Invalid Retro-Cues Can Eliminate the Retro-Cue Benefit: Evidence for a Hybridized Account

Acknowledgement: Supported by National Eye Institute (Grant R15EY022775) and the National Institutes of Health (COBRE Grant 1P20GM103650). We thank Gideon Caplovitz, Kevin Jones and Dwight Peterson for their comments and suggestions. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

It is not always clear ahead of time what things will become important later. Unlike a mystery program, our internal camera may not linger on seemingly unimportant details: the letter opener or the strangely placed frame that only become important later. The strict capacity limits of visual working memory (VWM) and its regular updating create a tension between encoding salient new items and maintaining relevant items. Any advantage, such as a cue, to improve VWM is beneficial. Attentional cues improve detection, discrimination, and recollection of cued items whether they are presented prior to stimuli (see Carrasco, 2011; Kastner & Ungerleider, 2000; Posner & Petersen, 1990, for reviews) simultaneously with stimuli (Jiang, Olson, & Chun, 2000) or even well after (≈1,000 ms) stimulus offset (Griffin & Nobre, 2003; Landman, Spekreijse, & Lamme, 2003). This third case of retrospective cueing interests us because VWM performance significantly improves without any additional sensory information. The difference in performance between neutral and valid trials is termed the retro-cue effect (RCE). Importantly, the retro-cue paradigm offers a window into attentional processing on items held in VWM.

There has been a debate about the mechanisms of the RCE. Compelling recent work has indicated that the validly retro-cued item is likely protected from subsequent decay, but at the cost of uncued items, which decay more quickly over time (Pertzov, Bays, Joseph, & Husain, 2013). These authors improved the standard retro-cue paradigm by employing a recall probe, thereby enhancing their ability to assess VWM in a finer grained manner (see also Bays, Gorgoraptis, Wee, Marshall, & Husain, 2011; Hollingworth & Hwang, 2013; Williams, Hong, Kang, Carlisle, & Woodman, 2013). This is consistent with retro-cue studies using recognition VWM probes to show significantly lower accuracy on invalid retro-cue trials when compared with performance with neutral or valid retro-cues (e.g., Astle, Nobre, & Scerif, 2012; Griffin & Nobre, 2003; Matsukura, Luck, & Vecera, 2007). Experiments probing the fate of uncued items require neutral, valid, and invalid cues. However, only a subset of retro-cue studies have reported data on invalid trials (Astle et al., 2012; Griffin & Nobre, 2003; Landman et al., 2003; Pertzov et al., 2013), although others have included a proportion of invalid trials that are not discussed in the findings (Nobre et al., 2004). We observed a robust RCE even when the majority of retro-cues were invalid and even after instructing participants to ignore the retro-cues (Berryhill, Richmond, Shay, & Olson, 2012). A second study found that there was no penalty to VWM precision when invalid retro-cues were incorporated and responses were probed using a recall measure (Hollingworth & Hwang, 2013). Hollingworth and Hwang (2013) used recall probes and found equivalently precise representations of the items in VWM across valid (80%) and invalid retro-cue (20%) trials. However, this pattern showing no negative consequence of invalid retro-cue trials is not universal. For example, other researchers have failed to identify an RCE when the retro-cue marked an item to forget and when there were rare invalid retro-cues (10% of trials), although the RCE was present when these directed forgetting retro-cues were always valid (Williams & Woodman, 2012). Williams and Woodman’s analyses found that this behavior was present beginning with the first block of trials. They concluded that participants ignored the directed forgetting retro-cues entirely when they had the small possibility of being invalid. In short, there is a lack of consistency with regard to the effect of invalid retro-cues on the RCE that is difficult to penetrate because of the paradigmatic differences across this handful of findings. We aimed to address this inconsistency by providing a bridge between different stimulus types and paradigms. Furthermore, although previous studies that included invalid retro-cues have reported whether they observed a significant RCE; they were not designed to compare the magnitude of the RCE when invalid retro-cues were present or absent. In other words, although invalid cues had been used before, what has not been systematically probed is whether the inclusion of invalid retro-cue trials changes the RCE magnitude. Here, we addressed that very question. If the presence of invalid cues does modulate the RCE, it may explain some of the variability in the retro-cue literature and it will inform several theoretical accounts of the RCE, namely, prioritization and protection (Matsukura et al., 2007). The prioritization account suggests that the cued item is compared with the probe item first, whereas uncued items remain accessible for subsequent comparisons. Importantly, all memoranda are subject to similar decay profiles. The protection account argues that the cued items are protected from decay through the focusing of internal attention, whereas the uncued items are subject to more rapid decay. A significant reduction in the RCE when invalid retro-cues are included would support the protection account, whereas consistent decay in performance across conditions without an interaction would support the prioritization account. Therefore, by including the invalidly cued trials, and comparing the RCE between with and without invalid retro-cue experiments, we aimed to have a better understanding on the mechanisms of the RCE.

This study asked whether the RCE is modulated by inclusion of invalid trials. We tracked VWM decay over an extended time period to better characterize the fate of cued and uncued items. Finally, we aimed to supply a missing bridge between the retro-cue literature using recognition and the newer findings employing recall. We completed two pairs of experiments varying probe type (Experiments 1a and 1b: recognition; Experiments 2a and 2b: recall), and cue type (Experiments 1a and 2a: neutral, valid; Experiments 1b and 2b: neutral, valid, invalid). We predicted that Experiments 1a and 1b would confirm the presence of the RCE over time with or without invalid trials. Experiments 2a and 2b, were included to provide a more sensitive, recall-based measure of the RCE. Interexperiment comparisons were included to test whether invalid retro-cue trials changed the magnitude of the RCE and whether they replicated previous decay patterns for uncued items.

These experiments tested the effect of invalid retro-cues on RCE magnitude and the decay of cued and uncued items over the full VWM range (100–24,000 ms). Previous work has demonstrated that the RCE shows some reduction with increased second delay durations of 2–15 s when neutral and valid retro-cue conditions are included (Nobre, Griffin, & Rao, 2008). Both experiments tested VWM using recognition. Experiment 1a consisted of valid and neutral retro-cue trials; Experiment 1b included invalid retro-cue trials. This between-subjects design was necessary due to the different trial types across experiments.

Participants

Experiments 1a and b tested groups of 20 volunteers from the university psychology pool (Experiment 1a: Mage = 22.25 year; range: 18–45 years; 15 female; Experiment 1b: Mage = 21.2 years; range: 18–43 years; 17 female) for extra course credits. The internal review board of the University of Nevada approved all protocols.

Stimuli

Stimuli and cues were presented on a uniform black background. The VWM display consisted of four unique equiluminant color disks (diameter: 5°) selected from a set of 10 disks. Disks were displayed equidistantly (6.5°) from fixation in a square pattern. Neutral retro-cues (“X”; 4° × 4°) and valid arrow retro-cue (4.5° × 4°) were white.

Experimental design and trial sequence

Participants sat 57 cm from a 17-in LCD monitor (Dell 1707 FPc or Samsung SyncMaster 172N). The experiment was programmed in E-Prime (Psychology Software Tools, Sharpsburg PA). Participants were instructed to remember color−location conjunctions. Participants were informed that the “X” cues were not informative and that the arrow cues were informative. In Experiment 1b, participants were informed that the arrow retro-cues were informative but sometimes misleading. To characterize RCE magnitude over time, we used second delay durations of 100, 800, 2,400, and 24,000 ms to investigate the effect of longer second delay durations as well as effect of inclusion of the invalid retro-cues. These delays were selected to bookend and extend our previous work that has shown the RCE magnitude is consistent at second delay durations ranging from 300–700 ms (Tanoue & Berryhill, 2012).

Trials began with the presentation of a white fixation cross (1,500 ms), followed by the VWM array (300 ms). After an unmasked first delay interval (1,000 ms), the retro-cue (100 ms) appeared. The retro-cue was followed by a second delay (100, 800, 2,400, or 24,000 ms; see Figure 1a). These durations were randomized and equally probable. Finally, a probe screen appeared containing a single probe item and three empty annuli. The task was to decide whether the probe item matched the original stimulus. Participants made an unspeeded response by pressing the “Y” key for a match (50%) and the “N” key for a nonmatch. In Experiment 1a, trials were randomized and equally split between valid (50%) and neutral (50%) trials; in Experiment 1b, trial types were randomized across valid (40%), neutral (40%), and invalid (20%) trials. Participants completed 200 trials in four blocks separated by rest breaks. In all experiments, participants performed an auditory suppression task by repeating different one-syllable words during each block of trials to minimize verbal encoding. Participants were audible through the testing room door and monitored for compliance. In all analyses, Greenhouse−Geisser corrections were applied where sphericity was violated and Bonferroni correction was applied to pairwise comparisons.

Experiment 1a

VWM accuracy was subjected to a 2 × 4 repeated-measures analysis of variance (ANOVA) with the within-subjects factors of retro-cue condition (valid, neutral) and second delay duration (100, 800, 2,400 or 24,000 ms). Confirming the RCE, performance was significantly better on valid retro-cue trials, F(1, 19) = 27.98, p < .001 (valid M = .84, SD = .02; neutral: M = .73, SD = .08; ηp2 = .59). There was a significant main effect of second delay duration, F(3, 57) = 5.43, p = .02; ηp2 = .22, driven by significantly reduced accuracy at the 24,000-ms second delay duration compared with the 800-ms second delay duration (p = .007). No other pairwise comparisons were significant (ps > .09). Importantly, there was a significant Retro-Cue Condition × Second Delay Duration interaction, F(3, 57) = 6.11, p = .001; ηp2 = .24. This interaction was driven by two components. First, accuracy on neutral retro-cue trials fell with longer second delay durations, as confirmed by a second ANOVA excluding valid retro-cue trials, F(3, 57) = 9.69, p = .001, with significantly lower accuracy after 24,000-ms delays than at the other durations (ps < .007). In contrast, performance on the valid retro-cue trials remained constant across delays (F < 1, p = ns; see Figure 2a.

Experiment 1b

Accuracy data were subjected to a 3 × 4 repeated-measures ANOVA including retro-cue (valid, invalid, neutral) and second delay duration (100, 800, 2,400, 24,000 ms). There was a significant main effect of retro-cue (neutral: M = .69, SD = .05; valid: M = .75, SD = .05; invalid: M = .65, SD = .07), F(2, 38) = 8.51, p = .001; ηp2 = .30, such that the valid retro-cue elicited more accurate VWM performance than neutral (p = .05) or invalid retro-cues (p = .004). Neutral and invalid retro-cues did not differ significantly (p = .41). As in Experiment 1a, there was a significant main effect of second delay duration, F(3, 57) = 11.05, p < .001; ηp2 = .36, indicating that extending the delay hurt accuracy. Accuracy was significantly lower in the 24,000-ms condition than in all other durations (ps ≤ .02); no other second delay durations differed significantly from each other (ps > .55). Unlike Experiment 1a, there was no significant Retro-Cue Condition × Second Delay Duration interaction, F(6, 114) = 1.02, p = .416; ηp2 = .05 (see Figure 2b).

Experiment 1a and 1b comparison

To directly compare the magnitude of the RCE between experiments, we calculated the RCE for each participant and conducted a repeated-measures ANOVA with the within-subjects factors of second delay duration (100, 800, 2,400, and 24,000 ms) and the between-subjects factor of experiment (1a and 1b). As expected, there was a main effect of second delay duration, F(3, 114) = 3.29, p = .023; ηp2 = .74. The main effect of experiment was not significant, F(1, 38) = 1.16, p = .288; ηp2 = .18, and neither was the Second Delay Duration × Experiment interaction, F(6, 114) = 2.11, p = .102; ηp2 = .53.

Experiments 1a and 1b tested VWM using recognition to determine whether the inclusion of invalid retro-cues interfered with RCE magnitude. The data showed that the presence of the invalid cues did not weaken the RCE. These results provide initial evidence that the presence of invalid trials does not interfere with RCE magnitude.

The second pair of experiments employed recall to probe VWM. We compared RCE magnitude across experiments with and without invalid retro-cue trials. We made two changes to the previous method. First, a preliminary experiment confirmed that replacing color patches with oriented bar stimuli would lead to the same conclusions as derived from Experiments 1a and 1b: including invalid retro-cue trials did not minimize the magnitude of the RCE. In Experiments 2a and 2b, we used oriented bar stimuli and extended the encoding time (e.g., Zhang & Luck, 2009). Second, we changed the second delay durations. In Experiments 1a and 1b, there was no significant difference between the 100-, 800-, and 2,400-ms second delay durations, so we selected 800 ms to maintain a consistent second delay duration across experiments. A recent study has shown that at second delay durations of 3,000 ms, invalidly cued items decay twice as much as validly cued items (Pertzov et al., 2013). We included the 3,000-ms second delay duration to make contact with their findings. Finally, we included a longer second delay durations (10,000 ms) to examine the consequences of delay and to extend the previous observations of Pertzov et al. We excluded the 24,000-ms second delay duration used in Experiments 1a and 2b because participants struggled with this long delay when asked for recognition responses and we were concerned that we would not obtain valid recall measures.

Participants

In Experiments 2a and 2b, two groups of 20 undergraduate volunteers participated (Experiment 2a: M = 23.6 years; range: 18–40 years; 13 female; Experiment 2b: M = 20.8 years; range: 18–33 years; 10 female) in exchange for extra course credits or $10.

Equipment and stimuli

Participants sat 57 cm from a 15-in MacBook monitor (MacBook Pro). The experiment was programmed in Matlab, using the Psychophysics Toolbox (Brainard, 1997; Kleiner, Brainard, & Pelli, 2007; Pelli, 1997). White bars (7.5° × 1.5°) were displayed equidistantly (6.5°) from fixation in a square pattern; cues were neutral (“X”; 4° × 4°) or valid or invalid arrow retro-cues (4.5° × 4°). In a VWM array, the four orientations differed by at least 10°.

Experimental design and trial sequence

The trial sequence was similar to that described above, except for expanding stimulus duration to 1,000 ms and the second delay durations (800, 3,000, 10,000 ms; see Figure 1b). At retrieval, participants rotated the single probe item using left and right arrow keys, with each key press incrementing rotation 1°. Pressing the space bar confirmed the response. Participants completed 210 (Experiment 2a) or 195 trials (Experiment 2b). In Experiment 2a, neutral and valid trials were equally likely; in Experiment 2b valid (40%) and neutral (40%) trials were twice as likely as invalid (20%) trials. The distance between the presentation orientation and the reported orientation was calculated as degree error.

Experiment 2a

Degree error was subjected to a 2 × 3 (Retro-Cue [valid, neutral] × Second Delay Duration [800, 3,000, 10,000 ms] repeated-measures ANOVA. The expected RCE was significant, F(1, 19) = 9.57, p = .006; ηp2 = .33. Error was significantly larger for neutral (M = 20.52, SD = 1.66) than valid (M = 17.28, SD = 1.56) trials. The main effect of second delay duration was also significant, F(2, 38) = 8.61, p = .001; ηp2 = 31. Greater error was observed in the 10,000-ms (M = 20.70, SD = 1.50) than in the 800-ms (M = 17.51, SD = 1.58; p = .005) or 3,000-ms (M = 18.48, SD = 1.67; p = .008) conditions. There was no significant difference between 800-ms and 3,000-ms (p = .763) conditions. The Retro-Cue Type × Second Delay Duration interaction was not significant, F(2, 38) < 1, p = ns; ηp2 = .02 (see Figure 3a).

Experiment 2b

The analysis included the invalid retro-cue condition. There was a significant main effect of cue, F(1.24, 23.67) = 5.99, p = .017; ηp2 = .24, but no significant RCE because there was no difference between valid and neutral conditions (p = .305). The effect was driven by borderline differences between the invalid (M = 28.70, SD = 2.52) and valid (M = 21.20, SD = 2.24; p = .056) and invalid and neutral (M = 23.70, SD = 2.03; p = .065) conditions. The main effect of second delay duration was significant, F(2, 38) = 19.50, p < .001; ηp2 = .50, with increased error evident between the 3,000-ms (M = 23.83, SD = 2.25) and 10,000-ms (M = 28.02, SD = 1.78; p = .005) delays and between the 800-ms (M = 21.75, SD = 1.86) and 10,000-ms delays (p < .001). The Retro-Cue × Second Delay Duration interaction was not reliable, F(4, 76) = 1.18, p = .323; ηp2 = .05 (see Figure 3b).

Experiment 2a alone identified a significant RCE in a recall VWM paradigm. Thus, the presence of invalid retro-cues significantly reduced the RCE. Not surprisingly, recall fidelity dropped over longer delays and there was no evidence that items were preserved from decay. The implications of these findings are discussed below.

In this study, we tested the possibility that including invalid retro-cue trials modulated the magnitude of the RCE and found that it did, under certain conditions. Notably, although we found a robust RCE using recognition probes with or without the invalid trials, this pattern was not evident when probing VWM with recall. The recall data from Experiments 2a and 2b showed the RCE did not persist when invalid trials were included. Thus, we failed to replicate the findings that the retro-cue serves to protect the cued item from VWM decay (Pertzov et al., 2013; Hollingworth & Hwang, 2013). These data challenge our goal of bridging findings from recognition and recall paradigms because the coarser recognition measure demonstrated a statistically reliable RCE, whereas the finer grained recall measure revealed an RCE only when the retro-cues were always valid. We were able to evaluate the decay of cued and uncued items over an extended time frame using long second delay durations.

The question remains what mechanism subserves the retro-cue effect in VWM. Two primary accounts of the RCE have received the most attention: protection and prioritization (Matsukura et al., 2007). As noted, protection states that the retro-cue focuses internal attention on the cued item preventing interference from competing representations in VWM. Prioritization proposes that the retro-cued item is the first VWM representation compared with the probe item at retrieval (Matsukura et al., 2007). A similar proposal suggests that the retro-cue optimizes the search through items in VWM (see Lepsien & Nobre, 2006, for a review). The strongest support for a protection account was apparent in Experiments 1a and 1b in which accuracy in neutral trials showed steady, significant decline and valid trials showed no loss of VWM accuracy over time, even with the inclusion of the invalid retro-cues. This suggested that the valid retro-cues prevented the cued item from experiencing decay. However, the remaining data sets did not support this interpretation, because decay functions across all trial types were similar, contrary to previous results (Pertzov et al., 2013). The prioritization account predicts that performance on all trials would drop steadily with increased delays, a pattern we observed in Experiments 2a and 2b. One possibility is to interpret these data using a hybridized unifying account (Pertzov et al., 2013). Accordingly, when the retro-cues were valid or neutral, protection could dominate, because there was no cost to focusing all resources on the cued item and protecting it from decay. However, in the presence of invalid retro-cues on a demanding probe like recall, protection would be detrimental and a prioritization approach was adopted. Going forward, it will be important for further research to include invalid retro-cue trials because only then will a full accounting of the RCE be possible.

We discuss several concerns and limitations for this study below. These concerns are limited to our stimuli and manipulation of the second delay durations. We used color disks for Experiments 1a and 1b and oriented bars for Experiments 2a and 2b. This difference raises concern that different stimuli might result in differences that did not result from our manipulations. We do not think that this is a serious concern because a pilot study using oriented bar stimuli revealed a significant RCE with and without invalid retro-cues across second delay durations up to 800 ms (see Footnote 1). Also, others have found significant RCE using a similar recall paradigm and testing bar, color, and shape stimuli (Pertzov et al., 2013). Thus, we believe that the present experiments serve as a bridge between recall and recognition paradigms even though the stimuli differ.

Another concern is that we used different second delay durations between experiments. We selected a single second delay duration from Experiments 1a and 1b (800 ms) to include in Experiments 2a and 2b. We felt comfortable doing so because our previous work has shown no significant differences in the RCE for second delay durations from 300–700 ms (Tanoue & Berryhill, 2012) and, in Experiments 1a and 1b, there were no significant differences between the shorter second delay durations (100, 800, and 2,400 ms). In contrast, we elected to shorten the longest second delay duration from Experiments 1a and 1b (24,000 ms) because trials were very long, participants found them difficult, and the data confirmed that their recognition performance was significantly less accurate. Experiments 2a and 2b required a more challenging recall response. As such, we elected instead to use a 10,000-ms second delay duration as our longest delay. The general robustness of the RCE provides some flexibility in timing, but we certainly acknowledge that a full parallel structure between Experiments 1 and 2 would be preferable.

On a related note, the retro-cue literature typically uses short second delay durations. By probing the RCE across longer second delay durations, we exposed ourselves to concerns about the contributions of long-term memory. Neuropsychological data has indicated that there is a maximum of about 40–45 s for short-term memory to operate (Damasio, Graff-Radford, Eslinger, Damasio, & Kassell, 1985; Tranel & Eslinger, 2000; Sidman, Stoddard, & Mohr, 1968). However, it is difficult to ever disentangle long-term memory from interacting with shorter forms of memory (e.g., Baddeley, 2012) and at the longer second delay durations we used we may have measured from “long-term working memory.” This term refers to the involvement of long-term memory activated by cues in working memory (Ericsson & Kintsch, 1995). For our purposes, we were interested in observing decay functions over the temporal window of working memory. We would argue that working memory was certainly engaged, but very possibly, at the longer second delay durations, long-term memory was playing some role as well.

Despite the aforementioned concerns and criticisms, we conclude that the present data extend recent findings testing VWM using recall retro-cue paradigms. They also provide additional support for a hybridized theoretical account of the RCE that is in greater accord with the prioritization account.

1  Two groups of 14 participants completed a pair of experiments using oriented line stimuli (1,000-ms presentations; six orientations: 0°, 30°, 60°, 90°, 120°, and 150°) at several second delay durations (200, 400, 600, 800 ms). In each experiment, there were 200 trials in a 6-AFC design such that all possible line orientations were available. Statistical analyses revealed a significant retro-cue effect (RCE) including invalid trials, F(1.19, 15.48) = 16.21, p = .001 (neutral: M = .56, SD = .03; valid: M = .66, SD = .02; invalid M = .45, SD = .05) or excluding them, F(1, 13) = 31.62, p = < .001 (neutral: M = .60, SD = .04; valid M = .75, SD = .03). However, over these shorter second delay durations, there was no main effect of second delay duration (ps > .12) and no Cue × Delay interaction (ps > .44), replicating our previous findings (Tanoue & Berryhill, 2012). We also compared the magnitude of the RCE between experiments. We conducted a repeated-measures analysis of variance with the within-subjects factors of second delay duration (200, 400, 600, and 800 ms) and the between-subjects factor of experiment (with invalid trials, without invalid trials). The main effect of second delay duration was not significant, F(3, 78) = 1.26, p = .293; ηp2 = .46, and neither was the main effect of experiment, F(1, 26) = 2.04, p = .164, ηp2 = .73. The interaction effect was also not significant, F(3, 78) = 3.05, p = .271, ηp2 = .69. In short, using bar stimuli instead of color patches lead to the same conclusion: including invalid retro-cue trials did not modulate RCE magnitude.

Astle, D. E., Nobre, A. C., & Scerif, G. (2012). Attentional control constrains visual short-term memory: Insights from developmental and individual differences. The Quarterly Journal of Experimental Psychology, 65, 277–294. doi:10.1080/17470218.2010.492622

Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual Review of Psychology, 63, 1–29. doi:10.1146/annurev-psych-120710-100422

Bays, P. M., Gorgoraptis, N., Wee, N., Marshall, L., & Husain, M. (2011). Temporal dynamics of encoding, storage, and reallocation of visual working memory. Journal of Vision, 11(10), 6. doi:10.1167/11.10.6

Berryhill, M. E., Richmond, L. L., Shay, C. S., & Olson, I. R. (2012). Shifting attention among working memory representations: Testing cue type, awareness, and strategic control. The Quarterly Journal of Experimental Psychology, 65, 426–438. doi:10.1080/17470218.2011.604786

Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10, 433–436. doi:10.1163/156856897X00357

Carrasco, M. (2011). Visual attention: The past 25 years. Vision Research, 51, 1484–1525. doi:10.1016/j.visres.2011.04.012

Damasio, A. R., Graff-Radford, N. R., Eslinger, P. J., Damasio, H., & Kassell, N. (1985). Amnesia following basal forebrain lesions. Archives of Neurology, 42, 263–271. doi:10.1001/archneur.1985.04060030081013

Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102, 211–245. doi:10.1037/0033-295X.102.2.211

Griffin, I. C., & Nobre, A. C. (2003). Orienting attention to locations in internal representations. Journal of Cognitive Neuroscience, 15, 1176–1194. doi:10.1162/089892903322598139

Hollingworth, A., & Hwang, S. (2013). The relationship between visual working memory and attention: Retention of precise colour information in the absence of effects on perceptual selection. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 368, 20130061. doi:10.1098/rstb.2013.0061

Jiang, Y., Olson, I. R., & Chun, M. M. (2000). Organization of visual short-term memory. Journal of Experimental Psychology Learning, Memory, and Cognition, 26, 683–702. doi:10.1037/0278-7393.26.3.683

Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention in the human cortex. Annual Review of Neuroscience, 23, 315–341. doi:10.1146/annurev.neuro.23.1.315

Kleiner, M., Brainard, D., & Pelli, D. (2007). What’s new in Psychtoolbox-3? [abstract]. Perception, 36(Supp.), 14.

Landman, R., Spekreijse, H., & Lamme, V. A. (2003). Large capacity storage of integrated objects before change blindness. Vision Research, 43, 149–164. doi:10.1016/S0042-6989(02)00402-9

Lepsien, J., & Nobre, A. C. (2006). Cognitive control of attention in the human brain: Insights from orienting attention to mental representations. Brain Research, 1105, 20–31. doi:10.1016/j.brainres.2006.03.033

Matsukura, M., Luck, S. J., & Vecera, S. P. (2007). Attention effects during visual short-term memory maintenance: Protection or prioritization?Perception & Psychophysics, 69, 1422–1434. doi:10.3758/BF03192957

Nobre, A. C., Coull, J. T., Maquet, P., Frith, C. D., Vandenberghe, R., & Mesulam, M. M. (2004). Orienting attention to locations in perceptual versus mental representations. Journal of Cognitive Neuroscience, 16, 363–373. doi:10.1162/089892904322926700

Nobre, A. C., Griffin, I. C., & Rao, A. (2008). Spatial attention can bias search in visual short-term memory. Frontiers in Human Neuroscience, 1. 4. doi:10.3389/neuro

Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437–442. doi:10.1163/156856897X00366

Pertzov, Y., Bays, P. M., Joseph, S., & Husain, M. (2013). Rapid forgetting prevented by retrospective attention cues. Journal of Experimental Psychology: Human Perception and Performance, 39, 1224–1231. doi:10.1037/a0030947

Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual Review of Neuroscience, 13, 25–42. doi:10.1146/annurev.ne.13.030190.000325

Sidman, M., Stoddard, L. T., & Mohr, J. P. (1968). Some additional quantitative observations of immediate memory in a patient with bilateral hippocampal lesions. Neuropsychologia, 6, 245–254. doi:10.1016/0028-3932(68)90023-7

Tanoue, R. T., & Berryhill, M. E. (2012). The mental wormhole: Internal attention shifts without regard for distance. Attention, Perception, & Psychophysics, 74, 1199–1215. doi:10.3758/s13414-012-0305-0

Tranel, D., & Eslinger, P. J. (2000). Effects of early onset brain injury on the development of cognition and behavior: Introduction to the special issue. Developmental Neuropsychology, 18, 273–280. doi:10.1207/S1532694201Tranel

Williams, M., Hong, S. W., Kang, M. S., Carlisle, N. B., & Woodman, G. F. (2013). The benefit of forgetting. Psychonomic Bulletin & Review, 20, 348–355. doi:10.3758/s13423-012-0354-3

Williams, M., & Woodman, G. F. (2012). Directed forgetting and directed remembering in visual working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38, 1206–1220. doi:10.1037/a0027389

Zhang, W., & Luck, S. J. (2009). Sudden death and gradual decay in visual working memory. Psychological Science, 20, 423–428. doi:10.1111/j.1467-9280.2009.02322.x

Submitted: April 3, 2014 Revised: June 16, 2014 Accepted: June 20, 2014