Leech et al. present a connectionist model of analogical reasoning based on relation priming, rather than on explicit structure-mapping processes. Their core idea is that priming is itself a mechanism for producing analogy, and from it ultimately emerges the relation that is critical for establishing the similarity between a pair of terms in one domain and a pair of terms in a second domain. This claim is compatible with recent work in language processing that emphasizes the role of “low-level” priming in the development of semantic representations. This is most apparent in work on dialogue, in which interlocutors prime each other to produce equivalent situation models that form the basis of mutual understanding (Pickering & Garrod Reference Pickering and Garrod2004). For example, interlocutors tend to repeat each other's choice of reference frames or ways of interpreting complex arrays (Garrod & Anderson Reference Garrod and Anderson1987; Schober Reference Schober1993). Clearly, alignment of analogical structures constitutes an important part of such situation models.
Critically, Pickering and Garrod's (2004) framework suggests that alignment takes place at many linguistic levels, and that repetition at low levels such as words enhances alignment at higher levels, such as the situation model. It follows that lexical repetition should enhance relational priming, and therefore analogical reasoning. Raffray et al. (Reference Raffray, Pickering and Branigan2007) directly addressed the issue of the effects of lexical repetition (of the head or modifier) on relation priming of noun-noun combinations such as dog scarf. Three expression-picture matching experiments investigated whether relation priming occurred in the context of head repetition, modifier repetition, or both, and allowed direct comparison of the effects of head and modifier repetition. Results showed that participants were more likely to interpret dog scarf as a scarf decorated with a picture of a dog (i.e., dog DESCRIBES scarf) than as a scarf worn by a dog (i.e., dog POSSESSES scarf) after interpreting another expression involving the description relation rather than the possession relation; but the priming was greater when one term was repeated (e.g., dog T-shirt or rabbit scarf) than if neither was repeated (e.g., rabbit T-shirt). In sum, while conceptual relations were independently primed, the level of activation that a given relation received was enhanced where there was repetition of lexical items between prime and target.
We propose that such “lexical boost” effects, similar to those found in syntactic priming studies (Pickering & Branigan Reference Pickering and Branigan1998), mean that priming of analogical relations should be enhanced by any repetition of terms. In Goswami and Brown (Reference Goswami and Brown1989), the participant infers that lemon is to cut lemon as bread is to cut bread. Importantly, the concept of a lexical boost within analogical reasoning only makes sense in the context of two- (or more) place relations. That is, to get a lexical boost we would need to consider analogies such as boy & ball is to boy kicks ball as man & stone is to man kicks stone. In this case, the lexical boost predicts that participants should find it easier to resolve analogies containing repeated terms, such as boy & ball is to boy kicks ball as man & ball is to man kicks ball, or similarly boy & ball is to boy kicks ball as boy & stone is to boy kicks stone. For more complex analogies, the prediction is that any repetition of concepts will enhance analogy. To take the authors' example, it should be easier to draw the analogy from World War II to World War I than to the Gulf War, because more of the objects (e.g., Germany) are repeated (see Table 3 of the target article, sect. 4.1.2). Whereas the analogy between Churchill orders_attack_of Germany and Bush orders_attack_of Iraq involves different Object 1s and Object 2s, the analogy between Churchill orders_attack_of Germany and Lloyd George orders_attack_of Germany involves different Object 1s but the same Object 2. If such analogy works like the priming effects we have discussed, then lexical repetition should facilitate analogical reasoning.
There is also evidence for a semantic boost to syntactic priming (Cleland & Pickering Reference Cleland and Pickering2003), so that priming is stronger when terms are semantically related than when they are not. For example, participants are more likely to describe a red sheep as The sheep that is red after hearing The goat that is red than after hearing The door that is red. It might similarly be the case that priming of analogical relations is enhanced by the inclusion of semantically related terms. That is, boy & stone is to boy kicks stone as man & pebble is to man kicks pebble might be easier to process than an analogy that contains semantically unrelated terms, such as boy & stone is to boy kicks stone as man & ball is to man kicks ball. Such effects should affect both the speed and the likelihood of obtaining a particular analogy.
Pickering and Garrod's (2004) model of dialogue assumes that processes by which interlocutors align their models of the situation are largely the result of automatic priming. In particular, repetition at one level of representation enhances repetition at other levels. For example, Branigan et al. (Reference Branigan, Pickering and Cleland2000) showed that interlocutors are more likely to repeat each other's grammatical choices if they also repeat each other's choice of verbs. More generally, we assume that if two people start to use the same words, they start to take on board the same conceptualizations. One of the functions of analogical reasoning is to assist in the process by which interlocutors end up with equivalent situation models, and therefore we predict that lexical and other repetition will enhance this process. Therefore analogical reasoning can be seen as part of the mechanism of alignment and will be affected by the processes affecting priming.
