Detecting and processing relations is a key task used to assess fluid intelligence (Gf), applied in such hallmark intelligence tests as Raven’s and Cattell’s. Previous studies identified the maximum complexity of relations (usually, five or six relation dimensions coped with at the same time) above which performance reaches floor and is no longer indicative of intelligence level. However, no study has tested the maximum simplicity of relations which yet strongly load on Gf. Knowing how simple relations yet involve Gf can shed light on the cognitive processes and mechanisms underlying this ability. In the series of studies, another hallmark Gf test, the transitive reasoning task, was gradually simplified in order to see how simple variant would no longer correlate with the Gf factor assessed with several standard intelligence tests. Originally, the task presents three premises, each binding two symbols with a transitive relation (e.g., AD, CB valid?). Such a task loads on Gf comparably to typical intelligence tests. In the first study (N=318), this task required to decide either about the relation between two unbound symbols, either separated by one (is AA valid?), or about two symbols that were already bound, but in the reversed order (is DGf.
Surprisingly, all the three task variants correlated with the Gf factor with the same strength, r = .55, exactly as the benchmark Gf tests. However, it was possible that even during premise reversing the full-blown model of the linear order was constructed, and thus even this variant involved Gf. So, the next study (N=181) applied only reversed premises as well as introduced an even simpler variant – exact repetitions of a premise (is B>D valid?). But still the reversed and repeated premise variants strongly correlated with the Gf factor, r = .59 and r = .68, respectively. In order to reject another possibility that sheer reversing might have turned the task into the Gf test, the third study (N=171) applied solely the exact premise repetition, and made sure that the instruction simply insisted on comparing the premises and response options on the screen. However, no drop in the Gf correlation was observed, r = .65. The fourth study (N=173) rejected a possibility that the Gf correlation resulted from processing of the relation term (< or >). Participants simply compared the spatial order of symbols (e.g., A C, B D, C D), and had to select the option that matched that order (B D instead of D B). Again, the Gf correlation was substantial, r = .50, and just slightly weaker than those for the remaining Gf tests. Only in the fifth study (N=110), when just several symbols were presented within the matrix, and two identical neighboring symbols were to be detected, the Gf correlation dropped to non-significant, r = .04. Concluding, complex relations are not necessary to capture Gf, and even presumably trivial relations, above sheer identity, strongly involve Gf, for reasons yet to be understood.
Cognition and Attention , Measurement and Psychometrics , Reasoning and Rationality
