Sensitivity to Acceleration in the Human Early Visual System

Abstract

It is widely believed that the human visual system is insensitive to acceleration in moving stimuli. This notion is supported by evidence that detection sensitivity for velocity modulation in moving stimuli is a lowpass function of the velocity modulation's temporal frequency. However, the lowpass function might be a mixture of detection by attention-based tracking and low-level mechanisms sensitive to acceleration. To revisit the issue of acceleration perception in relation to attentive tracking, we measured detection sensitivities for velocity modulations at various temporal frequencies (0.25–8 Hz) by using drifting gratings within long or short spatial windows that make the tracking of grating easier or more difficult respectively. Results showed that modulation sensitivity is lowpass for gratings with long windows but bandpass for gratings with short windows (peak at ~1 Hz). Moreover, we found that lowpass sensitivity becomes bandpass when we removed observer attention by a concurrent letter identification task. An additional visual-search experiment showed that a target dot moving with a velocity modulation at relatively high temporal frequencies (~2–4 Hz) was most easily detected among dots moving at various constant velocities. These results support the notion that high sensitivity to sluggish velocity modulation is a product of attentively tracking of moving stimuli and that the visual system is directly sensitive to accelerations and/or decelerations at the preattentive level.

Heading perception depends on time-varying evolution of optic flow

“…It is crucial to distinguish between acceleration sensitivity for drifting gratings and for moving dots (particularly for a single dot), as Weber fractions for accelerating dot motion are 5 to 10 times higher (77) (see SI Appendix for elaboration). Sensitivity to optic acceleration for drifting gratings peaks for 0.5°/s 2 phase speed change (74), which was similar to the phase speed changes in our time-varying phase motion stimulus (0.33°/s 2 , mean across patches and time; 0.61°/s 2 , mean across patches at end of stimulus video). Acceleration sensitivity is higher for radial than lateral dot motion patterns that share the same low-level features (78), suggesting that heading-tuned brain areas (e.g., medial superior temporal [MST], VIP) may be specialized for accelerating motion [although there is also physiological evidence to the contrary (79)].…”

“…Biphasic temporal responses to motion have been measured psychophysically (72) and in the responses of MT neurons (73) (i.e., computing the temporal derivative of image velocity). Similarly, psychophysical experiments reveal that humans are sensitive to optic acceleration (74), with Weber fractions of 0.05 for a 1-Hz sinusoidal temporal frequency modulation (i.e., 0.5°/s 2 phase speed change) of a drifting grating (74), similar to those for speed discrimination of a drifting grating, which ranged from 0.03 to 0.14 (75,76). It is crucial to distinguish between acceleration sensitivity for drifting gratings and for moving dots (particularly for a single dot), as Weber fractions for accelerating dot motion are 5 to 10 times higher (77) (see SI Appendix for elaboration).…”

Attention updates the perceived position of moving objects

“…To test this prediction, in Experiment 2, the participants were given an additional task of identifying digits embedded in a central stream of letters. This rapid serial visual presentation (RSVP) task restricts observers' attentional resources, including for attention-based motion processing (e.g., Motoyoshi, 2011;R. Nakayama & Motoyoshi, 2017).…”

Attention updates the perceived position of moving objects

“…From 0.2 sec before to 0.2 sec after the stimulus-presentation period (1.9 sec in total), a rapid-serial visual presentation (RSVP) display appeared concurrently in the center of the screen instead of the fixation point. We used the RSVP display in order to confine the observersʼ attentional resources throughout stimulus presentation [31]. In the RSVP display, 12 capital alphabetical letters (excluding I, O, Q, Y, Z) were serially presented every 80 msec and separated by a blank interval of 80 msec (6.3 Hz).…”