Words and their Internal Structure: Morphology (1)

Every speaker of a language knows words, indeed a great number of them. Part of our
linguistic knowledge thus consists of knowledge of words as well as a system of rules
regulating how words are formed.
To know a word is to know what it sounds like and to know what it means. As we have
discussed already, knowledge of the way a word sounds involves knowledge of a pair of
phonological representations (an Underlying Representation and a Surface Representation) as
well as knowledge of phonological rules. We wont discuss how the meaning of words are
coded in mental representations in much detail in this class (since it is a difficult topic about
which linguists still do not know all that much). Instead, we will focus on how the relation
between sound and meaning is established and what constitutes a word.
As we have discussed, speakers of English know, for example, that the UR of the word pen is
/pen/ and that its SR is [phe::n], as derived by the application of the Aspiration, Nasalization,
and Vowel Lengthening Rules. Speakers also know that this sound representation
corresponds to a particular meaning (that is, a concept or idea of some sort).
To explain this, we must postulate a level of linguistic representation where certain
phonological representations, such as /pen/, /bUk/, and /desk/, are matched together with
certain semantic representations (namely, a representation of meaning), for which we will
use the notation of double-square brackets (for example, PEN, where PEN, in all capital
letters, is simply a shorthand notation for whatever the semantic representation of the word
actually is, presumably in terms of some theory of semantic features, much like we use IPA
symbols as shorthand for sets of phonological features).
{/pen/, PEN}
{/bUk/, BOOK}
{/desk/, DESK}
A related fact is that when I say, for example, the sentence give me a pen please, this speech
event constitutes, acoustically, a continuous stream of sound. Yet speakers know that there
are five discrete words in the above sentence. In other words, they know that there are word
boundaries in such an expressing. We may be tempted to think that we have already solved
this problem by postulating discrete phonological representations in terms of discrete
phonological segments (ultimately sets of discrete features): Speakers can somehow translate
back-and-forth between the acoustic continuum and the phonological representation
[gwvmiphe::nphli:z], which is itself derived from the representation /gwvmipenphliz/ by way
of the application of various phonological rules, such as the Aspiration Rule, the Nasalization
Rule, and so on. But this does not tell us where one word starts and the next begins.
Ultimately, we need to indicate word boundaries in mental representation (we will use the
symbol # to indicate the beginning/end of a word):
[#gwv#mi##phe::n#phli:z#]

Can we say, then, that a word is simply the level of representation where there exists some
sort of systematic relation between sound and meaning? Obviously not, since there seem to
be several such levels. For example, there is clearly a systematic relation between the entire
string [#gwv#mi##phe::n#phli:z#] and some particular meaning. Suppose, however, that we
try to define words as the minimal level of linguistic representation where there is a
systematic relation between sound and meaning. Will that work?
Impossible! Why? Because, on one hand, it seems that impossible is indeed one word. On the
other hand, the word impossible can be further analyzed into two discrete parts, im and
possible, each with its own associated meaning.
That is to say that whatever the minimal unit of representation where sound is linked to
meaning turns out to be, impossible surely consists of two of them. At the same time, there is
an intuition that impossible is something (shall we call it a word?), as is possible, that im is
not! Consider also the following pairs of adjectives:
A
desirable
likely
happy
developed

B
undesirable
unlikely
unhappy
undeveloped

The meaning of each of the adjectives in the B-column is essentially the meaning NOT +
corresponding adjective in the A-column (for example, undesirable means not desirable,
unhappy means not happy, and so on). A popular English dictionary lists close to 3000
adjectives beginning with un, the meanings of all of which accord to the above pattern.
The minimal units of language that exhibit a systematic pairing of sound and meaning thus do
not correspond to words, but to in fact something more basic. These basic, minimal units of
sound and meaning are called morphemes. That, of course, now leaves us without a clear
definition of word. For the moment, however, lets keep to a very intuitive idea about what
words are, and focus on morphemes. Once we better understand the nature of morphemes,
and the rules regulating how morphemes can be combined, and once we further see how these
units relate to higher level units of linguistic expression (phrases/sentences, which will be
our next topic in this class), we will be in a better position to define words properly.
As it turns out, a word (or, at least for the time being, something we intuitively think of as
being a word) can consist of one, two, three, four, five, six, or perhaps even more,
morphemes:
1 morpheme:
2 morphemes:
3 morphemes:
4 morphemes:
5 morphemes:
6 morphemes:

desire
desirable
desirability
undesirability
unprofessionally
antidisestablishmentarianism

#desire#
#desire+able#
#desire+able+ity#
#un+desire+able+ity#
#un+profess+ion+al+ly#
#anti+dis+establish+ment+ari+an+ism#

The combinations of morphemes witnessed above are all fine, but we should note that there are
constraints on the way in which of morphemes can be combined to derive more complex words.
So, undesirable is a word, but *desirableun, *abledesireun, *ableundesire, *desireunable, and
*unabledesire are not. Also, as we will see below, a word consisting of more than one morpheme
has a very rich, articulate structure, which in fact results from the very way in which it was
derived. The study of the rules by which morphemes can combine to form words is called
morphology.
Free and Bound Morphemes
Some morphemes, such as desire, happy, man, child, and so on, are able to stand alone as
words on their own. Such morphemes are called free morphemes. Other morphemes,
however, such as -ish, -ness, -ity, un-, re-, etc., cannot function as words on their own, but
must combine with some other morpheme, in which case they constitute part of a
derived/complex word. Morphemes of the latter type are called bound morphemes.
Bound morphemes are also known as affixes. English has two different kinds of affixes:
prefixes and suffixes. Prefixes attach to the beginning of a word to form a new word,
whereas suffixes attach to the end of a word.
happy
unhappy

=
=

unhappiness

a free morpheme, or free-standing word.

a complex word derived by attaching the prefix un- to the
base (root) word, happy
a complex word derived by attaching the suffix -ness to the
previously derived word, unhappy.

Most languages of the world have suffixes and prefixes. In the language Isthmus Zapotec, the
prefix ka- attaches to nouns to make them plural.
zigi
chin
zike shoulder
diaga ear

kazigi
kazike
kadiaga

chins
shoulders
ears

In Turkish, the suffix -ak is attached to a verb to derive a noun with the meaning a place for
verb-ing:
dur
bat

stop
sink

durak a stopping place

batak a sinking place

Some languages have a type of affix that attaches not to the beginning or end of a word, but
rather is inserted somewhere in the middle. Such affixes are called infixes. In Bontoc, a
language spoken in the Philippines, the infix -um- is inserted after the first consonant of a
noun or adjective to form a verb:
fikas strong
kilad red
fusul enemy

fumikas
kumilad
fumusul

to be strong
to be red
to be an enemy

We have just seen many examples of morphology in languages where an affix is attached to a
word to create a new, more complex word. There are in fact several types of morphology, all
of which involve different types of morphological processes/rules and have different
characteristics. In this class, we will talk about three major types of morphology: derivational,
inflectional, and compound morphology. We will begin with derivational morphology.
Derivational Morphology
Consider again the morphologically complex word undesirability, which is derived as
follows:
Starting with the verb desire as the base/root word:
[V desire]
The suffix -able is attached to form the adjective desirable:
[A [V desire] able]
The prefix un- is then attached to this derived word to form the adjective undesirable:
[A un [A [V desire] able]]
Finally, the suffix -ity is attached to this complex derived word to form yet another derived
word, in this case a noun, undesirability:
[N [A un [A [V desire] able]] ity]
As we have seen, derivational morphology involves computational processes that take a word
as the input and derive a new, more complex, word as an output by attaching an affix (a
derivational prefix or suffix). One of the main characteristics of derivational morphological
rules is that the meaning of the output word differs systematically from that of the input word.
Also, it is very often the case that the category of the output word is different from that of the
input word; although we will introduce more categories later on, for now we can assume the
word categories (also known as lexical categories) to be: nouns (N), verbs (V), adjectives (A),
and adverbs (Adv).
For example:
[V desire]

[A [N desire] able]
Able-Suffixation Rule

In this example, the input to the rule is the word (a simple morpheme) desire, which is a verb,
whereas the output of the rule is the (complex) word desirable, which is an adjective. Furthermore,
these two words have clearly different, but systematically related, meanings. These properties of
rule are completely general and systematic: -able always attaches to a verb to yield an adjective
with a particular systematic meaning, as in adorable, likeable, readable, eatable, and so on.

Another derivational rule attaches -en to an adjective forming a verb with a particular meaning
(something like to make x have property A), as in darken. Below are some other derivational
affixes in English that give rise to changes in lexical category:
Noun Adjective
boy boyish
picture picturesque

alcohol alcoholic
virtue virtuous
Verb Noun

acquit acquittal
clear clearance

sing singer
predict prediction
Adjective Adverb

exact exactly
Noun Verb
moral moralize
vaccine vaccinate

brand brandish

All derivational rules must specify the category of the input and the output. As already mentioned,
in many cases of derivational morphology, the category of output differs from that of the input.
However, there is not always a change in the category. Adding a prefix in particular very often
does not give rise to any change in category. Some examples are:
amoral
reprint

a+moral
re+print

ex-husband
sub-minimal

ex+husband
sub+minimal

In the above examples, the category of the root and the category of the derived stem are identical.
Some suffixes also fail to give rise to any change in category, as the examples below attest:
longer
shortest
musician

long+er
short+est
music+ian

Tokyoite
faddist

Tokyo+ite
fad+ist

The application of derivational rules is recursive; the output of one derivational can serve as the
input to another:
desire
desirable
undesirable
undesirability

[Root desire]
[Stem [Root desire] able]
[Stem0 un [Stem [Root desire] able]]
[Stem00 [Stem0 un [Stem [Root desire] able]] ity]

Note that this sort of step-by-step, recursive derivation, where the output of one rule application
serves as the input to the next rule application, gives rise to a hierarchical structure. This notion of
recursive derivation, and the hierarchical morphological structure that it yields, turns out to be
crucial. An alternative theory that simply recognizes a base/root word and the affixes that are
attached to it, but assigns no hierarchical structure, or no intermediate levels of derived structure,
simply cannot work. Let us now see why.
Consider the following examples again:
undrinkable
undesirable

(a word consisting of the 3 morphemes un, drink, able)

(a word consisting of the 3 morphemes un, desire, able)

These are well-formed, which tells us that we can have words in English consisting of a free root
morpheme (in these examples, drink/desire), the prefix un-, and the suffix -able. The root
morpheme in both of the above examples is a verb. Indeed, wherever we find -able in a
well-formed word, it seems that we find it next to a verb. Apparently, -able cannot be attached
directly to an adjective or a noun, as the following impossible words suggest:
*readyable

*happyable

*studentable

*adultable

We can explain these facts by making an explicit statement in the Able-Suffixation rule to the
effect that -able can only be attached to a verb.
Next, note that the following words are ungrammatical, despite the fact that they are proper
subparts of the good examples undrinkable and undesirable above:
*undrink
*undesire

un+drink
un+desire

The prefix un- can be attached to adjectives, in which case the resulting word too is always an
adjective:
unready
unhappy

un+ready
un+happy

But then how can we account for the impossibility of *undrink above? We cannot resort to a rule
stating that un- can never immediately precede a verb, since such a rule would also predict that
undrinkable is bad.
Rather, to explain why undrink/undesire are impossible whereas undrinkable/undesirable are
possible, we have to refer to intermediate levels of structure. In other words, a theory that assumes
flat structures like the following, will not work:
[drink+able]
[desire+able]

[un+drink+able]
[un+desire+able]

*[un+drink]
*[un+desire]

Rather we need to assume hierarchical structures, with recursive embedding, like those below in
order to guarantee proper application of the affixation rules:
[un [drink+able]]

[un [desire+able]]

As we have seen, such embedded hierarchy follows from the nature of the derivation and the
recursive application of the rules, as follows:
Derivation of undrinkable:
Step 1:

Able-Suffix Rule
[V drink]

Step 2:

[A [V drink] able]

Un-Prefix Rule
[A [V drink] able]

[A un [A [V drink] able]]

Thus, if we state a rule to the effect that the prefix un- can attach to an adjective but not to a
verb, we can explain the impossibility of undrink/undesire, since in these ungrammatical
examples, un- has been prefixed to a verb. On the other hand, in the grammatical
undrinkable/undesirable, un- is prefixed not to the verbs drink/desire but to the previously
derived adjectives drinkable/desirable, within which the verbs drink/desire are embedded.
The hierarchical nature of the structure of complex words, especially those involving more
than one affix, can be visually/graphically witnessed more easily by using a tree-diagram
structure below, as opposed to the bracketing notion we have been using above (though it is
important to realize that these are simply notational devices with no formal differences):
undrinkable A
2
un
drinkable A
2
drink V able
Lets now examine a couple of derivational morphological rules in detail.
Able-Suffix Rule:
[V " ]

[A [V " ] [Suffix able]]

Able-Suffix Rule

Meaning of [A [V " ] [Suffix able]] = able to be "-ed

Ex.)

[V do ]

[A [V do ] [Suffix able]]

Meaning = able to be done

Un-Prefix Rule:
[A " ]

[A [Prefix un] [A " ]]

Un-Prefix Rule

Meaning of [A [Prefix un] [A " ]] = not "

Ex.)

[A happy ]

[A [Prefix un] [A happy ]]

Meaning = not happy

The existence of verbs like undo, unzip, unwrap, etc., may at first sight seem like an
exception to Un-Prefix Rule, as stated above. However, more careful examination leads to the
conclusion that there are actually two different un- prefixes, involving different rules, in
which case these examples do not constitute exceptions. Lets call the rule we just formalized
above the Un1-Prefix Rule and now state another rule, call it the Un2-Prefix Rule, as below:
Un2-Prefix Rule:
[V " ]

[V [Prefix un2] [V " ]]

Un2-Prefix Rule

Meaning of [V [Prefix un2] [V " ]] = to reverse the action of "-ing

Ex)

[V do ]

[V [Prefix un2] [V do ]]

Meaning = to reverse the action of doing

It is thus predicted that there are two different words pronounced as undoable and that there
have different meanings, one word meaning that something is such that it cannot be done, and
the other word meaning that something is such that it is possible to reverse the action of its
doing. Note that, even though the base/root morpheme do is the same in both words, this is
because there are two possible morphological structures/derivations, one involving un1 and
the other involving un2.
Two derivations of two different words undoable:
undoable1
[V do ] [A [V do ] [Suffix able]]
[A [V do ] [Suffix able]] [A [Prefix un1][A [V do ] [Suffix able]]]
Meaning = not able to be done
undoable2
[V do ] [V [Prefix un2] [V do ]]
[V [Prefix un2] [V do ]] [A [V [Prefix un2] [V do ]] [Suffix able]]
Meaning = able to reverse the action of doing

One final point needs to be mentioned regarding the Un2-Prefix Rule. Recall the
ungrammatical words undrink and undesire discussed above. We explained the fact that these
were impossible as words by appealing to the fact that un- can only prefix to an adjective not
a verb. But, as we have just seen, there are really two un-s in English, one of which (our un2
above) indeed attaches to verbs, not adjectives. So, what is wrong with undrink having the
structure/derivation [V un2 [V drink]], then? Lets think about the meaning. What would it
mean to reverse the action of drinking? Well, this is not an impossible thought; in fact, I
think many of us have in fact had such an experience (I know that I have!). Still, it seems that
human languages define the actions denoted by certain verbs as being reversible whereas
others are not; the actions denoted by verbs like do, zip, button, fasten, wrap, pack, etc. are
reversible in this sense; they can be prefixed by un2. Apparently, verbs like drink, desire,
eat, walk, etc., however, are not.