A six-process theory of analogical reasoning was tested by administering verbal analogy items to students in grades 3 through college. The items were classified according to five verbal relations: synonyms, antonyms, functional, linear ordering, and class membership. A new method of componential analysis that does not require precueing was used to investigate predictive efficiencies of the six processes. Response latencies and error rates were analyzed. Analogical reasoning ability was found to improve with grade level, and subsets of the theory successfully accounted for response latency and error data at various grade levels. Mapping and application were important variables in grades 3 and 6; encoding and index of association for the stem were the most important processes in ninth grade and undergraduate subjects. Performance also varied with type of verbal relation of the item. Functional and antonymous relations were easier to process than linear ordering, synonymous, and category name relations. The new method of breaking down component processes was found to be as successful as precueing at the adult level (the only level at which precueing has been used). Implications for intelligence testing are discussed. (Author/RD)