On Polylexy

On Polylexy D. A. Cruse A he two lead papers by Wierzbicka and Atkins agree on two major points: first, that dictionaries are an important language resource , and therefore that lexicography is a significant activity; and second, that linguistic theory, and lexical semantics in particular, is at least potentially of considerable relevance to lexicographers, whether as a means of "consciousness raising" or as a source of guiding principles that can contribute to practical decision making. However, within this broad framework of agreement, there are quite sharp differences both of opinion and emphasis. As a lexical semanticist with no practical experience in lexicography , I found Atkins's paper useful and illuminating in two principal ways. First, it draws attention to the many non-linguistic factors of which dictionary-makers must take account; second, it gives a lexicographer 's perception of areas where an input from linguistic theory would be relevant (these are presumably points where lexicographers intuitively feel uncertain of their ground). Atkins also throws down a practical challenge in the form ofa set ofcitations of the word taste for classifying into separate senses. I will return to this question below. Wierzbicka's paper is characteristically forthright and stimulating . I generally agree with her view that word meaning ought to have a central place in language study, and I concur in deploring its relative neglect in mainstream linguistics. I also agree that a semantics that has nothing to say to lexicographers should be regarded with the deepest suspicion. However, there are also a number of points where my views differ from those of Wierzbicka. Wierzbicka proposes definitions of word meaning in a style and format radically different from the traditional ones. However, I am puzzled about their linguistic status, and I have doubts concerning their appropriateness as models for dictionary definitions. On Polylexy89 With regard to the linguistic status of Wierzbicka's definitions, I find it difficult to grasp exactly what it is that they are supposed to represent. How do they relate to necessary and sufficient criteria of the classical variety, and to representations of prototypes? The examples offered seem to fall uncomfortably between the two, in that the components are not all logically necessary, but on the other hand they are insufficient to portray the prototype adequately. Take the example oipunüh. Wierzbicka argues that suffering on the part of the punished person should not be included in the definition because people who are punished do not always suffer; all that is necessary is that a desire to cause suffering should be present in the punisher. This suggests that the definition is intended to contain only logically necessary components. However, if this is the aim, it is not achieved, because not all the components in her definition ofpunish are logically necessary. Here is Wierzbicka's schema: X punished Y [for Z]. = (a)[Y did Z] (b)X thought something like this: (c)Y did something bad [Z] (d)I want Y to feel something bad because of this (e)it will be good ifY feels something bad because of this (f) it will be good if I do something to Y because of this (g)X did something to Y because of this Component (e), for instance, indicates that the punisher desires to cause suffering in the culprit. But this is not necessary: it is quite possible , and indeed morally correct, for a punisher to desire merely the hoped-for results of punishment, i.e., non-recurrence of the wrongdoing , or the upholding ofjustice. The punisher may actually hate having to cause suffering to achieve these aims. McKeown (1991) has shown that older children acquire vocabulary more successfully from definitions that present the prototype than from those that do not, and this might not be a bad aim for general dictionary definitions. A disadvantage ofsuch an approach is that it could lead to an open-ended commitment and definitions of indefinite length, because the features characterizing a prototype do not always have a clear cut-off point. One way of restricting the features might be to include only those that exhibit what in Cruse (1986) was called canonical necessity. Canonically necessary features are those necessary for well-formedness. For...