Divided Attention - Joseph & Nakayama, 1997 PDF

letters to nature 14. Bongaarts, J. in The Future Population of the World. What Can We Assume Today? (ed. Lutz, W.) 170– items increases. Stimulus attributes leading to this behaviour, such 195 (Earthscan, London, 1996). 15. Garenne, M. in The Future Population of the World....

letters to nature 14. Bongaarts, J. in The Future Population of the World. What Can We Assume Today? (ed. Lutz, W.) 170– items increases. Stimulus attributes leading to this behaviour, such 195 (Earthscan, London, 1996). 15. Garenne, M. in The Future Population of the World. What Can We Assume Today? (ed. Lutz, W.) 149– as stimulus orientation, are thought to be processed in parallel 169 (Earthscan, London, 1996). across space with unlimited capacity; hence they are called ‘pre- 16. Heilig, G. K. in The Future Population of the World. What Can We Assume Today? (ed. Lutz, W.) 196– 249 (Earthscan, London, 1996). attentive’ features, and are sometimes thought to be perceived 17. Cohen, J. E. How Many People Can the Earth Support? (Norton, New York, 1995). without the use of attention. In the case of stimulus orientation, 18. Zlotnik, H. in The Future Population of the World. What Can We Assume Today? (ed. Lutz, W.) 299–335 this fits well with the orientation selectivity of V1 neurons7,8, which (Earthscan, London, 1996). 19. Keyfitz, N. Applied Mathematical Demography (John Wiley, New York, 1977). could conceivably permit the perception of orientation differences 20. Rogers, A. Introduction to Multiregional Mathematical Demography (John Wiley, New York, 1975). regardless of the attentional state. 21. Lutz, W. (ed.)The Future Population of the World. What Can We Assume Today? Revised Edition. (Earthscan, London, 1996). If the perception of primitive features enjoys a special status in the 22. Lutz, W. Scenario Analysis in Population Projection (Working Paper WP-95-57, International Institute visual processing stream, avoiding any bottleneck of limited for Applied Systems Analysis, Laxenburg, Austria, 1995). resources, then performance in detection of a feature difference 23. Lee, R. D. Probabilistic approaches to population forecasting, in Rethinking Population Projections (eds Lutz, W. & Vaupel, J. W.) (International Institute for Applied Systems Analysis, Laxenburg, should be unaffected when attention is diverted elsewhere. We Austria) (submitted 1996). investigated the role of attention in the perception of ‘preattentive’ 24. Lutz, W. & Scherbov, S. Sensitivity Analysis of Expert-Based Probabilisitc Population Projections in the Case of Asutria (Working Paper, International Institute for Applied Systems Analysis, Laxenburg, orientation features with a dual-task procedure as depicted in Fig. 1. Austria, 1997). We used a competing task, that of reporting the identity of a single 25. Brass, W. in Biological Aspects of Demography (ed. Brass, W.) 69–110 (Taylor and Francis, London, white letter appearing in a rapidly changing stream of otherwise 1971). black letters at fixation. This rapid serial visual presentation (RSVP) Acknowledgements. This work has been conducted at IIASA (International Institute for Applied Systems is very demanding when presented at 12 letters s 2 1 and has been Analysis). shown to effectively consume attentional resources for periods up to Correspondence and requests for materials should be addressed to W.L. (e-mail: lutz@iiasa. ac.at). half a second9. A search array of oriented Gabor patches was presented for 150 ms, immediately followed by 150 ms high- contrast white-noise masks covering their locations. The lag between the onset of the white target letter in the RSVP stream and the onset of the orientation array was randomly varied to Attentional requirements in a examine the temporal extent of interference, if any. In the single- task condition, subjects were instructed to ignore the letters and ‘preattentive’ feature report only whether an orientation ‘oddball’, a uniquely oriented item, had been present. In the dual-task condition, subjects were search task instructed to report both the white letter and whether an orienta- tion oddball was present. Julian S. Joseph*, Marvin M. Chun† Severe impairments in performance in detecting orientation & Ken Nakayama‡ oddballs resulted when the attentionally demanding RSVP letter * Department of Psychology, University of Nevada, Reno, Nevada 89557, USA identification was additionally imposed (Fig. 2). In the condition of † Department of Psychology, Yale University, New Haven, Connecticut 06520, performing only the single task of orientation oddball detection, USA subjects performed well, averaging 94% correct. However, when ‡ Department of Psychology, Harvard University, Cambridge, performing the dual task of letter identification and orientation Massachusetts 02138, USA oddball detection, oddball detection accuracy was only 60 6 5% for......................................................................................................................... simultaneous letter and orientation onset (lag 0). Note that the It is commonly assumed that certain features are so elementary to chance level of performance is 50% in this task. Significant degrada- the visual system that they require no attentional resources to be tion in performance persists for several hundred milliseconds after perceived. Such ‘preattentive’ features are traditionally identified the target letter’s appearance, as a result of the attentional demands by visual search performance1–3, in which the reaction time for for processing the target letter9–11. For the longest lag of 667 ms, detecting a feature difference against a set of distractor items does dual-task performance recovered to the single-task level; thus the not increase with the number of distractors. This suggests an impairment reflects the temporal dynamics of attentional load unlimited capacity for the perception of such features. We provide rather than just a generic difficulty in encoding and retaining two evidence to the contrary, demonstrating that detection of differ- independent responses. The effects of condition and lag, and the ences in a simple feature such as orientation is severely impaired interaction between these variables were all significant (P , 0:01). by additionally imposing an attentionally demanding rapid serial One might speculate that we observed attentional effects in the visual presentation task involving letter identification. The same detection performance because these stimuli are unusual in some visual stimuli exhibit non-increasing reaction time versus set-size way and do not qualify as ‘preattentive’. Performance was quite high functions. These results demonstrate that attention can be critical when only the orientation oddball task had to be performed, but do even for the detection of so-called ‘preattentive’ features. these stimuli exhibit the standard experimental signature of so- One basic tenet of modern vision research is that certain attri- called preattentive perception, specifically reaction times that do butes of visual stimuli can be processed and detected in parallel not increase with the number of items? There was no reason to across the visual field4–6. Visual attributes such as orientation1,5,6, expect otherwise, because many studies1,3 have found this for colour, or size differences3 have been put forth as ‘preattentively’ orientation differences that are easily detected. Our stimulus dis- perceived stimulus properties, a concept introduced by Neisser4. play, however, was slightly different from those used in the usual Perhaps owing to the emphasis on a dichotomy between ‘preatten- visual search task in that our search array was on for only a short tive’ and ‘attentive’ processing, it is commonly assumed that fixed duration and was masked, whereas it is more customarily attentional resources are not necessary for the perception of such presented without a mask and for a longer duration up to the time image properties. This dichotomy stems from a long history of of the observer’s response. To remove any residual doubts, we did a research with the visual search paradigm in which the time to detect visual search reaction time experiment on visual stimuli that were a target is measured as a function of the number of display items precisely the same as those used in the first experiment, the only (reviewed in ref. 3). Stimulus attributes that require focal attention difference being that the number of oriented Gabor items was varied to be perceived exhibit positive slopes: reaction times increase with from trial to trial. Subjects were instructed to ignore the letters and increasing display size. In contrast, some search tasks show reaction respond correctly on the presence of a uniquely oriented item as times that remain flat or even decrease slightly as the number of rapidly as possible. The letter stream was presented as well, although Nature © Macmillan Publishers Ltd 1997 NATURE | VOL 387 | 19 JUNE 1997 805 letters to nature Figure 1 Schematic of visual stimuli. The Gabor items were oriented at either þ458 or 2 458 with respect to vertical. Half the trials contained one uniquely oriented item (an oddball) and half contained no oddball (all items identical). Subjects responded whether an oddball was present. An RSVP stream of letters (36 arcmin tall) was concurrently presented at fixation. In some blocks of trials, subjects had to report the white letter in addition to responding whether an oddball was present in the circular array (dual-task condition). In alternating blocks, subjects ignored the letters and responded only as to the presence of an orientation oddball (single-task condition). Responses were not speeded; subjects were instructed not to make their keypress responses until after the display sequence was completed. Figure 2 Percentage correct in the orientation oddball detection for the single- Figure 3 To check that the standard experimental signature of a preattentively task (oddball detection only) and dual-task (letter identification and oddball perceived attribute was exhibited by our stimuli, we measured the detection detection) conditions as a function of the lag between the onsets of the target reaction time as a function of the number of items in the display. The same stimuli letter and the orientation array. Plots here and in subsequent figures represent the were used as in the previous experiment, including the presentation of the letters, averages across subjects. Representative error bars indicate the s.e.m. across except that the number of oriented Gabor items was varied. Eight naive subjects subjects. In the dual-task condition, the percentage correct in the oddball task each performed 10 blocks of 60 trials each, with oddball presence, distractor was tabulated out of those trials in which the letter was identified correctly9 (letter orientation, lag and set-size counterbalanced within each block. The reaction accuracy was 83%; chance was 3.8%). time for orientation oddball detection did not increase with the number of items in the search, which is characteristic of so-called ‘preattentive’ features. The error rate showed no set-size dependence and averaged 7%. it was irrelevant to this task. Hence, on trials with set-size 12, the fixation13. To explore this issue with the RSVP letter identification stimuli were physically identical in every respect to those used in the task, we examined the detection of orientation differences in two- previous experiment. dimensional arrays of 36 oriented Gabor items (Fig. 4a) essentially The results (Fig. 3) show that the reaction time for detecting identical to those used by Braun and Sagi12, but with increased orientation oddballs does not increase with the number of distrac- contrast (53% instead of 36%) and with longer Gabor duration (33 tor items. In fact, there is a slight decrease, which is to be expected rather than 20 ms). A Gabor-mask stimulus onset asynchrony from the higher density at the larger set-sizes. The perception of the (SOA) of 200 ms was used. In all other respects the procedure was same stimuli that displayed attentional requirements in the first the same as in the first experiment. Profound attentional impair- experiment also exhibited the defining characteristic of a so-called ment was again observed (Fig. 4b) with the same qualitative pattern. ‘preattentive’ feature. The greatest impairment was at the shortest lag times after the target The circular arrays of oriented items discussed so far are the letter. We repeated this experiment with six different naive subjects simplest way of studying the attentional requirements of detecting a for an SOA of 133 ms and stimulus duration 100 ms, obtaining the feature difference, with each target and distractor having the same same effects (Fig. 4b). We also replicated Braun and Sagi’s null result eccentricity and hence comparable visibility. One might imagine, with the L/T discrimination task at a variety of orientation stimulus however, that perception of an orientation difference within a two- SOAs after subjects had received extensive practice in each single dimensional array of items could conceivably be part of a qualita- task and the dual task, although there was an effect in the initial tively distinct class of visual tasks with no attentional requirements blocks that quickly decreased as the tasks were learned. These whatsoever. Such arrays were tested by Braun and Sagi12,13 who observations suggest that extensive practice can greatly reduce the found no diminution of performance in highly practised subjects observable attentional effects, although the use of different atten- using a task of discriminating between the letters L and T at tional tasks, RSVP letter identification compared with L/T13,14 Nature © Macmillan Publishers Ltd 1997 806 NATURE | VOL 387 | 19 JUNE 1997  letters to nature Figure 4 a, Two-dimensional array of oriented Gabor items used to demonstrate conditions, as in the first experiment. The same stimuli were tested on another six that similar performance impairments occur in the perception of this type of subjects with an SOA of 133 ms and a stimulus duration of 100 ms. b, In the two- orientation difference. An orientation oddball, when present, was equally likely to dimensional arrays, detection of orientation differences was impaired by the appear in any location of the second concentric ring in the hexagonal array. The attentionally demanding letter identification task (letter accuracy 76% and 75% 33 ms array was followed by a 167 ms blank interval and a 100 ms array of masking respectively for the 200 and 133 ms SOA experiments). The effects of condition elements consisting of superimposed high-contrast Gabor items of different and lag, as well as the interaction between them, were all significant in each orientations12. Six naive subjects were tested in both single-task and dual-task experiment (P , 0:001). discrimination, might also have been partly responsible for the with 50% peak contrast and a standard deviation of 22 arcmin, with a cosine different results obtained in the present work. modulation of 110 arcmin wavelength. The Gabor items were regularly spaced In sum, our results show that visual feature search tasks, which around a circle at 5.38 eccentricity, with a random overall phase in the locations have been deemed independent of attentional resources, are of the array. Stimuli were binocularly viewed from a distance of 57 cm. Each impaired by a sufficiently demanding attentional load. We demon- trial began with a small dot appearing at fixation for 500 ms, followed by a strated that both flat reaction time vs set-size functions and rapid 500 ms blank interval and the beginning of the RSVP letter stream. Subjects discrimination performance are dissociated from unlimited- were first shown frozen displays, and then given 20 trials of practice in each the capacity detection of such ‘preattentive’ features. These findings single-task and dual-task conditions with feedback in the first 10 before the seem to rule out a conceivable architecture for the visual experiment began. system4–6,12,13 in which all feature differences are processed along a Received 25 November 1996; accepted 14 April 1997. pathway that has a direct route to awareness, without having to pass 1. Treisman, A. M. Preattentive processing in vision. Comput. Vis. Graph. Image Proc. 31, 156–177 through an attentional bottleneck. The results are consistent, how- (1985). ever, with the theoretical notion15 that all tasks are contingent upon 2. Treisman, A. M. & Sato, S. Conjunction search revisited. J. Exp. Psychol.: Hum. Percept. Perf. 16, 459– the availability of limited resources, while differing in their sensi- 478 (1990). 3. Wolfe, J. M. Guided search 2.0: A revised model of visual search. Psychonom. Bull. Rev. 1, 202–238 tivity to reductions in these resources. This view is supported by the (1994). results of Mack et al.16,17, in which the presence of an otherwise 4. Neisser, U. Cognitive Psychology (Appleton-Century-Crofts, New York, 1967). 5. Julesz, B. Textons, The elements of texture perception, and their interactions. Nature 290, 91–97 salient object in the visual field frequently goes unnoticed when it is (1981). unexpected and irrelevant to the task at hand. In the present study, 6. Julesz, B.. & Bergen, J. R. Textons, the fundamental elements in preattentive vision and perception of however, subjects are inherently constrained by limited attentional textures. Bell Sys. Tech. J. 62, 1619–1645 (1983). 7. Hubel, D. H. & Wiesel, T. N. Receptive fields and functional architecture of monkey striate cortex. J. resources, even when actively interrogating the visual stimulus for Physiol. (Lond.) 195, 215–243 (1968). the presence of a feature difference. These experiments argue against 8. Knierim, J. J. & Van Essen, D. C. Neuronal responses to static texture patterns in area V1 of the alert macaque monkey. J. Neurophys. 67, 961–980 (1992). a direct route from preattentive processing to perceptual report. By 9. Raymond, J., Shapiro, K. & Arnell, K. M. Temporary suppression of visual processing in an RSVP task: providing a demonstration of ‘preattentive’ information that an additional blink? J. Exp. Psychol.: Hum. Percept. Perf. 18, 849–860 (1992). cannot be overtly perceived without attention, we challenge the 10. Duncan, J., Ward, R. & Shapiro, K. Direct measurement of attentional dwell time in human vision. Nature 369, 313–315 (1994). current dichotomous view that assumes the existence of a separate 11. Chun, M. M. & Potter, M. C. A two-stage model for multiple target detection in rapid series visual privileged category of preattentive perception. presentation. J. Exp. Psychol.: Hum. Percept. Perf. 21, 109–127 (1995).......................................................................................................................... 12. Braun, J. & Sagi, D. Vision outside the focus of attention. Percept. Psychophys. 48, 45–58 (1990). 13. Braun, J. & Sagi, D. Texture-based tasks are little affected by second tasks requiring peripheral or Methods central attentive fixation. Perception 20, 483–500 (1991). In the first experiment, six naive subjects, recruited for pay from the university 14. Braun, J. Visual search among items of different salience: Removal of visual attention mimics a lesion in extrastriate area V4. J. Neurosci. 14, 554–567 (1994). community, each performed a 1-h session consisting of six blocks alternating 15. Norman, D. A. & Bobrow, D. G. On data-limited and resource-limited processes. Cogn. Psychol. 7, 44– between the single-task and dual-task conditions, with the order counter- 64 (1975). 16. Mack, A., Tang, B., Tuma, R., Kahn, S. & Rock, I. Perceptual organization and attention. Cogn. Psychol. balanced across subjects. Each block contained 80 trials, counterbalanced for 24, 475–501 (1992). the presence of an orientation oddball, distractor orientation and lag. Fourteen 17. Rock, I., Linnett, C. M., Grant, P. & Mack, A. Perception without attention: results of a new method. letters were presented after the target letter; between five and ten letters Cogn. Psychol. 24, 502–534 (1992). preceded it. No letter was repeated in a single trial. Luminances were 50 cd m 2 2 Acknowledgements. This work was supported by the University of Nevada, Reno College of Arts and for the white letter, 7 cd m 2 2 for the black letters, and 25 cd m 2 2 for the Science (J.S.J.), NIH (J.S.J., M.M.C.), and AFOSR (K.N.) and the McKnight Foundation (K.N.). We thank P. Cavanagh and J. M. Wolfe for helpful comments on the manuscript. background. Each letter in the stream was presented for 33 ms followed by a 50 ms blank gap. The Gabor functions were composed of a gaussian envelope Correspondence and requests for materials should be addressed to J.S.J. (e-mail: [email protected]). Nature © Macmillan Publishers Ltd 1997 NATURE | VOL 387 | 19 JUNE 1997 807