What do eye movements reveal about the role of memory in visual search?

Abstract

Horowitz and Wolfe (1998, 2003) have challenged the view that serial visual search involves memory processes that keep track of already inspected locations. The present study used a search paradigm similar to Horowitz and Wolfe's (1998), comparing a standard static search condition with a dynamic condition in which display elements changed locations randomly every 111 ms. In addition to measuring search reaction times, observers' eye movements were recorded. For target-present trials, the search rates were near-identical in the two search conditions, replicating Horowitz and Wolfe's findings. However, the number of fixations and saccade amplitude were larger in the static than in the dynamic condition, whereas fixation duration and the latency of the first saccade were longer in the dynamic condition. These results indicate that an active, memory-guided search strategy was adopted in the static condition, and a passive “sit-and-wait” strategy in the dynamic condition.

Acknowledgments

This research was supported by Deutsche Forschungsgemeinschaft Grant FOR 480/1–1.

Notes

¹ In a reanalysis of Horowitz and Wolfe's (1998) data set, Shore and Klein (2000) found fundamental differences between the static and dynamic conditions. In particular, when plotting target-absent RTs against display size, the slopes were steeper in the static than in the dynamic condition. In contrast, mean RT standard deviations and errors were overall larger in the dynamic than in the static condition. Shore and Klein concluded from this that different search strategies were at play in the two conditions. This conclusion receives further support from a recent meta-analysis of relevant studies by Klein and Dukewich (2006).

² A deliberately adopted sit-and-wait strategy would be but one of several reasons why observers may make fewer eye movements in the dynamic than in the static search condition. Another reason may be that the rapid and unpredictable changes of stimulus locations have an inhibitory effect on the oculomotor system, resulting in fewer saccades and longer saccadic latencies (Pannasch, Dornhoefer, Unema, & Velichkovsky, 2001; Reingold & Stampe, 2002). However, in effect, oculomotor inhibition may be just regarded as a variation of sit-and-wait behaviour.

³ The aim of the experiment was to examine for differential oculomotor strategies in the two search conditions, rather than to produce further evidence of memory processes operating in visual search. There is already ample evidence for the latter (see Klein & Dukewich, 2006, and Shore & Klein, 2000, for reviews), though memory capacity estimates vary dependent on whether RT (e.g., Jiang & Wang, 2004) or oculomotor measures (e.g., Gilchrist & Harvey, 2000, 2006; Hooge & Frens, 2001; Peterson, Kramer, Wang, Irwin, & McCarley, 2001) are taken into account.

⁴ Thus, if observers were to focus attention on only one, constant, display location, a sit-and-wait strategy would fail in approximately 72% of the trials. However, if attention is spread across a display region of, say, 4 × 4 locations (the size of the aperture used by von Mühlenen et al., 2003), then a sit-and-wait strategy would hardly ever fail (in only 0.23% of the trials).

⁵ The robust effects of display size on RTs may be taken to suggest that some kind of serial search process was at play even in the dynamic condition. However, these effects are better explained in terms of a sit-and-wait strategy when taking into account the differences in the number of fixations and fixation duration between the dynamic and static tasks. In the dynamic condition, while the number of fixations on target-absent trials was near-equivalent with 8-, 12-, and 16-element displays (0.51, 0.52, and 0.62, respectively), fixation duration increased as a function of display size (335, 321, and 359 ms). This pattern was reversed in the static condition, with the number of fixations increasing across display size (2.21, 2.97, and 3.53 for 8-, 12-, and 16-element displays, respectively) and fixation duration being equivalent in the three display size conditions (162, 163, and 162 ms).