Extended Data Fig. 8: Early peak during gaze fixations partly predicts the upcoming gaze. | Nature

Extended Data Fig. 8: Early peak during gaze fixations partly predicts the upcoming gaze.

From: Remote activation of place codes by gaze in a highly visual animal

Extended Data Fig. 8

(a) Coefficients of a linear mixed effects model that fit the early response (at +17 ms relative to saccade peak) as a function of eight behavioral variables (fixed effects), and allowed the intercept to vary for each cell (random effect). Four of the variables accounted for the gaze fixation preceding the saccade, and the other four accounted for the gaze fixation following the saccade. For the previous and next fixation, variables included the horizontal and vertical deviations from the target (which could be positive or negative), as well as the squared values of these deviations. Included are cells with place and gaze selectivity >0.5 (n = 278 cells). Asterisks indicate significant coefficients (p < 10−9 for all, p-value for the t-statistic of the hypothesis that the corresponding coefficient is different from zero, returned by MATLAB function fitlme. We then adjusted each p-value for multiple comparisons using a Bonferroni correction). Error bars indicate 95% confidence intervals for the coefficient estimates. The key conclusion is that the early response depended on the location of the next fixation. (b) Same as Fig. 3h, but only including trials where the previous fixation was 30–40° from the preferred target. Note that the early response still depends on the bird’s expectation of the light turning on, even though the previous fixation is clamped in the same narrow range of angles for all conditions. Due to the small subset of trials, the number of included cells is now smaller than in Fig. 3h (n = 387, 377, and 364 cells for the black, blue, and pink traces). (c) Coefficients of a model that fit the early response in the Blocked-trial task and included variables for whether the chickadee should expect the light to turn on in the trial, separately for catch and non-catch trials. In both cases, coefficients corresponding to these variables were significantly non-zero (* indicates p < 0.05, calculated as in panel (a)), indicating that the early response predicted whether the light would turn on. For (c), included cells were the same as in Fig. 3h (n = 402 cells).

Back to article page