Two experiments were designed to examine the relationship between three measures in visual search. (1) Reaction time (RT), (2) number of first saccades to target and (3) number of trials in which no saccade was made before target identification. It is typically assumed that as search becomes more difficult then there will be a concurrent increase in RT, a decrease in the number of first saccades to target and a decrease in no-saccade trials.

In both experiments the display consisted of four capital letters arranged symmetrically in a square around the fixation point and the participants' task was to decide whether either the target letter T or H was present in the display.

Experiment 1 had three conditions. In condition 1, participants searched for the targets among three capital letter O distracters. In condition 2, heterogeneous distracters (A, X, Y) were used and in condition 3 the target letter was displayed on its own without any distracters. Manual RT for the target decision was fastest for condition 3 (542 ms), slower for condition 1 (626 ms) and slowest for condition 2 (800 ms). Consistent with the RT data, percentage of target directed first saccades was highest in condition 3 (100%), lower in condition 1 (81%) and lowest in condition 2 (46%). For RT and first saccade to target, a comparison between condition 1 and 3 would suggest that the O distracters impart a cost. However, there were more trials that contained a saccade before response in condition 3 (96%) than in condition 1 (73%) and 2 (84%). We explain this surprising results in terms of lower-level factors i.e. the global effect (Findlay, 1982, Vision Research, 22: 1033-1045) and the automatic generation of a saccade to an isolated target.

In Experiment 2 there were 3 conditions. Condition 1 and 2 replicate conditions 1 and 2 in Experiment 1. In Condition 3 the target letter was present with the distracters S, U and O. Comparison of condition 2 and condition 3 is of special interest here. RT decreased between condition 2 (714 ms) and condition 3 (675 ms) as did the number of saccade trials (condition 2: 83%, condition 3: 73%). Likewise, the number of first saccades to target in Condition 2 was lower (48%) than in Condition 3 (53%). A further analysis of the direction of the initial saccades revealed that in condition 2 misdirected saccades were distributed equally over distracters but in condition 3 this was not the case. This suggests that higher-level factors such as item identity also influence search performance.

Together these experiments demonstrate the influence of both lower-level and higher-level factors in determining saccade and RT performance in search.