This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Copyright 1990 by the American Psychological Association, Inc. 0096-3445/90/S00.75
Journal of Experimental Psychology: General 1990, Vol. 119, No. 4, 397^*11
Attention Capture by Novel Stimuli William A. Johnston, Kevin J. Hawley, Steven H. Plewe, John M. G. Elliott, and M. Jann DeWitt University of Utah
In several experiments, observers were given glimpses of 4-word arrays. Accuracy of word localization was tested after each array. Some words, calledfamiliar, appeared many times across the series of arrays; others, called novel, appeared only once. The ratio of novel to familiar words in an array ranged from 0:4 to 4:0. When familiar and novel words were not intermixed (in 0:4 and 4:0 arrays), localization accuracy was higher for familiar words. However, when they were intermixed, especially in 1:3 arrays, accuracy tended to be higher for the novel words. This novel popout effect was the outcome of the suppressed localizability of the familiar words (relative to the 0:4 baseline) and the enhanced localizability of the novel words (relative to the 4:0 baseline). Novel popout may reflect an automatic orientation of attention away from more fluently unfolding regions of the perceptual field (familiar objects) and toward less fluently unfolding regions (novel objects).
the unexpected stimulus, making it relatively likely to be seen and yielding what we call novel popout. The other is that attention tends to be captured by and apportioned among the expected stimuli, making the unexpected stimulus relatively unlikely to be seen and yielding what we call novel sink-in. Novel popout appears to have a great deal of survival value because it would render organisms sensitive to unexpected intrusions into their familiar surroundings. More generally, novel popout might serve as an important counterweight to the pervasive tendency for cognition to be biased or primed by well-established expectancies and schemata. Thus, novel popout could produce a degree of vigilance to environmental change and mitigate the excessive knowledge entrenchment that might otherwise result from the self-fulfilling power of schemata or established neural networks. The possible adaptive utility of novel popout notwithstanding, much of the empirical and theoretical literature in cognitive psychology suggests that novel stimuli should sink into rather than popout from otherwise familiar scenes. The vast literature on perceptual priming and perceptual memory indicates that familiar or expected stimuli are more easily or fluently perceived than unexpected stimuli. For example, Jacoby and Dallas (1981) showed that accuracy of identification of words exposed at near-threshold durations is higher for those that had appeared previously in the experimental session than for those that had not. Consequently, if a scene is exposed so briefly that only one or two objects can be seen, then perceptual priming and memory effects should give the edge to the familiar objects and yield novel sink-in. Indeed, this idea is the basis of the effect theories of attention developed by William James (1890/1950) and others (see Johnston & Dark, 1986). Moreover, because the fluent perception of familiar stimuli is ensured by the stimulus restoration and graceful degradation features of most network models of perception, novel sink-in might be predicted by these models as well (McClelland & Rumelhart, 1986). Unfortunately, the empirical literature is somewhat mixed with respect to this issue. The voluminous research a few decades ago on the orienting reflex and exploratory behavior clearly established that unexpected objects elicit arousal and
When we gaze aimlessly at a natural scene, what captures our attention? The factors that govern attention capture are not well known, perhaps because the study of attention has concentrated on what might be called directed attention. Typically, observers are instructed either to look for prespecified targets (target detection) or to look at prespecified locations (focused attention). Because many naturalistic tasks involve directed attention, this line of research clearly has merit. However, attention often is relatively diffuse, or nondirected, initially but is captured suddenly by certain stimuli, as when the attention of a window shopper is caught by an item in a store window display or when a hiker's attention is diverted from the trail ahead to a rustle in nearby bushes. Inasmuch as directed attention and attention capture may have fundamentally different processing underpinnings, the paucity of research on attention capture may have left a significant gap in the understanding of attention and related processes. We sought to explore attention capture by investigating what stimuli happen to "pop out" from a brief, nondirected glance at a scene. In particular, we investigated the possible automatic capture of attention by novel stimuli, that is, by stimuli that are unlikely to occur in a particular context. The general question that we posed was this: When observers have only a glimpse of an array composed of a single unexpected stimulus and several expected stimuli, how likely is the unexpected stimulus to be seen? There are two interesting possibilities. One is that attention tends to be captured by Portions of this research were supported by Air Force Office of Scientific Research Grants 87-0212 and 89-0257 to William A. Johnston. Experiments 2 and 3 formed part of a doctoral dissertation submitted by Steven H. Plewe to the Graduate School of the University of Utah. We are grateful to Frances Friedrich, Irving Biederman, Tram Neill, and an anonymous reviewer for their insightful comments and suggestions with respect to an earlier version of this article. Correspondence concerning this article should be addressed to William A. Johnston, Department of Psychology, University of Utah, Salt Lake City, Utah 84112.
397
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
398
JOHNSTON, HAWLEY, PLEWE, ELLIOTT, DnWITT
investigatory activity in animals (e.g., Berlyne, 1960; Sokolov, 1963). In addition, when confronted with two visual patterns, one novel and one familiar, human infants tend to fixate more on the novel pattern (Fantz, 1964). More recently, studies of overt and nondirected scanning of naturalistic scenes by human adults indicate that unexpected, or incongruent, objects (e.g., an octopus in a barnyard scene) are looked at longer than are expected, or congruent, objects, and perhaps earlier and more frequently (e.g., Friedman, 1979; Loftus & Mackworth, 1978). Although the tasks used in many of these studies called for nondirected attention rather than directed attention, the scenes used were visible for at least several seconds. Thus, it is not clear that the results reflect the immediate and automatic seizure of attention by the novel objects. Rather than having their attention instantly captured by novel objects, observers might have encountered these objects somewhat late in their initial perusal of the scenes and only then directed their full attention to them. In fact, still more recent research indicates that the detection of unexpected objects is suppressed when only a glance at a scene is allowed. Biederman, Mezzanotte, and Rabinowitz (1982, Experiment 1) examined the effect of object incongruence on the detection of targets in scenes exposed for only 150 ms. A drawing of a naturalistic scene, such as a livingroom or street scene, was preceded by the name of a target object, such as COUCH, and was followed by a location probe. The observers' task was to say whether or not the target object had appeared at the probed location. Speed and accuracy of detection decreased to the extent that the object was incongruent with the rest of the scene. For example, a couch was more detectable when it was appropriately located in a livingroom scene than when it hovered above the buildings in a street scene. Moreover, the detectability of a congruent object was not suppressed by the presence of an incongruent object elsewhere in the scene. These findings argue in favor of the sink-in, rather than popout, of novel perturbations of familiar environments. Although the Biederman et al. (1982) study differed from prior research in its use of brief exposures, it differed also in its use of a directed-attention (viz., target-detection) task rather than a nondirected task. The present research examined the way that attention is distributed spontaneously between expected, or familiar, words and unexpected, or novel, words when the arrays are exposed briefly and the observers are encouraged to apprehend as many words as they can. In contrast to Biederman et al. (1982), we observed novel popout rather than novel sink-in; attention appeared to be captured by a single unexpected word when it was presented in the company of several expected words. In the remainder of this article, we summarize some preliminary observations of novel popout, describe some new explorations of the phenomenon, and consider related phenomena and possible accounts of novel popout.
Preliminary Studies Because our preliminary findings are not reported in easily accessible sources (DeWitt & Johnston, 1989; Farah, 1987; Johnston & Farah, 1986), they are collectively summarized in some detail here. Observers were asked to monitor a
succession of four-word arrays. Each word was exposed for 500 ms in one of four array locations and then backward masked for 50 ms by a row of Xs. The stimulus onset asynchrony (SOA) between arrays was 1,000 ms. Approximately 10% of the arrays were followed by a location probe in lieu of another array. A probe appeared for 3,000 ms as a ? in one of the locations of an otherwise empty array. The observers' task was to report aloud the word that had occupied the probed location. Two array compositions, or ratios of novel to familiar words—4:0 and 1:3—were manipulated between groups of observers. Novel words appeared only once in the entire series of arrays and familiar words appeared many times. Thus, the 4:0 arrays contained all novel words, and the 1:3 arrays contained one novel word along with three familiar words. The array locations of the words were determined randomly on each trial. For some of the observers of 1:3 arrays, the three familiar words were semantically related (e.g., MORNING, NOON, NIGHT); for other observers, the familiar words were not semantically related (e.g., ISLAND, TABLE, WEEK). In all, then, there were three groups, each composed of 36 observers: One group experienced only 4:0 arrays, one experienced only 1:3 arrays containing semantically related familiar words, and one experienced only 1:3 arrays containing semantically nonrelated familiar words. It is important to note that novelty and familiarity were defined contextually. The "novel" word in a 1:3 array was a familiar English word; however, unlike the "familiar" words with which it was displayed, its appearance could not be anticipated in the context of the experiment. Percentage correct responses to the location probes (report accuracy) is summarized in Figure 1. Novel popout can show up in two ways: within an array or between arrays. Withinarray popout is manifested as higher report accuracy for the novel words in 1:3 arrays than for the familiar words in those same arrays. Between-arrays popout is manifested as higher report accuracy for novel words in 1:3 arrays than for those in 4:0 arrays. As Figure 1 reveals, both forms of novel popout were observed, but the between-arrays effect was the larger of the two. 70
cn
\/X
g BO QUJ DC
I I Familiar Words
Novel Words
50 U UJ
rr rr o
40
CJ
a? 30
1:3 R 1:3 NR 4:0 NOVEL:FAMILIAR RATIO
Figure 1. Report accuracy in the preliminary research for novel and familiar words as a function of novel:familiar ratio and whether (1:3 R) or not (1:3 NR) the familiar words were preexperimentally related.
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
NOVEL POPOUT
Although the size of the within-array popout effect was comparable for the two types of 1:3 array, overall report accuracy was much higher and the between-arrays effect was much larger when the familiar words were related than when they were nonrelated. We recruit this observation later as support for the argument that the degree of novel popout in 1:3 arrays is a function of the fluency with which the familiar words in those arrays are perceived (see General Discussion). Suffice it to note at this juncture that, judging from the reportaccuracy data, perceptual fluency of familiar words appears to be greater if they are semantically related than if they are not. In summary, attention appears to be oriented rapidly and automatically to novel words in 1:3 arrays, especially when the familiar words are semantically related. This novel-popout phenomenon is somewhat curious, if not anomalous, because it appears to be directly at odds not only with the novel sinkin effect observed by Biederman et al. (1982), but also with the theoretical implications noted previously of perceptual priming and perceptual memory. As noted earlier, when individual words are exposed briefly, accuracy of identification is higher for those that had been seen before in the experiment (e.g., familiar words) than for those that had not (e.g., novel words). Thus, when 1:3 arrays of words are exposed briefly, it is curious that accuracy of identification is not higher for the familiar words than for the novel words. The new experiments sought to establish the boundaries and generality of novel popout and to explore its processing bases.
New Experiments Four new experiments are reported in this section. Experiment 1 tested whether the novel-popout effects would survive a number of task and procedural changes, compared novel popout with some well-known phenomena of directed attention, and tested for a complementary familiar popout effect in 3:1 (one-familiar) arrays. Experiment 2 traced the temporal development of novel popout and examined whether it still occurs under conditions that preclude eye movements. Experiments 3 and 4 investigated several possible bases of novel popout including figure-ground contrast, perceptual satiation of familiar words, and withdrawal of attention from familiar words.
Experiment 1 One of the changes introduced in Experiment 1 involved the nature of the probes. Rather than probing for the word that occupied a particular location, we probed for the location that a particular word had occupied. One effect of this change was to eliminate the guessing advantage for familiar words. In the preliminary research, familiar-word intrusions accounted for approximately 78% of the overt errors to the location probes. Apparently, when observers tried to guess what word was in the probed location, they were likely to pick one of the familiar words. This strategy should have inflated report accuracy for familiar words and attenuated the withinarray popout effect. With the new probe procedure, the ac-
399
curacy of localization yielded by pure guessing would approximate 25% for familiar and novel probes alike. Other procedural changes intended to enhance the power and efficiency of the design were that array composition was manipulated within rather than between observers and that probes were presented after every array rather than after just some of them. Besides these procedural changes, some new conditions were examined in Experiment 1. In addition to nondirected attention, two forms of directed attention were examined: target localization and focused attention. The inclusion of these tasks made it possible to compare the popout of novel words in nondirected attention with the popout of target and focal words in directed attention. In addition, the nondirectedattention task included 3:1 arrays as well as 1:3 and 4:0 arrays. The 3:1 arrays made it possible to test for a within-array familiar-popout effect.
Method Subjects and design. The observers were 513 United States Air Force (USAF) recruits who participated in the study as one of their regular assignments.' They were all high school graduates, 17 to 27 years of age, and approximately 88% of them were males. All observers performed under five different attention conditions: three nondirected-attention conditions (4:0,1:3, and 3:1 arrays) and two directedattention conditions (target localization and focused attention). Word type (novel/familiar, focal/nonfocal, or target/nontarget) was a nested factor in all but the 4:0 condition. Apparatus and procedures. The experiment was performed in the USAF Human Resources Laboratory at Brooks Air Force Base, San Antonio, Texas. Up to 30 observers were run at a time in separate testing carrels, each one equipped with a TERAK 8510A microcomputer, a television monitor, a keyboard, and an instruction booklet. The observers were able to progress through the experiment in an average time of about 1.5 hr. We describe the procedures for the nondirected-attention task first and then indicate the departures from those procedures for the directed-attention tasks. Words and other visual characters appeared in a four-cell array that was centered on the television screen. From a viewing distance of 60 cm, the array subtended visual angles of about 1.90 degrees vertically and 5 degrees horizontally. Alphanumeric characters subtended angles of up to 0.38 degrees vertically and 0.24 degrees horizontally. Each stimulus presented on a trial was centered in one of the array locations. A trial comprised a succession of five arrays: warning, attention, mask, probe, and feedback. These arrays are illustrated in Figure 2. Each location in a warning array contained a string of three asterisks. A warning array was exposed for 200 ms and was followed 600 ms later by an attention array. Each location in the attention array contained a different word. The attention array was presented for 383 ms and followed, 17 ms later, by a mask array, yielding a virtual exposure duration of 400 ms. The mask array contained strings of 9 Xs in lieu of words and was exposed for 100 ms. After a blank interval of 500 ms, a probe array was presented in which one of the words from the attention array reappeared in all four locations. The observers' task was to indicate in which location of the attention array
' A large number of observers was tested to permit analyses of individual differences in attention. However, these analyses fall outside the scope of, and are not summarized in, the present article.
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
400
JOHNSTON, HAWLEY, PLEWE, ELLIOTT, DEWITT
1. warning array (200 ms)
(600 ms) CHIN
*•
TOPIC (17 ms)
xxxxxxxxx 3. mask array
| XXXXXXXXX ] XXXXXXXXX |
Copyright 1990 by the American Psychological Association, Inc. 0096-3445/90/S00.75
Journal of Experimental Psychology: General 1990, Vol. 119, No. 4, 397^*11
Attention Capture by Novel Stimuli William A. Johnston, Kevin J. Hawley, Steven H. Plewe, John M. G. Elliott, and M. Jann DeWitt University of Utah
In several experiments, observers were given glimpses of 4-word arrays. Accuracy of word localization was tested after each array. Some words, calledfamiliar, appeared many times across the series of arrays; others, called novel, appeared only once. The ratio of novel to familiar words in an array ranged from 0:4 to 4:0. When familiar and novel words were not intermixed (in 0:4 and 4:0 arrays), localization accuracy was higher for familiar words. However, when they were intermixed, especially in 1:3 arrays, accuracy tended to be higher for the novel words. This novel popout effect was the outcome of the suppressed localizability of the familiar words (relative to the 0:4 baseline) and the enhanced localizability of the novel words (relative to the 4:0 baseline). Novel popout may reflect an automatic orientation of attention away from more fluently unfolding regions of the perceptual field (familiar objects) and toward less fluently unfolding regions (novel objects).
the unexpected stimulus, making it relatively likely to be seen and yielding what we call novel popout. The other is that attention tends to be captured by and apportioned among the expected stimuli, making the unexpected stimulus relatively unlikely to be seen and yielding what we call novel sink-in. Novel popout appears to have a great deal of survival value because it would render organisms sensitive to unexpected intrusions into their familiar surroundings. More generally, novel popout might serve as an important counterweight to the pervasive tendency for cognition to be biased or primed by well-established expectancies and schemata. Thus, novel popout could produce a degree of vigilance to environmental change and mitigate the excessive knowledge entrenchment that might otherwise result from the self-fulfilling power of schemata or established neural networks. The possible adaptive utility of novel popout notwithstanding, much of the empirical and theoretical literature in cognitive psychology suggests that novel stimuli should sink into rather than popout from otherwise familiar scenes. The vast literature on perceptual priming and perceptual memory indicates that familiar or expected stimuli are more easily or fluently perceived than unexpected stimuli. For example, Jacoby and Dallas (1981) showed that accuracy of identification of words exposed at near-threshold durations is higher for those that had appeared previously in the experimental session than for those that had not. Consequently, if a scene is exposed so briefly that only one or two objects can be seen, then perceptual priming and memory effects should give the edge to the familiar objects and yield novel sink-in. Indeed, this idea is the basis of the effect theories of attention developed by William James (1890/1950) and others (see Johnston & Dark, 1986). Moreover, because the fluent perception of familiar stimuli is ensured by the stimulus restoration and graceful degradation features of most network models of perception, novel sink-in might be predicted by these models as well (McClelland & Rumelhart, 1986). Unfortunately, the empirical literature is somewhat mixed with respect to this issue. The voluminous research a few decades ago on the orienting reflex and exploratory behavior clearly established that unexpected objects elicit arousal and
When we gaze aimlessly at a natural scene, what captures our attention? The factors that govern attention capture are not well known, perhaps because the study of attention has concentrated on what might be called directed attention. Typically, observers are instructed either to look for prespecified targets (target detection) or to look at prespecified locations (focused attention). Because many naturalistic tasks involve directed attention, this line of research clearly has merit. However, attention often is relatively diffuse, or nondirected, initially but is captured suddenly by certain stimuli, as when the attention of a window shopper is caught by an item in a store window display or when a hiker's attention is diverted from the trail ahead to a rustle in nearby bushes. Inasmuch as directed attention and attention capture may have fundamentally different processing underpinnings, the paucity of research on attention capture may have left a significant gap in the understanding of attention and related processes. We sought to explore attention capture by investigating what stimuli happen to "pop out" from a brief, nondirected glance at a scene. In particular, we investigated the possible automatic capture of attention by novel stimuli, that is, by stimuli that are unlikely to occur in a particular context. The general question that we posed was this: When observers have only a glimpse of an array composed of a single unexpected stimulus and several expected stimuli, how likely is the unexpected stimulus to be seen? There are two interesting possibilities. One is that attention tends to be captured by Portions of this research were supported by Air Force Office of Scientific Research Grants 87-0212 and 89-0257 to William A. Johnston. Experiments 2 and 3 formed part of a doctoral dissertation submitted by Steven H. Plewe to the Graduate School of the University of Utah. We are grateful to Frances Friedrich, Irving Biederman, Tram Neill, and an anonymous reviewer for their insightful comments and suggestions with respect to an earlier version of this article. Correspondence concerning this article should be addressed to William A. Johnston, Department of Psychology, University of Utah, Salt Lake City, Utah 84112.
397
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
398
JOHNSTON, HAWLEY, PLEWE, ELLIOTT, DnWITT
investigatory activity in animals (e.g., Berlyne, 1960; Sokolov, 1963). In addition, when confronted with two visual patterns, one novel and one familiar, human infants tend to fixate more on the novel pattern (Fantz, 1964). More recently, studies of overt and nondirected scanning of naturalistic scenes by human adults indicate that unexpected, or incongruent, objects (e.g., an octopus in a barnyard scene) are looked at longer than are expected, or congruent, objects, and perhaps earlier and more frequently (e.g., Friedman, 1979; Loftus & Mackworth, 1978). Although the tasks used in many of these studies called for nondirected attention rather than directed attention, the scenes used were visible for at least several seconds. Thus, it is not clear that the results reflect the immediate and automatic seizure of attention by the novel objects. Rather than having their attention instantly captured by novel objects, observers might have encountered these objects somewhat late in their initial perusal of the scenes and only then directed their full attention to them. In fact, still more recent research indicates that the detection of unexpected objects is suppressed when only a glance at a scene is allowed. Biederman, Mezzanotte, and Rabinowitz (1982, Experiment 1) examined the effect of object incongruence on the detection of targets in scenes exposed for only 150 ms. A drawing of a naturalistic scene, such as a livingroom or street scene, was preceded by the name of a target object, such as COUCH, and was followed by a location probe. The observers' task was to say whether or not the target object had appeared at the probed location. Speed and accuracy of detection decreased to the extent that the object was incongruent with the rest of the scene. For example, a couch was more detectable when it was appropriately located in a livingroom scene than when it hovered above the buildings in a street scene. Moreover, the detectability of a congruent object was not suppressed by the presence of an incongruent object elsewhere in the scene. These findings argue in favor of the sink-in, rather than popout, of novel perturbations of familiar environments. Although the Biederman et al. (1982) study differed from prior research in its use of brief exposures, it differed also in its use of a directed-attention (viz., target-detection) task rather than a nondirected task. The present research examined the way that attention is distributed spontaneously between expected, or familiar, words and unexpected, or novel, words when the arrays are exposed briefly and the observers are encouraged to apprehend as many words as they can. In contrast to Biederman et al. (1982), we observed novel popout rather than novel sink-in; attention appeared to be captured by a single unexpected word when it was presented in the company of several expected words. In the remainder of this article, we summarize some preliminary observations of novel popout, describe some new explorations of the phenomenon, and consider related phenomena and possible accounts of novel popout.
Preliminary Studies Because our preliminary findings are not reported in easily accessible sources (DeWitt & Johnston, 1989; Farah, 1987; Johnston & Farah, 1986), they are collectively summarized in some detail here. Observers were asked to monitor a
succession of four-word arrays. Each word was exposed for 500 ms in one of four array locations and then backward masked for 50 ms by a row of Xs. The stimulus onset asynchrony (SOA) between arrays was 1,000 ms. Approximately 10% of the arrays were followed by a location probe in lieu of another array. A probe appeared for 3,000 ms as a ? in one of the locations of an otherwise empty array. The observers' task was to report aloud the word that had occupied the probed location. Two array compositions, or ratios of novel to familiar words—4:0 and 1:3—were manipulated between groups of observers. Novel words appeared only once in the entire series of arrays and familiar words appeared many times. Thus, the 4:0 arrays contained all novel words, and the 1:3 arrays contained one novel word along with three familiar words. The array locations of the words were determined randomly on each trial. For some of the observers of 1:3 arrays, the three familiar words were semantically related (e.g., MORNING, NOON, NIGHT); for other observers, the familiar words were not semantically related (e.g., ISLAND, TABLE, WEEK). In all, then, there were three groups, each composed of 36 observers: One group experienced only 4:0 arrays, one experienced only 1:3 arrays containing semantically related familiar words, and one experienced only 1:3 arrays containing semantically nonrelated familiar words. It is important to note that novelty and familiarity were defined contextually. The "novel" word in a 1:3 array was a familiar English word; however, unlike the "familiar" words with which it was displayed, its appearance could not be anticipated in the context of the experiment. Percentage correct responses to the location probes (report accuracy) is summarized in Figure 1. Novel popout can show up in two ways: within an array or between arrays. Withinarray popout is manifested as higher report accuracy for the novel words in 1:3 arrays than for the familiar words in those same arrays. Between-arrays popout is manifested as higher report accuracy for novel words in 1:3 arrays than for those in 4:0 arrays. As Figure 1 reveals, both forms of novel popout were observed, but the between-arrays effect was the larger of the two. 70
cn
\/X
g BO QUJ DC
I I Familiar Words
Novel Words
50 U UJ
rr rr o
40
CJ
a? 30
1:3 R 1:3 NR 4:0 NOVEL:FAMILIAR RATIO
Figure 1. Report accuracy in the preliminary research for novel and familiar words as a function of novel:familiar ratio and whether (1:3 R) or not (1:3 NR) the familiar words were preexperimentally related.
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
NOVEL POPOUT
Although the size of the within-array popout effect was comparable for the two types of 1:3 array, overall report accuracy was much higher and the between-arrays effect was much larger when the familiar words were related than when they were nonrelated. We recruit this observation later as support for the argument that the degree of novel popout in 1:3 arrays is a function of the fluency with which the familiar words in those arrays are perceived (see General Discussion). Suffice it to note at this juncture that, judging from the reportaccuracy data, perceptual fluency of familiar words appears to be greater if they are semantically related than if they are not. In summary, attention appears to be oriented rapidly and automatically to novel words in 1:3 arrays, especially when the familiar words are semantically related. This novel-popout phenomenon is somewhat curious, if not anomalous, because it appears to be directly at odds not only with the novel sinkin effect observed by Biederman et al. (1982), but also with the theoretical implications noted previously of perceptual priming and perceptual memory. As noted earlier, when individual words are exposed briefly, accuracy of identification is higher for those that had been seen before in the experiment (e.g., familiar words) than for those that had not (e.g., novel words). Thus, when 1:3 arrays of words are exposed briefly, it is curious that accuracy of identification is not higher for the familiar words than for the novel words. The new experiments sought to establish the boundaries and generality of novel popout and to explore its processing bases.
New Experiments Four new experiments are reported in this section. Experiment 1 tested whether the novel-popout effects would survive a number of task and procedural changes, compared novel popout with some well-known phenomena of directed attention, and tested for a complementary familiar popout effect in 3:1 (one-familiar) arrays. Experiment 2 traced the temporal development of novel popout and examined whether it still occurs under conditions that preclude eye movements. Experiments 3 and 4 investigated several possible bases of novel popout including figure-ground contrast, perceptual satiation of familiar words, and withdrawal of attention from familiar words.
Experiment 1 One of the changes introduced in Experiment 1 involved the nature of the probes. Rather than probing for the word that occupied a particular location, we probed for the location that a particular word had occupied. One effect of this change was to eliminate the guessing advantage for familiar words. In the preliminary research, familiar-word intrusions accounted for approximately 78% of the overt errors to the location probes. Apparently, when observers tried to guess what word was in the probed location, they were likely to pick one of the familiar words. This strategy should have inflated report accuracy for familiar words and attenuated the withinarray popout effect. With the new probe procedure, the ac-
399
curacy of localization yielded by pure guessing would approximate 25% for familiar and novel probes alike. Other procedural changes intended to enhance the power and efficiency of the design were that array composition was manipulated within rather than between observers and that probes were presented after every array rather than after just some of them. Besides these procedural changes, some new conditions were examined in Experiment 1. In addition to nondirected attention, two forms of directed attention were examined: target localization and focused attention. The inclusion of these tasks made it possible to compare the popout of novel words in nondirected attention with the popout of target and focal words in directed attention. In addition, the nondirectedattention task included 3:1 arrays as well as 1:3 and 4:0 arrays. The 3:1 arrays made it possible to test for a within-array familiar-popout effect.
Method Subjects and design. The observers were 513 United States Air Force (USAF) recruits who participated in the study as one of their regular assignments.' They were all high school graduates, 17 to 27 years of age, and approximately 88% of them were males. All observers performed under five different attention conditions: three nondirected-attention conditions (4:0,1:3, and 3:1 arrays) and two directedattention conditions (target localization and focused attention). Word type (novel/familiar, focal/nonfocal, or target/nontarget) was a nested factor in all but the 4:0 condition. Apparatus and procedures. The experiment was performed in the USAF Human Resources Laboratory at Brooks Air Force Base, San Antonio, Texas. Up to 30 observers were run at a time in separate testing carrels, each one equipped with a TERAK 8510A microcomputer, a television monitor, a keyboard, and an instruction booklet. The observers were able to progress through the experiment in an average time of about 1.5 hr. We describe the procedures for the nondirected-attention task first and then indicate the departures from those procedures for the directed-attention tasks. Words and other visual characters appeared in a four-cell array that was centered on the television screen. From a viewing distance of 60 cm, the array subtended visual angles of about 1.90 degrees vertically and 5 degrees horizontally. Alphanumeric characters subtended angles of up to 0.38 degrees vertically and 0.24 degrees horizontally. Each stimulus presented on a trial was centered in one of the array locations. A trial comprised a succession of five arrays: warning, attention, mask, probe, and feedback. These arrays are illustrated in Figure 2. Each location in a warning array contained a string of three asterisks. A warning array was exposed for 200 ms and was followed 600 ms later by an attention array. Each location in the attention array contained a different word. The attention array was presented for 383 ms and followed, 17 ms later, by a mask array, yielding a virtual exposure duration of 400 ms. The mask array contained strings of 9 Xs in lieu of words and was exposed for 100 ms. After a blank interval of 500 ms, a probe array was presented in which one of the words from the attention array reappeared in all four locations. The observers' task was to indicate in which location of the attention array
' A large number of observers was tested to permit analyses of individual differences in attention. However, these analyses fall outside the scope of, and are not summarized in, the present article.
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
400
JOHNSTON, HAWLEY, PLEWE, ELLIOTT, DEWITT
1. warning array (200 ms)
(600 ms) CHIN
*•
TOPIC (17 ms)
xxxxxxxxx 3. mask array
| XXXXXXXXX ] XXXXXXXXX |
