Attentional Cues in Real Scenes, Saccadic Targeting, and Bayesian Priors

Abstract

Performance finding a target improves when artificial cues direct covert attention to the target's probable location or locations, but how do predictive cues help observers search for objects in real scenes? Controlling for target detectability and retinal eccentricity, we recorded observers' first saccades during search for objects that appeared in expected and unexpected locations within real scenes. As has been found with synthetic images and cues, accuracy of first saccades was significantly higher when the target appeared at an expected location rather than an unexpected location. Observers' saccades with target-absent images make it possible to distinguish two mechanisms that might mediate this effect: limited attentional resources versus differential weighting of information (Bayesian priors). Endpoints of first saccades in target-absent images were significantly closer to the expected than the unexpected locations, a result consistent with the differential-weighting model and inconsistent with limited resources being the sole mechanism underlying the effect.

Scene gist categorization by pigeons.

“…Given that scene gist recognition occurs within a single eye fixation (Eckstein et al, 2006;Torralba et al, 2006), and that fixation durations are generally tied to the processing times required for various visual recognition tasks (Nuthmann, Smith, Engbert, & Henderson, 2010;Rayner, 1998), it seems a reasonable assumption that pigeons may both require longer stimulus durations to recognize gist and tend make longer fixation durations than humans, both by roughly an order of magnitude.…”

Non-Persistent Effect of Prior Experience on Change Blindness: Investigation on Naive Observers

“…The expectation prior to performing the search that both the experienced and unexperienced changes would be presented, which was derived from the knowledge about the procedures of psychological experiments, could not explain the difference in the gaze bias between vision scientists and naive observers after they had confirmed that the location of the engine did not change. This was because the modulation of such expectation usually appears in the initial period of visual search (Eckstein, Drescher, & Shimozaki, 2006).…”

“…In this case, it was rational for the observers to persist in searching for the engine even after they met the display in which the change did not occur at the location of the engine. If only the vision scientists expected the location of the engine to change prior to performing a visual search, they would direct their eyes to the location of the engine first and then scan the other locations (Eckstein, Drescher, & Shimozaki, 2006;Torralba et al, 2006). To test these possibilities, we conducted the analysis of the time course of the gaze bias in each trial (Fig.…”

Modeling the integration of expectations in visual search with centre-surround neural fields

“…location represents where in space the saccade would be directed. The current paper compares these CSNF models to models used by Eckstein, Drescher, and Shimozaki (2006) and considers the integration of multiple targets in comparison to Findlay (1997), as well as Chen and Zelinsky (2006).…”

“…The CSNF model is a dynamic model in which each region in network excites nearby location and inhibits distant locations, thereby modeling competition for eye movements (saccades). The CSNF model was tested in a number of conditions analogous to a naturalistic search task in which the target was either (1) present in the expected location, (2) and Shimozaki (2006) found no significant differences. These discrepancies between model predictions and behavioural results are discussed.…”