Figure 5.
Effect of Trial-to-Trial Variation in the Noisy Motion Information on Choice
The analysis compares the leverage of information before and after a putative threshold crossing that terminates integration. Leverage is based on logistic regression of motion energy (right minus left; see STAR Methods), controlling for motion strength.
(A) Leverage of early- and late-motion energy (blue and red, respectively). The estimates are shown as a function of the inclusion window. Pairs of points include only trials with estimated termination times (tθ) near 400 ms. (Inset shows how the putative termination time tθ is acquired on each trial. It is the tSD from the clock setting minus the non-decision time estimated from the fits in Figure 2, top row). Filled symbols designate non-zero leverage (p < 0.05). Larger tolerances permit inclusion of more trials but blur the distinction between pre- and post-tθ. Error bars indicate SE.
(B) Difference in the leverage of motion energy before and after 400 ms requires tθ to be near 400 ms. The bars on the right side show the leverage values using tolerance of ±133 ms (same value and SE as the points in A marked by arrows). Bars on the left show the average leverages obtained by sampling the complementary trials with tθ outside this tolerance window (5,000 bootstraps of sets of 873 trials; error bars are average of the SE from the bootstraps). The distribution of differences (β1–β2) rarely exceeds the observed difference (p < 0.007). Combined data from subjects 1–4 using motions strengths 0%, ±3.2%, and ±6.4% coherence.