This analysis indicated that there was a marginally significant effect of condition (β = −.36, SE = .18, z = −1.94, p = .052), where participants were slightly more accurate in the phonological condition compared with the semantic condition.
We should note that a Wald test revealed that the parameter estimates for the fMRI activation variable were not statistically different between the phonological and semantic models, and only trending toward significance comparing the semantic and age models ( p = .07; fMRI parameter estimates: phonological = −11.811, semantic = −16.300, age = −0.7369) (Fig. 6 ).