Tema 4 – The concept of lexeme

The concept of lexeme and criteria for its identification. Polysemy and homonymy.

Session based on Murphy (pgs. 5-14 and 83-98)

WHAT IS A WORD?

DEFINING WORD

While the loose definition of lexical semantics is ‘the study of word meaning,’ the focus of this textbook is more precisely lexeme meaning. Never- theless, as a matter of fact, most of the cases discussed in this book (and most of what is traditionally considered to be lexical semantics) involve lexemes that are words rather than bound morphemes or multi-word lexemes (such as idioms). So, it is worthwhile to say a word about what words are. I have already hinted at one problem with using the term word as a technical term: word is ambiguous in that it could mean ‘lexeme’ as in (8) or ‘lexical unit’ as in (9):

8. Colour and color are two spellings of the same word.
9. There are eight words in the sentence Jo’s pet ferret hates Jo’s mother’s pet ferret.

But aside from the lexeme/lexical unit ambiguity of word, it can be tricky to determine which lexemes count as words in a particular language. Part of the reason for this is that the notion word can be defined in a number of ways, including:

• Orthographically – based on the written form
• Semantically – based on the meaning
• Phonologically – based on the word’s pronunciation
• Grammatically – based on positions in phrases

Ask random English speakers what a word is, and you are likely to hear definitions based on orthographic or semantic criteria. Probably the most common belief is that words are bits of language that have a space on either side of them in print. The most obvious problem with this orthographic definition is that it employs circular logic. The reason that we put spaces in writing is to signal the beginnings and ends of words; one must already know which bits of language are words before one can know where to put the spaces. For instance, we know to put a space between cat and nip in (10), but to run the two together in (11).

10. Tom saw the cat nip the baby.
11. Fluffy loves catnip.

Notice too that the spaces in writing do not necessarily represent spaces in speech. Cat+nip can be pronounced in just about the same way in sentences (10) and (11). Something other than spaces must be telling us which bits are words in these sentences. Another problem with the ‘space’ definition is that some orthographic systems do not put spaces between words. In writing the southeast Asian language Lao, for instance, spaces are only left at the ends of sentences. Nevertheless Lao speakers can identify the words in those sentences. If we want to talk about the bits of language with spaces on either side, we can call them orthographic words, but the orthographic definition doesn’t get to the essence of wordhood.

Once you give up on the orthographic definition of word, you might try the semantic definition, which states that words represent single, complete concepts. In that case, you could tell that cat+nip is two words in (10) because two meanings or ideas are represented, whereas in (11) catnip refers to just one thing. But this definition doesn’t work for a couple of reasons. First of all, how can one tell what a complete meaning is? For instance, one might think that policeman involves two concepts – being police and being a man – or that it is one concept involving the entire state of being a male police officer. Similarly, the same concept can be described by various numbers of words. For example, in comparing languages it is often the case that the same meaning is variously realized in different numbers of orthographic words, such as English post office versus Swedish postkontor versus French bureau de poste. The “semantic” definition of word thus leads us either to the illogical conclusion that these expressions meaning ‘post office’ must involve different numbers of meanings, since they involve different numbers of words, or to the conclusion that some languages have more spaces in their words than others. While the second conclusion might be true, we can do better than this definition.

A phonological word is a unit of language that is subject to the language’s particular word-based phonotactic constraints (i.e. which sounds can go next to which other ones) and phonological processes (e.g. variations in pronunciation of, say, the first consonant in a word or the main vowel in a word). For example, in English no word begins with the phoneme /t/ followed by /n/, so the first time you heard the word catnip, you didn’t have to consider the possibility that what you heard was two words, ca and tnip. A more general signal of phonological wordhood in English is that a phonological word has no more than one main stress. For long words, like homeostasis, there may be more than one stressed syllable, but there is one that is more stressed than the others: homeoSTAsis. So, we can tell that cat nip in (10) is two words because the two syllables are stressed evenly, but catnip in (11) is a candidate for wordhood because it has main stress just on cat. Phonological tests of wordhood do not, however, always coincide with our semantic and grammatical intuitions and our orthographic practice concerning word boundaries. For instance, because function words are usually not stressed in English, there is only one phonological word in the two- syllable form a cat – even though it is two orthographic words and arguably two semantic units. Phonological tests for wordhood are also entirely language- specific – for example, the stress test in English would not work for French, Chinese, or Norwegian. Thus the concept of phonological word is not terribly useful as a means of defining word in general, across languages.

Even though we have so far discussed words as form–meaning associations, the notion of wordhood can be considered to have more to do with grammar than with written/spoken form or meaning. A grammatical word is an expression that is uninterruptible,1 ‘movable,’ and that has a part of speech that can be identified on the basis of its morphological inflections and its distribution in phrases. To return to the cat+nip example, catnip is a noun, but cat nip in (10) has no single part of speech. It is not even a phrasal constituent (i.e. a complete part unto itself) in that sentence, as can be determined by various tests of constituency. For example, cat nip cannot be replaced by a single word that refers to the same thing in (10), whereas the noun catnip in (11) can be replaced by the pronoun it. Grammatical words cannot be split up within a sentence. Thus if we make the sentences in (10) and (11) into cleft sentences by moving a noun into an it is/was clause, we have to move cat separately from nip in the first case, but must move catnip as one unit, as (12) and (13) show.

12. • a. It was the cat that Tom saw nip the baby.
    • b. ∗It was the cat nip that Tom saw the baby.
13. • a. ∗It is cat that Felix loves nip.
    • b. It is catnip that Felix loves.

Furthermore, cat nip can be interrupted but catnip cannot, as shown in (14) and (15), respectively.

(14)   Don’t let the cat unintentionally nip the baby.

(15)   ∗Felix loves catwonderfulnip.

Using the grammatical or phonological criteria for wordhood, some expres- sions usually written with spaces within them count as words. These are com- pound words whose orthographical representations have not caught up with their lexical status. For example, we can tell that ice cream is a word because it has main stress on ice, it can be replaced in a sentence by it, we cannot put other words or morphemes in the middle of it (∗_ice minty cream_), and so forth.

The notion of ‘word’ is ambiguous, it can be interpreted as:

• Word as ‘word form’ or lexical unit

e.g., _There are seven words in this sentence._🡪Blank space between words.

• Word as ‘lexeme’

e.g., Colour and color are two spellings of the same word.

Different interpretation or manifestations of the concept ‘word’🡪a ‘supra-word’ and you have 2 versions corresponding to the American and the British forms.

The headword is the lexeme, on the basis on which we can construct lexical relations.

Words can be identified under the following criteria:

• Orthography
• Semantics
• Phonology
• Grammar

The problem is that these are often contradictory!

Idioms / phrasal verbs: ‘kick the bucket’ / ‘pass away’

Their elements form a semantic unit but are written separated orthographically.

OVERVIEW

In order to set the stage for exploring lexical semantics, this chapter defines basic terms and ideas in the field, particularly the notions of lexicon, lexeme, and word. It then describes and evaluates four methods for investi- gating the lexicon: dictionaries, corpora, intuition, and experimentation.

WHAT IS A LEXICON?

Word meanings are notoriously difficult to pin down – and this is well demonstrated in defining the basic terminology of lexical semantics. Semantics is the study of linguistic meaning, but it will take the next three chapters to discuss what meaning might mean in any particular theory of semantics. The lexical in lexical semantics refers to the lexicon, a collection of meaningful linguistic expressions from which more complex linguistic expressions are built. Such lexical expressions are often, but not always, words, and so lexical semantics is often loosely defined as ‘the study of word meaning,’ although the word word, as we shall see, is not the most straightforward term to use.

While many of the details of the structure and content of the lexicon are discussed in detail in later chapters, some general discussion of what the lexicon is and what it contains must come first. A lexicon is a collection of information about words and similar linguistic expressions in a language. But which information? Which expressions? What kind of collection? Whose collection? We’ll cover these issues in the following subsections, but first we must acknowledge the polysemy (the state of having multiple meanings) of the word lexicon. Lexicon can refer to:

• a dictionary, especially a dictionary of a classical language; or
• the vocabulary of a language (also known as lexis); or
• a particular language user’s knowledge of her/his own vocabulary.

For our purposes, we can disregard the first meaning and leave the study of such dictionaries to students of classical languages. The last two definitions are both relevant to the study of lexical semantics. In speaking of the lexicon, different scholars and theories assume one or the other or the interrelation of both, as the next subsection discusses.

WHERE IS THE LEXICON?

Some traditional approaches to the lexicon generally make claims about the vocabulary of a language, its lexis. Taking this perspective on vocabu- lary, the lexicon is “out there” in the language community – it is the collection of anything and everything that is used as a word or a set expression by the language community. Other linguistic perspectives, including those discussed in this book, focus on vocabulary “in here” – in the mind of a language user. The term mental lexicon is used in order to distinguish this more psychological and individualistic meaning of lexicon.

Clearly though, we have to take into account the fact that the “out there” and “in here” lexicons are interrelated; in order to communicate with each other, speakers of a language must aim to have reasonably similar ways of using and understanding the words they know – otherwise, if you said banana, I’d have no reason to believe that you didn’t mean ‘robot’ or ‘hallelujah.’ The lexicon of the language “out there” in our culture is the lexicon that we, as individuals, aim to acquire “in here” and use. This is not the same as saying that the lexicon of a language is a union of all the lexicons of all the language’s speakers. When linguists study a language’s lexicon, they tend to idealize or standardize it. For instance, say there’s an English speaker somewhere who, ever since being hit on the head with a mango, mistakenly uses the word goat to mean ‘pencil.’ Just because there’s someone who uses the language in this way does not mean that this fact about the use of the language needs to be accounted for in a model of the English lexicon – his use is clearly a mistake. So, in order to study the lexicon of a language, one needs to have a sense of what does and does not count as part of that language. Sometimes these decisions are somewhat arbitrary, but they are not simply decisions made on the basis of what is “correct” English in some school-teacherish (that is, prescriptive) sense. Non-standard words and uses of words are also part of the language that we want to explain, and we can pursue interesting questions by looking at them. For example, some people use wicked as slang for ‘especially good,’ which might lead us to ask: how is it that a word’s meaning might change so much that it is practically the opposite of what it originally meant?

Similarly, although mental lexicons exist in individual speakers’ minds, in studying the mental lexicon we do not necessarily want to investigate any one particular speaker’s lexicon (otherwise, we would have another few billion lex- icons to investigate after we finish the first one). Instead, the focus is usually on an imagined “ideal” speaker of the language, which again brings us back to the notion of a language’s lexicon. For an “ideal” mental lexicon, we imagine that a speaker has at her disposal the knowledge necessary to use the language’s lexis.

Most current approaches to the lexicon attempt to find a balance between the “out there” and the “in here.” While particular models of lexical meaning will be evaluated in this book on the basis of their psychological plausibility, part of what makes a theory psychologically plausible is whether it is consistent with (and engages with) the social facts of language acquisition and use. My continued use of the ambiguous term lexicon is an acknowledgment of the dual nature of the object of our study, but the terms mental lexicon and lexis are used wherever disambiguation is needed.

WHAT’S IN A LEXICON?

Having discussed the where of the lexicon, we move on to the what. The things that one knows when one knows a language can be divided into two categories: the lexical and the grammatical. A grammar is a system of rules or regularities in a language, and a lexicon is (at the very least) a collection of linguistic knowledge that cannot be captured by rules. The grammar accounts for linguistic issues like word order and regular morphological and phonological processes. For instance, our grammar tells us the difference between the sentences Bears trouble bees and Bees trouble bears, and that this is the same kind of difference as the difference between The kangaroo ate a flower and A flower ate the kangaroo. What the grammar cannot tell us is what bear and bee and trouble bring to the sentence. At some point in our acquisition of English, we learned that the sound [bi] and the spelling b-e-e are paired with a particular set of linguistic and semantic properties – like being a noun and denoting a kind of insect. The lexicon is the collection of those associations between pronunciations, meanings, and grammatical properties that had to be learned rather than produced by grammatical rules.

The lexicon is organized into lexical entries, much as a dictionary is organized into entries that pull together all the information on a headword (the word, typically in boldface type, at the start of a dictionary entry). Each of these lexical entries collects the appropriate information about a particular linguistic expression, called a lexeme. (Later we look at why it is more precise to use the term lexeme rather than word in the study of lexical meaning.) In the remainder of this subsection, we consider which expressions are lexemes and belong in the lexicon, then in §1.2.3 we’ll consider what information goes in a lexeme’s lexical entry. Section 1.2.4 goes into more detail on the notion of a lexeme as an abstract representation and the relationship between a lexeme (and its lexical entry) and actual uses of an expression. Let’s start with this description of lexeme:

A linguistic form (i.e., a bit of speech and/or writing) represents a lexeme if that form is conventionally associated with a non-compositional meaning.

In order to make more sense of this, let’s look more closely at the concepts of conventionality and (non-)compositionality, in turn.

That’s why in lexicology we prefer the notion of lexeme:

Murphy (page 6) defines lexeme as:

‘A linguistic form (i.e., a bit of speech and/or writing) that is conventionally associated with a non-compositional meaning’.

What does she mean by that?

• Conventional item: association of form and meaning, known to the entire linguistic community.

Lexemes, and the information about them in the lexicon, are conventional – that is, these form–meaning pairings are common knowledge among the speakers of the language, and we have had to learn these particular associations of form and meaning from other members of the language community. Compare, for example, a scream with a word. If you heard someone screaming, you would not know (without further information) what they were making noise about nor why they were making noise – after all, we scream with surprise, with delight, or with horror. But if the person yelled Spider! or Fire! or Jump!, you’d know what they were yelling about (but perhaps still not why they were yelling about it) because those words are used by members of our speech community to signal particular things.

• Non-compositional meaning: the meaning of a lexeme is not up the meaning of its parts.

Lexemes are non-compositional – that is, the meanings of these linguistic forms are not built out of (or predictable from) the meanings of their parts. For example, the word cat is non-compositional because its meaning is not evident from the sounds or letters that constitute the word. It’s not that the sound /t/ represents the tail of the cat or that the vowel /æ/ tells us that a cat has fur. The word cat and its meaning thus constitute an arbitrary pairing of form and meaning.

The meaning of black cat, however, is deducible if you know:

(a)  the word classes (adjective, noun) and meanings of the words black and cat

(b)  what it means in English to put an adjective in front of a noun.

Thus the meaning of the clause black cat is compositional; its meaning is built from the meanings of its parts. This means that black cat does not need to be included in the lexicon, but its non-compositional parts (black and cat) do.

So e.g., ‘houses’ is a word, but is not a lexeme.

Moreover, apart from simple words, this definition of lexeme includes:

• Bound morphemes (-er in driver)
• Complex words with unpredictable meaning (greenhouse)
• Set phrases and idioms (give up; pull sb’s leg)

Note that the Mental Lexicon may include compositional forms if they are frequently used.

So, a lexeme is identified on the basis of its meaning. But, many word-forms are ambiguous.

Lexemes and morphological complexity

While lexical semantics is often loosely defined as ‘the study of word meaning,’ the use of word in this definition is misleading, since lexical semantics is more accurately described as the study of lexeme meaning, and (a) not all words are lexemes and (b) not all lexemes are words.

Here, we need to pause and define a little terminology from morphology, the study of word structure. Morphemes are the smallest meaningful units of language, so a word is one morpheme if it is not built out of smaller meaningful parts. So, for example, language is a single morpheme: if we try to divide it into smaller parts (like l and anguage or langu and age), we don’t get meaningful parts of English. (And although we can see the word age in language, that is just accidental. Age is not a morpheme within language as ‘age’ is not part of the meaning of language.) On the other hand, sentimentally is composed of three morphemes: the noun sentiment and the suffixes –al (which is added to nouns to make adjectives and could be glossed as ‘marked by’) and –ly (which turns an adjective into an adverb: ‘in a certain manner’). This results in a complex word that means ‘in a manner that is marked by sentiment.’ Sentimentally is thus compositionally formed from sentiment plus two suffixes, and since it is compositional, in that the meaning comes straightforwardly from the combination of the parts, it does not count as a lexeme on our definition. However, sentiment and the suffixes -al and –ly act like lexemes, in that they have conventional meanings that cannot be derived from the meanings of their parts.

Suffixes, prefixes, and other bits of language that always attach to other mor- phemes are called bound morphemes: they must be bound to another linguistic form in order to be used. Words, on the other hand, are free morphemes, in that they can be used without being attached to anything else. Just as words can be arranged and rearranged to make new phrases with compositional meanings, morphemes can be arranged and rearranged to make new words with composi- tional meanings. The first time you come across a new morphologically complex word, like unputdownable or pseudoscientifically, you will be able to understand the word if you can understand its parts.

So far, we have seen that morphemes are lexemes in that they have conven- tional, non-compositional meanings. But more complex expressions may also be lexemes. For example, while greenhouse is a compound noun derived from two free morphemes, its meaning is not deducible from its parts, since a greenhouse is not actually green and it is debatable whether it is a house. A series of words can also be a single lexeme, as demonstrated by the bold expressions in the following examples:

(1) a. Look up that word in the dictionary!

b. Look up! There’s a dictionary in that tree.

(2) a. Her beloved pet is now pushing up daisies.

b. Fido was the apple of her eye.

In (1a), the phrasal verb look up is a single lexeme meaning ‘consult,’ as opposed to (1b), in which look and up each contributes its own meaning. Idioms like pushing up daisies (‘dead’) and apple of one’s eye (‘one’s beloved’) in (2) are also lexical expressions, since they are non-compositional, and thus have to be learned and mentally stored by a language speaker.

In summary, the term lexeme includes:

• simple words (free morphemes) that cannot be broken down into smaller meaningful parts, such as cup, Cairo, and contribute;
• bound morphemes, like un- as in unhappy and -ism as in racism;
• morphologically complex words whose meaning is not predictable from the meanings of the parts, including compounds like greenhouse (‘a glass building for growing plants in’) and needlepoint (‘a method of embroidery onto canvas’);
• set phrases whose meaning is not compositional, such as phrasal verbs like throw up (‘vomit’) and give up (‘quit’) and idioms like having the world on one’s shoulders and fly off the handle.

Are all lexemes non-compositional?

While linguists agree that a lexicon contains conventional, non-compositional form–meaning pairings, opinions differ as to whether one’s mental lexicon also contains some compositional expressions. In other words, just because we could understand the meaning of a complex expression from the meanings of its parts, doesn’t mean that we necessarily always go through the process of composing that complex expression every time we use it. In cases like those in (3), the expressions are so well-worn that they seem like idioms, in spite of having conventional meanings.

(3) a. happiness (happy + ness)

b. How are you?

c. It’s not whether you win or lose, it’s how you play the game.

Should the items in (3) be considered to be lexemes? One argument for including compositional expressions in the lexicon is that it is advantageous for a language user to have ready access to commonly used expressions. So, in a particular individual’s mind it might be convenient if complex and frequent but predictable expressions, like How are you? or I love you, were stored in the mental lex- icon. Having such expressions available both through grammatical assembly and through ready-made lexemes is redundant – and thus it appears to require needless extra effort to learn and store the expression in the mental lexicon. But while non-redundant models of language are more streamlined and elegant, they are not necessarily more realistic. After all, representing the same expres- sion (or access to the same expression) in different ways or different places in one’s mind could make language production and comprehension processes less likely to break down, since if one route to the expression fails, another might succeed.

Another argument for including compositional expressions in the lexicon is that some of them are particularly conventionalized – that is to say, people some- times rely on “ready-made” compositional expressions instead of composing new ones. The extremes of such conventionalization are seen in compositional cliche ́s like in this day and age, cry like a baby, or the example in (3c), but conventionalization of compositional expressions can be subtler too, as studies of collocations (particularly frequent word combinations) have shown. A case in point is example (4), which shows how the meaning ‘without milk added’ is indicated by different modifiers, depending on what is being modified.

(4) black coffee

black tea – however, this can also mean non-herbal tea, so people often prefer to say tea without milk or tea with no milk

plain chocolate, dark chocolate

dry cereal

Logically speaking, there is no particular reason why milkless coffee is black and not plain, nor why dark chocolate is not called black chocolate (after all, it is about the same color as black coffee). As English speakers, we’ve just learned that these adjectives and nouns are best used in some combinations and not others. Similarly, some nouns go with particular verbs. For example, in English one asks (or poses) a question, but one poses a riddle (one does not ask it) and one makes a query (one neither asks nor poses it). Such facts lead some to argue that we should be thinking of the lexicon as including larger and looser combinations than just single words and set idiomatic phrases, or that we should see the lexicon as a network of syntactic (grammatical) and semantic relations among lexemes. While these are very interesting issues, the focus in this book is particularly on the non-compositional forms (mostly words) that must be in the lexicon, as the emphasis on non-compositional meanings is what makes lexical semantics distinct from other kinds of semantics. Information about collocations is still of interest to us, though, as it may be used to determine differences in particular words’ meanings.

WHAT IS IN LEXICAL ENTRY?

Within the lexicon, the collection of information pertaining to a lex- eme is said to be its lexical entry, analogous to a dictionary entry. The information that must be stored about a lexeme is precisely that which is unpredictable, or arbitrary. At the very least, this means that we need to know the basics of the lexeme’s form and what meaning(s) it is associated with. When speaking of a word’s form we usually mean its pronunciation, but if we know it as a written word, then its spelling is part of its form, and if it is a part of a sign language, then its “pronunciation” is gestural rather than vocal. We only need to store in the lexicon the details of the lexeme’s form that are not predictable; so, for example, we do not need to store the facts that cat is made possessive by adding ’s or that the c in cat is usually pronounced with a slight puff of air – these facts are predictable by rules in the language’s grammar and phonology. As we shall see in the coming chapters, theories differ in what information about meaning is (or is not) included in the lexicon. For many modern theories, meaning is not in the lexicon, but is a part of general, conceptual knowledge. That is to say, the linguistic form is represented in the lexicon, but instead of its definition being in the lexicon as well, the lexical entry “points” to a range of concepts conventionally associated with that word. Other theories view the lexicon more like a dictionary, which provides basic definitions for words.

What other information is included in a lexical entry, again, differs from theory to theory. Most would say that a lexical entry includes some grammatical information about the word, for example its word class (or part of speech: noun, verb, etc.), and the grammatical requirements it places on the phrases it occurs in. For instance, the lexical entry for dine includes the information that it is a verb and that it is intransitive in that it takes no direct object, as shown in (5). Devour, on the other hand, is recorded as a verb that is transitive, so that it is grammatical with a direct object, but not without one, as shown in (6). Asterisks (∗) signal ungrammaticality.

(5) a. We dined.

b. ∗We dined a big bowl of pasta.

(6) a. ∗We devoured.

b. We devoured a big bowl of pasta.

We come back to these issues in chapter 7, where we consider whether the meanings of words might determine their word classes, or whether they are entirely arbitrary and need full specification in the lexicon.

We may also need information in the lexicon about which words go with which other words – for instance, the fact that stark collocates with naked but not with nude or the fact that the conventional antonym of alive is dead and not expired. We come back to some of these issues below and in chapter 6.

THE ABSTRACT NATURE OF LEXEMES

The last thing to say in this preliminary tour of the lexicon is that a lexeme is not the same as a word in real language use. Lexemes are, essentially, abstractions of actual words that occur in real language use. This is analogous to the case of phonemes in the study of phonology. A phoneme is an abstract representation of a linguistic sound, but the phone, which is what we actually say when we put that phoneme to use, has been subject to particular linguistic and physical processes and constraints. To take a psycholinguistic view, a phoneme is a bit of language in the mind, but a phone is a bit of language on one’s tongue or in one’s ear. So, the English phoneme /l/ is an abstract mental entity that can be realized in speech variously as, say, a “clear” [l] in land, or a “dark” [∼l] in calm. The phoneme is, so to speak, the potential for those two phones.

Similarly, when we use a word in a sentence, it is not the lexeme in the sentence, but a particular instantiation (i.e. instance of use) of that lexeme. Those instantiations are called lexical units. Take, for example, the lexeme cup. It is associated with a range of meanings, so that we can use it to refer to:

(a)  any drinking vessel, or

(b)  a ceramic drinking vessel with a handle whose height is not out of proportion to its width, or

(c) a hand shaped so that there is a depression in its upward facing palm, or

(d) the part of a brassiere that covers a breast.

But in a particular use, as in sentence (7), the lexical unit cup is assigned just one of those meanings – in this case, meaning (b).

(7) I prefer my coffee in cups, not mugs.

In (7), cup also occurs with a particular orthographic form and it is morphologi- cally marked as a plural. So, in this case we can say that the lexical unit cups in (7) is an instantiation of the lexeme cup, which has the plural form, the standard spelling and the (b) meaning associated with that lexeme.

The lexical entry provides the information needed to use the lexeme as a lexical unit; that is, it sets the parameters of that lexeme’s potential. So, as speakers or writers, we select particular lexemes to be realized as lexical units in our utterances because the parameters set in the lexical entry are consistent with what we want to do with that word – for example, for it to have a particular meaning and fill the verb slot in a sentence. As listeners or readers we recognize the lexical unit as being an instantiation of a particular lexeme and interpret the meaning of the lexeme in ways that are consistent with the lexical entry and the context in which the lexical unit occurs.

OVERVIEW

In this chapter we turn our attention to an issue we set aside earlier: the fact that in different contexts the same word form can be understood to have different semantic interpretations. For instance book can mean ‘a collection of pages that are bound in a rigid cover’ (hand me that book), ‘the information contained in a collection of pages that are bound in a rigid cover’ (that book is depressing) or ‘reserve’ (We want to book a table for four), and in any of its senses book can point to lots of different kinds of things. For example, the ‘collection of pages’ sense could refer to this book or to a phone book, and the ‘reserve’ sense of book can be used for booking a restaurant by phone or booking a hotel on-line. This chapter reviews three ways in which a single word form can have more than one interpretation: homonymy, polysemy, and vagueness. We’ll refer to these three phenomena collectively as cases of meaning variation. After defining the types of meaning variation, we consider whether or not they denote completely distinct phenomena. Next we focus on polysemy (when a single word has several senses) and discuss how words come to have multiple senses and how those senses relate to one another. Finally, we consider how polysemy is treated in two theoretical approaches: the componential approach of Pustejovsky and the frame-based approach of Fillmore.

POLYSEME, HONONYMY, VAGUENESS

The various interpretations of a word can be related to each other in different ways, and can come about in different ways. So, under the gen- eral heading of “meaning variation” we find three distinct phenomena, starting with:

1. If the word has one sense that is general enough that it can be applied to many different things, then the word has a vague, or indeterminate, sense.

In that case, the word can denote a variety of different things, but it gets to those things by means of a general “one sense fits all” meaning. For example, clock means ‘a device for measuring hours’ and therefore it can refer to a lot of different types and instances of clocks – digital clocks, analog clocks, alarm clocks, cuckoo clocks, and so forth.

The other two types of meaning variation arise when two uses of a word represent different senses. The state of having more than one possible sense is called ambiguity, and there are two types of lexical ambiguity:

1. If two form–meaning pairings involve two different lexemes that just happen to have the same spoken and/or written form, then it is a case of homonymy – that is, there are two lexemes that are each other’s homonym. For instance, the noun kind meaning ‘type’ and the adjective kind meaning ‘considerate’ are two different words that just happen to have the same spelling and pronunciation. If we want to talk just about the spoken form of the language, we can refer to homophones, which have the same pronunciation but not necessarily the same spelling, such as raze and raise. If we’re just looking at the written language, there are homographs, which are spelt the same but may or may not be pronounced differently – like the musical instrument bass /beis/ and the fish bass /bæs/.
2. If a single lexeme has two distinguishable senses associated with it, then we say that it is a polyseme or it is polysemous. The ‘bound pages’ and ‘information’ meanings of book are related to one another, so we would not want to conclude that we have two completely different words when we use the ‘text’ and ‘tome’ senses of book. So, we conclude that book is a polyseme with ‘text’ and ‘tome’ senses.

But how can you know if a word has one sense or two or five? And how can you tell if a particular word form represents one lexeme or two lexemes that just happen to look and sound exactly like each other? Let’s start exploring these issues by focusing on the difference between vagueness and ambiguity before differentiating polysemy and homonymy.

Considering the following word-form:

Port: 1. harbour. 2. kind of wine.

There are two possible analyses:

1. Two distinct lexemes (Homonymy)
2. One lexeme with two senses (Polysemy)

How do we decide?

• This is an example of homonymy: whenever a word has a meaning, immediately it becomes a lexeme.
• THIS IS THE THEORY OF THE MENTAL LEXICON: Each individual meaning corresponds to a lexeme.

VAGUENESS VS. AMBIGUITY

If an expression is vague then its sense is imprecise, but if it is ambiguous, it has at least two separate senses. We’ll look at three ambiguity tests – definition, contrast, and zeugma – and use them to show that friend is vague with respect to the sex of the friend, but it is not ambiguous between separate ‘male friend’ and ‘female friend’ senses.

In (1a), friend refers to someone male, and in (1b), it refers to someone female, so you might initially hypothesize that friend has two senses: ‘a male person with whom one has a relationship of mutual affection’ and ‘a female person with whom one has a relationship of mutual affection.’

(1) a. Ben is my friend. He’s a fine fellow.

b. Georgia is my friend. She’s a wonderful woman.

Well, we should identify discrete senses with relevant test:

• Definition test: However, the first indication that friend does not have two separate senses is the ease with which we can make a single definition that covers both of the cases in (1): ‘a person with whom one has a relationship of mutual affection.’ So, according to the definition test, friend is vague, not ambiguous.
• Try to make a general definition for the word in question that covers both senses.

The first indication that friend does not have two separate senses (female and male friend) is the ease with which we can make a single definition that covers both of the cases in (1): a person with whom one has a relationship of mutual affection.

So, according to the definition test, friend is vague, not ambiguous.

• Contrast test:

If friend is vague, then in some contexts it may refer to a person who happens to be female and in other contexts to a person who happens to be male. If this were not the case, that is, if friend had two gender-specific senses instead of one general sense, it would make sense to say something like (2) in order to mean (3):

(2)  # I have both friends and friends.

(3)  I have both male friends and female friends.

Compare (2) to (4), in which one can easily understand bats to mean two different things: a type of animal and a type of sporting equipment.

(4) I collect things that begin with B, so I have both bats and bats.

Because the two instances of bats have the same form, you cannot tell whether the first or the second refers to the animal (unless you can read the mind of the person who said (4)), but you can tell that there are two distinct senses here. So because the two uses of bat in (4) are not simply repetitive of each other, we can say that they semantically contrast, and therefore are ambiguous. Since two meanings of friend do not contrast in (2), we can conclude that it is vague.

Another way to test whether the word has two contrasting senses is to see if (in a single situation) one can answer either “yes” or “no” to questions like (5), depending upon which way the word is interpreted. Imagine that a coach is trying to organize a team to play baseball but the equipment hasn’t arrived. The coach asks:

(5) Are there any bats here?

Chances are that the players will answer “no” without even looking into the sky to see if any small flying mammals are about. But if they did notice the bats in the sky, they could answer:

(6) Yes, there are bats, but not bats.

On the other hand, if a child says (7) to you,

(7) Grace and Penny are my friends. Do you have any friends?

you would not assume that there are two sex-specific senses of friends even though the child only referred to female friends. It would thus be strange to say:

(8) # No, I don’t have any friends [= ‘female friends’], but Mark and Paul are my friends [= ‘male friends’].

So, since friend cannot be used to contrast with friend, we conclude that it is vague, not ambiguous.

• • To contrast the word in different contexts and see if it makes any sense when said in the same sentence.

If friend is vague, then in some contexts it may refer to a person who happens to be female and in other contexts to a person who happens to be male. If this were not the case, that is, if friend had two gender-specific senses instead of one general sense, it would make sense to say something like (2) in order to mean (3):

(2) # I have both friends and friends.

(3) I have both male friends and female friends.

• Zeugma test

Another way to tell the difference between ambiguity and vagueness is the zeugma test. A zeugma (or syllepsis) is a sentence in which two different senses of an ambiguous word are “activated” at the same time. Zeugmas sound weird because of the inherent conflict in the word that has to incorporate both senses, as in (9) and (10).

(9)  #Man waxes patriotic, truck (The Onion, November 14, 2001)

(10)  #John had a case of beer and another of measles.

Example (9) is a satirical newspaper headline, which mocks a common headline style in which two objects of a verb are separated by a comma instead of the conjunction and. But in this case, the oddness of using both patriotic and truck as objects of the same instance of wax indicates that the verb wax has two senses. To wax patriotic uses the sense ‘to speak or write in a specified manner’ and goes with an adjective that describes a manner, and to wax a truck uses the sense ‘to apply wax to’ and goes with a noun for a concrete object. In (10), the use of another of measles highlights the ambiguity of case, since another refers back to case, but is meant to be interpreted using a different sense of case (‘instance of a disease’) than the first sense we come across in the sentence (‘box containing a certain number of items’). The weirdness of these sentences indicates that two different senses are operating for wax and case, and thus these words are ambiguous. Compare this to the situation in (11), in which there is no problem in understanding both one and another as referring to friends, even though the referents differ in gender.

(11) One of my friends is a policewoman and another is a fireman.

This, then, provides further evidence that there is one sense of friend that is vague with respect to gender – if it were not, we would have a funny zeugmatic effect here.

Each discrete sense would correspond to one lexeme.

Consider the following example of Zeugma:

• The ship arrived at the port and then the captain drank one glass of it.

If you find this sentence a bit awkard that’s because we have activated both senses at a time, which is not possible.

More examples of Zeugma:

• John and his driving license expired yesterday
• They took the door off its hinges and went through it
• John declined a drink and Peter a verb

Do they sound OK to you?

There is something strange, if that happens is that we have two different lexemes on our mental lexicon for that same word, two different meanings for the same lexeme (sometimes that is used in comedy; to some, marriage is a word. To others, a sentence).

Note that there may be differences among you because individual lexicons are varied.

Each discrete sense would correspond to one lexeme.

Murphy also uses ‘etymology’ (pg. 88) as a criterion to identify lexemes.

POLYSEMT VS. HOMONYMY

Now that we have seen how vagueness differs from ambiguity, the next task is to differentiate between two sources of ambiguity: homonymy and polysemy. Sole in (12) is ambiguous.

(12) Hypermart is the sole distributor.
There are at least three possible interpretations of sole in this sentence, as listed in (13).

(13) _sole_1 ‘the bottom surface of a shoe’

_sole_2 ‘only’

_sole_3 ‘a type of flatfish’

Example (12) might be saying that Hypermart distributes shoe parts (_sole_1) or fish (_sole_3), or it could be the only company to distribute something else (_sole_2). Sentence (14) demonstrates that all three senses can be used without making one another redundant. This is another indication of ambiguity.

(14) Hypermart is the sole2 sole3 sole1 distributor.

In this case Hypermart is either the only distributor of shoe-bottoms for flatfish, or the only distributor of shoe-bottoms made of flatfish. Each sole makes its own contribution.

Sole is a homonym: that is, there are three lexemes that happen to have the form sole. We label the the different lexemes using subscripted numbers, as in (13), in order to refer to them unambiguously in our discussion. (Note that dictionaries also use sub/super-scripted numerals to differentiate homonyms.) But why should we believe that the three senses of sole indicate three entirely separate lexemes that happen to be called sole? There are a few clues. First of all, there would be no reason for a language to develop a single word that means all these three things at once, since they have nothing to do with each other. Even if we could come up with a story for why the senses might be connected, there is historical evidence that at least the first two are not related: _sole_1 comes from the Latin word solea ‘sandal,’ while _sole_2 derives from a different Latin word solus ‘alone.’ These unrelated words only started to look related in English because changes over the centuries robbed them of their final syllables. The fish and the shoe-part senses have some historical origins in common, since they both come from solea, possibly because the fish resembles a sandal. However, since most present-day English speakers don’t know about that link, it is not very relevant to their mental lexicons.

The grammatical properties of the words provide further evidence of their separate status. _Sole_2 is an adjective, while the other two are nouns. The two nouns have different morphological properties, making them seem even more dissimilar: the plural of _sole_1 is soles, but the plural of _sole_3 is sole. Thus, we can be fairly confident that the three _sole_s are homonyms, since English speakers do not seem to be using or perceiving them as a single lexeme.

On the other hand, the other source of lexical ambiguity, polysemy, involves a single word with several senses. This can come about because existing senses branch out to become new senses. So, in the case of polysemy, we expect that the different senses are related in some way – you can figure out why the word came to have this range of senses. This can be seen in the example of coat:

(15) coat a. ‘an outer garment with sleeves for wearing outdoors’

b. ‘an animal’s covering of fur’

c. ‘a covering of paint or similar material’

There are three senses here, but they seem related, since they all involve an outer layer on something. The ‘garment’ sense came first historically, and then, by extension, people also used it to refer to other kinds of coverings. One might wonder if instead coat is vague and has only one general sense that means ‘a covering.’ In that case, the specific uses of coat to refer to the types of things in (15) would follow from that general sense. But notice that these senses are special: each of these kinds of things that can be called coat is really different from the other kinds of things that can be called coat. These particular senses of coat have become conventionalized, so that we use coat more readily for these things than for other types of coverings. For example, we speak naturally of a coat of paint but not a coat of wallpaper (we instead call that covering a layer). If it were a case of a vague ‘covering’ sense that’s applied broadly, then we would expect it to be natural to call any covering a coat.

Sentence (16) demonstrates that coat has at least two senses:1

(16) The desk has three coats on it.

If coat just vaguely meant ‘covering’ in general, this sentence would not give you enough information to form a picture of what the desk looks like with three coats on it – the coats could be tablecloths or blotter pads or anything else that one might use to cover a desk. But, without any further contextual information, the most natural way to interpret this sentence is to see a desk with three garments draped upon it. With a little more contextual prodding, we might prefer to interpret coat in the ‘paint/varnish’ sense:

(17) The desk has three coats on it, but I still don’t think it looks finished.

But if we try to understand it as having both the garment and paint senses at once, we get a zeugma effect:

(18) # The desk has three coats on it – two parkas and one of walnut varnish.

Notice that in order to get the ‘outerwear’ or ‘paint’ senses in (16) or (17), you do not actually have to mention clothing or paint. That specific information has to be coming from the word coat, since there is nothing else in the sentence that indicates clothing or paint. But note that all of these senses of coat are still, in themselves, vague to some degree. For example, coat in its garment sense does not specify whether the garment in question is a fur coat or a lab coat or a raincoat.

Etymology is frequently used by lexicographers, but it is dangerous in lexicology.

Consider again the case of ‘port’:

Port: 1. harbour. 2. kind of wine

On the basis of Zeugma, we concluded it as a case of homonymy, but etymologically the two senses are related.

Port wine originally comes from the Portuguese city ‘Oporto’, which literally translates as ‘the port’.

So, etymology tells us that ‘port’ is polysemous, but do average speakers know this? Remember, we study the mental lexicon.

Note that polysemy should not be confused with vagueness (another type of ambiguity), but do average speakers know this. Remember, we study the mental lexicon.

A lexeme is vague if it is general in its application:

e.g., friend may refer to both a male or female friend.

Types of polysemy:

CONTEXTUAL POLYSEMY

Lexeme may also acquire different traits of meaning in context, not that it is polysemous but that depending on the context we highlight different senses of the meaning:

• The car needs washing (the bodywork)
• The car needs servicing (the engine)
• We can’t afford that car (the car’s prize)
• My car is not that fast (the car’s performance)
• The car crushed Arthur’s foot (the wheel)

We call this contextual polysemy / ambiguity.

REGULAR / SYSTEMATIC POLYSEMY

A variety of polysemy that gets a fair amount of linguistic attention is regular (also called systematic) polysemy. This refers to word senses that are distinct, but which follow a general pattern or rule in the language. For example, words for containers can generally refer to both a kind of container and the contents of the container, as can be seen in (19):

(19) ‘container’ sense: I put some sand into a box/bottle/tin/canister.

‘contents’ sense: I dumped the whole box/bottle/tin/canister onto the floor.

The relation between the ‘container’ and ‘contents’ senses is completely regular, which is to say it is entirely predictable. If we invent a new kind of container, we can be certain that the name of the container will also be able to denote its contents in some situations. Other cases of regular polysemy can vary in their regularity, however. For instance, the names of dances can also name a piece of music that accompanies such dances for some cases, as shown in (20), but not so much for others, as in (21).

(20) I’d like to hear a salsa/tango/waltz.

(21) #I’d like to hear a break-dance/disco.

Sometimes polysemy may be systematic

• Count/Mass alternations

The lamb is running in the field / John ate lamb for breakfast

• Container/containee alternations

Mary broke the bottle / The baby finished the bottle

• Product/producer alternations

The newspaper fired its editor / John spilled coffee on the newspaper

We call this regular polysemy.

INTERIM SUMMARY – DISTINGUSIHING TYPES OF MEANING VARIATION

In this section, we have identified three sources of meaning variation. If a word form is vague, then it has a general sense that can apply to various types of referents, but if it is ambiguous, then it has more than one sense and could be either a homonym or a polyseme. Figure 5.1 summarizes how to distinguish these phenomena.

While these three terms all refer to cases in which the interpretation of a form differs in different contexts, they reflect kinds of relations or properties of lexemes/senses:

• Vagueness is a property that a sense of a single lexeme can have, that is, the property of generality.
• Polysemy is a relation between senses associated with a single lexeme.
• Homonymy is a relation between different lexemes that are coinci- dentally similar in form.

Since vagueness, polysemy, and homonymy apply at different levels of consid- eration (single sense, single lexeme, different lexemes), it is sometimes the case that a single word form can illustrate all of these phenomena. So, the fact that sole is a homonym does not preclude the possibility that any of its senses might also be polysemous and that some of its senses may be vague. For instance, _sole_1 (‘bottom of shoe’) is homonymous with _sole_2 (‘only’), and it is also pol- ysemous in that it can mean both ‘bottom part of a shoe’ and ‘bottom part of a foot.’ Furthermore, the ‘bottom part of a shoe’ sense might be vague in some respects – for instance it does not specify the material the sole is made from, and so could be used to refer to both rubber soles and leather ones.

THREE CATEGORIES, OR A CONTINUUM?

Up to this point, we have taken a traditional view of the ways in which a form can have more than one “meaning,” including some tests to tell which of these cases involve homonyms, a polyseme, or just a vague sense. These kinds of distinctions are recorded in most dictionaries: homonyms are in separate entries; polysemes have several, usually numbered senses within the word’s entry; and vague senses are defined in suitably general ways. But in reality it is often difficult to judge whether two interpretations of a word form constitute a case of homonymy or polysemy or vagueness. Polysemes are supposed to have related senses, but different people have different views of what is related and how related two senses have to be in order to be “related enough” for polysemy. Dictionaries tend to rely on etymological criteria to determine which lexemes are homonyms, but (as already mentioned) historical facts are not always relevant to mental lexicons, since you do not have to know a word’s history to have a lexical entry for it. Vague terms are supposed to have a general definition, but, as we saw for coat, sometimes you can make a general definition covering several uses of the word (e.g. ‘a covering’), but other tests show the meanings to be more distinct. To give another example, I assumed for years that the word ear was a polyseme that could mean ‘an organ for hearing’ or ‘a cob (of corn).’ It seemed to me that cobs of corn came to be called ears (in North America at least) because they stick out from the stalk the same way that our ears stick out from our heads. But later I found out that (a) many people do not see a similarity between the two senses, and (b) the words for hearing organs and corncobs are not etymologically related. The historically different forms are better preserved in German (a relative of English), where they are Ohr and A ̈hre, respectively. On etymological grounds and in other people’s minds, ear is a homonym. But is it right to say that it’s necessarily a homonym in the lexicon of someone who sees them as related concepts?

These kinds of problems have led some to reject a firm distinction between homonymy, polysemy, and vagueness. David Tuggy (1993) has argued for treating these concepts as lying on a continuum – from maximal distinctness (homonymy) to maximal similarity, also known as monosemy, the state of having one mean- ing. Tuggy’s ideas are situated in the Cognitive Linguistics framework, and involve relations between semantic structures, phonological forms, and general- ized conceptual representations. This works similarly to the Conceptual Seman- tics approach to the lexicon that we saw in figure 4.1 in chapter 4, although the particulars of conceptual representation are different in Cognitive Linguistics and Conceptual Semantics. In order to get the gist of Tuggy’s ideas, they’re presented here in a less technical way than in his paper.

At one end of the meaning–variation continuum are cases like bat, which are truly ambiguous; bat can denote sticks or mammals and the senses do not overlap at all. In this case, we have two senses linked to one form, but any conceptual representation that could unite the two senses would have to be nearly empty because there is no common concept that covers the ‘stick’ and ‘mammal’ interpretations of bat. Perhaps it would indicate that it’s a concept for some kinds of things that can move or be moved fast through the air. Meanwhile, these two uses of bat are known in the speech community as distinct senses; those senses are conventionalized. In the illustrations below, the bigger/bolder the type, the more entrenched – that is, well-learnt and easily activated in the mind – a concept or word is. The figure in (22) shows that the conceptual information related to two senses of bat is distant and indistinct, while the two senses are in bold to show that they are well entrenched as conventionally distinct uses of the word.

Further along the continuum are pairs of meanings that are designated by the same word form but have separate senses for which a common conceptual rep- resentation is available, like coat. In (23) we can see two separate senses (at the semantic structures level) linked to a general, common notion, that of covering.

Vagueness involves two semantic structures that are not well entrenched as separate senses, and which have very strong, very close links to an elaborate shared concept, such as the case of friend in (24).

These three cases illustrate the beginning, middle, and end of the continuum, and we could add other examples that come between them, as shown in figure 5.2. All of these types of meaning variation are defined using the same types of struc- tures and relations between them, but differ in the strengths of entrenchment and semantic relatedness found in those structures. For example, the interpretations of book as ‘dictionary’ or ‘novel’ are more distinct from one another than the ‘female’ and ‘male’ interpretations of friend, since there are cases in which book is used to just denote novels. For instance, if you are going to the beach and want to bring a good book, you are likely to mean the type of book with a story, not a dictionary (no matter how good that dictionary is!). Even more distinct are the two senses of coat, for which the separate senses are more conventionalized.

Since we do not expect that all speakers in all English-speaking communities have grasped every lexeme or concept in the same way, we could suppose that some items (like ear) might lie at different points in the continuum for different speakers. Thus, viewing polysemy as a continuum seems to be both consistent with the types of meaning variation that we see in language and with a mentalistic approach to meaning.

MEANING VARIATION AND LANGUAGE CHANGE

Why is there so much meaning variation in language? Some of it serves particular purposes, and some of it is accidental. Vagueness exists because it is useful for us to have general ways of referring to things: the only way not to be vague would be to have a unique name for everything in the world, which would mean that we’d spend our whole lives learning our vocabulary and we would not be able to generalize about things by using words to refer to all members of the same category (since every word would denote a category of exactly one thing). So, most words (except for proper names) start out with some allowance for vagueness. Polysemy exists because it is fruitful to use old words in new ways, rather than having to learn new words all the time. We can usually extend words’ meanings in predictable ways, using a number of processes that we’ll discuss below, and thus people hearing an old word that is used in a new way will be able to appreciate the newly created meaning. Homonymy can develop through accident or through more severe processes in change of meaning. The following subsections consider the processes that give rise to homonymy and polysemy, starting with “co-incidental” sources of homonymy, and moving on to more regular processes of meaning extension.

SOURCES OF “COINDICENTAL” HOMONYMY

Homonymy arises in language mostly through coincidence or because the senses of a polyseme have become so separated from each other over time that we no longer perceive them as the same. One kind of coincidental homonymy arises from changes in the language that bring different words closer together in form. We have seen one example of this already: ear, whose body-part and grain (e.g. ear of corn) meanings have completely different historical roots and were originally pronounced differently. Similarly, the musical genre pop is a clipping from popular music and has nothing to do with the use of pop as an affectionate term for a father (from papa), or as the onomatopoetic verb for the action of burst- ing something like a balloon or a bubble. Each of these meanings of pop evolved in its own way and just happened to end up sounding and looking like one another.

The same situation holds when we borrow words from other languages that happen to be the same as existing words. For example, yen meaning ‘yearning’ (as in I have a yen for fine whiskies) already existed in English when yen, the currency of Japan, was borrowed into the language, so the two _yen_s are homonyms.

PROCESSES OF SEMANTIC CHANGE

We frequently assign new senses to old words. For instance, if some- one were walking around putting cherries on people’s heads, I could say to them Cherry me next! and we’d understand cherry to mean ‘put a cherry on someone’s head.’ A semantic change occurs when a new use of an old word becomes con- ventionalized through repeated use (which is unlikely to happen for that sense of cherry, I must admit). If the word retains its old sense, then it becomes pol- ysemous. If the two senses later drift apart (for instance, if we start substituting gooseberries for the cherries, but keep saying cherry), then the lexeme may divide into two homonyms. Sometimes, the old sense is overshadowed by the new sense (or falls out of use for some other reason), in which case polysemy is just a stage in the transition from old meaning to new meaning. For example, undertaker originally meant anyone who undertakes some business for someone else, and this could include funeral preparations. Gradually, the broader sense of undertaker has been forgotten and it has become specialized to mean ‘someone who undertakes funeral preparations for others.’ Since the original sense has essentially died out, undertaker is not polysemous between the ‘any job’ and ‘funeral’ senses for most English speakers today. But the fact that so many words are polysemous shows that semantic changes often add meanings to the language without subtracting any.

There are several ways in which an old word can develop a new sense, including: metonymy, metaphor, broadening/narrowing, conversion, and grammaticaliza- tion.

Metonymy is when a word is used to refer to something that is related to something else that the word can denote. For example, we might say that a farmer has three hands working for him. In this case, hands refers to laborers – i.e. people who use their hands – rather than to a body part. Similarly, we can refer to things or people by referring to the place where they are, as when we refer to a monarch as the throne or the American film industry as Hollywood. Another example involves using the same word to refer to plants and the food they produce, as can be seen in the two senses of carrots in (25):

(25) a. Grover ate the carrots. (= ‘edible root of the carrot plant’)

b. Grover watered the carrots. (= ‘carrot plant’)

Metonymy can be used productively to create figurative language in a particular context. For example, on discovering a case of double-parking, a car owner might exclaim Someone’s blocked me in! The me in this sentence means ‘my car.’ But if you look up me in a dictionary, you will not find the sense ‘the speaker’s car,’ because that particular interpretation is absolutely dependent on the context.

Lexical change, resulting in polysemy in the lexicon, occurs when the new sense becomes conventionalized, as it has for hand ‘laborer.’

Like metonymy, metaphor is a means of using language figuratively, which can either be used in nonce (one-off) conditions, or can be conventionalized to create new cases of polysemy. Metaphor involves seeing similarities between different things and describing one as if it were the other. We can see an example of conventionalized metaphor in another sense of hand: as in the hands of a clock. A computer mouse is so-called because it resembles a live mouse in terms of its shape and the way it “scurries.” When we put files into a computer folder, we can see the similarity to putting papers into a cardboard folder in a filing cabinet.

Another way in which words can have different senses is if they are autohy- ponyms that is, if one of the word’s senses is a more specific version of another of its senses. Historically speaking, this can happen through the broadening or narrowing of one of the word’s senses. For example, the verb drink can mean ‘consume liquid by mouth’ or ‘consume alcohol by mouth,’ as in (26) and (27), respectively.

(26)   After surgery, Jen could only drink through a straw.

(27)   After his liver transplant, George swore never to drink again.

In this case, the ‘consume alcohol’ sense in (27) is a subcategory of the ‘consume liquid’ sense in (26) – the original ‘liquid’ meaning has been narrowed. An example of broadening is Yankee, which in its original meaning denotes specifically people from the northern United States (in contrast to those from the South), but now can also denote someone from any part of the US, in contrast to those from other countries.

Finally, words can also take on new senses by changing their grammatical category, for example from noun to verb, or, on a grander scale, from content word to function word. If a word keeps the same form (that is, it doesn’t have a prefix or suffix and keeps the same pronunciation) when it changes from one category to another, then it has undergone a process known as conversion (or zero derivation). For instance, verbs expressing emotional or mental states can often be used as nouns for those states: love, hope, fear. In the noun-to-verb direction, all of the following names for liquids can also be verbs: paint, oil, water, milk. Have you noticed that one of these is not like the others? While the verbs paint, oil, and water all have senses relating to putting the liquid on/in something, the verb sense for milk involves taking the liquid out of something (an animal). In these cases, conversion from noun to verb has added some meaning, and we can see patterns in how lexical items in one category change when they are converted to other categories. Generally, conversions can happen from any lexical word class to another, though they tend to happen to morphologically simple words. For example, there is little point in making a verb out of the morphologically complex noun denial since there is already a morphologically simple verb deny doing the job.

You’ll know from experience that conversion happens quite easily in English. For instance, recent technologies have spurred on noun-to-verb conversions like to Google ‘to search for a word or phrase using the Google search engine,’ to friend ‘to select someone as your friend on a social networking site,’ to text ‘to send a text message.’ Much rarer and slower are cases of grammaticalization, in which lexical content words change to grammatical function words or functional morphemes – yet this is how languages get most of their function words. For example, the modal verb will, which we use as a future tense marker, has come to modern English from the Old English lexical verb willan, which meant ‘want (to do something).’

How did this happen? Well, if you want to do something, then you are probably talking about a future action. For example, if I say I want to eat lunch, I’m not eating lunch at the moment, but there is a good chance I will eat it in the near future. So the seeds of ‘futureness’ were already there in willan. The verb already appears before other verbs, so it is in the right position to be interpreted as an auxiliary verb. Over generations of learners, people paid less and less attention to the lexical ‘want’ aspect of the meaning and focused on the futureness of it – until the lexical meaning was mostly forgotten. This process is known as semantic bleaching, since the main force of the meaning has been washed away. The future marker will went a long way in its grammaticalization – losing all the grammatical markings of a lexical verb. Thus, as (28) shows, unlike the lexical verb want that willan was originally like, will does not appear with tense or agreement marking (the -s on wants) or an infinitive marker (to) on the following verb. And, again, unlike want, it cannot appear with a modal verb (in standard dialects of English).

(28)   ∗Ira wills to go home. (vs. Ira wants to go home.)

(29)   ∗Ira can will go home. (vs. Ira can want to go home.)

In other cases, we can see semi-grammaticalization – which may mean that we’re still on the path to grammaticalizing a form. For instance, Romaine and Lange (1991) have suggested that the use of like to introduce reported speech, as in (30), is on its way to being grammaticalized as a ‘quotative complementizer’ – that is, a grammatical morpheme that links a main clause to a quotation.

(30) Jody was like ‘I can’t believe it!’

Because it is still in the early stages of grammaticalization, the grammatical status of quotative like is hard to define (is it an adverb? a preposition? a com- plementizer?) and it still retains some of its comparative lexical meaning, in that we seem to be saying that Jody in (30) said something like I can’t believe it, but didn’t necessarily say it in those words or in the same manner as the reporter of the utterance has said it. But the prediction is that like will become much more regular in use and lose the association with comparison as it becomes more grammaticalized over the years. By the time that grammaticalization is finished, we should perceive the quotative like as a different word from the comparative preposition like in days like this. That is to say, it will become different enough from its original use that the relation between comparative like and quotative like will be a clear case of homonymy, rather than polysemy. For this reason too, we see the future marker will as a homonym of the ‘document for expressing wishes in the event of one’s death’ will, although historically they come from the same verb.

In this section, we have seen a lot of ways in which words expand their uses and how they sometimes split off into being different words altogether. Now that we have described the range of phenomena that are related to meaning variation, let’s look at how some semantic theories have tried to account for the fact that the relation between word form and word meaning is not one-to-one.