How does target duration affect object substitution masking?

Journal of Experimental Psychology: Human Perception and Performance 2010, Vol. 36, No. 5, 1267–1279 © 2010 American Psychological Association 0096-1523/10/$12.00 DOI: 10.1037/a0018733 How Does Target Duration Affect Object Substitution Masking? Angus Gellatly, Michael Pilling, Wakefield Carter, and Duncan Guest 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. Oxford Brookes University Object substitution masking (OSM) is typically studied using a brief search display. The target item may be indicated by a cue/mask surrounding but not overlapping it. Report of the target is reduced when mask offset trails target offset rather than being simultaneous with it. We report 5 experiments investigating whether OSM can be obtained if the search display is on view for a period of up to 830 ms but cueing of the target location is delayed. The question of interest is whether OSM must reflect the initial response of the visual system to target onset or whether it can arise in other ways, possibly during the transition from a pre-attentive representation of the target item to an attentional representation of it. Our results show that OSM decreases in strength as target duration increases. An explanation is suggested in terms of the object individuation hypothesis (Lleras & Moore, 2003). Keywords: object substitution masking, target duration, attention, object individuation colleagues (2000) suggested that demonstrations of (low level) OFM masking have frequently contained a substantial but unrecognized component of (higher level) OSM. Di Lollo et al (2000) and Enns (2004) proposed a theory of OSM that assumes that perception arises from recurrent communication between neurons at different levels within the visual system. Similarly, Lamme and colleagues (Landman, Spekreijse, & Lamme, 2003; Sligte, Scholte, & Lamme, 2008) also proposed that two-way communication between cells at different levels is necessary for conscious experience to arise. (Though see Macknik & Martinez-Conde, 2007, for an alternative view on the function of re-entrant fibers.) According to the theory of OSM, a newly appearing object stimulates lower level cells with spatially local receptive fields and geometrically simple stimulus requirements. In a feed-forward sweep, output from these cells activates higher level neurons that have larger receptive fields and are tuned to more complex stimulus properties. Competing pattern hypotheses are generated at the higher level. Resolution of competition between these hypotheses, and also binding of patterns to precise spatial locations, is thought to require feedback sweeps. Activations at higher and lower levels are compared for consistency, and after some number of cycles of forward and backward sweeps competitive interactions yield a stable percept. If the visual scene remains constant over the iterations required to achieve dynamic stability, the new object will be consciously perceived. However, if—as in the above example of OSM—a mismatch is detected between activation at the different levels, the iterative process will begin again based only on current sensory input. OSM is said to occur as a result of such a mismatch. Onset of target and mask sets up lower level activation leading to the hypothesis of target plus mask. If both offset simultaneously before the arrival of the feedback sweep, this hypothesis can still be matched to persisting but fading activity at the lower level. However, if the mask display continues after offset of the target, the hypothesis will mismatch strong sensory evidence that there is now only a mask present. Further iterations result in only the mask being consciously perceived. Perception of the target plus mask will have been substituted by perception of the mask alone. Focused attention is thought Object substitution masking (OSM) is a recently discovered form of masking that is conceptualized within a neurophysiologically inspired framework emphasizing competitive interactions between loops of neural activation (Bischof & Di Lollo, 1995; Di Lollo, Enns & Rensink, 2000; Enns, 2004; Enns & Di Lollo, 1997, 2000; Reiss & Hoffman, 2006, 2007). OSM is frequently demonstrated using common-onset four dot or outline square masking. A search array of items, letters for example, is briefly presented. One of the items, the target, is surrounded by an outline square (or four dot) mask, which also serves as a cue indicating the item to be reported. If the mask and target offset simultaneously, target identity can be reported fairly accurately. However, if the mask remains present after target offset then the ability to report target identity is greatly reduced. The performance reduction is an example of OSM. OSM supposedly involves substitution of one perceptual object (target plus mask) by another (mask alone). It contrasts with what Enns (2004) called object formation masking (OFM), which results from interference with the perceptual formation process involved in segmenting a target from the “camouflage” of background and other nearby objects. OFM is sensitive to factors such as contour abutment and overlap; it depends critically on the exact timing of target and mask onsets, and is little affected by manipulations of spatial attention. OSM, by contrast, is highly sensitive to attentional manipulations but not to the local spatiotemporal contour interactions thought to give rise to OFM. Di Lollo and Angus Gellatly, Michael Pilling, Wakefield Carter, and Duncan Guest, Department of Psychology, Oxford Brookes University. This research was supported by ESRC Grant RES– 000 –22–3087 to Angus Gellatly. We are grateful to Ioannis Argyropoulos for help with data collection for Experiments 1 through 3. Correspondence concerning this article should be addressed to Angus Gellatly, Department of Psychology, School of Social Science and Law, Oxford Brookes University, Clerici Building, Headington Campus, Gipsy Lane, Headington, Oxford OX3 0BP, England. E-mail: agellatly@ brookes.ac.uk 1267 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. 1268 GELLATLY, PILLING, CARTER, AND GUEST to modulate OSM by reducing the number of iterations required to achieve dynamic stability (cf. Neill, Hutchinson & Graves, 2002; Tata & Giaschi, 2004). Studies of visual masking have traditionally employed presentations of very brief target stimuli to obtain their effects. All studies of OSM of which we are aware have followed this tradition and, indeed, the theory of OSM postulates that masking results from interference with early processes involved in perception of a newly onset target object. However, a key condition for obtaining OSM is that attention should not be prefocused on the target (Di Lollo et al, 2000; Tata & Giaschi, 2004). This raises the question of whether it would be possible to obtain OSM with prolonged duration search displays if the indicative cue is delayed to prevent attention prefocusing on the target. A positive response to this question would show that OSM need not be a function of the initial response of the visual system to target onset but can arise in other ways, possibly during the transition from a pre-attentive representation of the target item to a fully attentive representation of it. A wealth of experimental evidence indicates how little observers can represent and report about the properties of objects that are not the focus of attention. In change blindness studies, for example, even the alternating onset and offset of an object in a “busy” array may be detected only after many cycles of the display if the associated luminance signals are masked by screen-wide luminance changes or by competing highly salient onsets and offsets of “mud-splashes” (O’Regan, Rensink & Clark, 1999; Turatto, Angrilli, Mazza, Umilta, & Driver, 2002). Studies of visual shortterm memory (VSTM) revealed that the properties of no more than about four items can be represented in VSTM at one time (Alvarez & Cavanagh, 2004; Luck & Vogel, 1997; Vogel, Woodman, & Luck, 2001) suggesting that change detection will occur only if the changing item happens to be one of the four selected items. And in a recent paper, Wolfe, Reinecke, and Brawn (2006) found that not even this limited amount of information was available from a visual array that had been presented for a longer duration than is typical in masking experiments. In their Experiment 2, Wolfe et al. presented an array of 32 left- and right-tilted bars for between 500 ms and 1,000 ms, after which one bar was occluded by a grey square. Despite the relatively lengthy presentation of the stimulus array prior to onset of the cue/mask, observers were unable to report the orientation of the occluded bar at better than chance. Studies such as those just cited suggest that, at best, only a very limited subset of items in an array can be represented in VSTM at any time, even when displays are presented for lengthy or repeated viewing. Given that this is the case, one may predict that if a search display is presented for several hundred milliseconds, rather than for the 10 ms to 50 ms duration usual in studies of OSM, it should still be possible to demonstrate a robust OSM effect associated with the items not represented in VSTM at the critical time. To summarize: If OSM is associated only with the early neural processes that follow the onset of a stimulus object, then OSM should not be obtained when display items have had a prolonged duration. Conversely, if OSM can arise in other ways, possibly associated with the changed representation of a previously unattended object when it comes to be attended, then prolonged duration display items should still exhibit susceptibility to OSM. Here, we report five experiments designed to test which of these outcomes obtains in fact. General Method Participants The various means by which participants were recruited are described below for individual experiments. All participants had normal or corrected-to-normal vision. The research project had approval from the Oxford Brookes University Research Ethics Committee, and participants were informed of their right to withdraw from the experiment at any point. Procedure All the experiments reported here involved presentations of search displays consisting of 12 letter stimuli presented at clock positions around a central fixation point on a virtual circle 3.2° in diameter when viewed from 140 cm (see Figure 1). Randomly on any trial, half the letters were “U”s and half were “H”s. Letters were 0.3° constructed from vertical and horizontal lines 0.01° in thickness. Search displays were presented for between 17 and 830 ms (see below for details). Trials began with a fixation point for 1 s after which the search display was added. In Experiments 1 to 3, the target item was indicated by a surrounding mask that was a 0.5° square. The mask was either present from the onset of the search display or subsequently added to the display. The search display then terminated, and the mask either terminated at the same time (control condition) or after a further period of 500 ms (trailing mask condition). Participants pressed the left slash key (“⶿”) if they thought the target was an “H” and the right slash key (“/”) if they thought it was a “U.” In Experiments 1 to 3, letter items and the mask square were always white (RGB values of 255, 255, 255). In Experiments 4 and 5, the target item was a different color to the items in the display or became a different color some time after display onset. The masking conditions in these two experiments will be described in due course. Instructions for all experiments emphasized that accuracy, not speed, of response was of importance. Participants in each experiment were initially shown slowed-down trial sequences to acquaint them with the task. They then undertook practice trials followed by blocks of experimental trials. The program allowed participants a break every block of 60 trials. For all the experiments, there were 40 trials per participant per data point. All the experiments used fully within-participant designs with order of conditions randomized. Experiments 1 to 3 were controlled by customized software written in Borland Turbo C⫹⫹. Experiments 4 and 5 were written in Matlab using the Psychophysics toolbox extensions (Brainard, 1997; Pelli, 1997). Experiments were presented on a monitor screen running at 60 Hz for Experiments 1 to 3 and 100 Hz for Experiments 4 and 5 and viewed from 140 cm in a dimly illuminated room. The common onset technique for demonstrating OSM (Di Lollo et al., 2000) is schematically represented in the top panel of Figure 2. In the control condition, target and mask have both common onset and common offset. In the trailing mask condition they have a common onset but the mask has a delayed offset. Because target and mask have common onsets in both conditions, any OFM between them should be equated for the two conditions. Subtracting performance in the masking condition from performance in the control condition is said to give the extent of OSM alone. In an alternative method, onset of the mask is delayed relative to target 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. TARGET DURATION AND SUBSTITUTION MASKING Figure 1. 1269 Stimulus sequence for Experiments 1 and 2. onset (see middle and lower panels of Figure 2). Provided the onset asynchrony is 100 ms or more, no spatiotemporal interactions (and so no OFM) are expected between target and mask elements. Once again, therefore, subtracting performance with a trailing mask from performance with a simultaneous offset mask gives the measure of OSM. Both methods were employed in the present experiments. Experiment 1 Participants There were 10 female and 4 male participants in Experiment 1, aged from 18 to 35 years. They were recruited by advertisement from among Oxford Brookes students and the general public, and received a small financial recompense. All were naı̈ve as to the purpose of the Experiment 1. Method The eight conditions of Experiment 1 are represented schematically on the left side of Figure 2. Note that stimulus durations, represented horizontally, are not drawn to scale. Target duration was 17 ms, 200 ms, or 500 ms. For the 17-ms targets, the common onset mask either offset simultaneously with the target (control condition) or remained present (trailed) for a further 500 ms (masking condition). These are the standard control and mask Figure 2. The stimulus conditions employed in Experiment 1 (left side) and the associated percentages of correct responses (right side) together. Error bars indicate ⫾ 1 standard error. T ⫽ target; M ⫽ mask; Sim ⫽ simultaneous. GELLATLY, PILLING, CARTER, AND GUEST 1270 conditions for the common onset method. For the 200-ms and 500-ms targets, there were similar standard conditions in which the mask was present for 200 or 700 ms, and for 500 or 1,000 ms. In addition, for each of these target durations there was a delayed onset mask condition in which the mask onset 17 ms before target offset and then trailed for a further 500 ms. In these conditions, “cue time” (when target and mask were simultaneously present) was the same as in the masking condition with the 17-ms target. 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. Results and Discussion Percentage correct responses and standard errors for each condition are shown on the right side of Figure 2. For the 17-ms targets, there is a strong OSM effect—a 20% difference in accuracy between the control and mask conditions, t(13) ⫽ 5.89, p ⬍ .001, but the analogous comparisons for 200-ms and 500-ms targets yields nonsignificant differences of 4% and 1%, maximum t(13) ⫽ 1.32, p ⬎ .05. However, if the comparison is made between the simultaneous onset/offset control condition and the delayed onset/offset masking condition for 200-ms and 500-ms targets, the OSM effect is 31% and 35%, respectively, minimum t(13) ⫽ 7.5, p ⬍ .001. At first blush, therefore, whether OSM decreases or increases with target duration depends on which method is used to calculate the size of the effect. However, neither comparison is appropriate for long duration targets because, unsurprisingly, cue time is a critical variable: The longer the cue time when both the target and the cue/mask are present simultaneously, the higher the level of performance. This consideration suggests that the size of the OSM effect should be based on comparisons in which cue time is equated. That is, the standard 17-ms target and mask control condition should be compared with each of the three masking conditions in which the cue time was also 17 ms. These comparisons give OSM effects of 20% as before, and 16% and 11% for target durations of 17 ms, 200 ms, and 500 ms, respectively. A one-way repeated-measures analysis of variance (ANOVA) on the individual percentage OSM figures indicated that a modest decline in OSM occurred with increasing target duration, F(2, 26) ⫽ 4.45, MSE ⫽ 32.80, ␩2p ⫽ .26, p ⬍ .05. However, not much can be concluded from these figures because comparing a 17-ms target control condition against a 200-ms or 500-ms target masking condition is less than satisfactory. Moreover, at 52% performance in the masked condition with 17-ms targets was barely above chance, which may have resulted in under estimation of the true masking effect in that condition. Building on these results from Experiment 1, therefore, Experiment 2 was designed to provide more equitable comparisons for calculating the size of the OSM effect for targets of different duration and also to avoid possible floor effects in the data. Experiment 2 Participants There were11 female and 7 male participants, as described for Experiment 1 and aged between 20 and 42 years. Method The six conditions of Experiment 2 are schematically represented on the left side of Figure 3. Note again that stimulus durations, represented horizontally, are not drawn to scale. The shortest target duration was increased from 17 ms to 50 ms in an effort to improve on the chance level performance obtained with 17-ms targets in the masked condition of Experiment 1. Target durations in Experiment 2 were 50 ms, 200 ms, and 500 ms. There was a control condition (simultaneous mask offset) and a masking Figure 3. The stimulus conditions employed in Experiment 2 (left side) and the associated percentages of correct responses (right side). Error bars indicate ⫾ 1 standard error. T ⫽ target; M ⫽ mask; Sim ⫽ simultaneous. TARGET DURATION AND SUBSTITUTION MASKING condition (delayed mask offset for each target duration. The mask was always present for the last 50 ms of the target presentation and offset either simultaneously with it or after a further 500 ms. 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. Results and Discussion Percentage correct responses and standard errors for each condition are shown on the right side of Figure 3. The amount of OSM (as indexed by the difference in percentages between the control and mask conditions i.e., the difference between grey and spotted bars for each target duration) for the three target durations is shown in Figure 4. There are several things to note about these results. First, trebling the duration of the shortest targets—and the duration of the cue/target overlap— had no effect on performance in either the control or masking conditions as compared with Experiment 1, nor, therefore on the size of the OSM effect. However, trebling the cue/target overlap time did improve performance in the 200-ms and 500-ms masking conditions as compared with the analogous conditions in Experiment 1. Second, increasing target duration to 200 ms and 500 ms resulted in the OSM effect being reduced and then effectively eliminated. Third, target duration did not affect control and masking conditions in exactly the same way. When the data were subjected to a 2 ⫻ 3 repeatedmeasures ANOVA, there were main effects of masking condition, F(1, 17) ⫽ 53.13, MSE ⫽ 64.49, ␩2p ⫽ .76, p ⬍ .001, and target duration, F(2, 34) ⫽ 33.79, MSE ⫽ 46.36, ␩2p ⫽ .67, p ⬍ .001, and also a significant interaction between the two factors, F(2, 34) ⫽ 12.76, MSE ⫽ 43.86, ␩2p ⫽ .43, p ⬍ .001. There was a significant, t(17) ⫽ 3.69, p ⬍ .01, increase in performance in the control condition with the 200-ms targets compared to with the 50-ms targets, but then with 500-ms targets control performance dropped back to the same level as for the 50-ms targets, maximum t(17) ⫽ 0.17, p ⬎ .05. Contrastingly, for the masking condition there was a significant increase in performance from the 50-ms to the 200-ms condition, t(17) ⫽ 7.43, p ⬍ .001, but no further significant change was found between the 200-ms and 500-ms conditions, t(17) ⫽ 0.8, p ⬎ .70. Figure 4. The size of the masking effect (the difference in the percentages of correct responses in mask and control conditions) in Experiment 2. Error bars indicate ⫾ 1 standard error. 1271 Absolute levels of performance on the task are difficult to interpret given the influence of contingent effects such as attentional capture and readiness effects that may occur with rapidly and abruptly changing visual displays. For example, enhanced performance in the 200-ms target conditions may have been due to a near-optimal readiness interval. That is, a delay of 150 ms between search display onset and cue/mask onset may be associated with a stronger perceptual or attentional alerting effect than obtains with either a zero or a 450 ms delay. Alternatively, it may have been due to a brief splitting of attention between several newly onset visual objects (Dubois, Hamker, & VanRullen, 2009; Gellatly & Cole, 2000), with enhanced processing peaking at around 200 ms after onset. The existence of such complicating factors underscores the importance of having matched masking and control conditions for each target duration. And given such potential complications, it is necessary to consider the broad pattern of data in terms of the central question: Is OSM found with targets that have had a prolonged presentation? Comparison of Experiments 1 and 2 indicates that a small increase in target duration, up to 50 ms and possibly more, has no effect on performance in either the control or masking conditions. However, they also show that, contrary to our original expectation, larger increases in target duration result in a reduction, and even elimination of, the OSM effect. Experiment 2 presents us, in particular, with a paradoxical finding when comparing the 50- and 500-ms target conditions: Target duration does not affect control condition performance but does affect performance in the masking condition, resulting in the elimination of the masking effect with the longer targets. This seems to imply that in the control conditions participants know no more about the items in a 500-ms display than they do about those in a 50-ms display— but what they know is no longer susceptible to OSM. Of the five experiments reported in this paper, Experiment 2 provides the simplest and most direct test of whether OSM occurs for prolonged duration targets. The apparent answer is that OSM is absent at longer target durations. It is important, however, to consider if any factor other than target duration itself could have caused this result. One possibility is that it is the delayed onset of the cue rather than target duration itself which leads to the observed reduction in OSM. Consider that in the standard common onset conditions—the short target duration conditions in Experiment 2—the cue (square) competes for visual attention with all the other new onset items in the display. By contrast, in the delayed onset conditions—the long target duration conditions in Experiment 2—the cue is a singleton onset that should very effectively attract attention to itself and the letter it surrounds. It is known that precuing attention to the target location strongly reduces OSM (Di Lollo et al., 2000). Is it possible, then, that the absence of OSM with the delayed onset cue could be due simply to attention being summoned so rapidly and effectively to the target location that OSM is prevented? If this were the case, it would be expected that performance in both control and masking conditions should be elevated relative to the standard common onset conditions, and this was not the case. However, we have already argued that absolute performance levels may be influenced by other—strictly irrelevant to our purpose—features of the different trial types and it would, therefore, be desirable to eliminate this possible explanation of our finding. GELLATLY, PILLING, CARTER, AND GUEST 1272 Experiment 3 was conducted to see if a similar effect of target duration can be found when the delayed onset cue is not an onset singleton. This was done by partially varying target duration within-trials rather than between-trials, so that every trial contained both short and longer duration display items. The short duration items always onset simultaneously with the cue, so the latter was never an onset singleton. Experiment 3 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. Participants There were 13 female and 5 male participants, as described above, aged 18 to 37 years. Method Whereas target duration in Experiments 1 and 2 had varied between trials, in Experiment 3 the duration of different items in the display was partially varied within trials. On each trial, following the fixation period, three “H”s and three “U”s appeared randomly positioned in the odd or even numbered clock positions in the display. After 17 ms, 200 ms, or 500 ms these “old” items were joined by six “new” items appearing in the even or odd positions. Simultaneously with onset of the new items, also three “H”s and three “U”s, a cue/mask square appeared round one of the old or one of the new items. After a further 50 ms, all the letters offset, and the square either offset simultaneously or remained for another 500 ms. New targets were therefore always displayed for 50 ms whereas old targets had durations of 67 ms, 250 ms, or 550 ms, and cue/target overlap was always 50 ms. In all other respects Experiment 3 was identical to Experiments 1 and 2. Results and Discussion Percentage correct performance and standard errors for each condition are shown in Figure 5 as a function of the stimulus onset asynchrony (SOA) between old and new items. For both old and new targets a strong OSM effect is evident. Also, as in Experiment Figure 5. Percentage correct performance in Experiment 3 for old and new targets in control and mask trials as a function of stimulus onset asynchrony (SOA) between old and new items. Error bars indicate ⫾ 1 standard error. 2, discrimination performance for old items in the unmasked, or control, condition shows an increase from the 50 ms (SOA ⫽ 17 ms) to the 200 ms (SOA ⫽ 200 ms) condition, followed by a decrease to the 500 ms (SOA ⫽ 500 ms) condition. This is the effect we suggested could be due either to optimal perceptual readiness occurring approximately 200 ms after search display onset or because of a brief enhancement of processing split across several newly onset objects and peaking at around 200 ms. That the enhancement was obtained for old objects only, and did not apply to new objects, supports the latter rather than the former account. An omnibus 2 ⫻ 2 ⫻ 3 repeated-measures ANOVA with factors of target type (new/old), masking condition (control/masked) and SOA (17 ms, 200 ms, 500 ms) was conducted on the percentage correct data. It yielded a significant main effect of mask condition, F(1, 17) ⫽ 40.13, MSE ⫽ 242.55, ␩2p ⫽ .70, p ⬍ .001, but nonsignificant effects of target type, F(1, 17) ⫽ 3.52, MSE ⫽ 87.86, ␩2p ⫽ .171, p ⬍ .1, and SOA, F(2, 34) ⫽ 2.91, MSE ⫽ 82.12, ␩2p ⫽ .15, p ⬍ .1. There were significant interactions between target type and SOA, F(2, 34) ⫽ 3.68, MSE ⫽ 110.49, ␩2p ⫽ .18, p ⫽ .035, and between masking condition and SOA, F(2, 34) ⫽ 6.96, MSE ⫽ 57.18, ␩2p ⫽ .29, p ⫽ .003, and a significant three way interaction, F(2, 34) ⫽ 4.84, MSE ⫽ 94.54, ␩2p ⫽ .22, p ⫽ .014. There was absolutely no interaction between mask condition and target type, F(1, 17) ⬍ 1, MSE ⫽ 61.45, ␩2p ⫽ .00. For both control and masking conditions, new target performance shows only a small influence of SOA, whereas SOA strongly modulates the reporting of old targets. Figure 6 shows how the size of the OSM effect (control condition minus masking condition for each target duration) varies with SOA for old and new targets separately. It can be seen from this figure that there is an interaction between old and new items in the amount of OSM they produced over changes in SOA. A 2 (target type) ⫻ 3 (SOA) repeated-measures ANOVA on the OSM effect data illustrated in Figure 6 showed this interaction to be significant, F(2, 34) ⫽ 4.84, MSE ⫽ 189.08, ␩2p ⫽ .22, p ⬍ .05. There was also a significant main effect of SOA, F(2, 34) ⫽ 6.96, MSE ⫽ 114.35, ␩2p ⫽ .29, p ⬍ .005, but not of target type, F(1, 17) ⬍ 1, MSE ⫽ 122.89, ␩2p ⫽ .00. Repeated-measures one-way ANOVAs on these data for new and old targets separately showed that for new targets there was no significant change in the amount of OSM across SOA, F(2, 34) ⫽ 1.09, MSE ⫽ 162.55, ␩2p ⫽ .06, p ⬎ .1. For old targets, however, there was a significant decrease in masking with increasing duration, F(2, 34) ⫽ 10.89, MSE ⫽ 140.88, ␩2p ⫽ .39, p ⬍ .001. It can be seen in Figure 6 that for old targets OSM reduces only a little across the two shorter SOAs but decreases significantly from the 200 ms to the 500 ms SOA, t(17) ⫽ 4.46, p ⬍ .001. Although a substantial OSM effect is found for target durations of 67 ms and 217 ms, it is absent for 517-ms duration targets. Just as for Experiment 2, the results of Experiment 3 demonstrate a large reduction in, and even elimination of, OSM for targets that have been present for about 500 ms. This was so even though the cue in Experiment 3 was never an onset singleton but always had to compete for attention with the simultaneous onsets of six new objects. We can, therefore, be assured that the absence of OSM for long duration targets, which was first clearly observed in Experiment 2, was not the result of the cue drawing attention more effectively to long duration than to short duration targets. TARGET DURATION AND SUBSTITUTION MASKING 1273 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. Experiment 4 Figure 6. The size of the masking effect for old and new targets as a function of stimulus onset asynchrony (SOA) between old and new items in Experiment 3. The masking effect equals the difference in the percentages of correct responses in mask and control conditions. Error bars indicate ⫾ 1 standard error. In the three previous experiments, the location of the target was cued by the same outline square that also functioned as a mask. Such a procedure confounds cuing and masking effects (Neill et al, 2000). It also confounds target duration with SOA between the target and the mask. The design of Experiment 4 allowed us to separate these different factors and further to test the generality of the findings from Experiments 1, 2, and 3. Specifically, we investigated whether OSM would still be eliminated for prolonged duration targets when, first, the target is cued by being or becoming a different color to distractor items; second, every item in the search display initially appears inside an outline square so that the presence of a square does not indicate the location of the target, and, third, SOA between target onset and mask onset is zero for all target durations. Pilot work indicated that cuing with the chosen color cue was not as effective as cuing with the square mask, and to compensate for this a longer cuing period of 70 ms was employed in Experiment 4. As a consequence, the target durations were slightly different from those employed in the previous studies and, in addition, an even longer target duration condition was introduced. Participants The results of Experiment 2 and for the old items of Experiment 3 are strikingly similar in general outline. In both data sets, accuracy in the control condition is an inverted-U function of target display duration whereas accuracy in the masked condition increases with display duration, particularly between the first and second durations. The similarity of the two data sets is impressive given that the stimulus sequences in Experiment 3 were more complicated than those in Experiment 2. On the one hand, the onset of new objects in the mixed displays of Experiment 3 might be expected to have captured attention away from old items (Yantis & Johnson, 1990); whereas, on the other hand, old items might have been expected to attract inhibition of the kind reported by Watson and Humphreys in visual marking studies (Watson & Humphreys, 1997). Either effect might be expected to have reduced performance at longer as compared to shorter durations because, in the first case, the relative “newness” of the new objects, and hence their tendency to capture attention would have increased, while, in the second case, inhibition, which is applied over the course of several hundred milliseconds (Watson & Humphreys, 1997) would increasingly have come into play. However, the data of Experiment 3 show little sign of either of these effects. Indeed, given that inhibition of old items is a supposedly voluntary strategy (Watson & Humphreys, 1997), which would be harmful to performance in the present task, as opposed to benefiting performance on the task employed in the visual marking studies, its apparent absence is perhaps not surprising. The question raised in the introduction to this paper was: Is OSM found with targets that have had a prolonged presentation? Despite the very different ways in which target duration was manipulated in Experiments 2 and 3, in both studies an initially strong OSM effect was found to decline markedly with increasing target duration. Experiment 4 was designed to test whether a similar result will also be found when different techniques are employed for cuing and masking the target item. There were 11 female and 3 male participants recruited from Oxford Brookes psychology students and who participated in partial fulfillment of course credit requirements. They were aged between 18 and 25 years. All were naı̈ve as to the purpose of the experiment. Method In Experiment 4, as in Experiments 1 and 2, target duration was varied across trials. The sequence of trial events was similar to, but different in certain ways from, that shown in Figure 1. Following the fixation frame, a display of six “H”s and six “U”s appeared exactly as described for Experiments 1 and 2, except that every item was surrounded by an outline square of the same white color. This display remained present for 70 ms, 210 ms, 520 ms, or 830 ms. In the first of these conditions, the target item was colored blue throughout the 70-ms presentation duration. In the other conditions, one of the items—the target— changed from white to blue after 140 ms, 450 ms, or 760 ms. Thereafter, either all the letters and squares disappeared (control condition) or all the letters and all the squares excepting the square around the target disappeared (masking condition). The single-masking square remained present for a further 500 ms before terminating. The blue target had a lower luminance (RGB of 130, 130, 255) than the white items. Results and Discussion Percentage correct responses and standard errors for each condition are shown in Figure 7, the size of the masking effect as a function of target duration is shown in Figure 8. Once again, control condition performance improves between the shortest (70 ms) and next shortest (210 ms) conditions. However, unlike in Experiments 2 and 3, control performance does not drop back to the lower level for the 520 ms target duration, though showing some sign of doing so in the 830-ms condition. However, the 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. 1274 GELLATLY, PILLING, CARTER, AND GUEST Figure 7. Percentage correct performance in Experiment 4 as a function of target duration for control and mask trials. Error bars indicate ⫾ 1 standard error. display change cuing attention to the target was different in Experiment 4 as compared with the previous experiments—a color change in an existing object as opposed to onset of a new square— which could account for the difference in control condition results. Moreover, cue duration was 70 ms rather than 50 ms, which may also have had an effect. The percentage correct data were entered into a 2 ⫻ 4 repeatedmeasures ANOVA with factors of masking (control/masking) and target duration (70 ms, 210 ms, 520 ms, and 820 ms). There were main effects of masking condition, F(1, 13) ⫽ 24.63, MSE ⫽ 87.92, ␩2p ⫽ .66, p ⬍ .001, and of target duration, F(3, 39) ⫽ 39.19, MSE ⫽ 47.91, ␩2p ⫽ .75, p ⬍ .001, and a significant interaction between them, F(3, 39)⫽ 3.0, MSE ⫽ 41.31, ␩2p ⫽ .19, p ⫽ .042. Performance was better in the control than in the masking condition for 70-ms targets, t(13) ⫽ 4.03, p ⬍ .001, 210-ms targets, t(13) ⫽ 3.67, p ⫽ .003, and the 520-ms targets, t(13) ⫽ 4.16, p ⬍ .001, but not for the 830-ms targets, t(13) ⫽ 1.35, p ⫽ .199. As in the previous three studies, in Experiment 4 OSM decreased with target duration. Whereas the previous studies found masking was absent for the 500-ms duration targets, this was not the case in Experiment 4, which produced reduced but still significant masking for a target duration of 520 ms. Fortuitously, an even longer target duration of 830 ms had been included in Experiment 4, and at this duration there was no longer a significant OSM effect. Why the absence of OSM should be delayed in Experiment 4 relative to Experiments 1 to 3 is not obvious but, as just noted, Experiment 4 used both a different target cue and a different cue duration, which may have affected the target duration at which OSM becomes absent. In addition, the presence of outline squares round every item in the search display may also have affected the characteristics of OSM. The results of Experiment 4 demonstrate that the elimination of OSM at longer target durations can be replicated when all items are surrounded by squares and the target is cued by color. Because this meant that target-mask SOA was zero in all conditions, the result also shows that the elimination of OSM can be a function of target duration alone and not of the SOA between target onset and mask onset. The topic of how and under what circumstances temporal separation of target and mask onsets may influence the occurrence of OSM is one to which we return later in the paper. For the moment, however, we turn attention to a possible criticism of Experiment 4, which is that because the target and (what would turn out to be the) mask onset simultaneously, its design confounds target duration with mask duration. Previous studies (Lim & Chua, 2008; Neill et al, 2002; Tata & Giaschi, 2004) found that if, prior to onset of the search display, masks surrounding the search item locations were briefly previewed, OSM was considerably reduced. In Experiment 4 the search items and their surrounding squares came on together, so the experiment did not involve mask preview in exactly the same sense as in the previous studies. However in the 210-ms, 520-ms, and 830-ms conditions, onset of the squares did precede onset of the indicative target cue, so the squares— one of which became the mask surrounding the target—were effectively being previewed. The decrease in OSM as target duration increased could, therefore, have been due to weakening of mask effectiveness. To eliminate this possibility, a final experiment was run that retained most of the design characteristics of Experiment 4 but with target duration and mask duration unconfounded. Experiment 5 Participants In Experiment 5 there were 9 female and 9 male participants, as described for Experiment 4. They were aged between 19 and 27 years. Method The design and procedure were identical to those of Experiment 4 except in two ways. First, onset of outline squares around each of the search display items always occurred simultaneously with onset of the color cue designating the target. That is, the squares Figure 8. The size of the masking effect as a function of target duration in Experiment 4. The masking effect equals the difference in the percentage of correct responses in mask and control conditions. TARGET DURATION AND SUBSTITUTION MASKING 1275 onset 70 ms before offset of the search display, at which time all the squares also offset except, in the masking condition, the one around the location of the target. Second, because pilot testing revealed that participants found target discrimination very difficult under these conditions, the colors of everything in the display sequence was reversed to make the task easier. Squares and distractor items were all blue, with the target white for the final 70 ms of its presentation. This meant that, when cued by the color change, the target increased in luminance to become the brightest item in the display whereas in Experiment 4 it had become the least bright. 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. Results and Discussion Percentage correct responses and standard errors for each condition are shown in Figure 9. The masking effect as a function of target duration is shown in Figure 10. The results are similar in important respects to those of Experiment 4. Once again, there is a substantial OSM effect at the shortest target duration, and this reduces as target duration increases, although it does not completely disappear even at the 830-ms duration. The increase in control condition performance between the shortest and next shortest durations, which was observed in all four previous experiments, is now all but absent. Control performance is more or less constant across target durations whereas performance in the masking condition improves as target duration increases. We cannot account for fine detail differences between experiments that differ in several small but potentially influential respects. As previously, we wish to concentrate on the overall pattern of data and what they say about our initial question. Once again the data demonstrate unequivocally that OSM reduces with target duration. The percentage correct data were entered into a 2 ⫻ 4 repeatedmeasures ANOVA. There was a significant main effect of masking condition, F(1, 17) ⫽ 52.77, MSE ⫽ 97.32, ␩2p ⫽ .76, p ⬍ .001, and the main effect of target duration approached significance, F(3, 51) ⫽ 2.74, MSE ⫽ 60.42, ␩2p ⫽ .14, p ⬍ .06. There was also a significant interaction between the two factors, F(3, 51) ⫽ 4.13, Figure 9. Percentage correct performance in Experiment 5 as a function of target duration for control and mask trials. Error bars indicate ⫾ 1 standard error. Figure 10. The size of the masking effect as a function of target duration in Experiment 5. The masking effect equals the difference in the percentage of correct responses in mask and control conditions. MSE ⫽ 40.96, ␩2p ⫽ .20, p ⬍ .05, reflecting the decrease in OSM with increasing target duration. Comparisons between control and mask conditions showed a significant difference for all four durations, although with t values diminishing as duration increased: with a 70-ms target, the t test comparison of the masking effect was, t(17) ⫽ 7.27, p ⬍ .001; with an 870-ms target, the comparison was t(17) ⫽ 2.81, p ⬍ .05. The results of Experiment 5 resemble those of Experiments 4 in showing a large reduction in OSM with longer duration targets. This strongly suggests that the reduction in OSM observed in Experiment 4 was due to target duration not mask duration. However, whereas in Experiments 2 and 3 OSM was eliminated for a 500-ms target duration and in Experiment 4 for a 830-ms target duration, in Experiment 5 a significant effect remained even at 830-ms duration, though considerably reduced from the size of the effect found with a 70-ms display duration. We can only conjecture why, unlike in Experiments 2 to 4, the OSM effect in Experiment 5 was only reduced and not eliminated at the longest target duration tested. One possibility is that on some trials when the target item changed color and increased in brightness, it was perceived as a new object rather than as an altered old object; brightness increases are particularly likely to produce such an effect (Rauschenberger, 2003). As a new object it would be subject to OSM in the same way as normal short duration targets, and this would act to dilute the reduction in OSM associated with the increasing nominal duration of the target. A further possibility has to do with the fact that display items in Experiment 5 were abruptly surrounded by squares at 0, 140, 450, and 760 ms following their own onset and simultaneously with the color change that cued the target. Attentional competition from these multiple onsets may well have delayed focusing of attention on the target. Because, as discussed in relation to the results of Experiment 2, OSM is most easily obtained when attention is diffuse, the delay in the arrival of attention at the target could have allowed OSM to persist for longer than in the previous experiments. GELLATLY, PILLING, CARTER, AND GUEST 1276 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. General Discussion We began by asking if OSM could be obtained with targets that had been presented for longer than the very brief duration typical of masking experiments. On the basis of a wealth of evidence that observers are able to represent and report very little about the properties of objects that are not at the current focus of attention (e.g. O’Regan et al., 1999; Turatto et al., 2002), and particularly in the light of the results from Wolfe et al. (2006, Experiment 2), we predicted that robust OSM would be maintained in such circumstances. The results of our experiments have falsified that prediction. Experiment 1 revealed the complications in deciding exactly what comparisons should be made to calculate OSM but indicated that the effect decreases when target duration increases from 17 ms to 500 ms. Experiment 2 confirmed this effect of duration, with OSM eliminated for targets of 500-ms duration. Experiment 3, using mixed displays of old and new items, again found OSM was eliminated at target duration of 500 ms. Experiments 4 and 5 cued the target by color and employed displays in which every search item was surrounded by an outline square either throughout its presentation (Experiment 4) or for the final 70 ms of that period (Experiment 5) with, on masking trials, the square mask around the target item trailing for 500 ms after display offset. In both cases, OSM decreased with increasing target duration, although the time course of the decrease was more extended than in Experiments 1 to 3, and in Experiment 5 a significant—though markedly reduced—OSM effect persisted even with a target duration of 830 ms. Our findings are contrary to what we had expected. They can be seen as providing support for the theory of Di Lollo et al. (2000), who interpreted OSM as an effect that arises during early perceptual processing of newly onset display items. It would seem from these five experiments that a visual object is strongly subject to OSM immediately following onset but either not at all or only weakly when a target has been present for some hundreds of milliseconds. In terms of our original question, this means that the transition from a pre-attentive representation of the target to an attentional representation of it does not give rise to OSM. However, before accepting this conclusion, account needs to be taken of another possible interpretation of the results. A reduction in OSM with increasing target duration might be unsurprising if it were simply supposed that longer target displays allow more items to be to be read into VSTM, and that once coded into VSTM items are immune from OSM. We can see two ways in which this proposal can be formulated but we contend there are both strong counter arguments and empirical evidence against both of them. The first form of the argument starts with only an assumption that the number of items that can be reported (from VSTM) should be an increasing function of display duration. This could account for the general pattern of data for the masking condition in all experiments because in every case performance does increase with display duration before reaching a plateau, at around 200 ms in Experiments 2 to 4 and somewhat later in Experiment 5. However, it equally follows from this argument that performance should increase in a parallel for the control conditions of each experiment, which clearly is not the case. Control performance is a broadly U-shaped function of duration in Experiments 2 to 4, and is almost flat in Experiment 5. Other factors, such as cue effectiveness, must also determine the number of these control condition items to be reported. However, as with the effect of duration on the number of items coded into VSTM, these other factors should equally apply to the masking condition as well as the control condition. The notion that longer display duration equates to more items being reportable from VSTM simply cannot account for the results of the experiments unless additional, and not immediately obvious, assumptions are added to it. A second, more elaborated version of the above proposal could go like this. Suppose, first of all, that in the short duration conditions of our experiments no items are ever coded into VSTM before display offset but that in the control condition the cued target can often be read out from iconic memory into VSTM and reported. In the masked condition, however, the target can rarely if ever be recovered from iconic memory and coded into VSTM. Transfer from iconic memory to VSTM is precisely what, on this account, OSM is assumed to prevent, so performance necessarily is close to chance. Now further suppose that in the longer duration conditions items do get read into VSTM before cue/mask onset, that the maximum capacity of VSTM is either 4 or 5 items, and that items in VSTM are not subject to OSM. Having four or five display items coded into VSTM prior to onset of the cue/mask leaves only eight or seven items that can be subject to masking in the long duration conditions in contrast with all 12 items being subject to masking in the short duration conditions. In the long duration condition there will be a 33% or a 42% chance of the target being one of the items in VSTM, and so being immune to OSM. On these assumptions, OSM for long duration displays should be reduced by 33% or 42% relative to the OSM obtained with short duration displays. How do these predictions compare to the results of our experiments? As already noted, Experiment 1 lacked appropriate control conditions for the 200-ms and 500-ms durations, and therefore these data cannot be tested against the predictions. For Experiments 2, 3, 4, and 5, however, comparing the percentage masking effects for the shortest and longest duration conditions—for example, a 20% OSM effect down to 4% in Experiment 2 (see Figure 4)—the obtained reductions in OSM were, respectively, 80%, 89%, 71%, and 53%. These figures considerably exceed even the prediction based on the generous assumption of a five-item capacity for VSTM. So far, we have presented arguments as to why our results cannot be explained on the simple assumption that longer duration displays (within the range studied here) allow for more items to be reported from VSTM. We think the fact that the control and masking functions in our experiments are so different illustrates the importance of a matched control condition for every experimental condition when studying OSM. In addition to our arguments, however, is empirical evidence against the assumption that a longer display duration means more items can be reported from VSTM. Previous investigators have found that, for the sort of durations employed in our experiments, extending display duration does not, in fact, lead to an increase in the number of items reported. Di Lollo & Dixon (1988) presented 15 letters in a circle around fixation for durations from 10 ms to 500 ms, one of which was subsequently cued for report with a radial bar occurring between 0 and 200 ms after display offset. For all cue delay conditions, including 0, the number of items reported actually decreased as display duration increased. Although, unlike in our experiments, cue and target did not temporally overlap, this result clearly contradicts the assumption that longer display durations 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. TARGET DURATION AND SUBSTITUTION MASKING necessarily result in more items entering VSTM. Similarly, in their influential paper, Luck and Vogel (1997) reported that extending display duration from 100 ms to 500 ms had absolutely no effect on the number of items held in VSTM for displays of four, eight, or 12 items. There are, therefore, both logical and empirical grounds for rejecting an explanation of our results as being due to longer duration displays resulting in more items being available in VSTM. These considerations take us back to the apparent paradox identified in the discussion of Experiment 2; namely, that although increased target duration does not appear to result in enhanced knowledge of the items in the display, it does seem to result in what knowledge there is becoming decreasingly susceptible to masking. How can we make sense of this finding? Vogel, Woodman, and Luck (2006) proposed that an abrupt onset object is subject to perceptual analyses at several levels that result in perceptual and conceptual representations in a variety of widely distributed perceptual and post-perceptual neural systems. Following Treisman (1988), these can be assumed to include representations of separate features such as color, orientation, and so forth as well as a token representation of the object itself. Object tokens (or “object files”) are hypothesized midlevel representational structures that mediate our perception of objects as they change over time and space (Kahneman & Treisman, 1984; Kahneman, Treisman, & Gibbs, 1992) There is both electrophysiological (Woodman & Luck, 2003) and imaging (Carlson, Rauschenberger, & Verstraten, 2007) evidence that various early representations are formed in relation to objects that, due to OSM, cannot be reported. OSM could, therefore, be thought of as interfering with the active consolidation of these early representations into VSTM. We shall consider in greater detail below what this might mean. However, we will discuss somewhat related ideas to the above that have been proposed by Sligte et al. (2008) building on recent demonstrations of retrocuing of brief visual displays. These studies revealed that more information about briefly shown display items is retained than had previously been appreciated (Griffin & Nobre, 2003; Landman et al., 2003; Lepsien & Nobre, 2007; Makovski & Jiang, 2007). Sligte et al. proposed that intermediate between iconic memory (Sperling, 1960), which is retinotopic, and what they call robust VSTM is a form of fragile VSTM storage, which is spatiotopic. Fragile VSTM seems to correspond to one or more of the distributed representations posited by Treisman (1988) and Vogel et al (2006). Landman et al (2003) and Sligte et al (2008) claimed that although the representations in fragile VSTM remain volatile unless consolidated into robust VSTM, they already contained information bound into coherent object representations. This implies that coherent object representations can arise in the absence of focused attention. A similar implication seems to be embedded in Di Lollo et al.’s (2000) theory of OSM, according to which focused attention reduces the iterations required for the visual system to lock onto an interpretation of an item, that is, suggesting that locking onto an interpretation can, if necessary, occur in the absence of focused attention. Perhaps, then, attention serves to consolidate the existing interpretation into robust VSTM, or to put it another way, serves to incorporate the unconscious interpretation into visual consciousness. The basis for this claim comes from a study by Landman et al (2003) in which participants had to detect a change in texture defined figures presented in two displays 1277 separated by a 1,600-ms grey interval. A spatial cue appearing during the grey interval enhanced performance relative to one appearing during presentation of the second display. Moreover, the cue was equally as effective when participants did not know whether the change would be of size or orientation as when they did know. Landman et al. interpreted this as showing that these two features were already bound into coherent object representations. However, the data also are consistent with the possibility that rather than being properly bound into coherent object representations, the features were only pre-attentively “bundled” together; that is, remaining somewhat independent of one another (Wolfe & Cave, 1999). For example, a large horizontal rectangle might be represented not explicitly as such but as “thingness” (object token), largeness, and horizontality, with possibly an implicit index of association between these. Gellatly, Pilling, Cole, and Skarratt (2006) observed OSM that was selective for either color or orientation; a difference in color between target and mask elements facilitated report of target color but not of target orientation, and vice versa. Observers were aware of an item having been presented at the target location—a token representation— but had limited and unequal access to information about the different features of the object. The latter would be an example of features being bundled but not bound together. Bundled features would, to a greater or lesser degree, be associated with the token representation. However, spatial attention is required to properly bind features into a type representation (Treisman & Schmidt, 1982), which itself requires binding to the appropriate token, a process thought sometimes to fail as, for example, in repetition blindness (Kanwisher, 1987). Drawing together the preceding ideas allows us to propose an explanation of our data. If masking is not reduced at longer duration because of an increase in awareness of item identities how else can the data be explained? Lleras and Moore (Lleras & Moore, 2003; Moore & Lleras, 2005) argued that OSM occurs because of interference at the level of object token representations of the mask and target. According to Lleras and Moore, OSM was a consequence of an updating process in which a single object token representing both target and mask had its stored features about the target “overwritten” by those of the mask. Thus, when a trailing mask surrounded a target, masking occurred because the visual system wrongly construed the trailing mask to be a continuation of the single object “target plus mask” and accordingly updated the representation of the object features, meaning that the original representation of the target features was lost. To put this another way, if the target disappears before its features have become bound together and bound to the single token established for target plus mask, then only the features of the mask get bound together and bound to the token. This is similar to but divergent from the ideas put forward by Di Lollo et al. (2000) in which masking was conceived as being a consequence of competition between different object representations rather than within a single object representation. The strongest evidence for the object updating account was the finding that OSM is produced even when the mask is presented at a distant location from the target, providing that the two objects are linked by an apparent motion signal that causes them to be perceived as a single object in motion, and having the properties of the mask. In the current experiments, our finding of substantial OSM with short duration targets, but weak or absent OSM with longer duration targets may be because of differences in the way object tokens are 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. 1278 GELLATLY, PILLING, CARTER, AND GUEST allocated to old and new items. With short target durations there may be a tendency for mask and target initially to be treated as a single perceptual object by the visual system (i.e., represented by a single object token) leading, as described above to the target features being overwritten (masked) when the target offset is trailed by the mask (Lleras & Moore, 2003; Moore & Lleras, 2005). When the target is presented on-screen for an extended duration before onset of the surrounding mask, the visual system is more likely to interpret the two as separate perceptual objects and thus represent them with separate object tokens, meaning that no overwriting occurs if the mask trails the target offset. The attenuation of OSM that is found when a mask is shown for a noticeable period before target onset (Neil et al., 2002; Tata & Giaschi, 2004) can be explained by a similar process: When the mask is shown well before the onset of the target the visual system may be more likely to individuate the two items (mask and target) with separate object tokens than when the two have similar onsets in time (see Lim & Chua, 2008, for a similar explanation of OSM mask preview benefit). Thus our findings suggest that having target and mask onsets that occur in a similar time window is a precondition for OSM. Our Experiment 4, however, shows that even this condition is insufficient for OSM with longer duration target displays. In Experiment 4 target and mask onset simultaneously in all duration conditions, yet OSM was evident only at the shorter stimulus durations. What this suggests is that the failure to individuate target and mask as separate objects occurs when the two have similar onsets and the target is presented only briefly. With longer presentations of the target, target and mask come to be allocated separate object tokens, meaning that target features are no longer vulnerable to being overwritten with mask features. Neill et al (2002) and Lim & Chua (2008) found the effect of mask preview did not differ for durations from 100 ms up to 1,500 ms. Their finding of no effect of mask preview duration contrasts with the present results that show target (and distractor) preview duration does affect OSM, with OSM decreasing as target preview increases up to as much as 760 ms (Experiment 5). This could be because preview of masks and preview of (potential) targets do, indeed, have asymmetrical consequences. However, we suspect the difference is more likely to reflect differences in experimental method and design. Neill et al. used displays of just two items and two masks. It is possible that spatially and temporally proximate items are more likely to be individuated the fewer of them that are present in a display. This would make sense if, for example, there is a limit on the number of object tokens that can be established and maintained at one time, or if attention is required to individuate items that are spatially and/or temporally close together. It is possible that with larger numbers of display items and masks, the duration of mask preview would influence OSM in the same way that preview of long duration targets did in the present experiments. Lim and Chua did employ larger displays than Neill et al., but the increase was only from two to four items. In addition, there is a problem with Lim and Chua’s study in that it did not include control conditions matched to each masking condition, as in our experiments, but instead compared performance across various masking conditions. This feature of their design renders their null effect for mask preview duration open to question. Further experiments are required to determine whether preview of mask objects and preview of search display objects have similar or different consequences for OSM when number of objects and other factors are adequately controlled. One puzzling aspect of the current findings is that they appear to conflict with Wolfe et al.’s (2006, Experiment 2) report that long duration targets can be very effectively masked. Several differences between their experiment and ours might account for the discrepant results. First, in Wolfe et al.’s study targets were cued only after offset, whereas in our experiments there was temporal overlap between target and cue. It is possible that having the cue/mask onset simultaneously with target offset may increase the likelihood that the mask is seen as a transformation of the target, so not necessitating a separate token and thereby reinstating masking with long duration targets. However, yet another explanation could lie in differences between the masks themselves. In Wolfe et al.’s experiment the mask (a grey rectangle) fully occluded the target. In our experiment the mask surrounded the target location without occlusion. It is possible that occlusion is a necessary condition for masking long duration targets. Finally, the set size of potential targets we presented was smaller than in Wolfe et al. (12 items vs. 32). Set size is a prime determinant of OSM (Di Lollo et al., 2000), and it may be that this difference is crucial, once again possibly because of limits on the number of tokens that can be established and maintained at one time. We are currently conducting experiments to investigate the influential factors that explain the difference in results. Concluding Comments We set out to ask if OSM is associated only with the early neural processes that follow the onset of a stimulus object or if it can arise in other ways, possibly being associated with the changed representation of a previously unattended object when it comes to be attended. Our data show clearly that OSM tends to be eliminated, or greatly decreased, the longer an object is present before being identified as a target. This demonstrates that it does not arise during the transition from a pre-attentive representation of the target to an attentional representation of it. The conclusion we reached is that the important factor in determining the presence or absence of OSM may be whether target and mask are associated with distinct object tokens (Lleras & Moore, 2003). Moore and Lleras (2005) showed that short duration targets are less vulnerable to masking if they differ in color from the surrounding mask because, it can be supposed, the two have a raised likelihood of being assigned separate object tokens rather than a shared token. Our results suggest that a temporal difference between target and mask can similarly serve to individuate the two display elements at the object token level and consequently reduce target susceptibility to masking (see also Neill et al., 2002; Tata & Giaschi, 2004; see also, Lim & Chua, 2008). OSM is, indeed, associated with early neural processes following display onset—if target and mask are represented by a single token— but can also be determined by later occurring neural processes if the target and mask subsequently come to be individuated and associated with separate tokens. References Alvarez, G. A., & Cavanagh, P. (2004). The capacity of visual short-term memory is set both by visual information load and by number of objects. Psychological Science, 2, 106 –111. 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. TARGET DURATION AND SUBSTITUTION MASKING Bischof, W. F., & Di Lollo, V. (1995). Motion and metacontrast with simultaneous onset of stimuli. Journal of the Optical Society of America A, 12, 1623–1636. Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10, 433– 436. Carlson, T. A., Rauschenberger, R., & Verstraten, F. A. J. (2007). No representation without awareness in the lateral occipital cortex. Psychological Science, 18, 298 –302. Di Lollo, V., & Dixon, P. (1988). Two for forms of visual persistence in visual information processing. Journal of Experimental Psychology: Human Perception and Performance, 14, 671– 678. Di Lollo, V., Enns, J. T., & Rensink, R. A. (2000). Competition for consciousness among visual events: The psychophysics of reentrant visual processes. Journal of Experimental Psychology: General, 129, 481–507. Dubois, J., Hamker, F. H., & VanRullen, R. (2009). Attentional selection of noncontiguous locations: The spotlight is only transiently “split” Journal of Vision, 9(5), 3, 1–11. Enns, J. T. (2004). Object substitution and its relation to other forms of visual masking. Vision Research, 44, 1321–1331. Enns, J. T., & Di Lollo, V. (1997). Object substitution: A new form of masking in unattended visual locations. Psychological Science, 8, 135– 139. Enns, J. T., & Di Lollo, V. (2000). What’s new in visual masking? Trends in Cognitive Science, 4, 345–352. Gellatly, A. R. H., & Cole, G. G. (2000). Accuracy of target detection in new object and old object displays. Journal of Experimental Psychology: Human Perception and Performance, 26, 889 – 899. Gellatly, A. R. H., Pilling, M., Cole, G. G., & Skarratt, P. (2006). What is being masked in object substitution masking? Journal of Experimental Psychology: Human Perception and Performance, 32, 1422–1435. Griffin, I. C., & Nobre, A. C. (2003). Orienting attention to locations in internal representations. Journal of Cognitive Neuroscience, 15, 1176 – 1194. Kahneman, D., & Treisman, A. (1984). Changing views of attention and automaticity. In R. Parasuraman & D. A. Davies (Eds.), Varieties of attention (pp. 29 – 61). New York, NY: Academic Press. Kahneman, D., Treisman, A., & Gibbs, B. J. (1992). The reviewing of object files: Object-specific integration of information. Cognitive Psychology, 24, 175–219. Kanwisher, N. (1987). Repetition blindness: Type recognition without token individuation. Cognition, 27, 117–143. Landman, R., Spekreijse, H., & Lamme, V. A. F. (2003). Large capacity storage of integrated objects before change blindness. Vision Research, 43, 149 –164. Lepsien, J., & Nobre, A. C. (2007). Attentional modulation of object representations in working memory. Cerebral Cortex, 17, 2072–2083. Lim, S. W. H., & Chua, F. K. (2008). Object substitution masking: When does mask preview work? Journal of Experimental Psychology: Human Perception and Performance, 34, 1108 –1115. Lleras, A., & Moore, C. M. (2003). When the target becomes a mask: Using apparent motion to isolate the object component of objectsubstitution masking. Journal of Experimental Psychology: Human Perception and Performance, 29, 106 –120. Luck, S. J., & Vogel, E. K. (1997). The capacity of visual working memory for features and conjunctions. Nature, 390, 279 –281. Macknik, S. L., & Martinez-Conde, S. (2007). The role of feedback in visual masking and visual processing. Advances in Cognitive Psychology, 3, 125–152. Makovski, T., & Jiang, Y. V. (2007). Distributing versus focusing attention in visual short term memory. Psychonomic Bulletin and Review, 14, 1072–1078. 1279 Moore, C. M., & Lleras, A. (2005). On the role of object representations in object substitution masking. Journal of Experimental Psychology: Human Perception and Performance, 31, 1171–1180. Neill, W. T., Hutchinson, K. A., & Graves, D. F. (2002). Masking by object substitution: Dissociation of masking and cuing effects. Journal of Experimental Psychology: Human Perception and Performance, 28, 682– 694. O’Regan, J. K., Rensink, R. A., & Clark, J. J. (1999). Change blindness as a result of “mudsplashes.” Nature, 398, 34. Pelli, D. G. (1997). The videotoolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437– 442. Rauschenberger, R. (2003). When something old becomes something new: Spatiotemporal object continuity and attentional capture. Journal of Experimental Psychology: Human Perception and Performance, 29, 600 – 615. Reiss, J. E., & Hoffman, J. E. (2006). Object substitution masking interferes with semantic processing: Evidence from event-related potentials. Psychological Science, 17, 1015–1020. Reiss, J. E., & Hoffman, J. E. (2007). Disruption of early face recognition processes by object substitution masking. Visual Cognition, 15, 789 – 798. Sligte, I. J., Scholte, H. S., & Lamme, V. A. F. (2008). Are there multiple visual short term memory stores? PLoS ONE, 3(2), 1–9. Sperling, G. (1960). The information available in brief visual presentations. Psychological Monographs: General and Applied, 74(11), 1–30. Tata, M. S., & Giaschi, D. E. (2004). Warning: Attending to a mask may be hazardous to your perception. Psychonomic Bulletin and Review, 11, 262–268. Treisman, A. M. (1988). Features and objects: The fourteenth Bartlett memorial lecture. Quarterly Journal of Experimental Psychology, 40, 201–237. Treisman, A. M., & Schmidt, H. (1982). Illusory conjunctions in the perception of objects. Cognitive Psychology, 14, 107–141. Turatto, M., Angrilli, A., Mazza, V., Umilta, C., & Driver, J. (2002). Looking without seeing the background change: Electrophysiological correlates of change detection versus change blindness. Cognition, 84, 1–10. Vogel, E. K., Woodman, G. F., & Luck, S. J. (2001). Storage of features, conjunctions and objects in visual working memory. Journal of Experimental Psychology: Human Perception and Performance, 27, 92–114. Vogel, E. K., Woodman, G. F., & Luck, S. J. (2006). The time course of consolidation in visual working memory. Journal of Experimental Psychology: Human Perception and Performance, 32, 1436 –1451. Watson, D. G., & Humphreys, G. W. (1997). Visual marking: Prioritizing selection for new objects by top-down attentional inhibition. Psychological Review, 104, 90 –122. Wolfe, J. M., & Cave, K. R. (1999). The psychophysical evidence for a binding problem in human vision. Neuron, 24, 11–17. Wolfe, J. M., Reinecke, A., & Brawn, P. (2006). Why don’t we see changes? The role of attentional bottlenecks and limited visual memory. Visual Cognition, 14, 749 –780. Woodman, G. F., & Luck, S. J. (2003). Dissociations between attention, perception and awareness during object-substitution masking. Psychological Science, 14, 605– 611. Yantis, S., & Johnson, D. N. (1990). Mechanisms of attentional priority. Journal of Experimental Psychology: Human Perception and Performance, 16, 812– 825. Received January 9, 2009 Revision received October 12, 2009 Accepted October 24, 2009 䡲