Generalizability of component processes in intelligence as revealed by latency measures

Abstract

In this study, we explored the relationships between the latencies of theoretically corresponding and noncorresponding component processes across a diverse task universe derived from Guilford's Structure of Intellect (SI) model. Seventy-two participants were presented a task universe that consisted of six componential tasks with a semantic or figural content and with an emphasis on either cognition, inductive reasoning, or evaluation operations. Latencies of 18 component processes were estimated from these componential tasks by means of multiple regression analysis and identified as encoding, searching, inference, mapping, application, comparison, and justification components. Canonical correlation analysis indicated that on a componential level the task universe showed to a certain extent diversity, suggesting that there was only a small bias in favor of finding convergent relationships. Multidimensional scaling and hierarchical cluster analyses revealed a three-cluster solution of component processes, which were tentatively interpreted as an encoding, a reasoning, and an evaluation cluster. These results provide initial evidence that theoretically corresponding component processes were generalizable (convergent validity) and theoretically noncorresponding component processes were not generalizable (discriminant validity) across the given task universe.