Structure mapping in analogy and similarity.

ions Keil, 1989 ; Rips, 1989) . For example, bats have the perceptual and behavioral characteristics of birds (they are similar to birds in this sense), but they are classified as mammals, because of important (though nonobvious) properties, such as giving birth to live young. On the basis of examples like this, similarity's role in categorization has been challenged ; it has been argued that category membership judgments are theory based rather than similarity based (Keil, 1989 ; Murphy & Medin, 1985) . The process of alignment and mapping points the way to a reconciliation of similarity-based and theorybased accounts (see also Goldstone, 1994a) . If we focus purely on perceptual similarity among objects, we are led to conclude that bats should be categorized with birds . On this view, theory-based knowledge (such as why bats are mammals) must intervene from elsewhere to overrule this assignment . However, if the similarity computation is assumed to be that ofstructural alignment, then the similarity between two instances will be based riot only on object-level commonalities but also on common relations such as common causal relations and common origins . Assuming that our representations include information about theory-based relations, such as that bats bear live young, as well as information about features, then the schism between similarity-based and theory-based categorization may be more apparent than real . Developmentally, if we assume that theoretical knowledge is acquired gradually, this view would account for the characteristic-to-defining shift (Keil & Batterman, 1984) in children's interpretations of word meaning from local object features (e.g ., a taxi is bright yellow and has a checkered sign) to deeper relational commonalities (e.g ., a taxi is a vehicle that may be hired to transport people) . Choice and decision . Structural alignment also sheds light on the processes underlying choice behavior. Medin, Goldstone, and Markman (1995) reviewed paral lels between phenomena in decision processing and phenomena in comparison processing that suggest an important role for structural alignment in decision making . Structural alignment influences which features to pay attention to in choice options . Research suggests that alignable differences are given more weight in choice situations than are nonalignable differences (Lindemann & Markman, 1996 ; Markman & Medin, 1995 ; Slovic & MacPhillamy, 1974) . For example, Markman and Medin (1995) asked participants to choose between video games and to justify their choices. Their justifications were more likely to contain alignable differences than nonalignable differences . As another example, Kahneman and Z'versky (1984) described to participants a hypothetical store in which a jacket could be bought for $125 and a calculator for $15 . They offered participants the opportunity to go to another store and save $5 on the total purchase. Participants who were offered ajacket for $125 and a calculator for $10 were more willing to make the effort to go to another store than those offered a jacket for $120 and a calculator for $15 . Even though the monetary reward for going to the other store was the same for both groups, participants were influenced by the alignable difference .