Howard E. Egeth

Overriding stimulus-driven attentional capture

Theeuwes (1992) found a distracting effect of irrelevant-dimension singletons in a task involving search for a known target. He argued from this that selectivity is determined solely by stimulus salience; the parallel stage of visual processing cannot provide top-down guidance to the attentive stage sufficient to permit completely selective use of task-relevant information. We argue that in the task used by Theeuwes, subjects may have adopted the strategy of searching for an odd form even though the specific target form was known. In Experiment 1, we replicated Theeuwes’s findings. Search for a circle target among diamond nontargets was disrupted by the presence of a diamond nontarget that was uniquely colored. In two subsequent experiments, we discouraged the singleton detection strategy, forcing subjects to search for the target feature. There was no distracting effect of a color singleton in these experiments, even with displays physically identical to those of Experiment 1, demonstrating that top-down selectivity is indeed possible during visual search. We conclude that goal-directed selection of a specific known featural identity may override stimulus-driven capture by salient featural singletons.

Searching for conjunctively defined targets

It has recently been proposed that in searching for a target defined as a conjunction of two or more separable features, attention must be paid serially to each stimulus in a display. Support for this comes from studies in which subjects searched for a target that shared a single feature with each of two different kinds of distractor items (e.g., a red O in a field of black Os and red Ns). Reaction time increased linearly with display size. We argue that this design may obscure evidence of selectivity in search. In an experiment in which the numbers of the two distractors were unconfounded, we find evidence that subjects can search through specified subsets of stimuli. For example, subjects told to search through just the Os to find the red O target do so without searching through Ns. Implications of selective search are discussed.

On the distinction between visual salience and stimulus-driven attentional capture

It is often assumed that the efficient detection of salient visual objects in search reflects stimulus-driven attentional capture. Evidence for this assumption, however, comes from tasks in which the salient object is task relevant and therefore may elicit a deliberate deployment of attention. In 9 experiments, participants searched for a nonsalient target (vertical among tilted bars). In each display, 1 bar was highly salient in a different dimension (e.g., color or motion). When the target and salient elements coincided only rarely, reducing the incentive to attend deliberately to the salient stimuli, response times depended little on whether the target was salient, although some interesting exceptions were observed. It is concluded that efficient selection of an element in visual search does not constitute evidence that the element captures attention in a purely stimulus-driven fashion.

Coordination of Voluntary and Stimulus-Driven Attentional Control in Human Cortex

Visual attention may be voluntarily directed to particular locations or features (voluntary control), or it may be captured by salient stimuli, such as the abrupt appearance of a new perceptual object (stimulus-driven control). Most often, however, the deployment of attention is the result of a dynamic interplay between voluntary attentional control settings (e.g., based on prior knowledge about a targets location or color) and the degree to which stimuli in the visual scene match these voluntary control settings. Consequently, nontarget items in the scene that share a defining feature with the target of visual search can capture attention, a phenomenon termed contingent attentional capture. We used functional magnetic resonance imaging to show that attentional capture by target-colored distractors is accompanied by increased cortical activity in corresponding regions of retinotopically organized visual cortex. Concurrent activation in the temporo-parietal junction and ventral frontal cortex suggests that these regions coordinate voluntary and stimulus-driven attentional control settings to determine which stimuli effectively compete for attention.

Made you blink! Contingent attentional capture produces a spatial blink

Previous studies have shown that the capture of attention by an irrelevant stimulus can be eliminated by foreknowledge of the spatial location of the relevant target stimulus. To explore whether spatial certainty is sufficient to eliminate capture, four experiments are reported in which the spatial location of the target is certain but the temporal position is uncertain. Subjects viewed a central rapid serial visual presentation stream in which a target letter was defined by a particular color (e.g., red). On critical trials, irrelevant color singletons appeared in the periphery. In Experiments 1 and 2, peripheral singletons produced a decrement in central target identification that was contingent on the match between the singleton color and the target color. Experiments 3 and 4 provided evidence that this decrement reflected a shift of spatial attention to the location of the distractor. The results suggest that spatial certainty is not sufficient to eliminate attentional capture and that attentional capture can result in a spatial “blink” that is conditional on top-down attentional control settings.

Parallel processing of multielement displays

Abstract The spatiotemporal characteristics of mechanisms that extract information from complex alphanumeric displays were investigated in a series of experiments using search and same-different detection tasks. Under several (but not all) experimental conditions the functions relating reaction time to the number of elements in the display were flat. Such data are consistent with a model in which individual elements are examined by independent parallel channels. Interestingly, this model was appropriate even in a search task in which the target was specified as any digit and the nontargets were a random assortment of letters.

It’s under control: Top-down search strategies can override attentional capture

Bacon and Egeth (1994) proposed that observed instances of attentional capture by feature singletons (e.g., color) were the result of a salience-based strategy adopted by subjects (singleton detection mode) and, thus, were not automatic. They showed that subjects could override capture by adopting strategies based on searching for specific target features (feature search mode). However, Theeuwes (2004) has recently argued that Bacon and Egeth’s results arose from experimental confounds. He elaborated a model in which attentional capture must be expected when salient distractors fall within a spatialwindow of attention. According to Theeuwes’s (2004) model, there exist two essential criteria for examining stimulus-driven capture. First, search latencies cannot increase with display size, since the search must be parallel; second, the salience of the irrelevant distractor must not be compromised by characteristics of the search display. Contrary to the predictions of Theeuwes’s (2004) model, we provide evidence that involuntary capture can be overridden when both of these criteria are met. Our results are consistent with Bacon and Egeth’s proposal.

Failure of spatial selectivity in vision

In a modified Stroop word-color paradigm, subjects were instructed to identify the color of a target patch in the center of a briefly exposed visual display and to ignore color names printed above and below the patch at distances of 1, 3, and 5 deg of visual angle. Earlier studies of selective attention to a specified location in the visual field have shown that the focus of attention subtends approximately 1 deg of angle. That is, when distractors associated with an incompatible response are spaced within about 1 deg of the target, reaction time to the target is slower than when the distractors are not associated with an incompatible response. Our results show interference with a 5-deg separation between target and noise elements. This finding casts doubt on the utility of the notion of attentional selectivity.

Toward a translational model of Stroop interference

Models of the Stroop effect that postulate single, centralized processing stations are fundamentally incorrect. A translational model is proposed in which parallel processing systems are hypothesized that are linked by a translation mechanism. Each system processes information using its own internal code specifically linked to the type of information the system processes. Information is passed between these systems via the translation mechanism. Evidence supporting this model comes from both the literature and the experiments reported in this article. Of primary importance is the demonstration that the pattern of interference that emerges is dependent upon the cognitive system used to process the information.

Do reaction time and accuracy measure the same aspects of letter recognition

Two experiments indicate that reaction time and accuracy are not always equivalent measures of the underlying processes involved in the recognition of visually presented letters. In conjunction with the results of previous work, our research suggests the following generalizations: (a) Under data-limited viewing conditions (the short exposure durations of the typical tachistoscopic task), response accuracy is sensitive to early perceptual interference between target and noise items, whereas reaction time is more sensitive to later processes involved in response interference. (b) Under resource-limited viewing conditions (the long exposure durations of the typical reaction time task), both accuracy and reaction time appear to be sensitive to processes occurring in the later rather than the earlier stages of processing. Since the two dependent measures do not always reflect the same perceptual processes, we suggest that the convergence of reaction time and accuracy within the context of a specific information processing model should be demonstrated empirically rather than assumed a priori.