Words and pieces of words

This chapter looks at the problems encountered in identifying and defining the notion ‘word’. It then discusses the identification and description of ‘morphemes’ (pieces of words). Finally, it looks at the way in which words can be assigned to ‘word classes’ (parts of speech).

The word appears to be a widespread concept. Even in primitive cultures, informants are often able to identify words. This is somewhat surprising, because nobody has yet proposed a satisfactory universal definition of the notion ‘word’, or provided a foolproof method of identification. People sometimes wrongly assume that a word is recognizable because it represents a ‘single piece of meaning’. But it can easily be shown that this view is wrong by looking at the lack of correspondence between words from different languages. In English, the three words cycle repair outfit correspond to one in German, Fahrradreparaturwerkzeuge. Or the six words He used to live in Rome are translated by two in Latin, Romae habitabat. And even in English, a word such as walked includes at least two pieces of meaning, ‘walk’ and ‘past tense’.

[]()Insight

The word appears to be a widespread concept, yet it is very difficult to define.

This chapter will deal with this matter. First, it will look at the problems of defining and identifying words. Second, it will consider pieces of words, or morphemes.

[]() Defining words

The best-known definition of a word is that proposed []()by the American linguist Bloomfield, who defined it as a minimum free form, that is, the smallest form that can occur by itself. This is fairly unsatisfactory, because words do not normally occur by themselves in spoken speech. Even if you ask a simple question, a normal-sounding reply often requires more than one word:


Who did that?	John did.
What’s that?	An oak tree.

Furthermore, some apparent words, such as did, the, and, are found alone only in exceptional circumstances, such as in answer to the question: ‘What does a-n-d spell?’

Bloomfield’s definition works best for written English, where we conventionally leave a space on either side. But linguists are concerned primarily with the spoken word, not the written, and the two do not necessarily coincide. For example, it seems to be purely accidental that the name of a certain type of snake, a boa constrictor, is written as two words rather than one, or that seaside appears as one word, but sea shore as two.

Why have linguists found it so hard to find a satisfactory definition of the notion ‘word’? The answer seems to be that there are different types of word. Consider the rhyme:

A flea and a fly in a flue

Were imprisoned, so what could they do?

Said the flea: ‘Let us fly’.

Said the fly: ‘Let us flee’.

So they flew through a flaw in the flue.

At the simplest level, this rhyme contains 36 written words. But some of these are repeated. If we decide to leave out repeats and count the number of different words []()(in technical terms, count word types instead of word tokens), we come up against several problems. Should fly (noun) and fly (verb) be counted as the same, since they sound the same, or as different, because they have different meanings? Should fly and flew be regarded as the same, because they belong to the same verb, or as different because they have different forms? These problems can be solved only if we decide what kind of ‘word’ we are talking about. It is important to distinguish between lexical items, syntactic words and phonological words.

If by ‘word’ we mean lexical item (the technical term for ‘dictionary entry’), then the sound sequence /flaI/ ‘fly’ represents two words, since most dictionaries have separate entries for fly (noun, N) and fly (verb, V):

[]()fly N: an insect with two wings.
[]()fly V: to move through the air in a controlled manner.

This is perhaps the most basic, and most abstract use of the word ‘word’. However, both of these lexical items have various syntactic forms associated with them. The insect could occur as fly (singular) or flies (plural), and the verb could occur as fly, flying, flies, flew, flown. So if we counted the various syntactic forms as different words, the overall total would be much higher ().


Lexical items	Syntactic words
fly N	fly flies
fly V	fly flyingfliesflew flown

[]()Figure 6.1.

A further complication occurs with a lexical

[]()

item such as

flaw

. This has the two syntactic forms

flaw

(singular) and

flaws

(plural). But the singular form

flaw

then has two different sound sequences associated with it, /fl

/ before a consonant, and /fl

r/ before a vowel (

[]()The flue had a flaw /fl <Image src="../images/f596-01.jpg" alt="Image" /> <Image src="../images/icon09.jpg" alt="Image" />/ which allowed the fly to escape.
[]()There was a flaw /fl <Image src="../images/f596-01.jpg" alt="Image" /> <Image src="../images/icon09.jpg" alt="Image" />r/ in the flue.

[]()Figure 6.2.

These examples show that we must not expect an exact overlap between different types of word. And in some other languages, the situation is far more complex than in English. In Latin, for example, the lexical item rosa, ‘rose’, has 12 different syntactic forms. In Welsh, the initial consonant of each word varies systematically, depending mainly on the preceding sound: the word for ‘father’ could be tad, dad, thad, or nhad. The last lines of the chorus in a well-known Welsh hymn have three different forms of the verb meaning ‘sing’: canu, ganu and chanu – and there is a fourth possibility, nghanu, which the hymn omits.

[]()Insight

The word ‘word’ is difficult to define because there are different kinds of word.

[]() Identifying words

For anyone working on an unknown language, it is important to identify these various types of word. There are two main stages in the analysis. First, []()finding chunks such as fly, flew, which recur as self-contained units. Second, deciding how many lexical items are covered by each chunk (as with fly, which covers two lexical items), and conversely, deciding how many different chunks belong to the same lexical item (as with fly, flew, where different syntactic forms belong to one lexical item).

For the first stage, finding chunks which behave as self-contained units, we look for sequences which are uninterruptible and mobile. These are useful guidelines in many languages. A sequence such as chickens cannot be interrupted. It is impossible to say \_chick-little-ens_, or \chicken-little-s. In addition, the sequence chickens can move about. It can occur next to different words, and in different parts of the sentence, as in: Chickens lay eggs, foxes eat chickens, the chickens clucked loudly, and so on.

[]()Insight

Chunks that are uninterruptible and mobile are likely to be syntactic words.

To take another example, suppose we had come across the sequence greentrousers, and wanted to know whether this was one or more words. We would begin by looking for sentences which included any part of the sequence greentrousers. We might find:

[]()Green leather trousers, Red trousers, Green shirts.

The fact that greentrousers can be interrupted by the word leather indicates that we are probably dealing with at least two words, green and trousers. This suspicion is confirmed by noting that both green and trousers occur with other words. But since green and trousers seem to be uninterruptible (we do not find \*trous-green-ers, for example), we surmise that each is a word.

At the end of this stage of the analysis, we have a rough list of ‘words’, though a list in which we are likely to have clumped together different lexical items which sound the same (homonyms), and to have separated different syntactic forms of the same lexical item.

For the second stage of the analysis, []()we need to consider the syntactic behaviour of these possible ‘words’, that is, their role in the overall sentence pattern. For example, fly N would show up as behaving differently from fly V, since each would fit into a different ‘slot’ in the sentences:

[]()The fly buzzed.
[]()Birds fly.

On the other hand, fly and flew would turn out to be somewhat similar, in that they would fit into the same general slot:

[]()They fly home on Sunday.
[]()They flew home on Sunday.

However, the syntactic behaviour of these different forms can be supplemented by an analysis of their make-up, or, in other words, the morphemes out of which they are constituted. Let us therefore go on to consider some basic facets of morphology.

[]() Morphemes

The smallest syntactic unit is the morpheme. Morphemes vary in size. Neither syllables nor length are any guide to their identification. The essential criterion is that a morpheme cannot be cut up into smaller syntactic segments.

The sentence in has eleven morphemes:

[]()Figure 6.3.

The, albatross, a, lullaby, are all single morphemes because none of them can be syntactically split up further. Alba- and -tross, for example, do not have any other []()role to play in the syntax of English: they exist only as part of the single unit, albatross. Chanted and dreamy, on the other hand, each consist of two morphemes: chant is found in words such as chanting, chants, and is also a word by itself, while -ed is found in wanted, batted and so on. Similarly, sleep-walking consists of three morphemes, because sleep, walk and -ing are all found elsewhere. In theory there is no upper limit to the number of morphemes per word: antidisestablishmentarianism, for example, has at least six: anti-dis-establish-ment-arian-ism.

[]() Recognition of morphemes

Linguists identify morphemes by comparing a wide variety of utterances. They look for utterances which are partially the same ():

[]()Figure 6.4.

[]()Insight

Morphemes are recognized by looking for partial similarities between words.

The partial similarity between sniffed, grunted, plodded and edged enables one to isolate the segment -ed. And the partial similarity between arrogantly and loudly, and between backwards and forwards, makes it possible to isolate -ly and -wards.

In Turkish, the similarity between adamlar, ‘men’, and kadınar, ‘women’, enables one to identify a plural suffix -lar, and the words for ‘man’, adam, and ‘woman’, kadın. In Swahili, the overlap between:


nitasoma	I will read
nilisoma	I read (past)
utasoma	you will read
ulisoma	you read (past)

allows us to identify soma, ‘read’; ni, ‘I’; u, ‘you’; ta, future tense; li, past tense.

Not all morphemes are as easily segmentable []()as these examples. But the identification of morphemes is done wholly by means of this one basic technique – the comparison of partially similar utterances.

[]() Types of morpheme

Morphemes such as albatross, chant, lullaby, which can occur by themselves as whole words are known as free morphemes. Those such as anti-, -ed, -ly, which must be attached to another, are bound morphemes. Bound morphemes are of two main types. Consider the sentence:

[]()The owl look-ed up at the cloud-y sky.

Superficially, both looked and cloudy have a similar make-up, consisting of one free morpheme, followed by a bound one. Yet the bound morphemes differ in nature: -ed on the end of look_ed_ is an inflectional morpheme, since it provides further information about an existing lexical item, look, in this case indicating that the looking occurred in the past. Other examples of inflectional morphemes are the plural, as in owls, and the possessive, as in Peter’_s_ car. However, -y on the end of cloudy behaves rather differently. It is a derivational morpheme, one which creates an entirely new word. Cloud and cloudy behave quite differently and fit into different slots in the sentence. Other examples of derivational morphemes are -ness as in happi_ness_, -ish as in green_ish_, and -ment as in establish_ment_.

In most cases, it is easy to tell the difference between inflection and derivation. Above all, inflectional endings do not alter the syntactic behaviour of an item in any major way. The word still fits into the same ‘slot’ in the sentence. Derivational endings create entirely []()new words. In addition, inflectional endings can be added on to derivational ones, but not vice-versa. That is, we find words such as establish-ment-s, but not \*establish-s-ment.

English has relatively few inflectional morphemes. These are on the whole easy to identify, though they sometimes present problems of analysis, as discussed below.

[]() Allomorphs

Sometimes a morpheme has only one phonological form. But frequently it has a number of variants known as allomorphs.

Allomorphs may vary considerably. Totally dissimilar forms may be allomorphs of the same morpheme. Cats, dogs, horses, sheep, oxen, geese all contain the English plural morpheme.

An allomorph is said to be phonologically conditioned when its form is dependent on the adjacent phonemes. An allomorph is said to be lexically conditioned when its form seems to be a purely accidental one, linked to a particular vocabulary item.

The English plural morpheme provides excellent examples of both phonologically and lexically conditioned allomorphs. Let us look at some of these.

[]() PHONOLOGICAL CONDITIONING

The study of the different phonemic shapes of allomorphs is known as morphophonology – sometimes abbreviated to morphonology.

/-z/ /-s/ /-iz/ are all phonologically conditioned allomorphs of the English plural morpheme. That is, each allomorph occurs in a predictable set of environments.

/-z/ occurs after most voiced phonemes []()as in dogs, lambs, bees. (A voiced phoneme is one in which the vocal cords vibrate, as in /b/, /d/, /g/, /v/, and vowels.)

/-s/ occurs after most voiceless phonemes, as in cats, giraffes, skunks. (A voiceless phoneme is one in which the vocal cords do not vibrate.)

/-iz/ occurs after sibilants (hissing and hushing sounds), as in horses, cheeses, dishes.

If we take /-z/ as basic, then we can say first, that /-z/ turns into /-iz/ after sibilants (), and second, into /-s/ after voiceless sounds ():

ø →

/ [+ sibilant] —— z. e.g. /h

sz/ → /h

[]()

Figure 6.5.

z → s/\[-voice] — —. e.g. /kætz/ →/kæts/

[]()

Figure 6.6.

Note that these ‘rules’ must be applied in the order given above. If the order was reversed, we would get forms such as *[d

∫s] instead of the correct [d

∫

z] for the plural of

dish

[]() LEXICAL CONDITIONING

Words such as oxen, sheep, geese present a problem. Although they function as plurals in the same way as cats, dogs, they are not marked as plurals in the same way. Such lexically conditioned plurals do not follow any specific []()rule. Each one has to be learnt separately.

Words such as oxen, sheep, geese can be identified as syntactically equivalent to the cats and dogs type of plural because they fit into the same ‘slot’ in a sentence ():

[]()

Figure 6.7.

Oxen, sheep and geese each contain two morphemes:

[]() ox + plural
[]()sheep + plural
[]()goose + plural

But only oxen is easily divisible into two:

[]()ox + /- <Image src="../images/f601-01.jpg" alt="Image" />n/ (-en)

Sheep can be divided into two if a zero suffix is assumed. A ‘zero suffix’ is a convenient linguistic fiction which is sometimes used in cases of this type. It is normally written /ø/:

[]()sheep + /ø/.

There is no obvious way to analyse

geese

. At one time, linguists suggested that the plural vowel /i

/ in /gi

s/ (

geese

) which replaces the /u

/ in /gu

s/ (

goose

) should be regarded as a special type of allomorph called a

replacive

. And they analysed the plural as:

[]()/gu <Image src="../images/icon09.jpg" alt="Image" />s/ + /i <Image src="../images/icon09.jpg" alt="Image" />/ ← (/u <Image src="../images/icon09.jpg" alt="Image" />/).

Here the formula /i

/ ← (/u

/) means ‘/i

/ replaces /u

/’.

But this is rather a strained explanation.

[]()

These days, most linguists simply accept that the form /gi

s/ (

geese

) represents two morphemes:

[]()goose + plural

and that these two cannot be separated. And a similar explanation is required for forms such as went, took, which represent:

[]() go + past tense
[]()take + past tense.

[]() Word classes

In every language, there are a limited number of types of lexical item. These different kinds of word are traditionally known as ‘parts of speech’, though in linguistic terminology the label word class is more common. Word classes are conventionally given labels, such as noun, verb, adjective.

Words are classified into word classes partly on account of their syntactic behaviour, partly on the basis of their morphological form. That is, words from the same word class are likely to fit into the same slot in a sentence, and to be inflected in similar ways. For example, the word class traditionally known as ‘verb’ can be recognized as a verb partly because it occurs after nouns (or phrases containing a noun), and partly because most verbs have an inflectional ending -ed to indicate the past:

[]()Arabella detested snails.
[]()Marianna smiled.

Careful analysis is needed, because in some cases, items which superficially appear to fit into the same slot in a sentence can turn out to be rather different in character. Consider the sentences:

[]()Charlie ate caviare.
[]()Charlie ate well.

At first sight, we might wrongly []()assume that caviare and well belong to the same word class. But a less superficial analysis reveals that they behave somewhat differently overall. If we tried altering the sentences around, we could say:

[]()Caviare was eaten by Charlie.
[]()What Charlie ate was caviare.

But we could not form the equivalent sentences with well:

[]()\*Well was eaten by Charlie.
[]()\*What Charlie ate was well.

These dissimilarities indicate that caviare and well are syntactically different, and belong to different word classes.

[]()Insight

Word classes (parts of speech) are recognized partly by their syntactic behaviour and partly by their morphological form.

It is not always easy to tell how many word classes a language contains. Many traditional textbooks claim that English has eight ‘parts of speech’. But this claim turns out to be based largely on old Latin grammars which were in turn translated from ancient Greek grammars, which mostly divided Greek words into eight word classes. If we look more closely, we find several discrepancies. For example, nouns and pronouns are traditionally classified as separate parts of speech, yet they have a large number of similarities:

[]()Max laughed.
[]()He laughed.

In fact, nouns and pronouns are more alike than the different types of word which are traditionally labelled adverbs. Words such as quickly and very are both usually classified as adverbs, []()but they behave quite differently:

[]()He ran quickly.
[]()\*He ran very.

The number of word classes varies from language to language. Some word classes, such as noun and verb, may be universal. But others vary. Nouns, adjectives and verbs are on a continuum. At one end are nouns, words which maintain their identity over time, such as tree, cat, river. At the other end are verbs, words which signify rapid change, as in walk, kick, push. In the middle come properties, such as large, beautiful, old. In English, these form a separate word class, that of adjectives. But this is not inevitable. Some languages treat them as a type of verb, so-called stative verbs, ones which denote a state. Where English says:

[]()Petronella is happy

a language such as Chinese might say, as it were:

[]()Petronella happies

using a verb instead of an adjective. English also sometimes flips between verbs and adjectives. Compare the archaic he ails (stative verb) with the modern day he is ill (adjective).

[]() Major word classes

English is sometimes considered to have four major word classes: noun (N), adjective (A), verb (V), preposition (P) ().

[]()Figure 6.8.

Of these four major classes, nouns, verbs and []()prepositions behave fairly differently from one another, though adjectives are somewhat strange, in that they have some noun-like qualities, and some verb-like ones. In Blessed are the _brave_, brave seems to have become a noun. And in Mavis is _asleep_, asleep seems fairly verb-like, since it fits into the same slot as sleeping in a sentence such as Mavis is sleeping.

It has been suggested that we should describe these four word classes in a manner parallel to the distinctive feature descriptions used for sounds, which can show shared similarities.


Noun	[+ N, − V]
Verb	[− N, + V]
Adjective	[+ N, + V]
Preposition	[− N, − V]

This seems to be a useful and economical way of capturing the similarities and differences between the major word classes.

The major word classes are known as lexical categories. Lexical categories contain content words, those with intrinsic meaning. They contrast with functional categories, which include ‘little words’ whose meaning is often difficult to specify, such as the, a, which are determiners (D), or the complementizer that in I know _that_ Paul is ill, often abbreviated to COMP or C. These function words are important for gluing pieces of sentences together into longer syntactic patterns.

This chapter has discussed words and morphemes. These fit into larger recurring patterns, which will be the topic of the next chapter.

[]()[]()THINGS TO REMEMBER

[]()The word ‘word’ is difficult to define because there are different kinds of word.
[]()Word tokens (total number of words) must not be confused with word types (number of different words).
[]()Lexical items (dictionary entries) need to be distinguished from syntactic words (words with different behaviour patterns) and from phonological words (different word shapes).
[]()Syntactic words are typically uninterruptible and mobile.
[]()Words have different roles within the sentence, and need to be assigned to word classes (parts of speech).
[]()Words can be assigned to word classes partly by assessing their syntactic behaviour (their slot in the sentence), and partly by looking at their phonological form (shape).
[]()In English, and many other languages, the major word classes are nouns, verbs, adjectives and prepositions.
[]()Morphemes are the smallest syntactic unit, though they vary in size, from single sounds to whole words.
[]()Variants of morphemes are known as allomorphs.
[]()Allomorphs are said to be phonologically conditioned when their form is dependent on the adjacent morphemes, and lexically conditioned when their form is linked to a particular lexical item.