What is assimilation?

Peter Jurgec
[email protected]

July 25, 2011

About this course

This course examines a very common cross-linguistic phenomenon—assimilation. Two
main questions are addressed. How can assimilation patterns differ from one another? Is
it possible to come up with a unified analysis all assimilation patterns? We will examine
many different types of assimilation, focusing on vowel, nasal, consonant harmony and
consonant–vowel interactions. This empirical survey motivates the analysis, which should
be able to capture the attested patterns and exclude the unattested ones. The analysis
is based on two well-established phonological theories: Autosegmental Phonology and
Optimality Theory.

Course Overview
Monday What is assimilation?
Tuesday Icy targets
Wednesday Transparency and blocking
Thursday Parasitic assimilation
Friday Positional effects

Today’s Outline
(1) a. What is assimilation?
b. What parameters are involved in assimilation?
c. How to capture these?

1 What is assimilation?
(2) Assimilation is an alternation involving at least two segments. One of these seg-
ments (the target) alternates in the presence of the other segment (the trigger),

1
Peter Jurgec Assimilation – EGG 2011

but not otherwise. The target acquires a phonological property of the trigger.
This phonological property can be characterized in terms of phonological features.
In the simplest of cases, a single phonological feature of a trigger affects a target.
(3) Example: Voicing assimilation in Russian (Padgett 2002:2)
a. ot-jehatj ‘to ride off’
ot-stupitj ‘to step back’
j
od-brosit ‘to throw aside’
b. pod-nesti ‘to bring (to)’
j
pot-pisat ‘to sign’
pod-ZetS ‘to set fire to’
(4) Some observations:
a. Assimilatory property: obstruent voicing
b. Segments involved: obstruents
c. Directionality: rightwards
(5) Three variables in assimilation:
a. Spreading feature
b. Targeted structure
c. Domain

2 Spreading feature
(6) The phonological property in assimilation can be construed in terms of one (or
more) spreading feature(s).
(7) If we look at assimilation patterns, we see that many features can spread.
(8) Example 1: Nasal harmony in Applecross Gaelic (Ternes 1973:134,135)
"ãh̃ũc̃¸ ‘neck’
"s̃Õh̃ı̃ ‘tame’
f̃r̃ı̃ã;ṽ ‘root.pl.’
kh O"ṽı̃ã;t ‘how much/many?’
tãṽ ‘ox, stag.pl’
˚
str̃ãı̃;G̃ ‘to be luxurious’
h ˚
"k Õı̃s̃paxk ‘wasp’
h
"t ãh̃ũs̃k ‘fool’
(9) Nasalization is the property of a stressed vowel which targets continuants right-
wards until the process is terminated by a stop. Similarly, nasalization also targets
the onset of the stressed syllable, but not if it is a stop.
(10) This pattern is similar to voicing assimilation in Russian in two respects. First,
the alternation is triggered by some phonological property—a feature of the trig-
ger. Recall that in Russian (3), this property is the value of voicing of an onset
obstruent. In Applecross, on the other hand, it is the nasality of a stressed vowel.
Second, the spreading feature affects adjacent segments. In Russian, voicing af-
fects all obstruents in the (immediately preceding) coda, but not obstruents in
the onset of the preceding syllable (cf. [pod-ZetS] vs. [pot-pisatj ]). In Applecross,
nasality affects all following continuants, but no segment across a stop. In other

2
Peter Jurgec Assimilation – EGG 2011

words, both processes involve a contiguous string of segments.

(11) Example 2: Emphasis spread in Southern Palestinian Arabic (Davis 1995:473−474)1
BALLAAS ‘thief’
èADD ˙ ‘luck’
˙˙
PABSAT ‘simpler’
BAAS ˙ ‘bus’
MAJAS ˙ SASiS ‘it didn’t become solid’
Tiin-ak˙ ˙ ˙ ‘your mind’
˙ Saan
QAT ‘thirsty’
˙
(12) Emphasis spread in Southern Palestinian Arabic (Davis 1995; Zawaydeh 1999;
Watson 1999, 2002; henceforth, SPalestinian) also affects a contiguous string
of segments. In this process, some consonants cause all preceding segments to
become pharyngealized.
(13) Example 3: Front/back harmony in Finnish (Ringen 1975/1988:77; Ringen &
Heinämäki 1999:305)
næh-kø:n ‘see-direct.sg’ tul-ko:n ‘come-direct.sg’
næk-ø ‘sight’ tul-o ‘coming’
pøytæ-næ ‘table-essive’ poutA-nA ‘fine weather-essive’
(14) Not all cases of assimilation involve a contiguous string of segments. Vowel har-
mony is an alternation which affects only vowels, while consonants are typically
ignored. Finnish (Ringen 1975/1988; Kiparsky 1981; Ringen & Heinämäki 1999)
suffix vowels alternate depending on the root vowels: suffix vowels are front after
front root vowels, and back after back root vowels. The feature for vocalic place
originates from the root and targets the suffix; it affects only vowels.

(15) In sum, different features assimilate.

3 Targeted structure
(16) So far I have looked at assimilation from the perspective of the trigger. Now
let us turn to targets by showing some of the variation with respect to what
segments can act as targets.
(17) The same spreading feature may target different classes of segments, which can
be characterized in terms of another feature. This suggests that assimilation
involves at least two variables: a spreading feature and a targeted structure.
(18) Example 1: Nasal harmony in Yaka (Hyman 1995:6,9)
a. tsub-idi ‘roam’ tsum-ini ‘sew’
kud-idi ‘chase’ kun-ini ‘plant’
kik-idi ‘obstruct’ wun-ini ‘murmur’
b. mak-ini ‘climb’ finuk-ini ‘sulk’
nik-ini ‘grind’ miituk-ini ‘sulk’
1
Pharyngealized segments are capitalized.

3
Peter Jurgec Assimilation – EGG 2011

(19) In (18-a), the perfective suffix in Yaka is usually realized as [-idi]. However, when
there is a nasal sonorant in the root, the suffix surfaces with a nasal sonorant as
[-ini]. This also happens when the triggering nasal is not at the right edge of the
root, which is shown in (18-b). As observed by Hyman (1995), nasality targets
only voiced consonants, ignoring all other segments.
(20) This pattern contrasts with the one found in Applecross, where intermediate
vowels are affected. The two languages are identical in terms of the spreading
feature, but they differ in what segments are targeted.
(21) Example 2: Tongue root harmony in Twi (Berry 1957:127−128,130)
biri ‘black’ o-biri ‘3p-black’
bIrI ‘red’ O-bIrI ‘3p-red’
firi ‘lend, borrow’ mi-be-firi-i ‘1p-fut-borrow-it’
fIrI ‘fail, miss’ mI-bE-fIrI-I ‘1p-fut-miss-it’
(22) In Twi (23), the affix vowels depend on the root vowels (Berry 1957; Painter
1973). Tense root vowels may occur with tense affix vowels, while lax root vowels
occur with lax affix vowels; consonants are unaffected.
(23) Recall that in SPalestinian emphasis spread (11), pharyngealized consonants af-
fect the preceding segments. Most analyses (Davis 1995; McCarthy 1997) assume
that the spreading feature in this case is the one responsible for tongue root re-
traction. In SPalestinian, this spreading feature affects all preceding segments
(consonants and vowels).
(24) If we compare Twi tongue root harmony and emphasis spread in SPalestinian,
we see that both involve the same spreading feature. The two languages crucially
differ in terms of what segments are targeted.
(25) Example 3: Palatalization
a. Czech (Rubach 2007:107)
plot ‘fence-nom.sg’ plotj -E ‘-loc.sg’
vod-a ‘water-nom.sg’ vodj -E ‘-loc.sg’
b. Irish (Nı́ Chiosáin 1994:97)
ahnj i:n ‘recognizes’ ahnj i:Nj -kj i:ro:g ‘a beetle recog.’
gan ‘without’ giNj -xj i:l ‘without sense’
c. Karaim (Kowalski 1929; Nevins & Vaux 2004:178–179)
suv-dAn ‘water-abl’ khj unj-dj Anj ‘day-abl’
boë-uS-uv-tSu ‘helper’ thj uzj -vj -tSj u-dj Anj ‘from the author’
(26) Recall that in Finnish (13), the root vowel determines whether the suffix vowel
will be front or back. The feature responsible for frontness/backness of root
vowels affects suffix vowels.
(27) Many other languages show alternations in which a front vowel affects a consonant
(25).
a. In Czech, a front vowel triggers an alternation that affects the secondary
articulation of the immediately preceding coronal. When followed by a front
vowel, coronals become palatalized.
b. In Irish, a palatalized dorsal consonant affects the preceding nasal.
c. Karaim exhibits palatalization of consonants, leaving intermediate (back)
vowels unaffected.

4
Peter Jurgec Assimilation – EGG 2011

(28) We can conclude that the same phonological feature is responsible for vocalic
frontness and secondary palatalization. Nevertheless, individual languages can
vary in terms of what segments are targeted.

(29) The three examples strongly suggest that the targeted structure in any assimila-
tion process is not predictable from the spreading feature. Thus, an analysis of
an assimilation process must include at least two variables.
These two variables are independent, as the relationship between them is not
entirely predictable.

4 Domain
(30) Prosodic and morphological domains are known to affect many phonological pat-
terns (McCarthy & Prince 1993b). So, it is unsurprising that domains also influ-
ence assimilation.
(31) So far, we have already seen how assimilation processes may differ with respect
to directionality.
(32) Nasal harmony in Sundanese (Robins 1957:91,95)
mãro ‘to halve’
ñı̃ãr ‘to seek’
Nũliat ‘to stretch (intr.)’
kumãh̃ã ‘how’
Nãjak ‘to sift’
mãwur ‘to spread’
mõlohok ‘to stare’
(33) In Sundanese, any vowel following a nasal sonorant is nasalized (Robins 1957;
Langendoen 1968; van der Hulst & Smith 1982; Cohn 1990, 1993; Piggott 1992;
Piggott & van der Hulst 1997; Benua 1997; Walker & Pullum 1999; Walker
1998/2000). Consonants cannot become nasal. Nasalization is triggered by a
nasal sonorant {m, n, ñ, N} and applies rightwards until it encounters a conso-
nant.
(34) Nasal harmony in Capanahua (Loos 1969:177,178)
põ̃ãn ‘arm’
bõõn ‘hair’
bı̃mi ‘fruit’
wWrãnwW ‘push it’
bãw̃ı̃n ‘catfish’
(35) In contrast, Capanahua displays leftward nasalization (Loos 1969; Halle & Vergnaud
1981; van der Hulst & Smith 1982; Safir 1982; Piggott 1987; Piggott & van der
Hulst 1997; Piggott 2003; Walker 1998/2000). This assimilation process is trig-
gered by a nasal sonorant stop and applies leftwards, targeting vowels and glides.
(36) By comparing Sundanese and Capanahua we see that assimilation may differ with

5
Peter Jurgec Assimilation – EGG 2011

respect to directionality. Some assimilation processes apply rightwards (Sun-

danese), while others apply leftwards (Capanahua).
(37) One way to look at this cross-linguistic variation is to say that directionality is a
separate parameter. This conclusion, however, is slightly misleading, as becomes
evident if we compare the following three types of grammars.
(38) Three alternative grammars
a. Grammar A contains both directionality and domains as assimilation vari-
ables.
b. Grammar B contains only domains (but no directionality)
c. Grammar C contains only directionality (but no domains).
(39) In light of the reviewed data, grammar C is not feasible. This is because many as-
similation patterns terminate at the boundary of some morphological or prosodic
domain. For example, Finnish vowel harmony (13) is restricted to prosodic words.
(40) This leaves us with grammars A and B.
(41) In an overwhelming majority of cases the two grammars do not make different
predictions. If so, parsimony prefers grammar B (that contains only domains)
over grammar A (that contains directionality in addition to domains).
(42) Furthermore, if directionality were an independent variable in assimilation, we
would predict at least one case of assimilation through any domain boundary
(total assimilation). For example, a nasal sonorant would trigger nasalization
of all subsequent segments (i.e., even across intonational phrase and sentence
boundaries). We know of no language with total assimilation.
(43) Grammar A can generate total assimilation since it can specify directionality
without reference to a domain. Grammar B, on the other hand, can only specify
directionality via a domain edge, and total assimilation is not restricted within
any domain. Hence, grammar A has to be rejected over grammar B.
(44) Directionality appears to be only epiphenomenal; assimilation is sensitive only to
specific prosodic/morphological boundaries (Nespor & Vogel 1986; Zec 1988/1994;
Peperkamp 1997).
(45) This conclusion about domains has at least one potential challenge. Many cases
of assimilation are bidirectional. Nasalization in Applecross (8) is of this type.
(46) Is bidirectional assimilation a combination of rightward and leftward assimila-
tion?
(47) Evidence for this comes from languages with directionality asymmetries.
(48) An example comes from Somali vowel harmony (Andrzejewski 1955; Saeed 1993,
1999; Krämer 2003).
(49) Example: Somali vowel harmony (Andrzejewski 1955:569,570; Saeed 1993, 1999;
Krämer 2003)
a. Leftward assimilation within an Intonational Phrase
wA: sA:n fArAs wæ: sæ:n dibi
dm hide horse.gen dm hide bull.gen
‘It is a horse’s hide.’ ‘It is a bull’s hide.’

6
Peter Jurgec Assimilation – EGG 2011

b. Leftward, not rightward, assimilation

hilib kA: kArI hilib kæ: i:bs8
meat dem cook meat dem buy
‘Cook that meat.’ ‘Buy that meat.’
c. Rightward assimilation to the following clitic
mA SAbE:l bA: mæ libæ:è bæ:
qm leopard foc qm lion foc
‘Is it a leopard?’ ‘Is it a lion?’
(50) What do the Somali data suggest?
(51) Directionality asymmetries
Somali exhibits disparities in leftward and rightward spreading. The Somali data
strongly suggest that bidirectional assimilation consists of two separate unidirec-
tional assimilations, each with its own domain of application.
(52) Domains
Somali exhibits vowel harmony that affects a domain much larger than the word.
Andrzejewski (1955) reports up to ten-word sequences with exclusively lax or
tense vowels. This pattern appears to be quite close to total assimilation dis-
cussed above. Somali appears to have directionality as independent variable that
can apply across any domain boundary.
However, such a conclusion turns out not to be true. In particular, Somali vowel
harmony never traverses pauses. Pauses are not random, but indicative of a
prosodic domain, which makes Somali directly parallel to similar characteristics
found in other languages. Nespor & Vogel (1986), Lahiri & Evers (1991) offer
additional evidence that these restricted positions of pauses relate to a prosodic
domain, such as the Intonational Phrase (IP). IPs are known to have effects on
assimilation and other alternations (Selkirk 1980; Lahiri & Evers 1991).
(53) Several other languages exhibit leftward/rightward disparities in assimilation:
a. Emphasis spread (pharyngealization) in Southern Palestinian Arabic is un-
bounded within a phonological word but only leftwards. In the opposite
direction, emphasis is blocked by {i, j, S, dZ} (Davis 1995, see also Zawaydeh
1999, Watson 1999, 2002).
b. In Epena Pedee, [nasal] spreads unboundedly rightwards, but only within
the same syllable leftwards (Harms 1985, 1994, Walker 1998/2000).
c. In Vata, [ATR] spreads from roots to suffixes. In the opposite direction,
[ATR] spreads optionally across word boundaries, but only to the first vowel
(Kaye 1982).
d. Ikwere nasal harmony shows a difference in the behavior of nasal sonorant
stops, which are regular targets in rightward, but not in leftward assimilation
(Clements & Osu 2003, 2005).
(54) Sometimes, assimilation applies across domains. In Catalan less formal speech,
the place of articulation of a non-continuant affects the preceding nasal both
morpheme internally, across morphemes and across word boundaries (Wheeler
1979, 2005).

7
Peter Jurgec Assimilation – EGG 2011

(55) Nasal place assimilation across word boundaries in Catalan (Wheeler 2005: 184)
só[m m]olts ‘they are many’
só[m p]ocs ‘they are few’
só[N g]rossos ‘they are large’
(56) The domain of assimilation is not immediately apparent. However, if our reason-
ing is correct, the relevant domain is larger than a prosodic word, for example a
phonological (or intonational) phrase. The reason why segments preceding the
word-final coronal nasal are not affected is because vowels terminate assimilation.
Thus the closest the place of articulation can get to the left edge of a phonological
phrase is one segment to the left (codas containing two nasal sonorants are illicit
in Catalan).

(57) To summarize, domains are the third parameter in assimilation.

5 Assimilation as feature spreading

5.1 Feature spreading
(58) One standard analysis of assimilation is autosegmental spreading.
(59) In Autosegmental Phonology (Goldsmith 1976, 1990; Clements 1976/1980, 1985a;
Kiparsky 1981), features are represented as autosegments that may be associated
with nodes.
(60) The representations are nonlinear. The highest mother-node is a root node, which
establishes linearity across segments. In (61), we see two features, [F] and [G],
associated with a single root node (×).
(61) An association line represents a relationship between a feature and a root node; a
segment consists of a root node and the features associated with that root node.
(62) Autosegmental representations
×

[F]
[G]
(63) Segmental alternations may also be represented in terms of autosegments. In
Autosegmental Phonology, assimilation is associating (or linking) a spreading
feature with a target root node.
(64) This process is also termed feature spreading: a feature spreads from a trigger to
a target.
(65) Assimilation as feature spreading
×1 ×2

[F]

8
Peter Jurgec Assimilation – EGG 2011

5.2 Assumptions about representations

(66) Subsequent work on autosegmental representations has claimed that features
are organized in a particular fashion (Goldsmith 1976; Halle & Vergnaud 1980;
Archangeli 1985; Clements 1985a,b, 1991; Clements & Hume 1995; Kaye et al.
1985; Steriade 1987, 1995; McCarthy 1988; Padgett 1991/1995; Odden 1991,
1994; Halle 1995; Sagey 1990; Harris & Lindsey 1995; Halle et al. 2000; Morén
1999/2001, 2003, 2006, 2007; Blaho 2008, among many others).
(67) In this model, some features are directly associated with the root node, while
others are not.
(68) Here, I assume a less restrictive model, in which features are not organized in
any particular fashion; what matters is whether a feature is associated with a
particular root node or not.
(69) Hence, the restrictions on feature spreading will never depend on organization of
features within a segment.
(70) I will use phonological features that are (i) phonetically motivated, (ii) universal
and (iii) privative.

6 Alignment constraints
(71) Autosegmental representations are referred to by OT constraints.
(72) Feature spreading is enforced by markedness constraints. The markedness con-
straint has to be able to capture the three parameters of assimilation (spreading
feature, targeted structure, and domain).
(73) One established approach to assimilation is to extend Generalized Alignment
(McCarthy & Prince 1993a) to segmental features (Kirchner 1993; Smolensky
1993; Cole & Kisseberth 1995; Itô & Mester 1995; Akinlabi 1996; Pulleyblank
1996; Golston 1996; McCarthy 1997; Ringen & Vago 1998; Archangeli & Pulley-
blank 2002, among many others).
(74) The logic behind such analyses is simple: an alignment constraint prefers an
output in which a feature is aligned with an edge of a phonological domain such
as a syllable, prosodic word, or phonological phrase.
(75) As an example, let us consider the constraint Align([nasal], R; PWd, R) in (76).
This constraint penalizes outputs containing oral segments after a nasal segment
within a Prosodic Word.
(76) Align([nasal], R; Prosodic Word, R)
For every [nasal] autosegment there must be a Prosodic Word, such that the
rightmost segment associated with [nasal] is also the rightmost segment of a
Prosodic Word.
(77) The constraint in (76) contains four variables: a single feature, a domain, and
two specified edges. However, the data reviewed suggest that feature spreading
actually involves two features.
(78) This empirical fact is not consistent with the one feature plus one domain con-
straint template above.

9
Peter Jurgec Assimilation – EGG 2011

(79) One solution would be to propose other constraints. For example, feature co-
occurrence constraints could exempt a class of segments from being targeted.
The problem with this solution is that it can exempt any segment, which predicts
many unattested patterns.
(80) An alternative would be to propose a revision of alignment constraints, and this
is the option I take. The revised version should include at least three variables:
one spreading feature, one targeted structure, and one domain. It turns out that
such a template has been proposed for prosody by Hyde (2008).
(81) Hyde (2008) proposes markedness constraints that have most characteristics of
classical alignment constraints. In particular, the constraints prefer outputs in
which two categories (features, domains) are aligned with one another.
(82) Hyde’s constraints assign violation marks to sets of violating pairs or triplets of
categories.
(83) This means that for a given input the number of violation marks will be dependent
on both the aligned categories and the offending categories.
(84) One type of Hyde’s alignment schema is presented in (85). This constraint assigns
a violation mark for every triplet hCat1, Cat2, Cat3i, if and only if Cat2 precedes
Cat3 within Cat1.
(85) Right edge distance sensitive alignment schema (Hyde 2008)
a. *hCat1, Cat2, Cat3i / Cat1
Cat2 Cat3
b. Assign a violation mark for every triplet hCat1, Cat2, Cat3i, iff
Cat1 is associated with Cat2 and Cat3
and
Cat2 precedes Cat3.
(86) The constraint in (85) consists of two parts. The first one is the violating triplet
hCat1, Cat2, Cat3i, while the second one is the arrangement of these categories.
(87) The relationship between Cat1 and the two other categories is that of association.
(88) The relationship between Cat2 and Cat3 can thus be characterized in terms of
precedence, which is a widely accepted temporal relation in phonology.
(89) As we have seen, assimilation patterns can be characterized in terms of three
parameters: a spreading feature, a targeted structure, and a domain.
(90) These three categories are entirely consistent with Hyde’s alignment schema,
which also contains three variables: the first variable is associated with the other
two, which are in turn in a precedence relation.
(91) In (92) we see an implementation of Hyde’s template that captures feature spread-
ing. The feature alignment constraint assigns a violation mark for every triplet
hDomain, [F], [G]i if and only if (i) the Domain is associated with [F] and [G],
and (ii) [F] precedes [G].
(92) Feature alignment
*hDomain, [F], [G]i / Domain
[F] [G]
(93) However, there is an important difference between a precedence relation between

10
Peter Jurgec Assimilation – EGG 2011

two root nodes and a precedence relation between two different features.
(94) In Autosegmental Phonology, precedence is established between like categories
(Goldsmith 1976). For any two root nodes, there is a unique precedence relation:
one always precedes the other. Similarly, for any two instances of the same
feature, one always precedes the other.
(95) For now, we will assume that a root node that is association excludes precedence.
(96) Featural precedence
[G] f-precedes [H], iff
(i) ∃×i associated with [G] but not with [H],
and
(ii) ∃×j associated with [H] but not with [G],
and
(iii) ×i precedes ×j .
(97) F-precedence is a crucial ingredient of feature alignment constraints. In (98), I
repeat the featural alignment template and complement it with a definition.
(98) Featural alignment
a. *hDomain, [G], [H]i / Domain
[G] [H]
b. Assign a violation mark for every triplet hDomain, [G], [H]i, iff
the Domain is associated with [G] and [H]
and
[G] f-precedes [H].
(99) I will show that alignment constraints similar to the one in (99) can model feature
spreading better than other approaches to feature spreading. This is despite the
fact that they require an additional concept of f-precedence.
(100) Another good argument in favor of alignment constraints based on Hyde (2008)
is purely formal. The constraint template in (100) is categorical rather than
gradient.
(101) Gradient constraints have been shown to generate many unattested patterns
(McCarthy 2003; Hyde 2008). One such example is the Midpoint Pathology,
which involves a pattern in which stress will fall on the syllable furthest apart
from both edges of a prosodic word.

11
Peter Jurgec Assimilation – EGG 2011

7 Applecross Gaelic nasal harmony

(102) Recall the data in (103).
(103) Nasal harmony in Applecross Gaelic (Ternes 1973:134,135)
"ãh̃ũc̃¸ ‘neck’
"s̃Õh̃ı̃ ‘tame’
f̃r̃ı̃ã;ṽ ‘root.pl.’
h
k O"ṽı̃ã;t ‘how much/many?’
tãṽ ‘ox, stag.pl’
˚
str̃ãı̃;G̃ ‘to be luxurious’
h ˚
"k Õı̃s̃paxk ‘wasp’
h
"t ãh̃ũs̃k ‘fool’
(104) The spreading feature in Applecross is [nasal] and the domain is the prosodic
word. The targeted structure is thus simply a root node.
(105) Hyde’s constraint schema allows for these three categories to be joined within
a single constraint, defined in (106). Notice that f-precedence is applicable to a
relationship between a root node and a feature, when they are not associated
with one another. For expositional purposes, I use an abbreviated notation
*PWd[nasal, ×] (“A root node must not f-precede [nasal] within a PWd”). I
will use similar abbreviations throughout this thesis.
(106) *PWd[nasal, ×]
a. *hPWd, [nasal], ×i / PWd
[nasal] ×
b. Assign a violation mark for every triplet hP W d, [nasal], ×i, iff
P W d is associated with [nasal] and ×
and
[nasal] f-precedes ×.
(107) In Applecross, the alignment constraint *PWd[nasal, ×] outranks the relevant
faithfulness constraint.
(108) From the viewpoint of Autosegmental Phonology what is being added when a
feature spreads is not a feature, but rather an association.
(109) Faithfulness constraints to associations are widely used in OT literature that
makes use of autosegmental representations—see Morén (1999/2001) for a full
discussion of faithfulness constraints for associations, and Blaho (2008) for an
extension to segmental features.
(110) In this particular example, the faithfulness constraint being violated by candi-
dates with spreading is a constraint against linking the feature [nasal] with a
root node. The constraint DepLink[nasal] in (111) is violated once by every
association line to [nasal], which is present in the output, but not in the input.
(111) DepLink[nasal] (cf. Itô et al. 1995; Myers 1997; Lombardi 1998; Morén 1999/2001;
Archangeli & Pulleyblank 2002; Blaho 2008)
Let ×i be an input root node and ×o its output correspondent. Assign a viola-
tion mark, iff ×o is associated with the feature [nasal] and ×i is not.
(112) The effect of the two constraints can be seen in tableau (113).

12
Peter Jurgec Assimilation – EGG 2011

(113) "ãh̃ũc̃¸ ‘neck’

[nas]
/ ã h u ç / *ω[nasal,×] DepLink[nasal]
[nas]
hω,[nas],hi! hω,[nas],ui,
a. ã h u ç hω,[nas],çi
[nas]
b. ã h̃ u ç hω,[nas],ui! hω,[nas],çi *
[nas]
c. ã h̃ ũ ç hω,[nas],çi! **
[nas]
d. ☞ ã h̃ ũ c̃¸ ***

8 Finnish vowel harmony

(114) One prediction of the current approach is that assimilation patterns may differ
in terms of the spreading feature, the targeted structure, and the domain.
(115) It is not hard to imagine many other constraints that are similar, but not iden-
tical to *PWd[nasal, ×]. One option is to modify the spreading feature. For
example, emphasis spread in SPalestinian (11) can be analyzed as regressive
spreading of [RTR] that targets all segments (Davis 1995). If so, the relevant
alignment constraint in SPalestinian contains [RTR] rather than [nasal].
(116) Another variable that can be changed is the targeted structure. I will now show
a constraint that differs from *PWd[nasal, ×] in two variables: the spreading
feature and the targeted structure. This is needed in the analysis of vowel
harmony in Finnish (13), where only vowels act as targets, while consonants
remain unaffected.
(117) Remember that Finnish—(13), repeated in (118) below—exhibits front/back
vowel harmony. Front root vowels come with front suffix vowels, while back
root vowels come with back suffix vowels; consonants are unaffected.
(118) Front/back harmony in Finnish (Ringen 1975/1988:77; Ringen & Heinämäki
1999:305)
næh-kø:n ‘see-direct.sg’ tul-ko:n ‘come-direct.sg’
næk-ø ‘sight’ tul-o ‘coming’
pøytæ-næ ‘table-essive’ poutA-nA ‘fine weather-essive’
(119) Here, I will analyze Finnish vowel harmony as spreading of the feature [back].
(120) Disclaimer: I will also slightly abstract away from the data and simply assume
that all vowels participate in vowel harmony (which is not the case). This will
be revised on Wednesday.

13
Peter Jurgec Assimilation – EGG 2011

(121) Comparison between Finnish and Applecross

Finnish Applecross
Spreading feature [back] [nasal]
Targeted structure vowel root node
Domain word word
Directionality root-to-suffix bidirectional
(122) The alignment constraint schema predicts these typological distinctions per-
fectly. In Finnish, the relevant constraint *PWd[back, vowel] in (123) contains
a vowel instead of a root node. Consonants never appear in the constraint-
violating triplets of this constraint.
(123) *PWd[back, vowel]
a. *hPWd, [back], voweli / PWd
[back] vowel
b. Assign a violation mark for every triplet hP W d, [back], voweli, iff
P W d is associated with [back] and vowel
and
[back] f-precedes vowel.
(124) Disclaimer: We will talk more about transparency on Wednesday.
(125) Tableau (125) shows the effect of *PWd[back, vowel].
(126) tulo ‘coming’
[bk]
/t u l-ø/ *ω[back,vowel] DepLink[back]
[bk]
a. t u l ø hω,[back],øi!
[bk]
b. ☞ t u l o *
[bk]
c. t u lG o **!

(127) We have now seen the effect of two similar, yet crucially different alignment
constraints—*PWd[nasal, ×] and *PWd[back, vowel]. They differ from one
another in two variables: the spreading feature (which may be [nasal] or [back])
and the targeted structure (which may be × or a vowel).
(128) The alignment constraint schema can be similarly modified further to include
other domains, other spreading features and targeted structures. This is consis-
tent with the cross-linguistic variation in assimilation, which was demonstrated
in the empirical part of this chapter.

14
Peter Jurgec Assimilation – EGG 2011

9 Conclusions
(129) Assimilation patterns may differ in three basic variables: the spreading feature,
the targeted structure, and the domain.
(130) I propose an analysis of feature spreading within Optimality Theory.
(131) This is based on a significant extension of a familiar approach that combines
alignment with faithfulness constraints specific to features.
(132) I demonstrate that all three basic parameters can be modeled using a single
class of markedness constraints.

References
Akinlabi, Akinbiyi (1996). Featural affixation. Journal of Linguistics 32. 239–289.
Andrzejewski, B. W. (1955). The problem of vowel representation in the Isaaq dialect of
Somali. Bulletin of the School of Oriental and African Studies, University of London
17. 567–580.
Archangeli, Diana (1985). Evidence for coplanar representation in nonlinear phonology.
Linguistic Inquiry 16. 335–372.
Archangeli, Diana & Douglas Pulleyblank (2002). Kinande vowel harmony: domains,
grounded conditions and one-sided alignment. Phonology 19. 139–188.
Benua, Laura (1997). Transderivational Identity: Phonological Relations between Words.
Ph.D. dissertation, University of Massachusetts Amherst, Amherst.
Berry, J. (1957). Vowel harmony in Twi. Bulletin of the School of Oriental and African
Studies, University of London 19. 124–130.
Blaho, Sylvia (2008). The syntax of phonology: A radically substance-free approach.
Ph.D. dissertation, University of Tromsø, Tromsø.
Clements, George N. (1976/1980). Vowel Harmony in Nonlinear Generative Phonology.
Bloomington: Indiana University Linguistics Club.
Clements, George N. (1985a). Akan vowel harmony: a nonlinear analysis. In D. L. Goy-
vaerts (ed.) African Linguistics: Essays in Honor of W. M. K. Semikenke, Amsterdam:
John Benjamins. 55–98.
Clements, George N. (1985b). The geometry of phonological features. Phonology 2.
225–252.
Clements, George N. (1991). Place of articulation in consonants and vowels: a unified
theory. In Working Papers of the Cornell Phonetics Laboratory, Ithaca: Cornell Uni-
versity, vol. 5. 77–123.
Clements, George N. & Elizabeth V. Hume (1995). The internal organization of speech
sounds. In John A. Goldsmith (ed.) The Handbook of Phonological Theory, Blackwell.
245–306.
Clements, George N. & Sylvester Osu (2003). Ikwere nasal harmony in typological per-
spective. In Patrick Sauzet & Anne Zribi-Hertz (eds.) Typologie des langues d’Afrique
et universaux de la grammaire, Vol. II , Paris: L’Harmattan. 70–95.
Clements, George N. & Sylvester Osu (2005). Nasal harmony in Ikwere, a language
with no phonemic nasal consonants. Journal of African Languages and Linguistics 26.
165–200.
Cohn, Abigail C. (1990). Phonetic and Phonological Rules of Nasalization. Ph.D. disser-
tation, University of California, Los Angeles, Los Angeles.

15
Peter Jurgec Assimilation – EGG 2011

Cohn, Abigail C. (1993). The status of nasalized continuants. In Marie K. Huffman

& Rena A. Krakow (eds.) Phonetics and Phonology 2: Nasals, Nasalization, and the
Velum, San Diego, CA: Academic Press. 329–367.
Cole, Jennifer & Charles Kisseberth (1995). An Optimal Domains Theory of vowel har-
mony. Studies in Linguistic Sciences 24. 101–114. Available on Rutgers Optimality
Archive, ROA 22, http://roa.rutgers.edu.
Davis, Stuart (1995). Emphasis spreading in Arabic and Grounded Phonology. Linguistic
Inquiry 26. 465–498.
Goldsmith, John A. (1976). Autosegmental Phonology. Ph.D. dissertation, Massachusetts
Institute of Technology, Cambridge.
Goldsmith, John A. (1990). Autosegmental and Metrical Phonology. Oxford: Blackwell.
Golston, Chris (1996). Direct Optimality Theory: Representation as pure markedness.
Language 72. 713–748.
Halle, Morris (1995). Feature geometry and feature spreading. Linguistic Inquiry 16.
1–46.
Halle, Morris, Bert Vaux & Andrew Wolfe (2000). On feature spreading and the repre-
sentation of place of articulation. Linguistic Inquiry 31. 387–444.
Halle, Morris & Jean-Roger Vergnaud (1980). Three dimensional phonology. Journal of
Linguistic Research 1. 83–105.
Halle, Morris & Jean-Roger Vergnaud (1981). Harmony processes. In Wolfgang Klein
& Williem Levelt (eds.) Crossing the boundaries in linguistics: Studies presented to
Manfred Bierwisch, Dordrecht: Reidel. 1–22.
Harms, Phillip Lee (1985). Epena Pedee (Saija) nasalization. In Ruth Brend (ed.) From
Phonology to Discourse: Studies in Six Columbian Languages, Dallas, TX: Summer
Institute of Linguistics. 13–18.
Harms, Phillip Lee (1994). Epena Pedee Syntax . Arlington, TX: The Summer Institute
of Linguistics and The University of Texas at Arlington.
Harris, John & Geoff Lindsey (1995). The elements of phonological representation. In
Jacques Durand & Francis Katamba (eds.) Frontiers of Phonology: Atoms, Structures,
Derivations, London, New York: Longman. 34–79.
Hyde, Brett (2008). Alignment continued: Distance-sensitivity, order-sensitivity, and
the midpoint pathology. Ms. Available on Rutgers Optimality Archive, ROA 998,
http://roa.rutgers.edu.
Hyman, Larry (1995). Nasal consonant harmony at a distance: The case of Yaka. Studies
in African Linguistics 24. 5–30.
Itô, Junko & Armin Mester (1995). The core-periphery structure of the lexicon and
constraints on reranking. In Jill Beckman, Suzanne Urbanczyk & Laura Walsh (eds.)
University of Massachusetts Occasional Papers in Linguistics 18 , Amherst: GLSA,
University of Massachusetts. 181–209.
Itô, Junko, Armin Mester & Jaye Padgett (1995). Licensing and underspecification in
Optimality Theory. Linguistic Inquiry 26. 571–631.
Kaye, Jonathan (1982). Harmony processes in Vata. In Harry van der Hulst & Norval
Smith (eds.) The Structure of Phonological Representations (Part II), Dordrecht: Foris.
385–452.
Kaye, Jonathan, Jean Lowenstamm & Jean-Roger Vergnaud (1985). The internal struc-
ture of phonological elements: a theory of charm and goverment. Phonology Yearbook
2. 305–328.
Kiparsky, Paul (1981). Vowel harmony. Ms. Massachusetts Institute of Technology.

16
Peter Jurgec Assimilation – EGG 2011

Kirchner, Robert (1993). Turkish vowel harmony and disharmony: An Optimality The-
oretic account. Ms. University of Alberta. Available on Rutgers Optimality Archive,
ROA 4, http://roa.rutgers.edu.
Kowalski, Tadeusz (1929). Karaimische texte im dialekt von Troki . Cracow: Nakladem
Polskiej Akademji Umieje˛ tności.
Krämer, Martin (2003). Vowel Harmony and Correspondence Theory. Berlin and New
York: Mouton de Gruyter.
Lahiri, Aditi & Vincent Evers (1991). Palatalization and coronality. In Carole Paradis &
Jean-François Prunet (eds.) Phonetics and Phonology 2: Volume 2. The Special status
of Coronals, Internal and External Evidence, San Diego, CA: Academic Press. 79–100.
Langendoen, Terrence D. (1968). The London School of Linguistics. Cambridge, MA:
MIT Press.
Lombardi, Linda (1998). Evidence for Max Feature constraints from Japanese. In Haruka
Fukazawa, Frida Morelli, Caro Struijke & Yi-Ching Su (eds.) University of Maryland
Working Papers in Linguistics 7 , University of Maryland. 41–62. Available on Rutgers
Optimality Archive, ROA 247, http://roa.rutgers.edu.
Loos, Eugene Emil (1969). The phonology of Capanahua and its grammatical basis.
Summer Institute of Linguistics of the University of Oklahoma.
McCarthy, John J. (1988). Feature geometry and dependency. Phonetica 43. 84–108.
McCarthy, John J. (1997). Process specific constraints in Optimality Theory. Linguistic
Inquiry 28. 231–251.
McCarthy, John J. (2003). OT constraints are categorical. Phonology 20. 75–138.
McCarthy, John J. & Alan Prince (1993a). Generalized Alignment. In Geert Booij &
Jaap van Marle (eds.) Yearbook in Morphology 1993 , Dordrecht: Kluwer. 79–153.
McCarthy, John J. & Alan Prince (1993b). Prosodic Morphology I. Ms., University of
Massachusetts and Rutgers University, Amherst, and New Brunswick, NJ. Available
on Rutgers Optimality Archive, ROA 482, http://roa.rutgers.edu.
Morén, Bruce (1999/2001). Distinctiveness, Coercion and Sonority: A Unified Theory
of Weight. New York, London: Routledge. Available on Rutgers Optimality Archive,
ROA 349, http://roa.rutgers.edu.
Morén, Bruce (2003). The parallel structures model of feature geometry. In J. Brugman &
A. Rhiel (eds.) Working Papers of the Cornell Phonetics Laboratory, Cornell University,
vol. 15. 194–270.
Morén, Bruce (2006). Consonant-vowel interactions in Serbian: Features, representations
and constraint interactions. Lingua 116. 1198–1244.
Morén, Bruce (2007). The division of labor between segment-internal structure and
violable constraints. In Sylvia Blaho, Patrik Bye & Martin Krämer (eds.) Freedom of
Analysis? , Berlin: Mouton de Gruyter. 215–235.
Myers, Scott (1997). OCP effects in Optimality Theory. Natural Language and Linguistic
Theory 15. 847–892.
Nespor, Marina & Irene Vogel (1986). Prosodic Phonology. Dordrecht: Foris.
Nevins, Andrew & Bert Vaux (2004). Consonant harmony in Karaim. In Anikó Csir-
maz, Youngjoo Lee & MaryAnn Walker (eds.) MIT Working Papers in Linguistics 46:
The Proceedings of the Workshop on Altaic in Formal Linguistics, Cambridge, MA:
MITWPL. 175–194.
Nı́ Chiosáin, Máire (1994). Irish palatalisation and the representation of place features.
Phonology 11. 89–106.
Odden, David (1991). Vowel geometry. Phonology 8. 261–289.

17
Peter Jurgec Assimilation – EGG 2011

Odden, David (1994). Adjacency paramaters in phonology. Language 70. 289–330.

Padgett, Jaye (1991/1995). Stricture in Feature Geometry. Stanford, CA: CSLI Publi-
cations.
Padgett, Jaye (2002). Russian voicing assimilation, final devoicing, and the problem of
[v]. Ms. University of California Santa Cruz. Available on Rutgers Optimality Archive,
ROA 528, http://roa.rutgers.edu.
Painter, Colin (1973). Cineradiographic data on the feature ‘covered’ in Twi vowel har-
mony. Phonetica 28. 97–120.
Peperkamp, Sharon (1997). Prosodic Words. The Hague: Holland Academic Graphics.
Piggott, Glyne (1987). On the autonomy of the feature Nasal. In Anna Bosch, Barbara
Need & Eric Shiller (eds.) Chicago Linguistic Society (CLS) 23-2: Parasession on
Autosegmental and Metrical Phonology, Chicago: Chicago Linguistic Society. 223–238.
Piggott, Glyne (1992). Variability in feature dependency: The case of nasality. Natural
Language and Linguistic Theory 10. 33–77.
Piggott, Glyne (2003). Theoretical implications of segment neturality in nasal harmony.
Phonology 20. 375–424.
Piggott, Glyne & Harry van der Hulst (1997). Locality and the nature of nasal harmony.
Lingua 103. 85–112.
Pulleyblank, Douglas (1996). Neutral vowels in Optimality Theory: A comparison of
Yoruba and Wolof. Canadian Journal of Linguistics 41. 295–347.
Ringen, Catherine (1975/1988). Vowel harmony: Theoretical Implications. New York:
Garland.
Ringen, Catherine & Orvokki Heinämäki (1999). Variation in Finnish vowel harmony:
An OT account. Natural Language and Linguistic Theory 17. 303–337.
Ringen, Catherine & Robert Vago (1998). Hungarian vowel harmony in Optimality
Theory. Phonology 15. 393–416.
Robins, R. H. (1957). Vowel nasality in Sundanese, a phonological and grammatical
study. In Studies in linguistic analysis, Oxford: Blackwell. 87–103.
Rubach, Jerzy (2007). Feature geometry from the perspective of Polish, Russian, and
Ukranian. Linguistic Inquiry 38. 85–138.
Saeed, John (1999). Somali . Amsterdam: John Benjamins.
Saeed, John Ibrahim (1993). Somali Reference Grammar . Kensington, MD: Dunwoody
Press.
Safir, Ken (1982). Nasal spreading in Capanahua. Linguistic Inquiry 13. 689–694.
Sagey, Elizabeth (1990). The Representation of Features in Non-Linear Phonology: The
Articulator Node Hierarchy. New York and London: Garland.
Selkirk, Elisabeth (1980). Prosodic domains in phonology: Sanskrit revisited. In
Mark Aronoff & Mary-Louise Kean (eds.) Juncture: A Collection of Original Papers,
Saratoga, CA: Anma Libri. 107–129.
Smolensky, Paul (1993). Harmony, markedness, and phonological activity. Paper pre-
sented at Rutgers Optimality Workshop 1, Rutgers University. Available on Rutgers
Optimality Archive, ROA 87, http://roa.rutgers.edu.
Steriade, Donca (1987). Locality conditions and feature geometry. In Proceedings of
NELS 17 , Amherst: GLSA, University of Massachusetts. 595–618.
Steriade, Donca (1995). Underspecification and markedness. In John A. Goldsmith (ed.)
The Handbook of Phonological Theory, Oxford: Blackwell. 114–174.
Ternes, Elmar (1973). The Phonemic Analysis of Scottish Gaelic: Based on the Dialect
of Applecross, Ross-shire. Hamburg: Helmut Buske Verlag.

18
Peter Jurgec Assimilation – EGG 2011

van der Hulst, Harry & Norval Smith (1982). Prosodic domains and opaque segments in
Autosegmental Theory. In Harry van der Hulst & Norval Smith (eds.) The Structure
of Phonological Representations (Part II), Dordrecht: Foris. 311–336.
Walker, Rachel (1998/2000). Nasalization, Neutral Segments, and Opacity Effects. New
York: Garland.
Walker, Rachel & Geoffrey Pullum (1999). Possible and impossible segments. Language
75. 764–780.
Watson, Janet C. E. (1999). The directionality of emphasis spread in Arabic. Linguistic
Inquiry 30. 289–300.
Watson, Janet C. E. (2002). The Phonology and Morphology of Arabic. Oxford: Oxford
University Press.
Wheeler, Max W. (1979). Phonology of Catalan. Oxford: Blackwell.
Wheeler, Max W. (2005). The Phonology of Catalan. Oxford: Oxford University Press.
Zawaydeh, Bushra (1999). The phonetics and phonology of gutturals in Arabic. Ph.D.
dissertation, Indiana University, Bloomington.
Zec, Draga (1988/1994). Sonority Constraints on Prosodic Structure. New York: Gar-
land.

19