Crystallized Intelligence, Fluid Intelligence, and Need for Cognition: Their Longitudinal Relations in Adolescence

Abstract

Investment theory and related theoretical approaches suggest a dynamic interplay between crystallized intelligence, fluid intelligence, and investment traits like need for cognition. Although cross-sectional studies have found positive correlations between these constructs, longitudinal research testing all of their relations over time is scarce. In our pre-registered longitudinal study, we examined whether initial levels of crystallized intelligence, fluid intelligence, and need for cognition predicted changes in each other. We analyzed data from 341 German students in grades 7-9 who were assessed twice, one year apart. Using multi-process latent change score models, we found that changes in fluid intelligence were positively predicted by prior need for cognition, and changes in need for cognition were positively predicted by prior fluid intelligence. Changes in crystallized intelligence were not significantly predicted by prior Gf, prior NFC, or their interaction, contrary to theoretical assumptions. This pattern of results was largely replicated in a model including all constructs simultaneously. Our findings support the notion that intelligence and investment traits, particularly need for cognition, positively interact during cognitive development, but this interplay was unexpectedly limited to Gf.

Keywords: crystallized intelligence; fluid intelligence; investment theory; investment traits; need for cognition.

Figure 1

Bidirectional Latent Change Score Analyses from Step 1. Note. *** < .001. ** < .01.

Figure 2

Latent Change Score Analyses from Step 2. Note. *** < .001. ** < .01. NFC = Need for cognition. Gf = Fluid intelligence. Gc = crystallized intelligence.

Figure 3

Latent Change Score Analyses from Step 3. Note. *** < .001. ** < .01. NFC= Need for cognition. Gf = Fluid intelligence. Gc = crystallized intelligence. Although the model estimation terminated normally according to Mplus, two Heywood cases appeared after the inclusion of the interaction term between Gf_t1 and NFC_t1. The prediction paths from Gf_t1 (β = 1.00) and NFC_T1 (β = 2.20) to Gc_LCS were greater than 1, although they were not significant.