Figure 5. Linear separability of object representations at the single-neuron level.
(A) Illustration of how a similar amount of view-invariant information per neuron can be encoded by object representations with a different degree of linear separability. The overlap between the response distributions produced by two objects across multiple views is similarly small in the top and bottom examples; hence, the view-invariant object information per neuron is similarly large. However, only for the distributions shown at the bottom, a single discrimination boundary can be found (dashed line) that allows discriminating the objects regardless of their specific view. (B) Linear separability of object representations at the single-neuron level (median over the units recorded in each area ± SE; *p<0.05, **p<0.01, ***p<0.001; 1-tailed U-test, Holm-Bonferroni corrected). The number of cells in each area is written on the corresponding bar. (C) Ratio between linear separability, as computed in (B), and view-invariant object information, as computed in Figure 4C, right (median ± SE). Significance levels/test as in (B). In both (B) and (C), .