COGNITIVE PSYCHOLOGY 12.97-136 (1980) A Feature-Integration Theory of Attention ANNE M.TREISMAN Universiry of British Columbia AND GARRY GELADE Oxford University A new hypothesis about the role of focused attention is proposed.The feature-integration theory of attention suggests that attention must be directed serially to each stimulus in a display whenever conjunc ons of more than on ure are neec to chara distinguish th e pos ble o A n nber or pre n a var or para both separable dimensions (shape and o)andoca nts or partsof (lines,curves,etc.in letters)as the features to be integrated into complex wholes The results were in general consistent with the hypothesis.They offer a new set o rationale fo When we open our eyes on a familiar scene,we form an immediate impression of recognizable objects,organized coherently in a spatial framework.Analysis of our experience into more elementary sensations is difficult.and appears subjectively to require ar n unusual type of pe ceptual activity.In contrast,the physiological evidence suggests that the visual scene is analyzed at an early stage by specialized populations of receptors that respond selectively to such properties as orientation,color, spatial frequency,or movement,and map these properties in different areas of the (Zeki,1976). he tr v ersy between analytic a theories perception goes b ack many years: the sociationists asserted that the experience of complex wholes is built by combining more elementary sensations,while the Gestalt psychologists claimed that the whole precedes its parts,that we initially register unitary objects and relationships and only later,if nece jects into thei e the 06 one t parts s or properties.This view is still active now (e.g.,Monahan ockhead,1977:Neisser,1976). The Gestalt belief surely conforms to the normal subjective experience ess【0Anne Treisman,Department of anada.W are grate ing Research Council.the Center for Adv Stanford,California,and the Spencer Foundation for financial support,to Melanie Meyer Martha Nagle,and Wendy Kellogg of the University of S nta Cruz for unning four of th Expenment V.and to Daniel Kahneman for many helpful comments and 0010-0285/80/010097-40505.00y0
COGNITIVE PSYCHOLOGY 12, 97-136 (1980) A Feature-Integration Theory of Attention ANNE M. TREISMAN University of British Columbia AND GARRY GELADE Oxford University A new hypothesis about the role of focused attention is proposed. The feature-integration theory of attention suggests that attention must be directed serially to each stimulus in a display whenever conjunctions of more than one separable feature are needed to characterize or distinguish the possible objects presented. A number of predictions were tested in a variety of paradigms including visual search, texture segregation, identification and localization, and using both separable dimensions (shape and color) and local elements or parts of figures (lines, curves, etc. in letters) as the features to be integrated into complex wholes. The results were in general consistent with the hypothesis. They offer a new set of criteria for distinguishing separable from integral features and a new rationale for predicting which tasks will show attention limits and which will not. When we open our eyes on a familiar scene, we form an immediate impression of recognizable objects, organized coherently in a spatial framework. Analysis of our experience into more elementary sensations is difficult, and appears subjectively to require an unusual type of perceptual activity. In contrast, the physiological evidence suggests that the visual scene is analyzed at an early stage by specialized populations of receptors that respond selectively to such properties as orientation, color, spatial frequency, or movement, and map these properties in different areas of the brain (Zeki, 1976). The controversy between analytic and synthetic theories of perception goes back many years: the Associationists asserted that the experience of complex wholes is built by combining more elementary sensations, while the Gestalt psychologists claimed that the whole precedes its parts, that we initially register unitary objects and relationships, and only later, if necessary, analyze these objects into their component parts or properties. This view is still active now (e.g., Monahan & Lockhead, 1977; Neisser, 1976). The Gestalt belief surely conforms to the normal subjective experience Address reprint requests to Anne Treisman, Department of Psychology, University of British Columbia, 2075 Wesbrook Mall, Vancouver, B.C. V6T lW5, Canada. We are grateful to the British Medical Research Council, the Canadian Natural Sciences and Engineering Research Council, the Center for Advanced Study in the Behavioral Sciences, Stanford, California, and the Spencer Foundation for financial support, to Melanie Meyer, Martha Nagle, and Wendy Kellogg of the University of Santa Cruz for running four of the subjects in Experiment V, and to Daniel Kahneman for many helpful comments and suggestions. 97 OOlO-0285/80/010097-40$05.00/O Copyright @ 19&l by Academic Press, Inc. All rights of reproduction in any form reserved
TREISMAN AND GELADE of perception.However the immediacy and directness of an impression are no guarantee that it reflects an early stage of information processing in the nervous system.It is logically possible that we become aware only of the final outcome of a complicated sequence of prior operations.“T p- do processing may de what we cons as a theory about perceptual coding it needs more objective support (Treis- man,1979). We have recently proposed a new account of attention which assumes that features come first in perception (Treisman,Sykes,Gelade,1977). In our model,which we call the featur egrati theory of attention features are regis ter ed early,automa n paral across the visual field,while objects are identified separately and only at a later stage,which requires focused attention.We assume that the visual scene is initially coded along a number of separable dimensions,such as color, orientation spatial fr requency,brightness, direction of ent.In order to e these separate representations a d to ensure the cor rect synthesis of features for each object in a complex display,stimulus locations are processed serially with focal attention.Any features which are present in the same central''fixation'of attention are combined to form a single object.Thus focal attention provides the "glue"which integrates the initially separable features into unit ary obje cts Once they have been correctly registered,the compound objects continue to be per- ceived and stored as such.However with memory decay or interterence. the features may disintegrate and "float free''once more,or perhaps recombine to form"'illusory coniunctions(Treisman.1977). We claim that,without focused attention,features cannot be related to each other. This poses a problem in ng phe omenal experie There seems to be no way we can consciously perceive an unattac shape without also giving it a color,size,brightness,and location.Yet unattended areas are not perceived as empty space.The integration theory therefore needs some clarification.Our r claim is that attention is neces sary for the correct on n of cor nctions,although unattended features are also conjo ine prior to conscious perc eption. The top-down processing of unattended features is capable of utilizing past experience and contextual information.Even when attention is directed elsewhere. we are unlikely to see a blue sun in a yellow sky.However,in the absence of focused attention and of cffective constraints on top-down processing. conjunctions of features could be formed on a random basis.These unat- tended co will ris to sory ons and physiol evide for the idea that stimuli are initially analyzed along functionally separa though not necessarily by physically distinct channels (Shepard,1964: Garner,1974;De Valois De Valois,1975).We will use the term "di- mension''to refer to the complete range of variation which is separately
98 TREISMAN AND GELADE of perception. However the immediacy and directness of an impression are no guarantee that it reflects an early stage of information processing in the nervous system. It is logically possible that we become aware only of the final outcome of a complicated sequence of prior operations. “Topdown” processing may describe what we consciously experience; as a theory about perceptual coding it needs more objective support (Treisman, 1979). We have recently proposed a new account of attention which assumes that features come first in perception (Treisman, Sykes, & Gelade, 1977). In our model, which we call the feature-integration theory of attention, features are registered early, automatically, and in parallel across the visual field, while objects are identified separately and only at a later stage, which requires focused attention. We assume that the visual scene is initially coded along a number of separable dimensions, such as color, orientation, spatial frequency, brightness, direction of movement. In order to recombine these separate representations and to ensure the correct synthesis of features for each object in a complex display, stimulus locations are processed serially with focal attention. Any features which are present in the same central “fixation” of attention are combined to form a single object. Thus focal attention provides the “glue” which integrates the initially separable features into unitary objects. Once they have been correctly registered, the compound objects continue to be perceived and stored as such. However with memory decay or interference, the features may disintegrate and “float free” once more, or perhaps recombine to form “illusory conjunctions” (Treisman, 1977). We claim that, without focused attention, features cannot be related to each other. This poses a problem in explaining phenomenal experience. There seems to be no way we can consciously “perceive” an unattached shape without also giving it a color, size, brightness, and location. Yet unattended areas are not perceived as empty space. The integration theory therefore needs some clarification. Our claim is that attention is necessary for the correcl perception of conjunctions, although unattended features are also conjoined prior to conscious perception. The top-down processing of unattended features is capable of utilizing past experience and contextual information. Even when attention is directed elsewhere, we are unlikely to see a blue sun in a yellow sky. However, in the absence of focused attention and of effective constraints on top-down processing, conjunctions of features could be formed on a random basis. These unattended couplings will give rise to “illusory conjunctions.” There is both behavioral and physiological evidence for the idea that stimuli are initially analyzed along functionally separable dimensions, although not necessarily by physically distinct channels (Shepard, 1964; Garner, 1974; De Valois & De Valois, 1975). We will use the term “dimension” to refer to the complete range of variation which is separately
ATTENTION AND FEATURE INTEGRATION 99 analyzed by some functionally independent perceptual subsystem,and "'feature to refer to a particular value on a dimension.Thus color and orientation are dimensions;red and vertical are features on those dimen- sions. Per ceptual dimension do to distinct physical dimensions Some relational aspects of physical attributes may be registered as basic features;for example we code intensity contrast rather than absolute intensity,and we may even directly sense such higher-order properties as symmetry or homogeneity.We cannot predict a priori what the e elementary words of the ptual language The existe pa r perceptual dimens from empirical criteria,such as those proposed by Shepard and by Garner.This paper will suggest several new diagnostics for the separabil- ity of dimensions.which derive from the feature-integration theory of attention.In this theory assume that integral featu are conjoined tion.Consequently,we can infer separability from a particular pattern of results in the preattentive and divided attention tasks to be described in this paper. We have stated the feature-integration hypothesis in an extreme form which s med to us initially implau It was portan th fore,to vary the par videl as pos sible,in order to max ize the gain from converging operations.We developed a number of different paradigms testing different predictions from the theory.Each experiment on its own might allow other interpretations,but the fact that all were derived as independent predictions from the same theory should alloy them if confirm d to strengthen it more than individually (1)Visual search.The visual search paradigm allows us to define a target either by its separate features or by their conjunction.If,as we assume,simple features can be detected in parallel with no attention limits.the search for t ets defined affected by vari such s(e.g or vert al)should he li ttle n the nun ors the display. Lateral interference and acuity limits should be the only factors tending to increase search times as display size is increased. perhaps by forcing serial eye fixations.In contrast,we assume that focal attention is necessary for the detection of targets that are defined by a coniunction of pron rties (e.g line in horizontal red and v gree n lines).Such should therefore be only after a serial scan of varying numbers of distractors. segregation.It scems likely that texture segregation and figure-ground grouping are preattentive,parallel processes.If so,they should be determined only by spatial discontinuities between groups of stimuli differing in separable features and not by discontinuities defined by coniunctions of features
ATTENTION AND FEATURE INTEGRATION 99 analyzed by some functionally independent perceptual subsystem, and “feature” to refer to a particular value on a dimension. Thus color and orientation are dimensions; red and vertical are features on those dimensions. Perceptual dimensions do not correspond uniquely to distinct physical dimensions. Some relational aspects of physical attributes may be registered as basic features; for example we code intensity contrast rather than absolute intensity, and we may even directly sense such higher-order properties as symmetry or homogeneity. We cannot predict a priori what the elementary words of the perceptual language may be. The existence of particular perceptual dimensions should be inferred from empirical criteria, such as those proposed by Shepard and by Garner. This paper will suggest several new diagnostics for the separability of dimensions, which derive from the feature-integration theory of attention. In this theory, we assume that integral features are conjoined automatically, while separable features require attention for their integration. Consequently, we can infer separability from a particular pattern of results in the preattentive and divided attention tasks to be described in this paper. We have stated the feature-integration hypothesis in an extreme form, which seemed to us initially quite implausible. It was important, therefore, to vary the paradigms and the predictions as widely as possible, in order to maximize the gain from converging operations. We developed a number of different paradigms testing different predictions from the theory. Each experiment on its own might allow other interpretations, but the fact that all were derived as independent predictions from the same theory should allow them, if confirmed, to strengthen it more than any could individually. (I) Visual search. The visual search paradigm allows us to define a target either by its separate features or by their conjunction. If, as we assume, simple features can be detected in parallel with no attention limits, the search for targets defined by such features (e.g., red, or vertical) should be little affected by variations in the number of distracters in the display. Lateral interference and acuity limits should be the only factors tending to increase search times as display size is increased, perhaps by forcing serial eye fixations. In contrast, we assume that focal attention is necessary for the detection of targets that are defined by a conjunction of properties (e.g., a vertical red line in a background of horizontal red and vertical green lines). Such targets should therefore be found only after a serial scan of varying numbers of distracters. (2) Texture segregation. It seems likely that texture segregation and figure-ground grouping are preattentive, parallel processes. If so, they should be determined only by spatial discontinuities between groups of stimuli differing in separable features and not by discontinuities defined by conjunctions of features
100 TREISMAN AND GELADE (3)Illusory conjunctions.If focused attention to particular objects is prevented,either because time is too short or because attention is di- rected to other obiects.the features of the unattended obiects are"free floating''with respect to one another.This allows the possibility of incor- s of featr ure more than nattended c ed.Such conj ed.o ple,the pitch and the loudness of dichotic tones are sometimes heard in the wrong combinations (Efron Yund,1974),and so are the distinctive features of dichotic syllables (Cutting,1976).In vision,subjects some- times wrongly recombine the case and the content of visual words pre- tad essively in the same cation (Lawrence. 19710.Trei (1977)obtained a large number of false-positive errors in a successive same-different matching task when the shapes and colors of two target items were interchanged in the two test stimuli.Each such interchange also added a constant to the correct response times,suggesting that the conju nction of features was che cked separately from presence of (4)Identity and location.Again,if focused attention is prevented,the features of unattended objects may be free floating spatially,as well as unrelated to one another.Thus we may detect the presence of critical features without knowing exactly where they are located,although we can certai ly ho in them rap ng a fea thi hypothe esis,be a separa e ope could logicall follow instead of preceding identification.However,the theory predicts that this could not occur with conjunctions of features.If we have cor- rectly detected or identified a particular conjunction,we must first have located it in order to focus attention on it and integrate its features.Thus t pr ident tio n for conjunction: but the o could be independent for features. (5)Interference from unattended stimuli.Unattended stimuli should be registered only at the feature level.The amount of interference or facilita- tion with an attended task that such stimuli can generate should therefore ions in There is considerable evidence in speech perceptic n that the meaning of unattended words can sometimes be registered without reach ng con scious awareness (e.g.,Corteen Wood,1972;Lewis,1970;MacKay, 1973;Treisman,Squire,Green,1974).Since words are surely defined by conjunctions,the evidence of word-recognition without attention ap- ears tocontradict our hypothesis.However,the data of these studies ses to med and relevant w ords on the attended channel trials.It may he ible a re- spons lly to be triggered by on ex pected word,without requiring exact specification of how these features
100 TREISMAN AND GELADE (3) Zllusory conjunctions. If focused attention to particular objects is prevented, either because time is too short or because attention is directed to other objects, the features of the unattended objects are “free floating” with respect to one another. This allows the possibility of incorrect combinations of features when more than one unattended object is presented. Such “illusory conjunctions” have been reported. For example, the pitch and the loudness of dichotic tones are sometimes heard in the wrong combinations (Efron & Yund, 1974), and so are the distinctive features of dichotic syllables (Cutting, 1976). In vision, subjects sometimes wrongly recombine the case and the content of visual words presented successively in the same location (Lawrence, 1971). Treisman (1977) obtained a large number of false-positive errors in a successive same-different matching task when the shapes and colors of two target items were interchanged in the two test stimuli. Each such interchange also added a constant to the correct response times, suggesting that the conjunction of features was checked separately from the presence of those features. (4) Identity and location. Again, if focused attention is prevented, the features of unattended objects may be free floating spatially, as well as unrelated to one another. Thus we may detect the presence of critical features without knowing exactly where they are located, although we can certainly home in on them rapidly. Locating a feature would, on this hypothesis, be a separate operation from identifying it, and could logically follow instead of preceding identification. However, the theory predicts that this could not occur with conjunctions of features. If we have correctly detected or identified a particular conjunction, we must first have located it in order to focus attention on it and integrate its features. Thus location must precede identification for conjunctions, but the two could be independent for features. (5) Interference from unattended stimuli. Unattended stimuli should be registered only at the feature level. The amount of interference or facilitation with an attended task that such stimuli can generate should therefore depend only on the features they comprise and should not be affected by the particular conjunctions in which those features occur. There is considerable evidence in speech perception that the meaning of unattended words can sometimes be registered without reaching conscious awareness (e.g., Corteen & Wood, 1972; Lewis, 1970; MacKay, 1973; Treisman, Squire, & Green, 1974). Since words are surely defined by conjunctions, the evidence of word-recognition without attention appears to contradict our hypothesis. However, the data of these studies indicate that responses to primed and relevant words on the unattended channel occurred only on 5-30% of trials. It may be possible for a response occasionally to be triggered by one or more features of an expected word, without requiring exact specification of how these features
ATTENTION AND FEATURE INTEGRATION 101 are combined.One study has looked at false-positive responses to rele- rds on u mor unattended channel (,1.They GSRs to words which sound ed similar to tha sho associated word when these were presented on the unattended than on the attended channel.This suggests either incomplete analysis of unattended items or incomplete sensory data. These predictions identify two clusters of results.corresponding to the rable fea and of onjunctions. able by parallel search;they are expected illusory conjunctions in the absence of attention;they can be identified without necessarily being located.and should mediate easy texture segre. gation;they can have behavioral effects even when unattended.Conjunc- tions,on the other hand,are expected to require serial search;they should perf ormance u nless ney sho ld vield highly correlated performance in the tasks of identification and location; they should prove quite ineffective in mediating texture segregation.Our aim was to test these predictions using two dimensions,form and color. whichare likely,both on physiological and on behavioral groundstobe separable.If the predi tions are confirmed a e to add sts to er s cnt ia,to form a more complete oral y1 diagnostic of separable or integral dimensions.Thus,if two physica properties are integral,they should function as a single feature in our paradigms,allowing parallel search,texture segregation,and detection without localization.If on the other hand,they are separable,their con- will r sed atter accur te perc eption e should result n illu sory conjune tions e may th paradigms to diagnose less clear-cut candidates for separability,such as the components of letters or schematic faces. The first three experiments are concerned with visual search:they compare color-shape onjunctions with disjunctive color and shape fea argets ey investigat effe of practice e and the role of feature discriminability in conjunction search,and test an alternative ac count in terms of similarity relations.Experiment IV explores the possi- bility that local elements of compound shapes (e.g.,letters)also function as separable features,requiring serial search when incorrect conjunctions could be formed.Exp tsV VI and VII ar concerned with tex ored shapes an etters as text el me periments VIII and IX explore the relation between identification and spatial localization,for targets defined by a single feature or by a con- lunction. EXPERIMENT I In an experiment reported earlier,Treisman et al.(1977)compared search for rget specified by a single feature ("pink" 0 and
ATTENTION AND FEATURE INTEGRATION 101 are combined. One study has looked at false-positive responses to relevant words on un unattended channel (Forster & Govier, 1978). They found far more GSRs to words which sounded similar to the shockassociated word when these were presented on the unattended than on the attended channel. This suggests either incomplete analysis of unattended items or incomplete sensory data. These predictions identify two clusters of results, corresponding to the perception of separable features and of conjunctions. Separable features should be detectable by parallel search; they are expected to give rise to illusory conjunctions in the absence of attention; they can be identified without necessarily being located, and should mediate easy texture segregation; they can have behavioral effects even when unattended. Conjunctions, on the other hand, are expected to require serial search; they should have no effect on performance unless focally attended; they should yield highly correlated performance in the tasks of identification and location; they should prove quite ineffective in mediating texture segregation. Our aim was to test these predictions using two dimensions, form and color, which are likely, both on physiological and on behavioral grounds, to be separable. If the predictions are confirmed, we may be able to add our tests to Garner’s criteria, to form a more complete behavioral syndrome diagnostic of separable or integral dimensions. Thus, if two physical properties are integral, they should function as a single feature in our paradigms, allowing parallel search, texture segregation, and detection without localization. If on the other hand, they are separable, their conjunctions will require focused attention for accurate perception, and its absence should result in illusory conjunctions. We may then use these paradigms to diagnose less clear-cut candidates for separability, such as the components of letters or schematic faces. The first three experiments are concerned with visual search; they compare color-shape conjunctions with disjunctive color and shape features as targets; they investigate the effects of practice and the role of feature discriminability in conjunction search, and test an alternative account in terms of similarity relations. Experiment IV explores the possibility that local elements of compound shapes (e.g., letters) also function as separable features, requiring serial search when incorrect conjunctions could be formed. Experiments V, VI, and VII are concerned with texture segregation, using colored shapes and letters as texture elements. Experiments VIII and IX explore the relation between identification and spatial localization, for targets defined by a single feature or by a conjunction. EXPERIMENT I In an experiment reported earlier, Treisman et al. (1977) compared search for targets specified by a single feature (“pink” in “brown” and
102 TREISMAN AND GELADE “purple"distractors in one condition,.“O'in“N'and“T'distractors in another)and for targets specified by a conjunction of features,a"pink O"'(Opink,in distractors Oeen and Npink).The function relating search when a single feature was target,but increased ir inearly w hen a conjunction of features was required.Experiment I replicates this study with some changes in the design,to confirm and generalize the conclusions.The most important change was in the feature search condition'subiects were now asked to search concurrently for two targets,each defined by different single fe color(blue) and a shape(S).Thus the were forced to attend to both dim sions in the feature condition as well as in the conjunction condition,although they had to check how the features were combined only when the target was a conjunction (T).The dis- tractors were identical in the two conditions(X and T. )to ensure that differences between feature and conjunction fromgrn a possibility in the previous experiment. fr ty o the distractors in the conjunction condition. Another question which has become important in evaluating information-processing hypotheses is how stably they apply across differ- ent stages of practice.Neisser,Novick,and Lazar (1963).Rabbitt (1967). and Shiffrin and Schneider(1977)have all showr alitative chang in performance as subjects repeatedly perform a particu task. Search ap- pears to change from conscious. limited capacity,serial decision making to automatic,fast,and parallel detection.LaBerge (1973)studied the effects of practice on priming in a visual successive matching task.He found that familiarity with the stimuli eventually made indepe dent of ex suggested that this was due to unitization of th features of highly fami iar stimuli.We propose that feature unitization may account also for the change with practice from serial to parallel processing in a display,in conditions in which such a change occurs.Thus the development of new unitary detectors for what were previously con nctions of features would fre eus from the constraints of focal attentior the re th 00 memory and in a phys cally present display ent I explored the possi ility that extended practice on a par ticular shape -color conjunction (Tgreen)could lead to a change from serial to parallel detection,which would suggest the possible emergence of a unitary "green T"detector. Method Stimuli.The stimulus displays were made by hand.using letter stencils and colored ink on white cards.The distractors were scattered over the card in positions which appeared random,although no systematic randomization procedure was used Four different display 15,and 30 items in cach An area s th f
102 TREISMAN AND GELADE “purple” distracters in one condition, “0” in “N” and “T” distracters in another) and for targets specified by a conjunction of features, a “pink 0” COpink in distracters O,,,, and Npink). The function relating search times to display size was flat or nonmonotonic when a single feature was sufficient to define the target, but increased linearly when a conjunction of features was required. Experiment I replicates this study with some changes in the design, to confirm and generalize the conclusions. The most important change was in the feature search condition: subjects were now asked to search concurrently for two targets, each defined by a different single feature: a color (blue) and a shape (S). Thus they were forced to attend to both dimensions in the feature condition as well as in the conjunction condition, although they had to check how the features were combined only when the target was a conjunction (Tg,.,,,). The distractors were identical in the two conditions (X,,,, and Thrown), to ensure that differences between feature and conjunction search could not result from greater heterogeneity of the distracters in the conjunction condition. (This had been a possibility in the previous experiment.) Another question which has become important in evaluating information-processing hypotheses is how stably they apply across different stages of practice. Neisser, Novick, and Lazar (1963), Rabbitt (1967), and Shiffrin and Schneider (1977) have all shown qualitative changes in performance as subjects repeatedly perform a particular task. Search appears to change from conscious, limited capacity, serial decision making to automatic, fast, and parallel detection. LaBerge (1973) studied the effects of practice on priming in a visual successive matching task. He found that familiarity with the stimuli eventually made matching independent of expectancy, and suggested that this was due to unitization of the features of highly familiar stimuli. We propose that feature unitization may account also for the change with practice from serial to parallel processing in a display, in conditions in which such a change occurs. Thus the development of new unitary detectors for what were previously conjunctions of features would free us from the constraints of focal attention to these features both in memory and in a physically present display. Experiment I explored the possibility that extended practice on a particular shape-color conjunction (T preen) could lead to a change from serial to parallel detection, which would suggest the possible emergence of a unitary “green T” detector. Method Stir?&. The stimulus displays were made by hand, using letter stencils and colored inks on white cards. The distracters were scattered over the card in positions which appeared random, although no systematic randomization procedure was used. Four different display sizes, consisting of 1,5, 15, and 30 items were used in each condition. An area subtending 14 x 8” was used for all display sizes, so that the displays with fewer items were less densely
ATTENTION AND FEATURE INTEGRATION 103 packed,but the average distance from the fovea was kept approximately constant.Each letter subten ded 0.8 ×0.6°.To ensure that the target locati across cond the arca cachcard w nd an in al hau tended 5 For each condition and each display size,eight cards were made,one with a target randomly placed in each of the resulting eight areas(top outer,top inner,left outer,left inner,right outer.etc.).Another eight cards in each condition and display size The distra and X each card as possible.The uarget in the coniunction condition was T it was either a blue letter or an S.The blue letter(Tor)matched half the distractors in shape,and the S(S orS)matched half the distractors in color.The fact that there were ou ougr conditio the conjun edure.The stimulus cards and RT were pide Development thre At the beginning of each trial,subjects viewed a plain white card in the tachistoscope,and each of their index fingers rested on a response key.The experimenter gave a verbal cady and pressec button to display a sccond whitc care and was then imm with the dominant handif arget and with the nondominnt and to respond as quickly as possible without making any errors.RT was recorded to the earest millis con on a digital timer [Advance Electronics.TC11],which was triggered by he onset of the earch array and stoppeo a response cey was Trials on whic dummytrial was given.the results of which were not recorded.Subje and whether or not they were correct after each trial:they were not however informed of the dummy trials procedure.the purpose of which was to exclude slow posterror responses from te on conjunc s and on ures in separate der conjunction targets.Six subiects did 3 blocks of 128 trials each in each condition.then two of these subjects volunteered to continue for another 4 blocks in the conjunction condition and wo for another 10 blocks,making 13 altogether(a total of 1664 trials).The mean RTs for on the cks approxin ed th e group means display sizes wasnd of differen d thus in alternative targets were,but did not know what the array size would be on any given trial. Each block contained 16 positive and 16 negative trials for each display size. Subjects.The six subjects,four men and two women,were members of the Oxford Subject Panel,ages between 24 and 29.Three of them had previously taken part in the Results Figure I shows the mean search times for the six subiects over the second and third blocks in each condition;the first block was treated as e details of linear regress t that search time inereseed linearly with dieplay ion analyses s on these
ATTENTION AND FEATURE INTEGRATION 103 packed, but the average distance from the fovea was kept approximately constant. Each letter subtended 0.8 x 0.6”. To ensure that the target locations did not vary systematically across conditions, the area of each card was divided into eight sections. This was done by superimposing a tracing of the two diagonals and an inner elliptical boundary, which subtended 8.5” x 5.5”. For each condition and each display size, eight cards were made, one with a target randomly placed in each of the resulting eight areas (top outer, top inner, left outer, left inner, right outer, etc.). Another eight cards in each condition and display size contained no target. The distracters in both conditions were Tbruwn and X,,,,, in as near equal numbers on each card as possible. The target in the conjunction condition was T,,,,,; in the feature condition, it was either a blue letter or an S. The blue letter (T hlue or X,,,,,) matched half the distracters in shape, and the S (Shruun or S g& matched half the distracters in color. The fact that there were four possible disjunctive targets in the feature condition (although the definition specified only “blue or S”), should, if anything, impair performance relative to the conjunction condition. Procedure. The stimulus cards were presented in an Electronics Development threefield tachistoscope and RT was recorded as described below. At the beginning of each trial, subjects viewed a plain white card in the tachistoscope, and each of their index fingers rested on a response key. The experimenter gave a verbal “Ready” signal and pressed a button to display a second white card bearing a central fixation spot, which remained in view for 1 set and was then immediately replaced in the field of view by a card bearing a search array. Subjects were instructed to make a key press with the dominant hand if they detected a target and with the nondominant hand otherwise, and to respond as quickly as possible without making any errors. RT was recorded to the nearest millisecond on a digital timer [Advance Electronics, TCll], which was triggered by the onset of the search array and stopped when a response key was pressed. Trials on which an error was made were repeated later in the testing session, and following each error a dummy trial was given, the results of which were not recorded. Subjects were told their RT and whether or not they were correct after each trial; they were not however informed of the dummy trials procedure, the purpose of which was to exclude slow posterror responses from the data. Each subject was tested both on conjunctions and on features in separate sessions following an ABBAAB order. Half the subjects began with the feature targets and half with the conjunction targets. Six subjects did 3 blocks of 128 trials each in each condition, then two of these subjects volunteered to continue for another 4 blocks in the conjunction condition and two for another 10 blocks, making 13 altogether (a total of 1664 trials). The mean RTs for these two subjects on the first 3 blocks closely approximated the group means. Within each block the presentation order of positive and negative trials and of different display sizes was randomized; thus in each block the subject knew what the target or the two alternative targets were, but did not know what the array size would be on any given trial. Each block contained 16 positive and 16 negative trials for each display size. Subjects. The six subjects, four men and two women, were members of the Oxford Subject Panel, ages between 24 and 29. Three of them had previously taken part in the search experiment described in Treisman et al. (1977). Results Figure 1 shows the mean search times for the six subjects over the second and third blocks in each condition; the first block was treated as practice. Table 1 gives the details of linear regression analyses on these data. The results show that search time increased linearly with display
104 TREISMAN AND GELADE SEARCH FOR COLORED SHAPES 2400T 200 COLOR SAPE 30 DISPLAY SIZE FiG.1.Search times in Experiment I. size in the conjunction condition,the linear component accounting for more than 99%of the variance due to display size.The ratio of the posi- tive to the negative slopes in the conju ondition was 0.43.which is quite close to half.These results suggest that search is serial and self- terminating with a scanning rate of about 60 msec per item.The variances increased more steeply for positive than for negative trials,and for posi- tives the root mean square of the RTs increased linearly with display size as predicted for serial self-terminating search. With the feature targ ets.the results we ery different.For the Dosi tive displays, rch times were rdly affect d by the number dis tractors,the slopes averaging only 3.1 msec.Deviations from linearity were signi cant,and the linear component accounted for only 68%of the variance due to display size.For the negatives,the linear component accounted for 96%of the variance due to display size,and departures from linearity did not reach significance.The slope was,however,less than
104 TREISMAN AND GELADE SEARCH FOR COLORED SHAPES - CONJUNCTION --- DISJUNCTION I 1 I I I 5 15 30 DISPLAY SIZE FIG. 1. Search times in Experiment I. size in the conjunction condition, the linear component accounting for more than 99% of the variance due to display size. The ratio of the positive to the negative slopes in the conjunction condition was 0.43, which is quite close to half. These results suggest that search is serial and selfterminating with a scanning rate of about 60 msec per item. The variances increased more steeply for positive than for negative trials, and for positives the root mean square of the RTs increased linearly with display size as predicted for serial self-terminating search. With the feature targets, the results were very different. For the positive displays, search times were hardly affected by the number of distractors, the slopes averaging only 3.1 msec. Deviations from linearity were significant, and the linear component accounted for only 68% of the variance due to display size. For the negatives, the linear component accounted for 96% of the variance due to display size, and departures from linearity did not reach significance. The slope was, however, less than
ATTENTION AND FEATURE INTEGRATION 105 TABLE 1 Linear Regressions of Reaction Times on Display Size in Experiment I Slope Positives 28.7 398 997 Conjunction Negatives 67.1 397 99.6 Feature Positives 3.1 448 67.9m mear Negatives 25.1 514 96.6 Feature color Positive 3.8 455 61.09 Feature shape Positive 441 78.5 half the slope for conjunction negatives.The ratio of positive to negative slopes with feature targets was only 0.12.In both conditions,all subjects showed the same pattern of results,with individuals varying mainly in the absolute values of slopes and intercepts. Errors in the feature condition averaged 2.2%false positives and 2.1% false negatives:for the conjunction condition there were 0.8%false posi- tives and 4.9%false negatives.There were no systematic effects of dis- play size on errors,except that false negatives in the conjunction condi- tion were higher for display size 30 than for 15,5,or 1(8.2%com 3.8%).The highest ear erro e conjunction condition and 3.5%in the feature condition. It is important to the theory that the difference between conjunction and feature conditions is present only when more than one stimulus is presented.The mean positive RT for display size 1 was 422 msec for the ed to 426 for and 446 tion.The negatives with display size 1were as faster in the conjunction than in the feature conditions,473 msec com- pared to 500 msec.Thus the difficulty of search for conjunctions arises only when more than one stimulus is presented. The effects of practice on conjunction search are shown in Fig.2.The positive slopes and intercepts decrease over the first 7 blocks and change little for the remaining 6 hlocks The ative slopes flu ctuate acros first 9 block <s and stab ize at ock 10.Both pos tive and negative remained linear throughout:the proportion of the variance with display
ATTENTION AND FEATURE INTEGRATION 105 TABLE 1 Linear Regressions of Reaction Times on Display Size in Experiment I Slope Intercept Percentage variance with display size which is due to linearity Positives 28.7 398 99.7 Conjunction Negatives 67.1 397 99.6 Feature Positives 3.1 448 67.9” mean Negatives 25.1 514 96.6 Feature color Positive 3.8 455 61.0” Feature shape Positive 2.5 441 78.5 ” Cases where deviations from linearity are significant at p < .Ol. The positive shape feature also deviates considerably from linearity, but the significance level here is only .08. half the slope for conjunction negatives. The ratio of positive to negative slopes with feature targets was only 0.12. In both conditions, all subjects showed the same pattern of results, with individuals varying mainly in the absolute values of slopes and intercepts. Errors in the feature condition averaged 2.2% false positives and 2.1% false negatives; for the conjunction condition there were 0.8% false positives and 4.9% false negatives. There were no systematic effects of display size on errors, except that false negatives in the conjunction condition were higher for display size 30 than for 15,5, or 1 (8.2% compared to 3.8%). The highest mean error rate for an individual subject was 5.5% in the conjunction condition and 3.5% in the feature condition. It is important to the theory that the difference between conjunction and feature conditions is present only when more than one stimulus is presented. The mean positive RT for display size 1 was 422 msec for the conjunction targets, compared to 426 msec for shape and 446 msec for color in the feature condition. The negatives with display size 1 were also faster in the conjunction than in the feature conditions, 473 msec compared to 500 msec. Thus the difficulty of search for conjunctions arises only when more than one stimulus is presented. The effects of practice on conjunction search are shown in Fig. 2. The positive slopes and intercepts decrease over the first 7 blocks and change little for the remaining 6 blocks. The negative slopes fluctuate across the first 9 blocks and stabilize at block 10. Both positive and negative slopes remained linear throughout: the proportion of the variance with display
106 TREISMAN AND GELADE SLOPES NEGATI POSITIVE FiG.2.The effects of practice on the slope and intercept of the function relating search size that was due to linearity was above 0.99 in every block except posi- tive blocks3and 1,when it wasand7,respectively.Thus there is little indication of any change in the pattern of s and no sign switch from serial to paralle I search over the 13 blocks of practice.The mean results for the two subjects who volunteered for this extensive practice were typical of the group as a whole on blocks 2 and 3(negative and positive slopes of 67 and 31,respectively,compared to the group means of 67 and 29;intercepts 423 and 389 compared to 397 and 398). Discussion rially,is the means by which the rect in egration of features into esetec especially on positive trials,fit these predictions well.Despite the major changes in the feature search condition between this experiment and the
106 TREISMAN AND GELADE P I , I I I I I I I 1 1 I I I INTERCEPTS 440 - NEGATIVE - POSITIVE - 360 - 320 - 280 - 240 : > I 1 I I I I 1 3 5 7 9 11 13 BLOCKS FIG. 2. The effects of practice on the slope and intercept of the function relating search time to display size. (The dotted lines are the data for the four subjects who did 7 sessions and the solid lines for the two subjects who continued for 13 sessions.) size that was due to linearity was above 0.99 in every block except positive blocks 3 and 12, when it was 0.98 and 0.97, respectively. Thus there is little indication of any change in the pattern of results and no sign of a switch from serial to parallel search over the 13 blocks of practice. The mean results for the two subjects who volunteered for this extensive practice were typical of the group as a whole on blocks 2 and 3 (negative and positive slopes of 67 and 31, respectively, compared to the group means of 67 and 29; intercepts 423 and 389 compared to 397 and 398). Discussion We suggested that focal attention, scanning successive locations serially, is the means by which the correct integration of features into multidimensional percepts is ensured. When this integration is not required by the task, parallel detection of features should be possible. The results, especially on positive trials, fit these predictions well. Despite the major changes in the feature search condition between this experiment and the