Figure 6.

Decoders weights and direction selectivity. A–C, Three example populations showing the normalized decoder weights plotted against the left-right selectivity for the standard decoders (left column, correlations present) and the decoders trained on shuffled data (right column, no correlations present). D, Decoder weights for all populations plotted against left-right selectivity, showing a strong relationship (left), especially when spiking correlations are removed (right). E, Decoder weights for all populations plotted against vertical meridian offset, showing a much weaker relationship (left), even when spiking correlations were not present (right).