Perceptual merging contributes to cueing effects

Abstract

Abstract

Abstract:

Abstract  An uninformative exogenous cue speeds target detection if cue and target appear in the same location separated by a brief temporal interval. This finding is usually ascribed to the orienting of spatial attention to the cued location. Here we examine the role of perceptual merging of the two trial events in speeded target detection. That is, the cue and target may be perceived as a single event when they appear in the same location. If so, cueing effects could reflect, in part, the binding of the perceived target onset to the earlier cue onset. We observed the traditional facilitation of cued over uncued targets and asked the same observers to judge target onset time by noting the time on a clock when the target appeared. Observers consistently judged the onset time of the target as being earlier than it appeared with cued targets judged as earlier than uncued targets. When the event order is reversed so that the target precedes the cue, perceived onset is accurate in both cued and uncued locations. This pattern of results suggests that perceptual merging does occur in exogenous cueing. A modified attention account is discussed that proposes reentrant processing, evident through perceptual merging, as the underlying mechanism of reflexive orienting of attention.

Introduction

A sudden onset in the periphery reduces the time it takes to respond to a target appearing in that same location compared to a target appearing in an alternative, equidistant location. This cueing effect, as it is usually called, is commonly used as a measure of the reflexive orienting of attention to the sudden onset (Corbetta & Shulman, 2002; Posner, 1980; Posner & Cohen, 1984). The cueing effect has been shown to be both short-lived (Posner, 1980; Posner, Nissen, & Ogden, 1978) and obligatory (Jonides & Yantis, 1988; Nakayama & Mackeben, 1989; Posner & Cohen, 1984), and these characteristics have therefore been ascribed to the exogenously driven shifts of attention. 

Attention has been suggested to facilitate information processing such that attended items reach conscious awareness more quickly (known as the law of prior entry, e.g., Titchener, 1908). In line with prior entry, exogenous attention has been studied using the speed of simple target detection, and the size of a reaction time (RT) advantage for cued targets is thought to reflect the advantage conferred by having attention focused on the target's location when it appears. For example, Posner and Cohen (1984) used a simple design with a central fixation box and two additional boxes aligned equidistantly at the horizontal midline. Boxes were “cued” by a brief brightening of the outline of the box border. Responses to targets appearing in cued boxes were faster than in uncued boxes. Many studies to date have confirmed this finding (e.g., Hein, Rolke, & Ulrich, 2006; Henderson, 1991; Jonides, 1980). The time it takes to shift this exogenous attention to a peripheral location is estimated to be between 70 and 175 ms (Carlson, Hogendoorn, & Verstraten, 2006; Müller & Rabbitt, 1989; Nakayama & Mackeben, 1989). Traditionally, attention has been compared to a moving spotlight, illuminating locations for deeper processing (Posner, Snyder, & Davidson, 1980). Extending the analogy, attention has been proposed to move like a spotlight, that is, in an analog manner such that it takes longer for attention to arrive at a location farther away from its current location than at a location closer to its current location (Shulman, Remington, & McLean, 1979; Tsal, 1983) and to “zoom” in and out like a lens (Eriksen & St. James, 1986). The spotlight model is an oversimplification with clear limits (e.g., Cave & Bichot, 1999; Driver & Bayliss, 1989), but it is still a commonly used way to describe the basic function and mechanism of attention. 

Although attention appears to be a very plausible explanation for the facilitative effects of exogenous cues on RT, at least three other explanations have been raised. First, cues could have low-level sensory effects. Unlike attention, which is typically characterized as a separate cognitive function that acts on perceptual processes by enhancing sensory signals, sensory advantages could be conferred by the cue through bottom-up changes in the properties of cells in the visual system. Posner and Cohen (1984) speculated that sensory advantages should be observed at multiple cued locations, and attention, a limited resource, could not be split to more than one cued location. Consistent with attention being responsible for cueing effects, they did not observe facilitation at a double-cue location as compared to an uncued location, and they also found that a dim cue was as effective as a bright cue. However, the control location used by Posner and Cohen was the uncued side in single-cue trials. Wright (1994), in contrast, observed cueing effects at multiple exogenously cued locations in an experiment in which targets could appear at eight different locations, so they were not restricted to only two locations as in Posner and Cohen's design. Wright's study showed faster RT to targets in cued locations in both single- and double-cue trials when these were separately compared to uncued trials under matched conditions. 

Recently, Albares and colleagues (2011) have put forward a second alternative explanation for exogenous cueing effects, which is that slower responses to uncued targets may reflect response inhibition. In their study, they included trials in which no cue preceded the target. When randomly intermixed with cued trials, targets with no preceding cue produced slower RTs than targets preceded by a cue, but this occurred whether that cue appeared in the same location as the target (“valid cues”) or opposite location (“invalid cues”). RTs to single targets within a block in which no cues ever appeared were faster than RTs to targets preceded by invalid cues but were not slower than RTs following valid cues. This pattern is inconsistent with an account based purely on shifts of attention because RTs to targets not preceded by a cue should be faster, on average, than RTs to targets in a different location than the cue (because the cue misdirects attention away from the target) and slower, on average, than RTs to targets in the same location as the cue (because the cue attracts attention to the target). They argue that their results suggest a role for response inhibition in cueing effects. That is, responses in a typical cue-target paradigm are inhibited to prevent erroneous responses to the cue, and this could play a role in producing cueing effects. 

As a third alternative, exogenous cueing effects on RT could be driven, at least in part, by perceptual merging. That is, the two peripheral events (the cue and the target) are more likely to be perceived as one event when they occur in the same location due to reentrant processing (Bouvier & Treisman, 2010; Enns & Di Lollo, 2000; Lamme & Roelfsema, 2000; Lamme, Zipser, & Spekreijse, 2002; Treisman, 1996). Reentrant processing in vision is generally described as a two-stage process. First, feed-forward processes initiate feature detection in striate and extrastriate visual areas. Later (i.e., after 80–120 ms; Lamme et al., 2002), higher stages of visual processing feed the signal back to these early feature identification stages to reduce noise and confirm, refine, and reanalyze the signal in a series of reiterative loops generally known as reentrant processes. Di Lollo et al. (2000) proposed reentrant processing to explain object substitution masking, whereby a stimulus that surrounds a briefly presented target can act as a mask when it persists after the target disappears. They propose that the presentation of the target and mask together initiates feed-forward processing, but reentrant processing is required to confirm the identity of the target. By the time the reentrant processing stage has begun, the input to early visual areas has changed, and the representation that once contained information about both the target and the mask now only contains feature information about the mask, so the target representation is transformed or replaced by the mask. 

By a similar logic, we propose that reentrant processes may operate in exogenous cueing: The cue elicits a feed-forward loop, and reentrant processes are needed to confirm the features of the object. By the time reentrant processing occurs, the target has appeared, leading to a transformation or replacement of the cue representation with the target. A head start on the process of target detection initiated by the cue onset could thereby lead to faster RT. An explanation based on reentrant processing is consistent with the time course of cueing effects, which tend to diminish quickly, are absent when the target follows the cue by more than 200 ms, and reverse at even later intervals (Posner & Cohen, 1984). Importantly, this account of RT benefits for cued targets predicts that the target and cue will share a single perceptual representation with a common onset. In other words, the perceived onset time of a cued target will be earlier than its actual onset time. 

Cueing has also shown an effect on perceived onset times using temporal order judgments (TOJ). Cued stimuli are judged to appear before uncued stimuli; that is, the point of subjective simultaneity (PSS) is biased such that cued stimuli must physically appear after the uncued stimuli for them to be judged as simultaneous (e.g., Bashinski & Bacharach, 1980; Downing, 1988; Handy, Kingstone, & Mangun, 1996; Pieron, 1952). Although response biases are a concern in many of these studies, Spence, Shore, and Klein (2001) still find a tendency to report the cued event as occurring earlier than the uncued event even when response biases are carefully controlled. PSS shifts in the cueing literature are typically ascribed to effects of attention. However, a study by Schneider and Bavelier (2003) found that exogenous cues at multiple locations still caused a shift in PSS similar to a single cue. Provided attention is a limited resource and that prior entry is a perceptual phenomenon caused by attention, then the shift in the PSS should have decreased with increasing numbers of exogenous cues. PSS shifts following endogenous cues (i.e., an arrow at fixation pointing to one of two peripheral locations) are also observed but are much smaller than those following exogenous cues (Spence et al., 2001). 

Other temporal distortions have been observed following exogenous cues. Yeshurun and Levy (2003) observed that exogenous cueing leads to an increased two-flash fusion threshold, which is the temporal gap between two flashes needed for them to be seen as separate events. Yeshurun and Levy found that two-flash fusion occurred for longer temporal separations at the cued relative to the uncued location. This finding suggests that exogenous cueing impairs detection of brief temporal intervals, which stands in contrast to the improvements in spatial discrimination that have been shown to accompany attention (e.g., Bashinski & Bacharach, 1980; Bonnel, Possamaï, & Schmitt, 1987; Cheal & Lyon, 1991; Downing, 1988; Henderson, 1991; Klein & Dick, 2002; Yeshurun & Carrasco, 1999). Yeshurun and Levy replicated these improvements in spatial discrimination at the cued location, demonstrating that they exist simultaneously with deficits in temporal discrimination. Similarly, Hein et al. (2006) showed that the ability to correctly indicate which of two stimuli occurred first was impaired if these two stimuli appeared after an exogenous cue at that location, and Montagna and Carrasco (2006) showed that exogenous cues increased the perceived flicker rate of subsequent targets. These studies are consistent with a general distortion in the perception of target onset following exogenous cues although it is important to note two caveats: First, an uninformative arrow cue (which also causes reflexive attention shifts; Tipples, 2002) led to TOJ distortions in Hein et al., suggesting that an event in the same location as the target may not be a necessary condition for temporal distortions; second, Chica and Christie (2009) found evidence for improved, rather than impaired, temporal discrimination at cued locations and criticized previous studies for not accounting for speed-accuracy trade-offs. 

Here we have designed a study that examines both RT to detect a target and the perceived onset time of a target to test predictions based on the traditional attention account as well as the three alternative explanations for cueing effects described above. First, we will outline the predictions for RT. We examine RTs to targets that follow a cue on the same side or the opposite side (precues), targets that precede a cue on the same or the opposite side (postcues), and targets alone (no cue). The accounts of cueing reviewed above make different predictions about the pattern of results we should obtain in this experiment, which are illustrated in Figure 1, left side. An attention-based account (here referred to as the traditional attention account) predicts that RTs to targets appearing after a same-side precue should have the fastest RTs, and targets appearing after an opposite-side precue should have the slowest RTs because attention is misdirected to the cue. No-cue and postcue conditions should produce similar RTs, intermediate between same side and opposite side precue RTs (Figure 1, top panel, left). Under both the sensory enhancement (Wright, 1994) and response inhibition (Albares et al. 2011) accounts, no-cue trials and postcue trials should produce the slowest RTs because no sensory enhancement or disinhibition of responses is elicited by the cue in these conditions (Figure 1, middle, left). Opposite-side precues could facilitate RT through a nonspatial enhancement of sensory processing or a general inhibition of all motor responses, and same-side cues could provide a more spatially specific sensory or response-related enhancement as observed by Wright (1994) and Albares et al. (2011). Finally, under the perceptual merging account, precues should produce the well-established cueing effect as described above. In the postcue condition, on the other hand, feed-forward processing will be initiated by the target, but by the time reentrant processing begins, the target has been replaced by the cue, which could interfere with target processing and slow responses (similar to object substitution masking, Di Lollo et al., 2000). Under perceptual merging, therefore, RTs to targets appearing in the same location as the postcue should be slower than to targets appearing in the opposite location (Figure 1, bottom panel, left). 

Figure 1

 

The prediction for (A) traditional attention, (B) response inhibition, and (C) perceptual merging accounts for MRT and POT. (A) The traditional attention account suggests that same-side precue target responses are facilitated (fastest MRT) and opposite-side precue target responses are prolonged (slowest MRT) whereas single-event and postcue responses fall between them. The POTs similarly reflect minimal error for no-cue and postcue targets but a perceived error ahead of the onset for precue targets (for simplicity, this prediction is based on the assumption that reported time is based on the time on the clock when attention shifts away from the center toward the cue and/or target, but we have also added fainter bars for an alternative prediction, see text for details). (B) Response inhibition and sensory enhancement predict fastest MRTs for same-side precue targets but also relatively facilitated RTs for opposite-side precue targets due to the disinhibitory effect of the cue. Single-event and postcue targets are slowest because no disinhibition has occurred before the target onset. POTs remain unaffected for all conditions. (C) Perceptual merging: As the onset of the target is merged with the cue onset, same-side precue targets have the fastest MRTs and are also perceived ahead of their onset. Precue opposite-side and single-event targets are unaffected by perceptual merging. The pattern reverses for the postcue trials: MRTs are largest for same-side targets because the cue interferes with target processing, and opposite-side targets and no-cue targets are unaffected by perceptual merging. Postcue target POT may be unaffected by the cue or may merge with it to be perceived later: Again, we have added fainter bars to indicate an alternative prediction that would also be consistent with the perceptual merging account. See the text for more details.

The second part of the experiment will measure the perceived onset time of the target directly by asking participants to indicate the time of the target onset using a clock presented at fixation. This method, borrowed from Carlson et al. (2006), eliminates the issues of response biases inherent in many time judgment tasks used in cueing studies (Spence et al., 2001). Here again, we can consider the predictions about what the results should be under the four different explanations for the cueing effect (Figure 1, right side). Under the traditional attention account, we assume that the precue would pull attention away from the clock and that there is not enough time for attention to be reoriented back to the clock before the target appears (if the cue-target interval were long enough for reorienting to occur, we should observe inhibition of return rather than cueing effects; Posner and Cohen, 1984). Given this, participants may use the last time they remember from the clock face when the target appears as an estimate of its onset time. If so, then targets will be judged to have occurred earlier than they actually occurred, particularly when preceded by precues, irrespective of whether those precues are in the same location as the target or not (Figure 1, right side, strong bars). Alternatively, participants will wait until the target appears and then reorient attention back to the clock in order to judge its onset time (Figure 1, right side, fainter bars). The law of prior entry suggests that attended targets will be perceived as occurring earlier than unattended targets. Therefore, provided that (a) attention shifts to the target after onset, then shifts back to the clock to gauge the time, and (b) prior entry speeds target processing, facilitating this shift of attention back to the clock, then reports of cued targets will be earlier than reports of uncued targets. In other words, the prediction based on traditional views on attention bifurcates on the basis of whether target onset judgments tend to be early (in which case there should be no difference between same- and opposite-side targets) or late (in which case same-side targets will be judged as earlier than opposite-side targets). We have illustrated both predictions in Figure 1, but to foreshadow the results, the onset judgments did tend to be early, consistent with subjects basing their judgments of the target onset on the time on the clock when attention first shifted away from it. 

Both response inhibition and sensory enhancement should have no effect on perceived onset time, so no effect of cue-target order or location is predicted. Perceptual merging predicts that the perceived onset of targets appearing in the same location as a cue will be bound to the cue onset time, resulting in perceived onset times for the precued targets that are earlier than their actual onset time. In the postcue condition, we propose that the target initiates feed-forward processing, but reentrant processing from the cue degrades the target representation as described above. This may result in targets that are perceived with delayed onset in addition to interfering with response times although the prediction here is not as clear cut (we will return to this issue in the Discussion). Targets appearing opposite to the cue should not be affected by cue-target order (Figure 1, bottom, right). 

Method

Participants

Eighteen participants (six female) participated in this experiment. The average age was 22.1, SD 2.4. All participants had normal or corrected-to-normal vision. Participants volunteered (three) or were reimbursed £5 for their time. 

Apparatus

An Apple Mac Mini running Mac OS X was used to run the two tasks, programmed using Psychtoolbox 3 (Brainard, 1997; Pelli, 1997) in Matlab 2009b and displayed on a 27-cm × 35-cm, 100-Hz CRT. A chin rest ensured a viewing distance of 43 cm to the screen throughout the experiment. 

Stimuli and procedure

The two tasks took 45 min to complete. Half of the participants participated in the manual RT task (MRT task) first and then the perceived onset time task (POT task) and the other half vice versa. 

The sequences of events in a trial were essentially the same in both tasks, as illustrated in Figure 2. At the beginning of each trial, two white boxes were displayed on the horizontal midline, 7.7° to the left and to the right of the center, subtending 3.1°. In the center, a clock of 4.6° diameter with 12 large and 60 small ticks was displayed, also in white. The background was black. White letters on the top of the screen (1.2°) informed the participant to click a mouse button to initiate the trial. Upon clicking the mouse button, the text disappeared, and the hand of the clock (1.2°) appeared and began rotating clockwise, making one revolution around the clock face per second. After a random interval between 500 and 800 ms, the first stimulus (which could be either the target or the cue) could appear in one of the two peripheral boxes and remained visible for 50 ms. At 110 ms after the first stimulus onset, the second stimulus appeared (which could also be either the target or the cue). The second stimulus also remained visible for 50 ms. The cue was a dimming of the outline of the peripheral boxes, that is, the white frame was replaced with a gray frame. A dimming was chosen as pilot experiments with a brightening of the peripheral box caused masking effects. The target stimulus was a red disk, subtending 1.5°, that appeared in the center of one of the two peripheral boxes. 

Figure 2

 

An example of the trial procedure for a postcue opposite-side trial. Note that the cue is a dimming of the peripheral box.

MRT task

Participants were instructed to press the space bar as soon as the red disk appeared and to simply ignore the clock. The display remained visible until a response to the target was made or for 4000 ms from the beginning of the trial if no response was made (e.g., on catch trials). This signaled the end of the trial, and the procedure started again. The cue and target could appear in the same or in different locations, and the cue could appear before the target (precue) or after the target (postcue). On one third of all trials, only the target appeared, and it appeared either at the first stimulus onset time or at the second stimulus onset time. Fifteen percent of all trials were catch trials in which the cue appeared but no target and no response was required to confirm that participants were waiting for the target to appear to press the button. Similarly, the cue in these trials could appear at the first or the second stimulus onset time. There were 240 trials in total; these were split into four blocks of 60 trials. 

POT task

Participants reported what time was on the clock when the red disk appeared (see also Carlson et al., 2006). The clock hand kept spinning for 2 s in total and then jumped to a random location so that the final position of the hand was never a predictor of the onset of the target. At the end of this 2-s interval, participants were instructed to use the left and right arrows to adjust the hand of the clock to the position that they perceived as having been on the clock when the red disk appeared. After adjusting the hand of the clock to the appropriate position, participants were instructed to hit the space bar to confirm their response. This ended the trial. 

The conditions in the POT task were the same as in the MRT task with the exception that catch trials were not included, resulting in 192 trials in total. These were split into three blocks of 60 trials and one block of 12 trials. 

Results

MRT task

MRT to target onset was measured by subtracting the time of the onset of the target from the time of the response. The median response for each participant in each condition was calculated. The first 20 trials were excluded as practice trials. Mean proportion of responses to catch trials was 3.5%, and no participant exceeded 10%. 

The results are displayed in Figure 3. A 2 × 2 repeated-measures ANOVA with the factors order (precue, postcue) and location (same side, opposite side) was conducted. The main effects of location, F(1, 17) = 1.19, and order (F < 1) were not significant, but, together, location and order qualified for an interaction, F(1, 17) = 23.06; p < 0.001. To examine this interaction, paired-samples t tests compared cueing effects for precues and postcues separately. In precue trials, same-side targets were responded to faster than opposite-side targets, t(17) = 2.99; p < 0.01. This pattern reversed for the postcue trials in which same-side targets had slower MRTs than opposite-side targets, t(17) = 2.54; p < 0.05, supporting the hypothesis that same-side targets merge with the cue event, leading to faster responses to targets on the same side as a precue but slower responses to targets on the same side as a postcue. Importantly, this latter result in the postcue responses is inconsistent with the predictions of the traditional attention account. 

Figure 3

 

The mean of the median manual RTs. The two bars in the no-cue condition represent the data from the target appearing at the first and second intervals, respectively. Significant differences between pairs are marked with an asterisk. Note that precue and postcue opposite side responses (in gray) did not differ significantly. Error bars reflect within subject variability (Morey, 2008).

Both the traditional attention and perceptual merging accounts predict faster responses to same-side precue targets than same-side postcue targets. For opposite-side targets, however, responses to precue targets should be slower than opposite-side postcue targets if attention is guided to the wrong location by the cue. Perceptual merging, on the contrary, predicts that opposite-side precue and postcue targets should not differ. A paired-samples t test between opposite-side precue and postcue targets revealed no difference between precue and postcue opposite-side targets, t(17) = 1.09; p = 0.298, consistent with the perceptual merging account. 

Perceptual merging and attention also differ in their predictions on the difference between precue opposite-side RT and postcue same-side RT: According to the traditional attention account, precue opposite-side responses should be delayed relative to postcue same-side responses, and perceptual merging predicts the opposite with postcue same-side targets delayed relative to precued opposite-side targets. Precue opposite-side and postcue same-side responses did not differ (t < 1), consistent with neither prediction. 

It was also investigated whether a response inhibition or sensory enhancement occurred in the current design as shown by Albares et al. (2011). A response inhibition account would predict that responses to precue targets should be faster than to no-cue targets, particularly when the target and cue appear in the same location, whereas postcue trials would not differ. As can be seen from Figure 3, this was not the pattern we observed. To assess this prediction, separate one-way repeated-measures ANOVAs were conducted on the data from the two target onset intervals (as described in the Method, two different intervals in the target-only condition were used in order to match target onset time in the no-target condition to target onset time in both the precue and the postcue conditions). Each ANOVA examined the effect of three target types (no cue, same side, opposite side) on median RTs. When the target appeared in the first interval, there were significant differences between the target types, F(2, 34) = 4.93; p < 0.05. The same analysis on targets appearing in the second interval was also significant, F(2, 34) = 3.92; p < 0.05. However, single-event targets differed from precue trials marginally for same-side targets, t(17) = 2.03; p = 0.058, but not for opposite targets (t < 1). Single-event targets in the second interval differed marginally for opposite-side targets, t(17) = 1.87; p = 0.078, but not for same-side targets (t = 1.1). Importantly, and as inspection of Figure 3 indicates, significant differences occur in both precue and postcue trials, a finding that is contrary to the response inhibition theory. 

POT task

POT was measured by subtracting the time on the clock when the target appeared (i.e., actual position of the hand) from the time on the clock when participants indicated that the target came on. The median response was calculated for each participant for each condition. The first 10 trials were excluded as practice trials. 

A 2 × 2 repeated-measures ANOVA on POT with order and location as factors revealed a main effect of order, F(1, 17) = 89.61; p < 0.001, reflecting that targets were perceived as having an earlier onset with a precue than with a postcue. The analysis further revealed a marginally significant main effect of location, F(1, 17) = 3.34; p = 0.085, reflecting that same-side targets were perceived earlier than opposite-side targets. The interaction of location and order was significant, F(1, 17) = 5.20, p < 0.05. A paired-samples t test confirmed what is clear from Figure 4: For precued targets, same-side targets were perceived as having an earlier onset than opposite side targets, t(17) = 2.72; p < 0.05. There was no difference between the perceived onset of the same-side and opposite-side targets with postcues, t(17) = 0.63; p > 0.05. This analysis indicates that precue targets were perceived as occurring earlier than their actual onset. Under the traditional attention account, this should have led to equivalent POTs for same- and opposite-side targets. Earlier POTs for same versus opposite sides is therefore in line with perceptual merging. However, precued opposite-side targets are also seen as earlier than their actual onset, which was not expected under the perceptual merging account. 

Figure 4

 

The POTs of targets. The two bars in the no-cue condition represent the data from the target appearing at the first and second intervals, respectively. Significant differences between pairs are marked with an asterisk. Error bars reflect within-subject variability (Morey, 2008).

Response inhibition would suggest that none of the perceived onsets should differ from zero nor should differences for the different target types occur. Matching the MRT analysis, an additional one-way ANOVA was conducted with target type (no cue, same side, and opposite side) for targets appearing in the first stimulus interval and separately on the second stimulus interval. A one-way repeated-measures ANOVA on the first stimulus onset interval revealed no differences between the target types (F < 1). In contrast, the same ANOVA conducted on the second stimulus onset interval revealed significant differences between the perceived onsets of the target types, F(2, 34) = 31.23; p < 0.001. Both same-side, t(17) = 9.52; p < 0.001, and opposite-side targets, t(17) = 4.49; p < 0.001, differed significantly from no-cue targets. The results can be seen in Figure 4. 

Discussion

While a pure perceptual merging account is unlikely, the findings do support a contribution from perceptual merging and suggest a novel account of automatic orienting to exogenous events. In the MRT task, typical cueing effects were observed in precue trials; that is, targets appearing in the same location as the precue were detected faster than targets in the opposite location. This effect reversed for postcue trials, in which responses to targets were delayed when appearing in the same location as the postcue relative to the opposite location in line with perceptual merging. According to the traditional attention account, opposite-side precue targets should have suffered a delay relative to postcue targets; this was not confirmed, but neither were they speeded relative to same-side postcue trials as the perceptual merging account predicted. 

The results from the perceived onset task also supported the perceptual merging account. The onset of precued targets was perceived to be earlier than actual onset whereas perceived onset of no-cue and postcued targets did not differ from their actual onset time. This finding is in line with the perceptual merging account in that the precue appears to have merged with the target onset, causing the target to be judged as earlier when they are in the same location. 

The difference between same- and opposite-side precued targets may, at first glance, also be interpreted through the traditional attention account. In the Introduction, we noted that attention has been suggested to expedite the arrival of stimuli in conscious awareness (e.g., Shore, Spence, & Klein, 2001; Titchener, 1908; for a review, see Spence & Parise, 2010). If attention speeds target processing, same-side precued targets should have been perceived as earlier than opposite-side precued targets. This result was observed, but it is important to note that precued targets were also reported to happen more than 40 ms before their actual onset (also no-cue targets were generally reported as happening before their actual onset, yet the effect was significantly smaller). Attention may speed target processing along to make it “arrive” faster in consciousness, but it should not be possible for attention, as it is described traditionally, to speed up the processing of a target that has not yet occurred. As such, attention-based prior entry cannot fully explain the observed data. 

The results seem to be more consistent with perceptual merging than with the other explanations for cueing effects. That said, two aspects of the data are difficult for perceptual merging alone to explain, and we believe a modified attention account can best capture the results. First, the opposite-side precued targets were judged to appear earlier than their actual onset. Perceptual merging predicted early judgments only for the same-side precued targets because merging would only occur with events at the same spatial location. Second, we expected perceptual merging to occur for both cue and target orderings. We proposed that the feed-forward processes initiated by a precue would facilitate target detection and cause it to appear earlier than its actual onset via reentrant processes, an effect we clearly observed. By the same token, the postcue in the same location as the target was predicted to slow target detection, not only relative to the postcue in the opposite location (which we did observe), but also relative to no-cue and opposite-side precued targets (which we did not). 

Together, these findings suggest that neither the traditional attention nor the perceptual merging account alone can explain the findings, but rather it appears that a third model is needed to coherently capture the results. 

A modified attention account

One such model for the automatic orienting of attention is the initiation of a reentrant loop that selectively integrates information from a location in space. In other words, attention is the selective processing of a relevant stimulus through continuous monitoring achieved by reentrant processing. Attention is needed to extract the time on the clock, and this attention must leave the clock position with the onset of the precue reflecting the negative POTs observed in the POT task and veridical POTs in the postcue trials. With the onset of the precue, a reentrant loop is initiated at the precue location, facilitating the onset of the target through perceptual merging and thus leading to even earlier target onset judgments for same-side precue trials and relatively delayed POTs at the opposite-side precue trials. This perceptual advantage in the same-side precue trials facilitates the manual response in the MRT task and interferes with responses in the postcue same-side trials, in which the cue is processed at the target location. The data from the POT postcue condition suggested that the postcue did not influence POTs. The modified attention account would predict that, in this case, POTs were veridical: The onset of the target draws attention away from the clock and, because the target is the first event, observers correctly assert the last remembered clock position as the onset time of the target. 

The modified attention account is consistent with other descriptions of attention (Cavanagh, Hunt, Afraz, & Rolfs, 2010; Hochstein & Ahissar, 2002; Spratling & Johnson, 2004) and is also consistent with results from object substitution masking (Enns & Di Lollo, 2000), in which masking effects increase with the amount of attention needed to detect the target. To explain this pattern, Enns and Di Lollo suggest the depth of processing needed to confirm the presence of the target increases the likelihood that reentrant processing will continue after the mask has replaced the target. In our experiment, the target was relatively easy to detect, and masking by the cue (in the sense of an inability to report the target at all) did not occur in the postcue condition. A cue following the target in the same location did slow RT, consistent with processing of stimuli in the target location persisting after the target disappeared, but the postcue did not change the POT of the target. Target processing could have progressed enough that the cue interfered with executing a response but did not replace the representation of the target. It is also important to keep in mind that the target/cue spatial configuration is not symmetric; it seems likely that a dim surrounding box that appears first will easily merge with a bright inner circle that appears second. When the bright inner circle appears first, it may less readily merge with a dim surrounding box that appears second. 

One important additional consideration is that attention, in addition to speeding RT, has also been proposed to elevate the clarity of perception (Titchener, 1908). Consistent with this, it is easier to identify the features of targets at cued locations over uncued locations (Bashinski & Bacharach, 1980; Bonnel et al., 1987; Cheal & Lyon, 1991; Downing, 1988; Henderson, 1991; Yeshurun & Carrasco, 1999; Yeshurun & Levy, 2003). Henderson (1991), for example, demonstrated increased discrimination accuracy of briefly presented letters (X and O for 67 ms) at exogenously cued locations. Klein and Dick (2002) showed that exogenous cues also improve sensitivity for digits presented within a rapid serial visual presentation (RSVP) of letters. The modified attention account we propose here does account for improved clarity of attended stimuli. That is, the reentrant loop does facilitate stimulus clarity but distorts temporal perception for a brief interval as confirmed by studies that have examined perception of temporal order after exogenous orienting (Hein et al., 2006; Montagna & Carrasco, 2006; Yeshurun & Levy, 2003). 

The current findings did not replicate Albares et al. (2011). Albares and colleagues suggested that cueing effects could be the result of the release of response inhibition. Response inhibition would predict that both single targets and postcued targets should have slower RTs than precued targets. This was not observed in the MRT task of the current design, in which single targets were no slower than cued targets. Response inhibition also cannot explain effects of cueing on perceived target onset times in our second task, nor can it explain other effects on target perception, such as the RSVP cuing results by Klein and Dick (2002) noted above. One possible explanation for the discrepancy is that Albares et al. used predictive cues whereas we used uninformative cues, which are typical in studies of exogenous cueing. Perhaps the informative nature of the cues in Albares et al.'s design introduced a stronger bias to respond to the cue and therefore a stronger need to inhibit responding to the cue. If this is the case, response inhibition may contribute to the effects of informative spatial cues, typically used in studies of voluntary orienting, as opposed to the reflexive and nonstrategic effects observed in exogenous cueing studies, which typically use uninformative cues. 

In summary, the current findings are a novel contribution to our understanding of the effects of exogenous cues on subsequent target processing. Here we have observed that cue and target, if presented at the same location, are perceptually merged. We propose that this merging reflects the activity of a reentrant process that is the underlying mechanism of reflexive attention. 

Acknowledgments

This work was supported by a grant of the James S. McDonnell Foundation to A. Hunt. 

Commercial relationships: none. 

Corresponding author: Hannah M. Krüger. 

Email: [e-mail]. 

Address: Centre Attention and Vision, Laboratoire Psychologie de la Perception, Université Paris Descartes, Paris, France.