FIGURE 10.

Texture and motion tuning of four example units in RFAMpa. (A) Texture tuning. Mean responses to examples from the seven MSCT classes. (B) Motion tuning. Response to translating binary random noise patterns at different speeds (x-axis) and directions (colors indicate direction of motion as show in the legend). (C) Same as (B), now for rotational motion. Error bars indicate standard error over examples. This figure shows that, although RFAMpa was trained only on static MSCT and only to reproduce the V2 population average responses in Figure 7A, it self-organized hidden units with diverse selectivity for the speed and direction of moving texture patterns.