74 and 0.35 for monkey M1 and M2, respectively (M1: p = 0.001; M2: p = 0.15; Fisher Z test; see Figure S4). Across all 34 3D-structure-selective sites, 22 sites (65%) contained at least one electrode position for which the MUA was significantly 3D-structure-selective buy MK0683 at each position in depth (p < 0.05, t test). Ten (75%)

of the remaining 12 sites contained at least one electrode position for which the MUA was significantly 3D-structure-selective for two positions in depth (p < 0.05, t test). In none of the 3D-structure-selective sites did we observe a significant reversal in structure preference at any position-in-depth (p > 0.05, t test). Hence, all 3D-structure-selective sites were characterized by only one 3D-structure preference. We tested whether stimulation in clusters containing MUA positions with significant selectivity for all positions-in-depth (putative completely invariant sites) caused larger microstimulation effects compared to stimulation in clusters with MUA positions that did not display significant structure selectivity at each position in depth (putative incompletely invariant sites). Stimulation in clusters with completely

invariant MUA positions caused significantly larger microstimulation Baf-A1 in vivo effects in monkey M1 (mean psychometric shift of 45% versus 19%; p = 0.005) but not in monkey M2 (mean psychometric shift of 12% versus 9%; p > 0.05), although a trend was and present. Given that we probably did not only stimulate completely invariant cells and given

the consistency of the microstimulation results, even in clusters with incomplete invariance (p < 0.003 for each monkey; t test for a significant psychometric shift toward more preferred choices), it seems possible that 3D-structure categorization does not solely rely on IT cells with complete tolerance for position-in-depth. Yet we cannot exclude the possibility that stimulation in 3D-structure selective clusters with incomplete invariance may have also stimulated nearby completely invariant structure-selective cells, from which we did not record, that caused the increase in preferred choices. Considering only the trials in which monkeys made a preferred choice, we observed significantly shorter average reaction times on stimulated compared to nonstimulated trials (Figures 6A and 6C; M1: average RT-difference: 3 ms; p = 0.04; M2: average RT-difference: 11 ms; p = 0.006; ANOVA). Furthermore, for non-preferred-choice trials, we noticed significantly longer average reaction times on stimulated compared to nonstimulated trials (Figures 6B and 6D; M1: average RT-difference: ∼5 ms; p = 0.002; M2: average RT-difference: ∼17 ms; p = 0.003; ANOVA).