Figure 4.
Similar attention-related effects on neuronal populations in two brain areas and two tasks. (A) In a change detection task, the effects of attention on the stimulus information that can be decoded from small populations of V4 neurons is similar to MT. The plot shows the ability of a cross-validated linear decoder to distinguish the original from changed stimuli (intermediate change amount) for both the Stimulus and Choice decoders (no SC data was available). Error bars represent SEM, gray lines are individual hemisphere-sessions (see methods). Attention significantly affected the performance of both the Stimulus and Choice decoders (N= 98 sessions, two-tailed paired t-tests, p=7.4×10−4 and p=1.1×10−15, respectively), but the attention-related improvement in the Choice decoder was greater than in the Stimulus decoder (N= 98 sessions, two-tailed paired paired t-test, p=2.1×10−8). (B) Decoding performance was typically better using the V4 responses from the cued condition and the Choice decoder weights from the uncued condition (y-axis) than using the V4 responses from the uncued condition and the Choice decoder weights from the cued condition (x-axis; N= 98 sessions, two-tailed paired t-test, p=0.0029; compare to Figure 3B). (C) The ability of a cross-validated linear decoder using V4 population responses to distinguish between stimulus configurations during a contrast discrimination task6 reveals no significant effect of attention (N= 17 sessions, two-tailed paired t-test, p=0.31). Plotting conventions as in A. Because of the details of the discrimination task (which did not include choices related to uncued stimuli), it was impossible to calculate a choice decoder using these data.