Solving four word analogy problems : the role of specificity and inclusiveness

The present work examined subjects' performance on eight types of four word analogy problems. Two critical dimensions distinguish among these analogy types: specificity and inclusiveness. Whole-part analogies such as hand : palm as foot : sole (read hand is to palm as foot is to sole) are specific because the association appearing in the two word pairs consist of spatial/functional relationships which are highly similar to each other. In contrast, analogies such as car : wheel as boat : mast are nonspecific because they use whole-part associations which are less similar to each other. Analogies are inclusive if they use relatively direct associations, as in the whole-part association illustrated by car : wheel. In contrast, noninclusive analogies require additional inferences between words, as illustrated in the part-part association bumper: wheel, which requires the object car to be inferred. Responses from undergraduate university subjects show that both inclusive and specific analogy problems were solved more quickly than their noninclusive and nonspecific counterparts, respectively. Experiment 1 illustrated these specificity and inclusiveness effects both in a recognition (multiple choice) paradigm, and a recall paradigm where subjects spoke their own answer choices aloud. Subsequent experiments were performed to examine the role of the association types and the role of word attributes in subjects' processing of these analogy problems. Experiment 2 attempted to prime subjects with the association type used in each block of analogy problems, but showed a very modest effect on solution latencies. In Experiment 3 reordering the words within analogy problems unexpectedly increased the latencies for many problems, apparently because different words appeared in the third word positions within them. Experiments 4 and 5 focussed directly on the study of specificity. Experiment 4 showed that the processing benefit found for specific analogies is due to the close match of word attributes between word pairs, not due to the attributes of the particular words used. Experiment 5 manipulated the taxonomic similarity of the subject matter addressed by the two pairs of words, and found that the use of word pairs from more taxonomically distant subject areas increased solution latencies for some analogy types. Experiment 6 required subjects to group analogy problems into categories they defined. This procedure validated six of the eight analogy types used in this thesis; the specificity distinction was not evident among the groups of problems formed by subjects. The discussion of these results supports a theoretical model of problem solving four word analogies which incorporates a stage-like, componential processing for nonspecific types, and a faster, more automatic processing for specific types. The discussions of empirical and theoretical work in this thesis also focussed more widely on its relevance to more practical uses of analogies in problem solving.Arts, Faculty ofPsychology, Department ofGraduat