Academic journal article Perception and Psychophysics

Dissociating the Effects of Similarity, Salience, and Top-Down Processes in Search for Linearly Separable Size Targets

Academic journal article Perception and Psychophysics

Dissociating the Effects of Similarity, Salience, and Top-Down Processes in Search for Linearly Separable Size Targets

Article excerpt

In two experiments, we explored the role of foreknowledge on visual search for targets defined along the size continuum. Targets were of large, medium, or small size and of high or low similarity relative to the distractors. In Experiment 1, we compared search for known and unknown singleton feature targets as a function of their size and similarity to the distractors. When distractor similarity was high, target foreknowledge benefited targets at the end of the size continuum (i.e., large and small) relatively more than targets of medium size. In Experiment 2, participants were given foreknowledge of what the target was not The beneficial effect of foreknowledge for endpoint targets was reduced. The data indicate the role of top-down templates in search, which can be "tuned" more effectively for targets at the ends of feature dimensions.

Models of visual search generally distinguish between top-down and bottom-up processes (e.g., Duncan & Humphreys, 1989; Treisman & Sato, 1990;Wolfe, 1994). Bottom-up processes are generally held to function by computing difference signals for input signals from items in the visual field, on the basis of their physical similarity (e.g., Bravo & Nakayama, 1992; Wolfe, 1998). This contrasts with top-down processes, which modulate input processes via a representation of the target. One way this may happen is that a representation of the target (e.g., foreknowledge of its color) biases processing resources toward items in the visual field sharing the representation's properties (Chelazzi, Miller, Duncan, & Desimone, 1993; Duncan & Humphreys, 1989).

Hodsoll and Humphreys (2001) produced evidence for differential effects of target foreknowledge in relation to "linear separability" effects found for size-defined stimuli. A target is linearly separable from distractors if a straight line can be drawn through a feature space separating the target and distractor feature values (see Figure 1A). Stimuli are nonlinearly separable if the target falls on the straight line drawn between the feature values for the distractors (Figure 1B). D'Zmura (1991) and Bauer, Jolicoeur, and Cowan (1996a, 1996b, 1998) characterized the effects of linear separability on search for targets defined along the color dimension. Search functions were fast and parallel for linearly separable targets and distractors. In contrast, if the target and the distractors were nonlinearly separable, search was more difficult and search functions were serial in nature. Similar effects in the size domain have been reported by Wolfe and Bose (1991) and Macquistan (1994). To investigate the effects of target foreknowledge under conditions of nonlinear separability, Hodsoll and Humphreys (2001) compared search for size targets when the target identity was known with those when the target was unknown and defined by being the singleton in the display (the only item with that particular feature value). If the standard search advantage for linearly separable targets was largely due to limitations on bottom-up computations (D'Zmura, 1991), there should be little effect of target foreknowledge on the linearly separable search conditions relative to on the nonlinearly separable search conditions (i.e., on search for large and small targets vs. search for medium targets). The data refuted this prediction. Target foreknowledge had the biggest effect on the linearly separable targets. Hodsoll and Humphreys (2001) proposed that when the target is known and linearly separable from the distractors, a template or linear operator can be easily "tuned" to the target's endpoint value along the feature dimension. This selectively benefits search in such a way that there is little effect of foreknowledge on targets that are nonlinearly separable from distractors.

Given that top-down processes seem to be important in search for linearly separable targets, we first ask here how their influence is modulated by the similarity relations between the targets and the distractors. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.