Feature Integration Theory and Behavioural Findings in the Visual Search Paradigm

Introduction

In the study of attention, the question of how we bind multiple visual inputs on the features of objects, including but not limited to colour, spatial orientation and shape, into objects that we recognise and identify has been an area of psychological inquiry for decades. Distinct modules of the visual cortex of the brain have been identified as areas where the independent features of objects are processed, such as V3 for shape processing, V4 for colour processing and V5 for motion processing (Zeki, 1993). This separate processing of features complicates the question of how we bind together these features to create concepts of identifiable objects.

When we visually search for features or objects (targets) in our environment we are drawn to either unique features that differ clearly from others around them or to targets incorporating combinations or conjunctures of multiple features. This visual search paradigm has been examined by testing reaction times to search for and identify targets with unique features versus targets characterised by a conjunction of features. Reaction times in feature searches were initially found to be lower than for conjunction searches which has been explained by the ease with which a unique feature can be spotted, using parallel processing in contrast to the sequential processing necessary to investigate in turn each component feature in a conjunction search (Treisman, 1985; Treisman & Gelade, 1980;). These results, among others, along with the neurological findings regarding separate modules corresponding to independent visual feature processing, have been explained by the Feature Integration Theory (FIT) set out by Anne Treisman (Treisman, 1985; Treisman, 1988; Treisman & Gelade, 1980;). This analysis will examine how FIT accounts for the findings from visual search paradigm investigations as well as how other visual search models have addressed the results.

Analysis

Over the course of her seminal works on visual search addressing identification of the foundation features of our observed environment and the binding of these features into recognised objects, Anne Treisman put forward FIT (Treisman, 1985; Treisman, 1988; Treisman & Gelade, 1980). The basic tenets of Feature Integration Theory recognise the visual search paradigm of distinct feature and conjunction visual searches and postulate that feature searches are based on parallel processing while conjunction searches rely on serial processing of multiple features. FIT takes as given that individual features are processed in distinct modules of the visual cortex (Treisman, 1988; Treisman and Gelade, 1980) and seeks to explain how this information is bound together. The theory holds that when attention is focused on a specific location in a ‘location map’ it recognises the various features present and retrieves the corresponding information from the feature maps, informed by the visual cortex modules. This feature information is then bound together into ‘object files’ which are subject to cross-referencing with stored object descriptions in memory which enables object recognition.

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!
Find out more about our Essay Writing Service

Drawing on the findings of Beck (1967), Treisman’s early findings substantiated that texture segregation between targets, based on straightforward characteristics such as colour or shape, is easier than segregation based on conjunction of separate characteristics (Treisman, 1985). Further, Treisman and Gelade (1980) found that unique features ‘pop out’ from dissimilar non-target distractors and do not appear to require focused attention to do so, suggesting parallel processing at an early visual stage. Conversely, in searches for targets that conjoin two features, reaction times increase in a linear manner related to the size of the display suggesting the requirement of turning attention to the individual features in a serial fashion (Treisman, 1985). Essentially, one location on the map is attended to at a time so focused attention shifts sequentially. In addition, cued experiments showed that valid cues had a beneficial effect on reaction time in conjoined conditions and little effect in unique feature conditions. These results supported the theory that attention is a necessary factor in the conjunction condition where the serial consideration of features is required and that parallel processing is occurring in the unique feature condition as it is unaffected by cues (Treisman & Gelade, 1980).

Other behavioural findings also factored into the development of FIT and are explained by it. For instance, visual search experiments found that reaction times were higher in conditions where subjects were asked to find targets with absent features than in conditions searching for present ones (Treisman & Souther, 1985). In essence, absent features were found to be more difficult to recognise than present ones, suggesting present features are subject to parallel processing versus a serial approach necessary to identify absent ones (Treisman & Souther, 1985). In another vein of visual search and attention inquiry, Anne Treisman and Hilary Schmidt (1982) addressed the question of attention overload or diversion and its effects on the incidence of incorrect or illusory conjunctions. Their experiments diverted attention by cuing to peripheral targets and found that colour and shape integration did not occur in this condition (Treisman & Schmidt, 1982). Overall, their series of experiments found that illusory conjunctions occurred where attention was constrained by resource limits or diversion, with colour and shape more susceptible to wrong feature recombination than solidity and size, and that illusory conjunctions were not dependent on the accuracy of feature perception or recall failure (Treisman & Schmidt, 1982). This reinforced the hypothesis that attention plays a central part in correctly identifying objects in a conjunction condition.

Many of the principal findings in the study of attention and feature binding have been addressed, examined, explained and in some cases challenged by alternative models of visual search. Wolfe, Cave and Franzel (1989) looked at the results found by Treisman and Gelade (1980) showing raised reaction times in the colour/form conjunction condition and steep reaction time x display size slopes which suggested that serial processing was at work. Their initial experiments found that in colour/form, colour/size or colour/orientation conjunction conditions, display size increases did not produce the linear increases in reaction times found in earlier studies (Wolfe et al, 1989). Their subsequent experiments found that these results were due to differences in the stimuli rather than parallel processing which would be the logical conclusion if FIT logic were to be applied (Wolfe et al, 1989). Further, they found that reaction times in the triple feature conjunction condition of colour/size/form were lower than in two feature conjunction conditions (Wolfe et al, 1989). They also identified that colour/form conjunctions can show evidence of parallel processing where colour is a central feature (Wolfe et al, 1989). The researchers draw the conclusion that not all conjunction condition searches are subject to serial search and that the conjunction search can draw on the first stage parallel feature searches to guide attention efficiently to prospective objects (Wolfe et al, 1989). Hence, the more features used to guide attention the more efficient the search. To explain their results and conclusions they propose the Guided Search Model as an alternative model of visual search (Wolfe, 1994; Wolfe et al, 1989).

The Guided Search Model holds that information on individual features from the feature maps envisaged in FIT guide attention to the location map areas where those features are present. This model is the inverse order to that of FIT where attention is serially directed to locations on the location map and only then accesses feature characteristics from the feature maps (Wolfe, 1994; Wolf et al, 1989). Wolfe et al (1989) findings regarding flat slope reaction times in conjunction search conditions which they attribute to guided searches can also explain Treisman’s (1988) behavioural findings that valid cues reduced reaction time in the conjunction condition since the Guided Search Model presumes that the more information available to guide attention the more efficient the search (Wolfe et al, 1989).

Wolfe’s Guided Search Model is built partially on the findings of James Hoffman (1979) which propounded a dual stage model for visual search. Hoffman surmised that in stage one targets are reviewed in parallel employing memory information on their features and are then subject to a stage two serial process, in rank order of similarity to memory items, for more complex comparisons (Hoffman, 1979). This theory can also explain the findings of Treisman’s later research (Treisman & Gelade, 1980) regarding the parallel processing of initial feature search and serial search for more complex conjunction searches. The first stage of Hoffman’s two-stage model also supports the texture segregation findings of Treisman (1985) which suggested an easier parallel processing stage for simple features.

The visual search paradigm findings of Treisman which led to the serial search presumption in the conjunctive condition are also challenged by other experimenters in the field. Nakayama and Silverman (1986) found evidence of parallel processing in conjunctive conditions in the case where one of the dimensions being processed is characterised by depth and another is two-dimensional. Similarly, MacLeod, Driver and Crisp (1988) found that parallel processing occurs in conjunction conditions among moving targets even when dispersed among dissimilar moving and similar stationary ones. Conversely, colour feature searches have been found to require serial searches in conditions where colour differences are small, evidenced by linear increases in search time relative to the number of non-target distractors, countering Treisman and Gelade’s (1980) findings related to parallel processing in feature search conditions (Nagy & Sanchez, 1990). It should be noted that Treisman and Gormican (1988) modified the Feature Integration Theory stance on colour feature processing to accept that such searches may be driven by serial processing when there are only slight differences in colour, calling into question a principal tenet of FIT and the visual search paradigm.

As discussed, key findings in the study of the visual search paradigm point to a dichotomous structure in visual search; feature search with findings principally supporting a parallel processing approach and conjunctive search implying serial processing. It is clear from the literature on visual search and attention that FIT, though challenged and updated, has maintained its position as a leading rationale for research findings. Nevertheless, John Duncan and Glyn Humphreys (1989) put forward an alternative visual search model based on an ongoing continuum of search efficiency calling into question the dichotomous feature/conjunction structure.

Duncan and Humphreys (1989) looked at findings of letter task experiments and identified inconsistencies when Feature Integration Theory is applied. They noted that the body of research in letter task experiments showed results varying from little or no effects of display size to larger effects, supporting the efficiency continuum approach based on similarities among targets (Duncan & Humphreys, 1989). They also found that display size showed increasing effects due to similarity of target/non-targets and non-target/non-targets (Duncan & Humphreys, 1989). The insertion of non-target similarity into experimentation and theoretical construction is a significant addition by Duncan and Humphreys to the literature and they point to the absence of this variable in prior research into visual paradigm research calling into question the findings from that research (Duncan & Humphreys, 1989). In addition, in their suite of experiments, they manipulated both target/non-target similarity and non-target/non-target similarity to produce a set of search efficiency results across a spectrum in both feature and conjunction conditions, supporting their alternative theory (Duncan & Humphreys, 1989).

In his pivotal work published in 1984, John Duncan put forward an alternative visual search model grounded in object-based attention rather than spatial/location based approaches. Duncan’s key experiment asked subjects to make single and dual judgements based on one object or dual judgements on the attributes of two objects (Duncan, 1984). He found that accuracy was high in the single object condition and lower in the two object condition suggesting an attentional capacity limitation related to the number of objects (Duncan, 1984). Consequentially, the object-based theory holds that attention can be focused on a whole object including its attributes rather than a specific location as in FIT. It also holds that stage one processes focus on the aspects of the objects themselves and in stage two attention can be focused serially on multiple objects due to capacity limitations (Duncan, 1984). Notwithstanding this alternative view of the role of attention in visual search paradigm investigation, Duncan (1984) notes that spatial/location-based and object-based theories may not be mutually exclusive, noting that Treisman (Treisman & Gelade, 1980) maintained that the focus areas for attention on location maps may be grouped into shapes indicative of objects due to Gestalt groupings (Duncan, 1984). Duncan’s acceptance of the findings from Treisman’s visual search paradigm research has been noted in subsequent research which also points to the integration of spatial/location-based and object-based theories (Muller and O’Grady, 2000).

Conclusion

As indicated in this analysis, the findings from investigation into the visual search paradigm have been examined over the course of decades. The main behavioural findings support the concept of a two stage process; feature search for unique features that ‘pop out’ from non-targets driven by parallel search processing and conjunction search for targets incorporating combinations of multiple features driven by serial processing whereby attention is directed to each feature at a location in turn. The parallel versus serial processing thesis has been examined by testing reaction times to search for and identify targets with unique features versus targets characterised by a conjunction of features. Much of the seminal experimentation and associated findings can be attributed to Anne Treisman and her associates.

Treisman’s FIT explains these findings by asserting that attention is focused on a specific location, recognises the features present and retrieves the corresponding information from the feature maps, informed by the visual cortex modules. This feature information is then bound together into ‘object files’ which are subject to cross-referencing with stored object descriptions in memory which enables object recognition. Subsequent research focusing on specific feature characteristics has indicated that not all feature search in parallel in nature nor is all conjunction search serial (Nakayama and Silverman, 1986; MacLeod, Driver and Crisp, 1988; Nagy & Sanchez, 1990). Further research on defining which features are driven by which processing models would add to exploration in the field.

Alternative visual search models also address the key findings associated with the visual search paradigm. The Guided Search Model explains these findings by postulating that information from feature maps guide attention to the location map areas where those features are present, rather than starting with attention focused on location maps before retrieving feature map information. (Wolfe, 1994; Wolf et al, 1989). Duncan and Humphreys (1989) similarity approach added to previous findings showing target/non-target similarity and non-target/non-target similarity yielding search efficiency results across a spectrum in both feature and conjunction conditions. As well, the Object-based attention theory put forward by Duncan (1984) builds on visual search paradigm findings asserting that attention can be focused on a whole object including its attributes rather than a specific location. It also holds that stage one processes focus on the aspects of the objects themselves and in stage two attention can be focused serially on multiple objects due to capacity limitations (Duncan, 1984).

Though role of attention in visual search and the binding of features has progressed significantly through the body of inquiry into the field, some open questions remain to be investigated. These include but are not limited to the specific features susceptible to parallel versus serial processing, the effects of degrees of similarity between targets/non-targets and non-targets/non-targets in individual feature recognition and whether the integration of FIT and Object-based attention theory leaves room for the dichotomous underpinning of the visual search paradigm. These studies, among others, would help to enrich the understanding of visual search and the role of attention.

References

Beck, J. (1967). Perceptual grouping produced by line figures. Perception & Psychophysics, 2(11), 491-495.

Duncan, J. (1984). Selective attention and the organization of visual information. Journal of experimental psychology: General, 113(4), 501.

Duncan, J., & Humphreys, G. W. (1989). Visual search and stimulus similarity. Psychological review, 96(3), 433.

Hoffman, J. E. (1979). A two-stage model of visual search. Perception & Psychophysics, 25(4), 319-327.

McLeod, P., Driver, J., & Crisp, J. (1988). Visual search for a conjunction of movement and form is parallel. Nature, 332(6160), 154-155.

Müller, H. J., & O'Grady, R. B. (2000). Dimension-based visual attention modulates dual-judgment accuracy in Duncan's (1984) one-versus two-object report paradigm. Journal of Experimental Psychology: Human Perception and Performance, 26(4), 1332.

Nagy, A. L., & Sanchez, R. R. (1990). Critical color differences determined with a visual search task. JOSA A, 7(7), 1209-1217.

Nakayama, K., & Silverman, G. H. (1986). Serial and parallel processing of visual feature conjunctions. Nature, 320(6059), 264-265.

Treisman, A. (1985). Preattentive processing in vision. Computer vision, graphics, and image processing, 31(2), 156-177.

Treisman, A. (1988). Features and objects: The fourteenth Bartlett memorial lecture. The Quarterly Journal of Experimental Psychology Section A, 40(2), 201-237.

Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive psychology, 12(1), 97-136.

Treisman, A., & Gormican, S. (1988). Feature analysis in early vision: evidence from search asymmetries. Psychological review, 95(1), 15.

Treisman, A., & Schmidt, H. (1982). Illusory conjunctions in the perception of objects. 1982, 14, 107-141.

Treisman, A., & Souther, J. (1985). Search asymmetry: A diagnostic for preattentive processing of separable features. Journal of Experimental Psychology: General, 114(3), 285.

Wolfe, J. M. (1994). Guided search 2.0 a revised model of visual search. Psychonomic bulletin & review, 1(2), 202-238.

Wolfe, J. M., Cave, K. R., & Franzel, S. L. (1989). Guided search: an alternative to the feature integration model for visual search. Journal of Experimental Psychology: Human perception and performance, 15(3), 419.

Zeki, S. (1993). A vision of the brain. Blackwell scientific publications.