The footprints of visual attention during search with 100% validand 100% invalid cuesMiguel P. Eckstein*, Binh T. Pham, Steven S. ShimozakiVision and Image Understanding Laboratory, Department of Psychology, University of California, Santa Barbara, CA 93106, USAReceived 26 August 2003; received in revised form 9 October 2003AbstractHuman performance during visual search typically improves when spatial cues indicate the possible target locations. In manyinstances, the performance improvement is quantitatively predicted by a Bayesian or quasi-Bayesian observer in which visualattention simply selects the information at the cued locations without changing the quality of processing or sensitivity and ignoresthe information at the uncued locations. Aside from the general good agreement between the effect of the cue on model and humanperformance, there has been little independent confirmation that humans are effectively selecting the relevant information. In thisstudy, we used the classification image technique to assess the effectiveness of spatial cues in the attentional selection of relevantlocations and suppression of irrelevant locations indicated by spatial cues. Observers searched for a bright target among dimmerdistractors that might appear (with 50% probability) in one of eight locations in visual white noise. The possible target location wasindicated using a 100% valid box cue or seven 100% invalid box cues in which the only potential target locations was uncued. Forboth conditions, we found statistically significant perceptual templates shaped as differences of Gaussians at the relevant locationswith no perceptual templates at the irrelevant locations. We did not find statistical significant differences between the shapes of theinferred perceptual templates for the 100% valid and 100% invalid cues conditions. The results confirm the idea that during searchvisual attention allows the observer to effectively select relevant information and ignore irrelevant information. The results for the100% invalid cues condition suggests that the selection process is not drawn automatically to the cue but can be under the observers’voluntary control.! 2004 Elsevier Ltd. All rights reserved.1. IntroductionA common finding in visual search is that humanperformance, measured either with response times oraccuracy, improves when a spatial cue indicates theprobable location of the target (Baldassi & Verghese,2002; Eckstein, 1998; Eckstein, Thomas, Palmer , &Shimozaki, 2000; Foley & Schwarz, 1998; Palmer, 1994;Palmer, Ames, & Lindsey, 1993; Posner, 1980; Solomon,Lavie, & Morgan, 1997; Verghese, 2001; Verghese &Stone, 1995). This result has been interpreted by some assuggesting that the cue allows the observer to allocateattentional resources to a single location rather thandistribute them across many locations and thereforeenhances processing at that cued (attended) location(e.g., Bashinski & Bacharach, 1980; Downing, 1988;Hawkins et al., 1990; Luck, Hillyard, Mouloua, &Hawkins, 1996). However, another hypothesis is thatwhen the target is presented among visually similar di-stractors, the presence of a 100% valid cue can benefitperformance by allowing the observer to ignore poten-tially confusable distractors. This benefit due to selec-tion is expected even without considering limitedattentional resources. This concept has been formalizedby the theory of signal detection (Green & Swets, 1966).In this theory, each element (target and distractors)elicits an internal response within the observer that issubject to noise. A distractor might be confused for thetarget because it occasionally elicits a stronger responsethan the target. In this context, attention improvesperformance by allowing the observer to select responsesfrom the relevant cued location and to ignore noisy re-sponses arising from irrelevant noise locations thatwould otherwise bring additional unnecessary variabil-ity into the decision.But how much improvement is expected in humanperformance based on these principles? One sensible*Corresponding author.E-mail address: [email protected] (M.P. Eckstein).URL: http://www.psych.ucsb.edu/~eckstein/lab/vp.html.0042-6989/$ - see front matter ! 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.visres.2003.10.026Vision Research 44 (2004) 1193–1207www.elsevier.com/locate/visresstarting point is to compare humans to an observer thatoptimally uses the cue to perform the task: the IdealBayesian observer. The optimal observer uses all priorinformation about the data, including that pr ovided bythe cue, to calculate the posterior probability of thevarious hypotheses considered (e.g., signal present vs.signal absent for a yes/no task; interval/location 1 vs.interval/location 2 for a 2 interval/alternative forcedchoice) and chooses the hypothesis with the highestposterior probability. The posterior probability is cal-culated as the product of the likelihood of the data giventhe hypothesis and the prior probability. In some cases(e.g., yes/no task in one of M locations), the Bayesianobserver decision rules are non-linear, cannot be calcu-lated using closed form expressions and require MonteCarlo computer simulations. Because running thesesimulations used to be time consuming given the avail-able computer power, historically investigators haveused other models that can be computed from closedform expressions, and are approximations to the opti-mal observer. One commonly used model bases itsdecision on the maximum response among the consid-ered responses (max model). In many instances, the idealBayesian observer results in performance improvementsof similar magnitude to the max model. Many studieshave used the set-size effects predicted by the maximummodel to compare to human performance. In manysituations, the set-size effects in human observers arecomparable to that expected from this attentionalselection model (Baldassi & Verghese, 2002; Eckstein,1998; Palmer et al., 1993; Solomon et al., 1997; Vergh-ese, 2001). In other instances where the tasks involvemore complex judgments (Poder, 1999), memory, or arapid temporal sequence of two possibly conflicting cues(Dosher & Lu, 2000a; Lu & Dosher , 2000), the set-sizeeffects present in humans were larger than predicted bythese models (Carrasco, Williams, & Yeshurun, 2002;Dosher & Lu, 2000b; Poder, 1999).A fundamental assumption of both the optimalBayesian observer and maximum response observer isthat for 100% valid cues, humans can perfectly
View Full Document
Unlocking...