Problems with definitions
So far we have been assuming that it is actually possible to formulate successful definitions for words in a significant number of cases. We should not take this for granted, however. One of the most frequent criticisms of definitional theories of semantics (as opposed, for example, to referential/ denotational ones: see 1.6.1) is that no satisfying definition of a word has ever actually been formulated. The scepticism about the existence of definitions is so widespread, in fact, that many researchers in disciplines closely related to linguistics, such as cognitive science and artificial intelligence, have completely abandoned the idea that definitions even exist (see Fodor 1998 for details): the emphasis on definition in linguistics strikes many from these other disciplines as misguided. The overall rejection of definition outside linguistics as a defensible mode of meaning analysis has various motivations. Many of them derive from the problems involved in a psychologistic interpretation of definitions as concepts, in which the structure of a definition reflects the structure of the underlying concept (so that, for example, the concept BACHELOR could be said to be the combination of the concepts UNMARRIED and MAN in just the same way as the definition of bachelor might be thought to be ‘unmarried man’).
Linguistics, however, is not necessarily committed to a conceptualist interpretation of definitions. As our initial typology of definition suggested, a definition can serve many purposes, of which revealing the putative underlying conceptual structure of a unit of language is only one. Nevertheless, the extreme difficulty of phrasing accurate definitions should be an embarrassment to any theory in which definitions are a privileged mode of semantic analysis.
A classic case of definitional inadequacy is the proposed ‘definition’ we have just mentioned of bachelor as ‘unmarried man’; this is the type of definition found in many popular dictionaries. One problem here is that there are many types of unmarried male, such as widowers, the Pope and Tarzan, whom we would not describe as bachelors. As a result, ‘unmarried male’ is not substitutable for bachelor salvaveritate, and the definition therefore fails. There are many other examples like this. Even though we typically think of many words as having concise definitions, and that these definitions accurately convey the word’s meaning, more detailed investigation reveals that our intuitions on this are in fact mistaken: when these definitions are subjected to scrutiny, it readily emerges that they are not, in fact, successful (for more examples, see 7.1.2).
A possible response here would be to claim that it is only the extreme brevity of the definition of bachelor which accounts for its inadequacy. If a definer tries hard enough, satisfactory definitions can be achieved: it is just that no one has yet taken the time to do so. This is exactly the point made by Wierzbicka (1996). According to her, the true definitions of most ordinary words are significantly longer than the brief statements we are used to reading in dictionaries. Here, for example, is her ‘imperfect first approximation’ definition of paint:
As Wierzbicka points out, this definition is both longer, and structurally different from the types of definition familiar from dictionaries. It is essentially a definition by context, specifying ‘a fairly complex scenario, with a number of temporal and causal links between the components’ (Wierzbicka 1996: 255). Even in this complex, detailed form, however, the definition’s accuracy can be challenged. If I have painted a car with red car enamel, there are many visible parts of the car which are not painted – the glass, the bumper bar, the headlights, the number plate, etc. Similar remarks apply to the sentence She painted her face with rouge. This invalidates component (e) of the definition. Components (i) and (j) are also problematic: if I painted a box with red paint, do we really want to say that I wanted the red paint to be like part of the box and that when I had finished the red paint was like part of the box? Maybe we do, maybe we don’t: in any case, it doesn’t seem clear enough to warrant inclusion in a definition. Similarly, the whole definition is equally appropriate as a definition of the word spray, as in the sentences I sprayed my arm with insect repellant or I sprayed the flower with water. I did this in order for my arm/the flower to look good (in the case of the arm, so that it would not be covered with ugly bites (component (c)), anyone looking at the repellant or the spray at that time, as at any other, could say what colour it was (component (g) – white, or transparent), the spray is clearly water-like (h) and I wanted the spray to become like part of the arm/the flower (components (i) and (j)). But since spray and paint are intuitively not synonyms, the definition of paint should exclude spray. (Note that spraying is also not a kind of painting; we cannot defi ne spray simply by adding some extra components to the definition of paint.) This objection is possible because of the definition’s description of the physical action of painting as putting stuff on the visible parts of the object. There are several ways of putting stuff on the visible parts of an object which observe all the conditions in (a) to (j) but which are not painting: as well as spraying, the process of dipping comes to mind. Apparently, then, increasing the length and detail of a definition does not necessarily increase its accuracy. (See Geeraerts 1993 for more discussion.)
One typical response to criticisms of definitions has been to claim that the definition captures the core, central, or prototypical part of the definiendum’s meaning, and that the respects in which the definition fails all involve special or peripheral aspects (for a discussion of the notion of prototype, see Chapter 7). Thus, Wierzbicka introduces components (b) and (c) into the definition of paint in order to take account of cases like painting something with a special solution in order to protect it against rust: this is not strictly done in order to make something look good, but it is like other acts of painting which do have this goal. Similarly, one could respond to the above criticism of component (e) by saying that the usual case when painting something is that the thing being painted be completely covered, and that other cases are somehow derived from this one. And to return to our earlier examples, a linguist could claim that ‘unmarried male’ is still a legitimate definition of bachelor in spite of the counter examples of widowers, etc., because it captures the core sense of bachelor.
These sorts of response are reasonable insofar as it does seem to be true that many words have a central set of typical applications, and a set of less typical ones. Bachelor, for instance, does indeed very often refer to unmarried males, even if not all unmarried males can be described as bachelors. We will consider this type of view of word meaning in Chapter 7. But with out a principled way of delimiting what does and does not count as a core meaning or a prototypical instance of a word, it is too easy for the definer to simply dismiss any counterexample to a definition as simply relating to a ‘non-core’ aspect of the definiendum’s meaning. Fodor puts this option with characteristic bluntness:
‘Core meaning’ and the like are not, however, notions for which much precise explication gets provided in the lexical semantics literature. The upshot, often enough, is that the definitions that are put on offer are isolated, simply by stipulation, from prima facie counterexamples.
This strikes me as a mug’s game, and not one that I’m tempted to play. I take the proper ground rule to be that one expression defines another only if the two expressions are synonymous; and I take it to be a necessary condition for their synonymy that whatever one expression applies to, the other does too. (1998: 48)
Under this view, a definition has to be substitutable for the definiendum in every single context. As we will see in the next chapter, given the infinite variety of language use, it would seem that any word can be used in any context (to see this, think of errors, joke contexts and fantasy): as a result, a comprehensive definition would seem unattainable.
Many linguists, however, would reject the argument that the heterogeneity of uses renders comprehensive definition impossible. For them, the fact that any word can be used in any context is only true in a trivial sense: there are clear differences between core and non-core uses, and definitions are possible for the former.
