Common attentional constraints in visual foraging

Abstract

Predators are known to select food of the same type in non-random sequences or "runs" that are longer than would be expected by chance. If prey are conspicuous, predators will switch between available sources, interleaving runs of different prey types. However, when prey are cryptic, predators tend to focus on one food type at a time, effectively ignoring equally available sources. This latter finding is regarded as a key indicator that animal foraging is strongly constrained by attention. It is unknown whether human foraging is equally constrained. Here, using a novel iPad task, we demonstrate for the first time that it is. Participants were required to locate and touch 40 targets from 2 different categories embedded within a dense field of distractors. When individual target items "popped-out" search was organized into multiple runs, with frequent switching between target categories. In contrast, as soon as focused attention was required to identify individual targets, participants typically exhausted one entire category before beginning to search for the other. This commonality in animal and human foraging is compelling given the additional cognitive tools available to humans, and suggests that attention constrains search behavior in a similar way across a broad range of species.

Figure 1. Example trials and foraging paths.

Panel A shows the feature foraging condition, where the task is to cancel all red and green circles while ignoring blue and yellow (or vice versa). Panel B, shows the conjunction foraging condition where the task is to cancel out all the red squares and the green circles (or vice versa). Panels C and D show typical foraging paths for the feature and conjunction conditions respectively. To explore the search space in these two conditions we suggest the reader locate the “Start” symbol in each example and follow through the sequence of symbols.

Figure 2. Hypothetical and actual distributions for the number of runs.

Panel A shows a sequence containing 7 runs and sketches three hypothetical distributions for overall foraging patterns. Note that these hypothetical distributions are simply caricatures meant to illustrate extreme strategies. If foraging were random we would expect a normal distribution of runs as the middle curve reflects. If participants shift repeatedly between targets during foraging the distribution to the right would be observed but if switching is minimized the distribution on the left would be observed. Panels C and D show the actual distributions obtained in the feature and conjunction foraging conditions, respectively, collapsed across participants. Note the change in the X-axis relative to Panel A and the difference in the Y-axis between the feature and conjunction panels. The histograms clearly show that the runs are much fewer (hence longer) in conjunction than in feature foraging.

Figure 3. Distance between consecutive taps and mean trial finishing time as a function of run length.

Panels A and B (feature foraging) and C and D (conjunction foraging) show the relationship between mean number of runs and (LEFT) distance travelled and (RIGHT) mean trial finishing time for each participant. The shaded areas represent 95% CI of the fitted lines.

Figure 4. Average run length as a function of trial for each participant in both feature and conjunction foraging conditions.

For most participants, the difference in run length (and therefore number of runs) between feature and conjunction search is clear. Note however, that participants 4, 8, 11 and 14 have essentially identical feature and conjunction patterns.