Fig. 2
Memory-guided microsaccades had reaction times and success rates consistent with being genuine responses to task instruction. a Reaction time distribution for memory-guided microsaccades in monkey N. The data shows all trials with target eccentricities <1o (Methods). The monkey had a reaction time distribution typical of saccadic responses. Error bars denote 95% confidence intervals, and the histogram was normalized by the total number of trials. b Cumulative probability of the same data, demonstrating a plateau of >80% success rate; the monkey successfully generated a memory-guided microsaccade (that is, within a reasonable reaction time; Methods) on the great majority of trials. Error bars denote 95% confidence intervals. c Same data as in b but separated by target eccentricity (inset legend). Even the smallest target eccentricities were associated with a majority of successful reactions. d–f Similar observations from monkey M. g–i Similar observations from monkey P. j–l Similar observations from our human subjects (Methods). m–x We repeated the same analyses above, but now for corrective, visually guided microsaccades occurring after target reappearance. Reaction times (m, p, s, v) were faster than for memory-guided microsaccades, and often exhibited express reactions (e.g. arrow in m). However, success rates were significantly lower than in the instructed memory-guided microsaccades (n, q, t, w; the horizontal blue line in each panel, and associated 95% confidence interval, replicates the plateau success rate from the corresponding panel in b, e, h, k). Success rates of corrective, visually-guided microsaccades also depended on target eccentricity (o, r, u, x) similarly to memory-guided microsaccades. n = 781, 1,602, 2,235, 7,402 memory-guided trials in monkey N, monkey M, monkey P, and humans, respectively. n = 2774, 4471, 3461, 10,866 visually guided trials in monkey N, monkey M, monkey P, and humans, respectively
