User:Nigar Azizli/sandbox

Feature-integration theory coined by Anne Treisman posits that attention is serially directed to different features of objects in order to distinguish them as meaningful wholes . This multi-stage theory proposes that single features (colour, edges, motion, shape, etc) of objects are first perceived in the visual field through parallel processing. Following this, these single features are combined into meaningful objects through serial processing. In order for features to correctly combine into a unitary whole, focused attention must be present. With the absence of focused attention at this stage, perception of objects could lead to illusory conjunctions. However, if objects are not confounded by illusory conjunctions and are instead perceived by top-down processing (with the use of prior knowledge and experience), faster parallel processing could also take place with little use of focused attention.

History

Objects in the world are all made up of distinct features (e.g. colour, orientation, spatial frequency, movement, etc) that are processed by different areas in the brain. The fact that we bind and perceive objects as meaningful wholes rather than separate features accounts for the perceptual binding problem and can be explained by the feature-integration theory . When perceiving an unusual scene (e.g. inside of a university), we have the ability to make sense of objects based on recognition (e.g. chairs, desks, cafeteria) and past experience without consciously paying attention to single features (e.g. edge of a desk). With this being said, it would be reasonable to conclude that we perceive wholes before perceiving individual features of objects. However, physiological evidence suggests that there are specific receptors which are responsible for analyzing visual scenes and processing each of these features at an early stage. The perception that we have of whole objects is the result of us not being consciously aware of complex operations at an early stage and only recognizing the final whole object .

This controversy dates back to years ago between Associationists and Gestalt Psychologists. Associationists state that unitary wholes are built by combining individual features . Whereas, Gestalt psychologists state that the unitary wholes precede the individual features. We first perceive the wholes and later come to actually analyze the individual parts of the whole . Anne Treisman's feature-integration theory overlaps the Associationists view of visual perception in which it also states that individual features of an object are perceived separately before they are fully combined and perceived as a unitary object.

Feature-integration theory

Feature-integration theory suggests that we perceive and recognize meaningful wholes in three different ways .

1.) One way is through focal attention. Once the separate features have been identified, they can be recombined with focused attention which is a serial processing of each feature in the visual field. Recombination of these features into wholes allows us to perceive a unitary object.

2.) Alternative way is through top-down processing. This route is used when our focused attention is unavailable (e.g. either its prevented by a brief moment or there is an overload). It works best when we perceive a familiar context since this kind of processing is highly dependent on prior knowledge and experience. A unitary whole could be perceived through this route if the task doesn’t require conjunctions to be specified and we can match the object to disjunctive features in our visual display .

Focal attention and top-down processing routes often operate together. However, they are also capable of operating independently of one another.

3.) And finally, features could also be recombined and perceived without focused attention. However, most often these wholes are randomly combined and therefore lead to illusory conjunctions. Illusory conjunctions are the incorrect perceptual combination of features such as; shape, colour, orientation, etc.

Illusory conjunctions are likely to occur when two objects in the visual field are closer to one another both in proximity and also feature similarity. The more features two objects share, the easier it is for random combination to occur with the absence of focused attention. It is also important to note that illusory conjunctions can occur when two objects are distant from one another

An example of an illusory conjunction where a red circle and a green square are mistakenly perceived as a green circle.

Stages of feature-integration theory (via focused attention)

Feature-integration theory is a multi-stage theory in which individual features of objects are perceived first and later on with the help of focused attention are recognized as unitary wholes. The initial stage is the preattentive stage. During this stage, perceived features in the environment are analyzed and registered automatically, without one’s conscious awareness. These features are registered in a parallel fashion across the visual field. Since parallel, the set size also does not matter in the preattentive stage. The specific primitive visual features of a stimulus are then registered in a “feature map”. Feature maps include the salient properties of objects such as their colour, orientation, shape, etc. Specific locations on the feature map become active depending on the feature. However there is no coordination of features that are in separate feature maps. For this coordination, high order processing, such as the focused attention stage is needed .

The focused attention stage is the next stage in the feature-integration model. It is in this stage that features that have already been registered are further identified using focused attention. For combination of specific features, stimulus locations are processed serially with focal attention. Since serial attention is needed for the correct combination of features, set size in this stage does matter. The larger the set size of features, the longer the visual search takes. At this stage, all the features that are available at the same “fixation” point of attention are combined into unitary wholes and perceived as objects. This stage works like a glue and combines all the single features into a meaningful whole. The resulting outcome of this stage is the final product which is consciously perceived by individuals.

The stages of Anne Treisman's Feature-integration theory.

Experiments

In a famous experiment conducted by Treisman and Glade in 1980, participants were assigned to two conditions (conjunction and disconjunction conditions). In both conditions participants were instructed to search for the target among a display of distracters (including brown T’s and green X’s). In the conjunction condition, participants had to search for a green T (the features of the target were therefore cojoined with the distracters). Whereas, in the disconjunction condition, participants had to search for either a blue letter or letter S. Therefore the target matched the distracters either in shape (could be blue T or blue X) or colour (blue S or brown S). With that being said, it was hypothesized that participants should have a more impaired performance in the disconjunction condition in comparison to the conjunction condition. Participants were instructed to press the response key when seeing the target. Results for this study showed that reaction time increased linearly relative to display size for the conjunction condition. This suggest that search is serial and self-terminating. However, for the disconjunction condition, when the target was present among the distracters, participants were able to detect it automatically despite the display size; but when it was not present, subjects tended to scan the display in a kind of increasing linear format relative to displace size .

Criticisms

One criticism of the feature-integration theory that has been proposed by Tsal, which states that the distinct stages in this theory can be misleading because in some cases single features could be combined at the preattetive stage, without focused attention. The example he gives for this is that when detecting stimulus in a display, we can sometimes perceive peripheral cues without diverting attention to it . Although feature-integration theory asserts the use of both preatentive and focused attention stages in order to perceive unitary wholes, it nevertheless accounts for alternative routes in the absence of focused attention. Since as humans it is our instinct to perceive and recognize objects as wholes rather than single features, our past experience and knowledge can be used in the absence of focused attention. However it is important to note that this alternative route can lead to illusory conjunctions .

An alternative criticism comes from Duncan and Humphreys, in which they state that in order for grouping to occur feature maps must be combined . Whereas, Anne Treisman argues that at the feature maps in the preattentive stage do not share features and are rather very distinct. However this distinction occurs at the preattentive stage and there for is not in our conscious awareness. In order for conscious grouping of the features to occur, features are combined in the “global object file”, rather than in the feature maps. Therefore although the different features are registered in different feature maps, these features are combined later on in the focused attention stage with our conscious awareness of the grouping .

Furthermore, Wolfe has also proposed a few critiques for the feature-integration theory . Based on his yearlong experiments, Wolfe has found that the distribution of visual search is unimodal. Opposing Anne Treisman’s theory, this suggests that the distribution of the search slopes of both serial and parallel search are not quiet distinct. Therefore, it is irrelevant to characterize search tasks as either parallel or serial based on their slopes.

Finally, Mcleod, Driver, and Crisp (1988) have also found criticism against the feature-integration theory . They have proposed that during the feature conjunction processes, the reaction time in some cases varies based on the specific features rather than set size. In their study, participants were instructed to search for the moving letter ‘X’, among a set of distracters that included a stationary letter ‘X’ and a moving letter ‘O’. They found that in both of the target present and target absent trials, participants' reaction time as a function of set of size of features was depicted as a flat line. This suggests that the search time for motion and shape occurs independently of the set size, which strongly contradicts Anne Treisman’s feature-integration theory.

Alternative models

A model proposed by Duncan and Humphreys takes an alternative approach to explaining visual perception. In their model, visual search is along a continuum rather than parallel or serial. The visual search in this model depends on how easily a target can enter the visual short-term memory. The similarity between the target and nontarget stimuli is crucial for this visual search because the extent to which the target matches the “attentional template” is what is important in the end. The model is set up in a way that as long as the nontarget and nontarget similarity is high, the nontarget and target similarity will be low and therefore will result in low visual search . Although very distinct from feature-integration theory, they nevertheless do hold a similarity. The transformation of feature maps into master map location with focused attention in feature-integration theory is virtually the same thing as Duncan and Humphreys' attentional template. In both models, this a way of establishing a correct template. However it is important to note that the feature-integration model only allows for the conjunction of target features, whereas Duncan and Humpreys' model allows room for the conjunction of distracters .

The second model that has been proposed by Wolfe takes an alternative approach to visual search. Almost like the fixation point in the feature-integration theory, this model claims that in order for the correct features to combine, it is the responsibility of the parallel processes to direct the “spotlight” of attention to these features . However, there lies a huge difference between this model and the feature-integration model. Feature-integration model does not allow for the serial processing in the focused attention stage to even consider information chosen during the parallel processing stage if a certain feature does not match the given target. Whereas, Wolfe’s model allows for this to occur.