The Exemplarbased View Of Concepts

• Categories are made up of a collection of instances or exemplars rather than any abstract description of these instances (e.g., a prototype summary description).

• instances are grouped relative to one another by some similarity metric.

• Categorisation and other phenomena are explained by a mechanism that retrieves instances from memory given a particular cue.

• When exact matches are not found in memory the nearest neighbour to the cue is usually retrieved.

given situation (see Panel 10.4) (Brooks, 1978; Erickson & Kruschke, 1998; Estes, 1976, 1993; Hintzman, 1986; Medin, 1975, 1976; Medin & Shaffer, 1978; Nosfosky, 1986, 1988, 1991; Nosofsky, Palmeri, & McKinley, 1994; Shin & Nosofsky, 1992).

As such, the exemplar-based view paints a very different picture of categories from the prototype view. Instead of there being some abstracted description of a bird which acts as a central prototype, the picture is one of a memory that stores millions of specific instances. So, instead of having a prototype for bird that is a list of all the characteristic features abstracted away from members of this category (e.g., has-wings, flies, etc.), one just has a store of all the instances of birds you have encountered in the past (e.g., the robin you see every morning, a crow, a chough, a penguin, etc). As we shall see, all the effects attributed to prototypes can be dealt with by this sort of account depending on what instance(s) come to mind in a specific context.

Evidence for the exemplar view

Much of the evidence that specifically seemed to support the prototype view can be explained by the exemplar view. Consider the effects of faster categorisation judgements for some members of a category than others. When asked "Is a robin a bird?" you can answer "yes" much faster than when asked "Is a penguin a bird?". Given that you have encountered many robins in the past, there are likely to be a lot more stored instances of robins than penguins. Therefore, a robin instance will be retrieved from memory much faster than a penguin instance, thus giving rise to the differences in judgement times. Similarly, typicality ratings are said to reflect the underlying pattern of instances in the category; a robin is a more typical instance of a bird than a penguin because there are many more stored instances of robins than penguins. Typicality gradients can be accounted for in similar ways.

The exemplar-based account is also more consistent with the recent research on prediction and conceptual instability. Recall that the research on prediction was really all about comparing one classified target instance with other instances of the category to make appropriate predictions about features of that target instance (e.g., Heit, 1992; Murphy & Ross, 1994). Similarly, effects like those involving changes in perspective and ad hoc categories are easier to explain in the context of a theory where one has instances that can be regrouped in different ways to meet the demands of specific task situation.

There is other more specific evidence that supports the exemplar view in opposition to the prototype view. The exemplar view preserves the variability of instances in the category, whereas a prototype is a type of average over the instances of the category that usually exclude this variability information. Rips and Collins (1993; Rips, 1989a) showed that this variability information could influence classification. Their task involved the categories pizzas and rulers; most pizzas are 12 inches in size but they vary a lot (i.e., anything from 2 inches to 30 inches in width) and rulers are also 12 inches in size but are much less variable (i.e., most of the time they are 12 inches long). Subjects were asked to make a judgement about a new object 19 inches in size, as to whether it was a pizza or a ruler. If people had a prototype then this judgement should reveal a 50-50 split between pizza and ruler, because the prototype average would be 12 inches for both. However, if the variability was used by people, then they should always say that the object was a pizza because it is much more likely to vary in size than a ruler.

The exemplar-based approach also preserves correlational information between instances of a category in ways that a prototype does not. Again, it has been found repeatedly that people use such knowledge in category learning and classification judgements (see Medin, Altom, Edelson, & Freko, 1982; Nosofsky, 1991).

Evidence against the exemplar view

For the most part, the exemplar view does better than the prototype view on many points. But, like the prototype view, there are some effects that it finds hard to explain. Typicality and category judgements should always co-vary; there should be no dissociation between the two. However, dissociations have been found. As we have seen Armstrong et al. (1983) have shown that people could make typicality judgements even when it was known the category had defining attributes. It is troublesome to the prototype and exemplar approaches to find that the causal link between the definition of a concept and the typicality measures can be called into question (but see Hampton, 1995). Furthermore, like the prototype view, the exemplar view depends on similarity. Hence, difficulties that arise in the treatment of similarity in prototype theory, tend to transfer to exemplar theories. Finally, these theories do not cope easily with class inclusion questions. For example, when people answer questions about the truth of a statement like "All birds are creatures" they seem to rely on general knowledge rather than specific examples. Yet the exemplar view has no good account of how such abstract knowledge comes into being.

Stop Anxiety Attacks

Stop Anxiety Attacks

Here's How You Could End Anxiety and Panic Attacks For Good Prevent Anxiety in Your Golden Years Without Harmful Prescription Drugs. If You Give Me 15 minutes, I Will Show You a Breakthrough That Will Change The Way You Think About Anxiety and Panic Attacks Forever! If you are still suffering because your doctor can't help you, here's some great news...!

Get My Free Ebook


Post a comment