Fig. 4: BCI training did not enhance stimulus responses in the subset of indirect neurons that matched their stimulus preference with one of the direct neurons.

a Example indirect neuron responses to 12 orientations at the preferred spatial frequency (0.04 cycles/°), before (VS baseline 2) and after BCI learning (VS post-learning). The mean across trials is shown in gray scale at the bottom (scale limits: 0–9 events). Scale bar: 10 μm. b Example 2-dimensional tuning profile of one DP neuron, and one indirect before and after BCI learning. Labeled contours of the Gaussian fit are 75%, 50%, and 25% of the peak response. c Spatial position of the subset of indirect neurons that were similarly tuned to DP#2 (red circle) in Mouse #2 before BCI training was initiated. Neurons were considered to be similarly tuned if their preferred orientation was within 30° and their spatial frequency within 0.03 cycles/° of a direct neuron, prior to BCI training. d The trial-by-trial response reliability to the preferred stimulus was the same before (VS baseline 2) and after BCI training (VS post-learning), the median Δ reliability = −0.06; Wilcoxon signed-rank test, p = 0.093, n = 39 neurons. The kernel density estimate (gray), median (white circle), and interquartile range of the distribution (black bar) are indicated. e Change in response amplitude at the preferred stimulus between VS baseline 2 and VS post-learning, same pool of indirect neurons as ‘d’. The response amplitude (failure trials removed) was the same before and after BCI training, the median Δ amplitude = 1.02 events/bin; Wilcoxon signed-rank test, p = 0.382, n = 25 neurons. Source data are provided as a Source Data file.