Lexeme formation: the familiar 39
3. Some speakers will fi nd the forms in (3b) odd, and will question their acceptability, but they are all attested
in the Corpus of Contemporary American English and discussed in Bauer, Lieber, and Plag ( 2013 ).
goes on. This morpheme (an a for the verb ‘love’ and an i for the verb ‘say’)
doesn’t mean anything, but still must be added before the inflectional
ending can be attached. The root plus this extra morpheme is the stem .
Thought of another way, the stem is usually the base that is left when the
inflectional endings are removed. We will look further at roots and stems
in Chapter 6 , when we discuss inflection more fully.
3.3 Affixation
3.3.1 Word formation rules Let’s look more carefully at words derived by affixation . Prefixes and suf-
fixes usually have special requirements for the sorts of bases they can
attach to. Some of these requirements concern the phonology (sounds) of
their bases, and others concern the semantics (meaning) of their bases –
we will return to these shortly – but the most basic requirements are often
the syntactic part of speech or category of their bases. For example, the
suffix -ness attaches freely to adjectives, as the examples in (3a) show and
sometimes to nouns (as in (3b) ), but not to verbs (3c) :
(3) a. – ness on adjectives: redness, happiness, wholeness, commonness,
niceness
b. – ness on nouns: appleness, babeness, couch-potatoness 3
c. – ness on verbs: *runness, *wiggleness, *yawnness
The prefix un- attaches to adjectives (where it means ‘not’) and to verbs
(where it means ‘reverse action’), but not to nouns:
(4) a. un – on adjectives: unhappy, uncommon, unkind, unserious
b. un- on verbs: untie, untwist, undress, unsnap
c. un- on nouns: *unchair, *unidea, *ungiraffe
We might begin to build some of the rules that native speakers of English use
for making words with -ness or un- by stating their categorial requirements:
(5) Rule for -ness (first version): Attach – ness to an adjective or to a noun.
Rule for un- (first version): Attach un- to an adjective or to a verb.
Of course, if we want to be as precise as possible about what native speak-
ers know about forming words with these affixes, we should also indicate
what category of word results from using these affixes, and what the
resulting word means. So a more complete version of our – ness and un-
rules might look like (6) :
(6) Rule for – ness (second version): – ness attaches to adjectives or nouns
‘X’ and produces nouns meaning ‘the quality of X’.
Rule for un- (second version): un- attaches to adjectives meaning ‘X’
and produces adjectives meaning ‘not X’; un – attaches to verbs
meaning ‘X’ and produces verbs meaning ‘reverse the action X’.
40 INTRODUCING MORPHOLOGY
If we’re really trying to model what native speakers of English know about
these affixes, we might try to be even more precise. For example, un- does
not attach to all adjectives or verbs, as you can discover by looking at the
next Challenge box.
Challenge
Look at the following words and try to work out more details of the
rule for un- in English. The (a) list contains some adjectives to which
negative un- can be attached and others which seem impossible or at
least somewhat odd. The (b) list contains some verbs to which
reversative un- can attach and others which seem impossible. See if
you can discern some patterns:
(a) unhappy, *unsad, unlovely, *unugly, unintelligent, *unstupid
(b) untie, unwind, unhinge, unknot, *undance, *unyawn, *unexplode,
*unpush
What the (a) examples in the Challenge box seem to show is that the negative
prefix un- in English prefers to attach to bases that do not themselves have
negative connotations. This is not true all of the time – adjectives like unself-
ish or unhostile are attested in English – but it’s at least a significant tendency.
As for the (b) examples, they suggest that the un- that attaches to verbs pre-
fers verbal bases that imply some sort of result, and moreover that the result
is not permanent. Verbs like dance, push, and yawn denote actions that have
no results, and although explode implies a result (i.e., something is blown up),
it’s a result that is permanent. In contrast, a verb like tie implies a result
(something is in a bow or knot) which is temporary (you can take it apart).
We have just constructed what morphologists call a word formation
rule , a rule which makes explicit all the categorial, semantic, and phono-
logical information that native speakers know about the kind of base that
an affix attaches to and about the kind of word it creates. We might now
state the full word formation rules for negative un- as in (7) :
(7) Rule for negative un- (final version): un- attaches to adjectives,
preferably those with neutral or positive connotations, and creates
negative adjectives. It has no phonological restrictions.
Now let’s look at two more affixes. In English we can form new verbs by
using the suffixes -ize or -ify . Both of these suffixes attach to either nouns
or adjectives, resulting in verbs:
(8) -ize on adjectives: civilize, idealize, finalize, romanticize,
tranquillize
-ize on nouns: unionize, crystallize, hospitalize, caramelize,
animalize
-ify on adjectives: purify, glorify, uglify, moistify, diversify
– ify on nouns: mummify, speechify, classify, brutify, scarify,
bourgeoisify
Lexeme formation: the familiar 41
We might state the word formation rules for -ize and -ify as in (9) :
(9) Rule for -ize (first version): -ize attaches to adjectives or nouns that
mean ‘X’ and produces verbs that mean ‘make/put into X’.
Rule for -ify (first version): -ify attaches to adjectives or nouns that
mean ‘X’ and produces verbs that mean ‘make/put into X’.
But again, we can be a bit more precise about these rules. Although -ize
and -ify have almost identical requirements for the category of base they
attach to and produce words with roughly the same meaning, they have
somewhat different requirements on the phonological form of the stem
they attach to. As the examples in (8) show, -ize prefers words with two or
more syllables where the final syllable doesn’t bear primary stress (e.g.,
trá nquil, hó spital ). The suffix -ify , on the other hand, prefers monosyllabic
bases ( pure, brute, scar ), although it also attaches to bases that end in a – y
( mummy, ugly ) or bases whose final syllables are stressed ( divé rse, bourgé ois ).
Since we want to be as precise as possible about our word formation rules
for these suffixes, we will state their phonological restrictions along with
their categorial needs:
(10) Rule for -ize (final version): -ize attaches to adjectives or nouns of
two or more syllables where the final syllable does not bear
primary stress. For a base ‘X’ it produces verbs that mean ‘make/put
into X’.
I leave it to you to come up with the final version of the word formation
rule for – ify.
3.3.2 Word structure When you divide up a complex word into its morphemes, as in (11) , it’s
easy to get the impression that words are put together like the beads that
make up a necklace – one after the other in a line:
(11) unhappiness = un + happy + ness
But morphologists believe that words are more like onions than like neck-
laces: onions are made up of layers from innermost to outermost. Consider
a word like unhappiness . We can break this down into its component mor-
phemes un + happy + ness , but given what we learned above about the
properties of the prefix un- and the suffix – ness we know something more
about the way in which this word is constructed beyond just its constitu-
ent parts. We know that un- must first go on the base happy . Happy is an
adjective, and un- attaches to adjectives but does not change their catego-
ry. The suffix -ness attaches only to adjectives and makes them into nouns.
So if un- attaches first to happy and -ness attaches next, the requirements
of both affixes are met. But if we were to do it the other way around, -ness
would have first created a noun, and then un- would be unable to attach.
We could represent the order of attachment as if words really were
onions, with the base in the innermost layer, and each affix in its own
succeeding layer: see Figure 3.2 .
42 INTRODUCING MORPHOLOGY
But linguists, not generally being particularly artistic, prefer to show
these relationships as ‘trees’ that look like this :
(12)
N
A
A
un happy ness
Similarly, we might represent the structure of a word like repurify as in (13) :
(13)
V
V
A
re pure ify
In order to draw this structure, we must first know that the prefix re-
attaches to verbs (e.g., reheat, rewash , or redo ) but not to adjectives ( *repure,
*rehappy ) or to nouns ( *rechair, *retruth ). Once we know this, we can say that
the adjective pure must first be made into a verb by suffixing -ify , and only
then can re- attach to it.
FIGURE 3.2 Words are like onions
Challenge
In English, the suffix -ize attaches to nouns or adjectives to form
verbs. The suffix -ation attaches to verbs to form nouns. And the suffix
-al attaches to nouns to form adjectives. Interestingly, these suffixes
Lexeme formation: the familiar 43
3.3.3 What do affixes mean? When we made the distinction between affixes and bound bases above, we
did so on the basis of a rather vague notion of semantic robustness; bound
bases in some sense had more meat to them than affixes did. Let us now
attempt to make that idea a bit more precise by looking at typical mean-
ings of affixes.
In some cases, affixes seem to have not much meaning at all. Consider
the suffixes in (14) :
(14) a. -(a)tion examination, taxation, realization, construction
– ment agreement, placement, advancement, postponement
-al refusal, arousal, disposal
b. -ity purity, density, diversity, complexity
-ness happiness, thickness, rudeness, sadness
Beyond turning verbs into nouns with meanings like ‘process of X-ing’ or
‘result of X-ing’, where X is the meaning of the verb, it’s not clear that the
suffixes -(a)tion, -ment , and -al add much of any meaning at all. Similarly
with -ity and -ness , these don’t carry much semantic weight of their own,
aside from what comes with turning adjectives into nouns that mean
something like ‘the abstract quality of X’, where X is the base adjective.
Affixes like these are sometimes called transpositional affixes , meaning
that their primary function is to change the category of their base with-
out adding any extra meaning.
Contrast these, however, with affixes like those in (15) :
(15) a. -ee employee, recruitee, deportee, inductee
b. -less shoeless, treeless, rainless, supperless
c. re- reheat, reread, rewash
These affixes seem to have more semantic meat on their bones, so to
speak: -ee on a verb indicates a person who undergoes an action; -less
means something like ‘without’; and re- means something like ‘again’.
Languages frequently have affixes (or other morphological processes, as
we’ll see in Chapter 5 ) that fall into common semantic categories. Among
those categories are:
• personal or participant affixes: These are affixes that create ‘people nouns’ either from verbs or from nouns. Among the personal affixes
in English are the suffix -er which forms agent nouns (the ‘doer’ of
can be attached in a recursive fashion: convene → convention →
conventional → conventionalize → conventionalization .
First draw a word tree for conventionalization . Then see if you can
find other bases on which you can attach these suffixes recursively.
What is the most complex word you can create from a single base
that still makes sense to you? Are there any limits to the complexity
of words derived in this way?
44 INTRODUCING MORPHOLOGY
the action) like writer or runner and the suffix -ee which forms patient
nouns (the person the action is done to). Also among this class of
affixes are ones that create other ‘people nouns’ other than the agent
or the patient in an event, for example, inhabitants of a place (like
Manhattanite or Bostonian ).
• locative affixes: These are affixes that designate a place. For example, in English we can use the suffix – ery or – age to denote a place where
something is done or gathered (like eatery or orphanage ).
• abstract affixes: These are affixes that create abstract nouns that denote qualities (like happiness or purity ) or statuses ( puppydom, advisor-
ship, daddyhood ) or even aspects of behavior ( buffoonery ).
• negative and privative affixes: Negative affixes add the meaning ‘not’ to their base; examples in English are the prefixes un- , in- , and non-
( unhappy, inattentive, non-functional ). Privative affixes mean something
like ‘without X’; in English, the suffix -less ( shoeless, hopeless ) is a priva-
tive suffix, and the prefix de- has a privative flavor as well (e.g., words
like debug or debone mean something like ‘cause to be without bugs/
bones’).
• prepositional and relational affixes: Prepositional and relational affixes often convey notions of space and/or time. Examples in
English might be prefixes like over- and out- or pre – and post – ( overfill,
overcoat, outrun, outhouse, preschool, preheat, postwar, postdate ).
• quantitative affixes: These are affixes that have something to do with amount. In English we have affixes like -ful ( handful, helpful )
and multi – ( multifaceted ). Another example might be the prefix re-
that means ‘repeated’ action ( reread ), which we can consider quan-
titative if we conceive of a repeated action as being done more
than once. Other quantitative affixes that we have in English
denote collectives or aggregates of individuals (e.g., acreage or
knickknackery ).
• evaluative affixes: Evaluative affixes consist of diminutives , affixes that signal a smaller version of the base (e.g., in English -let as in
booklet or droplet ) and augmentatives , affixes that signal a bigger ver-
sion of the base. The closest we come to augmentative affixes in
English are prefixes like mega- ( megastore , megabite ). The Native
American language Tuscarora (Iroquoian family) has an augmenta-
tive suffix – ʔ o ʔ y that can be added to nouns to mean ‘a big X’; for
example, takó:θ- ʔ o ʔ y means ‘a big cat’ (Williams 1976 : 233). Diminutives and augmentatives frequently bear other nuances of
meaning. For example, diminutives often convey affection, or
endearment, as we find in some words with – y or – ie in English (e.g.,
sweetie, kitty ). Augmentatives sometimes have pejorative overtones
(the Niger-Congo language Fula has a pejorative diminutive, for
example).
Note that some semantically contentful affixes change syntactic category
as well; for example, the suffixes -er and -ee change verbs to nouns, and the
prefix de- changes nouns to verbs. But semantically contentful affixes
Lexeme formation: the familiar 45
need not change syntactic category. The suffixes – hood and -dom , for
example, do not ( childhood, kingdom ), and by and large prefixes in English
do not change syntactic category.
So far we have been looking at suffixes and prefixes whose meanings
seem to be relatively clear. Things are not always so simple, though. Let’s
look more closely at the suffix -er in English, which we said above formed
agent nouns. Consider the following words:
(16) a. writer
skater
b. printer
freighter
c. loaner
fryer (i.e., a kind of chicken)
d. diner
All of these words seem to be formed with the same suffix. Look at each
group of words and try to characterize what their meanings are. Does -er
seem to have a consistent meaning?
It’s rather hard to see what all of these have in common. The words in
(16a) are indeed all agent nouns, but the (b) words are instruments; in
other words, things that do an action. In American English the (c) words
are things as well, but things that undergo the action rather than
doing the action (like the patient -ee words discussed above): a loaner is
something which is loaned (often a car, in the US), and a fryer is some-
thing (a chicken) which is fried. And the word diner in (d) denotes a loca-
tion (a diner in the US is a specific sort of restaurant). Some morphologists
would argue that there are four separate suffixes in English, all with the
form -er . But others think that there’s enough similarity among the
meanings of -er words in all these cases to merit calling -er a single affix,
but one with a cluster of related meanings. All of the forms derived with
-er denote concrete nouns, either persons or things, related to their base
verbs by participating in the action denoted by the verb, although some-
times in different ways. This cluster of related meanings is called affixal
polysemy .
Affixal polysemy is not unusual in the languages of the world. For
example, it is not unusual for agents and instruments to be designated by
the same suffix. This occurs in Dutch , as the examples in (17a) show (Booij
and Lieber 2004 ), but also in Yoruba (Niger-Congo family), as the examples
in (17b) show (Pulleyblank 1987 : 978):
(17) a. Dutch
spel-er ‘player’ ( spelen ‘play’)
Maai-er ‘mower’ ( maaien ‘mow’)
b. Yoruba
a-pànìà ‘murderer’ ( pa ‘kill’, ènìà ‘people’)
a-bẹ ‘razor, penknife’ ( bẹ ‘cut’)
46 INTRODUCING MORPHOLOGY
The Dutch suffix -er is in fact quite similar to the -er suffix in English in
the range of meanings it can express. The Yoruba prefix a- also forms both
agents and instruments.
3.3.4 To divide or not to divide? A foray into extenders , formatives , crans , and other messy bits In Chapter 1 we defined a morpheme as the smallest unit of language that
has its own meaning. We have now looked at affixes and bases, both free
and bound, and considered their meanings and how they combine into
complex structures. For the most part, the examples we have looked at are
simple and their analysis has been relatively clear. But often the closer we
look at the morphology of a language, the more complex it becomes.
There are many ways we can illustrate this with the morphology of
English , but we will choose just a few points in this section that compli-
cate our initial picture. I should point out in advance that my examples
here will come from English, but we should expect to find similar twists
and turns in any language whose morphology we examine in detail.
Consider the words in (18) :
(18) a. report, import, transport, deport, comport, export
b. cranberry, huckleberry, raspberry
c. Platonic, tobacconist, spasmodic, egotist
d. sniffle, snort, snot, snout
e. eggitarian, pizzatarian, pastatarian, fruitarian, flexitarian
Let’s start with the ones in (18a) . We have assumed so far that words can
be broken down into morphemes, which are pieces that have meaning.
The words in (18a) certainly look like they might be broken down because
they have recurrent parts, but if they are, what do the pieces mean? In
fact, English has dozens of words that are similar to what we might call
the – port family. See how many cells of Table 3.1 you can fill in.
Table 3.1
in- ex- con- re- trans- de-
-port
-mit
-ceive
-duce
-cede
-fer
-scribe
-gress
-sist
Lexeme formation: the familiar 47
One reason for our dilemma in analyzing these forms is that they are
not native to English. They were borrowed from Latin (or from French ,
which in turn is descended from Latin), where they did have clear mean-
ings: -port comes from the verb portare ‘to carry’, – mit from the verb mittere
‘to send’, -scribe from the verb scribere ‘to write’, and so on. But English
speakers (unless they’ve studied Latin!) don’t know this. Morphologists are
left with an unsatisfying sense that the words above somehow ought to be
treated as complex, but are nevertheless reluctant to give up the strict
definition of morpheme. One way of dealing with these pieces is to
acknowledge that they seem to be independent and recombinable in some
way, but that they are not morphemes in the normal sense. Bauer, Lieber,
and Plag ( 2013 : 16) call elements like these formatives , which they define
as “elements contributing to the construction of words whose semantic
unity or function is obscure or dubious.”
The items in (18b) illustrate a different type of formative that are some-
times called cran morphs , from the first bit of the word cranberry . The
second part of the word cranberry is clearly a free morpheme. But when we
break it off, what’s left is a piece that doesn’t seem to occur in other words
(except in recent years, words like cranapple that are part of product
names), and doesn’t seem to mean anything independently. There are
quite a few of these cran morphs in the names of other types of berries:
rasp- in raspberry , huckle- in huckleberry . In cases such as these we are even
more tempted than we were with – port, -ceive , and the like to divide words
into morphemes, even though we know that one part of the word isn’t
meaningful in the way morphemes usually are.
The examples in (18c) also display a puzzling characteristic. If we try to
break these words down into their component morphemes, what we find
is that each one consists of two obvious morphemes plus an extra sound
or two:
(19) Plato + n + ic (compare icon+ic )
tobacco + n + ist (compare accordian+ist )
spasm + od + ic (compare Celt+ic )
ego + t+ ist (compare clarinet+ist )
The question is what the extra bit is. Is it part of the base of the word or part
of the affix or part of neither? It seems pretty clear that it doesn’t mean
anything. And why do we get an / n / in Platonic , but / od / in spasmodic , and noth-
ing between the base and the suffix in heroic , or an /n/ in tobacconist , but a /t/
in egotist ? Morphologists don’t have a clear answer to these questions – part
Challenge
Do you think that units like -port, -mit, -ceive , and the like should be
considered morphemes? If so, what problems do they present for our
definition of morpheme ? If not, what should we do about the
intuition that native speakers of English have that such words are
complex?
48 INTRODUCING MORPHOLOGY
of the fun of doing morphology is that we can argue about the possibilities,
after all! – but we can at least give these puzzling bits a name. Following
Bauer, Lieber, and Plag ( 2013 ), we will call them extenders .
Next, let’s look at the examples in (18d) . These exhibit what is called
sound symbolism . All of the words begin with the consonant cluster /sn/
and seem to have something to do with the nose, but the sequence of
sounds /sn/ doesn’t mean anything by itself. Here morphologists are rela-
tively agreed that sound symbolic words cannot be broken into parts. For
one thing, the sequence /sn/ doesn’t refer to ‘nose’ everywhere it occurs
(consider words like snail, snap , or snit ), and for another, if we were to seg-
ment /sn/ in the words in (18) , what would be left over would neither have
meaning by itself nor recur elsewhere in English.
Our final group of odd bits is illustrated in (18e) . In these examples, the
first part of each word is clearly a free morpheme, but the second part is
not. Rather, it is what Bauer, Lieber, and Plag ( 2013 ) call a splinter ,
something which is split off from an original word, but which is not
really (yet!) a true suffix. In the examples in (18e) , the splinter is tarian ,
which is a bit broken off from the word vegetarian , and then used to create
new words meaning ‘one who eats X’. English has lots of splinters, among
them tastic , as in funktastic or fishtastic , which is used to form mostly ironic
words meaning ‘excellent or great in reference to X’, originally from fan-
tastic , or licious , as in bagelicious or bootielicious , which is used to form words
meaning ‘appealing in reference to X’, originally from the word delicious .
The difference between a splinter and a true suffix is that speakers under-
stand splinters in relation to the original word from which the ending
splits off. If these bits survive and continue to give rise to new forms,
though, they might someday be real suffixes!
3.4 Compounding
So far we have concentrated on complex words formed by derivation, spe-
cifically by affixation. Derivation is not the only way of forming new
words, of course. Many languages also form words by a process called
compounding. Compounds are words that are composed of two (or more)
bases, roots, or stems. In English we generally use free bases to compose
compounds, as the examples in (20) show:
(20) English compounds
compounds of two nouns: windmill, dog bed, book store
compounds of two adjectives: icy cold, blue-green, red hot
compounds of an adjective and a noun: greenhouse, blackboard,
hard hat
compounds of a noun and an adjective: sky blue, cherry red, rock hard
3.4.1 When do we have a compound? How do we know that a sequence of words is a compound? Surprisingly,
it’s not that easy to come up with a single criterion that works in all cases.
- 3 Lexeme formation: the familiar
- 3.3 Affixation
- 3.3.1 Word formation rules
- 3.3.2 Word structure
- 3.3.3 What do affixes mean?
- 3.3.4 To divide or not to divide? A foray into extenders, formatives, crans, and other messy bits
- 3.4 Compounding
- 3.4.1 When do we have a compound?
