Explanationbased Approach To Concepts

• Concepts can have attributes.

• But they also have relations between these attributes, which form explanatory connections between the attributes (e.g., that wings, feathers, and fight bones enable birds to fly).

• Concepts are not necessarily stored as static knowledge in memory, but may be dynamically constructed in working memory using attribute-definitions and other background knowledge (e.g., causal knowledge); hence, the phenomenon of ad hoc categories.

• Concept coherence and naturalness emerge from the underlying theoretical knowledge of concepts, not from similarity alone.

• Context effects on concept representations emerge from the way the concepts come to be constructed in working memory using background knowledge (e.g., a lifting or playing sentential-context for a piano, results in weight, and musical attributes becoming salient, respectively).

book-learned, scientific knowledge certainly tainly contains theories" (p. 290). Murphy and Medin, therefore, argue that even though similarity i is important it is not sufficient to determine which concepts will be coherent or meaningful. These arguments have informed a newly emergent view of concepts which has been termed the knowled e-based or explanation-based view.

The explanation-based view of concepts sees concepts as involving more than attribute-lists; concepts also contain causal and other background knowledge that might be represented by schemata (see Chapter 9). For example, living things with wings, feathers, and light bones are se en as forming a natural category because they re, according to a certain theory, manifestations of a single, genetic code; the category coheres because we have a theory that explains the co-occurrence currence of these attributes. Miller and Johnson-Laird (1976) were among the first to propose concept representations involved schematic knowledge, although others have made similar proposals (Cohen & Murphy, 1984; Keil, 1989; Lakoff, 1987). For the most part, this view of concep ts has been marked by several general statements of the view rather than many concrete realisation of it; see Lakoff (1987) on idealised cognitive models, Johnson-Laird's (1983) mental mode s account, and Medin and Ortony's (1989) psychological essentialism. Some of the main proposals of the view are shown in Panel 10.5.

Evidence for explanation-based views

There are several sources of evidence for explanation-based views. Some studies have shown that there is a dissociation between similarity and categorisation judgements, thus showing that similarity could not be the sole mechanism behind categorisation (Rips, 1989a). Other studies have shown how background knowledge (either causal or specific knowledge) can influence the application and learning of categories (see Ahn, Brewer, & Mooney, 1992; Malt, 1994; Medin, Wattenmaker, & Hampson, 1987; Pazzani, 1991; Wisniewski & Medin, 1994).

Rips (1989) has shown a dissociation between similarity judgements and categorisation in a study where one group of subjects were asked whether an object five inches in diameter was more likely to be a coin or a pizza, and a second group were given the same information and asked to judge the similarity of the object to either the coin or the pizza. Although the object's size was roughly midway between a large coin and a small pizza (as determined by prior norms), subjects in the categorisation group tended to categorise it as a pizza. However, the similarity group judged the object to be more similar to the coin. If categorisation was based on similarity alone, subjects' judgements in both groups should have tallied. The fact that they did not indicates that some other variable was at work, namely knowledge (or a theory) about the variability of the sizes of the objects in question. Coins have a size that is mandated by law, whereas pizzas can vary greatly in size (as we have seen earlier there is an alternative exemplar-based account of this effect).

Further evidence comes from Medin et al. (1987) who have shown that conceptual knowledge seems to drive the application of a family resemblance strategy in concept sorting. Recall that within the prototype view the typicality of a concept member is closely related to the family resemblance score for that instance; that is, the score that reflects the extent to which the instance's attributes are the same as those of other instances of the category. Medin et al. (1987) found that, in a sorting task, subjects persisted in sorting on the basis of a single dimension instead of using many dimensions, as a family resemblance account would predict. Medin et al. revealed that subjects abandoned this uni-dimensional sorting strategy in favour of a strategy that used several dimensions when the item had causally related, correlated properties. That is, when subjects were given conceptual knowledge that made interproperty relationships more salient, family resemblance sorting became very common. The moral being that correlated-attribute dimensions are really only used in sorting when there is some background knowledge or theory that connects them together.

Other studies have shown how background knowledge influences the categorisation process. One of the earliest findings in concept formation was that conjunctive concepts were easier to learn than disjunctive concepts (see Bruner et al., 1956). So, for example, it is easier for people to learn a concept called DRAF consisting of the conjoined features—black and round and furry—than when its features are disjunctive — "black OR round OR furry". Pazzani (1991) has demonstrated a reversal of this phenomenon when the disjunctive concept is consistent with background knowledge. In this study, groups of subjects were shown pictures of people (adults or children) carrying out actions (stretching or dipping in water) on balloons of different colours and sizes. One set of instructions required subjects to determine whether a given stimulus situation (e.g., a child dipping a large, yellow balloon in water) was an alpha situation. Another set of instructions required subjects to predict whether the balloon would inflate after the stimulus event. Groups receiving either of these instructions had to learn either a conjunctive concept (size-small and balloon-yellow) or a disjunctive concept (age-adult OR action-stretching-balloon). Pazzani established that most people know that stretching a balloon makes it easier to inflate and that adults can inflate balloons more easily than children (but note that this knowledge does not correspond directly to the disjunctive definition subjects had to learn).

Pazzani found that the alpha groups found the conjunctive concept easier to acquire than the disjunctive concept (see Figure 10.3). As in previous research, this result occurred because the background knowledge about inflating balloons was irrelevant to learning the alpha categorisation. However, in conditions receiving the instructions to predict inflation of the balloon the opposite was found; the group learning the disjunctive concept found it easier to learn than the conjunctive concept. This was due to the fact that subjects' background knowledge informed the formation of the disjunctive concept, but did not support the learning of the conjunctive-inflate concept (see also Pazzani, 1993).

In a similar vein, Wisniewski and Medin (1994) showed subjects children's drawings of people, asking them to form a category rule to describe the set, which could be extended to a new instance of the category. They introduced background knowledge to the task by labelling the pictures meaningfully (i.e., these are drawings by creative or non-creative children) or neutrally (i.e., these drawings were done by Group 1 or Group 2). Subjects given the meaningful labels categorised the drawings in a very different way from those given neutral labels; they tended to use more abstract features of the drawing (e.g., "action", "true to life", "bodily expression") rather than concrete perceptual features (e.g., "arms at the side", "pockets", "curly hair"). In the neutral group, the drawings were predominantly classified according to differences in concrete features. In short, the former group brought intuitive theories to the task, which extracted a very different

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