Feature Integration Theory Of Attention Research Articles

Perception as a Multi-Stage Process: A Reidian Account

The starting point of this paper is Thomas Reid's anti-skepticism: our knowledge of the external world is justified. The justificatory process, in his view, starts with and relies upon one of the main faculties of the human mind: perception. Reid's theory of perception has been thoroughly studied, but there are some missing links in the explanatory chain offered by the secondary literature. In particular, I will argue that we do not have a complete picture of the mechanism of perception of bodies. The present paper, relying, in part, on a particular theory in psychology – the feature integration theory of attention – will make a contribution in this regard.

Feature confirmation in object perception: Feature integration theory 26 years on from the Treisman Bartlett lecture.

The Treisman Bartlett lecture, reported in the Quarterly Journal of Experimental Psychology in 1988, provided a major overview of the feature integration theory of attention. This has continued to be a dominant account of human visual attention to this day. The current paper provides a summary of the work reported in the lecture and an update on critical aspects of the theory as applied to visual object perception. The paper highlights the emergence of findings that pose significant challenges to the theory and which suggest that revisions are required that allow for (a) several rather than a single form of feature integration, (b) some forms of feature integration to operate preattentively, (c) stored knowledge about single objects and interactions between objects to modulate perceptual integration, (d) the application of feature-based inhibition to object files where visual features are specified, which generates feature-based spreading suppression and scene segmentation, and (e) a role for attention in feature confirmation rather than feature integration in visual selection. A feature confirmation account of attention in object perception is outlined.

Unconscious Familiarity-based Color-Form Binding: Evidence from Visual Extinction.

There is good evidence that early visual processing involves the coding of different features in independent brain regions. A major question, then, is how we see the world in an integrated manner, in which the different features are "bound" together. A standard account of this has been that feature binding depends on attention to the stimulus, which enables only the relevant features to be linked together [Treisman, A., & Gelade, G. A feature-integration theory of attention. Cognitive Psychology, 12, 97-136, 1980]. Here we test this influential idea by examining whether, in patients showing visual extinction, the processing of otherwise unconscious (extinguished) stimuli is modulated by presenting objects in their correct (familiar) color. Correctly colored objects showed reduced extinction when they had a learned color, and this color matched across the ipsi- and contralesional items (red strawberry + red tomato). In contrast, there was no reduction in extinction under the same conditions when the stimuli were colored incorrectly (blue strawberry + blue tomato; Experiment 1). The result was not due to the speeded identification of a correctly colored ipsilesional item, as there was no benefit from having correctly colored objects in different colors (red strawberry + yellow lemon; Experiment 2). There was also no benefit to extinction from presenting the correct colors in the background of each item (Experiment 3). The data suggest that learned color-form binding can reduce extinction even when color is irrelevant for the task. The result is consistent with preattentive binding of color and shape for familiar stimuli.

Saliency based on cortex-like mechanisms

A visual attention system should preferentially locate the most informative spots in complex environments. Feature-integration theory of attention plays an important role in bottom-up computational model for visual attention. This point extremely decreases the detection accuracy and also impacts the performance of the automatic visual attention model. To improve the accuracy of saliency detection in human visual attention, we propose a model based on cortex-like mechanisms. Saliency Criteria are obtained from two pathways: Shannon's entropy and sparse representation. And our model not only substantiates the bottom-up model proposed by Itti and HMAX model by Paggio, but also enriches the theory of saliency detection. We demonstrate that the proposed model achieves superior accuracy in comparison to the classical approach in static saliency map generation on real data of natural scenes and psychology stimuli patterns.

Impaired Visual Search in Rats Reveals Cholinergic Contributions to Feature Binding in Visuospatial Attention

The visual search task established the feature integration theory of attention in humans and measures visuospatial attentional contributions to feature binding. We recently demonstrated that the neuromodulator acetylcholine (ACh), from the nucleus basalis magnocellularis (NBM), supports the attentional processes required for feature binding using a rat digging-based task. Additional research has demonstrated cholinergic contributions from the NBM to visuospatial attention in rats. Here, we combined these lines of evidence and employed visual search in rats to examine whether cortical cholinergic input supports visuospatial attention specifically for feature binding. We trained 18 male Long-Evans rats to perform visual search using touch screen-equipped operant chambers. Sessions comprised Feature Search (no feature binding required) and Conjunctive Search (feature binding required) trials using multiple stimulus set sizes. Following acquisition of visual search, 8 rats received bilateral NBM lesions using 192 IgG-saporin to selectively reduce cholinergic afferentation of the neocortex, which we hypothesized would selectively disrupt the visuospatial attentional processes needed for efficient conjunctive visual search. As expected, relative to sham-lesioned rats, ACh-NBM-lesioned rats took significantly longer to locate the target stimulus on Conjunctive Search, but not Feature Search trials, thus demonstrating that cholinergic contributions to visuospatial attention are important for feature binding in rats.

Parallel perceptual enhancement and hierarchic relevance evaluation in an audio-visual conjunction task

Attention directs limited-capacity information processing resources to a subset of available perceptual representations. The mechanisms by which attention selects task-relevant representations for preferential processing are not fully known. Triesman and Gelade's [Triesman, A., Gelade, G., 1980. A feature integration theory of attention. Cognit. Psychol. 12 , 97–136.] influential attention model posits that simple features are processed preattentively, in parallel, but that attention is required to serially conjoin multiple features into an object representation. Event-related potentials have provided evidence for this model showing parallel processing of perceptual features in the posterior Selection Negativity (SN) and serial, hierarchic processing of feature conjunctions in the Frontal Selection Positivity (FSP). Most prior studies have been done on conjunctions within one sensory modality while many real-world objects have multimodal features. It is not known if the same neural systems of posterior parallel processing of simple features and frontal serial processing of feature conjunctions seen within a sensory modality also operate on conjunctions between modalities. The current study used ERPs and simultaneously presented auditory and visual stimuli in three task conditions: Attend Auditory (auditory feature determines the target, visual features are irrelevant), Attend Visual (visual features relevant, auditory irrelevant), and Attend Conjunction (target defined by the co-occurrence of an auditory and a visual feature). In the Attend Conjunction condition when the auditory but not the visual feature was a target there was an SN over auditory cortex, when the visual but not auditory stimulus was a target there was an SN over visual cortex, and when both auditory and visual stimuli were targets (i.e. conjunction target) there were SNs over both auditory and visual cortex, indicating parallel processing of the simple features within each modality. In contrast, an FSP was present when either the visual only or both auditory and visual features were targets, but not when only the auditory stimulus was a target, indicating that the conjunction target determination was evaluated serially and hierarchically with visual information taking precedence. This indicates that the detection of a target defined by audio-visual conjunction is achieved via the same mechanism as within a single perceptual modality, through separate, parallel processing of the auditory and visual features and serial processing of the feature conjunction elements, rather than by evaluation of a fused multimodal percept.

Developing an Essentially Unidimensional Test With Cognitively Designed Items

The purpose of this study was to demonstrate how cognitive and measurement principles can be integrated to create an essentially unidimensional test. Two studies were conducted. In Study 1, test questions were created by using the feature integration theory of attention to develop a cognitive model of performance and then manipulating complexity factors within the model. It was hypothesized that the complexity factors predict item difficulty. Results indicated that some complexity factors predicted difficulty in a relatively small sample. In Study 2, items developed using the cognitive model were integrated with items measuring another factor to create a multidimensional test of spatial reasoning. Results were replicated in Study 2 with a sample of 460 participants. The test met the assumption of essential unidimensionality according to DIMTEST, was moderately correlated (r = .64) with the Bennett Mechanical Comprehension Test, and showed little evidence of differential item functioning. Implications are discussed.

Putting the Cheshire Cat Together in Amblyopia. Focus on “Spatial Resolution for Feature Binding Is Impaired in Peripheral and Amblyopic Vision”

A perceptual “object” can be defined as a spatiotemporal coincidence of its different attributes: for example, an object could be a large, red, moving square. If the different attributes occur at different times, or in different places, they do not define an object. The world would be a strange

Consciousness and Perceptual Attention

Our perception of external features comprises, among others, functional and phenomenological levels. At the functional level, the perceiver’s mind processes external features according to its own causal-functional organization. At the phenomenological level, the perceiver has consciousness of external features. The question of this paper is: How do the functional and the phenomenological levels of perception relate to each other? The answer I propose is that functional states of specifically perceptual attention constitute the necessary basis for the arising of consciousness in a perceiver.Widely studied within cognitive psychology, perceptual attention is still awaiting a thoroughgoing philosophical treatment. The paper presents and draws upon Anne Treisman’s feature-integration theory of attention (cf. A. Treisman & G. Gelade, “A Feature-Integration Theory of Attention,” Cognitive Psychology, 12, 1980. Pp. 97-136). According to this theory, attentional mechanisms are responsible for the binding of perceptual features into coherent and stable objects of perception. By itself, I will claim, the theory of feature integration does not allow a straightforward reduction of consciousness to the functional processing underlying it. However, on the basis of Treisman’s theory we can produce a methodological argument for endorsing the non-reductivist thesis that attentional states constitute the necessary basis for the arising of consciousness in a perceiver. The paper closes by presenting this argument, according to which the thesis is implied by a unified account of the common representational natures of attentional and conscious states.

Early vision and visual attention

The question whether visual perception is spontaneous, sudden or is running through several phases, mediated by higher cognitive processes, was raised ever since the early work of Gestalt psychologists. In the early 1980s, Treisman proposed the feature integration theory of attention (FIT), based on the findings of neuroscience. Soon after publishing her theory a new scientific approach appeared investigating several visual perception phenomena. The most widely researched were the key constructs of FIT, like types of visual search and the role of the attention. The following review describes the main studies of early vision and visual attention.

Attention mechanisms in visual search -- an fMRI study.

The human visual system is usually confronted with many different objects at a time, with only some of them reaching consciousness. Reaction-time studies have revealed two different strategies by which objects are selected for further processing: an automatic, efficient search process, and a conscious, so-called inefficient search [Treisman, A. (1991). Search, similarity, and integration of features between and within dimensions. Journal of Experimental Psychology: Human Perception and Performance, 17, 652--676; Treisman, A., & Gelade, G. (1980). A feature integration theory of attention. Cognitive Psychology, 12, 97--136; Wolfe, J. M. (1996). Visual search. In H. Pashler (Ed.), Attention. London: University College London Press]. Two different theories have been proposed to account for these search processes. Parallel theories presume that both types of search are treated by a single mechanism that is modulated by attentional and computational demands. Serial theories, in contrast, propose that parallel processing may underlie efficient search, but inefficient searching requires an additional serial mechanism, an attentional "spotlight" (Treisman, A., 1991) that successively shifts attention to different locations in the visual field. Using functional magnetic resonance imaging (fMRI), we show that the cerebral networks involved in efficient and inefficient search overlap almost completely. Only the superior frontal region, known to be involved in working memory [Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G., & Haxby, J. V. (1998). An area specialized for spatial working memory in human frontal cortex. Science, 279, 1347--1351], and distinct from the frontal eye fields, that control spatial shifts of attention, was specifically involved in inefficient search. Activity modulations correlated with subjects' behavior best in the extrastriate cortical areas, where the amount of activity depended on the number of distracting elements in the display. Such a correlation was not observed in the parietal and frontal regions, usually assumed as being involved in spatial attention processing. These results can be interpreted in two ways: the most likely is that visual search does not require serial processing, otherwise we must assume the existence of a serial searchlight that operates in the extrastriate cortex but differs from the visuospatial shifts of attention involving the parietal and frontal regions.

Visual feature integration and focused attention: response competition from multiple distractor features.

Predictions from Treisman's feature integration theory of attention were tested in a variant of the response-competition paradigm. Subjects made choice responses to particular color-shape conjunctions (e.g., a purple cross vs. a green circle) while withholding their responses to the opposite conjunctions (i.e., a purple circle vs. a green cross). The results showed that compatibility effects were based on both distractor color and shape. For unattended distractors in preknown irrelevant positions, compatibility effects were equivalent for conjunctive distractors (e.g., a purple cross and a blue triangle) and for disjunctive distractors (e.g., a purple triangle and a blue cross). Manipulation of attention to the distractors positions resulted in larger compatibility effects from conjoined features. These results accord with Treisman's claim that correct conjunction information is unavailable under conditions of inattention, and they provide new information on response-competition effects from multiple features.

The Role of Evaluation Research in Drug Abuse Policy

and stimulus similarity, Psychological Review, 96, 433-458 (1989); H. Pashler, Target-distractor dis criminability in visual search, Perception & Psycho physics, 41, 285-292 (1987). 5. M.L. Shaw, Attending to multiple sources of information, Cognitive Psychology, 14, 353-409 (1982); M.L. Shaw, Division of attention among spatial locations: A fundamental difference between detection of letters and detection of luminance in crements, in Attention & Performance X, H. Bouma and D.G. Bouwhais, Eds. (Erlbaum, Hillsdale, NJ, 1984); for contrasting results, see P.J. Bennett and P.D. Jaye, Letter localization, not discrimination, is constrained by attention, Canadian Journal of Exper imental Psychology (in press). For more general models, see N. Graham, P. Kramer, and D. Yager, Signal-detection models for multidimensional stim uli: Probability distributions and combination rules, Journal of Mathematical Psychology, 31, 366-409 (1987). 6. M.L Posner, C.R.R. Snyder, and B.J. David son, Attention and the detection of signals, Journal of Experimental Psychology: General, 109,160-174 (1980). 7. Shaw (1984), note 5. 8. A. Treisman and G. Gelade, A feature integration theory of attention, Cognitive Psychol ogy, 12, 97-136 (1980); D. Aiken and J. Palmer, Pursuing an alternative theory of the difference be tween conjunctions and disjunctions in visual search, unpublished manuscript, University of Washington, Seattle. 9. M. Pavel, A statistical model of preattentive visual search, paper presented at the meeting of the Psychonomic Society, New Orleans (November 1990); J. Palmer, Visual search latency: The influ ence of target-distractor discriminability on the mag nitude of set-size effects, paper presented at the meeting of the Association for Research in Vision and Ophthalmology, Sarasota, FL (May 1994) (tran script available from the author). 10. W.S. Geisler and K.L. Chou, The separation of low-level and high-level factors in complex tasks: Visual search, Psychological Review, 102, 356-378 (1995).

Response effects in the perception of conjunctions of colour and form.

Two experiments addressed the question whether visual search for a target defined by a conjunction of colour and form requires a central, serial, attentional process, but detection of a single feature, such as colour, is preattentive, as proposed by the feature-integration theory of attention. Experiment 1 investigated conjunction and feature search using small array sizes of up to five elements, under conditions which precluded eye-movements, in contrast to previous studies. The results were consistent with the theory. Conjunction search showed the effect of adding distractors to the display, the slopes of the curves relating RT to array size were in the approximate ratio of 2:1, consistent with a central, serial search process, exhaustive for absence responses and self-terminating for presence responses. Feature search showed no significant effect of distractors for presence responses. Experiment 2 manipulated the response requirements in conjunction search, using vocal response in a GO-NO GO procedure, in contrast to Experiment 1, which used key-press responses in a YES-NO procedure. Strikingly, presence-response RT was not affected significantly by the number of distractors in the array. The slope relating RT to array size was 3.92. The absence RT slope was 30.56, producing a slope ratio of approximately 8:1. There was no interaction of errors with array size and the presence and absence conditions, implying that RT-error trade-offs did not produce this slope ratio. This result suggests that feature-integration theory is at least incomplete.

Before you see it, you see its parts: Evidence for feature encoding and integration in preschool children and adults

Preschool children and adults were compared in two experiments examining the basic issue of whether perceptual representations of objects are built-up from independent features along the dimensions of size and brightness. Experiment 1 was a visual search experiment. Subjects searched for targets which differed from distractors either by a single feature or by a conjunction of features. Results from preschoolers were comparable to those from adults, and were consistent with Treisman and Gelade's (1980, Cognitive Psychology, 12, 97–136) feature-integration theory of attention. Their theory states that independent features are encoded in parallel and are later combined with a spatial attention mechanism. However, children's significantly steeper conjunctive search slope indicated a slower speed of feature integration. In Experiment 2, four mathematical models of pattern recognition were tested against classification task data. The findings from both age groups were again consistent with a model assuming that size and brightness features are initially registered, and then integrated. Moreover, the data from Experiment 2 imply that perceptual growth entails small changes in the discriminability of featural representations; however, both experiments show that the operations performed on these representations are the same developmentally.

Sequential Filtering In Connectionist Networks To Eliminate Crosstalk Problems>*

AbstractA common problem with massively parallel networks is crosstalk, unwanted mutual interference associated with distributed representations in connectionist networks. Many connectionist models have been constructed, but it has turned out to be very difficult to avoid crosstalk, especially in large scale models. In this paper, several current techniques to eliminate crosstalk are surveyed and a new sequential filtering approach is described which eliminates crosstalk in parallel connectionist networks. The new method combines the feature-integration theory of attention and the theory of connectionist systems. Two kinds of filtering techniques, attention focusing and WTA (winner-take-all) subnetworks, are applied to index features in a set of individuals. The system is implemented on the Rochester Connectionist Simulator.

Locus and persistence of capacity limitations in visual information processing.

Although there is considerable evidence that stimuli such as digits and letters are extensively processed in parallel and without capacity limitations, recent data suggest that only the features of stimuli are processed in parallel. In an attempt to reconcile this discrepancy, we used the simultaneous/successive detection paradigm with stimuli from experiments indicating parallel processing and with stimuli from experiments indicating that only features can be processed in parallel. In Experiment 1, large differences between simultaneous and successive presentations were obtained with an R target among P and Q distractors and among P and B distractors, but not with digit targets among letter distractors. As predicted by the feature integration theory of attention, false-alarm rates in the simultaneous condition were much higher than in the successive condition with the R/PQ stimuli. In Experiment 2, the possibility that attention is required for any difficult discrimination was ruled out as an explanation of the discrepancy between the digit/letter results and the R/PQ and R/PB results. Experiment 3A replicated the R/PQ and R/PB results of Experiment 1, and Experiment 3B extended these findings to a new set of stimuli. In Experiment 4, we found that large amounts of consistent practice did not generally eliminate capacity limitations. From this series of experiments we strongly conclude that the notion of capacity-free letter perception has limited generality.