4. Chapter 1
BASIC MORPHOLOGY
1.1 The lexeme concept
(1) cat, dog Two words
(2) cat, cats Two words? One word?
CAT LEXEME
(3)
kat kats word forms
Sg. Pl.
(4) a. Tom will walk to work
b. Tom walks to work
c. Tom is walking to work
d. Tom walked to work
(5) walk: {walk, walks, walking, walked}
1.2 Morphological processes - inflection and derivation
1.2.1 Derivation (derivational morphology) the creation of new lexemes
Often (but not always) changes grammatical category:
(6) V ⇒ N: (to) print ⇒ printer
V ⇒ A: print ⇒ printable
V ⇒ V: print ⇒ re-print
N ⇒ V: flea ⇒ de-flea
N ⇒ A: milk ⇒ milky
N ⇒ N: mother ⇒ motherhood
A ⇒ N: happy ⇒ happiness
A ⇒ V: thick ⇒ thicken
A ⇒ A: happy ⇒ unhappy
4
5. CHAPTER 1. BASIC MORPHOLOGY 5
Derivation typically adds a new lexical meaning component:
(7) printable: ‘such that can be printed’
motherhood: ‘property of being a mother’
thicken: ‘become or cause to become thicker’
Derivation is iterative (feeds into itself, sometimes referred to as recursion):
(8) in-de-cipher-abil-ity
1.2.2 Inflection (inflectional morphology)
Creates word forms of a lexeme
(9) CAT:
cat (Singular) cats (Plural)
(10) a. SING:
sing Base form
sings 3sg Present Tense
singing Present Participle
sang Past Tense
sung Past Participle (Perfect/Passive Participle)
b. WALK:
walk Base form
walks 3sg Present Tense
walking Present Participle
walked Past Tense
walked Past Participle (Perfect/Passive Participle)
(11) a. COLD:
cold Positive
colder Comparative
coldest Superlative
b. GOOD:
good Positive
better Comparative
best Superlative
1.2.3 Morphological operations
Morphological operation =def ‘concrete change made to a word form in order to signal a deriva-
tional or inflectional process’
(12) reprints: ‘3sg Present Tense RE[PRINT]]’ = ‘print again’
prefix root suffix
re print s
Affix = prefix or suffix
6. CHAPTER 1. BASIC MORPHOLOGY 6
Other languages permit more radical changes, e.g. infixation, circumfixation, replacement,
subtraction, reduplication, . . . Example of replacives morphology in Eng.:
Marx-ist ‘one who follows the tenets of Marxism’
Marx-ism
philosoph-y ∼ philosoph-er, nomin-ate ∼ nomin-ee.
Other operations in English:
Vowel change:
man ∼ men
sing ∼ sang ∼ sung
Sometimes this accompanies affixations:
break ∼ broke ∼ broken (= broke + en)
write ( ∼ wrote) ∼ written
Consonant change:
house [haUs] ∼ (to) house [haUz]
knife [naIf] ∼ knives [naIvz]
Stress shift:
contrast ∼ (to) contr´ st
´ a N∼V
(Languages with tones may use tone alternations to realize grammatical processes)
Conversion: word of one class treated as belonging to a different class without any overt mor-
phological operation:
N ⇒ V: paper ∼ to paper (the wall)
skin ∼ to skin a rabbit
head ∼ to head a department, an inquiry, a phrase
police ∼ to police a town, a regulation
V ⇒ N: walk ∼ go for a walk
fall ∼ take a fall
sleep ∼ get a good night’s sleep
Also phrasal verbs:
take off ∼ a smooth take off
put down ∼ a cruel put down
run through ∼ a quick run through (one’s lecture)
A ⇒ N: the good, the bad and the ugly
N ⇒ A: orange (balloon), primrose (wallpaper)
A ⇒ V: wet (the paper), dry (the dishes)
7. CHAPTER 1. BASIC MORPHOLOGY 7
1.3 Types of inflectional processes
1.3.1 Inherent inflection
Expressing functional categories which can be interpreted semantically (interpretable, mean-
ingful functional categories)
There are a variety of grammatical functions (or grammatical ‘meanings’) that can be ex-
pressed by inflectional means. These include:
• verb tense: e.g. past, present, future
• verb mood: e.g. indicative, subjunctive, imperative, ...
• noun: number, definiteness, case
• comparison of adjectives: positive, comparative, superlative
This type of inflection is close to derivation (and can be difficult to distinguish from deriva-
tion).
1.3.2 Contextual inflection
Expresses relations between words and phrases in a sentence. Doesn’t express a meaning
(uninterpretable functional categories). Two main kinds: agreement, government. (English
has little of either think of some examples).
Inflectional paradigms
The tables we have seen so far are all examples of paradigms, sets of the inflected word forms of
a language. Paradigms are defined by the grammatical distinctions which a language chooses
to code morphologically. E.g. nouns in English have to be inflected for Number, but not Case
or Possessor. Verbs in Russian are inflected for Tense, Mood, Voice and agree with their Subject,
but don’t agree with their Object.
There are often systematic relations between parts of a paradigm. Consider the paradigms for
a regular (‘weak’) and irregular (‘strong’) verb:
Form label weak verb strong verb
Base form climb swim
Present Participle climb-ing swimm-ing
3Sg climb-s swim-s
Past climb-ed swam
Perfect/Passive Participle climb-ed swum
Table 1.1: English verb paradigms
Notice that the Past form and the Perfect/Passive Participle are identical for climb. This is
true of all regular verbs. Such a coincidence of form is called syncretism. Most inflectional
paradigms exhibit some form of syncretism.
8. CHAPTER 1. BASIC MORPHOLOGY 8
1.4 Compounds
N
N N
(13)
morphology lecture
‘lecture which has something to do with morphology’
lecture = head of compound, morphology = modifier
Compound with head = endocentric
Meaning of compound determined regularly from meanings of elements
Compounding is recursive:
(14) morphology class room change announcement (procedures (review (committee (chair-
man (. . . )
(15) blackbird
N
A N
black bird
Non-compositional: meaning of whole compound can’t be inferred from component nouns.
[Why?]
Unheaded compounds (exocentric):
(16) lazybones
pickpocket
forget-me-not
Various birds’ names are like this: redcap, yellow hammer, . . .
1.5 Clitics
Words which can’t exist independently and need a host to ‘lean on’ (attach to phonologically):
(17) a. it’s it is, it has
b. could’ve could have
c. she’ll she will/shall
These contracted forms can attach to any category of word in principle, hence, they aren’t
proper affixes.
9. CHAPTER 1. BASIC MORPHOLOGY 9
(18) a. The man responsible’s been fired
b. The man responsible has been fired
(19) a. The ones over here’ll be fresher
b. The ones over here will be fresher
Phrasal affix - Possessive ’s (‘Saxon genitive’)
Here the clitic can’t be thought of as a reduced form of a full word. The clitic attaches to words
of any category, provided they’re on the right edge of the NP:
(20) a. Harriet’s hat
b. the man who Harriet met’s hat
c. the girl I’m speaking to’s hat
The clitic behaves just like an affix except that it attaches to the edge of the phrase rather
than to the syntactic head of the phrase (as a true Genitive Case ending would).
10. Chapter 2
English Inflection
2.1 Introduction - Functional categories and inflectional categories
As we saw in Chapter One inflections are a subset of the functional categories, which govern
syntactic relations in sentences. Functional categories are expressed as syntactic features, e.g.
the definiteness property of English noun phrases:
the cat [DEFINITE +]
a cat [DEFINITE -]
Here is a list of the main functional categories needed for English:
Grammatical relations
SUBJECT (or Nominative Case)
OBJECT (or Accusative/Objective Case)
ADJUNCT (or Adverbial)
POSSESSOR
MODIFIER (e.g. attributive adjective)
Nominal features
DEFINITENESS
NUMBER
PRONOUN
Pronominal features
PERSON
NUMBER
GENDER
Adjectival features
COMPARATIVE
SUPERLATIVE
10
11. CHAPTER 2. ENGLISH INFLECTION 11
Verbal features
TENSE {PAST, PRESENT, FUTURE}
ASPECT {SIMPLE, PERFECT, PROGRESSIVE}
VOICE {ACTIVE, PASSIVE}
MOOD {INDICATIVE, IMPERATIVE, INTERROGATIVE}
MODALITY {various modal auxiliary verbs}
POLARITY {Positive, Negative}
Agreement
SUBJAGR{3sg}
Some of these categories correspond to inflections:
[Number:{Sg., Pl}]
[Tense: {Past, NonPast}]
[SubjAgr:[Person:3, Number:Sg]]
Other functional categories are expressed in four main ways:
1. word order (e.g. for grammatical relations: ‘Tom saw Harriet’ ⇒ SUBJ [Tom], OBJ [Har-
riet])
2. function words (e.g. [DEF +] ⇒ the, OBJECT PRONOUN NUMBER SG, PERSON 3,
GENDER FEM ⇒ her)
3. combination of function word and specially inflected word form, e.g. passive ‘Harriet
was seen by Tom’: VOICE PASSIVE ⇒ appropriate form of auxiliary verb be + [Verb-
Form:Participle:Past] of lexical verb
4. inflections, e.g. NUMBER PLURAL ⇒ [Number:Plural] cats
Where functional categories are expressed by (2, 3) we say that we have an analytic (or pe-
riphrastic) construction; where they are expressed by (4), i.e. solely by inflections, we call it a
synthetic construction.
2.2 Noun inflection
2.2.1 Plurals
<Work out the different ways of forming an irregular plural>
Count vs. mass (coercion)
We customarily distinguish two types of noun on the basis of semantics: count and mass.
Count nouns denote individuated objects, while mass nouns denote stuff, substances or ag-
gregates that can’t be individuated. This distinction cuts across the other distinctions such as
proper/common or abstract/concrete. Examples:
12. CHAPTER 2. ENGLISH INFLECTION 12
Count: chair, idea, difficulty, chocolate
Mass: milk, justice, difficulty, chocolate
Only count nouns can take a plural form. However, many nouns seem to be in both categories:
(1) a. Your proposal faces several difficulties
b. Harriet ate several chocolates
a. The level of difficulty of the exercise was too high
b. Harriet’s fingers were covered in chocolate
This is common behaviour for other nouns, where a count or mass interpretation is forced on
a mass/count noun a process often known as coercion:
(2) You’ve got egg on your tie [count ⇒ mass]
(3) She ordered a milk and two coffees [mass ⇒ count]
Possessor
We can express possession analytically (by means of the preposition of) or synthetically by
means of the phrasal affix ‘s.
A peculiarity of the phrasal affix is that it can’t cooccur with a plural suffix:
(4) a. my friend’s book
b. my friends’ book [*my friend-s-’s book]
c. one of my friends
d. one of my friends’ book [= the book of one of my friends]
In other words a sequence of s-s is simplified to a single s. This phenomenon, where two
adjacent homophonous affixes or clitics, usually with different ‘meaning’, are simplified to
just one token, is quite common cross-linguistically. The technical term for this is haplology.
Nominal function words
Pronouns: Personal pronouns have special forms for SUBJ/OBJ, often referred to as Nomina-
tive/Accusative case forms. The designation is a little misleading, since the subject form is
only used for certain subjects, namely, when the pronoun is the sole exponent of the SUBJ fea-
ture/function. Everywhere else we get the obj form (hence, the obj form is the default subject
form). Compare:
(5) a. Tom went for a walk
b. Tom and I went for a walk [literary English only]
c. Tom and me went for a walk [normal colloquial English]
d. *I and Tom went for a walk
e. Me and Tom went for a walk
NB. Prescriptive grammarians usually try to ‘ban’ examples such as (5c, 5e) in favour of the
artificial construction type (5b). However, this is simply due to ignorance of the facts of English
and of the principles of linguistics.
13. CHAPTER 2. ENGLISH INFLECTION 13
(6) Who’s going for a walk?
Me.
*I
I am
*Me am (is)
(7) It’s me/*I
The obj form is used with prepositions:
(8) between you and me
The expression between you and I is becoming current. It started out as a straightforward gram-
matical error (an example of hypercorrection), originally from speakers with limited command
of written English who were called upon to speak in public. Now it’s entering the language as
a high register variant of between you and me. However, I haven’t heard anyone saying ‘between
I and NP’ (e.g. There’s nothing between I and my secretary).
Possessive pronouns
Note that there’s an adjectival and a pronominal use:
(9) a. This is my book
b. This (book) is (one of) mine
Singular Plural
poss adjective poss pronoun poss adjective poss pronoun
my mine our ours
thy thine your yours
his his
her hers their theirs
its its
one’s one’s
(This parallels demonstratives v. inf.)
Reflexives
These are formed by suffixing self/selves to a possessive adjective (1st/2nd person) or to the
object pronoun (3rd person):
my-self, thy-self, our-selves, your-selves
himself, herself, itself, oneself, themselves
Demonstratives
The only modifiers which have special plural agreement form: this/these, that/those.
Can be used as either a modifier or as the head of the nominal phrase (like possessives):
14. CHAPTER 2. ENGLISH INFLECTION 14
(10) I want that (book)
(11) I bought these (flowers)
2.3 Verb inflection
2.3.1 Analytic vs. synthetic forms: the tense/aspect system
Tense
There is a long standing controversy over the number of tenses in English. On the one hand
we may wish to distinguish Present, Past, Future:
Harriet runs, ran, will run
On the other hand, verbs only have two tense inflections. The solution is to recognize the
difference between morphological (inflectional) features and syntactic functional categories.
The syntactic functional feature TENSE has three values PRESENT, PAST, FUTURE, but the
morphological, inflectional feature just has two [Tense:Past, NonPast]. Thus, one of the TENSE
features has to be expressed analytically (by means of the modal auxiliary will + base form.)
Aspect
Both aspects are expressed analytically:
HAVE + -en Perfect
BE + -ing Progressive
In addition it makes sense to distinguish a Past Habitual tense/aspect:
Tom used to play the flute
This can just about cooccur with the Progressive:
Tom used to be making a nuisance of himself all the time
The Habitual doesn’t readily cooccur with modals:
*Tom may/could/would/can. . . used to
This should be distinguished from BE USED to V-ing:
Tom is used to getting up late
This might be thought of as a kind of Customary aspect. This occurs in other tense/aspect
forms:
was
Tom
had been used to getting up late
*is being
Whether we treat Customary and Past Habitual constructions as grammatical aspects depends
on the extent to which we think they’ve been grammaticalized.
15. CHAPTER 2. ENGLISH INFLECTION 15
Agreement
Only in 3sg for NonPast tense forms (but even then not for modal auxiliaries). BE: supernu-
merary agreement for Number in Past and Present:
Singular Plural
am
(art) are
is
was were
Participles
NB: Two systematic syncretisms:
Past Tense/Past participle syncretism [regular verbs only]
Passive/Perfect participle syncretism [all verbs]
These two syncretisms are rather different in kind. The first is found with all regular verbs and
some irregulars, but not all:
walk walked walked
bring brought brought
keep kept kept
send sent sent
cut cut cut
BUT
write wrote written
take took taken
ring rang rung
run ran run
The perfect/past participle syncretism is completely exceptionless and is therefore part of a
generalization that goes deeper than just morphological form.
(12) a. Tom has bought a book
b. The book was bought by Tom
a. Dick has rung the bell
b. The bell was rung by Dick
a. Harriet had taken the message
b. The message had been taken by Harriet
Even:
(13) a. Everyone had had a good time
b. A good time was had by all
16. CHAPTER 2. ENGLISH INFLECTION 16
Past participle as adjective
Participles are much more common as post-modifiers than as pre-modifiers:
??the rung bell the bell rung by the church warden
If the participle itself is modified, to form a kind of compound adjective, we get better results:
the recently-rung bell, freshly-mown hay, a seldom bought book, an oft-cited remark
However, a participle can’t have any genuine syntactic complements or adjuncts:
the book given to Harriet (by Tom)
*the given to Harriet (by Tom) book
the messages taken yesterday
*the yesterday taken messages
*the taken yesterday messages
-ing form
This is usually known as the Present Participle, though this is misleading in several respects.
It has three main uses:
• Formation of Progressive aspect (with aux. BE)
• Formation of Gerund/Verbal Noun
• Formation of Participle
1. Progressive aspect see above
The progressive aspect isn’t found with verbs which denote States (as opposed to ‘dynamic’
events which evolve through time such as Activities, Processes and so on)
*Tom is knowing the answer to these questions
*Tom is being tall nowadays
You can sometimes ‘coerce’ a special reading:
Tom is being stupid again
Harriet is being half a cow in this year’s pantomime
2. Gerund/Verbal Noun (VN) The Gerund/Verbal Noun is a nominalized form of the verb,
which, however, still keeps its argument structure (SUBJ, OBJ complements).
The term ‘gerund’ is sometimes used just for the adverbial use:
Walking home one night, I bumped into an old friend
With students taking more Linguistics courses, we’ll need more books for the library
Very often it’s interchangeable with the infinitive:
Taking
candy from a baby (isn’t always that easy)
To take
The Gerund/Verbal Noun functions as a clausal SUBJ (see above) or as a complement to the
17. CHAPTER 2. ENGLISH INFLECTION 17
verb or to a preposition:
Tom remembered/tried/advised closing the door quietly
the trick of closing the door quietly
after/instead of/by/despite closing the door quietly
Verbal properties of Gerund/VN
When the OBJ is realized in the manner of the OBJ of a verb, the Gerund/VN is modified by
adverbs (like a verb) not by adjectives:
Continually (*continual) playing loud music (is forbidden)
No def. art. is possible (though a possessor is possible):
Tom’s (continually) playing loud music
*the (continually) playing loud music
VN of Perfect aspectual forms (VN/gerund form of HAVE):
having said that, . . .
Tom having left, we started discussing Harriet’s new book
Tom’s having left early, we had to postpone the rest of the discussion
Tom remembered having closed the door
3. Participle
By ‘participle’ we mean the form when used as attributive modifier, i.e. functioning like an
adjective modifying a noun.
Pre-nominal:
The singing detective, a dripping tap
More natural as post-nominal modifier:
the girl reading a book
anyone claiming invalidity benefits
The use is similar to that of the -en (“past”) participle:
a broken window
a letter written by a small child in blue crayon
Defective forms and unusual forms: modal auxiliaries
Modal auxiliaries lack ing forms and past/perfect participles:
*Tom is musting open the door with his credit card
[cf Tom is having to open the door with his credit card]
*The door is musted open
[cf The door is needed open; the door must be opened]
*Tom has musted open the door with his credit card
[cf Tom has had to open the door with his credit card]
18. CHAPTER 2. ENGLISH INFLECTION 18
Modals lack a special 3sg form, the default form being used instead:
Tom can(*s) speak Russian
Harriet should(*s) leave early
Although there’s a base form for all auxiliaries, there’s no to-infinitive:
*To must leave early is annoying
*To can play the organ is very satisfying
*Tom expected to will be fired
Expressions like ‘to be able to’, ‘to have to’, ‘to be obliged to’, ‘to be about to’ are often used to
fill in the lacunae in the analytic paradigms with auxiliaries.
2.4 Adjectives
2.4.1 Comparative and superlative
Formed by adding er/est to monosyllabic adjectives or adjectives ending in an unstressed syl-
lable of a certain type (oversimplifying the facts somewhat):
long longer longest
green greener greenest
happy happier happiest
noble nobler noblest
common commoner commonest
narrow narrower narrowest
Suppletive cases:
good better best
many/much more most
little less least
bad worse worst
BUT:
*frequenter/est *pueriler/est *curiouser/est
2.4.2 Comparative and superlative - inflection or derivation?
The semantics of the comparative and superlative is rather complex. E.g. longer means ‘long
to a greater extent than some reference point’
I thought the play was longer
‘King Lear’ is longer than ‘As You Like It’
The average British soundbite is longer than the average American soundbite
The superlative means ‘longer than any other’.
The question arises whether these are to be regarded as inflected forms or derived forms.
This means asking whether longer, longest are forms of the lexeme LONG or whether they’re
separate lexemes in their own right. From the semantic point of view one might wish to say
that they’re different lexemes because of the significant meaning change. It’s not obvious that
19. CHAPTER 2. ENGLISH INFLECTION 19
we want to say that this meaning (roughly MORE and MOST) is actually grammaticalized
and hence is represented as a grammatical feature. Moreover, the comparative/superlative
forms have different complementation properties, because they take a than-phrase, whereas
the positive form doesn’t permit this.
20. Chapter 3
Clitics
3.1 Introduction
Clitics are functional elements (realize functional features/categories) which don’t have their
own stress or accent and for this reason cannot be phonologically independent (i.e. they can’t
appear as free independent words). For this reason, clitics are sometimes referred to as bound
words. Because they invariably realize functional features they are similar to inflectional af-
fixes and for that reason it’s appropriate to consider them here. A number of function words in
English can appear either as clitics or as fully fledged (accented) words. It is their intermediate
status between fully-fledged words and affixes which makes the behaviour of clitics especially
complex and interesting.
Clitics are obliged to ‘lean’ on a host word (the word clitic itself comes from the Ancient
Greek word meaning ‘lean’), to which they are attached phonologically. The fact that they are
bound elements makes them similar to affixes. Like affixes they can appear either before their
host or after it. Unlike affixes true clitics can attach to hosts belonging to any part of speech.
Affix Clitic
prefix proclitic
suffix enclitic
Zwicky (1977) outlines a helpful typology of clitics. There are three basic sorts. The first
type is a phonologically reduced form of a function word which can appear accented under
the right circumstances. The clitic surfaces in the same place in the linear syntactic string of
words that the full form word would appear in. Such clitics are called simple clitics. An ex-
ample would be the reduced form of the pronoun them /@m/ as in ‘I haven’t finished’em yet’.
This has exactly the same distribution as the full form of the pronoun but it’s phonologically
attached to the previous word. In other languages we find more complex situations in which
the placement of the clitics is determined by principles specific to those clitics. For instance,
many languages (for instance, the Romance languages and many of the Slavic languages) have
special clitic forms of pronouns but they appear in specially defined positions and don’t have
the same freedom of occurrence as full form pronouns. In some cases we find that the clitics
occur in positions from which full form pronouns are normally excluded. Where the place-
ment of a clitic can’t be given by the general principles of syntax which apply to that language
but has to be determined by special principles we speak of a special clitic. The distinction is
not always easy to draw, but it’s a useful starting point.
20
21. CHAPTER 3. CLITICS 21
3.2 English auxiliary clitics
In (1 - 4) we see examples of English auxiliaries:
(1) a. it is b. it’s
(2) a. could have b. could’ve
(3) a. she shall/will b. she’ll
(4) a. we had/would b. we’d
The full forms are found (in ordinary spoken English) when the auxiliary is stressed, to em-
phasise the polarity of the sentence: We HAD locked the door (after all). The reduced forms are
cliticized to the word to the immediate left.
The phonological shape of the clitic isn’t always easy to predict from the full form. The auxil-
iaries give us the following system:
am @m (m)/m will @l/l/l
are @ " would "
@d/d
is z should @d/d
has z
have @v/v
had @d/d
Table 3.1: Auxiliary clitic forms
In addition, there are reduced forms of other auxiliaries which, however, don’t really behave
like clitics: can /k@n/, do /dU/
Some of the clitics are given in two shapes. The form without any vowel is found exclu-
sively with vowel-final subject form personal pronouns. The form with the reduced vowel is
found everywhere else, including after a vowel-final word (v. inf.).
The clitic appears in the same syntactic position as the full form auxiliary, which means
that in principle it can attach to a word of any category (in (1, 3, 4) the clitic attaches to a
pronoun while in (2) it attaches to another auxiliary). In (5) we see further exemplification of
this ‘promiscuity’:
(5) Tom’s a linguist
A friend of mine’s a linguist
The girl we met yesterday’s a linguist
The man you were talking to’s a linguist
Similar examples can be constructed for other clitics:
(6) A friend of mine’ll do it
The men you were talking to’ve left
22. CHAPTER 3. CLITICS 22
It’s possible to have a string of clitics:
(7) The boys’ll’ve been playing football
The clitics have to be enclitics. This means that we can have a clitic auxiliary in a question, in
which the subject and auxiliary are inverted:
(8) Is Tom a linguist?
*’sTom a linguist
The cliticized subjects show restricted syntactic distribution. For instance, they don’t appear
after parenthetical phrases positioned after the subject:
(9) The man you were speaking to, according to Bill, is a linguist
*The man you were speaking to, according to Bill’s a linguist
(10) The boys, unless I’m mistaken, will’ve been playing football
*The boys, unless I’m mistaken’ll’ve been playing football
In addition, they don’t appear before such parentheticals:
(11) *The man you were speaking to’s, according to Bill, a linguist
*The boys’ll’ve, unless I’m mistaken, been playing football
What this shows is that it’s necessary for the cliticized subject to form an unbroken, continuous
phrase with the VP. It can’t be separated from it by a parenthetical phrase which creates its own
intonational phrasing.
I said that it’s impossible for an inverted auxiliary to be cliticized because the auxiliary
clitics are enclitic (suffixes). However, what if we have a sentence which begins with some
phrase other than the subject? Again, the clitic must be in the same intonational phrase as the
VP whose features it realizes, as shown in (12):
(12) Why is Tom a linguist?
*Why’s Tom a linguist?
These restrictions apply to the ’s clitic corresponding to has and is. However, the syllabic
clitics, i.e. those clitics which have a reduced vowel, such as /@v/, /@l/, /@d/ occur in a
wider set of contexts. In this respect the ’s clitic is uncharacteristic. For instance, we can have
examples such as (13), in which an inversion structure is possible with a syllabic clitic auxiliary
provided it has a host to its left:
(13) The boys haven’t finished their homework and neither’ve the girls
The journal articles’ll be easy to get and so’ll the books
Similarly, wh-question words can host the syllabic clitics:
(14) Why’d they left?
When’ll they come back
Which book’ve they read?
23. CHAPTER 3. CLITICS 23
It’s not possible to follow the cliticized word with a pause or a sentence gap (this is very clear
with ’s, though less so with some of the other clitics):
(15) Tom’s a linguist and Harriet is, too
*Tom’s a linguist and Harriet’s, too
(16) The girls’ll’ve been playing football but I don’t know whether the boys will’ve
*The girls’ll’ve been playing football but I don’t know whether the boys’ll’ve
(17) Tom’d do it, and Bill would, too
*Tom’d do it, and Bill’d, too
(18) Harriet’ll come but I don’t whether Tom will
*Harriet’ll come but I don’t whether Tom’ll
(19) Harriet’ll come but I don’t whether Tom will want to
Harriet’ll come but I don’t whether Tom’ll want to
*Harriet’ll (want to) come but I don’t whether Tom’ll
(20) Tom’s a linguist but I don’t know whether Harriet is
*Tom’s a linguist but I don’t know whether Harriet’s
3.3 Other types of clitic
Function words tend not to be accented in ordinary conversation, and for this reason pretty
well any function word is prone to become permanently unaccented and hence prosodically
dependent on some other word, i.e. a clitic. In this final section we look at other cases which
are commonly treated as cliticization in English. We’ll see that in some cases the clitics have
actually developed into affixes, just as in the case of the inflected subject pronouns.
3.3.1 Pronominal object clitics in English
The object forms of pronouns are also subject to reduction and hence cliticization:
Bake me /mI/ a cake
I’ve baked you /jU ∼ j@/ a cake
Bake ’im /Im/ a cake
Bake ’er /@(r)/ a cake
Bake us /@s/ a cake
Bake ’em /@m/ a cake
Occasionally, it appears that an object pronoun has become a clitic, in that it triggers idiosyn-
cratic allomorphy on its host:
Give me a break! Gimme a break /gImI@/
However, this only happens with a very small number of verbs, usually in specific idiomatic
contexts, so it’s too early to say that we’re dealing with genuine affixation yet.
24. CHAPTER 3. CLITICS 24
3.3.2 Prepositions and determiners
The articles are prosodically clitics except when they’re specially accented (when a different
allomorph is found):
He’s THE /Di:/ Noam Chomsky
I asked for A /eI/ watermelon, not a dozen of the things
However, the indefinite article shows affix-like behaviour in that it has an idiosyncratic al-
lomorphy before vowel-initial words: an. In British dialects the same is true of the definite
article: this is pronounced /Di:/ before consonants and /D@/ before vowels.
Prepositions which have a grammatical function rather than a lexical meaning tend to get
phonologically and accentually reduced to become clitics. This is most obvious with of, to, for
(it doesn’t seem to happen to by). The process applies to to and for when they appear as the
infinitive marker and a complementizer respectively.
a cup of tea k2p@
(Note the loss of the final consonant in of here.)
the man to meet mant@
tea for two ti:f@
The clitic form of of is enclitic, but the cliticization process is less pronounced with to, for which
can also be proclitic:
Where did they send it? To you (/t@ju:)
Who’s it for? For me (/f@mi:
In jocular English the cliticized expressions cup=of and pint=of seem to have become affixed
forms, as witnessed by the spellings ‘cuppa’ and ‘pinta’. Notice also the expression toofer (tu:f@)
as in ‘toofer the price of one’.
25. Chapter 4
COMPOUNDING
4.1 Introduction
In this chapter we look at cases in which combinations of words seem to have properties of sin-
gle lexemes. There are several kinds of such multi-word combinations. The most well-known
and well studied is the compound. However, it we’ll see that it’s appropriate to consider com-
pounds in the company of other types of construction such as idioms and lexicalized phrases
of various sorts.
An important theme in this chapter will be the distinction between syntactically constructed
phrases and morphologically constructed expressions such as compounds. We’ll discuss ways
of drawing this distinction as we proceed. Another important theme (which recurs throughout
the study of morphology and the lexicon) is that of lexicalization.
At various points I shall be referring to the analysis presented in the most influential current
descriptive grammar of English, namely, Huddleston and Pullum (2002), which I shall refer to
as ‘CGEL’.
4.2 The internal structure of compounds
English shares with many languages the ability to create new words by combining old words:
houseboat, boathouse, penknife, bread knife, blackbird, and thousands of others. Although these
would appear to be combinations of two words, and hence effectively phrases, we will see
later in this chapter that in many cases such expressions behave more like single words as far
as the syntactic principles of English are concerned. It is expressions of this sort, that clearly
consist of two lexemes (content words) but which behave like a single word with respect to
syntax, that we call compound words or compounds.
I shall restrict my attention almost entirely to English compounds in this chapter (a good
survey of compounding cross-linguistically can be found in Bauer 2009). In this section we’ll
look at the way compounds are built up and how their component words relate to each other.
We often find that compounding types derive historically from types of syntactic constructions
and so we shouldn’t be surprised to find that compounds are often similar in their structure to
ordinary syntactic phrases. We’ll look in much more detail later on at ways of distinguishing
compounds from phrases. Like phrases, many compounds have structure in which there is a
main word, the head, and a non-head usually functioning as a modifier. These are the headed
or endocentric compounds, such as coffee table or blackbird. The crucial property of such com-
25
26. CHAPTER 4. COMPOUNDING 26
pounds is that they have the semantic relation of hyponymy with respect to their heads. In
simpler terms this means that a blackbird is a kind of bird (i.e. blackbird is a hyponym of bird,
or conversely bird is a hypernym of blackbird). However, it is also usual to distinguish a series of
non-headed or exocentric compounds, in which the modifier-head structure is lacking. Parade
examples are pickpocket or yellow hammer, which are not instances of pockets or hammers.
There are other types of compound which don’t fit easily into the headed/non-headed cat-
egorization. They tend to be named with the terms used by Sanskrit grammarians for such
expressions over two millenia ago. The dvandva (= ‘two-and-two’) type of compound is es-
sentially the name of a composite entity, in which both parts get named. A number of proper
names are like this: Austria-Hungary, Morgan-Grenfell, Time-Warner, . . . . Austria-Hungary was
a country/empire consisting of two smaller entities, Austria and Hungary. In a number of
cases such compounds are formed so as to serve as modifiers of another noun, as in mother-
daughter (relationship).
A similar type is the coordinative compound, illustrated by player-manager. This doesn’t
normally denote two distinct people but rather one person who satisfies both descriptions.1
Another type with a Sanskrit label is the bahuvrihi (‘having much rice’), exemplified by
lazybones, birdbrain, redhead. Here the compound can be thought of as a (usually metaphorical)
expression of the form ‘having N that is A/like that of N’, e.g. ‘ a person with a red head/brain
like that of a bird/. . . ’.
4.2.1 Headed (endocentric) compounds
In this section we look at compounds which can be plausibly analysed as consisting of a head
and a non-head. There are very few compounds headed by verbs in English2 , so we will con-
centrate mainly on noun-headed compounds and then look more briefly at adjective-headed
compounds.
A blackbird is a type of bird, a windmill is a mill, a coffee table is a table and so on. We
say that bird, mill, table are heads, and headed compounds are called endocentric. The other
part of the compound is a modifier. Thus, in (32) house is the modifier, while in (1b) boat is the
modifier:
(1) a. N b. N
N N N N
house boat boat house
It is possible to form compounds out of compounds. For instance, we can have coffee table
book, coffee table book cover, coffee table book cover design, coffee table book cover design fashion, etc, as
illustrated in (2):
1 Of course, player-manager could also be used as a dvandva, as in the expression player-manager tensions, = ‘tensions
between players and the manager’.
2 There are very many languages which do permit regular compounding with verb heads, though, see footnote 9.
27. CHAPTER 4. COMPOUNDING 27
(2) N
N N
N N cover
N N book
coffee table
There is no logical (i.e. linguistic) limit to the lengths of such compounds. This possibility of
allowing a process to feed itself ad infinitum is called recursion and we say that compounding
in English is recursive. This is an important property which makes compounding resemble
some sort of syntactic process. I shall return to this point later when we see that in most cases
it’s probably best to think of the process as being syntactic rather than morphological.
We can combine adjectives with nouns, or nouns with nouns (coffee table). We can also
combine nouns with adjectives (canary yellow, iron hard, sky blue). We can also form adjective +
adjective combinations (dark blue, icy cold). I shall return to the adjective-headed constructions
in due course. Examples such as swearword (verb + noun) and babysit (noun + verb) are rare
examples in which a verb is part of the compound. Such V N compounds are represented by
just a handful of cases and the construction is unproductive. There are many cases in which it
may appear that a modifier is a verb but this is because so many verbs in English also double
as nouns. Thus, raincoat is a N N compound not a V N compound. Finally, there are very
sporadic instances of V V compounds, generally appositional in their semantics, and usually
used as modifiers of nouns, as in drink-drive (campaign) (an advertising campaign warning
against the hazards of driving while under the influence of alcohol), stop-go (policy).
The verb-headed compounds of the kind babysit, proof-read, arm-twist and so on are almost
without exception backformations. Backformation is the creation of new words by virtue of
analogy with existing words, as opposed to word formation through the operation of regular
and productive morphological processes. Thus, English has a productive N N compound-
ing rule but no N V compounding rule (in other words, English is not a noun incorporating
language). Now, we sometimes find that words enter a language despite their being no pro-
ductive rule or principle which licenses their form. Instead, the word appears as the result of a
misanalysis of a previous word. For instance, in the history of English the verb peddle appeared
somewhat later than the word peddlar. Originally, peddlar was treated a single morpheme. As
the word became an accepted part of the vocabulary it came to be reanalysed as consisting of
peddle + er by analogy with hundreds of other words of this structure. Thus, the verb peddle
came into existence by a kind of reverse word formation process. Later in the history of En-
glish exactly the same thing happened to the word editor. This word is a loan from Latin. After
it had become established the verb edit was created, by the same analogical move which gave
rise to peddle.
Exactly the same thing has happened with babysit. Originally, a N N compound was formed
from baby and sitter meaning ‘one who sits with the baby’ or some such. But sitter clearly comes
from suffixation of the verb sit: sit + er ‘one who sits’, so speakers have permitted themselves
the licence of provided babysitter with a second parsing: [[babysit]er], which can now be treated
as deriving from a verb babysit.
In effect, words such as edit, peddle arise through a false morphological parsing of the word.
Speakers assume that the final component is an affix rather than just part of the root. This kind
of false analysis is known as folk etymology. A recent example of this is given by the word
hamburger. This has been etymologized as ham + burger. As a result the component burger has
28. CHAPTER 4. COMPOUNDING 28
come to mean ‘sandwich in bun containing cooked filling’ and has given rise to cheeseburger,
baconburger, chickenburger (and possibly other burgers which have not entered into the culinary
horizon of the author), as well as the generic word burger itself. As a piece of linguistic history
this is unexceptional, though as a matter of fact it rests on a false etymology. The original
hamburger is supposed to contain beef, not ham, of course, and the word itself comes from the
German word meaning ‘of or pertaining to the town of Hamburg’.
The example of babysit is not purely a case of false etymology, however, Unlike the case with
hamburger or edit, speakers are still aware that babysit contains baby and sit (in some sense).
Thus, they have effectively reanalysed the word, but in a fashion that violates the normal
rules of language. This is often done as a form of wordplay, and is found in poetic language.
If the idea catches on and more N V compounds are formed in this fashion then we may
reach a situation in which N V compounding becomes part of the language and hence a bona
fide morphological process. This hasn’t yet happened in English, though it’s interesting that
in technical or specialist vocabulary of various kinds we tend to find far more of these N V
compounds than in ordinary language. Maybe the N V compound process is acquiring the
mark of specialist vocabulary, in which case the process may not spread. On the other hand if
it were to become linguistically very fashionable to use specialist vocabulary in non-specialist
contexts then N V compounding might ultimately spread into ordinary language.
This patterning is summarized in Table in 4.1, where unproductive types are given in
square brackets:
Modifier/Head Noun Adjective Verb
Noun houseboat colour-fast [babysit]
Adjective blackbird red-hot slow-cook
Verb [swearword] tamper-proof [drink-drive]
Table 4.1: Compound types in English
Some of these types are much more productive and frequent than others.
Some examples of compounds whose meaning can only be figured out given the right context:
elephant gun, speed camera
Cf also the following paradigm:
sunflower oil, olive oil, whale oil, cod liver oil, . . . , baby oil
4.2.2 Adjective-headed compounds
CGEL ch. 19 §4.3 p. 1658f
There are compounds in which an adjective is modified by a noun as in sky blue, rock hard
or by another adjective as in dark blue or icy cold or in the case of fail-safe, tamper-proof by a verb.
CGEL categorizes them in the following way:
Intensifying: bone-dry, dirt-cheap, feather-light
Measure terms: ankle-deep, week-long
Incorporated complement/modifier: accident-prone, burglar-proof, user-friendly
Self -compounds: self-confident, self-evident, self-righteous
29. CHAPTER 4. COMPOUNDING 29
However, although CGEL describe these as compounds we could just as easily say that
they are adjectives with noun or adjective modifiers. CGEL: 547 describes various ways of
modifying an adjective syntactically, including by NP, as in a [two inch long] nail or a [day long]
trip. It’s not clear, then, why we should treat the regular and productive types of adjective-
headed construction as anything but syntactic phrases. We’ll return to the question of how to
distinguish compounds from phrases later in the chapter.
CGEL provides instances such as foot-loose, headstrong, threadbare, top-heavy as examples of
particularly lexicalized compounds.
4.2.3 Non-headed (exocentric) compounds
Not all compounds are headed. Although the word lazybones clearly consists of lazy and
bones, neither word is the head of the compound. An unheaded compound of this sort is
called exocentric and there are several subtypes. One type is represented by examples such
as Austria-Hungary, parent-teacher (association), mother-daughter (relationship) and with adjectives
blue-green. Here the compound is just two nouns combined with equal status and so we can
call them coordinate compounds. Where we have a semantically transparent compound we
can again say that the compound is made up of two fully-fledged lexemes, neither of which is
the sole head.
In (3, 4) we see two further examples of exocentric compounds:
(3) N (4) N
V N V
pick pocket V Particle
take off
The example in (4) actually shows a case of conversion, of a particle verb (phrasal verb) (to) take
off into a noun (a) take-off. The result is a noun which consists of two words but which lacks a
head in the ordinary sense. Although verbs rarely enter into headed compound constructions
they are rather more common in unheaded compounds such as (3, 4). However, exocentric
compounding other than the coordinate compound type is unproductive in English.
4.2.4 Morphophonological properties of compounds
Standard examples of compounding involve fully-fledged words such as coffee or pick. English
also has compounds consisting entirely of bound morphemes. These are often known as neo-
classical compounds. Some examples are given in (5):
(5) anthropology logorhea erithrocyte
anthropomorphic rheostat cytoplasm
morphology elasmobranch hydrogen
Such words are formed from Greek (sometimes Latin) roots which are not generally used on
their own. Sometimes, one of these roots does correspond to a word, as in biosphere (sphere) or
30. CHAPTER 4. COMPOUNDING 30
psychology (psycho). In the case of psycho (and many other such examples) the word is formed
by clipping of a fuller neo-classical compound (e.g. psychosis or psychotic). The examples in (5)
can be segmented as follows:
(6) anthrop-o-log-y log-o-rhe-a erithr-o-cyte
anthrop-o-morph-ic rhe-o-stat cyt-o-plasm
morph-o-log-y elasm-o-branch hydr-o-gen
It is difficult to know how best to analyse the components of neo-classical compounds. On the
one hand they are not really affixes, since many of them can appear at the beginning or the
end of the word (e.g. cyte). On the other hand, they are not obvious examples of roots because
they generally can’t appear unless they are attached to some other similar form (e.g. erithr-
or -plasm). In some cases the element may have a preference for the beginning of the end of
the word. Bauer (1983) refers to these respectively as Initial Combining Forms (ICF) and Final
Combining Forms (FCF). For example, erithr- doesn’t seem to occur finally, so this would be
an ICF. Many, however, are like cyte in functioning both as ICF’s and FCF’s.
Returning to the compounds in (5/6) we see that in each case the two elements are sepa-
rated by an -o- element (see later for discussion). Two examples are illustrated in (7):
(7) N
? intermorph ?
erithr -o- cyte
cyt plasm
In other languages it’s not uncommon for the elements of most or all compounds to be sepa-
rated from each other by special linking elements of this kind and in languages such as German
or Dutch we find a variety of such elements being used, depending on the words combined,
their phonological or morphological structure and so on. Such elements are sometimes called
intermorphs (or intermorphemes) as shown in (7). They are another example of meaningless
morph(eme).
Compounds in other languages may exhibit other morphological properties. In some lan-
guages endocentric compounds are left-headed rather than right-headed as in English, while
other languages may make much use of exocentric (non-headed) compounds.
4.2.5 Morphological properties of compounds
English compounds in the majority of cases adhere to a principle which is fairly widespread
in compounding processes throughout the world’s languages: the internal (non-head) com-
ponents can’t be inflected. For verbs this is not easy to demonstrate because compounding
doesn’t productively allow a verb to be a modifier. Nonetheless, in those rare examples of V
X compounds we never find uncontroversial inflection, i.e. Past Tense forms or 3sg agreement
forms. Thus, alongside swearword we would never find *sworeword or *swearsword. The situa-
tion is rather different with ing participles, however, which can be interpreted as nouns. Thus,
compounds such as riding boots or training centre are not actually V N compounds but N N
compounds. Thus, we can conclude that inflected forms of verbs never appear as the modifier
in an endocentric compound. A similar situation obtains with adjectives: when we have an A
31. CHAPTER 4. COMPOUNDING 31
N compound the adjective doesn’t inflect for comparative and superlative forms. Indeed, it’s
difficult to find even lexicalized compounds based on comparatives or superlatives.
The situation is slightly different with plurals of nouns. We’ll just consider N N compounds
and to start with we’ll have to consider the meaning of a noun in more detail. A concrete noun
used in sentence will generally be able to refer to a concrete object or collection of objects, i.e.
it will be part of a phrase which is referential:
(8) a. The cat is sitting on the mat
b. A cat is sitting on the mat
c. (Some) cats are sitting on the mat
The functional system of English (i.e. the set of functional features including inflections and
function words) forces us to make explicit information about number and definiteness. How-
ever, the noun cat on its own doesn’t express such information, it simply conveys the general
concept of cathood. The noun can only denote a specific, definite, indefinite, etc. cat or cats
when it appears in a specific inflected form modified by specific function words as part of a
syntactic phrase. Now, in a N N compound a lexeme such as cat isn’t accompanied by any
exponents of functional features. It therefore can’t refer to any particular cat or cats, but can
only denote the general concept. For this reason, the ‘cat’ in the compound catfood doesn’t pick
out any concrete cat at all. We say that cat fails to refer, i.e. that it is non-referential3 .
Given that the modifier component of a N N compound is non-referential it isn’t surprising
that it fails to attract those types of inflection which would make it referential, such as num-
ber marking. If [Number:Plural] is simply the morphological exponent of a syntactic feature
[NUMBER PLURAL] this is not surprising. Thus, catfood may be thought of as ‘food for cats’,
but actually it’s better to think of the meaning as more like ‘food for the generic cat’. The vast
majority of N N compounds in English are like this. Because N N modifiers are non-referential
it’s rare to find proper names inside compounds – the proper name in its canonical usage has
to refer to a concrete individual, which means that it’s inappropriate as a noun modifier.
However, there are cases in which plurals do appear inside compounds. In some cases this
is simply because the plural form of the word has undergone semantic drift or specialization
and is therefore no longer merely a form of the noun lexeme. For instance, the noun systems
in systems analyst isn’t just the plural of system. Rather, it’s acquired a specialist, technical
meaning, and hence, effectively, has become a new lexeme. The semantic drift has, in other
words, turned an inflected form into a derived lexeme. In other (admittedly rather rare) cases,
though, we encounter what appear to be bona fide plurals in compounds: teeth marks, parks
commissioner, prisons inspectorate and so on. In each case, however, we can argue that the plu-
ral has been reinterpreted as a kind of collective noun and hence has undergone some slight
degree of semantic drift. The crucial point is that we never find is a systematic singular-plural
distinction inside compounds. A hypothetic example of such a distinction would be the fol-
lowing. Suppose there were a folder which was designed to store individual compact disks,
and suppose there was also a folder which stored more than one disk at a time. We wouldn’t
distinguish the two types of folder by calling the first a disk folder and the second a disks folder.
Rather, we’d use the term disk folder for each and if we wished to distinguish them we’d prob-
ably resort to calling the first type a single disk folder or some such. It is in this sense that we
can’t productively make a number distinction inside compounds (in English).
A feature of compounds that is often cited is that it’s impossible for the modifying ele-
ment to be a syntactically formed phrase. For instance, we can have compounds such as book-
shop but not [expensive book]shop (meaning ‘a shop for expensive books’) or [books about music]
3 Recall the discussion of referring expressions in the ‘Meaning’ segment of the module.
32. CHAPTER 4. COMPOUNDING 32
shop (meaning ‘a shop selling books about music’). This is sometimes expressed as the ‘No
Phrase Constraint’, which, interestingly, violates itself, because No Phrase is itself, apparently,
a phrase. What does seem to be possible is for a noun to be modified by a phrase which has
been lexicalized to some degree. Thus, although there is no compound *expensive book(s) shop
but we can say second-hand book shop, in which second-hand appears to be a syntactic modifier of
book. Likewise, we can have examples such aerial acrobatics team, infectious diseases specialist, but,
joking apart, we can’t have *[dangerous acrobatics] team or [disgusting diseases] specialist, with the
bracketings shown (and with the intended interpretations ‘team performing dangerous acro-
batics/specialist in disgusting diseases’). Of course, this raises the question of what counts as
‘lexicalized’.
4.3 Compounds vs. phrases
On the face of it a compound is very similar to a phrase: both are higher level units consisting of
more than one word. Indeed, a productive (as opposed to a lexicalized) compound of necessity
consists of more than one lexeme. A consequence of this is that a string consisting of, say, a
Determiner + Adjective + Noun can in principle be given two distinct analyses, in which the
[A N] constitutes a phrase (NP) as in (9) or a compound noun (AN) as in (10):
(9) DP (10) DP
D NP D NP
a N’ a N
AP N A N
A bird black bird
black
However, compounds are generally said to have a number of properties that distinguish them
from phrases, including phonological, semantic and syntactic properties. We’ll briefly survey
these properties in turn.
4.3.1 Phonological properties of compounds
The main phonological property distingushing compounds in English from phrases is stress
or accent. An ordinary syntactically formed phrase spoken in a pragmatically neutral context
typically bears a main accent on its rightmost lexical word. ‘Pragmatically neutral context’
means that there’s no special emphasis or focussing, so that the phrase is spoken as it would
be when the speaker wishes to make the most general statement in which all the content words
are equally informative. For instance, in response to a question such as ‘what is the cat looking
out of the window at?’ we might say:
(11) that large black BIRD
As indicated, the main accent will then fall on the word ‘bird’, because we aren’t emphasizing
that the bird is specifically large or black. However, if the conversation had already introduced
the topic of black birds of various sizes and the speaker wished to emphasize that the cat was
looking at the large one so as to correct the misapprehension that the cat was looking at the
small one she might say (12):
33. CHAPTER 4. COMPOUNDING 33
(12) No, he’s looking at that LARGE black bird, (not the small one)
If it was the colour of the bird that was at issue then the speaker might say:
(13) No, he’s looking at that large BLACK bird, (not the brown one)
The principle that says that it’s the final lexical (content) word which is assigned accent
in neutral discourse we’ll call the principle of Phrasal Accent or Phrasal Stress. The point
of Phrasal Accent is that compounds often behave differently. Thus, the normal way of pro-
nouncing the compound word blackbird in a neutral context would be with stress on black:
(14) The cat’s looking at that large BLACKbird
This pattern of initial stressing is called Compound Accent or Compound Stress.
Now, this would be a very good diagnostic for identifying compounds if it were without
exceptions. Unfortunately, there are two respects in which the phonological criterion is less
than fully helpful. First, we sometimes find that the normal Phrasal Accent puts the stress on a
word other than the final lexical word. However, this is very unusual with, say, NPs consisting
of AP + N, so this ambiguity will seldom arise. More significantly, we frequently find cases
in which we have a fairly clear example of a compound but with phrasal accent. Attempts
have been made to determine when a compound will appear with phrasal stress but so far
none have successfully accounted for all cases. In particularly, there are some contrasts which
appear to be irreduceably lexicalized. Why, for instance, do we say town h´ ll but t´ wn house?
a o
Some well-known examples involve streets and cakes. Consider the compounds in (15 -
16):
(15) a. Abbey Road
´ (16) a. mince p´e ı
b. Penny L´ ne
a b. treacle t´ rt
a
c. Peyton Pl´ ce
a c. custard cr´ am
e
d. ´
Fifth Avenue d. apple turnover
´
... ...
´
e. Oxford Street e. c´ rrot cake
a
All of these examples look rather like compounds (and pass the syntactic tests for compounds
to be discussed below). In particular, they are all N N combinations, a typical compounding
structure. However, only the last examples, headed by street and cake actually have compound
stress. US and British English even differ on the stressing of some cases, e.g. UK ice cr´am
e
(phrasal stress) vs. US ´ce cream (compound stress).
ı
4.3.2 Semantic properties of compounds
The productively formed compounds we’ve seen so far are semantically compositional. The
semantic interpretation rules may be somewhat complex but they are no more so for phrases
than for compounds. As we’ve seen, the interpretation of a nonce compound will generally
depend substantially on pragmatic factors, but the same can be said of phrase interpretation.
But how do we interpret the semantically transparent compound formations? We’ll return to
this question below in §4.3.4
However, not all compounds are compositional. Many of the compounds that have entered
into common use have become lexicalized. A well-known and oft-cited example of such a lex-
icalized compound is blackbird. As a consequence of lexicalization its semantic interpretation
34. CHAPTER 4. COMPOUNDING 34
is not related in a straightforward way to the meanings of its constituent nouns, that is, the
semantic interpretation is not compositional. In the case of ‘blackbird’ there are two senses in
which this is true.
First, we find that the denotation is semantically restricted. In the modern language the
compound blackbird doesn’t denote just any entity which is both a bird and is black. A crow
is not a blackbird, and neither is a black swan (though both are black birds, of course). In fact,
the term ‘blackbird’ applies solely to the species Turdus merula. In other words, the term only
applies to a proper subset of the things it would apply to if its interpretation were completely
free, as in the case of the phrasal construction. For this reason, it’s possible to say without
contradiction ‘This black bird is not a blackbird’.
A more drastic deviation from compositionality is seen when we observe that female black-
birds are actually brown. Thus, it’s perfectly logical to say ‘This blackbird is not black’. Even
without brown blackbirds, we could easily imagine an albino blackbird, or a blackbird that had
been dyed green, and so on, in which case we could say without contradiction ‘This blackbird
is white/green/. . . ’.
Of course, not all compounds show these deviations from compositionality. This happens
only when the compound has been fixed in the lexicon of speakers and has been allowed to get
semantically restricted, or to ‘drift’ semantically, or both. In this respect, blackbird is different
from, say, seabird. A seabird is any bird that has some appropriate connection with the sea. The
term doesn’t denote a particular kind of seabird (though in principle it could have done, of
course). In the case of seabird we can reasonably say that the meaning of the whole is derived
from that of the parts, much as in the case of morphology lecture. In the case of blackbird, however,
we have to say that the semantic component is no longer ‘visible’ or ‘active’ in the compound.
All we have is the phonology. But equally, we have no particular reason to provide the black
component with a syntactic label, either. Thus, we must say that the structure of blackbird is as
in (17):
SEM: [BIRD: Turdus merula]
SYN: N
PHON: blak + b@:d
(17)
SEM: [BIRD]
[PHON: blak] SYN: N
PHON: b@:d
As a kind of shorthand I’ve indicated that the semantic representation for ‘blackbird’ (the sci-
entific name Turdus merula) is a subtype of the semantic class of birds. This allows us to infer
(correctly) that a blackbird is a bird but doesn’t allow us to draw any other conclusions. More-
over, the compound has to have its own lexical entry with its own semantic representation
because the component Turdus merula can’t be inferred from either of the components. Since
the modifying part black doesn’t have any syntactic or semantic properties it can only add its
phonology to the construction as a whole.
The fact that the ‘black’ component of blackbird lacks any semantics means that it is a mean-
ingless morpheme (like the cran- of cranberry). The fact that it resembles the adjective black
shouldn’t fool us. Any compound which has been subject to semantic restriction or drift and
which therefore is not semantically compositional (or at any rate is not as semantically compo-
sitional as a syntactic phrase would be) will have the same properties. It turns out that most
A N compounds are either like blackbird or are exocentric, like yellow hammer4 . But if this is so,
then we can hardly say that English has A N compounding at all as a productive process.
4a different species of bird, Emberiza citrinella.
35. CHAPTER 4. COMPOUNDING 35
It’s not just A N compounds that show the effects of lexicalization. There are innumerable
N N compounds whose semantics is non-compositional (or which have a specialist meaning
fixed in the dictionary which is non-compositional, alongside a compositional interpretation).
Those non-compositional readings have to be treated exactly like blackbird. For instance, a
word such as textbook would appear to consist of TEXT and BOOK. However, a textbook, whilst
clearly a book, is a special type of book. But this semantic restriction can’t be derived from a
knowledge of (any of the) meanings of the word text. Rather, we have to say that text is another
cranberry element. Such cases are rather common.
In many cases the head of the compound will retain its meaning but even this isn’t guaran-
teed. The word crayfish looks as though it consists of the cranberry element cray- + the lexeme
FISH , but for many people (including me) it’s very odd to call a crustacean a fish.5 Thus, in this
case we would have to say that both components were meaningless forms. Semantic drift pro-
vides many comparable examples. When hobbyhorse means ‘pet theme or activity’ it doesn’t
denote a kind of horse (not even a toy one). On the other hand, for me the noun scuttle doesn’t
exist outside of the compound coal scuttle.6 Now a coal scuttle is indeed a container for coal, so
the modifier component is interpreted compositionally, but since the head word scuttle doesn’t
exist outside the compound we’d have to regard the modifier coal as contributing no additional
meaning but merely repeating a component of meaning that is already implicit (this is called
pleonasm).
The upshot of this discussion is that a large number of compounds in English contain mean-
ingless elements and not a few consist entirely of meaningless elements (a fact which is not
always fully recognized in discussion of the subject).
4.3.3 Syntactic properties of compounds
I said earlier that compounds exhibit the property of lexical integrity, in that syntactic pro-
cesses are unable to gain access to the individual words which make up the compound. In
this respect, a true compound has the same syntactic properties as a monomorphemic word.
However, compounds often arise historically from earlier forms of syntactic construction, and
such constructions may undergo grammaticalization to different degrees. As a result, it often
turns out to be very difficult to draw a clear dividing line between the syntactic properties of
compounds and phrases.
In straightforward cases compounds clearly differ from phrases in their syntactic structure:
in the phrasal structure in (9) the adjective in effect heads an adjective phrase and this AP is
attached to the noun to form a syntactic unit which is no longer categorially a word in any
sense, but a genuine multiword phrase. There are two ways in which this is reflected in the
case of black bird. First, we can replace the single adjective with a full adjective phrase in (9)
without significantly changing the overall syntactic structure. In particular we can modify the
adjective with a degree term such as very, somewhat or surprisingly:
(18) a [N [AP rather black] bird]
Similarly, we can interpose a phrase such as another AP or a parenthetical expression between
the adjective black and the noun it modifies in the phrasal structure:
(19) a [N [AP rather black] and [unusually large] bird]
5 The origin of the word is folk etymology - it arises from a mispronunciation of the Old French word escrivisse (Mod. Fr.
ecrevisse)
´
6 My Chambers dictionary gives the definition ‘a shallow basket; a vessel for holding coal’. I never use the word in the
sense of ‘shallow basket’.
36. CHAPTER 4. COMPOUNDING 36
(20) a black, or possibly just very dark brown, bird
These manipulations are impossible with the compound word. Similarly, neither the head of a
NN compound nor its modifier can be pronominalized:
(21) a. *We saw a blackbird and a sea one
b. *By the sea we saw many it-birds
We can summarize these observations by saying that a compound seems to have the syntactic
distribution of a single word. In the terminology of syntax we can say that they behave like X°
categories rather than phrasal categories.
The composite nominal construction
We should distinguish lexical compounding of this kind from a very similar construction
which CGEL (ch. 5 §14.4 p. 448f) refers to as the composite nominal. In this construction
a noun is modified by another noun but the combination is regular and productive and is best
regarded as part of the syntax. Consider for instance morphology lecture. This has the same
structure as a compound noun such as coffee table, but its syntactic properties are entirely dif-
ferent. A true compound behaves as though it were a single word with respect to syntax. It
exhibits, in other words, what is often called lexical integrity: it can’t be split up and its parts
can’t be referred to independently elsewhere in the sentence.
CGEL (p. 449) outlines five diagnostics for identifying composite nominals.
1. coordination in the modifier
2. coordination in the head
3. ‘delayed right constituent coordination’
4. modification within the modifier
5. modification within the head
These are illustrated by various types of manipulation we can perform on the expression
London colleges:
(22) Coordination in the modifier: various [London and Oxford] colleges
(23) coordination in the head: various London [schools and colleges]
(24) delayed right constituent coordination: [two London and four Oxford] colleges
(25) modification within the modifier: two [south London] colleges
(26) modification within the head: two London [theological colleges]
All of these manipulations are typical of Adjective + Noun syntactic phrases (i.e. phrases like
(a) black cat). They are not found with true compounds:
(27) Coordination in the modifier: *various [coffee and dining] tables
(28) coordination in the head: *various coffee [tables and mugs]
(29) delayed right constituent coordination: *[two coffee and four dining] tables
37. CHAPTER 4. COMPOUNDING 37
(30) modification within the modifier: *two [decafinated coffee] tables7
(31) modification within the head: *two coffee [wooden tables]
4.3.4 Compounds and composite nominals: structure
If the composite nominal construction is part of the syntax of English we expect it to be com-
pletely productive. In fact, of course, the formation of composite nominals is governed in large
part by their meaning and semantic plausibility, as any other phrase. Nonetheless, it’s worth
considering just how easy it is to construct composite nominals of the form N N in ordinary
conversational English. For instance, imagine you’re decorating or moving house and explain-
ing to people where things have to be stored. Suppose you have three tables piled with boxes
of possessions. One table has boxes of books, another boxes of computer equipment and a
third has boxes of kitchen utensils. Under such circumstances you might easily find yourself
asking someone to put one of your potplants on the book table. You might even ask someone
to fetch a table lamp from under the ‘kitchen table’ even though the table is not a table which is
in a kitchen but rather the table which happens at the moment to have boxes of kitchen things
on it.
There’s no set way of determining exactly what a composite nominal means on a given
occasion of use when it’s a nonce formation (i.e. when the compound is coined in a specific
context as in our house moving scenario). The way we interpret such compounds is by us-
ing common sense and figuring it out from the context and from the assumptions and beliefs
we share with our interlocutors (Downing 1977). In other words we use pragmatic strategies
to identify the meaning of the compound word. If these assumptions are false then misun-
derstandings can occur. For example if I believe that the boxes with books in them (or ‘book
boxes’) actually contain videos then I may not succeed in determining the referent of the nonce
term ‘book table’. And, of course, I may mistakenly go to the kitchen to look for the ‘kitchen
table’ (again, Downing 1977 discusses this in some detail).
So how are such NN compounds and composite Ns interpreted? The standard assump-
tion is that the compounding/composition process introduces an empty semantic relationship
between the two nouns. This relation is often denoted by the symbol ‘R’ (or variants thereof,
such as R or ℜ). Thus, Downing’s example of bike girl would be represented semantically
as ℜ<bike, girl> meaning ‘girl who has some pragmatically specified relation to the concept
‘bike”. The relation could be ‘girl who comes to work on a bike’, ‘girl who we saw mending
her bike while we were out walking’, ‘girl who works in the bike shop’, . . . , or even ‘girl who
refuses to ride a bike’.
Now, attempts have been made to analyse all N N combinations (whether composite nomi-
nals or compound nouns) in terms of the operation of a fixed set of semantic relationships. For
instance, we often find taxonomies of N N compounds which distinguish between modifiers
meaning ‘made out of a material’ (iron bar, brick wall), ‘location’ (kitchen sink, rabbit hutch), ‘part
of’ (car engine) and so on. However, the semantic primitives used are sometimes ambiguous
(should with think of ‘car’ as expressing the whole of which ‘engine’ is a part or as the location
of the engine, for instance), and in some cases hopeless vague, such as ‘for’. Even with com-
pletely unspecified semantic primitives such as ‘for’ we encounter difficulties in providing a
semantic representation for some compounds (do we really want to gloss elephant gun as ‘a gun
for elephants’, for example?). Some attested compounds defy any kind of principled analysis.
A particularly clear example of this is the British English term speed camera. This means ‘cam-
era used for filming motorists in order to provide evidence of violation of speed restrictions’.
7 It’s
actually very hard to imagine how to construct an example of this sort from coffee table, because the meaning of the
whole compound is rather idiosyncratic. This, of course, is exactly what we expect from a compound.
38. CHAPTER 4. COMPOUNDING 38
The compound is perfectly easy to interpret from its components in the appropriate context,
but there’s no way that this interpretation can be said to follow from straightforwardly gram-
maticalized semantic processes.
Part of the reason for this indeterminacy in interpretation is, of course, to be found in the
origins of compound nouns. They start out typically as composite nouns, in which the se-
mantic relationship between modifier and head is determined by general knowledge or by
the context of first utterance. Essentially we are simply expected to use our common sense to
figure out the meaning of a compound, and if that fails (because we aren’t familiar with the
origins of the word) we have to learn its meaning the same way we would learn the meaning
of a monomorphemic word.
We can conclude that in English there are semantically regular examples of compounds (as
opposed to composite nouns form in the syntax) that are formed out of existing lexemes which
have a recognizable meaning, and the meaning of the compound as a whole is derived from
the meaning of its parts, even if this has to be mediated by pragmatics. In representing such
compounds it’s therefore necessary to treat them as structurally comprised of two lexemes, as
shown in (32):
(32) N
N N
HOUSE BOAT
Actually, this is better thought of as a three-dimensional representation, combining phonology,
syntactic structure and semantics. Let’s say that HOUSE and BOAT have the two lexical entries
in (33) (ignoring the irrelevant MORPH feature):
SEM(HOUSE): [BUILDING FOR INHABITATION]
(33) a. SYN(HOUSE) N
PHON(HOUSE): /haUs/
SEM(BOAT): [WATER-GOING VESSEL]
b. SYN(BOAT): N
PHON(BOAT): /boUt/
Recall that what the representations in (33) are telling us is that there are two lexemes with la-
bels or ‘addresses’ HOUSE and BOAT respectively. The PHON component of the HOUSE lexeme
is /haUs/, its SYN label is ‘N’ and so on.
We can now ask ourselves how a compound word is formed. There are three main aspects.
First, we add the basic phonological representation of the first lexeme to that of the second:
/haUs+boUt/. Second, we need to represent the fact that the compound consists of two nouns
in accordance with the language-specific principle of English which licenses such N N com-
pounding (bear in mind that not all languages permit such compounds, with the types of
interpretation that English allows - the Romance and the Slavic languages, for instance, don’t
have such compounds). Clearly, we also need to be able to say that we combine the semantic
representations in some way in order to produce a new semantic representation enriched with
real world information about universities and their structure, about linguistics and so on, to-
gether, perhaps, with contextually determined beliefs and assumptions. Moreover, there is a
crucial additional semantic fact, which is that the compound lexeme as a whole denotes some-
thing of the same semantic type as the second component, whose precise meaning is obtained
by modifying it with the meaning of the first lexeme. In other words, we need to reflect the
39. CHAPTER 4. COMPOUNDING 39
semantico-syntactic fact that the second noun is the head of the compound and the first is its
modifier.
The overall structure is shown in (34):
[BOAT] related to concept [HOUSE]
N
/house+boUt/
(34)
[HOUSE] [BOAT]
N
N
/haUs/ /boUt/
4.3.5 Synthetic compounds
There’s one important class of compounds that remains to be discussed. Consider the exam-
ples in (35 – 37):
(35) train driver
(36) a. horse riding
b. slum clearance
c. soil management
(37) a. hand-painted
b. pan-fried
c. alcohol-dependent
d. government-sponsored
Example (35) is a case of a synthetic compound (also called a verbal compound or verbal
nexus8 ). Examples such as (36a) would be universally regarded as belong to the same class
of compounds as (35). In the case of (36b–d) matters are a little more controversial, though
the similarity in structural relations is certainly striking. Some linguists would also regard the
examples of (37) as synthetic compounds, though this is less widely accepted. Other types of
compound, in which the head is not a deverbal noun, are known as root compounds.
These compounds are all related to syntactic phrases, as in (38–40):
(38) a. drive a train
b. driver of trains
(39) a. ride/riding a horse
b. clear slums
clearance of slums
c. manage the soil
management of the soil
(40) a. painted by hand
b. fried in a pan
c. sponsored by the government
8 CGEL: 1652 fn. 24 objects to the use of the term ‘synthetic compound’, for good reason, but unfortunately it’s become
entrenched in the linguistics literature.