e. those with the highest t/Z values for the Cyclopamine chemical structure words vs. baseline contrast. As this comparison (words vs. baseline) is orthogonal to both of the variables investigated (lexical category, abstractness), the strategy applied for selecting ROIs follows recent recommendations to avoid “double dipping” ( Kriegeskorte, Simmons, Bellgowan & Baker, 2009). In this data-driven analysis, average activation values within each of these 2 mm-radius spheres for each subject and each of the four word categories were entered into a repeated-measures ANOVA with the factors ROI x lexical category (2) × semantics/abstractness

(2). Note that, because 2 × 2 × 2 mm voxels, 8 mm smoothing kernel and 2 mm ROI radius were chosen, the half maximum width of each ROI was 12 mm. This allowed us to keep overlap between ROIs to a minimum while at the same time compensating for some of the spatial variance caused by the projection of individual brains to the averaged MNI template. Where appropriate, Huynh–Feldt correction was applied to correct for sphericity violations. In this case, epsilon values and corrected p values are reported throughout. Whereas psycholinguistic properties Anti-infection Compound Library solubility dmso were matched between word groups (see Methods, Appendix

B), results of the semantic rating study executed prior to the fMRI experiment revealed significant differences in the semantic variables of imageability, arousal, action-relatedness, concreteness, visual-relatedness, colour-relatedness and form-relatedness (see Appendix B). For all of these features, 2-way ANOVAs revealed significant interaction effects and, in most cases, additional main effects. The interactions of all object-related features, including concreteness, imageability, form- and visual-relatedness, showed, as expected, highest values for concrete nouns towering over

all other word groups. For arousal and action-relatedness, which both reflect semantic action features, concrete verbs achieved the highest ratings and concrete nouns the lowest. In addition, object-related semantic ratings were TCL higher for nouns than for verbs and higher for concrete items than for abstract ones; with regard to action-relatedness, verbs dominated over nouns and, again, concrete over abstract items. Statistical tests for word groups, including interactions and main effects, are displayed in Appendix B. Pairwise comparisons between stimulus groups showed that the abstract noun category was indeed significantly less imageable (t(78) = −14.028, p < .001), less concrete (t(78) = −16.812, p < .001), less related to visual objects (t(78) = −15.145, p < .001), and less form/shape-related (t(78) = −10.443, p < .001) than concrete nouns. Likewise, abstract verbs were significantly less imageable (t(78) = −8.613, p < .001), less concrete (t(78), and less action-related (t(78) = −3.018, p < .005) than concrete verbs.