Abstract
It has long been known that a brief target can be rendered invisible if followed by a brief mask. Two general patterns of backward masking have been observed when the strength of the target percept is plotted against the SOA between the target and mask (a masking function). For some kinds of masks, the masking function increases monotonically as the SOA increases from zero. For other kinds of masks, the masking function is u-shaped, with a bottom at around 50 ms. We now propose that there is a more general principle than type of mask that describes whether a monotonic or u-shaped masking function appears. Namely, at the shortest SOAs the target and mask integrate into a single percept and the visibility of the target features in the integrated percept determines performance in the masking task. A monotonic or u-shaped masking function occurs when the integrated percept hides or facilities the target's features, respectively. We tested this hypothesis by running a backward masking experiment with four types of targets and five types of masks. On each trial the observer identified the location of a known target in a field of three distracters. Some target/mask combinations produced monotonic masking functions, while others produce u-shaped masking functions. A second experiment verified that target identification at the shortest SOAs was related to the visibility of the target in the integrated target/mask percept. The target and mask stimuli were presented together in a visual search experiment, which measured RT for detecting the presence or absence of the target among the distracters. Across the different target/mask combinations RT correlated strongly with percent correct identification of the target at the shortest SOAs in the masking experiment (r=−0.90, −0.87, −0.87) for each of the three observers. This relationship suggests that the shape of the masking function is determined by the effect of temporal integration of the target and mask.