The grammar of words: words and word parts

Can words contain other meaningful elements that are not words?

5.1 Introduction

In this chapter, we discuss how linguistic meaning is encapsulated in different levels of patterning of linguistic form, looking in turn at words, sounds and sentences. We will do this by looking at language data that illustrate the particular points about linguistic meanings that we wish to highlight. Much of the data will be taken from English, because this is the language that we share: we are writing this book in English and you are reading it in English. Do keep in mind, however, that English is one particular language among thousands of other languages around the world, and that this book deals with how to analyse language in general, not particular languages. This means two things.

First, the language data used in this book are to be taken as representative examples. Clearly, no one can use an infinite amount of data in any book. We have chosen the best examples we could think of, to help us bring our points across in the clearest possible way. Here’s an example of what we mean. If you had to explain to us what an insect is, you could choose to show us samples, like a beetle or an ant, as representative members of a class of beings called “insects”, and then tell us to work out, from these samples, what is it that characterises insects. You could even add that a spider or an earthworm are not insects, to help us better understand what can rightfully be called an insect and what cannot. This is exactly what we expect you to do from the examples that we give.

Second, the concepts that we exemplify using the English language are meant to apply to other languages as well. Think of our English examples as you would think of using the example of a pendulum, or of a coin dropping to the ground, to explain the general law of gravity. It wouldn’t make sense to conclude that the law of gravity applies only to coins or to pendulums, just because we use coins and pendulums to demonstrate the effect of the law.

One very important way in which we human beings understand the world around us and learn to talk about it is by means of generalisation. We find certain properties in certain objects, and we extend those properties to other objects that we perceive as similar. This is why some people say that “everybody” knows that “women” enjoy shopping, or that “snakes” are dangerous. No one has ever polled every single woman to check their shopping enjoyment, or every single snake around the world to check their threatening behaviour. Nor every single human being about their opinions on women or snakes. Rather, generalisations such as these are based on observation of a subset of the population of women and snakes, respectively. Here’s another example to bring home the same point. If you taste cheese for the first time and find it disgusting, you may conclude that any kind of cheese will taste disgusting. Generalisation is a vital component of any scientific explanation, but with one proviso, that is equally vital. Just as you would have to change your mind about cheese if you decide to try a Camembert and find it delicious, so scientists will narrow down their generalisations, and change the labels by which they call things, when they find counterexamples, examples that contradict what they previously thought. In the same way, what we will have to say in this book about English applies generally, across different languages. A good way of reminding yourself of this throughout this book is to try to apply the concepts that we describe here to any other language(s) that you may be familiar with, as soon as they are described. At the same time, you should also be looking out for counterexamples, whether in English or in other languages that you are familiar with. A few activities and exercises in the book will help you keep these two very important points in mind!

5.2 The word “word”

Morphology can be defined as the study of words. Let’s start by checking out what a word is.

You may or may not agree that all three examples above are words. If you feel any discomfort about calling each of them a “word”, you are not alone. Word is in fact one of the concepts in linguistics that defies precise definition. Since everyone talks about words, we might assume that this is a well-understood concept. The truth is that it is not. What is meant by a word varies greatly from language to language. Some languages, like Mandarin, have largely monosyllabic words, which are words that consist of a single syllable. Other languages, like Malay, allow the stringing together of several syllables, or several words, to form larger words. This is also the case in Welsh, where the word in (5.1) contains other words, llan, fair, pwll, etc., just like many words of English do too. For example, the English word handbag contains the word hand and the word bag. The difference is that Welsh allows the spelling together of more words than English does. If English had the same spelling rules as Welsh, the English translation of the Welsh word above could be spelt like this:

Trying to work out the meaning of a word so spelt is not much worse than trying to work out the meaning of chemical compounds or of medical terms, like this well-formed English word that we found listed in the Oxford English Dictionary:

What is meant by a “word” also varies within a single language, including whether we are talking about speaking words or writing them. Perhaps you would want to say that karaoke-singers is two words, because of the hyphen separating karaoke from singers in the spelling? Or would it satisfy you better to say that it is a single word, because it has a unique meaning, just like say, sopranos does? This is a single word whose meaning can be rephrased as ‘opera-singers’. The best definitions of the word word take either orthography (spelling) or rhythm into account. Orthographic definitions take word as a unit that is separated by a blank space on each side, in a printed text. Rhythmical definitions take it as a unit of speech that can be separated by an optional pause, meaning that a word can be pronounced on its own, preceded and followed by silence. The blank spaces on a page represent the possible silences between words. Robot-characters in some science-fiction films pronounce sentences in just this way, word by word.

Try to pronounce the following sentences in a robot-like manner, word by word: I saw a black bird by the green house. I saw a blackbird by the greenhouse. How many words do you count in each sentence?

The fact that we do understand what robot-like sentences mean, despite their stilted delivery, reveals one very important feature of human speech. This is that speech is a stream of ordered units, as already discussed in Chapter 1. Human speakers, and robots that attempt to replicate their speech, must order the units that make up their utterances in a particular way, so that the utterances so formed make sense. That is, linguistic units must occur in particular positions along the stream of speech, surrounded by other units. This observation helps us clarify one concept that is central to linguistic analysis, the concept of distribution.

5.3 Distribution

Linguistic units, like people or objects, show up in predictable places, or contexts. You wouldn’t expect to find the morning newspaper tucked away in the fridge, or a cat sitting next to you watching a movie at the cinema, or a palm tree in full bloom in the North Pole. Linguistic units pattern in the same predictable way.

The context of a particular linguistic unit is given by the linguistic units of the same type that surround that unit. For example, the context of a particular word is given by the words that precede and follow it. The same holds for morphemes (see section 5.4.2 below for morphemes) or sounds. Analysis of context in these terms is a natural consequence of the fact that speech occurs along the dimension of time: sounds follow sounds, words follow words. The distribution of a particular linguistic unit is then the set of contexts in which that unit is found to occur. By the same token, you can also work out your own distribution, if you list all the places in which you are likely to be found.

The distribution of linguistic units may be represented by a distributional frame. Given a form like, say, XYZ, the context of the unit Y is given by a distributional frame with the general form:

This representations is similar to a more familiar one like, say, a + b = c, where each letter stands for a number. In both cases, the letters represent a variable that can be replaced by something else, according to conventions that all users of these representations have agreed upon. The conventions in distributional frames are:

• Each symbol X, Y, Z, represents one linguistic unit of the same type, e.g. a morpheme or a sound. These units occur in the given sequential order: X precedes Y, and both precede Z.
• The blank represented by indicates the context in which Y, the unit in question, occurs. In this case, Y follows X and precedes Z.

For example, given the phrase the brown cat, we say that the distribution of the unit (word) brown is given by the distributional frame the _ cat, where the blank indicates the occurrence of brown. Likewise, the distribution of the in the same phrase is given by the frame _ brown. Or, given the word cat, pronounced [kæt], we say that the distribution of the unit (sound) [æ] is given by the distributional frame [k] _ [t] (see Chapter 3 for conventions on the representation of speech sounds).

5.4 Morphological units

There are two chief levels of word patterning that interest morphologists. One deals with patterns of words to form phrases or sentences, based on our observation that different words behave differently when chained together with other words. The characteristic behaviour of particular groups of words allows us to classify words into different word classes. The other level of morphological analysis deals with patterns within the words themselves, that is, with the internal grammatical structure of words.

We now discuss these two analytical levels in turn. Notice, however, that both analytical levels work together to provide us with insight about word patterning.

5.4.1 Word classes

Check the following data:

The cat sleeps on the mat.

*The sleeps cat on the mat.

*The cat sleeps the on mat.

Recall, as said in Chapter 1, that an asterisk preceding a form indicates that that form does not occur in the language under analysis. All three sentences in (5.2) contain the same words. The only difference is that the words pattern differently, resulting in one well-formed sentence and two ill-formed ones. For example, we see that the word the can precede the word cat, but not the words sleeps or on. Observations of this kind lead us to assign words to different word classes, according to their relative positions in a phrase or a sentence, that is, according to which words can or cannot follow or precede other words. This chain ordering of words tells us about their distributional properties, by showing us which particular positions are grammatical for which types of words. In the linguistics literature, word classes are sometimes called parts of speech, lexical classes, grammatical categories, grammatical classes. For our purposes, you can take all of these terms as equivalent.

Words can be broadly divided into two main word types – lexical words and grammatical words – according to their distribution and their meaning. Within each of the two types, several word classes can be further distinguished, because each word class patterns in characteristic ways.

Below, we give a number of principles, or criteria, that can help us identify different word classes. As you read through them and think about their application to different words in the same word class, you should keep in mind that none of the current criteria for defining word classes is watertight, including the ones that we suggest here. This means that once you understand how these criteria apply, you will be able to come up both with words that fit the criteria as well as with words that fail the criteria. The latter are counterexamples, showing that the criteria represent generalisations about what holds true most of the time, rather than all of the time. We will give a few examples of these ‘bad-behaved’ words ourselves, where relevant. This is not a problem for our analysis: this book presents only a very elementary set of principles to help us deal with language (languages are very, very complex things!), and professional linguists themselves go on being baffled by the complexity and quirkiness of language. One of them, Edward Sapir, once famously said that “all grammars leak” (Sapir 1921: 38). These “leaks” are precisely what makes us want to go on trying to understand how language works, so that we can fix them to make our analyses “flow” in a satisfactory way. In this sense, a good grammarian is like a good plumber. Take the criteria that we offer here as typical criteria, that do useful work in helping us identify word classes, but only in the majority of cases rather than in all cases.

Lexical word classes

Lexical words represent a specific referent in the world of our experiences. They refer to objects and substances that we can see, sensations that we can feel, qualities and events that we can observe. Lexical words therefore form the largest group of words in languages. The word classes to which they belong are open classes, in that the overwhelming majority of new words that become part of a language are of this type. In the linguistics literature, the terms lexical words, content words and open-class words are sometimes used interchangeably. These terms reflect the fact that whenever a new interesting thing is created, invented or found, we immediately create, invent or find a new lexical word to go with it, so that we can talk about it. When the authors of this book were growing up, there was no email, SMS or blogs, and so there were no words for these things either. There were, however, telegrams, telex and vinyl LPs, and the words for them were part of daily life at the time. New words that restock the vocabulary of languages, and old words that fade away for lack of linguistic demand are all lexical words. We can say that languages keep themselves alive and working through the comings and goings of their lexical words.

Lexical meanings are generally referential and, as such, they are arbitrary and idiosyncratic. Four lexical word classes can be distinguished in English according to their distributional properties. These are nouns (N), verbs (V), adjectives (Adj) and adverbs (Adv). We now discuss each one in turn.

Noun (N)

Nouns are the only lexical word class that can be followed by a mark of plural, in English. Plural forms of nouns are often represented in spelling by – (e)s at the end of the word in so-called regular plurals, where (e) may or may not occur, depending on the spelling of the singular form. For example, the word cup is a noun in the singular form, representing a single object that we call “cup”. The word cups is also a noun, in the plural form that indicates more than one of those objects. The same is true of a noun like brush and its plural brushes. Regular forms are forms that are productive, or active, in languages. Taking English nouns as example, this means that whenever a new noun comes into the language, its plural will be formed by adding -(e)s to it, as in faxes, emails or modems.

By our distributional criterion, if you can fit a word in the following frame, then that word is a noun:

The regular plural marker is given in brackets to allow for irregular plural nouns like foot-feet, tooth-teeth, and goose-geese, as well as nouns which have the same form for singular and plural, e.g. sheep, deer, fish, and fruit. We do not otherwise deal with irregular morphology in this book.

Our distributional frame shows that a word like cow (or foot) is a noun. However, we need to point out that saying that “only nouns can be pluralised” does not mean that ‘all nouns can be pluralised’. English nouns pattern in two different ways. Nouns like cow can be pluralised because their referents can be counted. These nouns are therefore called count nouns. A word like milk cannot fit the given frame, because phrases like *one milk and *several milks are not well-formed for most speakers of English. But words like milk fit other distributional frames associated with nouns, e.g. patterning after a determiner (see below). The word milk is therefore a noun too, although its referent cannot be counted. The reason is that nouns like milk refer to shapeless substances, and counting applies only to referents with well-defined dimensions. This being so, we can propose a frame that provides a shape- giving context for non-countable nouns:

The first blank can be filled with countable, shape-giving words like box, packet, loaf, glass, bowl (sometimes called measure words), much in the way that classifiers are used before nouns in Chinese and other Asian languages. The non-countable noun fills the second blank. For example, a glass of milk is a well-formed English phrase. “Shapeless” nouns like milk are called mass nouns.

Besides plural forms, another feature that characterises nouns is that they can be followed, in writing, by -’s, a mark that is used to indicate possession, or belonging, and that is called possessive or genitive in the literature. When used with a noun followed by another noun, the possessive mark indicates that the second noun in some way is part of, or belongs to, the first one. For example, when we refer to the cat’s tail we are talking about the tail of a particular cat.

Nouns name entities of various kinds. Nouns that name people, places, institutions and brands, like Jane, Malaysia, Telecom or Stradivarius, are called proper nouns (or proper names) and are also treated as nouns in the literature. All other nouns are common nouns. Like common nouns, proper nouns can be followed by the possessive ’s. One example is a phrase like Malaysia’s climate.

However, the remaining properties of proper nouns are different from those of common nouns. First, proper nouns cannot be said to have well- established referents across the board. The name India, for example, refers to a particular country, but the word India does not “mean” the country named by this word, nor does it mean a set of objects with perceivable shared properties. Rather, it is a label that the country goes by, just like Mary is a label that people called Mary go by. It would be hard to find a common feature of meaning among all people named Mary, parallel to the feature of meaning that allows us to designate all cups by the name “cup”. What all the individuals named Mary have in common is that someone decided to call them “Mary”.

Second, these words fit only marginally in both of the distributional frames given above. They do not designate substances, and we can “count” Jimmies and Chloes in particular groups (children love to do this), or Ugandas and Englands if we want to highlight, say, striking features of different parts of the same country (adults love to do this), but only in a marginal sense.

Third, proper nouns pattern equally marginally with other words that can precede nouns, like determiners and adjectives (see below): expressions like That Japan is my favourite one or This is a green Matthew are, to say the least, unusual. Given these provisos, we follow here the traditional classification of proper nouns as nouns.

Verb (V)

Verbs can be followed by a mark of past tense, often represented in spelling by -(e)d in so-called regular verbs. As noted above, we are only concerned with regular morphology. A word like bake fits this frame, and is therefore identified as a verb:

Some sentences contain only one verb. Examples are I tripped or My neighbours have two dogs, where tripped and have are the verbs. Other sentences contain several verbs, from two up to a maximum of five, in English. Examples are Janet is singing and The laundry would have been being washed, where is singing and would have been being washed are all verbs. Some people doubt the “correctness” of the latter string of verbs, but there’s nothing grammatically wrong with it: it is simply a less common construction.

We call the last verb in strings like these the main verb, because this is the verb that carries the referential meaning that is being talked about. The other verbs are called auxiliary verbs, or auxiliaries, because their function is to help specify the time and duration of the action or state indicated by the main verb. In sentences with only one verb, this verb is of course the main verb. For example, the sentence Janet is beautiful has only one verb, is, and this is the main verb.

Three English verbs, be, do and have, can function either as main or auxiliary verbs. In the sentence My neighbours have two dogs the verb have is the main verb, whereas it is an auxiliary in the sentence The laundry would have been being washed. Because of this dual function, these verbs are very common in English, and because they are so common they are also very irregular. We saw above that newcomer words to a language, like email, that are therefore just becoming common in it, follow regular patterns, not irregular ones. Words are like clothes, the more you use them the more shapeless they get and the less they look like new clothes off the rack. The verb be is in fact the most irregular verb in English, in that it can appear in eight different forms. Just to satisfy your curiosity, these forms are: am, are, is, was, were, being, been and be itself!

Verbal forms that vary according to tense (and/or person, see below) are called finite forms. For example, forms like are, were, baked, bakes, are finite. If a sentence contains only one verb, the form of that verb is finite. The verb have is finite in the sentence My neighbours have two dogs, because it changes to had when we talk about the past, or to has when we say My neighbour has two dogs. Verbal forms that remain unchanged regardless of tense and person are non-finite forms. For example, (to) speak, speaking, spoken, and (to) be, being, been. In a sentence with main verb and auxiliaries, the first auxiliary is finite and all the other verb forms are non-finite. In the sentence The laundry would have been being washed, only the auxiliary would is finite.

1. I am doing my homework.
2. I am a teacher.
3. I have been doing my homework.
4. I have seen that film.
5. I did my shopping yesterday.

Adjective (Adj)

Adjectives pattern in two typical ways. They can immediately precede nouns, or they can follow forms of a verb like be. The distributional frames for these two patterns are:

If a word can pattern in these two ways, then it is an adjective: brown is one example. Typical adjectives do indeed pattern in these two alternative ways, although a few do not. Some adjectives can be used only before nouns. For example, you can say the current president, where current is an adjective, but you cannot say * the president is current . Conversely, other adjectives pattern only after be -like verbs: you can say dinner is ready , but not * a ready dinner .

The function of adjectives is to modify nouns. This means that adjectives tell us something about a noun, specifying a property or a quality of that noun. If we say the green book or that large whale, we’re indicating which particular book or whale we mean: green, not red, and large, not tiny. This qualification of nouns can also be a matter of degree. If we see first a large whale, and later a very, very large whale, we can express this by saying that the second whale is larger than the first. If we then see a third, absolutely enormous whale, we can say that this one is the largest of them all. In both cases we are comparing sizes, and grading them: one is larger than another, like one can be smaller than another, or one can have the greatest (or smallest) size compared to all the others. The technical names for these uses of adjectives are comparative and superlative, respectively.

Degree forms of adjectives vary according to the length of the adjective word. There are two rules. For short adjectives with one or two syllables, we add -(e)r and -(e)st at the end of the word. For long adjectives, we add separate words, more and most before the adjective. We can say that dolphins are more intelligent than whales, or that they are the most intelligent marine mammals, but we don’t say that they are *intelligenter than whales, or the *intelligentest of all. By the same token, we say easier, not *more easy, and easiest, not *most easy. (And yes, this short word vs. long word rule has several exceptions: short adjectives like tired and ready are examples.)

Adverb (Adv)

Traditionally, the “class” of adverbs is a sort of ragbag: if morphological criteria cannot clearly identify the class of a lexical word, the solution is often to call it an “adverb”.

This is not simply an easy way out of a difficulty in classifying certain words. It underscores the difficulty itself instead. Similar problems of classification usually arise for two main reasons: either we are basing our classification framework on insufficient data, or the classification framework itself needs revamping. In other words, either the number of clearly identifiable words that we have observed, large though it may be, is not enough to help us decide the word class of a particular word; or perhaps we should start thinking about discarding the class “adverb” altogether, and create new word classes that better explain what makes “adverbs” special. Zoologists, for example, had to create a new zoological class to account for platypuses, the very odd mammals that lay eggs instead of giving birth to live young as “regular” mammals do. Facts, about living beings or language, cannot be changed to suit a theory that doesn’t explain them. It is the theory, no matter how respectable or how popular, that must change in order to serve our understanding of these facts. Adverbs may well be the platypuses of the grammatical zoo.

The facts are that attempting to identify adverbs on formal or distributional grounds is not straightforward, in that they may vary widely in shape as well as in patterning within phrases and sentences. The most common criteria used in the definition of adverbs are of two kinds. One, a referential criterion, states that the meaning of adverbs can modify the meaning of different words or phrases, including other adverbs. That is, their meaning attributes some quality to the meaning of another word. For example:

In (5.3), very modifies the adjective pretty by intensifying its meaning. In (5.4), the word fast modifies the verb speaks by specifying the way in which the speaking takes place, and very in turn modifies fast. Both very and fast are therefore adverbs.

The second criterion is distributional, and states that adverbs are generally mobile words. This means that they may occur in different positions within an utterance without loss of grammaticality. For example:

Sadly, she is an idiot. She, sadly, is an idiot.

She is, sadly, an idiot. She is an idiot, sadly.

The mobility of sadly identifies it as an adverb. Note that mobility is further demarcated by pauses, in speech, and by commas, in print. Note also that this criterion would fail to identify very as an adverb.

Grammatical word classes

We said above that lexical word classes express content in terms of concepts of various kinds. In contrast, grammatical words represent distributional relationships between lexical words. Grammatical words belong in closed classes, in that new words of a language rarely are of this type. Across languages, the number of words in grammatical word classes is therefore much smaller than in lexical word classes. In the literature, other terms that designate grammatical words are closed-class words and function words.

The meanings of grammatical words can be said to be structural and systematic, rather than referential and idiosyncratic. These meanings are the “glue”, as it were, that helps us express and understand the meaning relationships between lexical words. Four classes of grammatical words are usually described for English. These are determiners (Det), pronouns (P), conjunctions (C) and prepositions (P), as follows.

Determiner (Det)

Determiners can precede both N and Adj: this cow and this large cow are both well-formed in English, where this is a determiner. A frame for Det might then be, where the brackets indicate an optional word:

This frame identifies words like the, a, these, my, your, both, as determiners.

The word class determiner includes different types of words that fit the frame above. Would the distribution of determiners in the following data allow us to say that there are different types of determiner in English? Why?

Pronoun (Pr)

Pronouns can replace sequences of an optional determiner, followed by an optional adjective, followed by a noun. Using the symbol ≈ to indicate distributional equivalence, we can formalise the distribution of pronouns in this way:

In the three sentences in (3.6), the word sequences in italics can all be replaced by a word like they, identifying this word as a pronoun:

Those black cats are so annoying.

Black cats are so annoying.

Cats are so annoying.

Pronouns are interesting words, because they work by proxy, as it were. Hence their name, pro-nouns. Personal pronouns, for example, are used to replace direct reference to ourselves and our interlocutors. Instead of referring to ourselves and our conversation partners by name, we use pronouns. The pronoun I, for example, is used by speakers to refer to themselves, while the same speakers use you to refer to their listeners. When listeners in turn become speakers, they use the two words in exactly the same way. These are therefore called first person and second person pronouns, respectively – the speaker comes “first”. English has first person pronouns that have singular and plural forms, I to refer to the speaker only, and we to refer to a group including the speaker. In contrast, the English second person pronoun you has the same form to refer to one or more than one interlocutor. Third person pronouns refer to what the conversation is about. In English, these pronouns have three forms for the singular, he, she and it, the first two usually referring to sexed beings, and the latter to inanimate referents or beings whose sex is irrelevant or unknown. The plural form is the same for all three, they.

Other pronouns are also used in relation to the participants in an exchange. Possessive pronouns (e.g. his, hers, mine, theirs, yours) indicate whether something belongs to or is a characteristic of those participants. Demonstrative pronouns (e.g. this, those) show the distance between the speaker and the referent that is being talked about. Take, for example, the sentence:

This sentence contains two pronouns, these and yours. The first is demonstrative, and signals a referent that is near the speaker. (In contrast, the pronoun those signals a referent further away from the speaker.) The second pronoun is possessive, indicating that the listener owns whatever the speaker is referring to.

What is the word class of the underlined words in these sentences? Explain how you reached your decisions.

1. That man really loves boiled squid.
2. I can’t understand that.
3. My plate of barbecued squid is much tastier than his.
4. I find his choice of food very funny.

Conjunction (Conj)

Conjunctions are linking words, “conjoining” other words or phrases in order to enable multiple occurrences of the same word class or phrase. Conjunctions link units that are of the same type, e.g. an adjective with an adjective or a pronoun with a pronoun (section 7.5 deals with conjunctions in greater detail). In these two examples, the italicised conjunction and joins two pronouns in the first sentence, and two sequences of Adv Adj in the second sentence:

You and I need to talk seriously.

My cat is very beautiful and very stupid.

Preposition (P)

Prepositions are linking words that typically show relationships of space and time between other words or phrases. They are followed by (Det) (Adj) N sequences, or by pronouns, which replace these sequences. The distributional frames for a preposition are:

These frames identify words like in, under, despite, through, as prepositions. The meanings expressed by prepositions can be referential, like those of lexical words. For example, the word under regularly means a location tucked below something on a higher level in space. Or a preposition like during typically refers to a period of time. Other prepositional meanings are not fully lexical. For example, a word like in does not mean anything that can be usefully generalised from uses like in the house, in a moment, in conclusion, in fact, in neat rows, in Swahili. Similarly to the problems raised in the classification of adverbs, discussed above, these observations about prepositional meanings raise other problems for our analytical framework. In this case, we may question the adequacy of a watertight distinction between “lexical” and “grammatical” word meanings, that leaves prepositional meanings scattered between both.

Once particular words have been safely assigned a word class, we can use those words as shortcut tests of the class of other words. This is the substitution criterion. By this criterion, any word that can replace the noun cow in contexts where cow is found is also a noun. The same principle applies to the other word classes.

Here is a summary of the word classes discussed in this chapter.

Word class

Examples