The study of saccadic eye-movements during visual search has revealed much about the underlying processes that allow targets to be selected amongst distracters. This project has been greatly influenced by seminal work on reaction times (RT's) and visual search, with much of the recent research being concentrated on comparing oculomotor behaviour between parallel and serial search conditions. The current study aimed to characterise scan path patterns in serial search in more detail and look for qualitative difference across display conditions.

In Experiment 1 we compared three search tasks all of which involved detecting the presence or absence of an E. Condition 1 was a pop-out search task (search for an E amongst O's). Condition 2 was a serial search task in which the distracters where homogenous and similar to the target (search for an E amongst F and L's). Condition 3 was also a serial search task but with highly heterogeneous distracters (search for an E amongst 24 different letters).

For measures of RT, number of saccades and errors, condition 3 was most difficult and condition 1 was least difficult. This supports the idea that both distracter heterogeneity and target-distracter similarity determine task difficulty in search (Duncan & Humphreys, 1989, Psychological Review, 96: 433-458). For conditions 2 and 3 we found strong evidence for stereotypical, strategic scanning patterns; this effect was strongest in condition 3. There is also a clear relationship between strategic scanning and RT: this can be used to account for large differences in RT between trials.

In Experiment 2 we investigated whether the increased use of the strategy in condition 3 compared with condition 2 was a result of distracter heterogeneity or task difficulty. It appears that task difficulty is the overriding factor.

Overall these experiments support the idea that as search becomes more difficult there is a systematic shift from one scanning behaviour to another.

Supported by Engineering and Physical Sciences Research Council UK.