Figure 9.
Effects of intention estimation on velocity modulation across all training runs. Analysis was restricted to channels that were originally velocity modulated, had a well correlated linear model, and were valuable to offline decoder performance (n = 37 channels over 12 days, Monkey W; n = 16 channels over 7 days, Monkey N). The raw change in modulation after intention estimation is applied to the original velocities is shown binned in 2 spike/s intervals. In both subjects intention estimation improved modulation compared to a randomized bootstrap analysis (p < 1 × 10−7 Monkey W, p < 0.01 Monkey N; one-sided two sample t-test).