Training Analysis

Data from each training run and each participant were separately analyzed at a first level analysis using FSL [26]. Visually distinct and visually similar correct trials were separately modeled as two predictors. Each stimulus onset time was convolved with a hemodynamic response function and entered into a general linear model with their temporal derivatives to estimate beta weights. Contrasts of interests were set that tested for differential activation to the two types of categorization trials. Even after performance becomes equated, we focused on data from runs 4, 5, and 6 to explore whether participants engage distinct processes for visually similar vs. visually distinct categories (the runs after subjects could perform the task with proficiency for all three categories: see Figure 4). Data from the three runs within each subject were combined at a second level while using fixed-effects analysis. Data across participants were then combined at the third level using a random-effects analysis (FLAME 1). Figure 7 depicts regions that were more engaged during visually distinct trials over visually similar categorization trials (blue), and vice-versa (red). Individual voxels were considered to be active when reaching a *Z* > 1.8 and surviving a whole-brain cluster size threshold set at *p* < 0.05 [27]. This threshold was used based on the exploratory nature of our small sample pilot experiment, and the reported results were interpreted for the purposes of motivating Experiment 2.

**Figure 7.** Univariate contrasts of visually similar > visually distinct (Red) and visually distinct > visually similar (Blue) during training displayed in (**a**) sagittal, (**b**) coronal, and (**c**) axial views. Red: dorsal lateral and inferior frontal areas along with hippocampus were engaged significantly more during rule application compared to perceptual similarity analysis. Blue: Fusiform gyrus and lateral occipital cortex were engaged significantly more during perceptual similarity analysis compared to rule application.

The superior and inferior frontal gyri were engaged significantly more on visually similar categories when compared to the visually distinct category (red clusters, Figure 7). The right hippocampus, a region associated with declarative recall, was also engaged during the classification of visual similar categories, which is consistent with prior work indicating the role of hippocampus in rule-based categorization [28]. During classification to the visually distinct category, the lateral occipital cortex and fusiform gyrus were engaged significantly more as compared to visually similar categories (blue clusters, Figure 7), consistent with what might be expected for similarity-based categorization. Tables 1 and 2 illustrate a summary of the top 11 regions associated with each condition.


**Table 1.** Cluster location and size for similar > distinct contrast in blocks 4, 5, and 6.

**Table 2.** Cluster location and size for distinct > similar contrast in blocks 4, 5, and 6.

