The purpose of this study is to use the rule space procedure to develop a general set of cognitive and linguistic attributes that account for performance on the Analogy Section of the SAT I Verbal. The overall research strategy was to develop a set of cognitive and linguistic attributes on one form of the Analogy Section, the experimental form, and then to examine the generalizability of these by applying the same set of attributes to a different form of the test, the generalizability form. Results produced a set of 13 cognitive and linguistic attributes, with 6 interaction attributes, that were able to classify 97% of test-takers into their latent knowledge states on the experimental form; the same set of attributes were able to classify 90% of test- takers into their latent knowledge states on the generalizability form. The analysis provides a strong indication of the important determinants of success on the SAT Analogy task, and provides evidence that these are relatively stable across forms.
