Standard-oddball explorer legend

Heatmaps show baseline-subtracted context or matched-control signals, or their difference, in the selected modality's native units. Teal and grey traces show context and control means; bands are across-source SEM, not across mice. Individual-source means have no trial uncertainty band. Interactive activity is aligned to native stimulus-table start_time over −1.5 to +1.5 s; the static figure retains its original window. Tan/grey shading marks median actual neighboring/mismatch presentation bounds, not the 500-ms test windows. Neighboring trials retain their actual stimuli; controls have randomized orientations. Recorded samples only contribute to means, without interpolation across acquisition pauses. Source identities, tests and recorded-trial ranges are available to assistive technology; complete counts are in the frozen data.

Q1 compares the deviant with the preceding standard; Q2 compares context and matched-control baseline-subtracted responses. The exploratory screen uses raw p<0.05 and |MI|>0.1, either sign. Both is their intersection. Q2-nonsignificant is not proof of absent visual activity. Frozen context-only Rastermap orders are filtered, never refitted. SLAP2 rows remain grouped by DMD.

Cohort bars pool responsive counts over eligible jointly testable observations. Points identify mice; Neuropixels/mesoscope intervals are 95% mouse-cluster bootstrap bounds. The two-mouse SLAP2 subset has no bootstrap intervals. SLAP2 circles/squares mark mice 828408/829704. Exact counts and mouse IDs are retained in the embedded data; the barplot has an accessible count summary. Population selections do not change cohort classifications or denominators. These separate recordings are not matched cells or tests of cross-modality differences.