Serial Attention Mechanisms in Visual Search: A Direct Behavioral Demonstration

Abstract

In visual search, inefficient performance of human observers is typically characterized by a steady increase in reaction time with the number of array elements-the so-called set-size effect. In general, set-size effects are taken to indicate that processing of the array elements depends on limited-capacity resources, that is, it involves attention. Contrasting theories have been proposed to account for this attentional involvement, however. While some theories have attributed set-size effects to the intervention of serial attention mechanisms, others have explained set-size effects in terms of parallel, competitive architectures. Conclusive evidence in favor of one or the other notion is still lacking. Especially in view of the wide use of visual search paradigms to explore the functional neuroanatomy of attentional mechanisms in the primate brain, it becomes essential that the nature of the attentional involvement in these paradigms be clearly defined at the behavioral level. Here we report a series of experiments showing that highly inefficient search indeed recruits serial attention deployment to the individual array elements. In addition, we describe a number of behavioral signatures of serial attention in visual search that can be used in future investigations to attest a similar involvement of serial attention in other search paradigms. We claim that only after having recognized these signatures can one be confident that truly serial mechanisms are engaged in a given visual search task, thus making it amenable for exploring the functional neuroanatomy underlying its performance.

The serial-parallel dilemma: A case study in a linkage of theory and method

“…The fundamental postulate of the methodology is that positive, or same, matches can take different amounts of time (usually shorter) than negative, or different, matches. This assumption has been verified many times (for a relatively early landmark study, see Bamber, 1969;Bricolo et al, 2002, also is pertinent). It is also necessary to postulate that processing can selfterminate when that is allowed for a definite correct response, but the data can also be assayed to test this assumption (and see Townsend & Van Zandt, 1990;Van Zandt & Townsend, 1993).…”

“…Similarly, in the case of serial selfterminating processing, the survivor function for the minimum time in k trials will be the equation at the bottom of this page. It can be seen that the probability that the target is processed first will decrease, in general, as n gets larger, thus affecting the distribution on the minimum (as was noticed, e.g., by Bricolo et al, 2002). Once the properties of the distributions are known, a number of established strategies for data analysis can be utilized.…”

“…Assume that processing is orga-nized so that the items (indexed by, say, location) are operated on in exactly the same sequence on each trial, and assign Serial Position 1 to the first location processed, 2 to the second and so on. Then target-present RTs, plotted as a function of serial position, must fall precisely on the target-absent RTs as a function of n (e.g., Bricolo, Gianesini, Fanini, Bundesen, & Chelazzi, 2002;Townsend & Van Zandt, 1990). The reasoning is clearly that for, say, the third position, observers will always process the target third and, hence, their RT must equal that for n ϭ 3 on target-absent trials, and so on.…”

An Algorithm for Automatically Choosing Distractors for Recognition Based Authentication using Minimal Image Types~!2009-05-27~!2009-09-10~!2009-12-24~!

“…It is impossible, therefore, to predict how long it will take for someone to locate their own image. Various aspects will impact on the efficiency of the process, including the visual complexity of the image [16], the number of, and overall, colours the images share [17,18], the image size (which enhances discriminability [19]), the number of images in the challenge set [20], the similarity of the images to each other (both in terms of semantics and syntax) [21,22], the age of the viewer [23], the genre and task being carried out [24].…”

Postural costs of performing cognitive tasks in non-coincident reference frames

“…In which case, random search for a target requires on average only half of the display items to be checked if the target is present, but requires all items in the display to be checked if the target is absent [282]. On a parallel account, the linear increase in RTs is explained in terms of a capacity limitation in reallocating attention to the target following parallel detection of all the display items [37]. Regardless of the precise explanation as to the cause of search efficiency, both accounts agree that increasing set size increases task-load [282].…”