When one looks around a display, the eye makes a fixation on a target of interest, and then makes a very fast (<30 msec) ballistic saccade to the next target, where it typically 'stops on a dime' (without jitter) and either remains where it has landed, or makes another fast corrective saccade to center the new target more exactly. Programming saccades takes time, and usually latencies for the next sacade are slow - around 300 msec is typical. However, some years ago Fischer discovered in monkeys that if the current target is turned of before the next saccade target is turned on, saccades speed up. He called this the 'gap' effect, the gap being the time from offest to onset; a gap of 200 msec is the most effective for both monkey and man.

    We have investigated several explanations for the 'gap effect' . In our displays, the first target was a fixation spot, presented at the center of the screen. The second target (which we call 'the target' for short) appeared at random in one of 4 locations (top, bottom, left, right of the display screen, at 10 deg eccentricity). We measured both the 'saccade reaction time' (SRT) to the target, using an eye tracker, and the 'motor reaction time' (MRT), or time to press a button to indicate perception of the second target. Three trained observers pressed a manual key at target onset (MRTtarg), or pressed a key at fixation offset (MRTfix), or made a single saccade to the target (latency = SRT).

    The fixation spot was continuous (no gap), or (gap) was blanked, or dimmed, or brightened, or changed into 4 spots, just 200 msec before the target appeared.

    As potential explanations of the gap effect, we exclude, by data analysis, (1) supression of microsaccades, (2) early triggering of pre-programmed saccades, and (3) express saccades. We also exclude (4) speed-accuracy trade-offs --faster saccades were less accurate within, but not across, conditions, (5) foveal capture of attention, (6) changes in the shapes of the SRT distributions, and (7) salience of the fixation target, in that mean MRTs and mean SRTs were independent across conditions. We conclude that the gap effect reflects both (8) an overall warning effect and (9) the disengagement of saccade-specific central attention.

    Previous researchers have rejected (8) the notion that attention is relevant to the gap effect, but we think mistakenly. Our data suggest that release of attention from the fixation helps one to speed up programming the next saccade.

    The application to warning displays is that if one wants to speed processing of the next target, one should terminate the previous target early -- as long as encoding the previous target has already been ensured.