Figure 1.
Flash lag illusion and monkey behavioral task.
(A) Illustration of the flash lag illusion. A flash that is presented in vertical alignment with a moving bar appears to lag behind the moving bar. (B) The schematic shows stimuli (not drawn to scale) and events in a trial. Stimuli were presented on CRT monitors with a gray background. Subjects fixated for 300 ms after which the stimulus period began. A bright moving bar started to move at a constant speed (10, 14 or 20 °/s) from the left or right end of the monitor display. After the onset of the moving bar, another bar of the same size and luminance was flashed below the moving bar for one video frame (10 ms) at a fixed location. The flash onset time was adjusted to provide different horizontal spatial offsets. For each trial a single spatial offset was chosen randomly from a range of offsets. After a delay of 200 ms from the flash onset, subjects were allowed to report if the flash was located on the right or left side of the moving bar, using a joystick or by saccading to visual targets (illustrated as two black dots in the upper visual field). The next trial began after 1200 ms of inter-trial period.
Table 1.
Behavioral training and reward scheme for monkeys.
Figure 2.
Monkeys perceive the flash lag illusion.
(A) Psychometric function from a single session of monkey B for a moving bar speed of 20°/s. The abscissa shows the veridical spatial offsets of the stimuli. The ordinate shows the probability of the monkey reporting that the flash lagged behind the moving bar. Four different points along the psychometric function are illustrated with cartoons of veridical stimuli (filled vertical bars) and hypothetical perceived stimuli (open vertical bars). The point of subjective equality (PSE) is the spatial offset at which the monkey reports the flash as lagging in 50% of trials (i.e. he perceived the two bars as vertically aligned). In this session the PSE was −0.31°, indicating that the flash had to be placed 0.31° ahead of the moving bar to make them perceptually aligned. Ns: average number of trials per suprathreshold spatial offset data point (spatial offset magnitude >1.5°); Np: average number of trials per subthreshold spatial offset data point (spatial offset <1.5°). Error bar: 95% bootstrap percentile-based plug-in estimate of confidence interval. (B) Psychometric function fit of the pooled responses (5–8 sessions) of monkeys B and H. For clarity, data from only two speeds (10 and 20 °/s) are shown. Ns and Np: as in A. For the speed that is not shown (14 °/s): for monkey B, Ns = 4061 and Np = 628; for monkey H, Ns = 852 and Np = 108. Error bars as in A.(C) Perceived spatial lag for monkey B (open circles) and monkey H (filled circles), as a function of the speed of the moving bar. Each data point is the PSE estimated from the psychometric function fit (shown in panel B) of responses pooled from five to eight sessions. Error bars as in A.(D) Slope of the psychometric functions measured at the threshold point (0.5 on the ordinate). Coding and error bars as in C.
Figure 3.
(A) Example psychometric functions from a naïve (left panel) and a non-naïve (right panel) human subject measured at two speeds of the moving bar. Each data point shown was computed from 120 (subject AL) or 126 (subject SP) trials pooled from three to four sessions. Error bars: 95% bootstrap percentile-based plug-in estimate of confidence intervals.(B) Perceived spatial lags (left panel) and slopes (right panel) of the psychometric function measured at different speeds of the moving bar from eight human subjects. Each data point represents the relevant parameter extracted from psychometric functions fitted (as in panel A) on responses pooled across two to five sessions. Open symbols: naïve subjects; filled symbols: non-naïve subjects. Error bars as in A. Slope error bar upper limits for subject TS are cropped: values are 6.5 and 6.1 for speeds 10 and 14 °/s respectively).
Figure 4.
Comparison of illusion characteristics in monkeys and humans.
(A) Perceived lag in monkeys (open circles) and humans (filled circles) averaged over subjects as a function of speed of the moving bar. Error bars: ±1SEM.(B) Slope of the psychometric function averaged over subjects. Coding and error bars as in A.
Figure 5.
Change in perceived lag in monkeys and humans over time.
Perceived lag measured at three speeds of the moving bar over the course of testing on multiple days in humans (panel A) and monkeys (panel B). (A) Each data point represents one to two sessions (sessions within a day were pooled). Subject AL and SS showed a significant increase and decrease, respectively, in perceived lag over testing days. Error bars: 95% bootstrap percentile-based plug-in estimates of confidence intervals.(B) Each data point represents a single session for monkey B and up to three consecutive non-overlapping sessions for monkey H (sessions were pooled if necessary, to meet a criterion of a minimum of five trials per stimulus condition). Error bars as in A.(C) Perceived lag (left panel) and slope (right panel) of psychometric function estimated from only the first one to three testing days in monkeys and humans and averaged over subjects. Error bars: ±1SEM.