o])
Affrication: A voiced palatal fricative aimilate in manner feature to a pre
ceding top
(becoming an affricate). (The exact formulation of thi rule depend on the feat
ure one
ue for affricate, top and fricative.) (n-o1o jdo1o)
Naal Deletion: A naal delete before an alveopalatal fricative. (njiko iko)
What are the crucial ordering? Illutrate them with derivation (howi
ng correct and incorrect reult
reulting from alternate ordering.
194
Continuative Completive
1. koko pgoko wallow
2. o1o nzo1o do, make
3. tii ndii become inoperative
4. ncndo ncndo come back
5. tju 1u jdu 1u put into
6. jiko iko ee
7. o1o jdo1o pa
195
Feedback for Chapter 34
189
Totonac
Devoicing mut precede h-Deletion.
Correct order: Incorrect order (note wrong reult):
uncle and uncle and
Underlying
kuku kukuh
Underlying
kuku kukuh
Devoicing
kuku
h-Deletion
kuku
h-Deletion
kuku
Devoicing
kuku kuku
Surface
kuku kuku
Surface
kuku * kuku
34.1 Serbo-Croatian
190
For the data in the firt part: a-Epenthei, motivated by the need to incorpor
ate a tray onorant.
For the data in the econd part: l-Vocalization. The cononant l become o whe
n it occur yllable-final.
(The revere i formally poible, given thee data, but le plauible phonolog
ically.)
Ordering: a-Epenthei mut precede l-Vocalization. The oppoite order give in
correct reult.
plump (mac.) plump (mac.)
Underlying
obI

Underlying
obI
(yllabification)
(ob) l
(yllabification)
(ob) l
a-Epenthei
oboI
l-Vocalization
obo
l-Vocalization
oboo
a-Epenthei ----Surface
oboo
Surface
* obo
In the incorrect order derivation, the l-Vocalization rule would have to be formul
ated omewhat
differently ince the l i not part of any yllable at thi point. An alternativ
e (which would not recue the
olution, of coure) would be that l become o in word-final poition. For more
dicuion, ee Kentowicz
1994:90ff.
34.2 Lamba
191
Palatalization:  i palatalized (to j) before i.
192
Ordering: Vowel Harmony mut precede Palatalization. Demontration:
Underlying
ko - iIo
Harmony
kocIo
Correct reult with thi order.
Palatalization
Surface
kocIo
196

Underlying
ko - iIo
Palatalization
kojiIo
Harmony
kojcIo
Incorrect form with thi order.
Surface
* kojcIo
193
The rule of Harmony bleed the rule Palatalization.
34.3 Ocotepec Mixtec
194
Crucial Ordering:
Affrication mut precede Voicing (to keep j and from acting identically).
Underlying n jiko
n o1o
Incorrect: n jiko
Affrication
ndo1o
Voicing niko
Voicing niko

Affrication ndiko
Naal Deletion iko

Naal Del.

Naal Place Aimilation pdo1o Naal Place Aim. pdo1o

Degemination Degem.
Surface iko
pdo1o
Surface * pdiko
Affrication mut precede Naal Deletion (for the ame reaon).
Underlying n jiko
Incorrect: n jiko
Affrication
Naal Del. jiko
Voicing niko
Affrication
Naal Deletion iko
Voicing
Naal Place Aimilation Naal Place Aim.
Degemination Degem.
Surface iko
Surface * jiko
Voicing mut precede Naal Deletion to change j to before the naal delete.
Underlying n jiko
Incorrect: n jiko
Affrication
Affrication
Voicing niko
Naal Deletion jiko
Naal Deletion iko
Voicing
Naal Place Aimilation Naal Place Aimilation
Degemination Degem.
Surface iko
Surface * jiko
197
Summary and Review Quetion for Section 4
In thi ection we have dicued a variety of important iue.
In order to capture important generalization and at the ame time i
mplify phonological
repreentation, it ha been propoed that phonological repreentation
are not fully pecified a far a
feature go. Predictable feature (either from context or from other feature) a
re omitted.
In order to capture other important generalization, it ha alo been
propoed that phonological
repreentation mut be enriched with information that i different from feature
: namely, proodic tructure
in the form of organization into yllable, etc. Thi information i often direc
tly relevant to procee which
delete or inert ound.
Finally, we have een that phonological rule may interact in interet
ing way. In order to keep rule
imple, it may be neceary to have them apply in a crucial order.
If you have matered the material in thi ection, you hould be able to
(a) explain why ome underlying form may not contain all of the fea
ture that occur in urface
form
(b) recognize when phonological procee are enitive to the location
of a yllable, word, or
utterance boundary
(c) decribe the major internal part of a yllable
(d) pare tring of ound according to propoed maximal yllable template
(e) examine a et of data and propoe an adequate maximal yllable template
(f) provide account of data which may be analyzed in more than one way, depend
ing on how the
feature are linked to the yllable tructure (affricate, glide, etc.)
(g) recognize that phonological procee may be enitive to the preence or a
bence of tre
(h) account for allomorphy by propoing epenthei and deletion rule
(i) explain the relationhip of many intance of epenthei and deletion to th
e yllable tructure
of the language
(j) explain and argue for rule ordering relationhip
For Further Reading:
Goldmith, John A. 1990. Autoegmental & metrical phonology, Oxford, Ca

mbridge, Ma.; Bail

Blackwell.
Goldmith, John A., ed. 1995. Phonological theory, Cambridge, Ma., Oxford: Ba
il Blackwell.
Harri, Jame W. 1983. Syllable tructure and tre in Spanih: a nonlinear ana
lyi, Cambridge, London;
MIT Pre.
It, Junko. 1986. Syllable theory in proodic phonology, Ph.D. Diertati
on, Univerity of Maachuett,
Amhert.
van der Hult, Harry and Norval Smith. 1982. The tructure of phonolo
gical repreentation I, II,
Dordrecht; Fori.
Yip, Moira. 1987. Englih vowel epenthei. Natural Language and Linguitic Theo
ry 5:463-84.
Review Quetion
The following quetion are to help you review the material in the preceding ec
tion.
195
198
1. (T or F) The interaction of phonological rule ometime reult in there be
ing a urface form that
appear to be a counterexample to one of the rule.
2. (T or F) The mot common yllable type among the language of the world i C
V.
3. (T or F) In ome language the yllable nucleu i optional.
4. What proce i illutrated by the following data: {obom} [obo ]
5. (T or F) The underlying form of a morpheme mut be the allomorph that occur
when there are no
affixe.
6. Name the two major part of a yllable: _______________, _________________.
7. Name the two major part of the econd part of a yllable: ______________, __
___________.
8. Circle the equence which might be analyzed in ome language a a
ingle cononant
t
phonologically: jt
nd
d
pt
t
tr
mg
kj
kh
9. Circle the word which may be pared by the following yllable tem
plate and retriction:
[CCVC], Onet obligatory.
joIt ont pt trimp
miIcpk pooi Oin tjuz tjo
10. Draw a yllable tree tructure for the word [p+int|.
Feedback for Review Quetion
195
1. T 2. T 3. F 4. coalecence 5. F 6. Onet & Rhyme
7. Nucleu & Coda 8. nd, d, t, kj, kh
9. pt, Oin, tjuz, tjo (tj could be an affricate)
10.

/
\
O
R
/ \
/ \
| |
N Co
| |
| / \
p + i n t
199
Section 5
Phonological Rule:
Supraegmental Propertie

In thi ection we examine certain feature of language that are typi

cally not linked to a particular
cononant or vowel but to the yllable or word or a larger unit. While thi may
alo be true of feature like
[naal] or [voice] a well a tre and pitch, the former are more
commonly cloely aociated with
particular cononant or vowel in underlying form, unlike the latter.
200
Chapter 35 - Stre
Stre refer to the relative degree of prominence placed on yllable within an
utterance. Thi i ometime
manifeted by loudne (greater amplitude of the acoutic ignal), but
many language manifet tre by
lengthening the treed yllable and/or by raiing the pitch of the
voice on the treed yllable. (People
who claim they are tone deaf hould not be unable to ditinguih the noun permit
from the verb permit. If
they can ditinguih them a peaker typically do, baed on the pitch, they are
not tone deaf.) Stre may
alo be manifeted in increaed mucular activity involved in articulat
ory movement.
a
Englih ue a
combination of thee feature to ignal tre: a treed yllable i
often pronounced louder and with a
higher pitch; the vowel of the yllable i uually longer, too. Stre i alo r
ealized in feature uch a the
apiration of voicele top at the beginning of treed yllable.
Although tre i actually realized on the yllable a a whole, it
i mot evident in the vowel, which
ha greater intrinic prominence anyway, becaue of it reonance. For
thi reaon, linguit have often
aigned the feature [tre] to the vowel of the yllable in particular, uing
a rule that begin a follow:
(259) V [+tre] ...
Elaborate mean have alo been devied to generate the variou level of tre
that decriptive work
on Englih have differentiated. For example, the following repreentation i ba
ed on the work like that of
Kenyon and Knott:
b
(260) 3 4 1 0
relaxation
Current theoretical repreentation of tre have been quite varied. Some appro
ache label yllable a
trong (S) or weak (W),
c
or ue other device which do not make ue of the feature [tre]. In the nota
tion
with aterik,
d
the greater number indicate a higher degree of tre.
(261)
S
W
S
*
*
*
S W S W W
* *
* * *
me tri ca li ty metri cality
Both of the notation hown above indicate that ome of the yllable are more p

rominent than other, and

they indicate their relative prominence.
Much of the work within metrical phonology attempt to find the factor that are
relevant to the rule of
tre placement in a language, on the aumption that peaker ue r
ule to aign tre to word rather
than memorize where tre i aigned to the word (apart from ome or many irr
egular word). Some of
the factor are reviewed below.
Boundary information
Some tre rule make reference to root boundarie and ome make reference to w
ord boundarie. The firt
kind place tre on a certain yllable of the root and the tre tay on tha
t yllable throughout the entire
paradigm, regardle of how many affixe are added. Stre rule in many languag
e work thi way.
(262) Seri: Stre the firt yllable of the root.
a

Peter Ladefoged (1982) A coure in phonetic, New York: Harcourt Brace Jovanovi
ch, p. 225.
b
J. Kenyon and T.A. Knott (1944) A pronouncing dictionary of American Englih, S
pringfield, Ma.: G.C. Merriam
Co. Thi work i cited in Chomky and Halle (1968) The ound pattern of Englih
, New York, Harper and Row, for
example.
c
For example, Bruce Haye (1980) A metrical theory of tre rule, Doctoral di
ertation, MIT, Cambridge.
d
For example, ee Morri Halle and Jean-Roger Vergnaud (1987) An eay on tre
, Cambridge, Ma., MIT Pre.
201
The following data how thi for Seri (with the root underlined).
e
Regardle of whether a yllable i
added at the beginning or at the end of the word, the tre alway remain in t
he ame place, on the firt
yllable of the root.
(263) tanpx Did he go home?
ihpyanpx I went home.
ihpyomanpx I didn t go home.
tanipxat Did they go home?
matanipxat Did you (plural) go home?
canpxiha S/he i going home.
In other language, the tre rule refer to the edge of the word
(either the beginning or the end).
Therefore tre doe not alway appear on the ame yllable of the root. In fac
t, it may not appear on the
root at all. Note how tre move around in Englih: dfect, defctive;
prope, propotion; nnene,
nonnical. Some tre rule refer to initial yllable, other to final, other
to penultimate (next to lat),
among other poition. (One of the goal of metrical phonology i to account
for all of thee cae and to
exclude what are believed to be impoible cae.) The following rule illutrat
e two imple cae:
(264) Chimalapa Zoque: Stre the penultimate yllable of the word (or only yll
able).

Finnih: Stre the firt yllable of the word.

f
Syllable weight
Stre placement alo often depend on yllable tructure or yllable
weight (perhap irrepective of
poition in the word, perhap in addition to thi information). A claic exampl
e i Latin where tre wa
aigned to the antepenultimate (third from end) yllable unle the p
enultimate yllable wa a heavy
yllable (one whoe rhyme wa not a imple Veither VV or VC); if the penultimate
yllable wa heavy, it
wa treed.
(265) Light penultimate yllable: rficit
Heavy penultimate yllable: ref:cit (long vowel)
refctu (cloed yllable)
Thi tre aignment rule i till reflected to ome degree in Romance languag
e and Englih. Compare,
for example, alminum (antepenultimate tre becaue of light penultimate
yllable), conndrum (heavy
penultimate yllable becaue it i cloed) and cerbrum (heavy penultimate attrac
t tre becaue it ha the
long vowel [ij] and not a hort vowel uch a [i]).
Other type of grammatical information may alo be neceary. In Engli
h, tre ometime depend
on whether the morpheme i a noun, adjective, or verb.
(266) prfect (Adjective) - perfct (Verb)
cnvert (Noun) - convrt (Verb)
In Spanih, the tre pattern for noun and adjective i quite different from
the one for verb.
g
Of coure, mot language have exception to any general rule of tre placemen
t. One way to handle
thee i by including the tre of thee exceptional word in the underlying re
preentation. Current theorie
of tre have propoed other mean of accounting for irregular tre
which are more contrained than
direct marking, but we cannot go into thee here. While uch example are often
cited in earlier literature to
illutrate that tre i phonemic (i.e. not completely predictable), they hould
not deter the analyt from
looking for the regular tre pattern in the language.
e

In thi chapter, a occaionally elewhere, we ue acute accent to indicate pr

imary tre on a word, rather than the
IPA ymbol of a raied vertical troke preceding the tre yllable.
f
John A. Goldmith (1990) Autoegmental and metrical phonology, Cambridge, Ma.
, Bail Blackwell, p. 114.
g
Jame W. Harri (1983) Syllable tructure and tre in Spanih: a
nonlinear analyi, Cambridge, London, MIT
Pre.
202
Key Concept
general tre pattern tre placement enitive to root or word boundarie
tre placement enitive to yllable weight
Potcript for Teacher
Thi chapter barely introduce the topic of tre. Stre rule that
do not refer to the edge of word (but

rather to root) do not get much attention in the current literature, which i 
ubtantial. One point that need
to be emphaized i the ditinction between the approach preented her
e and that commonly ued in the
tructuralit tradition. In that tradition, if tre wa not everywhere predict
able from the edge of the word,
then it wa conidered phonemic and therefore tre wa aumed to b
e neceary in the phonemic
repreentation of every word. Under thi view, there i nothing to ay
about tre in a language like Seri
(other than it i phonemic)or even mot language, which i a eriou
mitake. The exitence of
exception thu wa often allowed to override the potulation of a general tre
 rule within the tructuralit
tradition.
35.1 Try it for yourelf with Dakota Sioux
Give a tre rule to account for the placement of tre in the word below.
196
1. to ni to be old 6. o moni to walk on
2. kok o to eparate by triking 7. ptu jo bent over
3. mni kujo alt 8. tjik tc I kill you
4. mo joktc you kill me 9. wi tjojoktc you kill them
5. o moniu koIc
walking about, he looked for hore
35.2 Try it for yourelf with Ithmu Zapotec
Give a tre rule to account for the placement of tre in the word below.
197
1. bo ndogo leaf 6. ri bi it
2. bcndo fih 7. zo bi will it
3. gcto tortilla 8. ri ono give birth
4. gudu bizo lat year 9. zo ono will give birth
5. gi fire 10. gu ono gave birth
203
Feedback for Chapter 35
35.1 Dakota
196
Stre occur on the econd yllable of the word.
35.2 Ithmu Zapotec
197
Stre fall on the penultimate (or only) yllable of the root.
204
Chapter 36 - Introduction to Pitch
Since peaker of a language do not peak in a complete monotone, th
ere are alway variation in the
fundamental frequency of the ound they make. Change in fundamental
frequency are perceived a
change in pitch, which i often ued for linguitic purpoe. That i, it i no
t alway irrelevant to the hearer.
And ince we oberve that different people peak at different pitch level (comp
are adult with children, for
example), we know that it i not abolute pitch that i ignificant,
but the relative pitch of one yllable
compared with the yllable around it.
Tone
Pitch variation that can affect the lexical meaning of a word are called tone.
A language that ue pitch for
lexical ditinction i called a tone language. The majority of the world  lang
uage fall into thi category.
Nupe i a tone language, a can be een from the following word with
High (H), Mid (M), and Low (L)

tone.
(267) bo
high tone (H)
to be our
bo
mid tone (M)
to cut
bo
low tone (L) to count
Pitch i being ued here for a lexical purpoe, to contrat the mean
ing of different word. The three
eparate level of pitch are ignificant and mut be included in the lexical rep
reentation of the word.
Intonation
Pitch difference are alo ued in language to mark off larger phonological and
grammatical unit than the
word, uch a phrae and claue, and to ditinguih between differen
t type of uch unit. Here the
variation in pitch do not affect the lexical meaning of utterance,
but convey other type of information
uch a quetioning, anger, affection, politene, etc. Thi phenomenon i calle
d intonation.
Englih ue a falling pitch (High to Low) equence at the end of neutral declar
ative phrae:
(268)
H
L
John i from Lo Angele.
A riing pitch (Low to High) equence often indicate a ye-no quetion.
(269)
L
H
I John from Lo Angele?
If the final yllable i treed, the entire riing pitch equence i mapped on
to that yllable, reulting in a
gliding pitch equence.
(270)
LH
I he here?
While we have dicued intonation data from non-tonal language (Englih) here,
tonal language alo
may diplay intonation, a we how in the following chapter.
Key Concept
relative pitch tone intonation
205
36.1 Try it for yourelf with Ithmu Zapotec
What concluion do you draw from the following data regarding the role of pitch
in thi language?
198
1. no ndo? hot (low pitch, high pitch)
2. no ndo? bitter (low pitch, riing pitch)
3. jogo cheek (low pitch on both yllable)
4. jigo chachalaca (bird) (low pitch, riing pitch)
36.2 Try it for yourelf with Mende
What concluion do you draw from the following data regarding the role of pitch
?
199
1. mbu owl (high falling pitch)
2. mbo rice (low riing pitch)
3. pc Ic houe (high pitch on both yllable)
4. bc Ic trouer (low pitch on both yllable)
36.3 Try it for yourelf with Englih
Conider variou way in which you might change the intonation on a
phrae uch a Good night!. What
doe each intonation pattern indicate? How would you teach thi to a non-native
peaker?
Feedback for Chapter 36
36.1 Ithmu Zapotec
198

Thi i a tonal language. Pitch i being ued for lexical purpoe. The word (
or at leat the root) mut
include tonal information in their underlying form.
36.1 Mende
199
Thi i a tonal language. Pitch i being ued for lexical purpoe. The word m
ut include tonal
information in their underlying form.
206
Chapter 37 - Intonation
Intonation i like pice in food, adding incredible (and incredibly important)
variety to the way word are
ued. Conider the following different way in which the word tamale can be ai
d (S = uper high):
(271) M H L
tamale Simple tatement
L H H
tamale Quetion
L S L
tamale Surprie
L L L
tamale Dilike
L S M
tamale Surprie plu dibelief (horror?)
Maybe you dont agree with thee repreentation exactly. Sometime a major differ
ence between dialect
i a much intonation a it i light difference in the pronunciatio
n of the vowel and cononant.
Regardle, ome of thee intonation pattern are very ubtle and difficult to r
epreent. Punctuation device,
uch a comma, quetion mark, exclamation mark, underlining, and ita
licization, are only imprecie
indication of ome of the major intonation pattern in language.
An important feature of intonation, a oppoed to lexical tone, i the fact that
it i ditributed over an
entire phrae, even though one may till talk about the different tone
. A good illutration of thi i
provided by the following et of example,
a
which tart from monoyllable and end with polyyllabic
phrae, uing the ame pair of intonation tone:
(272) Exclamationfalling tone Surpriefalling-riing tone
Tom! Tom!?
Tommy! Tommy!?
telephone number! telephone number!?
Studie of intonation very often refer to variou part of the domain of a parti
cular tone, of which the
mot important i the nucleu.
For example, in a imple reading of
the following entence, the word
tamale i the nucleu and carrie the falling tone.
(273) Were going to eat tamale tonight.
One tudy of certain dialect of Englih in the United Kingdom found
that difference included the
following: (a) where treed yllable occur in relationhip to the bae-line
pitch (higher or lower), (b) the
inclination of the bae line (flat or gradually dropping), and (c) the amount of
pitch movement that occur
on treed yllable.
b

All language have intonation, even tone language,

c
although they ue it in different way and ome
reportedly tend to have impler intonation ytem.
d
The ue of intonation i an important part of what
native peaker have learned about the language they peak, becaue within a lan
guage community, the ue
a

Kenneth L. Pike (1945) The intonation of American Englih, Ann Arbor, Univerit
y of Michigan Pre, p. 26.
b
Gillian Brown, Karen L. Currie and Joanne Kenworthy (1980) Quetion
of intonation, London: Croon Helm, pp.
19-20.
c
Peter Hawkin (1984) Introducing phonology, London; Hutchinon, p. 193. But th
i doe not mean that they all ue
intonation in the ame way.
d
Paul Tench (1996) The intonation ytem of Englih, London, New York, Caell,
p. 6.
207
of intonation are hared. And correct intonation i one of the apect of langua
ge that non-native peaker
often fail to acquire for lack of attention.
One kind of change in pitch which might be mentioned here under into
nation ha been called
declination.
e
Thi i the gradual fall in pitch that occur from the beginning
of an utterance over ome
pan of word. In many tone language thi reult in ucceive tone
becoming phonetically lower and
lower in pitch until, at the end of the phrae, the high tone could be phonetic
ally a low or even lower than
the low tone at the beginning of the phrae.
f
Thi pitch decrement erve a ueful linguitic purpoe in
ignaling claue and entence boundarie,
g
which eem to be one function of intonation generally in
nontonal language, of coure.
Trancription
Many way of trancribing intonation have been ued, but apparently there i ti
ll little or no agreement on
the bet method even for Englih, much le for broad cro-linguitic ue. Some
ytem are analogou to a
phonetic trancription, and other are obviouly phonemic in that they r
equire more information to
actually pronounce them accurately, or require a ue of capitalization
that cannot be ued for tandard
phonetic trancription. Some ample are given below:
h
Line:
i
he gone to the of-fice
Number:
j

2
he gone to the
3
o
1
ffice
M
H L
Letter:
k
he gone to the office
Contour:
l
he gone to the office (not illutrated)
Diacritic:
m
he gone to the \ office
v. he gone to the / office?
he gone to the office
v. he gone to the office?
In all language, intonation ha a function, or a variety of function
. Ladefoged tate that all
language ue pitch difference to mark the boundarie of yntactic un
it. In nearly all language the
completion of a grammatical unit uch a a normal entence i ignaled by a fall
ing pitch.
n
We will look at ome common ue of intonation and contrat Englih
intonation with that of other
language o that the importance of intonation contour may be appreciated.
e

It ha alo been called downdrift, although thi term alo i ued t
o refer to omething lightly more pecific,
dicued in a later chapter.
f
John J. Ohala (1978) The production of tone, Tone: a linguitic 
urvey, ed. Victoria A. Fromkin, New York, San
Francico, London: Academic Pre, p. 31.)
g
Ohala (1978), p. 32.
h
Hawkin (1984), pp. 195f.
i
C. C. Frie (1940) American Englih grammar, New York, Appleton Century; Pike
(1945).
j
Pike (1945); G. L. Trager and H. L. Smith (1951) An outline of Englih tructu
re, Studie in linguitic, Occaional
Paper 3, Norman, Oklahoma; Mara Beatriz Fontanella de Weinberg (1980) Three inton
ational ytem of Argentinian
Spanih, in Linda R. Waugh and C. H. van Schooneveld, ed., (1980) The melody of
language, Univerity Park Pre,
Baltimore, pp. 115-126.
k
Mark Liberman and Alan Prince (1977) Proodic form and dicoure function, Ling
uitic Inquiry 8:249-336.
l
Lilia E. Armtrong and Ida C. Ward (1931) A handbook of Englih intonation, Ca
mbridge: W. Heffer & Son. Thi
graphical preentation of intonation i ued in many work a part of the phonet
ic repreentation.
m

Tench (1996), uing diagonal; Hawkin (1984), uing accent. Other ytem of
diacritic may be found in W. R.
Lee (1960) An intonation reader, London, Macmillan. The underlined word indicate
 the place where the intonational
contour i realized.
n
Peter Ladefoged (1982) A coure in phonetic, New York, Harcourt Brace Jovanovi
ch, p. 227.
208
Statement
Englih imple declarative tatement are characterized by a falling to
ne pattern on the nucleu of the
utterance.
o
The nucleu i often the lat lexical item. For example, in the exchange, A: Wh
at happened?
B: He fooled him, the falling intonation on the repone occur on the word fooled.
The pronoun him,
not being a noun, verb or adjective, doe not erve a the locu of the intonati
on pattern.
In a imple entence like He fooled him, in repone to the quetion
What happened?, falling
intonation occur on fooled (the pronoun him not counting a a lexical item). A
prepoitional phrae uch a
with a wink of hi eye at the end of thi entence would attract the falling in
tonation to the word eye: He
fooled him with a wink of hi eye.
In Seri, declarative are marked by falling pitch that begin at mid
level (repreented below a ML).
The fall occur on the treed yllable of the lat lexical item.
(274)
ML
Ciyaitim.
S/he fooled him/her.
ML
Mim.
S/he leeping.
ML
Ctam quih haxz com ciyaitim.
The man fooled the dog.
ML
Sate.
No.
Ye-No Quetion
If we change the intonation pattern on Englih word, we can change the meaning
of the utterance from a
imple tatement to a quetion. If we ay tamale with a riing intonation, we a
re aking a ye-no quetion,
which might be Are thee tamale?, or Do you want tamale?, but not What are tam
ale?.
Thi kind of imple witch of intonation to change a tatement into
an quetion i not poible in all
language. In Seri, for example, thi jut wouldnt make any ene with
out a correctly formed verb or
otherwie appropriate morphology. But language do typically have a different in
tonation pattern for ye-no
quetion. In Seri, thee are marked by a falling pitch equence that begin at
high level (indicated below by
HL).
(275)
HL
Citaitim?
Did /he fool him/her?
HL
Tim?
I /he leeping?
Apparently thi i imilar to the pattern found in Haua (a tonal language), for
example, where one effect of

quetion intonation i to raie the lat high tone of the phrae to

an extrahigh pitch with a harp fall.
p
Notice how different thee pattern are from Englih, and therefore how odd Seri
and Haua would ound if
pronounced with Englih intonation.
Note that thi riing intonation pattern i not ued in content quet
ion in Englih (ee below), nor in
tag quetion which are looking for confirmation, uch a at the end
of Clae tarted yeterday, didnt
they?
It ha been claimed that ye-no quetion almot invariably have either
a terminal rie or in ome
way a higher pitch than the correponding tatement pattern.
q
o
p

Hawkin (1984), p. 205.

Ruell G. Schuh (1978) Tone rule, in Tone: a linguitic urvey, New York: Aca
demic Pre, p. 245.
q
Cruttendon (1986), p. 162, citing finding preented in Dwight L. Bolinger (197
8) Intonation acro language, in J.
P. Greenberg, C.A. Ferguon and E. A. Moravcik (ed.), Univeral of human lang
uage, Vol. 2: Phonology, Stanford,
Stanford Univerity Pre.
209
Content Quetion
Simple quetion which expect an anwer other than ye or no have a cou
ple of common intonation
pattern in Englih:
Where did he go? (falling) and Where did he g
o? (falling-riing). In fact,
cro linguitically, both pattern are commonly found.
r
In Seri, uch content quetion have a falling intonation pattern whic
h appear on the quetion word,
the ret of the claue being rather flat
(276) H L
Quihya citaitim?
Who fooled him/her?
who
/he.fooled.him/her
Focu of information
In Englih we can ue variation on the intonation pattern to draw a
ttention to a particular element of an
utterance.
A entence like the following can be uttered without emph
ai on a particular word. The
intonation will be that of a imple declarative.
(277) Maybe on Sunday we hould eat fih and chip omewhere.
But numerou variation on thi entence would be appropriate in different conte
xt, a you can verify by
pronouncing it with a falling intonation on the underlined word.
(278) a. Maybe on Sunday we hould eat fih and chip omewhere.
b. Maybe on Sunday we hould eat fih and chip omewhere.
c. Maybe on Sunday we hould eat fih and chip omewhere.
d. Maybe on Sunday we hould eat fih and chip omewhere.
d. Maybe on Sunday we hould eat fih and chip omewhere.
e. Maybe on Sunday we hould eat fih and chip omewhere.
f. Maybe on Sunday we hould eat fih and chip omewhere.
In other language, thi ue of intonation may or may not be poible. I dont bel
ieve that uch a imple

ue of intonation i poible in Seri.


In that language, focued element are moved to a pecial place in the
entence and there marked with a pecial morpheme and intonation. The contructi
on in Seri i omewhat
more like the (awkward) Englih It i fih and chip that maybe on Sunday we ho
uld eat omewhere.
Other
In addition to the lexical tone found in Vietnamee, Thompon tate
that the final yllable of the
utterance carry the mot intonation information. If the tone of the final yll
able begin and end at a lower
than normal pitch, thi indicate certainty. If the tone of the final yllable
begin and end at a higher than
normal pitch, thi indicate uncertainty.
t
Intonation figure into Englih in facinating way that we have not dicued,
a the example at the
beginning of thi chapter partially how. There i much more to decribe here.
u
Tench 1996 lit ix major
function of intonation:
r

Cruttendon (1986), p. 165, citing finding preented in R. Ultan (1978) Some gen
eral characteritic of interrogative
ytem, in J. P. Greenberg, C.A. Ferguon and E. A. Moravcik (ed.), Univeral
 of human language, Vol. 4: Syntax,
Stanford, Stanford Univerity Pre.

According to Tench (1996:10), Haua alo doe not have the option o
f emphaizing particular element by
intonation, and Cruttenden (1986:149) claim that French and Portuguee alo do
not.
t
Laurence Thompon (1965) A Vietnamee grammar, Seattle: Univerity of Wahingto
n.
u
The variou analye of Englih intonation tetify to it complexity. For examp
le, ee Hawkin (1984), chapter 7 (a
uccinct preentation); M. A. K. Halliday (1967) Intonation and grammar
in Britih Englih, Mouton, The Hague,
Pari; and Elizabeth Couper-Kuhlen (1986) An introduction to Englih proody
, Edward Arnold, Baltimore, London;
Pike (1945); Tench (1996).
210
(1) the organization of informationthe grouping of thought, indication of new
information v. given
information;
(2) the realization of communicative functioncommand, tatement, quetio
n, peruade, requet,
etc., plu indication of deference and authority;
(3) the expreion of attitude, which i an extremely important but
very under-decribed apect of
language; how doe one how politene or anger in language X?;
(4) yntactic tructurefor example, the difference between She wahed and bruh
ed her hair (where
the intonation make it clear that he wahed her hair), and She wa
hed, and bruhed her hair
(where the intonation ugget that he wahed up (hand, faced, whatev

er) and alo bruhed her

hair;
(5) textual tructurethe way in which larger unit uch a paragraph are indic
ated phonologically;
(6) the identification of peech tyleditinguihing kind of language event,
uch a converation,
prayer, formal reading, etc.
It i alo probably true that thee different pattern are ome of t
he mot difficult thing about a
language for a peron to ue correctly when peaking another language.
Except in rare cae, they are
hardly ever taught, depite their great importance. Thi i a matter for languag
e learner to be well aware of.
37.1 Try it for yourelf on your language
Take five imple entence in your language and trancribe alternative intonatio
n pattern that may occur
with them. Practice reading your trancription back to yourelf. Then ak a cla
mate to read them.
Experiment with way to preent thi trancription uing ymbol that are common
ly available.
37.2 An exercie in trancription
200
Ue one of the method for trancribing intonation to correctly ditinguih the
following utterance. (Some
method have been incompletely repreented, o you may have to improvie.)
I ve already paid Mary. (neutral)
I ve already paid, Mary. (direct addre)
I ve already paid Mary. (emphai)
Ive already paid Mary? (quetion)
I ve already paid Mary, ... (beginning of lit)
Have another tudent read your trancription and ee that he or he can accurat
ely reproduce the entence.
37.3 Look at contrating pattern
201
Examine the following utterance and explain what "tone" are neceary
to repreent the intonation
pattern. Dicu the way in which thee interact with tre of the nucleu.
(3 = High)
Pattern 1 Pattern 2
2
Good
31
dy.
2
Good
13
dy.
2
Good
3
mr1
ning.
2
Good
1
mr3
ning.
2

Good
2
af2
ter31
non.
2
Good
2
af2
ter13
non.
2
Hap2
py
3
h1
li1
day.
2
Hap2
py
1
h3
li3
day.
211
Key Concept
intonation pattern nucleu declination
Feedback for Chapter 37
37.2 An exercie in trancription
200
1. I ve already paid Mary. (neutral)
2. I ve already paid,
Mary. (direct addre)
3. I ve already paid Mary. (emphai)
4. I ve already paid
Mary, ... (beginning of lit)
37.3 A look at contrating pattern
201
One pattern ha falling intonation and one ha riing. The tone begin on the 
treed yllable of the
nucleu. If the nucleu ha only one yllable, that yllable carrie the entire
tone; thi i alo true if the
nucleu doe not have any untreed yllable after the treed yllable (afte
rnon). Otherwie, the tone i
ditributed over the nucleu, with the firt two yllable carrying the change i
n pitch.
212
Chapter 38 - Tone
In chapter 36 we aw data from Nupe which howed that the pitch wa relevant to
the lexical meaning, jut

like egmental information. Thi i the defining ituation for a tone language.
Grammatical ue of tone
Many tone language alo ue tone to make change in grammatical mean
ing. That i, a morpheme may
conit olely of a tone (underlyinglyince uperficially the tone will be pronou
nced with ome egment,
of coure). The following data from Atatlahuca Mixtec illutrate thi,
howing that Preent tene i
conitently a High tone on the firt yllable.
(279) Future Preent
ndu ku ndu ku look for
ku nu ku nu weave
ki i ki i enter
Glide
Many tone language are bet analyzed a having imple tone underlyin
gly even though glide may
ometime occur phonetically. For example, a High tone in ome language might re
ally be in many cae a
phonetically riing tone. The claim i that the phonetic change in the tone hav
e no phonological role, and
are bet analyzed imply a phonetic correlate of the tone and not a ditincti
ve feature.
In Chiquihuitln Mazatec there i contrat between four tone.
(280) tjho (High) I talk
tjho (Higher Mid) difficult
tjho (Lower Mid) hi hand
tjho (Low) he talk
Low tone i phonetically a low falling tone when it i the only ton
e on an utterance-final yllable.
Therefore, the tone on the word for he talk i actually a glide (Low to extra L
ow) when the word i poken
in a lit, for example.
Other apparent glide in ome tone language are due to adjacent tone-bearing un
it with different level
tone. For example, in Ocotepec Mixtec a Mid-Low glide occur when two vowel oc
cur next to each other,
the firt having a Mid tone, the econd having a Low tone.
(281) ML ML compare: M L
M L
[o:] bird /oo/ [iko]
cloud
/wiko/
It i not neceary that the vowel be long in order for there to be more than on
e tone aociated with it.
The following data are alo from Ocotepec Mixtec. The Continuative mor
pheme can be analyzed a a
prefix coniting olely of a High tone that mut ubequently be a
ociated with the verb tem. Thi
aociation reult in a glide when the following tone i a Low tone.
(282) Continuative + Stem
H
L M
H LM
| |
| |
koku
koku
ecape
H
L H
H L H
| |
| |
tiwi
tiwi
become inoperative
213
Some language allow more than one tone to be linked to a ingle tone-bearing un
it in underlying form.
For example, a (high) Falling tone might be analyzed (omewhat informally) a fo
llow.
a
(283) H L

V
Such tone are referred to a contour tone. Chinee i a language that ha cont
our tone.
(284) ma high mother
ma riing hemp
ma falling cold
ma falling-riing hore
Retricted on the ditribution of tone
While in ome tone language the tone are aigned to each yllable
on a relatively free bai, in other
language evere retriction are found.
In a few language, tone pattern are aigned to morpheme or word
a a whole, rather than to
individual yllable. For example, a word might have the pattern HL realized on
one yllable a a HL glide,
on two yllable a H-L, on three yllable a H-L-L, on four yllable a H-L-L
-L, etc.
b
Other tone language limit the number and poition of tone within a word or mor
pheme. Thee have
ometime been called pitch-accent language. Pitch-accent language are
like the tone language
dicued above in that tone may be contrative. At the ame time th
ey are like tre language. One
yllable per word i marked a prominent and a tone i linked to it
firt (it i accented). The pitch on the
remaining yllable i then predictable from thi accent.
Japanee i an example of a pitch-accent language. At firt, it may look like a
imple tonal language,
ince it ha word uch a h fire (with high pitch), and h day (with lo
w pitch). Thee fact have not
been analyzed a true tone ince the ytem i o retricted.
Japanee i analyzed a having an accent: a word may have one accent or none at
all. On one yllable
word, therefore, there are two pattern (accented or not accented). O
n two yllable word there are three
pattern: neither yllable accented, firt yllable accented, or econd
yllable accented. On three yllable
word, there are four pattern: firt, econd, third, or no yllable
accented. The accented yllable mut be
indicated a being uch in underlying form ince where it goe i not predictabl
e.
The yllable are poken at high pitch until an accented yllable i reached, wh
ich i alo pronounced at
high pitch; and then the pitch drop. One extra fact: in Tokyo Japanee (hown b
elow) the firt yllable of a
word i pronounced at a lower pitch if it i not accented. Note the following pa
ttern (the accented yllable
i indicated with an acute accent):
(285) kbuto helmet HLL
kokro heart LHL
otok man LHH (next yllable i Low)
katati form LHH (next yllable i High)
Key Concept
tone language tone for grammatical meaning contour tone
pitch accent
a

The formal analyi of contour tone ha been controverial for a l

ong time. Are contour tone unanalyzable (e.g.,

Riing), or are they a equential ordered equence of pecial tone combination

(e.g. Low-High), and bet analyzed a
affricate?
b
See William R. Leben (1973) Supraegmental phonology, Ph.D. Diertati
on, MIT; William R. Leben (1978) The
repreentation of tone, Tone: a linguitic urvey, ed. Victoria A. From
kin, New York, London: Academic Pre; and
Larry M. Hyman (1987) Proodic domain in Kukuya, Natural Language and Linguiti
c Theory 5:311-33.
214
Chapter 39 - Tone Rule
Tone undergo procee imilar to thoe een in earlier chapter for
egment. They aimilate to
neighboring tone (tone feature pread), they diimilate, and they un
dergo change at the edge of
domain (epecially at the end of an utterance).
Tone are not imply feature of the vowel. They are now well etablihed a in
dependent entitie
autoegmentthat are aociated with vowel or yllable. (Some other pr
opertie of peech, perhap
many, have thi characteritic, but tone are the mot famou autoegment.)
In perhap the implet of all ituation, tone and the unit which
bear themTBU, tone-bearing
unitline up one-to-one. A two yllable word would have two tone, a
three yllable word would have
three tone, etc., and the tone and TBU would match up. The urface form of o
me word in a language
like thi would be:
(286) H L L H
p a t a t a p a
The underlying form of thee morpheme would not need to tell which
tone went where. The
information {pata, HL} would be ufficient, and by convention the tone would a
ociate with the correct
yllable.
In other ituation, and very commonly, the tone and TBU do not line up one-toone. The number of
tone may be le than the number of TBU; o omething ha to happe
n. Quite commonly, the nearet
tone pread to the TBU that lack a tone. Thi i hown by the dotted line belo
w.
(287)
H L
p a t a t a
Thi might be epecially obviou if an affix i underlyingly tonele and alway
receive it tone from the
morpheme adjacent to it.
If the number of tone i greater than the number of TBU, omething
alo ha to happen. Quite
commonly, the extra tone link with the nearet TBU.
(288) H
L H

 a z a
In both of the preceding example, the reult hown are obtained onl
y if the convention for aociating
tone begin the linking from the left ide. Thi i the mot common (but not un
iveral) ituation. Otherwie
the following reult would have been obtained.

(289)

H L H L H

p a t a t a  a z a
But two other poibilitie exit when the number of tone doe not
match the number of TBU. If
there i an extra TBU, a tone could be inerted, ay the default tone for the la
nguage. If there i an extra
tone, it might be deleted.
It i alo poible that a particular lexical item ha the tone or
one of the tone pre-linked in the
lexicon. For example, a morpheme might have the tone pattern HL, have three TBU,
but have the L prelinked to the lat TBU.
(290)
H L
i  u z a
Thi kind of pre-linking circumvent the mot general application of the aocia
tion convention. And a a
reult, the firt two yllable in the preceding example would end up with high
tone.
215
Contour Tone
Many propoal have been made in order to account for contour tone.
Yip propoe the following for
Cantonee, uing her feature of Regiter and Pitch.
a
(291) Regiter Pitch
55 i poem H h
33 i try H l
53 i ilk H hl
35 i caue H lh
22 i affair L h
(unatteted) L l
21 i time L hl
24 i city L lh
Edge Phenomena
Kukuya ha been analyzed a having two tone: High and Low. Neverthele, when a
High tone occur at
the end of an utterance, it i lowered lightly, to Mid. Therefore, a word uch
a m i pronounced with
the tone pattern Low-High uually, but with the tone pattern Low-Mid before pau
e. Therefore, we have the
following rule:
(292) H M / ___ ]

The tem blg fence i pronounced with High tone, except before paue wher
e it i pronounced
with Mid tone. The analyi given for thi propoe, typical of many
current tone analye, that yllable
with imilar pitch hare the ame tone.
b
If the word blg i repreented a
(293)
H
ba la ga
then the rule lowering High to Mid before paue will give the correct reult, a
hown below.
(294)
H
M
ba la ga
ba la ga
Aimilation (Spreading)
Tone alo aimilate to neighboring tone. In Peole Mixtec, the follo

wing tone pattern occur on

common native two yllable word:
c
(295) H H
H L
M M
L L
Thi might look like clear evidence for three tone. But if there ar
e three tone in thi language, then one
hould have found more poible combination of tone on two yllable word, un
le there are ome very
bizarre retriction. A there i not good evidence for contrat betwe
en three tone, we might attempt to
analyze thee data with only High and Low. One might take the M M pattern a th
e phonetic realization of
the tone equence L H. If thi i the cae, a High i omewhat lower after a Lo
w, and a Low i omewhat
higher before a Lowimple cae of aimilation.
In Gwari a Low pread to a following High, creating a contour tone phonetically
.
a

Moira Yip (1995) Tone in Eat Aian language. In John A.Goldmith, ed. Phonolo
gical Theory, Cambridge, Ma.,
Oxford: Bail Blackwell, pp. 476-494.
b
Larry Hyman (1975) Phonology: theory and analyi, New York, Holt, Rinehart and
Winton.
c
Some word are pronounced omewhat differently in utterance-final poition (ee
chapter 26).
216
(296) /o kpo / length L H
|
|
[o kpo ]
ok po
In Kikuyu, a Low pread to a following High, cauing the following High to be l
ot from that yllable.
(297) L
H H
x
goi - iic
bought (Immediate Pat)
One kind of aimilation which i atteted in many African language i known by
the pecial name of
downdrift (which mut be ditinguihed from imple declination, dicue
d in the chapter on intonation).
Downdrift produce what i eentially a tone that i midway between a high tone
and a low tone. Ignoring
the matter of how thi hould be handled in term of tone feature,
we might ay that the equence L H
become L M phonetically, and that thi lowered high tone count a th
e pitch level for high tone until
another equence L H occur, when the high tone i lowered again.
Hyman (1975) give the following example from Haua.
d
Note firt of all that the word are analyzed
a imply having low tone and high tone. The number indicate (very
roughly) the pitch at which the
yllable are pronounced, with 6 being the highet. Note how the pitche gradual
ly go lower and lower. The
lat high tone i at a lower pitch than the firt low tone. In ome way, thi l
ook like what one might expect

declination to look like, but it i a bit more pecial in that it i triggered b

y the low tone and i not imply a
property of the tretch of yllable.
(298) Ba la: da She: hu za: u zo:
Tone: L H L H L H L H
Pitch:
4
6 3
5 2 4
1
3
Independence of Tone
Much of the recent literature on tone ha emphaized the view that t
one operate quite differently than
feature of the vowel uch a [+high], that tone are autoegmental.
One evidence that tonal feature are
unlike true vowel feature i the way in which they perit when vow
el are deleted. The floating High
tone of Ocotepec Mixtec hown in (282) i found attached to a vowel in related l
anguage. What happened
hitorically i that the vowel wa lot and the tone wa retained a
the ole indicator of the Continuative
morpheme.
In Margi the deletion of the root vowel in the following example doe
 not mean the lo of the Low
tone that it wa bearing; intead, a contour tone (low-riing) i the reult.
(299) L
H
L H
tIo + wo
tIwo
In Bakwiri, a language game in which two yllable of a two yllable word are re
vered, the tone are
left intact. Only the cononant and vowel change poition:
(300) kc Ii falling

Ii kc (in word game)

Some two-tone language alo illutrate the independence of tone by a
phenomenon which i called
downtep, which i related to the phenomenon of downdrift mentioned above. Con
ider the phonetic tone
pattern hown below:
(301) L L
L M
analyzed a L H (cf. downdrift)
H L
H H
H M
d
Larry M. Hyman (1975), p. 226.
217
The problem i the lat pattern. If there are two tone and two yl
lable, we hould have four poible
pattern, not five. And the fact that we have H M but not M H i alo puzzling.
Thi fact, together with the
fact that we alo do not have a contrat between M and H after L, ugget that
we hould not propoe three
tone for thi ituation. So what to do?
The propoal that ha been adopted by numerou analyt i that the H M i actua
lly a pecial kind of
downdrift where the yllable which contained the triggering low tone ha been lo
t, but the downdrift effect
caued by the low tone ha been retained. So the H M i analyzed a H LH, and u
ually written H !H (high
downtepped-high).
Sometime there i evidence that the word ha a low tone underlyingly in that po
ition. But ometime
there i not, in which cae downtepped high i in eence a kind of
third tone, but one which i

predictably retricted in it ditribution and which doe not contrat

with a following high.
e
That i, once
the High pitch ha been re-et to a lower level by downtep, High pitch i now p
honetically lower for that
utterance.
Tone Feature
Our dicuion of tone o far ha been quite informal, referring to unit uch a
 H (high), L (low), and M
(mid). Thi i much like referring to the cononant p and k in eg
mental phonology and never getting
around to thinking of thee ound in term of their compoite feature (a erio
u mitake). There have been
numerou attempt at characterizing tone with feature, although we do not dic
u thee here.
Key Concept
tone change at edge of domain tone aimilation tone feature
downdrift downtep
e

For more dicuion, ee Hyman (1975), pp. 226ff; and Stephen R. An
deron (1978) Tone feature, in Victoria R.
Fromkin, Tone: a linguitic urvey, New York, etc.: Academic Pre, pp. 141ff.
39.1 Try it for yourelf with Mende (I)
According to the ource of thee data (Leben 1978:186), the majority o
f Mende monomorphemic word
have the tone pattern H, L, HL, LH, and LHL ....
Draw aociation line to link tone and vowel in the following exam
ple in order to obtain the correct
reult. (The cononant and vowel ymbol are not phonetic.)
202
HL
1. mbu mbu owl (Falling)
H
2. howomo howomo waitline
LHL
3. mbo mbo
companion (Riing-Falling)
LH
4. ndovuIo ndo vu Io ling
218
Anwer the following typological quetion baed on what you have jut learned a
bout Mende tone:
203
a. The number of tone (may)(may not) exceed the number of TBU in a word.
b. The number of TBU (may)(may not) exceed the number of tone in a word.
c. Tone aociation begin on the (left)(right).
39.2 Try it for yourelf with Mende (II)
Propoe an analyi that ue the fact from the preceding exercie to account f
or the fact below. What do
you propoe for the underlying tone of the morpheme -hu? Explain why.
204
in ...
1. k5 k5 hu war
2. mbu
mbu hu owl
mbo

3.

mba hu rice
4. pc Ic pc Ic hu houe
5. bc Ic bc Ic hu trouer
6. ngi
Io
7.

ngi Io hu dog
njo ho

njo ho hu woman
39.3 Try it for yourelf with Mende (III)
Propoe an analyi of thee fact.
205
____ buine (compound)
1. k5 k5 hi ndo war
2. mbu mbu hi
ndo owl
3. mbo
mbo hi ndo rice
4. pc Ic pc Ic hi ndo houe
5. bc Ic bc Ic hi ndo trouer
6. ngi Io ngi Io hi ndo dog
7. njo ho
njo ho hi ndo woman
39.4 Try it for yourelf with Mende (IV)
Aume that the following rule applie:
A final High tone of the v
erb tem i dropped when the verb i
conjugated in the Pat Negative. Show how you would account for each
form in the econd column,
remembering the way tone work in thi language, a hown by the pre
viou exercie. You may need to
make one new propoal, but maybe not. More than one analyi i poible.
206
Pat Negative
1. g5 nd5 g5 nd5 ni tarve
2. njo mu njo mu ni become bad, ugly
3. mc Ii mc Ii ni graze
4. gbo wo gbo wo
ni
howl
5. hi tc
hi tc ni come down
219
39.5 Try it for yourelf with Ewe
If Ewe ha only three tone (H, M, L), how might you account for the fact below
? What are the underling
tone of each morpheme?
207
1. c to mc in a mountain (mountain in) (Low-High Low)
2. c to mc in a buffalo (buffalo in) (Low-Low Low)
3. c to mc
in a mortar (mortar in)
(Mid-Mid Falling)
39.6 Try it for youelf with Marinahua

Thee data have been trancribed uing the tradition of uing high nu
mber for high tone; therefore 5 =
highet, 1 = lowet. They are a cloe interpretation of the fact a originally
preented.
A.
Determine the number of tone needed to account for the followin
g data. Firt do an analyi which
imply aume thi i a tonal language. (At that point, you may wan
t to check your anwer before
proceeding.) Be ure to give an analyi of the firt word (which ha a phonetic
ally long vowel and a falling
tone).
Procedural hint: Alway tart with the idea that tone i not contrative at all
. Will that work for one-yllable
word? Two yllable word? Three yllable word? Then go with the idea that two
tone will uffice. Will
that work? If not, then try three contrative tone, etc.
Write up thi analyi informally, howing how the tone that you poit account
for the phonetic fact.
208
B.
Then propoe a pitch-accent analyi. What fact are better handl
ed under one analyi or the other?
There are no imple word with the tone pattern to
21
to
3
,
to
21
to
3
to
3
, and to
21
to
2
to
3
.
How doe your
analyi account for thee gap? Hint: your pitch accent analyi hou
ld appear quite different from your
tone analyi.
209
1. no
51
dead
2. i
2
mi
1
blood
3. o
2
no
1
over there
4. ti
2
po

1
back of a canoe
5. i
5
jmi
1
buzzard
6. ko
3
jo
3
to hit
7. wo
51
to place
8. mo
3
po
3
head
9. to
5
po
1
root
10. tji
3
fire
11. ko
5
Ii
2
ti
1
arrow
12. to
2
chigger
13. Io
3
i
3
path
14. Io
5
i
1
rie in a river
15. Io
2
i
1
cornfield
16. mo
3
to
3
ti
3
hammer

17. to
2
ko
1
ro
1
chicken
18. o
5
ro
2
pi
1
paddle
19. po
2
pi
4
o
1
paper
(uncommon tone pattern on three
yllable word; perhap only on
loanword)
20. po
3
to
4
o
1
threw out
(uncommon tone pattern on three
yllable word)
It may or may not be relevant, but apparently word with (contratively) long vo
wel and no contour tone,
uch a to
3
, do not exit.
220
Phonetic note: Each phonological word i a rhythm unit with a nuclear yllable w
hich i longer than other
yllable of that word. (Pike and Scott 1975, p. 197)
Feedback for Chapter 39
39.1 Mende I
202
In 1, HL both attach to the only yllable. In 2, H attche to all three yllabl
e. In 3, LHL all attach to the
one yllable. In 4, L attache to the firt yllable, and H attache to the eco
nd and third.
203
a. The number of tone may exceed the number of TBU in a word.
b. The number of TBU may exceed the number of tone in a word.
c. Tone aociation begin on the left. [Thi i hown by example 4. Low i lin
ked to the firt
yllable, High to the econd, and then High i linked to the remaining yllable
in the word.)
39.2 Mende II
204
It hould be analyzed a being tonele underlyingly. It i attached to the roo

t, and the tone melody of the

root provide the tone. Where a HL melody (a in 2) appear on one yllable in t
he firt column, the ame
melody i ditributed to two yllable in the uffixed word in the econd column
.
39.3 Mende III
205
The morpheme for buine ha a low tone that i linked only to the econd ylla
ble. The firt yllable of
thi morpheme receive it tone from the tone melody of the other root.
39.4 Mende IV
206
Firt olution: the uffix -ni ha underlying low tone. In 1, L (root melody)+
L (uffix tone) account for
the low tone in the word. In 2, the dropping of the final high tone of the root
remove all tone from the
root melody and leave imply L (from the uffix). In 3, the dropping of the fin
al high tone of the root
leave a tone melody of imply L, and the correct reult are obtained. In 4, t
he underlying tone HL+L
link in traightforwardly (left to right). In 5, the tone LHL+L link left to ri
ght. Preumably the extra L tone
i deleted; there are only three tone bearing unit in the pat negative of come
down.
Second olution: the uffix -ni ha no underlying tone. We would poit a defaul
t low tone that i inerted in
the abence of any other tone. In 1, with a L melody, L pread to all yllable
. In 2, with a H melody, the H
i dropped in the negative, and default L i inerted and pread to all yllabl
e. In 3, with a LH melody, the
H i dropped in the negative, and L pread to all yllable. In 4, with a HL me
lody, the L pread to the
third yllable. In 5, with a LHL melody, the melody i linked to the three ylla
ble a uual.
39.5 Ewe
207
One might poit that mortar ha a (floating) High tone in it underlying form.
Mountain i LH, buffalo i
LL (or imply L), and mortar i MMH (or imply MH, perhap, depending on how th
ing work), with High
tone marked a being unattached or unattachable (floating) with repect to the r
oot. It attache to the next
available tone bearing unit.
39.6 Marinahua
208
Two tone are ufficient to account for the data. There are only two pattern
on one-yllable word, only
three pattern on two-yllable word (two tone would give four poible pattern
), and only five pattern
on three-yllable word. High tone ha the following variant: 5 on firt yllab
le when before Low, 4
elewhere when before Low, 3 otherwie. Low tone i 1 following a L, 1 word-fin
ally (actually, phrae221

finally) following another tone, and 2 elewhere (i.e., following H and word-med
ially and when the only
tone). (Actually, it i not word-final, but phrae-final. All of the phonetic f

act are not hown here.) Some

ample word: [no
51
] dead
/no o
/ (HL), [i
21
mi
1
] blood /i
mi
/ (LL).
Setting aide the rare word, one might even propoe that there are only three m
elodie poible in
Marinahua: H, L, and HL, with left-to-right linking of tone.
Neither propoal o far account for the lack of imple tone on long vowel. One
might ugget that
word uch a dead are phonologically /no/ (HL) and that the length i entirely
the reult of the linking of
two tone to a monoyllabic word.
209
Pitch Accent Analyi: A pitch accent analyi appear to be poible. One mig
ht propoe that there i
one or no accented yllable per word and that pitch i high until the accented 
yllable, and then low after
that; pitch i low if there i no accent on the word.
Under thi analyi, the following fact are accounted for:
One yllable:
L No accent.
H Accent.
Two yllable:
LL No accent on either yllable.
HL Accent on the firt yllable.
HH Accent on the econd yllable.
*LH Not a poible pattern ince there i no other logical poibility.
Three yllable:
LLL No accent on any yllable.
HLL Accent on the firt yllable.
HHL Accent on the econd yllable. (For ome reaon, not common; ee ex. 20)
HHH Accent on the lat yllable.
*LHH, *LLH Not poible pattern ince there i no other logical poibility.
However, LHL hould alo not be poible, but it doe occur rarely, perhap only
in loanword; ee ex. 19.
Loanword may or may not fit it in with native accentual pattern.
Thi analyi doe not account for the apparent lack of long vowel with a impl
e tone.
222
Review Quetion for Section 5
For Further Reading:
Bolinger, Dwight. 1972. Intonation: elected reading. Penguin, Middleex.
Brend, R. M. 1975. Studie in tone and intonation by member of the
Summer Intitute of Linguitic. S.
Karger: New York, London, etc.
Crytal, David. 1969. Proodic ytem and intonation in Englih. Cambridge: Cam
bridge Univerity Pre.
Cruttenden, Alan. 1986. Intonation. Cambridge Univerity Pre: London, New Yor
k, etc.

Durand, Jacque. 1990. Generative and non-linear phonology. New York, London: Lo
ngman.
Fromkin, Victoria A., ed. 1978. Tone: a linguitic urvey. New York, London: Aca
demic Pre.
Goldmith, John A. 1979. Autoegmental phonology. [1976 Ph.D. Diertation, MIT.
] New York: Garland
Pre.
Goldmith, John A. 1990. Autoegmental and metrical phonology. Cambridge
, Ma., Oxford: Bail
Blackwell.
Goldmith, John A., ed. 1995. Phonological theory, Cambridge, Ma., Oxford: Ba
il Blackwell, pp. 476494.
Halle, Morri and Jean-Roger Vergnaud. 1988. An eay on tre. Cambridge, Ma
.: MIT Pre.
Hammond, Michael. 1988. Contraining metrical theory: a modular theory
of rhythm and detreing.
[1984 Ph.D. Diertation, UCLA.] New York: Garland Pre.
Hawkin, Peter. 1984. Introducing phonology, London; Hutchinon.
Haye, Bruce. 1980. A metrical theory of tre rule. Ph.D. Diertation, MIT.
Hogg, Richard and C. B. McCully. 1987. Metrical phonology: a coureboo
k. Cambridge, London, New
York: Cambridge Univerity Pre.
Hyman, Larry. 1975. Phonology: theory and analyi. New York, London,
Sydney: Holt, Rinehart and
Winton.
Kager, Ren. 1995. The metrical theory of word tre. In John A. Gold
mith, The handbook of
phonological theory, pp. 367-402. Blackwell: Cambridge, Ma. and Oxford.
Leben, William. 1973. Supraegmental phonology. Ph.D. Diertation, MIT.
Liberman, Mark. 1979. The intonational ytem of Englih. [1975 Ph.D.
Diertation, MIT.] New York:
Garland Pre.
Odden, David. 1995. Tone: African language. In John A. Goldmith, The
handbook of phonological
theory, pp. 444-475. Blackwell: Cambridge, Ma. and Oxford.
Pike, Kenneth L. 1945 The intonation of American Englih, Ann Arbor, Univerity
of Michigan Pre.
Pike, Kenneth L. 1948. Tone language. Ann Arbor: Univerity of Michigan Pre.
Pulleyblank, Dougla. 1986. Tone in Lexical Phonology. Dordrecht: Reidel.
Tench, Paul. 1996. The intonation ytem of Englih. Caell: New York, London.
van der Hult, Harry and Keith Snider. 1993. The phonology of tone: the
repreentation of tonal regiter.
Berlin and New York: Mouton de Gruyter.
Yip, Moira. 1980. The tonal phonology of Chinee. Ph.D. Diertation, MIT.
Yip, Moira. 1995. Tone in Eat Aian language. In John A. Goldmith, ed. The ha
ndbook of phonological
theory, Blackwell: Cambridge, Ma. and Oxford, pp. 476-494
223
Review Quetion
The following quetion are to help you review the material in the preceding ec
tion.
210
(1) (T or F) Stre i the technical name for the relative degree of loudne
of a yllable.
(2) The two boundarie to which tre i mot commonly enitive cro-lingui
tically are
root and
____________.
(3) A heavy yllable i one which ha extra weight in the (a) onet (b) rhyme
.

(4) (T or F) If a language ha two word uch a badg and bdag, tre i not p
redictable and mut
be included in the underlying repreentation of all word.
(5) The linguitic ue of pitch to ditinguih lexical item i known a _____
_________.
(6) The linguitic ue of pitch to how doubt, quetion, politene,
etc. i known a
_______________.
(7) (T or F) The tone of a word and the number of vowel or yllable in the
word alway match up
one-to-one.
210
Feedback for Review Quetion
1. Fale (Stre i more than amplitude.) 2. word 3. rhyme
4. Fale (Stre might be irregular on only one of thoe word and generally pr
edictable in mot cae.)
5. tone 6. intonation 7. Fale
224
Appendix A - Feature
Sound are compoed of feature, a hown below (in an abbreviated fahion).
p
b
t

[labial] [-voice]
[labial]
[+voice] [coronal] [-voice]
Feature have ix important ue. Firt, they ditinguih between ound that oc
cur in underlying repreentation or in lexical repreentationthe phoneme. Traditionally thi ha been
called their claificatory
function. The phoneme of a language mut differ in at leat one feature. For ex
ample, /p/ and /b/ are two
phoneme of Englih and are ditinguihed by the feature [voice]. When ther term
feature i ued in thi
way, it i typically modified by the word ditinctive, o we peak of
the ditinctive feature of a
language.
Second, feature are ued to decribe the phonetic ound of a langua
ge. In ome dialect of Spanih
there are two way to pronounce the phoneme //, depending on the context: [],
a in pi[]ta and [z], a in
de[z]de. Thee two pronunciation are characterized by a difference in the value
for the feature [voice]. We
alo need feature to decribe the phonetic difference between the t of Spani
h and the t of Englih, ince
they are not the ame phonetically. Thi ue of feature complicate
the election of the feature that we
want to include in the univeral et. The reult i that there i till ome di
agreement among linguit with
repect to feature. Feature theory i till grappling with thee and
other iue. In fact, anything we ay
here i outdated already.
Third, feature are ued to define natural clae of ound. Many time a group
of ound i found in a
pecific context or undergoe a common change or i related to a cer
tain phonological proce. For
example, the three top /p/, /t/, and /k/ of Englih have apirated
variant in certain context. With very
few feature we can iolate thee top from the ret of the cononant of Engli
h in order to indicate which
are the cononant that participate in thi phonological proce. We r

eturn to the topic of natural clae

below.
Fourth, feature are ued to decribe phonological procee. The fact that in a
given language a vowel
i naalized before a naal cononant mut be captured formally in an adequate w
ay.
Fifth, (ditinctive) feature have been ued to attempt to explain why certain 
ound exit and other do
not.
Sixth, feature have been ued a part of a implicity metric; that
i, they have been ued to formally
meaure the implicity of a phonological rule. A natural proce (one that i co
mmonly found in language)
hould be tatable with fewer feature than a proce which i unnatu
ral.
a
Although earlier verion of
generative phonology had difficulty in making the evaluation metric work well, g
ood progre in thi area
ha been made with the ue of feature geometry (dicued below).
The choice of feature i difficult becaue one mut chooe the feature that e
rve the bet for all of the
ue mentioned above. For thi reaon there i alo coniderable diagreement am
ong linguit with repect
to thi matter.
The ditinctive feature in their claificatory function are ued only with two
pecified value: plu (+)
and minu (-), or in a privative way (the feature i either preent or it i not p
reent). The ue of binary
value ha it bai in variou obervation. For example, many langua
ge have oral vowel but no
naalized vowel. Other language have oral vowel and naalized vowel. But the
re i no language, to our
knowledge, that ha oral vowel, naalized vowel, and emi-naalized v
owel a phoneme.
b
If the
a

For early argument along thee line, ee Morri Halle (1962) Phonology in gen
erative grammar, Word 18:54-72.
b
The one example that ha been brought forward i from Palantla Chinantec, dicu
ed in Ladefoged and Maddieon
(1996, pp. 299-300) where it i hown that an adequate analyi can be given tha
t doe not require a trinary ditinction
in naality.
225
ditinctive feature in their claificatory function can have only two value,
thi fact i explained. In other
cae, uch a with vowel height, the binary nature of the feature
predict that ome vowel ytem are
more common and natural than other. (In ome current work on feature
, ome feature have only one
value, but the feature may be abent altogether. The feature i called a privati
ve feature. If the feature may
be plu or minu, it i called equipollent.)
In their function of decribing the phonetic reality, the feature can
have calar value, although thi
ytem ha not been propoed in any detail. For example, the firt 

in abeto can be pronounced with

varying degree of voicing, depending on tyle, peed, dialect, etc. It i not 
ufficient to decribe with only
[+voice] or [-voice]. If we ue a cale from 0 to 1 for voicing at the phonetic
level, then we can give a more
precie decription of thee fact. The  can be [.1 voice] (with li
ttle voicing), [.6 voice] (with more
voicing), or perhap [.9 voice] (with ignificant voicing). At the phon
etic level one mut alo indicate
variou grade of naalization in vowel.
After the firt et of univeral ditinctive feature wa propoed, everal othe
r et have been uggeted.
Some have been acoutically baed,
c
and other articulatorily baed, which we generally follow here.
d
Although there i till ome diagreement over the et of univeral f
eature and in the definition of
thee feature, we preent below the feature that are among the currently mot
ued and needed.
Major Cla Feature
The feature [onorant] i of primary importance becaue it divide ound into t
wo major clae which are
referred to repeatedly: the cla of onorant and the cla of obtruent.
e
onorant [on] Sonorant are the ound that are produced in uch a way that th
ere i reonance in the oral
cavity. The ound which are not onorant are called obtruent.
The onorant include the vowel, emivowel, liquid, and naal.
f
Oral top, fricative, and affricate
are obtruent.
Manner Feature
With the following manner feature we can define other important clae of oun
d.
cononantal
[con]
Cononantal ound are produced with at leat mo
derate contriction of the
upralaryngeal vocal tract (i.e., that portion of the vocal tract that
i above the larynx). The only
noncononantal ound are the vowel, the glide (uch a [j]), [h], and glottal
top [?].
c

See Roman Jakobon, Gunnar Fant and Morri Halle

to peech analyi, Cambridge, Ma.,
MIT Pre.
d
Noam Chomky and Morri Halle (1968) The ound pattern of
rper & Row.
e
The feature [yllabic] ha been ued in the pat
feature to ditinguih between yllabic and
nonyllabic ound, [i] and [j], for example. Thi feature
nce of a ound in the yllable. There are
cae in which a cononant can be [+yllabic] phonetically
m). Thi feature ha been dipened with
in the verion of phonological theory that we are adopting
repreentation of word with yllable
tructure, CV keleta, etc.
f

(1952) Preliminarie

Englih, New York, Ha

a

a major cla

refer to the promine

(e.g. the /m/ in pri
here by enriching the

In ome language the laryngeal eem to group with thee ound, and for that
reaon are conidered onorant by
ome linguit.
226
continuant
[cont]
A ound i [+continuant] if there i not complet
e blockage of the air tream at the
primary contriction in the vocal tract. Fricative, emivowel (glide)
, and vowel are
[+continuant], wherea top (oral or naal) and affricate are [continu
ant]. Liquid are not a
homogeneou cla with repect to thi feature. It ha been claimed that vibrant
 (flap, tap, and
trill) are continuant, while [I] i not, and evidence from variou language 
eem to upport thi.
g
One piece of evidence come from Spanih: after naal and [I], the voiced coron
al obtruent i a
top (example: don[d]e, bal[d]e), wherea following vowel, fricative,
and flap r, they are
fricative (example: lo[]o, de[]e, bar[]a). If we aume that the featu
re [+continuant] i
preading, the flap r i [+continuant], but l i not. Another piece of evidence
come from Englih.
Conider the word attitude, where the firt t i a flap in American Englih. The
t in particle i alo
flapped, but the t in altitude i not. The r i [+continuant] like the vowel and
provide the context
for flapping, but the l doe not.
naal
[na]
If air leave through the naal cavity, the ound i
claified a [+naal]. Naal onorant
cononant and naalized vowel are [+naal].
voice [voi] The feature [voice] i ued to indicate the preence or abence of
vibration of the vocal cord.
h
g
h

Chomky and Halle (1968) The ound pattern of Englih.

There are other feature ued to decribe le common ound. We do not go into
thee detail here.
Other laryngeal feature beide [voice] are alo ued to decribe pho
nologically ignificant modification (ee
Stephen Anderon [1975] The organization of phonology, New York, Academic Pre,
p. 301 and Noam Chomky and
Morri Halle [1968] The ound pattern of language, New York, Harper and Row, p.
321): [pread glotti] (the glotti
i open, permitting the air to ecape but inhibiting the vibration of the vocal
cord), [contricted glotti] (the arytenoid
cartilage are kept tene o that the vocal cord can vibrate only a
t one end), and [heightened ubglottal preure].
Thee feature can be ued a follow:
[pread glotti] [contricted glotti] [voice] [heightened u
bglottal preure]
voiced ound +
voicele obtruent +
voicele vowel +
h + +
apirated cononant + +
glottal top +

breathy voiced vowel + +

murmured cononant + +
glottalized cononant + +
laryngealized (creaky) ound + +
imploive cononant + +/
The click, which are produced with ingreive velar air, are claified a [+ve
lar uction] by Stephen Anderon (1975)
The organization of phonology. The feature [glottal uction] ha alo
been propoed to ditinguih between the
glottalized cononant and the imploive cononant.
227
Place of Articulation for Cononant
Major Place Feature
In current feature theory, three place feature node (a oppoed to terminal fea
ture) are poited to account
for the major point of articulation of cononant: labial, coronal, and doral.
Labial
The ound which are produced uing the lower lip have a la
bial feature node in their feature
compoition. Thi include bilabial, labiodental, and rounded cononant.
Coronal The ound which are produced uing the front (or crown) of the tong
ue have a coronal feature
node. Such ound include interdental, dental, alveolar, and alveopalatal conon
ant.
Doral
The cononant which are produced by activating the body (dor
um) of the tongue have a doral
feature node. They include palatal, velar and uvular cononant.
Thee three feature node are ufficient to differentiate between the
mot common point of
articulation:
[Labial] [Coronal] [Doral]
bilabial, interdental, palatal,
labiodental dental, velar,
alveolar, alveopalatal uvular
Thee major place feature node make certain prediction. Firt, it i
common for a language to have
either a bilabial fricative or a labiodental fricative, but much le common to
have both. It i common for a
language to have a dental top or an alveolar top, but le common for one to h
ave both. In other word,
the feature are deigned to reflect general tendencie in language.
Feature Dependent on Place Feature
When the major place feature node cannot adequately ditinguih betwee
n the place of articulation
that a language utilize, other feature mut be invoked. For example, the featu
re [trident] could be ued to
ditinguih between interdental fricative and dental or alveolar fricative (e
e below). Other feature which
are ued to fine tune the point of articulation are given below. Some (or all) o
f thee feature are dependent
on one of the major place feature node. That i, only [labial] ound can be [r
ound], only [coronal] ound
can be [anterior], etc.
Feature Dependent On Labial
round [rnd] When the lip are rounded in the articulation (either a the prima
ry articulation (e.g. o) or a a
econdary articulation (e.g. k:), the feature [round] i preent.
Feature Dependent on Coronal
ditributed
[dit]
If the tongue tip i ued, the ound i [-dit
ributed] (apical articulation); if the tongue
blade i ued, it i [+ditributed] ince more of the urface i ue

d in the articulation (laminal

articulation).
anterior
[ant]
The feature [anterior], a currently ued, ditingui
he between dental and alveolar
([+anterior]) and alveopalatal and retroflex cononant ([-anterior]).
lateral [lat] If one or both ide of the midection of the tongue are lowered
, permitting the air to ecape in
that fahion, the ound i [+lateral].
i

Thee feature were deigned to account for phonological oppoition, not nece
arily all phonetic difference. In thee
matter, it i important to keep abreat of the latet reearch; thi textbook i
 not attempting to do thi.
i
Patricia Shaw ha argued (colloquium, Univerity of Arizona, December
1988) that the interdependency between
lateral and the coronal node may be formalized in a different way, y
ielding deirable reult in the decription of
certain language. See alo Juliette Blevin (1994) A place for lateral in the
feature geometry, Journal of Linguitic
30:301-48.
228
trident [trd] The trident ound are fricative or affricate which are cha
racterized by more noie than
other fricative or affricate.
j
[trd]
interdental
0
dental/alveolar
+  t
alveopalatal
+ j tj
Feature Dependent on Doral
high [hi]
If the tongue body i raied in the production of a ou
nd, the ound i [+high]. The [+high]
cononant are the palatal and the velar,
k
while the uvular are [-high].
back [bac]
If the tongue body i retracted, the ound i [+back].
Velar and uvular are [+back], while
palatal are [-back]. Palatalized ound have a doral node which i pecified a
 [-back].
Thee feature group the cononant in the following way:
Labial Sound: [round]
bilabial
p
rounded bilabial
+ p
labiodental
I
Coronal Sound: [anterior]
interdental
+ 0
dental
+ 
alveolar
+ 
alveopalatal
j
Doral Sound: [high] [back]
palatal
+
j
velar
+
+ k
uvular

+ q
Place of Articulation for Vowel
The following feature are ued to ditinguih baic vowel qualitie:
back The central and back vowel are [+back], while the front vowel are [-back

].
Front Central
Back
[bac]

+
+
The reaon for making the plit thi way i that it i not important phonologica
lly whether a vowel i central
or back phonetically ince other feature (uch a [round]) can alway
be ued to ditinguih between
central and back vowel.
high and low
The tongue body i lowered in the production of a [+
low] ound. The feature [high] and
[low] are ued in combination to define three vowel height.
j

In previou work, the feature [trident] wa ued to ditinguih bil

abial and labiodental articulation. The feature
[ditributed] ha alo been propoed to make uch ditinction. If the contrict
ion involve more area, a in the bilabial
articulation, the ound i [+ditributed]. Dental and alveolar cononant, whic
h are not commonly found in contrat in
a ingle language, uually can be ditinguihed with the feature [dit
ributed] if there i a ditinction between apical
(ue of tongue tip) ([-ditributed]) and laminal (ue of tongue blade)
([+ditributed]) articulation. Otherwie, the
appropriate feature characterization i unclear. The retroflex articulati
on are characterized a [-ditributed]. For
further general dicuion of feature uch a [ditributed], ee Jacqu
e Durand (1990) Generative and non-linear
phonology, London, Longman.
k
Thi ue of [high] only for doral cononant i different from that in earlier
generative theory.
229
[high] [low]
Cloe and near-cloe
+
i u
Cloe-mid, mid, and open-mid

c o
Near-open, open

+ u a
Round vowel are characterized a [+round]. It i much more common fo
r a cloe back vowel to be
rounded than unrounded. In ome language, however, cloe back unrounded vowel
are alo found.
The naalized vowel are decribed with the feature [naal] that ha already bee
n mentioned.
Other vowel modification
ATR
The feature [advanced tongue root] ha been propoed for diting
uihing between cloe ([+ATR])
and near-cloe vowel ([-ATR]), and between cloe-mid ([+ATR]) and open-mid ([-A
TR]) vowel.
Thi feature replace the feature [tene] that ha been ued in ome work to ma
ke thi ditinction.
In the view of ome phonologit, thi ue of the feature [ATR] i not alway ju
tifiable.
Natural Clae
One can peak of natural clae in two way: formally and functionally. Formall
y, the number of feature
that are needed to define a natural cla i alway le than the n
umber of feature that are needed to
decribe any of it member. And the fewer the feature, the larger the cla
. The mot intereting natural
clae are thoe which function a uch in the phonology of a language. That i
, a will be een repeatedly,

a certain group of ound i ingled out by ome phonological rule.

Cla Feature and/or Node
p t k b d g m n [-continuant]
b d g m n [-continuant], [+voice]
b d g [-onorant], [-continuant], [+voice]
p t k [-continuant], [-voice]
 z j tj d [Coronal], [+trident]
Feature Organization
Recent work ha argued that feature are organized hierarchically.
l
That i, rather than a repreentation
uch a hown below for a ound like [p], a wa typical of earlier generative t
heory,
+con
-on
-cont
-na
-lat
-trd
-cor
etc.
l

George N. Clement (1985) The geometry of phonological feature, Phonology Year

book 2:223-250; Elizabeth Sagey
(1986) The repreentation of feature and relation in nonlinear phonology, Ph.D
. Diertation, MIT; John McCarthy
(to appear) Feature geometry and dependency, articulatory organizationfrom
phonology to peech ignal, O.
Fujimura, ed., Bael, S. Karger. Alo ee G. N. Clement and Elizabeth V. Hume (
1995) The internal organization of
peech ound, in John A. Goldmith, ed., The handbook of phonological
theory, Blackwell, Cambridge, Ma. and
Oxford, pp. 245-306. The important innovation of the latter have not been incor
porated into thi text yet.
230
the repreentation hould be omething along the following line, with
ome feature grouped together
under certain node:
C
( = [p] )

[-voice]
[-continuant]
Place
Labial
Thi organization of feature, known a feature geometry,
i baed o
n a hierarchical view of feature
organization which i more fully illutrated below.
m
C/V
Root Node

Tonal Node
Laryngeal Node

(Regiter, Mode)
n
[hi] [Hi]
[voice]

Supralaryngeal Node
[onorant]
[cononantal]
[continuant]
[naal]
Place Node

Labial Node
o

[round]
Coronal Node

Doral Node
[trident]
[anterior]
[ditributed]

Tongue Root Node

[ATR]

[back]

[low]
[lateral]
[high]
The eential claim i that certain feature may work together in bundle, not
alway individually. See,
for example, the dicuion of place aimilation in chapter 11.
m

The organization hown here i an adaptation of variou propoal. Where each f

eature appear i a theoretical iue.
It i more important to know why a feature might appear in a certain place than
where a given feature appear.
n
Thi i one view of tone feature, baed on Keith Snider (1990) Tonal uptep in
Krachi: evidence for a regiter tier,
Language 66:453-474.
o
If the ditinction between labial and labiodental articulation i to be made u
ing the feature [trident] or the feature
[ditributed], thi hierarchy mut be revied.
231
Appendix B - Orthography Deign
One of the reaon for which phonological analyi i done i to pre
pare to develop a practical writing
ytem (orthography) that will be ued by native peaker of a particular langua
ge. Many factor enter into
the creation of an orthographic ytemome of which are directly related to the p
honological analyi and
many of which are not. In the following ection ome of thee factor are dicu
ed.
For thoe of u who have had experience with only the Roman alphabet, it i impo
rtant to realize that
an adequate writing ytem doe not have to ue thi alphabet. If an orthography
were being developed for a
new language found in Greece, it would be reaonable to expect that
an orthography baed on the Greek
alphabet would be ued. But we mut think even further than thi. Some language
ue a yllabary, where
each ymbol repreent a yllable rather than a ingle phoneme. Some
orthographie, uch a that of
Chinee, are really not baed on the ound ytem at all. (Thi i
not o trange, really; it i alo true of
number like 8 and ymbol uch a & and $.) We cannot take the time here to dicu a
dvantage or
diadvantage which are found in each of thee ytem (including the alphabetic
approach), but point out
that flexibility mut be exercied in thi area. The dicuion that
follow center around an alphabetic
approach, however.
Linguitic factor
There are two baic view with repect to the viual repreentation of utteran
ce. One, which we will call

the phoneme-baed view, emphaize the importance of a fairly traightforward bi

directional ound-to-letter
correpondence.
By bidirectional, we mean that the relationhip i X Y. For example, the letter
k would be used to
represent the voiceless velar stop phoneme and nothing else, and the voiceless v
elar stop phoneme would be
always be written k. English does not have an orthography with this bidirectiona
l correspondence, as is well
known. The sound [k] is sometimes written c, (as in cool), sometimes as cc (as
in hiccough), sometimes as
k (as in keep), sometimes as ck (as in pick), sometimes as ch (as in school), an
d who knows what else.
a
And
these symbols sometimes are used for other sounds. The letter c is s
ometimes pronounced [s] as in ice.
This lack of one
to
one bidirectional correspondence is arguably part of the pro
blem of literacy in English
since children cannot automatically pronounce [k] and know that they should writ
e c, nor can they see c and
automatically know that they should pronounce it as [k].
Phoneme
based orthography
The following examples from Seri show that the orthography for that language is
(by and large) a phoneme

based orthography, although the practical orthography uses symbols which are not
those of the International
Phonetic Association, of course.
Phonemic
representation
Orthography
stone
1ot
hat
woman
kmoom
cmaam
roadrunner
joop
zaap
your hair
miIit
milit
ironwood
1n
heen
Orthographie of many major language violate the phoneme-baed view ince they
have long hitorie
and often there are no mechanim in place for regular updating of 
pelling convention. Examine the
following data from Spanih (general Wetern Hemiphere dialect, pecifi
cally Mexican) and determine
whether the pelling i trictly phonemic.
a

Name, uch a Iraq, add further ymbol. Labialized k (or /kw/, depending on t
he analyi) i written qu.
232
Phonemic
Repreentation

Orthography
floor
pio
pio
zone
ono
zona
ky
jcIo
cielo
ice
jcIo
hielo
plater
jco
yeo
hand
mono
mano
brother
crmono
hermano
cheee
kco
queo
cow
boko
vaca
drink
bcbc
bebe
live
bibc
vive
Although Spanih i often mentioned a a language which ha a reaona
ble orthography, even a
phoneme-baed one, it i clear upon examination that it doe not foll
ow thi principle with any trictne
(epecially a it i poken by the majority). A the preceding example how, /b
/ i ometime repreented
a b and ometime a v; /j/ i repreented by both
i and y.
b
H i ued at time although it i not
pronounced at all. // i ometime written a c, ometime a , and ometime
a z. /k/ i both c and qu
(not to mention k in a few word like kilograma).
A phoneme-baed orthography eem relatively traightforward, and in many (if no
t mot) cae, newly
orthographie have been uccefully developed baed on thi principle.
Unfortunately, there are ome
problem. A phoneme-baed orthography require a phonemic analyi, and hence o
me clear idea of what
a phoneme i. Thi book ha attempted to how ome of the complexity
of phonological analyi, and it
hould be evident that the notion of phoneme i not entirely traightforward.
Conider two example from Seri. Firt, the phoneme /m/ aimilate to
the point of articulation of a
following cononant when it occur in the coda of an untreed yllable. So [p]
i an allophone of /m/ (not
of /n/ ever, a a matter of fact), and preumably all word that ar

e pronounced with [p] hould be written

with m.
That work for 99% of the cae, but there are ome peci
al word. Firt, there i a word like
[ ktoopk], which mean ome kind of duck, which inexplicably ha a ve
lar naal even in a treed
yllable. Thi word, preumably onomatopoetic, cannot be pelled phonemi
cally. Second, there are a
number of demontrative that have the underling form {1im-kop}, {1imkix}, {1im-kom} and {1imkoi}, which have velar naal phonetically even though the firt yllable i tr
eed (compare {1im-intika}
which ha [m]). Thee word alo cannot be written phonemically.
Another example from Seri i alo illutrative of a problem. Vowel le
ngth i contrative in the
language, but only in treed yllable. In other yllable, one can hear hort
and (very) long vowel, but
thi length i not contrative. Thu a word uch a {1pm} white-tailed deer i pho
netically [ 1p::m].
We analyze thi word a being /1pm/ phonemically.
c
The complication arie in a mall detail about the
lengthening rule. Suffix vowel are never lengthened. Since phonemic re
preentation and practical
orthographie typically do not include morphological information in them, there
are a very few ituation in
which the phonemically-written word doe not give enough information about how t
o pronounced it. If you
b

If one wanted to ay that the phonemic repreentation of [j] hould be /i/, the
point i till true.
c
An alternative phonemic analyi would claim that the lengthened vowel in the 
econd yllable i alo a long vowel,
given that long vowel are phonemic. Thi would mean that hundred or
thouand of word have predictably long
vowel.
233
were to ee a Seri word written a patoc, you wouldnt know if it were a ingle mo
rpheme and pronounced
[pot:o:k], or a root plu uffix, pat-oc and pronounced [pot:ok] (with
hort o). Depite thi problem, the
current orthography doe not how vowel length in the econd yllable of uch wo
rd.
Morpheme-baed orthography
The econd approach to orthography, which we will label the morpheme-b
aed view, emphaize the
importance of a fairly uniform repreentation of each morpheme, regardl
e of the phonologicallyconditioned allomorph it may actually have. A very imple example: th
e firt morpheme of the word
grandon and granddaughter i pelled the ame even though in the econd word th
ere probably are not two
ditinct d  in that word. But no language that we know of follow thi view ent
irely, although ome do it
partially. The following example from Englih how that it orthograph
y i morpheme-baed to ome
extent. (The morpheme in quetion are underlined.)
Englih
cjn ane

d
niti
anity
ojn
ign
igneI
ignal
Iuz
loe
I5t
lot
The difference between a phoneme- and a morpheme-baed orthography how up when
a morpheme
ha two or more allomorph. Allomorphy (above the phonetic level) invariably mea
n different pelling in
a phoneme-baed orthography, but not in a morpheme-baed orthography.
Certain factor affect the appropriatene of a morpheme-baed approach in a giv
en cae. Phonological
change that are not regular would have to be indicated omehow. It would be ina
ppropriate, for example,
to pell the plural of leaf a leaf jut to maintain morpheme invariance. The v
oicing of the /f/ in the plural
i not part of the regular ound pattern of Englih, a we noted in chapter 5. T
he ue of a morpheme-baed
approach to orthography uually require more intruction. We return to thi i
ue below.
Thee two baic approache to orthographic repreentation will be comp
ared below with repect to
three feature of an optimal orthography propoed by Edward Klima: low
degree of arbitrarine,
expreivene, and economy.
e
Low degree of arbitrarine
Since the correlation of meaning to ound i already arbitrary, there i no rea
on to introduce an additional
apect of arbitrarine. That i, it eem reaonable for an orthography to refl
ect the phonological tructure
of morpheme. Thu the morpheme for a canine animal hould be repreented by om
e ymbol, uch a do-g ince peaker already know how the word i pronounced. A cloe correlation
between ound (at the
underlying, lexical, or urface level) and pelling therefore reult i
n a low degree of arbitrarine. One
might argue that anytime there i not a bidirectional correlation betw
een phonological repreentation and
orthographic repreentation, the language i diplaying ome degree of arbitrari
ne.
Arbitrarine in Englih orthography i illutrated by the ue of the letter a i
n ome monoyllabic word
to repreent the phoneme //, a in bad, wag, cat, but to repreent the phoneme /
o/ in other monoyllabic
word, a in wad, what, and watt.
a
// /o/
d

The final e of thi pelling i actually inconitent with the morpheme-baed a

pproach.
e
Edward Klima (1972) How alphabet might reflect language, in Language

by ear and by eye: the relationhip

between peech and reading, J. F. Kavanagh and I.G. Mattingly (ed.), Cambridge,
Ma.: MIT Pre, pp. 57-80.
234
Arbitarine in Spanih orthography i illutrated by the ue of the
letter h (which ha no
pronunciation), and the repreentation of /b/ by the letter v in ome word and
the letter b in other.
Arbitrarine i alo reflected by the ue of more than one orthographic repree
ntation per phoneme, a
in the ue of oa and o...e to repreent the ame ound, a in boat and hope.
oa o...e
/o/
Thi principle doe not imply that orthographie hould adhere to the ymbol th
at are ued by linguit
for phonetic trancription. For example, if long vowel are ditinctive, that l
ength could be indicated well
by a number of convention. Some common one are: doubling (aa), diacritic lett
er (ah, a...e), and diacritic
(a). We return to thi topic below.
Expreivene
Another characteritic of an optimal orthography i expreivene. When two wor
d uch a the verb lead
[lijd] and lead [lcd] have the ame orthographic repreentation, the orthography
i not a expreive a the
ound ytem. Homophone which are pelled differently, uch a meat,
mete, and meet illutrate the fact
that Englih orthography i more expreive than the (preent-day) oun
d ytemwhich might be
conidered a poitive feature in that emantic difference are thereby
reflected in the orthography. Thi i
achieved only at an increae in the degree of arbitrarine, however. (T
he Englih orthography i alo expreive in that it preerve written contrat which are maintained i
n only ome dialecte.g. Mary,
merry, marry.) A language may ytematize a way to increae expreivene, a i
n the ue of capital letter
(e.g., Bill v. bill) or accent (e.g., Spanih m more v. ma but). French orthogra
phy i expreive in
that it indicate plural with x or  even in context where it i
not pronounced, following a morphemebaed principle. Thu, le belle femme the beautiful women ha three occurrence
of , none of which i
pronounced in thi particular utterance.
Economy
A third feature of an optimal orthography i economy. The economy pri
nciple i that the orthography
hould not duplicate the predictable variation that the ound ytem h
a. Phoneme-baed and morphemebaed orthographie both hold to the economy principle, but they differ with re
pect to the degree to which
they hold to it. A phoneme-baed orthography would agree to economize
in the orthography by not
repreenting variation in a ound which i everywhere determined by i
t environment. Thu all of the
allophone of /t/ in Englihapirated, unreleaed, unapirated, glottalize
d, and flappedare repreented
by only one ymbol, t.
A morpheme-baed orthography would want to economize in the orthography
by not repreenting in

the orthography any change that i anywhere predicted by a regular phonological

rule. Thu uch pelling
a rotate [t] : rotation [6], paradigm [] : paradigmatic [g], column []: columnar
[n] are viewed a the
proper application of the economy principle, depite the difference in pronunci
ation. The reader learn to
aociate only one viual equence with each morpheme; the reading proce there
by avoid a trip through
the phonology. The maximizing of the economy principle come at the expene of e
xpreivene in ome
cae, however. For example, the pelling for [Jcjjjo] could be motiva
ted to be ratio by deriving it from
rate plu the uffix -io. The pelling of the word ratio and patio now conceal
their different pronunciation
ince Englih orthography, like mot orthographie, doe not include ymbol for
morpheme boundarie.
One ituation in the Seri orthography where the economy principle eem to be ap
propriately applied i
with certain allomorphy of the future prefix i-. The i often delete, even befo
re cononant, o many future
verb form begin with  followed by cononant. When  precede the a
lveopalatal fricative z, however, it
aimilate to it. Thi aimilation i apparently automatic and except
ionle, although there are no other
place in the language where the ituation arie. The future of the verb -zip i
 pronounced [jjip]. The mot
obviou choice for writing thi are zzip (writing each phoneme eparately), and
zip (preerving the viual
cue of  for future tene).
f
f

It will be intereting to ee what Seri writer do. If they ound out the word, p
reumably they will write zzip. If they
235
The economy principle, regardle of how it i applied, ugget that
an optimal orthography i not
characterized by unneceary variation in the hape of the morpheme.
Therefore it i conidered
advantageou if a morpheme ha the ame hape everywhere. There are way that th
i poitive feature can
be maintained without uing the ame ymbol for two different ound
(a t for /t/ and /j/, in rate : ratio
above). The ue of diacritic ymbol (like the ilent e of Englih) a
nd (true) diacritic are two uch way.
Conider the following infamou pair of word:
divine [divojn] divinity [diviniti:]
erene [oJijn] erenity [oJcniti:]
profane [pJoIcjn] profanity [pJoIniti:]
cone [kown] conical [konikoI]
The long i, e, and a, which are phonetically the diphthong [oj], [ij], and [cj] o
f the unuffixed form,
are indicated in the Englih orthography by the ilent e that occur at the end
of the word (functioning a a
diacritic ymbol).
g
Thee vowel occur hort (phonetically lax) in other form, uch a when uff
ixe are
added, but orthographically the root have nearly the ame hape a i
n the iolated form. In thi way

Englih follow the economy principle quite well. Compare thi approach with one
which would repreent
the morpheme very differently. If each phoneme had a eparate repreentation in
Englih orthography, the
word might written like divayn, divinitiy, eriyn, erenitiy, profeyn, profanit
iy. (Actually, who know how
the chwa would be written!)
Englih alo write certain very common morpheme with a morpheme-baed approach
. The pat tene
uffix i -ed whether it i pronounced [id] or [t] or [d] (heated, paed, rolle
d).
Ithmu Zapotec ha a ucceful practical orthography that i phonemebaed except in a few
ituation. One of thee i how it write the poeive prefix [j-].
Wherea the phoneme /j/ i uually
written xh, and the phoneme // i written x in thi language, the poeive pref
ix i written imply x when
it precede a cononant (in which poition there i no contrat poible). The r
eaon eem to be imply that
xt adequately ignal the phonetic value of [jt] and that the h in xht would be
uperfluou.
h
It i not hard to find example of how a language doe not follow
the variou principle dicued
above. Englih and French are often cited a having terrible pelling ytem, a
nd yet in ome way they are
very good. The Greek orthography wa not an optimal one in variou repect. Cer
tain cononant cluter,
which cannot be analyzed a ingle cononant, were written with pecial ymbol
: /p/ a and /ks/ as .
Because of this, man morphemes had alternating shapes unnecessaril . F
or e
ample, if the future suffi

{-s} was added to the root {pcmp-} send, the result was written as ra
thr than as . Thi
violate the economy principle ince a morpheme i written with more
than one hape unnecearily. The
phoneme /h/ (if it i correctly analyzed a uch) occurred only wordinitially and wa repreented by a
pecial mark above the firt vowel. While thi i a good idea, ince thee word
pattern in certain way like
vowel-initial word, the orthography alo indicated (by mean of another pecial
mark) when the word did
not begin with an /h/. Finally, Greek alo violated the economy principle
by having two hape for lower
cae //: one hape () wa ued word-internally, and the other () was usd word-fi
nall .
Comarison
You hav likl alrad com to th conclusion that an orthograh wh
ich is strictl honm-basd or
strictl morhm-basd is lss than idal. Clarl thr is rason to
qustion ithr of ths xtrm
ositions. It is gnrall agrd that at th initial stags of radi
ng and writing, lss ffort is rquird to
larn an orthograh that is honm-basd. (This is an imortant oint to rmm
br if schooling to tach

think futur, rsumabl th will writ szi.

g

Altrnativ sllings using diacritic lttrs or s mbols would b divihn, divin

iti and divn, diviniti.
h
This ma suggst that th wrong dcision was mad fort som ars
ago whn th currnt orthograh was
adotd. It might hav bn bttr to hav x rrsnt /j/ and xh r
rsnt //. Howvr, right or wrong, th currnt
orthograh is wll usd and thr is vr littl liklihood that it would b ch
angd at this oint.
236
rading and writing is scarc or not availabl.) Nvrthlss, it is
also claimd that a morhm-basd
orthograh allows for a gratr facilit in rading at latr stags. Aftr th
initial stags th radr cass
to sound out words and instad maks a form-to-maning connction. Comar th fol
lowing rrsntations of tlgrah / tlgrah .
Phonm-basd Morhm-basd
tlugraf
tlgraf
tulgrufi
tlgrafi
Whras th honm-basd rrsntation ma b asir for nw radrs, it rs
nts som disadvantags
for mor skilld radrs sinc it forcs thm to rcogniz vr sign
ificant honological altrnation, no
mattr how rgular.
i
W should also oint out that a honm-basd orthograh for English
would b difficult to
imlmnt in viw of th radicall diffring dialcts of English. Th
currntl usd orthograh dos not
distinguish btwn dialcts, whras a honm-basd on would of ncssit hav
 to rflct on crtain
dialct in th slling of ach word.
Othr honological factors
Th sound s stm itslf ma b imortant to considr if crtain sound
s air u in k honological
rocsss. Suos thr wr fiv vowls, both long and short. It would b ras
onabl to choos s mbols
for th long vowls which ar basd on th short vowls, rgardlss
of th actual hontics. Likwis, an
asiratd sris of stos is rasonabl rrsntd as a modification of th sim
l sris, ithr b digrah
or diacritic; a novl charactr is not rquird for ach distinct honm.
S llabl structur ma b vr rlvant. For xaml, th digrah th
for th honm /O/ will not b
aroriat in a languag that rgularl has consonant clustrs that ma includ
th clustr /th/, of cours. If
th honm // and th clustr /g/ both occur, thn on will not
automaticall roos using ng to
rrsnt th formr without at th sam tim considring how th lattr will b
rrsntd.
On ma also tak into considration th frqunc with which a honm occurs.
For xaml, Sri
has two kinds of latrals:
/I/ (voiclss fricativ) and /I/. Th v
oicd vrsion is xtrml rar, but th
voiclss on is not uncommon. Th dcision was mad to us lain l
to rrsnt th voiclss fricativ,
and an undrscord l for th rar on, and this has smd to work wll.
Morhm altrnations also motivatd th choic of c for /k/. Thr is an intima
t connction in th

languag btwn th honm /k/ and th squnc /ko/, as th formr is oftn
drivd from th lattr b
(non-automatic) ruls. So th words /1ooko/ hous and /1ookt/ houss ar writtn ha
aco and haact,
rsctivl . Th o-dirsis rsrvs visual similarit btwn th full o and
th labialization.
All languags rsumabl show som honological ffcts of fast sch or casual
sch, as oosd
to carful sch. In English, casual sch must b distinguishd fro
m informal sch as wll. Th
contraction of is not to isnt is a charactristic of informal sch, but informa
l sch ma b carfull or
slowl articulatd. And so English orthograh rrsnts th distinction b
twn is not and isnt. On th
othr hand, th alatalization which taks lac in squncs such as did ou, w
hr it coms out [ didj],
is charactristic of casual sch and vn of som lvls of carful
sch. This alatalization is not
rrsntd in English orthograh , vn though th chang from [d] to
[d] is from on honm to
anothr.
Ps cholinguistic factors
Th rvious sction outlind som facts that nd to b considrd w
hn dvloing an orthograh .
Thos facts wr rimaril basd on th linguistic dscrition of th
languag. It was shown that som
anal sis of morhm altrnations is a rrquisit sinc crating an
orthograh involvs mor than th
i

Thr ar two honological ruls surssd in th morhm-basd r

rsntation abov. On is th rul which
tnss final vowls. Th othr is th rul which rducs unstrssd vowls in c
rtain ositions, oftn to schwa, which
w rrsnt hr as bing an allohon of th sam vowl as in th word but.
237
choic of s mbols for th contrasting sounds. For this rason, th ma
ttr of crating an orthograh is
somthing which cannot siml b tossd into th hands of a novic,
whthr a nativ sakr of th
languag or not. In this sction w will brifl considr som s cholinguistic
factors that ar involvd in
crating an orthograh .
On imortant considration is th attitud of nativ sakrs to th
writtn form of thir languag,
sciall if thr xists an indtrminac in th linguistic anal sis
. Sakrs knowldg of thir
languag ma rovid valuabl inut to th orthograh cration rocss b hli
ng to answr crtain qustions, whthr dirctl or indirctl . Should a word brak b ut hr
?
How should this b writtn?
Thrfor, whil thr is no substitut for a carful honological anal sis of t
h languag, sakr raction
to roosd sllings nds to b carfull considrd. (Nvrthlss,
grat caution must b xrcisd in
this ara sinc man xtranous factors could b affcting th sakr s answr,
such as a dsir not to mak
th languag look at all diffrnt from th national languag.)
Anothr factor to considr is th imortanc of a articular honological fatur
 in th languag. This

is sciall rlvant whr strss or ton has onl a marginal contr

astiv function. Considr English, for
xaml, whr strss lacmnt contrasts such noun/vrb airs as cntrast / cont
rst and rmit / rmt.
Dsit this contrastiv function of strss in English, strss is not rrsnt
d in th orthograh . Likwis
in Sri, although strss is not comltl rdictabl, it is not r
rsntd in th orthograh . (An
altrnativ would b to rrsnt strss onl whr it is not rdictd b som
siml rul, as in Sanish.)
Somtims th honmic rincil has bn st asid at tims for orthograhis o
f minorit languags
whr th hontic forms corrsond to honmic forms in th national
languag. For xaml, in som
varitis of Mixtc, th honm /t/ is ronouncd with a nasal rlas, [t
n
], whn it rcds a nasalizd
vowl. Th nasal rlas is not honmic, but th salinc of this fatur and t
h rction of it as a ral n,
has ld nativ writrs to rfr writing it as tn. Similarl , in lan
guags whr voicd consonants ar
allohons of th voiclss ons, thr ma b rssur to writ both
variants diffrntl if th national
languag writs thm diffrntl . And finall , whn a grat dal of borrowing t
aks lac into a languag,
a non-honmic distinction in nativ words ma hav to b writtn bc
aus of th influnc of th
honmic distinctions in th borrowd words.
Th dcision to us a articular orthograhic s mbol rathr than anothr also in
volvs considration of
various facts, including distinguishabilit . Th honm /0/ in English
is rrsntd b th squnc th
rathr than an ntirl diffrnt s mbol. (This is onl ossibl bcaus th ho
nms /t/ and /h/ rarl occur
togthr, and thn onl in comounds, .g. outhous.) An altrnativ rrsntat
ion such as a d with a bar
through it would robabl b judgd lss acctabl sinc it is visuall so much
lik th standard d, and th
rrsntation & might b rjctd bcaus it is too bizarr. For sim
ilar rasons, th honm /u/ in a
languag might b rrsntd as (umlaut u) rathr than (umlaut i), sinc th la
ttr is distinguishd from
siml i onl b a singl dot.
On should not assum that an alhabtic s mbol must b basd on som
 hontic chart, such as th
IPA. IPA s mbols man nothing mor to a nw radr and writr than
an othr s mbol. Considr th
following wa s in which th honms /j/ and /j/ ar rrsntd in th followin
g languags:
/j/ /j/
English sh Portugus nh
Frnch ch Italian gn
Sri z Sanish
Mixtc x
Isthmus Zaotc xh
Grman sch
Sociolinguistic factors
Orthograhis ar not dsignd for linguists but for th sakrs of
th languag, sakrs who liv in
social grous and who invariabl hav diffrncs in sch habits. An orthogra

h that is dsignd to b
acctabl ovr a wid dialctal ara will b diffrnt from on that
is dsignd for a narrow linguistic
grou. Th broadr th linguistic grou, th lss likl it is that
a honm-basd orthograh will b
utilizabl. Th advantags of an orthograh that is mor morhm-bas
d (such as rmitting gratr
238
cross-communication) must b wighd, howvr, against th disadvantag
that much mor instruction in
th orthograhic s stm will robabl b rquird. If a linguisticall divrs s
ch communit is not abl
or willing to mak srious comromiss in th orthograhic rrsntation of th
ir sch, thr will not b
succss in ths attmts to bridg dialcts.
It is likl that sakrs of a minorit languag will also rad and writ a maj
orit or national languag.
It has oftn bn considrd advantagous if th s mbol-honm corrsondncs
in th two languags ar
as much in agrmnt as ossibl. This will hl a rson who is li
trat in on languag to mor asil
bcom litrat in th othr. Clos conformit ma actuall b rquird b th g
ovrnmnt of th countr .
In Mxico at on tim, th us of c and qu to rrsnt /k/ was i
n fact rquird. Such conformit is not
without disadvantags, howvr. As th rsult of orthograhic convntions in Sa
nish, th Sri rfix {k}
is slld c- bfor back vowls and qu- bfor front vowls. This i
s a comlication for Sri radrs, of
cours, but b larning it for Sri, th also larn it for Sanish, and vic v
rsa.
j
Th wa that words look in a dominant languag ma affct th convntions that s
akrs of a minorit
languag will acct. For xaml, in countris whr th major langua
g is not tonal or dos not writ
ton, it is going to b mor difficult to convinc sakrs of a minorit ton l
anguag that ton nds to b
writtn in thir languag. (And indd it is not alwa s th cas tha
t it must b. Linguists ma objct to
having ton omittd from a ractical orthograh , but sakrs will argu that t
h know what th tons ar
and so th dont want to bothr with thm. (Th sam kind of argumnt is mad for
strss, vowl lngth,
and othr contrasts. And th argumnt cannot b dismissd asil ; in fact, it i
s vr snsibl in som cass.)
Whil oftn th sakrs th minorit grou want thir writtn languag
 to look lik othr languags
hld in stm (which ma man th us of crtain s mbols, th avoidanc of
othrs, or th rfrnc to
hav short words), it is also tru that somtims th minorit grou
want thir languag to look diffrnt
from th national languag and will choos crtain s mbols as a sign of th diff
rnc. Som grous insid
th U.S. hav dlibratl chosn to adot orthograhis that ar unlik English
(and mor lik Sanish) in
ordr to distanc thmslvs culturall in this s mbolic wa . Govrnmnt olic
ma actuall insist on at
last crtain diffrncs btwn th national languag and minorit la
nguags. In Mxico toda , for

xaml, th rssur is just th oosit from rvious riods; k i
s a s mbol that Indian languags ar
xctd to hav to show that th ar diffrnt from Sanish.
Of cours, if th s mbols of th local languag and th national lan
guag ar at odds, thr ma b
roblms, sciall if thr is going to b massiv borrowing or cod-switching
tak lac in th writtn
languag. Sri dos us a fw s mbols which hav diffrnt valus from Sanish,
as shown blow, but ths
hav not bn a roblm. An of us who hav larnd to rad and wr
it mor than on languag hav
larnd to handl diffrnt s stms.
Practical factors
Som factors baring on th cration of an orthograh ar rlatd to such matt
rs as what is availabl on
rinting rsss and t writrs in th countr . Comutr solv this 
roblm onl for th rich and
j

Th c/qu issu is quit diffrnt from th v/b on of Sanish, sinc all radrs an
d writrs of Sanish larn how
to us c and qu corrctl . Th distribution is comltl rul-govrn
d. Howvr, th distribution of v and b is
arbitrar , and nw writrs of Sanish frquntl mak mistaks with ths lttr
s.
k
Actuall , it is oftn rtroflx, so th s mbol rhas should b .
l
Taxi, Taxco, Mxico, Xochimilco (som ronunciations), rsctivl .
S mbol Sri valu Sanish valu in Mxico
z
j
k
s
h
1
(null)
x
ks, s, x, j

l
f
I

c
239
comutr-litrat. Thr is no rason to introduc a hontic s mbol into t
h orthograh th honms
//, /j/, and /_/ ar rrsntd as , z, and x in Sri, for xaml, bcaus th
lattr s mbols ar asir to
tach and rroduc than th formr.
Othr Mattrs
Orthograh dsign is mor than just a mattr of choosing s mbols. A rathr knot
t roblm in man cass
is whr to mak word braks. Som morhms, such as rositions or ostositi
ons, ma b unstrssd
bcaus th ar not nouns or vrbs, and thrfor do not aar in
isolation (s chatr 8). In som
languags, ths ar writtn as sarat words, and in othrs th ar
 writtn as if th wr rfixs or
suffixs. Whil som ol (linguists and nativ radrs) rfr to f

ollow th rincil whn in doubt,

string it out (writ morhms as sarat words), this aarntl dos
not alwa s work. In othr cass,
thr aars to b a tndnc to writ all unstrssd morhms (whthr trul
affixs or not) attachd to
som othr word.
Comounds rsnt anothr kind of comlication, vn in English whr
thr ar no good ruls for
dciding whn to writ somthing as on long word, on h hnatd word, or two w
ords. All thr otions
ar usd. Othr languags hav comounds and similar issus will aris
. It is actuall a bit wors in Sri,
whr contrastiv vowl lngth is lost whn th s llabl is unstrssd. In Sri
comounds, it is th scond
art that rtains its lngth. So whn th word ctooml occurs as th
first art of a comound, it will b
ronouncd with a short vowl. But how should it b writtn?
Anothr roblm that cannot b ignord has to do with small (or grat) dialct d
iffrncs. How should
words b slld whn ol sa thm diffrntl ? English orthograh is vagu
nough that diffrncs
btwn th ronunciation of roof, rout, and crk, for xaml, ar siml ign
ord. Somtims, a slling
s stm fars bttr if it is slightl mor abstract, if it ls off
crtain honological ruls. For xaml,
Sri has a rul that changs a // to /k/ (th lattr is writtn c in Sri) whn
it immdiatl follows a k.
So whn th rfix /k/ is addd to a word such as oof if s/h arrivs, th rsul
t is [kkoo] (that
is, [k:o:], for most sakrs. Nvrthlss, som sakrs actuall ronounc t
h word as [koo].
In othr words, som sakrs assimilat th // to th /k/ of th rfix, and
som do not. Th orthograh
of Sri has ignord this rul and lts ol ronounc th squnc as th wis
h. Th word if s/h arrivs
to him/hr/it is slld coof and adquatl srvs both sub-dialcts.
Such solutions do not alwa s
work, but th ar worth considring.
Loanwords will alwa s b a difficult bcaus th rgularl introduc nw s
ounds and nw attrns
that ma b incororatd with difficult into th writing s stm. It
sms aroriat to anticiat ths
roblms, but how? For xaml, in Sri thr ar honms /t/ and /j/, and ths
 ma occur in th clustr
/tj/ at tims (not vr commonl ). Th national languag of Mxico, Sanish, has
an affricat /t j/ which is
writtn as ch. Sinc Sri dos not hav this affricat, but it has th hontica
ll similar squnc /tj/, on
can alwa s writ a nam such as Chla, but not with th sam s mbols as Sanish
(it would b Tzla in Sri
(with an undrlind l to show th voicing which is not common in S
ri latrals). Th Sri orthograh is
adquatl xrssiv, but th roblm is that such a siml thing as th writin
g of nams will b diffrnt
in th Sri orthograh from Sanish orthograh . At this oint it sms unavoid
abl sinc othrwis on
would hav to allow th honolog and th orthograh of th national
languag to dictat how th
minorit languag will b writtn.
Political factors

Of cours on wants to think about all of th issus laid out abov
, and t in th nd th languag
communit itslf has to mak th dcision. It will do this in on or mor wa s.
First, th communit ma hav alrad dcidd that crtain s mbols or convntion
s will b followd,
sciall if thr is an acadm of som sort for th languag or languag famil
. This acadm ma or
ma not hav mad what on ma considr to b th bst choic, but
th dcisions ar a ralit not to b
ignord nor disaragd. Ths ma b influncd b a long histor of
govrnmnt olic , of attituds, of
inut b othr linguists, tc. Th ma not b immutabl, but th ar ral. Ev
n if thos dcisions lad to
vr unsatisfactor rsults, th still cannot b ignord. (For xaml, th dc
ision ma hav bn mad to
rrsnt two common honms with th sam s mbol.)
240
Scond, th communit ma not hav mad a dcision, but it will considr its ot
ions. It ma considr
that th dcisions b outsidrs (including thos of rlatd languags,
thos of govrnmnt-sonsord
committs, and thos of linguists) ar irrlvant. Th olitical landm
ins ar ral as on intracts with
ol who ar making dcisions and with thos who ar going to b
actuall affctd b thm. In ths
cass, it is imortant to undrstand how communitis mak dcisions in that cult
ur.
Third, th communit ma nominall acct th convntions that ar roosd b o
utsidrs or insidrs
(whthr wll-rasond or not), but dmonstrat that th ar not adquat b i
thr not larning thm, b
not using thm, or b changing thm ovr tim. What is rlvant in th long run
is what th writrs in th
communit us in futur gnrations and not what is roosd b linguists or com
mitts.
241
Tr it for ourslf
Considr th following honm list (basd on Zaotc languag of Mxi
co) and mak roosals for an
orthograh of this languag basd on thr diffrnt scnarios. Ths ar roo
sals that ou ar making to
sakrs of th languag for thir considration, so ou will nd to hav raso
ns for our suggstions.
Th s llabl onst ma hav two consonants and th nuclus ma hav two sgmnts
(if two, on of thm is
a glid). (S saml words blow.) Th kinds of comlx which occur ar /sC/, /
jC/, or /NC/ (whr N =
nasal, C = consonant).
First, roos an orthograh that might b aroriat for this langu
ag if English wr th languag of
ducation and communication for th countr in which it is found. Sc
ond, roos an orthograh that
might b aroriat if Frnch wr th languag of ducation. (If ou dont know
Frnch, choos anothr
major languag that ou do know somthing about, rhas Sanish.) Thi
rd, roos an orthograh that
dos not tak th national languag into considration, but which is
constraind b th non-ngotiabl
s mbols that a govrnmnt committ has imosd (as shown). Discuss our roosa
ls. For xaml, if ou

roos that ch should rrsnt /tj/ in th English-influncd orthograh , tl

l wh .
211
For ach orthograh , show how th following words would b writtn:
1. so music
(vowl with crak voic)
2. wogo rat
3. gi fir
4. bi w fla
5. gjc1 tortilla (vowl with glottalizd articulation)
6. jjo1 custom of
7. mbjo1jo clown
8. scj comutr
9. rusidubi wars out
10. nd hard driv crash
242
Phonm English Frnch committ


b
b
t
t
d
d
tj
d
k
k
g
g
s
s
z
z
j
x

xh
m
m
n
n


I
l
r
r
j

w
w
o
a
c

i
i
o
o

u
u

crak voic on
vowls
glottalization on
vowls (honticall
oftn lik a vowl
followd b glottal
sto)
a
243
211
Orthograh dsign xrcis
English-orintd orthograh (on roosal, taking English orthograh sriousl
, but rcognizing
that th rsults ma look odd at first):
Unlss thr is a rason not to, rsumabl th following honms would b writ
tn as shown bcaus of
th dirct corrsondnc to th most common English slling for ths honms
.
Phonm
 b t d tj d k g s z j m n I j w
Orthograh  b t d ch j k g s z sh m n l w
Mor difficult cass: Sinc English dos not hav a siml wa to rrsnt th
sound // sinc this sound
is not a honm of English, on might roos somthing analogous to th digra
h usd for /j/, naml zh.
Similarl , English dos not hav th alatal nasal as a honm. Prhas n woul
d b good to us. Th fla
is a roblm bcaus in English this is an allohon of /t/, and th r of Englis
h is quit diffrnt.
Nvrthlss, I would roos using r for th fla.
Th vowls ar difficult. Sinc th s llabls of this languag ar all on, it
would b good to look at
similar s llabls in English. Th corrsondncs ar oftn digrahs in English,
howvr.
/i/ , a s, a
// a ,  da , h (bsids igh in nigh)
/o/ o, o go, ho (bsid ow in low)
/u/ oo, u too, du (bsids o in do and o in sho)
/o/ (dosnt occur in on s llabls in monos llabic words xct for
intrjctions, as in ha! and bah!, and a fw words lik ma)
// (dosnt occur in on s llabls in monos llabic words, but othrwis it is
writtn as a, as in ha )
Thos facts dont mak th dcisions vr as . Th // is sciall difficult b
caus th squnc /j/ is
also ossibl in th languag, and so // cannot (?) b writtn a sinc /j/ wo
uld com out as a . It might
b nic just to roos th rgular fiv vowls lus for th sixth vowl. But th
at would b a co-out for
this xrcis that is suosd to mak us rall think about altrnativs basd
on English slling, so I
would roos:
Phonm i  o u
o
Orthograh   o oo ah a
A crak vowl could b rrsntd b , and a glottalizd vowl b .
244

1. so music sahah
2. wogo rat wahgah
3. gi fir g
4. bi w fla bw
5. gjc1 tortilla g 
6. jjo1 custom of sh ah
7. mbjo1jo clown mb osho
8. scj comutr s
9. rusidubi wars out roosdoob
10. nd hard driv crash nja
Frnch-orintd orthograh (on roosal)
Unlss thr is a rason not to, rsumabl th following honms would b writ
tn as shown bcaus of
th dirct corrsondnc to th most common Frnch slling for ths honms.
Phonm
I
 b t d k g s j m n 
Orthograh  b t d c, qu g, gu s j ch m n gn l
Of th abov cass, th us of c,qu for /k/ and g, gu for /g/ ar otntial rob
lms (oliticall , but not
linguisticall ) sinc th do carr ovr som comlication from Frnch orthogra
h (c bfor back vowls,
qu bfor front vowls).
Frnch slls /w/ as hu and /j/ as hi, but ths sllings would b roblmatic
in Zaotc bcaus th glids
ma occur aftr consonants and also at th nd of s llabls. Thr would b no a
utomatic roblm in
slling ths honms siml as u and i, howvr.
Sinc /s/ and /z/ ar rrsntd b comlicatd convntions in Frnch (somtim
s s, somtims ss, tc.),
w will roos that z b usd to rrsnt [z] in Zaotc, and that siml s b
usd to rsnt /s/.
Mor difficult cass: Sinc Frnch dos not hav th honms /tj/ and /d/, on
might roos th
digrahs tch and dj, rsctivl , basd on th rrsntations of th fricativ
s.
Th fla could b rrsntd as r, although it will hav a rathr diffrnt ho
ntic valu from Frnch.
For th vowls, on might roos th following, basd on th most common Frnch
quivalnts.
/i/ i
/c/ ai (or rhas siml , dsit th diffrnt hontic valu)
/o/ au (or rhas o, dsit th diffrnt hontic valu)
/u/ ou
/o/ a
// a (a difficult dcision bcaus this sound dos not occur in on s llabls i
n Frnch)
A crak vowl could b rrsntd b , and a glottalizd vowl b .
1. so music saa
2. wogo rat uaga
3. gi fir gui
4. bi w fla biiu
245

5.
6.

gjc1 tortilla gui

jjo1 custom of chia

7.

mbjo1jo clown mbiocho

8. scj comutr si
9. rusidubi wars out rousidoubi
10. nd hard driv crash ndja
Sanish-orintd orthograh (on roosal)
Unlss thr is a rason not to, rsumabl th following honms would b writ
tn as shown bcaus of
th dirct corrsondnc to th most common Sanish slling for ths honms
.
Phonm
I r
 b t d k g tj s m n 
Orthograh  b t d c, qu g, gu ch s m n l r
Of th abov cass, th us of c,qu for /k/ and g, gu for /g/ ar otntial rob
lms (oliticall , but not
linguisticall ) sinc th do carr ovr som comlication from Sanish orthogra
h (c bfor back vowls,
qu bfor front vowls).
Sanish slls /w/ as hu but this slling would b roblmatic in Zaotc bcau
s th glids ma occur
aftr consonants and also at th nd of s llabls. Thr would b no automatic 
roblm in slling /w/
siml as u, howvr. In fact, an analogous dcision could b mad for /j/ vn
though on could us .
Sinc Sanish dos not hav /z/, som roosal nds to b mad. Th s mbol z in
Sanish orthograh
rrsnts onl [s] in th Amricas. Dsit this roblm and otntial comlica
tion, w will roos that z
b usd to rrsnt [z] in Zaotc.
Mor difficult cass: Sinc Sanish dos not hav th honms /j/, // and /d/, o
n might roos sh, zh
and dzh (novl uss of ths combinations), or xh, x, and dx (variations on on
hontic valu of x in
Mxican Sanish influncd b indignous lac nams), or x, ll, and dx (diffri
ng rimaril in th us of ll
which has th ronunciation [] in rgional Sanish). Each of ths otions has b
n adotd and rfrrd
b diffrnt Zaotc languag grous. For th sak of this xrcis, w will ado
t ll.
For th vowls, on might roos th following, basd on th most common Frnch
quivalnts.
/i/ i
/c/ 
/o/ o
/u/ u
/o/ a
// (utilizing th dirsis which is usd lswhr in Sanish)
A crak vowl could b rrsntd b , and a glottalizd vowl b .
246

1. so music saa
2. wogo rat uaga
3. gi fir gui
4. bi w fla biiu
5. gjc1 tortilla gui
(Problm: how would [guj] b distinguishd from [gi]? Undr th roosals hr,
ths would both b
slld gui, wouldnt th ? This roblm would suggst that somthing has to b ch
angd.)

jjo1 custom of xia
mbjo1jo clown mbioxo
8. scj comutr si
9. rusidubi wars out rusidubi
10. nd hard driv crash nll
Committ-influncd orthograh
Th following consonant-lttr corrsondncs ar givn b th committs roos
al, and th will b
followd.
Phonm
 b t d k g s z j m n
 I r j w
Orthograh  b t d k g s z x xh m n l r w
Th committ has not bn a roosal for th affricats /tj/ and /d/. Th roo
sal might b mad to us a
digrah and trigrah basd on th corrsonding stos and fricativs: tx and dx
h. Th trigrah might b
simlifid b omitting th rdundant h, laving dx. Ths would nd to b tst
d, of cours.
Th vowl s mbols ar all rovidd xct for //. Thr ar a numbr of roosals
that might b mad,
including and a. W will suggst th lattr (to avoid th diacritic).
For crak voic, w might roos using  or . If w had // with crak voic,
th lattr would
rsult in aa, which might b dmd lss asthticall lasing than aa. Nv
rthlss, in th absnc
of good rasons not to, w will roos .
1. so music saa
2. wogo rat waga
3. gi fir gi
4. bi w fla biiw
5. gjc1 tortilla g 
6. jjo1 custom of xia
7. mbjo1jo clown mb oxo
8. scj comutr s
9. rusidubi wars out rusidubi
10. nd hard driv crash ndxa
247
Andix C - Phonolog Writ-us
Th rsults of linguistic rsarch should b ublishd in an aroriat lac,
such as a rofssional journal.
In man cass, straight dscritions of th honolog of a languag a
r not aroriat for such journals,
howvr. Man radrs of journals ar mor intrstd in thortical i
ssus and data that bar on thm.
Nvrthlss, thr is a lac for good dscritiv work if its rs
ntation is adatd to th audinc and,
whn aroriat, som attntion is aid to issus of currnt concrn. In this a
ndix w rsnt an outlin
of toics which ar basic to an honological dscrition. Th outlin
is not ncssaril aroriat for a
ublishd articl, but th information it covrs should b includd somwhr. I
t is also crtainl ossibl
that othr toics must b includd.
In ordr to mak th dscrition usful to a wid audinc, thor -a
rticular formalism should b
downla d, or at last accomanid b a carful ros dscrition. In
addition, ach claim should b
backd b a carfull rsntd arra of data. Thr things ar ssntial: good
writing, good data, and good
argumntation.
6.
7.

Th following xamls ar somwhat mor sgmnt-orintd than fatur-orintd,
sinc at this lvl
of dscrition w ar aiming at a mor gnral audinc. In som cass, this ma
lad to lss than dsirabl
rsults, in which cas som additional discussion is ncssar .
Invntor of Phonms
Includ a list of th honms of th languag, rhas in chart for
m. Us standard linguistic s mbols in
this list rathr than thos which ar usd b litratur for nativ sakrs (tj
rathr than ch, for xaml).
Rgardlss of what ou sa , ol a attntion to how words ar s
lld. You dont want to miscommunicat.
If a roosd honm occurs onl in loanwords, or lss than a dozn morhms,
or idohons, list it
saratl or mark it clarl .
a
Saml Consonant Chart
Consonants
 t k 1 (Oral Stos)
I s j x (Fricativs)
m n (Nasal Stos)
I (Liquid)
r occurs in loanwords from Sanish.
Saml owl Chart
owls
i (Clos Front Unroundd)
o (Mid Back Roundd; functions honologicall as high)
(On Front Unroundd)
o (On Cntral Unroundd; functions honologicall as back)
owls occur both short and long; s discussion.
Tons: High, Low
a

Most languags hav borrowd words from othr languags, and som mo
r than othrs. Nvrthlss, a skwd
ictur of th cor honolog can b givn if loanwords ar not considrd sar
atl .
248
Evidnc for Phonmic Contrasts
Prsnt data which stablishs th xistnc of th honms or distin
ctiv faturs claimd. W suggst
that data b givn in honmic rathr than hontic transcrition at
this oint, and organizd b class of
sounds. For xaml, data rgarding  and b nd to b carfull rsntd, but
on dos not worr whthr
 and t ar distinct honms. Thrfor labials nd to b contrast
d, not voiclss stos. Othr classs
includ coronals, vlars, nasals, liquids, tc. Of cours, on should adat th
rsntation to th facts of th
languag. It is not ossibl to roos a cannd st l for rsnting data from
all languags.
Prsnt th data with glosss. Giv xamls with th sounds in vario
us ositions in th word if
ossibl, although th focus of concrn hr is not th distribution
of th honms, but th xistnc of
contrast.
If a crtain contrast is hard to stablish, mak this fact vr clar. (An tim
a contrast is stablishd b
onl a fw xamls, on should b susicious.) Avoid rsnting a mi

slading ictur that will mak it

difficult for th radr to think about altrnativs.
If thr ar unrsolvd issus (and thr alwa s ar!), don t sw t
hm undr th rug. B u front
about thm, although of cours th don t nd to occu an inordinat amount of
sac or attntion.
S llabl Structur
In this sction of a writ-u, discuss kinds of s llabls that occur in th lang
uag. In lin with th viw of
s llabls rsntd in this book, on might start b rsnting th maximal s ll
abl tmlat and illustrativ
xamls of s llabl t s that it covrs. (Again, b carful to dist
inguish s llabls which aar onl in
loanwords.)
Saml S llabl Structur Ovrviw
Maximal S llabl Tmlat: [CC]
CC xamls
CC xamls
C xamls
C xamls
 xamls
C xamls
C xamls
If an of ths s llabl t s is rstrictd to som osition in th word, not
this. For xaml,
(302) S llabls without onsts occur onl word-initiall .
(303) S llabls with codas occur onl if strssd.
If scial s llabls occur at th dg of words, includ discussion of
ths s llabls. For xaml,
s llabls nding in [gz] occur onl at th nd of ol morhmic words
in English, th z bing ithr th
lural suffix (as in frogs), th third singular suffix (lags) or th ossssiv
morhm (igs).
Undr this sction also discuss comlx sgmnts, such as affricats,
and th glid/vowl issu, as
ncssar .
Saml Discussion of Comlx Sgmnts, Glids, tc.
[tj] is anal zd as an affricat (rathr than a clustr) sinc thr ar no unam
biguous xamls of
CCC s llabls. (Provid xamls hr.)
Nons llabic i is anal zd as [j] at th bginning of roots sinc rfix allomor
h shows that such
roots attrn with consonant-initial roots. (Provid data hr.)
Distribution of Phonms
Som honms ma hav strict distributional constraints. Prsnt such constrain
ts.
249
Examls of distributional constraints
/h/ occurs onl s llabl-initiall .
A nasal is alwa s homorganic with th following consonant.
Th following vowl/glid squncs do not occur: ij, ji, uw, wu, ow, wo.
If a s llabl bgins with two consonants, th first consonant must b s.
If

[ C C
Then [+contin]
[coronal]
[-voice ]
Voiced cononant do not occur yllable-finally.

Phonetic Detail
Decribe the phonetic nature of the egment and tone. Thi can be
done by cla of egment when
appropriate.
Sample Dicuion of Phonetic Detail
The phoneme p, t, and k are baically voicele unapirated top (bilabial, al
veolar, and velar).
When a phoneme ha more than one allophone, decribe the allophone and jutify
the rule with adequate
data. For example:
Sample Dicuion of Allophone
The voicele top have an optional voicele naal releae in utterance-final
poition. Thi naal
releae i homorganic with the top.
m i phonetically a naalized labiovelar approximant when it follow a tautoyll
abic velar top.
/kmoom/ [kw o o m] woman (tautoyllabic), /ikmono/ [ikmono] andpiper (not tauto
yllabic). It
i a velar naal when it occur in an untreed yllable and immediately preced
e a back
cononant (k, x, or y). Otherwie it i a labial naal.
It i quite diturbing to read a phonological decription which make
a certain claim for which
inadequate data i preented. If you make a claim, be ure to provide ytematic
illutrative data to back up
the claim and to how that an alternative analyi i not adequate.
Morphological Structure
In preparation for dicuion of phonological rule applying within wor
d, preent the morphological
tructure of word in the language. Thi dicuion hould include wor
d tructure rule which pecify the
order of morpheme within verb, noun, etc. Of coure, for ome language much
more mut be included
here than for other. A phonological decription which i complete in thi area
may be quite long.
Sample Word Structure Rule
Verb Structure:
V Subject Agreement - Tene - Stem - Plural
Phonological Rule
Apart from thoe already dicued in the ection on phonetic detail
above, ome phonological rule may
apply when word are juxtapoed. (Thi i epecially true of tone an
dhi.) Mot of the phonological rule
relevant here, however, are thoe which apply inide of word when mo
rpheme are combined. The rule
250
hould be given, with evidence for each part of the rule, and an ad
equate number of example to jutify
them. Thi part of a phonological decription may alo be quite long; in additi
on to imply preenting the
rule and upporting evidence, jutify the underlying form, the direction and g
enerality of the rule, dicu
and jutify rule ordering, and dicu rule interaction.
Stre and Intonation
Give an account of the general tre pattern of the language, or at leat pree
nt the baic fact. Similarly,
decribe major intonation pattern.
Sample Informal Dicuion of Stre
Stre generally occur on the firt yllable of the root, although there are nu
merou exception. In

compound, the econd part receive primary tre. No clear econdary tre i
perceived.
Sample Informal Dicuion of Intonation
Declarative entence are uually characterized by a gradually riing pitch cont
our which reache a
High level on the lat treed yllable and then drop to Low on ucceeding yl
lable. (Give
example.)
Interrogative ye-no quetion are characterized by a gradually riing pitch con
tour which reache
an extra High level on the lat treed yllable and then drop to Low on ucce
eding yllable.
(Give example.)
Interrogative content quetion are characterized by a pitch contour that tart
at a High level and
fall gradually to Low on ucceeding yllable. (Give example.)
251
Appendix D - Symbol Table
Table 1: Common Cononant (plu a few more)
Bilabial LabioInterDental or
PalatoPalatal Velar
Uvul
ar
Pharyngeal
Glottal
dental dental Alveolar
alveolar
Stop
p
t
c
k
q
?
b
d
j
g
G
Affricate
t 
t j
d z
d
(lateral)
t
Fricative


I
0
j

x
_
h
h

j
y
n
\
(lateral)
I
Naal
m
n
j
p

Liquid
(lateral)
I
(tap, flap)
i
(trill)
r
Glide
J
j
w
Double Articulation:
w kp gb k g (labiovelar)
Modification:
Apirated: t Imploive: 5 u g
Glottalized: t Voicele Sonorant: m
Laryngealized: t
Palatalized: t Labialized: t
Prenaalized:
n
d
252
Table 2: Common Vowel (plu a few more)
a
Front
Central
Back
unround round
unround round
unround round
cloe
i
y
I
u
u
u
near cloe
i
.
u

cloe-mid
r
mid
open-mid

o
o
o

near open

n
open
a
u
n
Modification:
Voicele: a
Laryngealized (glottalized, creaky voiced): a
Naalized: a
Long: V: or V
.
or VV
a

For a comprehenive treatment of ymbol, ee Geoffery K. Pullum and

William A. Laduow (1986) Phonetic
ymbol guide, Chicago, Univerity of Chicago Pre.
253
Appendix E - Language Index and Source
Information
a
Agutaynen (Philippine) Source: Stephen Quakenbuh , p.c. [Exercie on page 14
3.]
Arabic
Cairo Arabic (Egypt) Source: Mot of the data come from Robert T. Harm (1968)
Introduction
to phonological theory, Englewood Cliff, New Jerey, Prentice-Hall. The origina
l ource of the
data there were Walter Lehn and Peter Abboud (1965) Beginning Cairo A
rabic, Autin, Texa,
Multilithed. The ource of the other data i unknown. [Exercie on page 101.]
Bakwiri (Bantu, Cameroon) Source: Jacque Durand (1990) Generative and
non-linear phonology,
London, New York: Longman. The original ource wa J.-M. Hombert (1973
) Speaking
backward in Bakwiri, Studie in African Linguitic 4:227-35. [Data cited on pa
ge 216.]
Baque (Spain) Source: Jo Ignacio Hualde (1991) Unpecified and unmarked vowel,
Linguitic Inquiry
22:205-9. [Exercie on page 184.]
Berber (Morocco) Source: Mohamed Guerel (1986) Glide in Berber and yllabicit
y, Linguitic Inquiry
17:1-12. [Data cited on page 170.]
Bua (Niger-Congo; Nigeria) Source:
Klau Wedekind (1972) An outline
of the grammar of Bua
(Nigeria). Diertation, Kiel. [Exercie on page 125. Data cited on page 125.]
Campa
Aheninca Campa (Peru). Source: Donald A. Burquet and David L. Payne (1993) Pho
nological
Analyi: A Functional Approach. Dalla: SIL and Univerity of Texa at Arlingto
n. [Exercie on
page 129 and 179.]
Capanahua (Panoan, Peru)
Source:
Eugene Loo (1967) The phonology o
f Capanahua and it
grammatical bai, Ph.D. diertation, Univerity of Texa, Autin. [Data cited
on page 107 and
107.]

Cahinahua (Peru)
Source: Kenneth M. Keninger (1963) The phonological
hierarchy of Cahinahua,
Studie in Peruvian Indian language: I, Mexico City, Summer Intitute of Lingu
itic, pp. 207217. [Exercie on page 131.]
Chatino (Zapotecan, Otomanguean, Mexico) Source: Lelie and Kitty Pride,
p.c. [Data cited on page 57
and 83.]
Chinantec
Lalana Chinantec (Chinantecan, Otomanguean, Mexico) Source: Merrifield (1987). [
Exercie on
page 89.]
Lealao Chinantec (Chinantecan, Otomanguean, Mexico) Source: Jim Rupp, p.
c. [Exercie on
page 105.]
Chinee. Source: Larry Hyman (1975) Phonology: theory and analyi, New
York, Holt, Rinehart and
Winton. [Exercie on page 66. Data cited on page 213.]
Min Nan Chinee (Taiwan). Source: Merrifield (1987). They were original
ly provided by D.
Chuang and C. Rench. [Exercie on page 66.]
Chol (Mayan, Mexico). Source: Viola Warkentin and Ruby Scott (1980) Gra
mtica chol, Mexico City:
Summer Intitute of Linguitic. [Exercie on page 24.]
a

Merrifield (1987) i William R. Merrifield et al., ed. (1987) Laboratory manua

l for morphology and yntax, Dalla,
SIL.
254
Chontal
Oaxaca Chontal (Tequitlatecan, Hokan, Mexico) Source: Kenneth L. Pike (1947) Ph
onemic: A
technique for reducing language to writing. Ann Arbor, Univerity of M
ichigan Pre; data
originally provided by Viola Waterhoue and May Morrion. [Exercie on page 127.
]
Czech. Source: unknown. [Exercie on page 174.]
Daga (Papua New Guinea) Source: John and Elizabeth Murane (1972) Voca
lic yllabicity in Daga,
Phonetica 25: 19-26, Elizabeth Murane (1974) Daga grammar: from morphem
e to dicoure,
Univerity of Oklahoma, Norman, SIL. [Exercie on page 128.]
Damana (Chibchan, Colombia) Source: Cindy S. William (1993) A grammar
ketch of D m na, M.A.
Thei, Univerity of North Dakota. [Exercie on page 48, 116, and 130.]
Doyayo (Cameroon) Source: Marinu Wiering and Eliabeth Wiering (1994) Phon
ological decription of
Doyayo (Poli dialect), The Doyayo language: elected tudie, Arlington, Texa:
Summer Intitute
of Linguitic and The Univerity of Texa at Arlington. [Data cited on page 126
.]
Ejagham (Bantu, Nigeria & Cameroon) Source: Donald A. Burquet and Dav
id L. Payne (1993)
Phonological analyi: a functional approach. Dalla: SIL and Univerity
of Texa at Arlington;
data originally provided by K. Watter. [Exercie on page 91.]
Englih [Exercie on page 66 and 205.]
American Englih [Exercie on page 80, 81, and 174.]
Southern Britih Englih [Exercie on page 263 and 264.]

Ekimo
Kukokwim Ekimo (Alaka, U.S.) Source: unknown. (Some phonetical detail ha bee
n omitted.)
[Exercie on page 115.]
Ewe (Kwa, Togo) Source: George N. Clement (1978) Tone and yntax in Ewe, Elem
ent of tone, tre,
and intonation, ed. Donna Jo Napoli, Wahington, D.C., Georgetown Univerity
Pre. [Exercie
on page 219.]
Flemih (Teralfene dialect).
Source: Juliette Blevin (1994) A place
for lateral in the feature geometry.
Journal of Linguitic 30:301-48. The ource of the data were cited a Willem J
. de Reue (1984)
Cluter palatalization and vowel fronting in the Flemih dialect of Te
ralfene, manucript, The
Univerity of Texa at Autin, and C. McClemore (1987) Palatalization
and lateral pread in
Teralfene Dutch, manucript, The Univerity of Texa at Autin. [Data cited on
page 99.]
German. Source: Jerzy Rubach (1990) Final devoicing and cyclic yllabif
ication in German, Linguitic
Inquiry 21:79-94. [Data cited on page 145.]
Pennylvania German (U.S.) Source: Unknown. [Exercie on page 149.]
Greek (Modern)
Source: Andrea Koutouda (1962) Verb morphology of M
odern Greek: a decriptive
analyi. International Journal of American Linguitic 28.4. Publication
24 of the Indiana
Univerity Reearch Center in Anthropology, Folklore, and Linguitic. [Exerci
e on page 86.]
Guanano (Colombia & Brazil) Source: Nathan and Carolyn Waltz (1967) Gu
anano phoneme, Phonemic
ytem of Colombian language, Mexico City, Summer Intitute of Linguitic. [E
xercie on page
116.]
Gwari (Kwa, Nigeria) Source: Larry Hyman (1975) Phonology: theory and
analyi, New York, Holt,
Rinehart and Winton. The original ource wa Larry Hyman (1972) Naal
 and naalization in
Kwa, Studie in African Linguitic 3:167-205. [Data cited on page 215.]
255
Haua (Sudan) Source: Eugene Nida (1949) Morphology: the decriptive an
alyi of word, Ann Arbor,
Univerity of Michigan Pre. The original ource wa Carleton T. Hodg
e (1947) An outline of
Haua grammar, Language Diertation 4, Baltimore: Linguitic Society of America
. [Exercie on
page 141. Data cited on page 208.]
Houailou (Oceanic, New Caledonia) Source: Frantiek Lichtenberk (1978) A ketch
of Houailou grammar,
Working Paper in Linguitic, Univerity of Hawaii Vol. 10, 2:73-116.
Thi work i baed on
Jacqueline de La Fontinelle (1976) La langue de Houailou (Nouevelle-Cald
onie). Pari: Socit
d tude linguitique et anthropologique de France. Langue et civiliation tra
dition orale, 17.
[Exercie on page 47.]
Hungarian. Source: Merrifield (1987). The data were checked and the trancriptio
n adjuted with the help
of Emee Lng. [Data cited on page 93ff and 99.]
Inga (Quechuan, Colombia)
Source:
Stephen H. Levinohn (1976) The I

nga language, The Hague and

Pari, Mouton. [Data cited on page 107.]
Italian. Source: dictionary and micellaneou other ource. [Exercie on page 1
29.]
Japanee. Source: Jame D. McCawley (1978) What i a tone language?, in Victori
a A. Fromkin, Tone: a
linguitic urvey, New York, London; Academic Pre; Junko Ito and Ral
f Armin Meter (1986)
The phonology of voicing in Japanee: theoretical conequence for morp
hological acceibility,
Linguitic Inquiry 17:49-73; and Larry Hyman (1975) Phonology: theory and analy
i, New York,
Holt, Rinehart and Winton. [Data cited on page 56 and 213.]
Javanee (Java) Source: unknown. [Exercie on page 67 and 72.]
Karuk (Hokan, northern California; alo known a Karok) Source: Willia
m Bright (1957) The Karok
language, Univerity of California Publication in Linguitic 13, Berke
ley and Lo Angele:
Univerity of California Pre. [Exercie on page 83, 114, and 168.]
Kikongo (Bantu, Zaire). Source: Benjamin Ao (1991) Kikongo naal harmon
y and context-enitive
underpecification, Linguitic Inquiry 22:193-6. [Data cited on page 88.]
Kikuyu (Bantu, Tanganyika) Source: Merrifield (1987); original ource wa
 B. Hooley, p.c.; and Larry
Hyman (1975) Phonology: theory and analyi, New York, Holt, Rinehart
and Winton. The
original ource wa Mary Pratt (1972) Tone in ome Kikuyu verb form,
Studie in African
Linguitic 3:325-378. [Data cited on page 14ff and 216.]
Korean. (Korea) Source: Victoria Fromkin and Robert Rodman (1983) An introducti
on to language, New
York, London, Sydney: Holt, Rinehart and Winton; and Chang-Hai Park (
1961) An intenive
coure in Korean, Yonei Univerity, Seoul. [Exercie on page 147. Dat
a cited on page 16 and
107.]
Kukuya (Bantu, Congo) Source: Larry M. Hyman (1987) Proodic domain in Kukuya,
Natural Language
and Linguitic Theory 5:311-33. The analyi wa originally due to Chri
tiane Paulian (1974) Le
Kukuya: langue Teke du Congo. [Data cited on page 215.]
Kuman (Chimbu, Papua New Guinea) Source: Juliette Blevin (1994) A pla
ce for lateral in the feature
geometry. Journal of Linguitic 30:301-48. The ource of the data wa
cited a J. Lynch (1983)
On the Kuman liquid, Language and Linguitic in Melaneia 14:98-112. [Data cite
d on page
104.]
Lamba (Zambia & Zaire) Source: Michael Kentowicz and Charle Kiebert
h (1979) Generative
phonology: decription and theory, New York, London, Academic Pre. Th
e data are originally
from C. Doke (1938) Textbook of Lamba grammar, Johanneburg, Witwatert
rand Pre.
[Exercie on page 96 and 193.]
Latin Source: dictionary. [Exercie on page 173. Data cited on page 201.]
Lithuanian. Source: unknown. [Exercie on page 57 and 72.]
256
Madija (Arauan, Peru & Brazil) Source: Paty Licln and Stephen A. Marl
ett (1990) Madija noun

morphology, International Journal of American Linguitic 56:102-120; Pat

y Adam (Licln)
(1987) Gender agreement in Madija, Native American language and gramma
tical typology,
paper from a conference at the Univerity of Chicago, Univerity of Chicago, Ap
ril 22, 1987, ed.
Paul D. Kroeber and Robert E. Moore, 1-18, Indiana Univerity Linguit
ic Club; and Paty
Licln, p.c. [Exercie on page 21 and 25. Data cited on page 30, 44, 46, 46, 10
0, 107, 151, and
170.]
Manam (Autroneian, Manam and Boea Iland, off the coat of Papua New Guinea)
Source: Frantiek
Lichtenberk (1983) A grammar of Manam, Univerity of Hawaii Pre.
[
Exercie on page 10
and 10. Data cited on page 29.]
Manobo
Aguan Manobo (Philippine) Source: Daniel and Marilou Weaver (1963) Th
e phonology of
Aguan Manobo (with pecial reference to ), Paper on Philippine Language No. 1,
Intitute for
Language Teaching, Univerity of the Philippine and Summer Intitute of Lingui
tic, Manila, pp.
1-6. [Exercie on page 117.]
Maori (New Zealand) Source: Kenneth Hale (1973) Deep-urface canonical
diparitie in relation to
analyi and change: an Autralian example, Current trend in linguiti
c 11, Diachronic, areal,
and typological linguitic, The Hague and Pari, Mouton, pp. 400-458. [Exercie
on page 177.]
Margi (Nigeria) Source: Jacque Durand (1990) Generative and non-linear
phonology, London, New
York: Longman. The original ource wa C. Holffmann (1963) A grammar of the Marg
i language,
London: Oxford Univerity Pre. [Exercie on page 216.]
Marinahua (Peru) Source: Eunice Pike and Eugene Scott (1975) The phono
logical hierarchy of
Marinahua, in Ruth M. Brend, ed., Studie in tone and intonation, S. Karger: Ba
el, London, New
York, etc., pp. 196-211. Reprinted from Phonetica 8:1-8 (1962). [Exercie on pa
ge 154, 158, and
219.]
Mazatec
Chiquihuitln Mazatec (Popolocan, Otomanguean, Mexico) Source: Allan Jamie
on (1977)
Chiquihuitlan Mazatec tone, Studie in Otomanguean phonology, ed. Willia
m R. Merrifield,
Dalla, SIL and Univerity of Texa at Arlington. [Data cited on page 212.]
Mende (Bantu, Sierra Leone). Source: William R. Leben (1978) The repreentation
of tone, in Victoria A.
Fromkin, Tone: a linguitic urvey, New York, Academic Pre, pp. 177-219. [Exer
cie on page
205, 217, 218, 218, and 218.]
Mengen (Papua New Guinea). Source: Daniel D. Rath (1993), Phonologie of Autron
eian language No.
2, ed. John M. Clifton, Ukarumpa, SIL. [Exercie on page 24.]
Miwok
Sierra Miwok (California) Source: L. S. Freeland (1951) Language of the Sierra M
iwok, Indiana
Univerity Publication in Anthropology and Linguitic. [Exercie on page 117.]

Mixtec
Atatlahuca Mixtec (Mixtecan, Otomanguean, Mexico). Source: Ruth Mara Alex
ander (1980)
Gramtica mixteca: mixteco de Atatlahuca, Mexico City, Intituto Lingtico
de Verano. [Data
cited on page 212.]
Huajuapan Mixtec (Mixtecan, Otomanguean, Mexico) Source: John H. Cowan
and Eunice V.
Pike (1967) Huajuapan Mixtec phonology and morphophonemic, Anthropological Ling
uitic 115. [Exercie on page 20 and 263.]
Nuyoo Mixtec (Mixtecan, Otomanguean, Mexico) Source: Larry Harri, p.c. [Exerci
e on page
68, 72, 76, and 82.]
257
um Mixtec (Mixtecan, Otomanguean, Mexico)
Source: Laura Gittlen and Ste
phen A. Marlett
(1985) um Mixtec yllable tructure and morphology. Work Paper of the
Summer Intitute of
Linguitic, Univerity of North Dakota, 29:175-94. [Data cited on page 125.]
Ocotepec Mixtec (Mixtecan, Otomanguean, Mexico) Source: Ruth Mary Alexan
der, p.c.
[Exercie on page 194. Data cited on page 212 and 216.]
Peole Mixtec (Mixtecan, Otomanguean, Mexico) Source: John Daly, p.c. See alo Jo
hn P. Daly
(1977) A problem in tone analyi, Studie in Otomanguean phonology, ed. William
R. Merrifield,
Dalla: SIL and Univerity of Texa at Arlington. [Exercie on page 215.]
Mokilee (Mokil Atoll and the iland of Ponape, near the Marhallee Iland) So
urce: Sheldon P. Harrion
with the aitance of Salich Y. Albert (1976) Mokilee reference grammar, Honol
ulu, Univerity
of Hawaii Pre. [Data cited on page 66.]
Nabak (Papua New Guinea) Source: Edmund and Grace Fabian, p.c. [Exercie on pag
e 67 and 76. Data
cited on page 71.]
Ngemba (Bantoid, Cameroon) Source: Charle L. Eatlack (1968) The noun clae o
f Ngemba, Journal of
African Language 7:33-40. [Exercie on page 23.]
Nahuat
Sierra Nahuat (Nahuatl, Uto-Aztecan, Mexico) Source: Harold and Mary Ke
y (1953) The
phoneme of Sierra Nahuat, International Journal of American Linguitic
19:53-6. [Exercie on
page 129. Data cited on page 146, 146, and 146.]
Nahuatl
Hueyapan Nahuatl (Nahuatl, Uto-Aztecan, Mexico) Source: R. Joe Campbell (1976) U
nderlying
/p/ in Hueyapan Nahuatl, International Journal of American Linguitic 42:46-50
. [Data cited on
page 190.]
North Puebla Nahuatl (Nahuatl, Uto-Aztecan, Mexico) Source: Earl Brockwa
y (1963) The
phoneme of North Puebla Nahuatl, Anthropological Linguitic 5:14-17. S
ee revied analyi in
Earl Brockway (1979) North Puebla Nahuatl, in Ronald W. Langacker, ed., Studie
in Uto-Aztecan
grammar 2, 141-98, Dalla: SIL and Univeriity of Texa at Arlington. [Exercie
on page 69, 72,
123, 148, 148, and 154. Data cited on page 104 and 123.]

Southeatern Puebla Nahuatl (Nahuatl, Uto-Aztecan, Mexico) Source: Ralph

Reed, p.c.
[Exercie on page 174.]
Tetelcingo Nahuatl (Nahuatl, Uto-Aztecan, Mexico) Source: Kenneth L. Pi
ke (1947)
Phonemic: A technique for reducing language to writing. Ann Arbor, Un
iverity of Michigan
Pre (data originally provided by Richard S. Pittman); Richard S. Pittman (1961
) The phoneme
of Tetelcingo Nahuatl (Morelo) Nahuatl, A William Cameron Townend en
el vigimoquinto
aniverario del Intituto Lingtico de Verano, Mexico City: ILV, pp. 643-651; and
David Tuggy,
p.c. [Exercie on page 130 and 165.]
Nupe (Nigeria) Source: Larry Hyman (1975) Phonology: theory and analyi
, New York, Holt, Rinehart
and Winton. The original ource wa Iaac George (1970) Nupe Tonology
, Studie in African
Linguitic 1:100-22. [Exercie on page 204.]
Oodham
Tohono O odham (Tepiman, Uto-Aztecan, Arizona) Source: Jane Hill and Of
elia Zepeda,
Tohono O odham plural, Paper preented at the Univeridad de Sonora,
Hermoillo, Sonora,
Mexico. October 1990. [Exercie on page 45.]
Pare (Arawakan, Brazil) Source: Orland Rowan (1967) Phonology of Pare (Arawakan)
Acta Linguitica
Hafrienia: International Journal of Structural Linguitic 10:2, 201-210. [Exer
cie on page 84.]
258
Pengo (India) Source: Michael Kentowicz and Charle Kieberth (1979)
Generative phonology:
decription and theory, New York, San Francico, and London, Academic
Pre. [Exercie on
page 62 and 76.]
Popoluca
Sierra Popoluca (Mixe-Zoquean, Mexico) Source: Merrifield (1987); and Benjamin E
lon (1947)
Sierra Popoluca yllable tructure, International Journal of American Li
nguitic 13:13-17.
[Exercie on page 67 and 148.]
Portuguee
Brazilian Portuguee (Brazil) Source: unknown. [Exercie on page 124.]
Quechua
Corongo Quechua (Quechuan, Peru) Source: Daniel Hintz (1990) Ditinctive
 of Corongo
Quechua: hitorical and ynchronic perpective, M.A. Thei, Grand Fork, Unive
rity of North
Dakota. [Exercie on page 49 and 162. Data cited on page 127, 151, and 177.]
Pataza Quechua (Quechuan, Peru) Source: Charlotte Zahn (1978) Materiale
 para etudio
fonolgico (vol. 1), ed. Eugene Loo, Yarinacocha, Peru, SIL. [Exercie
on page 68, 72, and
77.]
Quichua
Eatern Quichua (Quechuan, Ecuador)
Carolyn Orr and Bety Wriley (19
65) Vocabulario
quichua del oriente del Ecuador, Quito, Intituto Lingtico de Verano. [Exercie o
n page 117.]
Salaaca Quichua (Quechuan, Ecuador) Source: Sheldon Wakoky (1990) The

phonology of
Salaaca Quichua, M.A. Thei, Univerity of North Dakota. [Data cited on page
90 and 139.]
Selepet (Papua New Guinea) Source: K. A. McElhanon (1972) Selepet gram
mar, Part I: from root to
phrae, Pacific Linguitic, Serie B, 21, The Autralian National Univ
erity. [Exercie on page
10.]
Serbo-Croatian Source: Michael Kentowicz (1993) Phonology in generative
grammar, Cambridge,
Ma., Blackwell. The original ource of the data i not known. [Exercie on pag
e 173.]
Seri (Serian, Hokan, Mexico) Source: Edward W. Moer and Mary B. Moe
r (1965) Cononant vowel
balance in Seri (Hokan) yllable, Linguitic 16:50-67; and Stephen A. Marlett
(1981/1995) The
tructure of Seri, Ph.D. Diertation, Univerity of California, San Di
ego. [Exercie on page 4,
20, 21, 22, 38, 37, 114, 147, 164, 169, and 270. Data cited on page 7, 12ff, 15
, 16ff, 18, 29, 29,
36, 37, 64, 71, 93, 99, 104, 113, 115, 125, 128, 139, 140, 146, 152, 157ff, 160,
168, 170ff, 185ff,
200ff, 208ff, 237, 239 and 231ff.]
Sioux
Dakota Sioux (Siouan, North and South Dakota) Source: Patricia Shaw (1
989) The complex
tatu of complex egment in Dakota, Theoretical perpective on Nativ
e American language,
ed. Donna B. Gerdt and Karin Michelon, Albany, SUNY Pre. [Exerci
e on page 202. Data
cited on page 83.]
Spanih. Source: Jame W. Harri (1983) Syllable tructure and tre
in Spanih: a nonlinear analyi,
Cambridge, London, MIT Pre; and other common ource. [Data cited on page 12,
17, 29, 90,
107, 114, 122, 126, 145ff, 161, 168, 170, 171, 176, and 231ff.]
Sre (Viet Nam) Source: unknown. [Exercie on page 178.]
Swahili (Bantu, Zaire) Source: Jim Meyer, p.c. [Exercie on page 31ff.]
259
Tabaru (Indoneia) Source: Edward A. Kotynki (1987) Tabaru phonology and morph
ology, M.A. thei,
Univerity of North Dakota; Edward A. Kotynki (1995) Glottal top and
the naal prefix in
Tabaru and other North Halmaheran language, in Wyn D. Laidig, ed., D
ecriptive tudie in
language of Maluku, part II, 1-17, NUSA 38:Jakarta; Edward A. Kotynk
i (1998, m.) TabaruIndoneian-Englih dictionary. The thei alo appeared complete in Work
paper of the Summer
Intitute of Linguitic: Univerity of North Dakota Seion (1988) vol. 32, pp.
143-216. [Exercie
on page 269.]
Tagalog (Philippine) Source: Tereita V. Ramo (1971) Tagalog tructure
, Honolulu: Univerity of
Hawaii Pre. [Data cited on page 168.]
Tairora (Papua New Guinea) Source: Alex and Loi Vincent (1962) Introd
uctory note on Tairora verb
morphology and yntax, Studie in New Guinea linguitic, pp. 4-27, Oc
eanic Linguitic
Monograph 6, Univerity of Sydney, Autralia. [Exercie on page 263.]

Tarahumara (Taracahitan, Uto-Aztecan, Mexico) Source: Eugene Nida (1949)

Morphology: the
decriptive analyi of word, Ann Arbor, Univerity of Michigan Pre. [Exerci
e on page 111
and 142.]
Tlapanec (Tlapanecan, Otomanguean, Mexico) Source: Mark Weather, p.c. [Data cit
ed on page 162ff.]
Tojolabal (Mayan, Mexico) Source: Julia Supple and Celia M. Dougla (
1949) Tojolabal (Mayan):
phoneme and verb morphology, International Journal of American Linguit
ic 15:168-174.
[Exercie on page 118.]
Totonac (Totonacan, Mexico) Source: Herman P. Achmann (1946) Totonaco
phoneme, International
Journal of American Linguitic 12:34-43. [Exercie on page 147 and 189.]
Miantla Totonac
Source: Juliette Blevin (1994) A place for latera
l in the feature geometry.
Journal of Linguitic 30:301-48. The ource of the data wa cited a
C. J. MacKay (1991) A
grammar of Miantla Totonac, Ph.D. diertation, The Univerity of Texa at Aut
in. [Exercie on
page 43.]
Trio (Surinam) Source: Morgan W. Jone (1972) Language of the Guiana, ed. Joe
ph E. Grime, Mexico
City, Summer Intitute of Linguitic, pp. 42-46. [Exercie on page 124.]
Twana (Bantu, outhern Africa) Source: unknown, but the phonetic fact are ba
ed on C. M. Doke (1954)
The outhern Bantu language, London, New York, Cape Town, Oxford Univerity Pre
 for the
International African Intitute. [Data cited on page 121.]
Turkih
Some of the data are from the following ource: Michael Ke
ntowicz (1993) Phonology in
generative grammar, Cambridge, Ma., Blackwell; and George N. Clement
and Engin Sezer
(1982) Vowel and cononant diharmony in Turkih. In Harry van der Hu
lt and Norval Smith,
ed., The tructure of phonological repreentation, Part 2, pp. 213-255. [Exer
cie on page 97.]
Tzeltal (Mayan, Mexico) Source: Marianna C. Slocum (1948) Tzeltal (Mayan) noun a
nd verb morphology,
International Journal of American Linguitic 14:77-86. Alo ee Marianna C. Slo
cum, Florencia
L. Gerdel, and Manuel Cruz Aguilar (1999) Diccionario del tzeltal de
Bachajn (Chiapa).
[Exercie on page 105.]
Tzotzil (Mayan, Mexico) Source: Judith L. Aien (1987) Tzotzil claue tructure
, Dordrecht and Boton,
Reidel. [Exercie on page 22.]
Ukrainian. Source: Michael Kentowicz and Charle Kieberth (1979) Generative p
honology: decription
and theory, New York, San Francico, and London, Academic Pre. [Exercie on pa
ge 84.]
Venda (South Africa).
Source:
C. M. Doke (1954) The outhern Bantu
language, London, New York,
Cape Town, Oxford Univerity Pre for the International African Intit
ute. [Exercie on page
116.]
Vietnamee. Source: Laurence Thompon (1965) A Vietnamee grammar, Seatt
le, Univerity of
Wahington. [Data cited on page 209.]

260
Walmatjari (Pama-Nyungan, Autralia). Source: Joyce Hudon (1978) The core of W
almatjari grammar,
Canberra, Autralian Intitute of Aboriginal Studie; and New Jerey, Humanitie
Pre. [Exercie
on page 262.]
Wantoat (Papua New Guinea). Source: Merrifield (1987); originally provided by D.
Davi and E. Deibler.
[Exercie on page 65.]
Waorani (Ecuador) Source: Merrifield (1987). [Data cited on page 88.]
Xavante (Brazil) Source: Merrifield (1987); originally provided by E. B
urge and J. Hall. [Exercie on
page 101.]
Yaqui (Taracahitan, Uto-Aztecan, Mexico) Source: Contantino Martnez Fabin (1990)
La reduplicacin
en la categora mayore del yaqui, Paper preented at the Univeridad
de Sonora, Hermoillo,
Sonora, Mexico. [Exercie on page 83.]
Yoruba (Niger-Congo, Nigeria) Source: Diana Archangeli and Dougla Pulleyblank (
1989) Yoruba vowel
harmony, Linguitic Inquiry 20:173-217; and Dougla Pulleyblank (1988) V
ocalic
underpecification in Yoruba, Linguitic Inquiry 19:233-70. [Data cited on page
95 and 125.]
Zapotec
Coateca Alta Zapotec
(Zapotecan, Otomanguean, Mexico) Source: Joeph
Benton, p.c.
[Exercie on page 23.]
Ithmu Zapotec (Zapotecan, Otomanguean, Mexico) Source: Merrifield (1987
) (originally
provided by John and Jean Alop and Velma Pickett) and Stephen A. Ma
rlett and Velma Pickett
(1987) The yllable tructure and apect morphology of Ithmu Zapotec, Internat
ional Journal of
American Linguitic 53:398-422. [Exercie on page 47, 171, 202, and 205. Data
cited on page
151, 177, and 235.]
Quiegolani Zapotec (Zapotecan, Otomanguean, Mexico) Source: Sue Regnier (1993) Q
uiegolani
Zapotec phonology, 1993 Work Paper of the Summer Intitute of Linguitic, Univ
erity of North
Dakota Seion 37. [Data cited on page 161.]
Quioquitani Zapotec (Zapotecan, Otomanguean, Mexico) Source: Michael Ward
, p.c. [Exercie
on page 166.]
Yatzachi el Bajo Zapotec (Zapotecan, Otomanguean, Mexico) Source: Inez M. Butl
er H. (1980)
Gramtica zapoteca: zapoteco de Yatzachi el Bajo, Mexico City, Intituto Lingtico d
e Verano.
[Exercie on page 58.]
Zoque
Chimalapa Zoque (Mixe-Zoquean, Mexico) Source: Lyle Knudon (1975) A na
tural phonology
and morphophonemic of Chimalapa Zoque, M.A.Thei, Grand Fork, Univer
ity of North
Dakota. See alo the dicuion in John A. Goldmith (1990) Autoegmen
tal & metrical
phonology, Oxford, Cambridge, Ma., Bail Blackwell. [Exercie on page
124. Data cited on
page 146 and 201.]

Copainal Zoque (Mixe-Zoquean, Mexico). Source: Roy and Margaret Harrion

and Ctulo
Garca H. (1981) Diccionario zoque de Copainal, Mexico City: Intituto Lingtico de
Verano;
William L. Wonderly (1951) Zoque II: Phoneme and morphophoneme, International
Journal of
American Linguitic 17:105-23. [Exercie on page 59, 68, 82. Data cited on pa
ge 88 and 176.]
261
Appendix F: Open-ended Exercie
Thi appendix preent data et which will be ueful for dicuing important i
ue that face phonologit.
They do not necearily have pat anwer and model olution. If you are intere
ted in dicuing the
problem or the olution, write to me at teve_marlett@il.org.
Other exercie of thi nature may be found on-line oon at the SIL North Dakota
webite: www.nd.il.org.
262
Walmatjari (Pama-Nyungan; Autralia)
The publihed data have been trancribed here phonetically according to
IPA convention. The language
doe not have any fricative. Some ound, including the emi-cononant
w, j, and
(ditinct from the
flap/trill which occur in the data below) do not occur word finally.
[n] i an apico-alveolar, [n] i an
apical pot-alveolar, and [n] i a lamino-alveolar. Similar convention
are ued for the other coronal
cononant. [r] i an apico-alveolar vibrant / flap (depending on context thi p
honetic detail ha not been
included). Iue to look at: voicing, place aimilation (i our feature yt
em adequate?), alternation that
do not eem to be phonologically motivated, morpheme that dont alternate, upple
tive v. non-uppletive
analye.
Locative Ergative Nominative Dative
1. tirpto tirptu tirp tirpku mark on meage tick
2. mon indo mon indu mon in mon ingu woman
3. Iombon d o Iombon d u Iombon Iombon gu mall one
4. mid iI o mid iI u mid iI mid iI ku newborn babe
5. pud uI o pud uI u pud uI pud uI ku frog (pecie)
6. kun d oIo kun d oIu kun d oI kun d oIku daughter
7. kun oro kun oru kun or kun orku dog
8. pit pit t o pit pit t u pit pit pit pitku lightning
9. woIokt o woIokt u woIok woIokku frog (pecie)
10. ponon d o ponon d u ponon ponon gu old woman
11. mopopo mopopu mopo mopowu girl
12. por pupo por pupu por pu por puwu father
After word of three or more yllable, the following pattern of uf
fixe occur (*the allomorphy of the
Dative uffix i not affected by polyyllabicity, however):
13. pobuI uI o pobuI uI u pobuI u pobuI uwu* iter
Other data of importance:
pori boy
poriwon d i boy
konoI on won d i other
piI opgirwon d i blanket
puro camp
purogod i camp (Allative)
263
Southern Britih Englih

a
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.

teo n tone
teo ni tony
+ok rock
+oki rocky
wiIeo willow
wiIeo i willowy
g+o: gra
g+o:i gray
oIt alt
oIti alty
bie beer
bie +i beery
to: tar
to:+i tarry
w5:te water
w5:t+ i ~ w5:te+i watery
Ia: fur
Ia:+i furry
tce tear (v.)
tce+ip tearing

The trancription i baed on the Collin Cobuild Englih Language D

ictionary (1987) London and Glagow:
Collin. Some adaptation to the tandard IPA have been made.
264
Southern Britih Englih (monoyllable)
b
g+o:, po:
gra, pa
bo:n, ko:m
barn, calm
ho:t
heart
o:nt, pIo:nt
aunt, plant
to:
tar
----Ic
le
+cn
wren
Ict
let
+cnt
rent
----+ce
rare
g, m
ga, ma
+n
ran
+t
rat
+nt, pnt
rant, pant
----- ----Ii:
leae
Ii:n
lean
bi:t
beet
----bi:

bee
----ki
ki
kin
kin
kit
kit
Iint
lint
----Iie
fear
Iai 
lice
Iai n
line
Iai t
light
pai nt
pint
pai
pie
Iai e , pai e
fire, pyre
Iu:
looe
pu:n
poon
Iu:t
loot
----Iu:
loo
----pju:
puce
tju:n
tune
kju:t
cute
----kju:
cue
----boj
buh
----Iot
foot
----- ----Ioe , kjoe
lure, cure
Ici 
face
Ici n
lane
Ici t
fate
Ici nt, pci nt

faint, paint
+ci
ray
----deo 
doe
teo n
tone
teo t
tote
weo nt
c
wont
teo
toe
----Io
lo
don
don
Iot, wot
lot, what
Iont
font
----- ----b\
bu
+\n
run
+\t
rut
+\nt
runt
----- ----na:
nure
Ia:n
learn
ha:t
hurt
Ia:nt, ba:nt
d
learnt, burnt
Ia:
fur
----g5:z
gauze
I5:n
lawn
n5:t
nought
h5:nt
haunt
I5:, d5:
law, door
----Iao 
loue
gao n

gown
gao t
gout
kao nt
count
kao
cow
ao e
hour
v5i 
voice
d5i n
join
k5i t
e
quoit
d5i nt
joint
d5i
joy
----b
The trancription i baed on the Collin Cobuild Englih Language Dictionary (1
987) London and Glagow: Collin.
Some adaptation to the tandard IPA have been made.
c
Mot example in thi cell would be polymorphemic or pecial in ome way. For e
xample, wont i an archaic word.
Other word include: wont and dont.
d
Both of thee example are polymorphemic. No monomorphemic example were found.
e
Thi may be a unique word of thi pattern.
265
Huajuapan Mixtec (Otomanguean, Mexico)
The following data are given in a broad phonetic trancription. Each biyllabic
word ha tre on the firt
yllable. Geminate vowel could be analyzed in different way, of coure, a cou
ld element uch a [nd],
[kw] and [kj]. In mot, if not all, Mixtec language, what i trancribed a [p]
i actually a naalized palatal
glide. Alo, in mot Mixtec language, the degree of naalization whic
h i heard adjacent to
naal
cononant i much le than what i heard elewhere. Make ome careful hypothe
e about the phoneme
of thi language. Conider carefully the ditribution of the cononant and vowe
l. And pay cloe attention
to naalization. If you need ome prompting about what to ee in the
e data, look at the quetion which
follow the data.
1. o1o LM good
2. c1c MM houe
3. itji MM now
4. iji LL weet
5. iko LL cloud
6. tjcIo HL calf
7. tji i HL fingernail
8. tjiko ML banana
9. tjo o MM work

10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
266
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.

tjo o LL hen
o 1mo ML clothing
o mo MM will change
oji LL nephew
c1c LM on
c 1c LL lard
itji LL noe
iko LL will ell
iko HH i elling
iko LL neck
i 1o HH buzzard
i ni LL head
ito LL tortilla
ito ML uncle
i 1i LM leg
o1o LM ear
o 1no LL hirt
ykjy LL niece
koo MM will lie down
ko 1o HH want
ko 1o LL will talk
kotji MM cotton
kojo HH hard thing
kjco MM will neeze
kji ni MM poor quality
kjyji HH white
koko MM will wallow
ko po LM meat
koo LL nake
ko1o LL plate
ko mi LH four
ko ni MM yeterday
kwoIi HL hore
kwo 1o LH yellow
kwo1o LH red
kwcc HH low
kjctc MM will dig
kjiti ML animal
kjyo LL year
kjyko LL comb
kjykjy HM i ewing
kjykjy LM will ew
Ioo HL bone
Icndc ML navel
Ickwo LL eyebrow
ndo1o ML hand
ndcc HH i caring for
ndc1c HH i watching
ndc1i LL mud
ndii MM pulque
ndio LL andal
ndoo LL ugarcane
ndoko ML zapote (fruit)
ndoto MM plit wood
ndytji ML bean
po mo LL corn huk
po ni ML brother
pc 1c HM i cratching
po o ML town

69. ni no LM above
70. no 1o LM tooth
71. no 1o ML fire
72. no mo LL oap
73. oi HH i cloing
74. oo LL new
75. o1o LL foot
76. o 1ni HH i tying
77. o no HH i opening
78. jc c HM i buying
79. jito LM bed
80. jito LL oven
81. jii LL grandfather
82. ji 1i LL muhroom
83. joo MM griddle
84. toto LL medicine
267
85. tc1ndc HL i cutting
86. tc1ndc LL will cut
87. tci ML chair
88. titji LL avocado
89. tiIo LM bird
90. ti no LM dog
91. tiji LM tomach
92. ti oko LMM fih
93. to mi HM feather
94. tyty ML paper
95. oo HL tongue
96. okwo MM crooked
97. c1c HH door
98. iko ML furrow
99. ito ML tree
100. oko LL team
101. oo ML water jug
102. oo LL grinding tone
103. o1o ML rope
104. y1o LL thread
105. y1y ML mouth
106. yy LL rock
107. 1io MM our
108. 1i1i HH raw
109. 1i i LL alt
110. 1i i LL nine
111. 1i i MM one
112. 1i o ML thorn
113. 1i ni MM warm
114. 1i no LL ix
115. 1oo LL bitter
116. 1yo LL even
Quetion relating to the Huajuapan Mixtec data.
1. I naalization contrative?
2. Conider the ditribution of all cononant with repect to naalized vowel
? Which one are defective
in their ditribution?
3. Conider the ditribution of naalized vowel. I there anything to comment
on?
4. Conider how glottal top hould be analyzed.
5. How will you analyze [kj], [kw], and [tj]?
268
Tairora (Papua New Guinea)

All verb in Tairora conjugate like one of the following verb. (For the purpoe
 of thi problem, analyze
the uffixe a monomorphemic chunk.)
peak go eat
Pat 3
tibo bibo noibo
Dubitative 2 tirccro birccro noirccro
Perfective 3
tiro biro noiro
Imperative Sg ticno buono noono
Imperative Pl ticto buoto nooto
Avolitional 1 ticrooro buorooro noorooro
Abilitative 3
ticnoroo buonoroo noonoroo
Pat 1
turouko burouko nourouko
Perfect 1 tunoro bunoro nounoro
Far Pat 3
turo buro nooro
Far Pat 2
tunoro bunoro noonoro
Avolitional 2 tirooro birooro nccrooro
Avolitional 3p tibooro bibooro nccbooro
Neutral ti bi ncc
Cutomary 3p tiko biko nccko
269
Tabaru (Indoneia)
Monomorphemic word fall into two major pattern if they conit of more than on
e yllable. Examine all
of the data below and propoe a ingle, imple tre rule. Thi rule interact
with other phonological rule
which you may wih to propoe. Hint: the data at the bottom of the page are a i
mportant a thoe at the top.
Word with penultimate tre
goo
buttock
po pokc
illne
moo
feel
bc Icko
houlder
1owc
thread
o moko
watermelon
1uru
mouth
ko kowo
ah
iro
firt
mo didi
two
rio
footprint
tjo ono
pant
biro
rice
mo doko
marry
gcri
brother-in-law
do wopi
and

Word with antepenultimate tre (alway with identical vowel in lat two yll
able)
odomo
eat
owoko
fruit
1okcrc
water
wckoto
wife
gogcrc
it
ouku
hot
nooko
fih
obiri
night
koono
ultan
oii
urinate
go omopo
crocodile
tumu diipi
even
gu mutuku
rope
bo 1ouku
grave
If a word ha a uffix, tre alway fall on the penultimate yllable.
kuIo
give
ku Ioi
till give
tikc
look for
ti kci
till look for
1okcrc
drink
1okc roi
till drink
mokc
ee
mo kcwo
not ee
tu 1uru
follow
tu1u ruwo
not follow
bo dito
have mifortune
bodi towo
not have mifortune
pidiIi
releae
pidi Iuwo
not releae
odomo

eat
odo muwo
not eat
po 1oono
beat
po1oo nuwo
not beat
punuu
be full
punu uwo
not be full
270
Seri (Mexico)
if I ... if you ... if /he ... did I ... did you ... did /he ...
i1 poo1o im poo1o i poo1o i1 to1o in to1o i to1o
ee it/him/her
i1 poojt im poojt i poojt i1 tojt in tojt i tojt
tattoo him/her
i1 poom im poom i poom i1 tom in tom i tom
wallow it
i1 pom im pom
pom i1p toom in toom
toom
wallow (unpec.)
i1 poitom im poitom
poitom i1p toitom in toitom
toitom
talk
i1 poop im poop
poop i1p top in top
top
arrive
i1 pooim im pooim
pooim i1p toim in toim
toim
laugh
i1 pootxo im pootxo
pootxo i1p ttxo in ttxo
ttxo
tagger
i1 pookto im pookto i pookto i1 tookto in tookto i tookto
look at it/him/her
i1 poonI im poonI i poonI i1 toonI in toonI i toonI
tir it
i1 potx im potx
potx i1p totx in totx
totx
arie
i1 pii im pii i pii i1 tii in tii i tii
hear it/him/her
i1 piim im piim
piim i1p tiim in tiim
tiim
leep
i1 pooi im pooi i pooi i1 tooi in tooi i tooi
do it
i1 poojy im poojy
poojy i1p toojy in toojy
toojy
be fat
i1po pii impo pii ipo pii i1 tpii in tpii i tpii
tate it
i1po ponjy impo ponjy po ponjy i1p tponjy in tponjy
tponjy
run
i1po jyok impo jyok ipo jyok i1 tjyok in tjyok i tjyok
hack it off
po mooI
tmooI
be yellow
i1po onx impo onx ipo onx i1 tonx in tonx i tonx
carry him/her on back
i1po onx impo onx po onx i1pto onx into onx to onx
carry on back
(unpecified)
i1 piip im piip i piip i1 tiip in tiip i tiip
carry it on head
i1 pp im pp
pp i1p tp in tp
tp

carry on head
(unpecified)
i1po ijx impo ijx ipo ijx i1 ttijx in ttijx i ttijx
grind it
i1po ijx impo ijx po ijx i1pto ijx into ijx to ijx
grind (unpecified)
271
Seri (negative of previou page):
i1po mo1o impo mo1o ipo mo1o i1 tkw o 1o in tkw o 1o i tkw o 1o
ee it/him/her
i1po mojt impo mojt ipo mojt i1 tkw o jt in tkw o jt i tkw o jt
tattoo him/her
i1po mom impo mom ipo mom i1 tkw o m in tkw o m i tkw o m
wallow it
i1po moom impo moom po moom i1p tkw o o m in tkw o o m
tmoom
wallow
i1po moitom impo moitom po moitom i1p tkw o i tom in tkw o i tom
tmoi
tom
talk
i1po mop impo mop po mop i1p tkw o p in tkw o p
tmop
arrive
i1po moim impo moim po moim i1p tkw o im in tkw o im
tmoim
laugh
i1po mtxo impo mtxo po mtxo i1p tkw txo in tkw txo
tmtxo
tagger
i1po mookto impo mookto ipo mookto i1 tkw o o kto in tkw o o kto i tkw
o o kto
look at it/him/her
i1po moonI impo moonI ipo moonI i1 tkw o o nI in tkw o o nI i tkw o o
nI
tir it
i1po motx impo motx po motx i1p tkw o tx in tkw o tx
tmotx
arie
i1po mii impo mii ipo mii i1 tkw i i
in tkw i i
i tkw i i
hear it/him/her
i1po miim impo miim po miim i1p tkw i i m in tkw i i m
tmiim
leep
i1po mooi impo mooi ipo mooi i1 tkw o o i in tkw o o i
i tkw o o i
do it
i1po moojy impo moojy po moojy i1p tkw o o jy in tkw o o jy
tmoojy
be fat
i1pom pii impom pii ipom pii i1tkom pii intkom pii itkom pii
tate it
i1pom ponjy impom ponjy pom ponjy i1ptkom ponjy intkom ponjy tom ponjy
run
i1pop jyok impop jyok ipop jyok i1tkop jyok intkop jyok itkop jyok
hack it off
pom mooI tom mooI
be yellow
i1pon onx impon onx ipon onx i1tkon onx intkon onx itkon onx
carry him/her on
back
i1pomo onx impomo onx pomo onx i1ptkw o onx intkw o onx tmo on
x
carry on back
(unpec.)
i1po miip impo miip ipo miip i1 tkw i i p in tkw i i p i tkw i i p
carry it on head
i1po mp impo mp po mp i1p tkw p in tkw p
tmp
carry on head

(unpec.)
i1pom mijx impom mijx ipom mijx i1tkom mijx intkom mijx itkom mijx
grind it
i1pomo ijx impomo ijx pomo ijx i1ptkw o ijx intkw o ijx tmo ijx
grind (unpec.)
272
Seri (affirmative paive of firt page)
i1po po1o impo po1o po po1o i1p tpo1o in tpo1o
tpo1o
ee
i1po pokto impo pokto po pokto i1p tpokto in tpokto
tpokto
look at
i1po p impo p po p i1p tp in tp
tp
hear
po poii
tpoii
do
po1 pii to1 pii
tate
po1 jyok to1 jyok
hack off
i1po1 onx impo1 onx po1 onx i1pto1 onx into1 onx to1 onx
carry on
back
po pp
tpp
carry on
head
po1 ijx to1 ijx
grind
Negative paive of firt page:
i1pom po1o impom po1o pom po1o i1ptkom po1o intkom po1o tom po1o
ee
i1pom pokto impom pokto pom pokto i1ptkom pokto intkom pokto tom pokto
look at
i1pom p impom p pom p i1ptkom p intkom p tom p
hear
pom poii tom poii
do
pomo1 pii tomo1 pii
tate
Micellaneou:
mo1 to1o
did I ee you?
mo to1o
did /he ee you?
ki1jo n to1o
who did you ee?
ojjo i to1o
what did he ee?
273
Seri (continued)
I ...! you ... ! /he ... ! I did ... you did ... /he did ...
i1 yoo1o ir yoo1o i yoo1o i1 joo1o ip joo1o i joo1o
ee it/him/her
i1 yoojt ir yoojt i yoojt i1 joojt ip joojt i joojt
tattoo him/her
i1 yoom ir yoom i yoom i1 joom ip joom i joom
wallow it
i1p yom ir yom
yom i1p jom ip jom
jom
wallow
i1p yoitom ir yoitom
yoitom i1p joitom ip joitom
joitom
talk

i1p yoop ir yoop

yoop i1p joop ip joop
joop
arrive
i1p yooim ir yooim
yooim i1p jooim ip jooim
jooim
laugh
i1p yootxo ir yootxo
yootxo i1p jootxo ip jootxo
jootxo
tagger
i1 yookto ir yookto i yookto i1 jookto ip jookto i jookto
look at
it/him/her
i1 yoonI ir yoonI i yoonI i1 joonI ip joonI i joonI
tir it
i1p yotx ir yotx
yotx i1p jotx ip jotx
jotx
arie
i1 yii ir yii i yii i1 jii ip jii i jii
hear
it/him/her
i1p yiim ir yiim i yiim i1p jiim ip jiim i jiim
leep
i1 yooi ir yooi i yooi i1 jooi ip jooi i jooi
do it
i1p yoojy ir yoojy
yoojy i1p joojy ip joojy
joojy
be fat
i1yo pii iryo pii iyo pii i1jo pii ipjo pii ijo pii
tate it
i1pyo ponjy iryo ponjy yo ponjy i1pjo ponjy ipjo ponjy jo ponjy
run
i1yo jyok iryo jyok iyo jyok i1jo jyok ipjo jyok ijo jyok
hack it off
yo mooI jo mooI
be yellow
i1yo onx iryo onx iyo onx i1jo onx ipjo onx ijo onx
carry him/her
on back
i1pyo onx iryo onx yo onx i1pjo onx ipjo onx jo onx
carry on back
(unpec.)
i1 yiip ir yiip i yiip i1 jiip ip jiip i jiip
carry it on
head
i1p yp ir yp
yp i1p jp ip jp
jp
carry on head
(unpec.)
i1yo ijx iryo ijx iyo ijx i1jo ijx ipjo ijx ijo ijx
grind it
i1pyo ijx iryo ijx yo ijx i1pjo ijx ipjo ijx jo ijx
grind
(unpec.)
274
Negative of previou page
i1yo mo1o iryo mo1o iyo mo1o i1jo mo1o ipjo mo1o ijo mo1o
ee it/him/her
i1pyo mop iryo mop yo mop i1pjo mop ipjo mop jo mop
arrive
i1yo mookto iryo mookto iyo mookto i1jo mookto ipjo mookto ijo mookto
look at
it/him/her
i1pyo motx iryo motx yo motx i1pjo motx ipjo motx jo motx
arie
i1yo mii iryo mii iyo mii i1jo mii ipjo mii ijo mii
hear

it/him/her
i1pyo miim iryo miim yo miim i1pjo miim ipjo miim jo miim
leep
i1yom pii iryom pii iyom pii i1jom pii ipjom pii ijom pii
tate it
i1pyom ponjy iryom ponjy yom ponjy i1pjom ponjy ipjom ponjy jom ponjy
run
Affirmative Paive of previou page
i1pyo po1o iryo po1o yo po1o i1pjo po1o ipjo po1o jo po1o
ee
i1pyo pokto iryo pokto yo pokto i1pjo pokto ipjo pokto jo pokto
look at
i1pyo p iryo p yo p i1pjo p ipjo p jo p
hear
yo1 pii jo1 pii
tate
Negative Paive of previou page
i1pyom po1o iryom po1o yom po1o i1pjom po1o ipjom po1o jom po1o
ee
i1pyom pokto iryom pokto yom pokto i1pjom pokto ipjom pokto jom pokto
look at
i1pyom p iryom p yom p i1pjom p ipjom p jom p
hear
yomo1 pii jomo1 pii
tate
275
Seri (continued)
I ... you ... /he ... I will ... you will ... /he will ...
i1 mii1o im mii1o i mii1o i1 ii1o in ii1o i ii1o
ee it/him/her
i1 miijt im miijt i miijt i1 iijt in iijt i iijt
tattoo him/her
i1 miim im miim i miim i1 iim in iim i iim
wallow it
oom
i1p moom im moom
moom i1p oom in oom
wallow
i1p miitom im miitom
miitom i1p iitom in iitom
iitom
talk
iip
i1p miip im miip
miip i1p iip in iip
arrive
i1p miiim im miiim
miiim i1p iiim in iiim
iiim
laugh
iitxo
i1p miitxo im miitxo
miitxo i1p iitxo in iitxo
tagger
i1 mookto im mookto i mookto i1 ookto in ookto i ookto
look at it/him/her
i1 moonI im moonI i moonI i1 oonI in oonI i oonI
tir it
i1p motx im motx
motx i1p otx in otx
otx
arie
i1 mii im mii i mii i1 ii in ii i ii
hear it/him/her
i1p miim im miim
miim i1p iim in iim
iim
leep
i1 mooi im mooi i mooi i1 ooi in ooi i ooi
do it
i1p moojy im moojy
moojy i1p oojy in oojy
oojy
go fat
i1m pii mim pii im pii i1 pii in pii i pii
tate it

i1pim ponjy mim ponjy im ponjy i1p ponjy in ponjy

ponjy
run
i1p jyok min jyok in jyok i1 jjyok ip jjyok i jjyok
hack it off
im mooI
mooI
be yellow
i1n onx min onx in onx i1 onx in onx i onx
carry him/her on
back
i1pmo onx immo onx mo onx i1po onx ino onx o onx
carry on back
i1 miip im miip i miip i1 iip in iip i iip
carry it on head
i1p mp im mp
mp i1p p in p
p
carry on head
i1m mijx mim mijx im mijx i1 ijx in ijx i ijx
grind it
i1pmo ijx immo ijx mo ijx i1po ijx ino ijx o ijx
grind
276
Negative of previou page
There ore no negotive form of thi tene.
i1 kw o 1o in kw o 1o i kw o 1o
ee it/him/her
i1p kw o p in kw o p
mop
arrive
i1 kw o o kto in kw o o kto i kw o o kto
look at it/him/her
i1p kw o tx in kw o tx
motx
arie
i1 kw i i in kw i i i kw i i
hear it/him/her
i1p kw i i m in kw i i m
miim
leep
i1kom pii inkom pii ikom pii
tate it
i1pkom ponjy inkom ponjy om ponjy
run
Affirmative Paive of previou page
po1o
i1pim po1o mim po1o im po1o i1p po1o in po1o
ee
i1pim pokto mim pokto im pokto i1p pokto in pokto
pokto
look at
p
i1pim p mim p im p i1p p in p
hear
imo1 pii o1 pii
tate
Negative Paive of previou page
i1pkom po1o inkom po1o om po1o
ee
i1pkom pokto inkom pokto om pokto
look at
i1pkom p inkom p om p
hear
mo1 pii
tate
277
Seri (continued)
who ... who will ... Imperative that I ... that you ... that /he ...
ii1o
ki1o
ko1o
1ii1o
mii1o
ii1o

ee it/him/her
kijt
kojt 1iijt
miijt
iijt
iijt
tattoo him/her
iim
kim
kom
1iim
miim
iim
wallow it
koom
om i1 jom ip jom
jom
oom
wallow
iitom
koitom
koitom 1iitom
miitom
iitom
talk
kop
kop
1iip
miip
iip
iip
arrive
iiim
koim
koim 1iiim
miiim
iiim
laugh
ktxo
ktxo
1iitxo
miitxo
iitxo
iitxo
tagger
ookto
kookto
1ookto 1i 1ookto mi 1ookto i 1ookto
look at it/him/her
koonI
1oonI 1i 1oonI mi 1oonI i 1oonI
oonI
tir it
otx
kotx
otx i1 jotx ip jotx
jotx
arie
kii
1ii 1i 1ii mi 1ii i 1ii
ii
hear it/him/her
iim
kiim
1iim 1i 1iim mi 1iim i 1iim
leep
kooi
ooi 1ooi 1i 1ooi mi 1ooi i 1ooi
do it
oojy
koojy
oojy i1 joojy ip joojy
joojy
be fat
kpii
pii i1 pii 1i pii mi pii i pii
tate it
ponjy i1 ponjy 1i ponjy mi ponjy i ponjy
kponjy
run
kjyok
jjyok i1 jyok 1i jyok mi jyok i jyok
hack it off
mooI i mooI
kwooI
be yellow
konx
onx i1 onx 1i onx mi onx i onx
carry him/her on back
ko onx o onx o onx i1jo onx ipjo onx jo onx
carry on back (unpec.)
kiip
1iip 1i 1iip mi 1iip i 1iip
iip
carry it on head
p
kp
p i1 jp ip jp
jp
carry on head (unpec.)
kkijx
1iijx
miijx
iijx
ijx i 1ijx
grind it
ko ijx o ijx o ijx i1jo ijx ipjo ijx jo ijx
grind (unpec.)
278
Negative of previou page
mo1o
i mo1o
kw o 1o 1i mo1o mi mo1o i mo1o
ee it/him/her
i mop
kw o p 1i mop mi mop i mop
mop
arrive
mookto
i mookto
kw o o kto 1i mookto mi mookto i mookto
look at
it/him/her
mii
i mii
kw i i 1i mii mi mii i mii
hear it/him/her

i miim
miim
kw i i m 1i miim mi miim i miim
leep
mooi
i mooi
kw o o i 1i mooi mi mooi i mooi
do it
im pii om pii kom pii 1im pii mim pii im pii
tate it
im ponjy om ponjy kom ponjy 1im ponjy mim ponjy im ponjy
run
im mooI om mooI im mooI
be yellow
in onx on onx kon onx 1in onx min onx in onx
carry him on
back
Affirmative Paive
po1o 1i po1o mi po1o i po1o
1o po1o
ee
1o pokto
pokto 1i pokto mi pokto i pokto
look at
p 1i p mi p i p
1o p
hear
1o poii
poii i poii
do
1o pii o1 pii i1o pii
tate
1i1o onx mi1o onx i1o onx
carry on back
Negative Paive
im po1o om po1o 1im po1o mim po1o im po1o
ee
im pokto om pokto 1im pokto mim pokto im pokto
look at
im p om p 1im p mim p im p
hear
im poii om poii im poii
do
imo1 pii mo1 pii imo1 pii
tate
279
Seri (continued)
what I ... what you ... what he ...
I can ...
a you can ... /he can ...
1o koo1o mo koo1o o koo1o i1 tkw o 1o in tkw o 1o i tkw o 1o
ee it/him/her
1oojt
moojt
oojt i1 tkw o jt in tkw o jt i tkw o jt
tattoo him/her
1oom
moom
oom i1 tkw o m in tkw o m i tkw o m
wallow it
i1p tkw o i tom in tkw o i tom
tmoitom
talk
i1p tkw o p in tkw o p
tmop
arrive
i1p tkw o im in tkw o im
tmoim
laugh
1o okto mo okto o okto i1 tkw o o kto in tkw o o kto i tkw o o kto
look at
it/him/her
1o onI mo onI o onI i1 tkw o o nI in tkw o o nI i tkw o o nI
tir it
i1p tkw o tx in tkw o tx
tmotx

arie
1ii
mii
ii i1 tkw i i
in tkw i i i tkw i i
hear it/him/her
i1p tkw i i m in tkw i i m
tmiim
leep
i1 jooi ip jooi
jooi i1 tkw o o i in tkw o o i i tkw o o i
do it
i1p tkw o o jy in tkw o o jy
tmoojy
be fat
1o pii mo pii o pii i1tkom pii intkom pii itkom pii
tate it
i1ptkom ponjy intkom ponjy tom ponjy
run
Negative of above
1i mo1o mi mo1o i mo1o
(nonexitent) ee
1i mookto mi mookto i mookto
look at
1i mii mi mii i mii
hear
1i mooi mi mooi i mooi
do
1im pii mim pii im pii
tate
a

Verb in thi tene are alo generally followed by another morpheme (1o), thu
making them ditinct from otherwie
homophonou verb form. Thu, I can ee it i actually [i1 tkw o 1o1o]
280
Seri (continued)
to ... to (paive) ...
i 1o1o iko po1o
ee
i 1ojt iko pojt
tattoo
i 1om iko pom
wallow
i koitom
talk
i kop
arrive
i koim
laugh
i ktxo
tagger
i 1okto iko pokto
look at
i 1onI iko ponI
tir
i kotx
arie
i 1ii iko p
hear
i kiim
leep
i 1ooi iko poii
do
i koojy
be fat

i1o pii iko1 pii

tate
iko ponjy
run
i1o jyok iko1 jyok
hack off
iko mooI
be yellow
i1o onx iko1 onx
carry on back
iko onx
carry on back (unpecified)
i 1iip iko pp
carry on head
i kp
carry on head (unpecified)
i1o ijx iko 1ijx
grind
iko ijx
grind (unpecified)
281
Seri (continued)
*i1o mo1o (to not ee), *ikom po1o (to not be een) (i.e., there are no negati
ve infinitive)
moi 1ojt to tattoo you
The data below are given only a ome item that may affect how you look at ome
of the verb data.
my... your... hi/her... Unpoeed
1i Iit mi Iit i Iit 1o Iit
head
1ii
mii
ii
1oo
noe
1i tn mi tn i tn 1o tn
mouth
Other data which may be helpful (at leat omebody thought o at one time):
ko ikx
who cover one houlder (with jacket)
joop
roadrunner
po xoo
quid
nop
mountain lion
ki joomxk
/he took it to him/her
The marking of tre in thee data hould not be taken a clearly delineating 
yllable boundarie.
The yllable which are marked [o i] may be pronounced [ oi| by many people and
perhap in mot
ituation.
282
Appendix G: Gloary of Key Concept
b
ablaut rule
a rule which change one vowel into another by changing one or m
ore of it feature
alienable cla
In many language, the difference between noun
that do not have to be poeed by
omeone or omething (the alienable cla) and thoe that are typicall
y poeed by omeone or
omething (the inalienable cla) i important. Kinhip term (uch a niece) an

d body part word

(uch a arm) are the two mot common clae of word that are typically referr
ed a belonging to
the inalienable cla of noun.
Crytal 1991: If a poeed item i
een a having only a
temporary or non-eential dependence on a poeor, it i aid to b
e alienable, wherea if it
relationhip to the poeor i a permanent or neceary one, it i inalienable
.
allomorphic evidence
evidence ued in determining underlying form
 (including contrative feature)
that conit of the variou allomorph that a given morpheme exhibit
; contrated with
ditributional evidence
allomorph
When a morpheme ha more than hape, it ha i aid to have more
than one allomorph. Two
allomorph of the pat tene uffix in Englih are [t] (a in wept) and [d] (a
in eized). Two of the
allomorph of the noun for leaf in Englih are [lijf] and [lijv] (the latter occur
ring when the word
i plural).
allophone
the phonetic realization of a phoneme
arbitrary word clae
Sometime the vocabulary of a language divide into g
roup for one reaon or
another; if the clae do not relate to any phonological, emantic, or yntacti
c feature of the word
in quetion, the clae are aid to be arbitrary and are omething jut memoriz
ed by the language
peaker. The o-called trong verb of Englih (thoe that form their pat tene i
n irregular way,
uch a peak/poke, get/got, rie/roe, do/did, etc.) are a mall arbitrary wor
d cla of Englih.
aimilation
any proce by which one ound become more like another in it
environment
[ATR]
advanced tongue roota feature which ditinguihe a tene vowel uch a [
i], in which the
tongue root advance, with a lax vowel uch a [i] in which it doe not
[back]
a feature which characterize ound produced by backing the tongue
body, uch a velar and
uvular, and back vowel
bleeding order
a relationhip between ordered rule in which the rule (A) t
hat applie firt remove a
tructural repreentation to which another rule (B) would otherwie have applied
(Crytal 1991)
C
The ymbol C i often ued a a horthand for any cononant (non-yllabic ou
nd).
coalecence
the fuion of two adjacent ound which reult in a ound whic
h hare the propertie of
the two original ound
coda
the part of a yllable which follow the nucleu, uch a the t of the
yllable kat
complementary ditribution
When ound occur in mutually excluive environm
ent, they are aid to
occur in complementary ditribution.
cononant mutation rule
a rule that change one cononant into another by al
tering one or more of it
feature
contraint on rule
principle by which the application of rul
e i affected (blocked or retricted or
permitted) without complicating the rule in quetion by the addition

of pecific extra feature,

morpheme boundarie, etc.
b

Some gloe cite David Crytal (1991) A dictionary of linguitic and phonetic
, Bail Blackwell: Cambridge, Ma.
and Oxford, England.
283
[continuant]
Sound that are [-continuant] include thoe in which the air t
ream i completely blocked at
the primary contriction in the vocal tract (uch a b, n, l). The incluion of
lateral in the group of
non-continuant i adopted by Chomky & Halle in The Sound Pattern of
Englih. Many other
dicuion of thi feature define it to exclude l.
contour egment
a way of decribing affricate, prenaalized t
op, and imilar ound which are
phonetically complex and yet function phonologically a ingle ound
contour tone
a tone which i viewed a being inherently non-level, uch a
Falling or Riing
contrat in identical environment
evidence for a ditinctive feature which
i found in context which
differ only in that feature (minimal pair)
contrat in noninfluencing environment
evidence for a ditinctive feature w
hich i found in context
which may alo differ in other, but uppoedly irrelevant, feature
contrative feature
phonological feature which minimally ditingui
h ound which contrat in a
language (mot commonly known a ditinctive feature)
[Coronal]
the major place feature node which cover ound which are pronou
nced uing the front or
blade of the tongue (dental, alveolar, alveolapalatal); many dicui
on of phonology have
included palatal cononant with the coronal.
declination
the gradual fall in pitch that occur from the beg
inning of an utterance over ome pan of
word.
deletion
the lo of a feature or egment in a given context
derivation
the formal repreentation of how an underlying form (or concatena
tion of underlying form)
undergoe the phonological rule of the language; the lat tage of a phonologic
al derivation i the
urface form of the word, which i often the phonetic trancription (or omethin
g cloe to it)
derived environment
an environment for rule application which i
not found in a phonological
repreentation a it occur in the lexicon. One common kind of derived environme
nt i that created
when word formation rule join morpheme together.
diimilation
the change by which one ound become le alike a neighborin
g ound
ditinctive feature
(ee contrative feature)
ditributional evidence
evidence for deciding whether feature are ditinc
tive or not ditinctive, baed
on the context in which the feature occur (a oppoed to allomorphic evidence)
[Doral]
the feature place node that cover ound which are p
roduced uing the back of the tongue
(palatal, velar, uvular).
downdrift
the phonetic lowering of a high tone after a low tone
downtep
the lowering of a high tone after a low tone which i not part of t

he phonetic repreentation.
(The low tone which trigger the downtep may or may not be independently attet
ed.)
epenthei
the inertion of a ound in a particular context,
uually motivated by yllable tructure
pattern
feature
(in phonology) a property of the ound. Sound are viewed a a com
bination of variou feature
uch a [voice] and [naal].
feature preading
the formalization of aimilation rule by drawing new a
ociation line between a
feature and a node
feeding order
a relationhip between ordered rule in which the
rule (A) that applie firt create a
tructural repreentation to which another rule (B) i applicable (Crytal 1991)
formative
The term formative i often ued a a lightly more for
mal way to refer to a morpheme.
Crytal 1991: A formally identifiable, irreducible grammatical element wh
ich enter into the
contruction of larger linguitic unit, uch a word and entence.
284
free variation
the ubtitutability of one ound for another in a given env
ironment, with no conequent
change in the word meaning by a peaker (Crytal 1991)
[high]
If the tongue body i raied in the production of a doral ound, t
he ound i [+high] (palatal and
velar, high vowel). (Caution: The definition and ue of thi feature in curre
nt theory i different
from earlier generative theory.)
inalienable cla
See alienable cla.
intonation
the ditinctive ue of pattern of pitch, or melody (
Crytal 1991) that doe not relate to
lexical ditinction
[Labial]
the major place feature node that cover ound which
are pronounced uing the lip a
articulator
labialization
a econdary articulation involving any noticeable lip-rounding (
Crytal 1991)
[low]
If the tongue body i lowered in the production of a vowel, it i [+lo
w].
major place feature
the feature node currently ued mot commonly to diti
nguih between the major
point of articulation are [Labial], [Coronal], and [Doral]
minimal pair
two utterance which are ditinguihed only by one feature or
phoneme
minor rule
rule which apply to a pecially marked ubet of word or morph
eme ince the majority of
cae do not undergo the rule (even they may qualify tructurally to do o). The
rule that change
foot into feet in the plural form i a minor rule of Englih which applie to a
mall ubet of noun.
morpheme the minimal ditinctive unit of grammar, and the central conce
rn of morphology (Crytal
1991). A word uch a tree i compoed of one morpheme, and the word tree of
two.
morphological conditioning
When the ditribution of uppletive al
lomorph i determined by the
preence or abence of another morpheme in the word, that uppletive allomorphy
i aid to have
morphological conditioning.

morphological ubclaification
the diviion of the lexicon of a language b
aed on morphological fact,
uch a whether a given root may co-occur with a certain affix. For example, o
me adjective in
Englih may combine with the prefix
un- (e.g. happy, certain, able),
and ome may not (funny,
melancholic, proud, tall, ad). Thi fact point to one morphological ubclai
fication of Englih
adjective.
morphology
the area of linguitic that deal with word formation
multiple function formative
a formative that conflate more than one fairly
obviou ditinct emantic or
yntactic feature. The cae ending of Greek conflate gender, number and cae
. Some language
conflate ubject peron and tene, o that one prefix indicate both
who did the action and alo
whether the action wa pat, preent or future. A pronoun uch a u could be vi
ewed a multiple
function formative that conflate peron (firt), number (plural), and
cae (accuative). A
portmanteau (q.v.) i a more radical cae of a multiple function formative.
naalization
the effect on ound which are produced while the
oft palate (or velum) i lowered,
allowing air to ecape through the noe
[naal]
the feature decribing ound which are produced with the oft pala
te lowered, allowing air to
ecape through the noe
natural cla
a grouping of ound (in the cae of phonology) which i baed
on ome inherent feature()
of the ound. For example, the ound that have the feature [+voice] form a na
tural cla, and the
ound that have the feature [-voice] form another natural cla. The ound [p]
, [z], and [a] do not
form a natural cla becaue they do not hare any feature.
non-arbitrary word clae
See arbitrary word clae.
nucleu
the core part of a yllable, which may conit of one or more v
owel and diphthong, or even
ome cononant-like ound (cf. yllabic naal, liquid, etc.)
285
obtruent
any ound which i not a onorant i an obtruent, characterized
by a low degree of reonance
(oral top, fricative, and affricateregardle of whether they are voiced or v
oicele)
onet
any egment which occur before the nucleu of the yllable
palatalization
an articulation (mot commonly, a econdary articulation) in
volving a movement of the
tongue toward the hard palate (Crytal 1991)
parability of a tring
the ability of a tring of ound to be properly 
yllabified
phoneme
the minimal unit in the ound ytem of a language,
according to traditional phonological
theorie (Crytal 1991). In theorie which embrace the concept of diti
nctive feature a the
minimal unit, phoneme are conventionalized grouping of thoe feature.
phonetic
the area of linguitic that deal with (epecially) the pronuncia
tion of ound
phonological allomorph
allomorph of a morpheme which are derived from the
ame underlying form
by the application of a phonological rule (a oppoed to uppletive allomorph,
which have ditinct

underlying form)
phonological conditioning
When the ditribution of uppletive all
omorph i determined by
phonological feature of adjacent morpheme, the allomorphy i aid to
have phonological
conditioning.
phonological rule
a rule that applie to the phonological repreentat
ion altering the value of
phonological feature
phonology
the area of linguitic that deal with the ound a
nd (epecially) the ound ytem of
language
pitch
an auditory phonetic feature, correponding to ome degree
with the acoutic feature of
frequency (Crytal 1991) which i then analyzed a either intonation or tone
pitch accent
the prominence given to one yllable in a word which i tonal
in nature
Place
the node which dominate the node [Labial], [Coronal], and [Doral]
which i ued in rule where
all of the place feature pread
place aimilation
change in one ound by which it become more like an ad
jacent ound with repect
to place feature
portmanteau
a ubtype of multiple function formative which conflat
e two or more morpheme which
otherwie occur independently in the language.
For example, wa could
be conidered a
portmanteau in Englih that conflate the verb be and the Pat tene
(which i elewhere
manifeted a -d). The morpheme du in French i a portmanteau in Fre
nch that conflate de of
and le the (mac. g.).
potpoition
A potpoition i analogou to a prepoition in Englih (uch a
 for, to, from) except that it
follow rather than precede the word() with which it form a phrae
reduplication
a morphological proce that repeat ome part or p
attern of a morpheme typically a a
way of indicating omething like repetitive or plural
rewrite rule a rule that formally change one ymbol into another or other, 
uch a PP P NP and uch
a  z
rhyme
the contituent of the yllable compoed of the nucleu and the coda
onorant
a ound which i high in reonance, produced with a relatively free
airflow, and a vocal cord
poition uch that pontaneou voicing i poible, a in vowel, liqu
id, naal and lateral
(Crytal 1991) Glide / emivowel are alo onorant. The claification of gl
ottal top ha been a
matter of ome debate. In ome language it function like a onorant
and in other it apparently
doe not.
[onorant]
the feature which formally ditinguihe the cla of onorant (
ee above) from the cla of
obtruent
286
tray cononant
a cononant which cannot be included in a yllable during 
yllabification due to a limit
on the number of egment allowed in the yllable or due to retriction on feat
ure
tructure-preerving rule
the output of a tructure-preerving rule i imil

ar to the kind of tructure

that occur a input to the rule; for example, uch a rule doe not
produce ound which do not
occur in underlying form
uppletive allomorph
When a formative ha more than one hap
e in the lexicon, it i aid to have
uppletive allomorph. The plural morpheme in Englih ha (among other
) the uppletive
allomorph
z (a in dog[z]), en (a in oxen), and zero (a in heep).
See alo yntactic
conditioning, phonological conditioning and morphological conditioning.
upiciou pair
phone which, becaue of their phonetic im
ilarity, might be analyzed a being
allophone of a ingle phoneme
yllable
a phonological unit which conit of at leat a prom
inent element (nucleu, mot typically a
vowel) and optionally with preceding element (onet) and following element (co
da)
yllable template
the pattern by which tring of phoneme are evaluated to
determine how they may be
organized into yllable; the maximal yllable template pecifie the l
arget type of yllable
permitted
yntactic conditioning
Suppletive allomorphy i ometime (but not commonly)
dependent on yntactic
factor of the word. In that cae it i aid to have yntactic conditioning.
yntax
the area of linguitic that deal with the combination of word into
phrae and entence
tone
the ue of pitch to ditinguih lexical item
underlying form
the hape that a morpheme ha in the lexicon (bef
ore any phonological rule have
applied to it)
underpecification of feature in underlying form
the omiion of any valu
e for a given feature in the
lexical repreentation of a morpheme (becaue it i predictable from o
ther feature or from the
context in which the morpheme occur)
utterance
a tretch of peech about which no aumption have been made in ter
m of linguitic theory
(Crytal 1991) An utterance may be a ingle yllable or word, uch a Wow!, a phr
ae, a claue,
a entence, or a peech. (Becaue of the wide range of poible mean
ing for thi word, ome
linguit prefer to ue more pecific term uch a phonological phrae or inton
ational unit.
V
The ymbol V i often ued a a horthand for any vowel (yllabic ound).
[voice]
thi feature ditinguihe ound like [k] from ound like [g]; in
the former, the vocal fold are
not vibrating wherea in the latter (which are [+voice]) they are
voicing aimilation
the proce by which the value of the feature [voice] o
f one ound affect another
ound
vowel harmony
the aimilation of one vowel to another vowel
in the ame word (typically in an
adjacent yllable)
word
a grammatical unit, the internal characterization of which i typically
viewed a being within the
domain of a theory of morphology and the ditribution of which i vi
ewed a being within the
domain of a theory of yntax

word tructure rule

a rule that give information about the order of morpheme
 within a word, which are
optional Thee are very theory-pecific in form, but imple one are N N
tem
- Poeor, and V
Subject Agreement - Tene - (NEGATIVE) - Verb
tem
287
Appendix H: Topic Index
[anterior], 227
[ATR], 95, 229
[back], 94, 228
[cononantal], 225
[continuant], 90, 226
[high], 94, 228
[low], 94, 228
ablaut, 35
alienable cla, 12
allomorph, 13, 14
allophone, 71
anterior, 227
arbitrary clae, 14
arbitrary word clae, 15
aimilation, 74
autoegmental, 216
back, 228
bleeding order, 189
coalecence, 177
complementary ditribution, 121
cononant mutation rule, 35
continuant, 126
contour egment, 160
contour tone, 213
contrat in identical environment, 113
contrat in noninfluencing environment, 114
coronal, 64, 227
deletion, 176
derivation, 35, 56
derived environment, 80
diimilation, 104
ditributed, 227
ditribution, 113
doral, 64, 227
edge phenomena, 145
equipollent, 225
feature, 40, 55
feature geometry, 230
feature, 40, 65, 71, 74, 113, 217, 224
feeding order, 189
formative, 2, 7
free variation, 122
heavy yllable, 201
inalienable cla, 12
intonation, 204
labial, 64, 227
labialization, 84, 124
lateral, 227
lexicon, 1
link, 160
major cla feature, 225

major place feature, 64

manner aimilation, 126
manner feature, 225
maximal yllable template, 151
minimal pair, 113
minor rule, 35, 37
morpheme, 2
morpheme cut, 8
morphological conditioning, 18
morphology, 2, 7
multiple function formative, 29
mutually excluive environment, 121
naal, 226
naalization, 88, 125
natural clae, 40
noncontrative, 121
notation, 74
Obligatory Contour Principle, 104
obtruent, 61, 225
OCP, 104
onet, 152
optional element, 1
orthography, 231
palatalization, 83, 124
paradigm, 8
parenthee, 1
phoneme, 71
phonetic trancription, 40
phonological allomorph, 55
phonological conditioning, 16
pitch, 204
pitch-accent, 213
Place, 75, 139
place aimilation, 64, 99, 123
portmanteau, 29
privative, 225
reduplication, 44
rewrite rule, 1
rhyme, 152
round, 227
rule ordering, 188
lah dah notation, 16
onorant, 61, 225
pecial yllable, 153, 154
pread, 214
preading, 74
tray cononant, 172
tre, 168
Stre, 200
trident, 227, 228
tructure preerving rule, 79
uppletion, 19
uppletive allomorph, 13, 55
uppletive allomorphy, 15, 16, 18
upiciou pair, 134
yllabification, 152
yllable, 138, 145, 146, 151, 152, 154
yllable tructure contraint, 156
yntax, 1
timing keleton, 160

timing tier, 160

tone, 212, 214
tone feature, 217
tone language, 204
tone, 204
tranitional, 83, 125, 171
underlying form, 36, 56, 182
underpecification, 139
utterance, 145
voice, 40, 55, 226
voiced, 40
voicele, 40
voicing aimilation, 55, 56
vowel feature, 94, 140
vowel harmony, 93
word, 145
word tructure, 8
word tructure rule, 8
y-diagram, 121