Problems with prototype theory
As we noted at the outset of this chapter, it has been argued that prototype theory is inadequate as a theory of knowledge representation. In this section, we briefly review some of the objections, as well as consider whether Rosch and her colleagues intended their findings to be interpreted directly as a model of knowledge representation.
We begin with a number of criticisms discussed by Laurence and Margolis (1999), who present a survey of the criticisms that have been levelled against prototype theory in the literature. The first criticism, which Laurence and Margolis describe as the problem of prototypical primes, concerns the study of ODD NUMBERS that we discussed earlier (Amstrong et al. 1983). Recall that this study found that even a ‘classical category’ of this nature exhibits typicality effects. Armstrong et al. argue that this poses potentially serious problems for Prototype Theory since such effects are not predicted for classical categories.
The second criticism that Laurence and Margolis identify is that, like the classical theory, prototype theory also suffers from the problem of ignorance and error: it fails to explain how we can possess a concept while not knowing or being mistaken about its properties. The basis of this criticism is that a concept with prototype structure might incorrectly include an instance that is not in fact a member of that category. The example that Laurence and Margolis use to illustrate this point is that of a prototypical GRANDMOTHER, who is elderly with grey hair and glasses. According to this model, any elderly grey-haired woman with glasses might be incorrectly predicted to be a member of this category. Conversely, concepts with a prototype structure may incorrectly exclude instances that fail to display any of the attributes that characterise the prototype (for example, a cat is still a cat without having any of the prototypical attributes of a cat).
The third criticism that Laurence and Margolis discuss is called the missing prototypes problem: the fact that it is not possible to describe a prototype for some categories. These categories include ‘unsubstantiated’ (non-existent) categories like US MONARCH and heterogeneous categories like OBJECTS THAT WEIGH MORE THAN A GRAM. In other words, the fact that we can describe and under stand such categories suggests that they have meaning, yet prototype theory as a model of knowledge representation fails to account for such categories.
Finally, Laurence and Margolis describe the problem of compositionality, which was put forward by Fodor and Lepore (1996). This is the criticism that prototype theory provides no adequate explanation for the fact that complex categories do not reflect prototypical features of the concepts that con tribute to them. To illustrate this point, Laurence and Margolis cite Fodor and Lepore’s example of PET FISH. If a prototypical PET is fluffy and affectionate and a prototypical FISH is grey in colour and medium-sized (like a mackerel), this does not predict that a prototypical PET FISH is small and orange rather than medium, grey, fluffy and affectionate.
As this brief discussion of the criticisms levelled against prototype theory indicates, Rosch’s findings have often been interpreted directly as a theory of knowledge representation (a theory about the structure of categories as they are represented in our minds). Indeed, Rosch explored this idea in her early work (albeit rather speculatively). Consider the following passage:
[A prototype can be thought of] as the abstract representation of a category, or as those category members to which subjects compare items when judging category membership, or as the internal structure of the category defined by subjects’ judgments of the degree to which members fit their ‘idea’ or ‘image’ of the category. (Rosch and Mervis 1975: 575)
Rosch retreats from this position in her later writings. As she later makes explicit, ‘The fact that prototypicality is reliably rated and is correlated with category structure does not have clear implications for particular processing models nor for a theory of cognitive representations of categories’ (Rosch 1978: 261). In other words, while typicality effects are ‘real’ in the sense that they are empirical findings, it does not follow that these findings can be directly ‘translated’ into a theory of how categories are represented in the human mind. In other words, experiments that investigate typicality effects only investigate the categorisation judgements that people make rather than the cognitive representations that give rise to these judgements.
This point is central to Lakoff’s (1987) discussion of Rosch’s findings. Lakoff argues that it is mistaken to equate prototype or typicality effects with cognitive representations. Rather, typicality effects are ‘surface phenomena’. This means that they are a consequence of complex mental models that combine to give rise to typicality effects in a number of ways. Typicality effects might therefore be described in intuitive terms as a superficial ‘symptom’ of the way our minds work, rather than a direct reflection of cognitive organisation. Lakoff (1987) therefore attempts to develop a theory of cognitive models that might plausibly explain the typicality effects uncovered by Rosch and her colleagues. As we will see in the next section, Lakoff’s theory of cognitive models avoids the problems that we summarised above which follow from assuming Prototype Theory as a model of knowledge representation.
