Npgrj_NN_1874 423..425

Accurate timing over the subsecond scale is essential for a range of human perceptual and motor activities, but the mechanisms for encoding this time scale remain poorly understood. Current research is beginning to call into question traditional ideas that events are timed by a centralized supramodal clock, in favor of distributed, modalityspecific mechanisms1. Two recent studies from visual psychophysics are particularly challenging for centralized event timing: local adaptation of the visual field by fast-moving stimuli selectively reduces the apparent duration for stimuli presented to that position2, and saccadic eye-movements drastically decrease the apparent duration of visual but not auditory events3. Although these results speak against a centralized supramodal clock, it remains unclear whether visual event timers show a genuine spatial tuning in external space, which would implicate highlevel sensory mechanisms. In this study, we provide evidence that the neural mechanisms that time visual events are spatially selective in realworld coordinates, allowing them to separately time real objects localized in space. Observers adapted to a grating drifting within a 121 diameter patch while fixating on a clear black spot to its lower left. After a suitable adaptation period, they made a saccade to a target that appeared 151 right of fixation (see Fig. 1a and ref. 4). Eight hundred ms after the display of the target, a test grating was presented for 600 ms in one of three randomly chosen positions: the same retinotopic position, the same spatiotopic position or in one that was neither (control condition). In separate sessions, subjects maintained fixation after adaptation, and the test was presented in the same (retinotopic and spatiotopic) position as the adaptor (full adaptation condition). Five hundred ms after the offset of the test, a probe grating of variable duration was presented at a fourth position, and observers indicated whether it appeared to be of shorter or longer duration than the test. The point of subjective equality (PSE) of the test and probe duration was estimated from the median of the best fitting cumulative Gaussian psychometric function (Fig. 1b). As adaptation to high speeds is known to decrease apparent speed5, and the apparent duration of moving stimuli depends on speed6, we first matched the apparent speed of the probe to that of the test (separately for all subjects and conditions using standard staircase techniques), then used this speed-matched probe for subsequent measurement of perceived duration. For two subjects (triangles in Fig. 1c), we matched speed by increasing the speed of the test and leaving the probe at 10 Hz (although this had the disadvantage of lowering the apparent contrast of the test, a problem avoided with the other technique). In practice, speed-matching affected mainly the full and retinotopic adaptation conditions, with little change being necessary for the spatiotopic condition in six out of eight subjects. We then estimated the PSE of the test stimulus duration for the four adaptation conditions (results for representative naı̈ve observer RA, Fig. 1b). Under all conditions, the psychometric functions rose smoothly from 0 to 1, with an s.d. around 80 ms. PSEs for both the control and the retinotopic adaptation conditions were near 600 ms, the actual duration of the test. Full adaptation reduced the PSE to 526 ms and spatiotopic adaptation to 456 ms. We plotted the percentage reduction in perceived duration (relative to control) for spatiotopic and retinotopic adaptation against that for full adaptation (results for all subjects, Fig. 1c). There was some variability in the magnitude of the effect, particularly for the full adaptation condition, but all eight subjects showed strong reduction in the spatiotopic condition and virtually none in the retinotopic condition. The average PSEs (Fig. 2) confirm this: retinotopic adaptation had little effect, with an average PSE of 591 ± 7 ms (near the control value of 600 ms), whereas the PSE for full adaptation was 476 ± 31 ms and 443 ± 22 ms for spatiotopic adaptation. We also measured the apparent duration with the probe moving at the same physical speed as the 10-Hz test (Fig. 2). Under these conditions, the effect was strong for retinotopic adaptation (33%) and strongest for full adaptation (41%). Notably, the spatiotopic effect was virtually identical to the speed-matched condition (27% reduction), which was to be expected as the conditions were very similar, but shows that the effect does not depend on speed matching. The average PSEs for control runs with no adaptation had no substantial differences, confirming that the results cannot be attributed to spatial position, temporal order or the saccades themselves. Finally we measured the effect dichoptically for three of the subjects, adapting the right eye and testing with the left (using ferroelectric goggles; Figs. 1c and 2). Although the test and probes had identical speed in this case, there was very little retinotopic adaptation (bootstrap t-test, P 1⁄4 0.39, not significant), whereas the spatiotopic adaptation was similar to the full adaptation, both significantly different from the control (P 1⁄4 0.001 and P 1⁄4 0.002, respectively). These results suggest that at least two mechanisms are involved in adaptation of duration, one retinotopic and the other spatiotopic. The