Abstract
Attention is known to modulate sensory inputs; however little is known on how the interactions between top-down and bottom-up cues dynamically evolve. Here, we show that auditory-visual perception changes systematically and non-monotonically from an interaction of sensory and attentional factors during the continuous viewing of a physically invariant stimulus. The stimulus was a visual target (flash) revolving about fixation. On every rotational cycle (60 cycles/trial x 534 ms per cycle = 32.04 s), a sound occurred in synchrony with the same angular position of the flash. For each cycle, observers (n=10) had to report on the perceived location of the flash coincident with the perception of sound (audio-visual synchrony). The initial estimate of the location of audio-visual synchrony was slightly behind veridical audio-visual synchrony, but then over subsequent cycles, gradually moved forward in the direction of the motion. This steady drift was followed by a sharp reset back to the initially judged position. This perceptual sequence was qualitatively common across observers. When we manipulated attention to specific positions of the cycle on a trial, the location of audio-visual synchrony moved and remained there. The results show a striking parallel to the continuous report data suggesting the transitions are mediated by an attractive effect on audiovisual timing exerted by attention. Motion seems to drag spatial attention forward from the location of veridical audio-visual synchrony. As attention drifts far beyond veridical, its effect is overruled by the near-veridical sensory synchrony signals and the perceived position of audio-visual synchrony snaps back to near-veridical.