Serial Visual Search Processed Efficiently

Across Multiple Depth Planes

(presented at ARVO 1998)

Return to front page.

Visual search tasks have traditionally been classified as either serial (attentive) or parallel (preattentive) based reaction time trends as a function of the number of distractors. Serial searches show an increase in reaction time as the number of distractors increases, while parallel searches are unaffected by the number of distractors. When targets are distinguished solely on basis of a single basic feature such as color, size, or orientation, searches tend to be parallel in nature.

Previous research has been mixed in its support that disparity can lead to parallel searches (e.g. Triesman, 1986; Nakayama and Silverman, 1986; He and Nakayama, 1995; O'Toole and Walker, 1997). Intertwined with this issue is the question of whether or noti ndividuals can selectively attend to a single depth plane. Refuting the Nakayama and Silverman (1986) conclusion, He and Nakayama's (1995) results suggest that such selectivity is not possible; iso-depth attention can only be directed when there are enough elements to form the appearance of a well-formed surface. O'Toole and Walker showed that preattentive searches defined by disparity were possible, but were dependent upon the relative disparity of targets and distractors and the global surface context.

The current experiment further investigates these issues by using a serial search task and manipulating the number of distractors across depth planes. In contrast to the above-mentioned research, the task was serial within the target plane as well as across depth planes.



Based on the theory that disparity is processed by multiple channels, three hypotheses seem possible (Melton and Scharff, 1998):

1. If disparity is processed independently in parallel channels, it should be possible to perform a serial search at one depth without interference by distractors placed at other depths. If so:

2. Individual channels may be accessed simultaneously, i.e. in parallel, but attention may not be focused within specific disparities. If so:

3. If disparity is processed in a serial manner:

Additionally, if searches are influenced by attention drawn to well-defined surfaces, then searches for targets with a fixed number of distractors in the target plane should be slowed when distractors at other depth planes are more dense, and thus likely to form the perception of a global surface.



Six participants performed three visual search task experiments (single depth plane, blocked depth, and randomized depth). Experiment order was counter-balanced across participants. For all experiments, participants viewed stimuli through a front-surface mirror stereoscope and responded by pressing designated keys on a keyboard.

Stimuli consisted of left and right eyes' views of a color and size conjunctive search task. The target was always the large green square, while distractors were large red and small green squares. Prior to each trial, participants fused a nonious fixation cross. There were 10 trials for each combination of stimulus variables and each participant completed each experiment five times, with the entire first set regarded as practice trials.

Single Depth Experiment: The number of distractors (5, 10,15, 20, 25, 30) and target presence and absence were manipulated. Although viewed stereoscopically none of the elements contained any disparity, so the entire stimulus appeared to be within the plane offixation.

Blocked Depth Experiment: Three variables were manipulated: Target Depth, Condition, and Target Presence/Absence. Target Depth refers to the possible location of the target: front plane (disparity12 arcmin), middle plane (disparity 6 arcmin), back plane (disparity 0 arcmin). The condition variable contained the manipulation of the number of distractors by varying the number of distractors present in the target depth plane and the remaining depth planes. The total number of distractors always equaled 30. There were six conditions so that the target plane either had 5, 10, 15, or 20 distractors while the remaining planes had some multiple of five distractors. For this experiment, trial order was blocked by depth, so that the participants knew ahead of time in which depth plane to look for the target. See Fig. 1 for an example.

Figure 1: Left and right eyes' views; target is the large,green square.

Randomized Depth Experiment: The stimuli were identical to those presented in the Blocked Depth Experiment; however, presentation of the trials was completely randomized.


All trials with incorrect responses were excluded from the analyses. For each subject, medians were calculated for each combination of variables for each session. A completely-within ANOVAwas performed for each subject for each experiment. Additionally, an ANOVA was performed for each subject to compare responses between the blocked-depth and randomized-depth experiments.

In the single-plane experiment all subjects showed a significant increase in reaction time as the numbers of distractors increased, and all responded to present trials significantly faster than absent trials; all but one showed a significant interaction with the reaction times to the absent trials increasing more rapidly than those for the present trials. However, post-hoc analyses and a reviewof the graphs of the interactions (yellow lines in Figure 2) show that, for most subjects, the increases in reaction times were only apparent for the larger numbers of distractors. This is especially true for the target-present trials.

For both the Blocked and the Randomized Depth Experiments, there was no significant or systematic main effect of depth plane. In most cases, interactions with depth did not exist; those that did were non-systematic across experiments and participants. For both experiments, all participants showed significantly faster responses to present trials than absent trials. Across participants, there were either non-systematic or no significant effects for the manipulation of the number of distractors within the target plane.

Each participant showed a significant effect for knowledge of depth plane; however, half of the participants responded faster when depth was blocked while the other half responded fastest when depth was randomized. In most cases, knowledge did interact with target presence / absence, so that knowledge most greatly affected the target-absent trials.

Figure 2:

Figure 2 shows each participant's mean reaction times and standard errors as a function of the number of distractors. The data from the Blocked and Randomized Experiments (blue and red lines, respectively) have been transformed so that the two conditions which had five distractors in the target plane have been averaged, and the same done for the two conditions with ten distractors in the target depth plane. Thus each line for these two experiments only has four points.Note that each of these trials contained a total of 30 distractors (across all depth planes), and for most participants the target-present reaction times were more rapid than those seen in the single-plane experiment condition containing 30 distractors. This was also true for half the participants in the target-absent trials.


Overall, a comparison between the Blocked and Randomized Experiments shows similar trends: reaction times do not depend upon the number of distractors located in the depth plane with the target. This may in part be due to the ease with which the task is performed for small numbers of distractors; the single-plane experiment also showed little effect of number of distractors until there were 25 and30 distractors in the same plane.

Importantly however, target-present reaction times tend to be more rapid when distractors are spread across multiple depth planes than when they are all located with in a single depth plane. This suggests that individuals may more efficiently process information across multiple depth planes, and it gives support to the hypothesis that information in various depth channels may be processed simultaneously. While half the participants also efficiently processed the target-absent trials, the remaining participants seem to revert to a purely serial search (within and across planes) when the target was not present.

Across participants, knowledge did not systematically affect theresponses. Thus, there is mixed support for the hypothesis that individuals may independently attend to information at particular depths. Finally, there seems to be no evidence that the more well-defined depth planes (those with more distractors) drew attention away from less-defined planes. For example, for all participants across all depths, the (5T, 20, 5) condition was processed more rapidly than the (5T, 10, 15) condition.

Return to front research summary page.