Fig. 3: Conflict neuron time–depth organization.
From: Functional architecture of executive control and associated event-related potentials in macaques
a Average spike rate (top) and recruitment (bottom) of broad- (dark) and narrow-spiking (light) neurons on canceled (thick) and latency-matched no-stop-signal (thin) trials, aligned on SSRT (left) and tone (right), normalized to 95th percentile within respective intervals. SSRT-aligned recruitment was the difference between trials indicated by p(diff > 0). Tone-aligned recruitment was the difference in spiking on canceled trials (Acanceled) relative to baseline (BL) lowest spiking value ±500 ms from the tone. Modulations after tone were not analyzed. Post-saccadic spiking on no-stop-signal trials before tone can exceed that on canceled trials. b Time–depth plot showing latency and recruitment across depth from perpendicular penetrations. Symbols mark the beginning of modulation for broad- (triangles) and narrow-spiking (stars) neurons. Color map indicates percentage through time at each depth relative to sampling density. Solid horizontal line marks L3–L5 boundary. The lower boundary of L6 is not discrete. c Modulation on canceled (thick) relative to latency-matched no-stop-signal (thin) trials for lower, intermediate, and higher p(NC|SSD) of two representative neurons: n1 had broad spikes in L6; n2 had narrow spikes in L5 (n2) (identified in b). Shaded interval highlights significant differences in spiking across conditions. d Model comparison table listing each tested model. The heatmap shows the difference in BIC values (ΔBIC) for each model compared to the model with the lowest BIC value (black fill) with hotter colors corresponding to lower ΔBIC values. Asterisks (*) indicate models with a significant main effect. The green circle indicates the best-fit model; the white indicates candidate models (ΔBIC < 2). Spike rate variation after SSRT was best predicted by the conflict model. Full statistics in Supplementary Table 3. e Significant variation of spiking (residualized and adjusted for spiking across neurons) as a function of p(NC| SSD) proxy for conflict (normalized z-scale). Each point plots average spike density and mean p(NC|SSD) across all trials in early-, mid-, or late-SSD bins for each of 75 neurons. For 11 neurons no reliable estimate of spike density for the late-SSD bin was obtained due to too few trials. These data points were not included.
