Perceptual grouping and attention in visual search for features and for objects

This article explores the effects of perceptual grouping on search for targets defined by separate features or by conjunction of features. Treisman and Gelade proposed a feature-integration theory of attention, which claims that in the absence of prior knowledge, the separable features of objects are correctly combined only when focused attention is directed to each item in turn. If items are preattentively grouped, however, attention may be directed to groups rather than to single items whenever no recombination of features within a group could generate an illusory target. This prediction is confirmed: In search for conjunctions, subjects appear to scan serially between groups rather than items. The scanning rate shows little effect of the spatial density of distractors, suggesting that it reflects serial fixations of attention rather than eye movements. Search for features, on the other hand, appears to independent of perceptual grouping, suggesting that features are detected preattentively. A conjunction target can be camouflaged at the preattentive level by placing it at the boundary between two adjacent groups, each of which shares one of its features. This suggests that preattentive grouping creates separate feature maps within each separable dimension rather than one global configuration.     

Evidence for auditory feature integration with spatially distributed items

Recent auditory research using sequentially presented, spatially fixed tones has found evidence that, as in vision for simultaneous, spatially distributed objects, attention appears to be important for the integration of perceptual features that enable the identification of auditory events. The present investigation extended these findings to arrays of simultaneously presented, spatially distributed musical tones. In the primary tasks, listeners were required to search for specific cued conjunctions of values for the features of pitch and instrument timbre. In secondary tasks, listeners were required to search for a single cued value of either the pitch or the timbre feature. In the primary tasks, listeners made frequent errors in reporting the presence or absence of target conjunctions. Probability modeling, derived from the visual search literature, revealed that the error rates in the primary tasks reflected the relatively infrequent failure to correctly identify pitch or timbre features, plus the far more frequent illusory conjunction of separately presented pitch and timbre features. Estimates of illusory conjunction rate ranged from 23% to 40%. Thus, a process must exist in audition that integrates separately registered features. The implications of the results for the processing of isolated auditory features, as well as auditory events defined by conjunctions of features, are discussed.     

Motion coherence and conjunction search: implications for guided search theory.

Feature integration theory has recently been revised with two proposals that visual conjunction search can be parallel under some circumstances--either because items with nontarget features are inhibited, or because items with target features are excited. We examined whether excitatory or inhibitory guidance controlled conjunction search for an X oscillating in one direction among Os oscillating in that direction and Xs oscillating in another. Search was affected by whether items oscillated in phase with each other, and it was exceptionally difficult when items with target motion moved out of phase with each other and items with nontarget motion moved out of phase. The results suggest that conjunction search can be guided both by excitation of target features and by inhibition of nontarget features.     

Motion coherence and conjunction search: Implications for guided search theory

Feature integration theory has recently been revised with two proposals that visual conjunction search can be parallel under some circumstances—either because items with nontarget features are inhibited, or because items with target features are excited. We examined whether excitatory or inhibitory guidance controlled conjunction search for an X oscillating in one direction among Os oscillating in that direction and Xs oscillating in another. Search was affected by whether items oscillated in phase with each other, and it was exceptionally difficult when items with target motion moved out of phase with each otherand items with nontarget motion moved out of phase. The results suggest that conjunction search can be guided both by excitation of target features and by inhibition of nontarget features.     

Serial and parallel search in pattern vision?

The nature of the processing of combinations of stimulus dimensions in human vision has recently been investigated. A study is reported in which visual search for suprathreshold positional information--vernier offsets, stereoscopic disparity, lateral separation, and orientation--was examined. The initial results showed that reaction times for visual search for conjunctions of stereoscopic disparity and either vernier offsets or orientation were independent of the number of distracting stimuli displayed, suggesting that disparity was searched in parallel with vernier offsets or orientation. Conversely, reaction times for detection of conjunctions of vernier offsets and orientation, or lateral separation and each of the other positional judgements, were related linearly to the number of distractors, suggesting serial search. However, practice has a significant effect upon the results, indicative of a shift in the mode of search from serial to parallel for all conjunctions tested as well as for single features. This suggests a reinterpretation of these and perhaps other studies that use the Treisman visual search paradigm, in terms of perceptual segregation of the visual field by disparity, motion, color, and pattern features such as colinearity, orientation, lateral separation, or size.     

Visual search at different spatial scales

Treisman and Gelade's (1980) feature-integration model claims that the search for separate ("primitive") stimulus features is parallel, but that the conjunctions of those features require serial scan. Recently, evidence has accumulated that parallel processing is not limited to these "primitive" stimulus features, but that combinations of features can also produce parallel search. In the experiments reported here, the processing of feature conjunctions was studied when the stimulus features of a combination were at different spatial scales. The patterns in the search array were composed of three cross-shaped or T-shaped (local) elements, which formed an oblique bar (the global pattern) 45 deg or 135 deg in orientation. When the target and distractors differed from each other at one spatial scale only (either in the bar orientation or in the shape of the local elements), target detection was independent of the number of distractors, i.e., the search was parallel. In the conjunction task, in which the target and distractors were defined as the combinations of the bar orientation and the element shape, i.e., both spatial scales were relevant, the detection of the target required slow serial scrutiny of the search array. It is possible that the conjunction search could not be performed in parallel because switches between the two scales (or spatial frequency channels) are linked to attention and the task required the use of both scales in order to find the target.     

Emergent features, attention, and object perception

The perceptual processing of arrows and triangles and of their component angles and lines was explored in a number of different tasks. The results suggest that some analysis of shapes into simpler parts occurs preattentively, because these parts can recombine to form illusory conjunctions when attention is divided. The presence of "emergent features," such as closure or arrow junctions, was inferred from predicted correlations in the pattern of performance across tasks and across individual subjects. Thus triangles (for most subjects) and arrows (for some subjects) behave as if they had a perceptual feature that is absent from their parts and that mediates parallel detection in search and easy texture segregation. For some subjects, circles could apparently supply the additional feature (presumably closure) required to form illusory triangles from their component lines, whereas for other subjects circles had no effect. The fact that triangle lines can form illusory conjunctions with another shape makes it unlikely that triangles are perceived holistically and strengthens the interpretation that relies on emergent features.     

Search, similarity, and integration of features between and within dimensions.

Three experiments test the claim that conjunction search is difficult only because the target resembles each distractor, whereas the distractors are highly discriminable from each other. The results show that when similarity is controlled, there is an additional difficulty created by the need to conjoin features. In addition, a target with standard values (blue and vertical) is found more easily than targets with nonstandard values (e.g., violet and tilted). Similarity may result in shared components in the functional codes that represent the targets and the distractors. A hypothesis that is based on coarse coding of features values relates the difficulty of feature search with nonstandard targets to problems in coding conjunctions of features within dimensions. Consistent with this account, illusory targets are reported not only in the usual conjunction displays but also in displays containing different features that may share the same underlying components.     

Stimulus-specific dynamics of learning in conjunction search tasks

We investigated the time course of learning in visual search for within-dimension conjunctions in homogeneous (inverted V in Vs) or heterogeneous distractors (L in 90° counterclockwise-rotated Ls and 180° rotated Ls), and for across-dimension conjunctions search (green\in green/s and red\s). Two learning sessions of 10 and 5 blocks each took place three months apart. In the first session, the reduction in mean search time was greater for within-dimension conjunctions with heterogeneous distractors. Learning endured from one session to the other for all stimuli, suggesting that learning can be long-term in search for feature conjunctions. However, when the positional regularity between elements was perturbed, the effect of prior learning was reduced for homogeneous within-dimension conjunctions. This suggests that, for these stimuli, learning facilitates grouping by positional regularity. Finally, the learning effect with homogeneous withindimension conjunctions disappeared when both target and distractors were rotated to modify the orientation of conjunction patterns. This suggests that learning improves a grouping mechanism selective for the relative orientations of combined form features.     

Distinct mechanisms account for the linear non-separability and conjunction effects in visual shape encoding

In a series of visual search experiments involving simple 2D shapes, Arguin and Saumier (2000) showed that targets that were made of conjunctions of distractor features or that were a linear combination of distractor features were searched at significantly slower rates than single-feature linearly separable targets. The present study assessed whether these conjunction and linear nonseparability effects can be attributed to distinct mechanisms. Specifically, we studied the impact of target-distractor similarity on the search rates for single-feature, conjunction, and linearly nonseparable targets. The results replicate the conjunction and linear nonseparability effects obtained by Arguin and Saumier. They also show that the conjunction and linear separability effects are differently modulated by variations in target-distractor similarity. This dissociation demonstrates that both effects are based on distinct mechanisms. The possible nature of these mechanisms is discussed.     

Guided search for triple conjunctions

A key tenet of feature integration theory and of related theories such as guided search (GS) is that the binding of basic features requires attention. This would seem to predict that conjunctions of features of objects that have not been attended should not influence search. However, Found (1998) reported that an irrelevant feature (size) improved the efficiency of search for a Color × Orientation conjunction if it was correlated with the other two features across the display, as compared to the case in which size was not correlated with color and orientation features. We examined this issue with somewhat different stimuli. We used triple conjunctions of color, orientation, and shape (e.g., search for a red, vertical, oval-shaped item). This allowed us to manipulate the number of features that each distractor shared with the target (sharing) and it allowed us to vary the total number of distractor types (and, thus, the number of groups of identical items: grouping). We found that these triple conjunction searches were generally very efficient—producing very shallow Reaction Time × Set Size slopes, consistent with strong guidance by basic features. Nevertheless, both of the variables, sharing and grouping, modulated performance. These influences were not predicted by previous accounts of GS; however, both can be accommodated in a GS framework. Alternatively, it is possible, though not necessary, to see these effects as evidence for “preattentive binding” of conjunctions.     

Target and nontarget grouping in visual search

Results recently reported by Driver, McLeod, and Dienes (1992) are used to contrast three accounts of visual search—in particular, their mechanism for easy conjunction search. In the Driver et al. study, the target was defined by a conjunction of form and movement; the key manipulation was phase in both target and nontarget motion sets. Mechanisms working separately on each display element (inhibition from nontarget features, facilitation from target features) are unable to explain large effects of phase, since this is defined only by relationships between one element and another. As implemented in the guided search model of Cave and Wolfe (1990), local suppression between similar elements is also unable to account for the results. More promising is an approach based on perceptual grouping: Elements moving in phase can be selected (target motion) or rejected (nontarget motion) as a group. Rather than a bias against elements that are similar to or grouped with their neighbors, there is a bias to treat grouped elements together.     

Conjunction of color and form without attention: evidence from an orientation-contingent color aftereffect

According to feature-integration theory (Treisman & Gelade, 1980), separable features such as color and shape exist in separate maps in preattentive vision and can be integrated only through the use of spatial attention. Many perceptual aftereffects, however, which are also assumed to reflect the features available in preattentive vision, are sensitive to conjunctions of features. One possible resolution of these views holds that adaptation to conjunctions depends on spatial attention. We tested this proposition by presenting observers with gratings varying in color and orientation. The resulting McCollough aftereffects were independent of whether the adaptation stimuli were presented inside or outside of the focus of spatial attention. Therefore, color and shape appear to be conjoined preattentively, when perceptual aftereffects are used as the measure. These same stimuli, however, appeared to be separable in two additional experiments that required observers to search for gratings of a specified color and orientation. These results show that different experimental procedures may be tapping into different stages of preattentive vision.     

Learning efficient visual search for stimuli containing diagnostic spatial configurations and color-shape conjunctions

Visual search is often slow and difficult for complex stimuli such as feature conjunctions. Search efficiency, however, can improve with training. Search for stimuli that can be identified by the spatial configuration of two elements (e.g., the relative position of two colored shapes) improves dramatically within a few hundred trials of practice. Several recent imaging studies have identified neural correlates of this learning, but it remains unclear what stimulus properties participants learn to use to search efficiently. Influential models, such as reverse hierarchy theory, propose two major possibilities: learning to use information contained in low-level image statistics (e.g., single features at particular retinotopic locations) or in high-level characteristics (e.g., feature conjunctions) of the task-relevant stimuli. In a series of experiments, we tested these two hypotheses, which make different predictions about the effect of various stimulus manipulations after training. We find relatively small effects of manipulating low-level properties of the stimuli (e.g., changing their retinotopic location) and some conjunctive properties (e.g., color-position), whereas the effects of manipulating other conjunctive properties (e.g., color-shape) are larger. Overall, the findings suggest conjunction learning involving such stimuli might be an emergent phenomenon that reflects multiple different learning processes, each of which capitalizes on different types of information contained in the stimuli. We also show that both targets and distractors are learned, and that reversing learned target and distractor identities impairs performance. This suggests that participants do not merely learn to discriminate target and distractor stimuli, they also learn stimulus identity mappings that contribute to performance improvements.     

Perceptual integration of motion and form information: evidence of parallel-continuous processing

In three visual search experiments, the processes involved in the efficient detection of motion-form conjunction targets were investigated. Experiment 1 was designed to estimate the relative contributions of stationary and moving nontargets to the search rate. Search rates were primarily determined by the number of moving nontargets; stationary nontargets sharing the target form also exerted a significant effect, but this was only about half as strong as that of moving nontargets; stationary nontargets not sharing the target form had little influence. In Experiments 2 and 3, the effects of display factors influencing the visual (form) quality of moving items (movement speed and item size) were examined. Increasing the speed of the moving items (> 1.5 degrees/sec) facilitated target detection when the task required segregation of the moving from the stationary items. When no segregation was necessary, increasing the movement speed impaired performance: With large display items, motion speed had little effect on target detection, but with small items, search efficiency declined when items moved faster than 1.5 degrees/sec. This pattern indicates that moving nontargets exert a strong effect on the search rate (Experiment 1) because of the loss of visual quality for moving items above a certain movement speed. A parallel-continuous processing account of motion-form conjunction search is proposed, which combines aspects of Guided Search (Wolfe, 1994) and attentional engagement theory (Duncan & Humphreys, 1989).     

Inhibitory guidance in visual search: The case of movement–form conjunctions

We used a probe-dot procedure to examine the roles of excitatory attentional guidance and distractor suppression in search for movement-form conjunctions. Participants in Experiment?1 completed a conjunction (moving X amongst moving Os and static Xs) and two single-feature (moving X amongst moving Os, and static X amongst static Os) conditions. "Active" participants searched for the target, whereas "passive" participants viewed the displays without responding. Subsequently, both groups located (left or right) a probe dot appearing in either an occupied or an unoccupied location. In the conjunction condition, the active group located probes presented on static distractors more slowly than probes presented on moving distractors, reversing the direction of the difference found within the passive group. This disadvantage for probes on static items was much stronger in conjunction than in single-feature search. The same pattern of results was replicated in Experiment?2, which used a go/no-go procedure. Experiment?3 extended the go/no-go procedure to the case of search for a static target and revealed increased probe localisation times as a consequence of active search, primarily for probes on moving distractor items. The results demonstrated attentional guidance by inhibition of distractors in conjunction search.     

Efficient visual search without top-down or bottom-up guidance

Two types of mechanisms have dominated theoretical accounts of efficient visual search. The first are bottom-up processes related to the characteristics of retinotopic feature maps. The second are top-down mechanisms related to feature selection. To expose the potential involvement of other mechanisms, we introduce a new search paradigm whereby a target is defined only in a context-dependent manner by multiple conjunctions of feature dimensions. Because targets in a multiconjunction task cannot be distinguished from distractors either by bottom-up guidance or top-down guidance, current theories of visual search predict inefficient search. While inefficient search does occur for the multiple conjunctions of orientation with color or luminance, we find efficient search for multiple conjunctions of luminance/size, luminance/shape, and luminance/topology. We also show that repeated presentations of either targets or a set of distractors result in much faster performance and that bottom-up feature extraction and top-down selection cannot account for efficient search on their own. In light of this, we discuss the possible role of perceptual organization in visual search. Furthermore, multiconjunction search could provide a new method for investigating perceptual grouping in visual search.     

Parallel search for conjunctions with stimuli in apparent motion

A series of experiments was conducted to determine whether apparent motion tends to follow the similarity rule (i.e. is attribute-specific) and to investigate the underlying mechanism. Stimulus duration thresholds were measured during a two-alternative forced-choice task in which observers detected either the location or the motion direction of target groups defined by the conjunction of size and orientation. Target element positions were randomly chosen within a nominally defined rectangular subregion of the display (target region). The target region was presented either statically (followed by a 250 ms duration mask) or dynamically, displaced by a small distance (18 min of arc) from frame to frame. In the motion display, the position of both target and background elements was changed randomly from frame to frame within the respective areas to abolish spatial correspondence over time. Stimulus duration thresholds were lower in the motion than in the static task, indicating that target detection in the dynamic condition does not rely on the explicit identification of target elements in each static frame. Increasing the distractor-to-target ratio was found to reduce detectability in the static, but not in the motion task. This indicates that the perceptual segregation of the target is effortless and parallel with motion but not with static displays. The pattern of results holds regardless of the task or search paradigm employed. The detectability in the motion condition can be improved by increasing the number of frames and/or by reducing the width of the target area. Furthermore, parallel search in the dynamic condition can be conducted with both short-range and long-range motion stimuli. Finally, apparent motion of conjunctions is insufficient on its own to support location decision and is disrupted by random visual noise. Overall, these findings show that (i) the mechanism underlying apparent motion is attribute-specific; (ii) the motion system mediates temporal integration of feature conjunctions before they are identified by the static system; and (iii) target detectability in these stimuli relies upon a nonattentive, cooperative, directionally selective motion mechanism that responds to high-level attributes (conjunction of size and orientation).     

Integration of local features as a function of global goodness and spacing

In two experiments, the accuracy with which subjects detected a conjunction of features was examined as a function of the spacing between items and the goodness of the axis along which they were located. In each array, two items were arranged along a vertical, a horizontal, or a diagonal axis. Based on the well-established oblique effect, the vertical and horizontal axes were considered to be good global patterns and the diagonals were considered to be poor. In Experiment 1, the two items in an array could be two horizontal lines, two vertical lines, a vertical and a horizontal line, or a plus sign with one of the single lines. In Experiment 2, a positive- and a negative-diagonal line were used as the individual features, and an "X" was used as the conjunction. The results from Experiment 1 indicated that global goodness influenced only the rate of illusory conjunctions, and not of feature errors. Illusory conjunctions of vertical and horizontal line segments were more likely to occur in vertical and horizontal arrangements. The results from Experiment 2 revealed a reversal of the effect of global goodness on the rate of illusory conjunctions: Illusory conjunctions of negative- and positive-diagonal line segments were more likely to occur in diagonal arrangements. The results of both experiments taken together showed the existence of an important and new factor that influences the likelihood that features of shape will be conjoined: the ease with which line segments conjoin when they are translated along their extent toward each other. In both experiments, greater spacing between items produced more feature-identification errors and fewer feature-integration errors than did less spacing.     

Perceptual Grouping and Motion Coherence in Visual Search

Search for a conjunction of form and motion is greatly affected by manipulations of phase in the target and nontarget motion sets. To test whether this finding can be best explained by perceptual grouping, we moved a random set of dots in phase or counterphase with target or nontarget motion. Perceptual grouping was found to have a dramatic effect on search performance. We propose that this interaction between perceptual grouping and visual search is governed by three general rules. Our data also provide convincing evidence of the preattentive organization of a visual display into surfaces defined by common motion.