Consolidating working memory: Distinguishing the effects of consolidation, rehearsal and attentional refreshing in a working memory span task

Journal of Memory and Language 81 (2015) 34–50

Contents lists available at ScienceDirect

Journal of Memory and Language journal homepage: www.elsevier.com/locate/jml

Consolidating working memory: Distinguishing the effects of consolidation, rehearsal and attentional refreshing in a working memory span task Donna M. Bayliss a,⇑, Jade Bogdanovs a, Christopher Jarrold a,b a b

Neurocognitive Development Unit, School of Psychology, University of Western Australia, Australia School of Experimental Psychology, University of Bristol, United Kingdom

a r t i c l e

i n f o

Article history: Received 17 March 2014 revision received 20 December 2014

Keywords: Short-term consolidation Working memory Attentional refreshing Articulatory rehearsal

a b s t r a c t In a series of experiments, we demonstrated that manipulating the opportunity that individuals had to consolidate each memory item produced systematic differences in working memory span performance. In young adults, presenting an unfilled delay interval immediately following the presentation of each to-be-remembered item and before the onset of a distractor processing activity produced enhanced working memory performance relative to when the same delay interval was presented after the processing activity. In addition, the beneficial effect of providing an opportunity for consolidation was unaffected by manipulations of processing difficulty (Experiment 1), processing pace (Experiment 2), and articulatory suppression (Experiment 3). Finally, we demonstrated that RT functions consistent with a process of short-term consolidation are evident at longer item presentation times more commonly associated with working memory span tasks (Experiment 4). Together, these results suggest that the process of consolidation is separable from articulatory rehearsal and attentional refreshing. Moreover, these results are difficult to account for in terms of cognitive load, temporal distinctiveness, and/or distractor removal and suggest that current models of working memory may need to be modified to take into account the temporal parameters associated with the initial consolidation of memory items. Ó 2015 Elsevier Inc. All rights reserved.

Introduction Working memory is thought to be responsible for the active maintenance and management of information required to complete current task goals (Baddeley, 1986), and is commonly measured using working memory span tasks that require both the maintenance of a series of tobe-remembered items and the completion of a concurrent processing activity (Daneman & Carpenter, 1980). The factors influencing the maintenance of information in ⇑ Corresponding author at: School of Psychology, Mailbag M304, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia. Fax: +61 (0) 8 6488 1006. E-mail address: [email protected] (D.M. Bayliss). http://dx.doi.org/10.1016/j.jml.2014.12.004 0749-596X/Ó 2015 Elsevier Inc. All rights reserved.

working memory have been the subject of extensive research in the 40 years since the seminal model of Baddeley and Hitch (1974) was first proposed. A central tenet of the Baddeley and Hitch model was that verbal information was maintained in working memory through a process of rehearsal (Baddeley, Lewis, & Vallar, 1984), and extensive research has been devoted to understanding this mechanism (Awh et al., 1996; Baddeley, 1986; Baddeley, Thomson, & Buchanan, 1975; Tam, Jarrold, Baddeley, & Sabatos-DeVito, 2010; Tan & Ward, 2008). However, more recent conceptualisations of working memory have begun to suggest that other processes in addition to rehearsal may be important for the successful maintenance of information in memory (Camos, Lagner,

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

& Barrouillet, 2009; Cowan, 1999; Tam et al., 2010). In particular, a number of researchers have argued for a process of attentional refreshing that acts to maintain information stored in working memory by focusing domain-general attention on to-be-remembered items (Barrouillet, Bernardin, & Camos, 2004; Barrouillet, Portrat, & Camos, 2011; Johnson, 1992), with considerable evidence amassed to support this view (Barrouillet, Bernardin, Portrat, Vergauwe, & Camos, 2007; Camos et al., 2009; Raye, Johnson, Mitchell, Greene, & Johnson, 2007; Raye, Johnson, Mitchell, Reeder, & Greene, 2002). Additionally, some recent investigations have suggested that the process of consolidating information into working memory may also play a role in determining performance (Nieuwenstein & Wyble, 2014; Ricker & Cowan, 2014; Vergauwe, Camos, & Barrouillet, 2014). However, the nature of the consolidation process in the context of working memory remains relatively underspecified and, importantly, little is known about how it relates to the more established processes of rehearsal and attentional refreshing. The current study aimed to address these issues by examining the effect of providing an opportunity for consolidation in a working memory span task, and differentiating this effect from those associated with attentional refreshing and articulatory rehearsal. Short-term consolidation refers to the processing involved in transforming fragile, transient sensory input into more durable memory representations (Chun & Potter, 1995; Jolicoeur & Dell’Acqua, 1998) and has been distinguished from the more basic sensory and perceptual encoding involved in the detection and identification of a stimulus (Jolicoeur & Dell’Acqua, 1998; Nieuwenstein & Wyble, 2014; Ricker & Cowan, 2014). The transmission of information during sensory and perceptual encoding is thought to be fast and to occur in parallel, with representations formed during these stages subject to rapid forgetting unless they undergo a process of consolidation. In contrast, the process of consolidation, which is thought to occur after these basic encoding stages, is argued to be time-consuming and reliant on central attentional mechanisms (Jolicoeur & Dell’Acqua, 1998). Consolidated memory representations are thought to be available for later report in the absence of ongoing sensory input (Chun & Potter, 1995; Jolicoeur & Dell’Acqua, 1998), to be able to withstand interference from new information entering the system (Nieuwenstein & Wyble, 2014), and to be more resistant to forgetting (Ricker & Cowan, 2014). Ricker and Cowan (2014) have likened encoding to the sensory activation of features in long-term memory and consolidation to the entry of these features into Cowan’s (1988, 1995) focus of attention. A number of frameworks have proposed a similar distinction between initial sensory encoding and the subsequent formation or consolidation of integrated representations in short-term memory (e.g., Chun & Potter, 1995; Jolicoeur & Dell’Acqua, 1998; Massaro, 1975; Nieuwenstein & Wyble, 2014) and we will follow suit in using this distinction and terminology. However, we remain open to the possibility that what we and others have termed consolidation may in fact reflect the operation of other post-encoding processes and will return to this issue in the General Discussion.

35

Jolicoeur and Dell’Acqua (1998) were one of the first to provide a systematic investigation of the process of shortterm consolidation. They developed a paradigm for investigating the time course of consolidation by combining a visual memory task with an auditory forced-choice reaction time task. Participants were presented with a visual stimulus item (a letter or symbol) which was quickly masked, followed by the forced-choice reaction time task, which involved judging tones as being either high or low in pitch by making an appropriate keypress. Finally, participants were required to recall the initial visual memory stimulus. The time interval between the presentation of the initial memory stimulus and the onset of the tone was varied systematically. The hypothesis was that if the memory stimulus was still being consolidated when the tone was presented, reaction times to the tone would be slowed relative to a tone presented after consolidation had finished, when an individual’s central attentional mechanisms would again be available. Results consistently showed that reaction times to the tone were slower when the interval between the initial memory stimulus and the tone was short, and became increasingly faster as this interval increased. This was taken as evidence of a process of consolidation. Jolicoeur and Dell’Acqua (1998) went on to demonstrate that this process of consolidation was under conscious control, by including a condition in which participants were still presented with the memory items, but were instructed to ignore these and to just respond to the choice reaction time task, as recall was not required at the end of each trial. For this condition, reaction times to the tone were unaffected by temporal proximity to the initial stimulus, suggesting that the slowed reaction times in the former condition were specifically associated with the need to remember information and not simply the perceptual characteristics of the task. Jolicoeur and Dell’Acqua (1998) concluded that consolidation is an important process involved in transferring information into short-term memory, and that it requires central processing resources. Although considerable research has been devoted to establishing the basic parameters of the consolidation process (Jolicoeur, 1999; Jolicoeur & Dell’Acqua, 1999; Nieuwenstein & Wyble, 2014; Stevanovski & Jolicoeur, 2007, 2011), it is not currently known whether the process of consolidation is important for performance on a typical working memory span task. However, converging evidence from studies that have manipulated the time available for post-encoding processing in other paradigms suggests that it is likely to be a contributing factor. For example, a recent study by Barrouillet, Plancher, Guida, and Camos (2013) showed that increasing the time available for the encoding of memory items in a serial recall task, by presenting memory items at a slow (i.e., 5000 ms per item) rather than a fast (i.e., 500 ms per item) pace, led to better recall of the memory items. Moreover, recall of the memory items presented at a slow pace was more resistant to an increase in the number of distractors presented during retrieval, suggesting that the memory traces may have been protected to some extent from event-based interference, but were still affected by an increase in the attentional demand of the distractors. Barrouillet et al. (2013) attributed the difference between the fast and slow paced conditions to

36

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

the increased time allowed for consolidation of the memory items. However, the paradigm used by Barrouillet et al. (2013) did not prevent articulatory rehearsal during presentation of the memory items and so it is unclear whether the beneficial effects reported in their study really reflect greater consolidation of the memory items or just a greater opportunity for rehearsal. Similarly, Ricker and Cowan (2014) showed that equating the time available for post-encoding processing, but not basic encoding time, led to equivalent rates of forgetting in a visual probe recognition paradigm. In their paradigm, participants were presented with three unfamiliar visual characters either simultaneously or sequentially, followed by a mask and then a variable retention interval. When only the time available for sensory and perceptual encoding was equated, the rate of forgetting across the retention intervals was greater for the simultaneous condition than the sequential. However, when blank intervals were inserted into the simultaneous condition (by presenting the whole memory array multiple times) to also equate the free time available across the two conditions, the difference in forgetting rates was eliminated. Ricker and Cowan (2014) argued that this equated the amount of time available for the consolidation of each memory item into memory, and concluded that consolidation time was an important determinant of working memory performance. Finally, using a Brown–Peterson paradigm in which participants are presented with a series of items to remember followed by a distractor processing activity, Vergauwe et al. (2014) showed that response times for the first processing item increased linearly with memory load (cf. Jarrold, Tam, Baddeley, & Harvey, 2011). This increase in response times for the first processing item was evident for both verbal and spatial memoranda and spatial and numerical processing tasks, and was estimated to be around 250 ms per item. Vergauwe et al. (2014) attributed this increase to the time required for consolidation of the memory items, which they argued was attentionally demanding and led to a postponement of the processing activity. Response times on subsequent processing items (i.e., all processing items apart from the first within each processing episode) also showed a linear increase with memory load, however, the magnitude of this increase was estimated to be around 50 ms per item, which Vergauwe et al. (2014) attributed to the cost associated with attentional refreshing. There is, therefore, good reason to believe that the process of consolidation may be important for working memory performance, but crucially none of these studies have used a working memory span task in which the distracting processing activity is interspersed between the memory items. The rapid switching that occurs between the encoding of each memory item and completion of an interleaved processing activity in a working memory span task would suggest that a consolidation process may be particularly susceptible to interruption in these tasks (see Jarrold et al., 2011, for evidence that memory performance is affected by the location of the distractor processing activity). If the opportunity for consolidation does affect performance on commonly used working memory tasks, it will be important to take this into consideration when

designing future studies, as unintentionally impeding the consolidation process may in turn affect working memory span performance and relationships found between working memory and other higher-level cognitive abilities. Moreover, studying consolidation within the context of a working memory span task will have important theoretical implications for current models of working memory that do not incorporate such a process. Indeed, a key issue that is yet to be addressed is whether the process of consolidation differs from other processes such as attentional refreshing and articulatory rehearsal that are known to be important for working memory performance; it could certainly be argued that consolidation is simply an extension of the same maintenance processes that are known to operate once items are initially encoded in memory. There is some evidence to suggest that this is not the case. Stevanovski and Jolicoeur (2007) combined the Jolicoeur and Dell’Acqua (1998) paradigm described above with an articulatory suppression task and showed a similar pattern of results as in the earlier Jolicoeur and Dell’Acqua study. That is, RTs to a choice RT task decreased as the time between the presentation of the memory item and the choice RT task increased. This suggests that the process of consolidation is independent of articulatory rehearsal. In addition, Stevanovski and Jolicoeur (2007) showed that the slope of the consolidation function was unaffected by an increase in memory load from one to three items, even though RTs increased overall. This suggests that although there is a cost associated with the concurrent maintenance of a memory load, the efficiency of the consolidation process appears to be independent of this cost. However, in terms of the relationship between consolidation and attentional refreshing, no study has attempted to distinguish these processes from one another, though different timescales have been suggested to be associated with each (e.g., see Vergauwe et al., 2014). The current series of experiments was therefore designed to investigate whether short-term consolidation is important for working memory span performance and to examine the distinction between consolidation, articulatory rehearsal and attentional refreshing. The first of these aims was addressed by manipulating the opportunity that participants had to consolidate memory items in a working memory span task. The tasks used here were modified versions of the computer-paced complex span tasks used by Barrouillet et al. (2004, 2007) and Lépine, Bernardin, and Barrouillet (2005), as these allow for the timing of operations to be manipulated precisely. Each task involved a series of memory items, with an interleaved processing activity that was to be completed after the presentation of each memory item. In one condition, an opportunity for consolidation was provided by presenting a blank delay interval immediately following the presentation of each memory item and before the onset of the processing activity. The rationale was that in this condition, participants would be able to consolidate the memory representation to some extent before switching their attention to the processing activity. In contrast, in a second condition, the processing activity was presented immediately following the presentation of each memory item, thus minimising

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

any opportunity for consolidation of the memory representation prior to completing the processing activity. Importantly, this condition also included a delay interval to equate the overall duration of the tasks, but this was presented after the processing activity and before the next memory item. If the process of consolidation is important for working memory, then performance should be better in the condition that provides an opportunity for this to occur prior to the presentation of the processing activity. To address our second aim of differentiating between consolidation, rehearsal and attentional refreshing, we controlled the opportunity that participants had to engage in each of these processes and examined whether this influenced any effects attributed to the process of consolidation. In Experiment 1, we manipulated the opportunity participants had for attentional refreshing by varying the difficulty of the processing activity. Within the Time-Based Resource Sharing (TBRS) model of Barrouillet et al. (2004, 2007, 2011), the ‘cognitive load’ of a given working memory task is defined as the proportion of the total task time during which the processing activity occupies attention and as a consequence, prevents attentional refreshing of the memory items. In support of this claim, Barrouillet et al. (2004, 2007) have shown that more difficult processing tasks take longer to complete (thus purportedly decreasing the time available for attentional refreshing and increasing the cognitive load of the task) and result in poorer working memory performance. In Experiment 2, we used a different manipulation of processing difficulty, and in Experiment 3, we blocked articulatory rehearsal by requiring participants to complete the tasks under articulatory suppression. In all experiments, the manipulations of processing difficulty were crossed with the manipulation varying the opportunity for consolidation described above to create four working memory tasks. This enabled us to examine whether any effects of consolidation were evident independently of any effects of attentional refreshing and/or articulatory rehearsal. A final experiment was also conducted to provide more direct evidence of the involvement of a process of short-term consolidation in working memory span performance.

Experiment 1 Experiment 1 was designed to examine whether shortterm consolidation is important for performance on a typical working memory span task by systematically varying the opportunity for consolidation. Processing difficulty, and therefore, the cognitive load of the task, was also manipulated, allowing us to investigate whether consolidation and attentional refreshing are separate processes. By manipulating processing difficulty, we were also able to address the question of whether memory traces that have had time to consolidate are more resistant to interference, as suggested by the results of Barrouillet et al. (2013). The continuous operation span task used by Lépine et al. (2005) was modified in this experiment to include a manipulation of the position of the processing activity, which was presented either immediately after the presentation of the memory item (immediate condition), or after

37

a blank delay interval (delayed condition). This manipulation was combined with the manipulation of processing difficulty used by Camos et al. (2009, Experiment 1) to vary the opportunity for attentional refreshing, and thus, the cognitive load of the memory task. This processing difficulty manipulation involved either reading or performing simple mathematical calculations. It was hypothesised that allowing time for consolidation would be beneficial for working memory performance, resulting in better performance in the delayed processing position condition overall. In addition, if consolidation provides a protective effect for memory items against interference from the concurrent processing task, then any effect of processing difficulty may be moderated by the position of the processing task. That is, memory items that have had more opportunity for consolidation may be less affected by an increase in the difficulty of the processing task. Further to this, our experimental design allowed for consolidation and attentional refreshing to be disentangled. If consolidation and attentional refreshing are separable processes, it was expected that there would be main effects of both processing difficulty and processing position, reflecting the influence of attentional refreshing and consolidation, respectively. However, if short-term consolidation is not important for working memory and performance can be best explained in terms of attentional refreshing, only a main effect of processing difficulty would be expected, as the time available for attentional refreshing overall remained constant across both levels of processing position. Method Participants Sixty-two undergraduate university students (18 males) aged between 17 and 52 years (mean age = 20.53 years) participated in the study. Participants received either course credit or a small payment for participation. Tasks and procedure Working memory tasks. Four working memory tasks were created by crossing two levels of processing difficulty (easy, difficult) with two levels of processing position (immediate, delayed). Each working memory task involved the recall of lists of letters in the same order that they were presented. The pool of letter stimuli comprised 18 consonants (all consonants excluding Y, W, and Q). Lists were between three and seven letters in length, with four trials completed at each length, for a total of 20 trials for each working memory task. An additional three practice trials were presented at the beginning of each working memory task, all with a length of two items. Letter stimuli were randomly selected without replacement for each trial. Experimental trials were presented in a randomised order for each participant. The presentation of the letters was interspersed with the processing task, which involved either reading digits aloud (easy condition) or performing simple calculations aloud (difficult condition). All working memory tasks were computerised and presented using the Revolution Dreamcard programme.

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

Each trial of the working memory tasks began with a fixation cross presented in the centre of the computer screen for 750 ms, followed by a letter presented in Arial font approximately 30 mm in height, which the participants were required to read aloud. This letter was presented for 500 ms, followed by a blank screen for 200 ms and a variable delay interval before the processing task. In the delayed condition, a blank screen was presented for a further 2400 ms before presentation of the first processing item, while in the immediate condition, the first processing item was presented directly following the 200 ms inter-stimulus interval, with no delay. Participants then completed the processing task. A single digit between 1 and 9 inclusive was displayed in the centre of the screen, followed by the consecutive presentation of two simple operations randomly selected for each trial (i.e., +1, 1, +2, or 2). These digits were the same size as the letters, but presented in colour, alternating between red and green for each successive operation. Each digit or operation was presented for 1000 ms, with an inter-stimulus interval of 200 ms, during which the screen was blank. Participants were required to read the initial number aloud, and then perform the calculations presented, for example, ‘‘4, plus 2 is 6, minus 1 is 5’’. In the easy condition, the answer to the arithmetic problem was presented next to the operation, so participants only needed to read what was on the screen (i.e., 4, +2, 6, 1, 5). In the difficult condition, the answer was not presented, and participants needed to perform the calculations themselves. Responses to the arithmetic problems in the difficult conditions were recorded by the experimenter. Following the processing task was a second variable delay (2400 ms in the immediate condition, and 0 ms in the delayed condition), before the next letter in the trial was presented. This second delay ensured that the overall time for each trial remained constant across conditions. The storage and processing sequence was repeated between three and seven times, depending on trial length. Immediately following the trial, participants were presented with a screen containing all possible letters, and were asked to verbally recall the letters in the same order that they were presented. The experimenter clicked on the corresponding letters as they were recalled, with letters recalled accurately and in the correct serial position scored as correct. The proportion of letters recalled correctly was calculated for each task. Procedure. Participants completed all tasks over two sessions with two working memory tasks completed in each session. Tasks were counterbalanced across participants, with the restriction that a participant complete tasks of the same difficulty level within the same session. Participants were instructed that both parts of the working memory task (i.e., remembering the letters and reading/ performing the calculations) were equally important.

they achieved perfect performance on the processing task (for a maximum score of 20). Two participants (one male, one female) were excluded from the analysis for poor performance on the processing task, more than three standard deviations below the mean for the sample. The remaining 60 participants had a mean age of 20.52 years, and ranged from 17 to 52 years old. Data analysis Means and standard errors for the four working memory tasks are presented in Fig. 1. A 2  2 repeated measures ANOVA with the factors of processing position (immediate, delayed) and processing difficulty (easy, difficult) revealed significant main effects of processing position, F(1, 59) = 4.24, p = .04, gp2 = .067, with better performance in the delayed condition (M = .51 and .54 for the immediate and delayed conditions respectively), and processing difficulty, F(1, 59) = 62.04, p < .01, gp2 = .513, with poorer performance in the difficult processing condition (M = .59 and .45 for the easy and difficult conditions respectively). No significant interaction was found, F(1, 59) = 0.97, p = .33, gp2 = .016. Discussion Results from this first experiment showed that both the processing difficulty and processing position manipulations affected performance on the working memory task. Higher processing demands in the difficult processing condition resulted in lower memory performance, and providing a delay directly after the presentation of the to-beremembered item (i.e., before the subsequent processing activity) resulted in better memory performance. This second result is consistent with the concept of short-term consolidation described by Jolicoeur and Dell’Acqua (1998), as providing a delay directly after the presentation of the storage items, and thus allowing time for these items to be consolidated, did result in better memory performance (see also, Nieuwenstein & Wyble, 2014). The finding that the manipulations of processing position and processing difficulty – though both significant – did not interact,

Easy

0.8

Difficult

Proportion Correct

38

0.6

0.4

0.2

0.0

Results

Immediate

Delayed

Processing Position Data screening Participants were screened based on their performance on the processing task for the difficult conditions, that is, participants were awarded one point for each trial where

Fig. 1. Mean proportion of items correctly recalled for each working memory task formed by crossing processing position (immediate, delayed) with processing difficulty (easy, difficult). Error bars represent standard errors.

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

suggests that the processes of consolidation and attentional refreshing are separate. According to Barrouillet et al. (2011), the cognitive load of a task depends on the ratio of the time during which the processing activity captures attention to the total processing time available, with more executively demanding processing tasks requiring attention for longer. In the current experiment, this cognitive load ratio remained constant regardless of the position of the processing activity (i.e., immediate or delayed), and was only affected by the difficulty of the processing task. Nevertheless, a main effect of processing position was found, with a delay immediately following the storage items resulting in better overall working memory performance relative to if the same delay was presented after the presentation of the processing items. If the opportunity for attentional refreshing was the only factor affecting working memory performance, we would not expect to find this main effect of processing position, as the opportunity for refreshing was the same in both processing position conditions. Thus, the results of the current study strongly suggest that the process of consolidation is separate from attentional refreshing and constrains performance on working memory span tasks. In terms of the process of consolidation itself, memory traces that have been consolidated are assumed to be in some way ‘protected’ from interference from other information subsequently entering the cognitive system (e.g., see Nieuwenstein & Wyble, 2014). While this could potentially explain the better memory performance in the delayed processing position conditions overall, if the memory items were indeed more resistant to interference due to being consolidated, then one would expect to see a reduced effect of processing difficulty in the delayed conditions. This was not evident. Items in the delayed conditions suffered the same degree of memory decrement from the increase in processing difficulty as items in the immediate conditions that had not had a chance to be consolidated. One potential explanation for this is that the manipulation of processing difficulty that we used may not have actually increased the amount of interference experienced because participants were required to name the same items in each condition and so, the number of interfering events was the same. Thus, it could be argued that the real difference between the two conditions was that in the difficult condition, participants were required to generate their own answers, and so, the time available for refreshing would have been reduced in this condition relative to the easy condition. This is consistent with suggestion of Barrouillet et al. (2013), that memory traces that have had a longer time to consolidate are more resistant to interference from distractors, but are still susceptible to increases in attentional demand. Thus, while providing an opportunity for consolidation in a working memory span task did lead to better memory performance, consistent with a consolidation account, the consolidated memory traces were still susceptible to increases in attentional demand. However, at this point, it is worth noting that the predicted reduction in the effect of processing difficulty following an opportunity for consolidation may have been masked due to certain aspects of the methodology used

39

in our initial experiment. For example, the processing component of the working memory task used was very fast paced, faster even than the ‘fast paced’ condition in the very similar processing task used by Camos et al. (2009). That being the case, it is possible that the memory traces did enjoy some degree of protection from subsequent interference as a result of being consolidated, but that any evidence of this protection was obliterated by the intensely demanding processing activity. A second aspect to consider was that the delay interval for the delayed processing position condition was relatively long at 2400 ms, compared to the maximum 1600 ms interval used by Jolicoeur and Dell’Acqua (1998). A 2400 ms interval was used to ensure that all participants completing the delayed processing position conditions had time to consolidate the information presented. However, it is possible that this interval was also long enough to encourage rehearsal of the memory items, either overtly or covertly. There was anecdotal evidence to suggest that this was the case, at least for some participants. If participants were able to use verbal rehearsal during the delay interval, this adds a possible confound to these conditions in that rehearsal of the memory items straight after their presentation (i.e., the delayed processing position condition) might have been more beneficial for memory than rehearsal of the memory items following completion of the processing activity (i.e., the immediate processing position condition), when the memory traces may have either decayed or been overwritten to some extent. Thus, while the opportunity for rehearsal (and forgetting) was the same in both processing position conditions, in reality, early rehearsal may have been more useful than late rehearsal. These concerns were addressed in the following two experiments.

Experiment 2 In Experiment 2, four working memory tasks were created by crossing two levels of processing position (immediate, delayed) with two levels of a pace manipulation (fast, slow) in which a simple digit naming task was presented at either a fast or slow pace (cf. Barrouillet et al., 2004). A pace manipulation was used in this experiment rather than a difficulty manipulation, because it enabled the use of simpler processing stimuli and thus, a more straightforward manipulation of the attentional refreshing involved in the task. The use of the same processing activity (i.e., reading digits) in both the fast and slow paced conditions also allowed for a comparison of performance on the processing component of the task, across the manipulations of both pace and processing position. According to the TBRS model, there is less opportunity for attentional refreshing when participants are required to complete processing items at a fast pace. Thus, we expected a main effect of pace such that memory performance would be poorer in the fast paced condition. In addition, in this experiment, the length of the delay interval presented either directly after the memory item or after the processing activity was reduced to 1000 ms. This was done to reduce the opportunity for participants to engage in

40

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

articulatory rehearsal, while still providing an opportunity for consolidation to occur. In line with Experiment 1, we expected that providing participants with an opportunity to consolidate their memory directly after the presentation of the stimulus would lead to better memory performance. Furthermore, if attentional refreshing and consolidation are separable processes, as was suggested in Experiment 1, then we would expect both the pace (slow or fast) and processing position (immediate or delayed) manipulations to affect recall performance on the working memory task independently. Having said this, if the process of consolidation does in some way insulate the memory representation from interference from subsequent information entering the cognitive system (e.g., Nieuwenstein & Wyble, 2014), then the effect of pace should be reduced in the delayed processing position conditions compared to the immediate processing position conditions. Although the manipulation of pace is not a difficulty manipulation per se, reading the presented digits at a fast pace is more difficult (Barrouillet et al., 2004), and recent modelling investigations of this effect have suggested that presenting processing items at a fast pace might act to reduce the time available for other processes that offset interference, resulting in greater interference in this condition (Oberauer, Lewandowsky, Farrell, Jarrold, & Greaves, 2012). Although this predicted interaction was not evident in Experiment 1, it was thought that employing a simpler processing task might enable some evidence of this protective effect to emerge. Having said this, if the lack of an interactive effect is replicated in this experiment, this would advance our understanding of the nature of the consolidation process thought to be operating within the short time-frame associated with working memory. Participants’ accuracy in the processing task was also analysed to assess for any effects of processing position on processing performance.

rate. Digits used were between one and nine inclusive, and the letters used were the same as in Experiment 1. Items were selected randomly without replacement for each trial (for the letters) or memory item (for the digits). All working memory tasks were computerised and presented using the Revolution Dreamcard programme. Each trial began with the presentation of a fixation cross in the centre of the computer screen for 750 ms, followed by a letter presented in black Arial font approximately 30 mm in height. This letter was presented for 500 ms, and followed by a blank screen for a delay interval of 100 ms for the immediate conditions, and 1000 ms in the delayed conditions. After the delay interval the processing task began, which involved the presentation of six individual digits, in red Arial font, also approximately 30 mm in height and presented in the centre of the computer screen. Each digit was presented for 500 ms, with an inter-stimulus-interval of 100 ms for the fast conditions, and 500 ms for the slow conditions. Directly following the presentation of the final digit of each processing set, a blank screen was presented for 1000 ms in the immediate conditions and 100 ms in the delayed conditions to keep the overall length of the immediate and delayed versions constant. Following each trial, participants were presented with a screen containing all possible letters, and were required to recall the letters in the same order that they were presented, by clicking on them using the computer mouse. Responses were scored by calculating the proportion of letters recalled accurately in the correct serial position.

Method

Procedure. Participants completed all tasks over two sessions with two working memory tasks completed in each session. Task presentation was counterbalanced across participants. Participants were instructed that both parts of the working memory task (i.e., remembering the letters and reading the digits correctly) were equally important. Following the working memory tasks in the second session, participants completed the Raven’s Standard Progressive Matrices – Plus Version.

Participants Sixty-nine undergraduate Psychology students (24 males) aged between 17 and 52 years (mean age = 21.37 years) participated in the study. Participants received credit points or a small payment in return for taking part in the experiment. Tasks and procedure Working memory tasks. Working memory tasks were created by crossing two levels of processing pace (fast, slow) with two levels of processing position (immediate, delayed). As in Experiment 1, all working memory tasks involved recalling lists of letters in the same order that they were presented. In this case, lists were between two and six letters in length, with 20 trials in total (four at each list length). In addition to these, three practice trials were given at the beginning of each working memory task, with a list length of two memory items. As for Experiment 1, the order of presentation of the experimental trials was randomised for each participant. The presentation of the letters was interspersed with a processing task, which involved reading single digits as they appeared at a fixed

Fluid reasoning tasks. The Ravens Standard Progressive Matrices – Plus Version was administered using the standardised instructions, with participants given a 20 min time limit to complete as many items as possible.1

Results Data screening Participants were screened on their performance in the processing task. Performance on the processing task was calculated by determining the number of trials (out of 20) in which the participant correctly named all digits as they were presented on the computer screen. This processing score was calculated for each working memory task separately, and any participants with a score more than three standard deviations below the mean were excluded 1 The Raven’s Matrices were included for the purpose of comparing participants across Experiments 2 and 3.

41

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

from all subsequent analyses, as they were deemed to have not completed the task correctly. Two female participants were excluded on the basis of their processing performance. The remaining 67 participants had a mean age of 21.37 years, with a range of 17–52 years. Data analysis Descriptive statistics for all working memory variables are presented in Fig. 2. Results for the Ravens Progressive Matrices showed a mean score of 39.31 with standard deviation 4.52. A 2 (processing position: immediate, delayed)  2 (pace: fast, slow) repeated-measures ANOVA was conducted on recall accuracy. This revealed significant main effects of processing position, F(1, 66) = 14.75, p < .01, gp2 = .18, with better performance in the delayed condition (M = .55 and .61 for the immediate and delayed conditions respectively), and of pace, F(1, 66) = 58.06, p < .01, gp2 = .47, with poorer performance in the fast pace condition (M = .63 and .54 for the slow and fast paced conditions respectively). No significant interaction was found, F(1, 66) = 0.53, p = .47, gp2 = .01. Further to this, an analysis of accuracy on the processing task was conducted to investigate any differences between the conditions in performance on the processing task. A 2 (processing position: immediate, delayed)  2 (pace: fast, slow) repeated-measures ANOVA revealed a significant main effect of pace, F(1, 66) = 20.47, p < .01, gp2 = .24, with poorer performance in the fast paced condition (M = .98 and .94 for the slow and fast paced conditions respectively), but no significant effect of processing position, F(1, 66) = 0.62, p = .43, gp2 = .01, and no significant interaction between pace and processing position, F(1, 66) = 2.00, p = .16, gp2 = .03.

The results from Experiment 2 are consistent with the findings from Experiment 1. Decreasing the opportunity for attentional refreshing by increasing the pace of the processing activity did lower recall performance, consistent with the findings of Barrouillet and colleagues (Barrouillet et al., 2004; Camos et al., 2009). However, providing a delay interval directly after presentation of the storage items improved memory recall performance, consistent with a process of consolidation. These findings cannot easily be explained in terms of a trade-off between the memory and processing components as performance on the processing activity was high (>90% in all conditions) and unaffected by processing position. Additionally, as in Experiment 1, providing an opportunity for consolidation in the delayed conditions did not fully protect recall from an increase in attentional demand as the effect of processing pace was comparable across both the immediate and the delayed processing conditions. Thus, the processes of consolidation and attentional refreshing appear to be separable and both appear to be important for working memory performance. Even with the reduced opportunity for articulatory rehearsal in this experiment, it is still possible that some participants were able to rehearse during the delay interval, and that this may have contributed to the improved performance on the delayed conditions of the working memory task. That is, the effect of processing position in Experiments 1 and 2 may reflect the beneficial effect of articulatory rehearsal performed early in the trial rather than any true effect of consolidation. Having said this,

Experiment 2

Experiment 3

(no articulatory suppression)

(articulatory suppression)

0.8

0.6

0.6

0.4

0.4

0.2

0.2

Immediate

Processing task

Proportion correct

Recall task

0.8

Discussion

Delayed

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

Immediate

Delayed

Slow Fast

Immediate

Delayed

Immediate

Delayed

Processing position Fig. 2. Mean proportion of items correct for the recall task and mean proportion of trials correct for the processing task in each condition with standard error bars for Experiments 2 and 3.

42

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

Stevanovski and Jolicoeur (2007, Experiment 3) showed that consolidation functions similar to those observed in the earlier work of Jolicoeur and Dell’Acqua (1998) were evident even when participants were required to perform an articulatory suppression task, suggesting that the consolidation and rehearsal of information involves separable processes (see also, Sun, Zimmer, & Fu, 2011, though it is unclear in their paradigm whether they are measuring basic encoding, consolidation or a combination of the two). Moreover, Camos et al. (2009) and Jarrold et al. (2011) have shown that independently manipulating participants’ ability to engage in articulatory rehearsal and attentional refreshing produces separable effects on memory performance. Taken together, these studies suggest the possibility of three independent processes that contribute to the maintenance of verbal information in working memory, namely, attentional refreshing, articulatory rehearsal, and consolidation. Experiment 3 was designed to investigate this possibility in the current experimental paradigm, by impeding participants’ ability to engage in articulatory rehearsal through the use of articulatory suppression. If providing a delay interval immediately following the presentation of each storage item is beneficial for memory performance, even when rehearsal is blocked by articulatory suppression, this will provide further support for the independence of the consolidation process from rehearsal processes, and the importance of memory consolidation for working memory performance. Experiment 3 In Experiment 3, participants completed the same four working memory tasks as in Experiment 2. However, in this experiment, we blocked articulatory rehearsal by having participants engage in articulatory suppression during the blank delay intervals that occurred between the presentation of the memory item and the processing activity in the delayed conditions, and between the presentation of the processing activity and the next memory item in the immediate conditions. If the enhanced memory performance evident in the delayed conditions of the previous experiments is due to the beneficial effect of early rehearsal (as opposed to rehearsal performed after the presentation of the processing activity when the memory traces may have become degraded), then the memory advantage for the delayed processing position conditions should be eliminated in this experiment. If, however, consolidation and articulatory rehearsal are separable processes, then we should replicate the pattern of performance evident in Experiment 2. Participants’ accuracy on the processing task was again analysed to assess any effects of processing position on processing performance. Method Participants Forty undergraduate Psychology students (10 males) aged between 17 and 30 years (mean age = 19.85 years) participated in this study. All students received credit points or a small payment in return for taking part in the experiment.

Tasks and procedure The tasks and procedure for this study were identical to those used in Experiment 2, with the exception that articulatory suppression was also completed during each working memory task. This additional articulatory suppression task required participants to repeat ‘Monday, Monday’ in the 1000 ms delay period either immediately before or immediately after the processing task, depending on the processing position condition. Working memory tasks and the Ravens Standard Progressive Matrices – Plus Version were administered according to the procedure described in Experiment 2. Results Data screening Participants were screened on the basis of their performance on the processing task, as in Experiment 2. Following this procedure, three female participants were excluded from subsequent analyses on the basis of their processing performance. The remaining 37 participants had a mean age of 19.69 years, with a range of 17 years to 30 years. Data analysis Descriptive statistics for all working memory variables are presented in Fig. 2. Performance on the Raven’s Progressive Matrices task for these participants resulted in a mean score of 38.89 with a standard deviation of 5.81. A 2 (processing position: immediate, delayed)  2 (pace: fast, slow) repeated-measures ANOVA conducted on recall accuracy revealed a significant main effect of processing position, F(1, 36) = 15.66, p < .01, gp2 = .303, with better performance in the delayed condition (M = .43) than the immediate condition (M = .36), but no significant effect of pace, F(1, 36) = 1.56, p = .22, gp2 = .042, and no significant interaction between processing position and pace, F(1, 36) = 0.04, p = .85, gp2 = .001. An analysis of accuracy on the processing tasks revealed significant main effects of pace, F(1, 36) = 13.09, p < .01, gp2 = .267, with better performance in the slow pace condition (M = .92 and .86 for the slow and fast paced conditions respectively), and processing position, F(1, 36) = 17.86, p < .01, gp2 = .332, with better performance in the delayed condition (M = .85 and .93 for the immediate and delayed conditions respectively). There was no significant interaction between processing position and pace, F(1, 36) = 1.99, p = .17, gp2 = .052. Cross-group comparison The performance of the participants in Experiments 2 and 3 was compared to investigate the overall effect of articulatory suppression, processing position and pace on working memory performance. To ensure that there was no difference in the ability of the two groups, performance on the Ravens Progressive Matrices was compared across experiments, with no significant difference found, t(102) = 0.41, p = .682. Following this, a 2 (experimental group: no articulatory suppression (Experiment 2), articulatory suppression (Experiment 3))  2 (pace: fast, slow)  2 (processing position: immediate, delayed) mixed ANOVA was conducted. This revealed a significant main

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

effect of experimental group, F(1, 102) = 39.07, p < .01, gp2 = .277, with poorer performance when articulatory suppression was required (M = .58 and .40 for Experiment 2 and Experiment 3 respectively). Significant main effects were also found for pace, F(1, 102) = 26.51, p < .01, gp2 = .206 (M = .52 and .46 for the slow and fast pace conditions respectively), and processing position, F(1, 102) = 28.17, p < .01, gp2 = .216 (M = .46 and .52 for the immediate and delayed conditions respectively). The only significant interaction effect found was between experimental group and pace, F(1, 102) = 7.43, p < .01, gp2 = .068. As already shown above, increasing the pace of the processing activity had a detrimental effect on memory performance for the participants in Experiment 2 who were not engaging in articulatory suppression. The same increase in pace had no reliable effect on the memory performance of participants in Experiment 3 who were engaging in articulatory suppression, even though overall memory performance for this group was poorer than that of participants in Experiment 2, indicating that articulatory suppression was having an effect. Discussion Consistent with Experiments 1 and 2, the findings from this third experiment again showed that providing a delay after the presentation of each memory item resulted in better memory performance relative to when the same delay interval was provided after the presentation of the processing activity. Importantly, the beneficial effect of providing a delay after each memory item was evident even though participants were engaging in articulatory suppression during the delay intervals in this experiment, and thus, their ability to rehearse the memory items was severely restricted. This provides strong evidence to suggest that the enhanced memory performance observed in the delayed condition of this experiment, and the previous two experiments, is due to participants having the opportunity to consolidate their memory before completing the processing component of the task, and is not due to a beneficial effect of rehearsal performed directly after presentation of the memory item (as opposed to rehearsal performed after the processing activity). Articulatory suppression did reduce memory performance overall, indicating that the manipulation was effective and that rehearsal processes appear to contribute to performance on these working memory tasks. However, the finding that the effect of processing position was unaffected by the introduction of articulatory suppression suggests that the processes involved in rehearsal and consolidation are independent (cf. Stevanovski & Jolicoeur, 2007; Sun et al., 2011). Surprisingly, in this experiment, the pace manipulation did not affect memory performance. This is inconsistent with the results of Experiments 1 and 2, and also with the findings of many previous studies (e.g., Barrouillet et al., 2004; Camos et al., 2009). The fact that the concurrent performance of articulatory suppression eliminated the effect of pace could be taken to suggest an overlap in the mechanisms or resources required for articulatory rehearsal and attentional refreshing. However, given the

43

previous evidence to the contrary (Camos et al., 2009; Tam et al., 2010), this seems unlikely. Instead, it is possible that by requiring participants to perform articulatory suppression during the delay interval, we have increased the cognitive load of the task to the point where retention duration becomes important and the longer retention duration in the slow paced condition has reduced the recall advantage usually evident in this condition (see Barrouillet et al., 2004, p. 90, for a description of the conditions under which such an effect might occur). Further research is needed to clarify the limits of the effect of processing pace on working memory performance. Additional evidence to support a process of consolidation comes from performance on the processing activity of the working memory task. Participants were more accurate on the processing task in the delayed processing position condition than the immediate condition. This is consistent with the idea that individuals in the immediate condition were still engaged with consolidating the memory item when the processing activity was presented, resulting in impaired performance on the digit naming task. Jolicoeur and Dell’Acqua (1998) found similar evidence of an impairment in Task 2 accuracy when Task 1 required the encoding of a memory item (i.e., their Experiment 7), however, these effects were highly variable and were only evident at the shortest SOAs in their study. Given that the effects of processing position on processing accuracy were also variable in the current experiments (i.e., no significant effect in Experiment 2), further research is required to determine the nature and extent of any effects that consolidating information has on the accuracy of subsequent processing. Nonetheless, taken together, the results of Experiments 2 and 3 suggest that the process of consolidation is separable from both the process of rehearsal and attentional refreshing, and that all three are important for working memory performance.

Experiment 4 Experiments 1–3 have shown a beneficial effect of providing a blank delay interval after the presentation of each memory item, which we have interpreted as evidence of a consolidation process. However, it could be argued that the timings that we have used in our experiments are relatively long compared to the timings used by Jolicoeur and Dell’Acqua (1998) to assess short-term consolidation. In their experiments, Jolicoeur and Dell’Acqua showed that the consolidation of a single letter appeared to be largely complete between 500 and 700 ms post-stimulus onset. In our experiments, in the immediate conditions, each letter was presented for 500 ms followed by a 100 or 200 ms blank delay interval before the onset of the processing activity (these timings, though brief for a working memory span task, were chosen to try to capture the consolidation process whilst still ensuring some generalisability to other working memory tasks). Consequently, it could be argued that any consolidation would be complete prior to the onset of the processing activity, even in our immediate conditions. The final experiment was designed to address this question.

44

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

In Experiment 4, participants completed a serial recall task with an auditory choice RT task (judging whether a presented tone was high or low in pitch) inserted between each presentation of a memory item. Consistent with the paradigm used by Jolicoeur and Dell’Acqua (1998), the tones were presented at varying delay intervals following the presentation of each memory item, and participants were required to respond as quickly as possible to the tones without sacrificing accuracy on the tone judgement task. If participants are still consolidating the memory item when the tone is presented, then we should see reaction time functions consistent with those demonstrated by Jolicoeur and Dell’Acqua (1998). That is, reaction times to the tone should be slower when the interval between the offset of the memory item and the presentation of the tone is short, and should become faster as this delay interval increases. However, if consolidation is already complete by the time that the tones are presented, then reaction times should be unaffected by variation in the delay interval between presentation of the memory item and the tone. By examining these reaction time functions across serial position, we will also be able to address the question of what happens to these reaction time functions as the number of items maintained in memory increases, which will provide a better understanding of the processes that are operating within a working memory task. Method Participants Twenty-one undergraduate university students (9 males) aged between 17 and 28 years (mean age = 19.24 years) participated in the study. Participants received either course credit or a small payment for participation. Task and procedure The serial recall task involved remembering lists of letters and responding to an auditory choice RT task presented after each letter. Each trial began with the presentation of a fixation cross in the centre of the computer screen for 1000 ms, followed by a letter presented in black Arial font approximately 22 mm in height for 500 ms. Following a variable delay interval of either 50, 150, 350, 750 or 1550 ms post stimulus offset, a tone that was either high (1000 Hz) or low (500 Hz) in pitch was presented for 100 ms and participants were required to decide if it was low or high and respond by pressing the ‘A’ and ‘L’ keys respectively, as quickly as possible but without making mistakes. The inter-letter interval was fixed at 3550 ms. This ensured that there was adequate time to respond to the tone at the longest delay interval and that the overall duration of each trial was the same for all participants. At the end of each trial, participants were presented with a screen containing all possible letters and were asked to verbally recall the letters in the same order that they were presented. The experimenter recorded their responses and entered them into the computer. Responses were scored by calculating the proportion of letters recalled accurately in each serial position. In addition, participants were required to complete articulatory suppression throughout each trial by repeating the word

‘‘Monday’’ continuously from the onset of the fixation cross until they were asked to recall. The letter stimuli were the same as in the previous experiments and were randomly selected without replacement for each trial. Each trial was five letters in length, with a total of 50 trials. An additional four practice trials were presented at the beginning of the task. The delay interval that followed the presentation of each letter was randomised with the constraint that each of the five delay intervals was used once within each trial, and each delay interval was presented an equal number of times in each serial position. There were an equal number of high and low tones presented for each delay interval in each serial position. The tones were presented through Sennheiser headphones at a clearly audible level. The task was programmed using LiveCode and presented using a standard computer and monitor. Results Data screening Participants were screened based on their recall accuracy performance. One female participant was removed on this basis as their recall accuracy was less than 20%, which was more than 2.5 standard deviations below the mean for the sample. Reaction times for correct responses on the tone task were trimmed to remove any responses greater than 2000 ms as this was the maximum time allowed for responses at the longest delay interval. These responses were coded as a timed-out error. Each participant’s individual reaction times were then assessed for outliers for each delay interval at each serial position with none identified. Accuracy on the tone task was also evaluated with all participants performing above 70%. The remaining 20 participants had a mean age of 19.30 years, with a range of 17–28. Data analysis Mean RTs for the tone task are presented in Fig. 3. A clear reduction in RTs with increasing delay interval is evident. This was confirmed by a 5 (serial position)  5 (delay interval) repeated measures ANOVA, which revealed a significant main effect of delay interval, F(4, 76) = 16.40, p < .01, gp2 = .463 (M = 813.73, 799.86, 770.65, 748.79 and 718.51 for the 50, 150, 350, 750 and 1550 ms delay intervals respectively), a significant main effect of serial position, F(4, 76) = 8.55, p < .01, gp2 = .310 (M = 768.60, 722.50, 755.31, 793.62 and 811.50 for serial positions 1–5 respectively), and no significant interaction (p > .10). Follow-up contrasts revealed no significant difference between the 50 and 150 ms delay intervals (p > .10), but a significant reduction in RTs from 150 to 350 ms, F(1, 19) = 11.96, p < .01, gp2 = .386, from 350 to 750 ms, F(1, 19) = 4.79, p = .04, gp2 = .201, and from 750 to 1550 ms, F(1, 19) = 5.00, p = .04, gp2 = .208. For serial position, contrasts showed that RTs were significantly faster for the second serial position relative to the first, F(1, 19) = 5.83, p = .03, gp2 = .235, but then showed a progressive slowing with slower RTs for the third serial position relative to the second, F(1, 19) = 9.64, p < .01, gp2 = .337, the fourth serial position relative to the third, F(1, 19) = 15.80,

45

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

950

Serial Position 1 Serial Position 2

Mean Reaction Time (ms)

Serial Position 3 Serial Position 4

850

Serial Position 5

750

650

0

0

500

1000

1500

2000

Delay Interval Fig. 3. Mean reaction time to tones presented at each delay interval for each serial position in Experiment 4. Error bars are omitted for clarity.

p < .01, gp2 = .454, and no significant difference between the last two serial positions (p > .10). Descriptive statistics for accuracy on the tone task are presented in Table 1. A 5 (serial position)  5 (delay interval) repeated measures ANOVA revealed a significant main effect of serial position, F(4, 76) = 3.44, p = .01, gp2 = .153, but no significant main effect of delay interval and no significant interaction (all p > .10). Follow-up contrasts showed that accuracy was better for the second serial position (.97) relative to the first (.94), F(1, 19) = 6.69, p = .02, gp2 = .260, but no other significant differences between subsequent serial positions were evident (all p > .10). Descriptive statistics for recall accuracy are also presented in Table 1 and show a typical serial position curve.

for a process of short-term consolidation. Importantly, we have demonstrated that similar RT functions are evident over longer time frames such as those used in the current experiment, where each memory item was presented for 500 ms and RTs were still decreasing up to 1550 ms poststimulus offset. This suggests that the process of consolidation may be operating over a longer time period than originally proposed by Jolicoeur and Dell’Acqua (1998). Indeed, in the more recent study by Stevanovski and Jolicoeur (2007, Experiment 3), RTs to a secondary tone task presented at varying intervals following a memory item continue to decrease between 500 and 1600 ms poststimulus onset, similar to the findings of the present experiment. Moreover, by combining the paradigm used by Jolicoeur and Dell’Acqua (1998) with a serial recall task to create a novel working memory task, we have revealed another intriguing finding, namely, that RTs increased across serial position (with the exception of the first position), but without an interaction with delay interval. While the slower RTs evident for the first serial position may reflect a cost associated with initiating a task set at the beginning of each trial, the overall increase in RT across the remaining serial positions suggests there is also a cost of maintaining information in memory. Crucially though, serial position did not interact with delay interval, suggesting that the processes associated with initiating a task set and maintaining items in memory are separable from those involved in consolidating memory items. Finally, as participants completed articulatory suppression throughout the task, the findings from this experiment provide further support for the separability of consolidation from the process of articulatory rehearsal. Thus, we have demonstrated that RT functions consistent with a process of short-term consolidation are evident at the item presentation times used in Experiments 1–3, which provides converging evidence to support the interpretation of the effects of processing position demonstrated in the previous experiments in terms of a process of consolidation.

Discussion General discussion The results from Experiment 4 show a clear pattern of decreasing RTs as the delay interval between the presentation of the memory item and the tone increases. This finding is consistent with those reported by Jolicoeur and colleagues (Jolicoeur & Dell’Acqua, 1998; Stevanovski & Jolicoeur, 2007), which they have argued provide support

The current experiments were designed to investigate whether short-term memory consolidation contributes to performance on working memory span tasks. In addition, we aimed to differentiate between consolidation and two previously identified processes that contribute to the

Table 1 Mean proportion correct and standard deviations for recall accuracy across serial position and tone accuracy for each delay interval across serial position. Variables

Serial position 1

Recall accuracy Tone accuracy 50 150 350 750 1550

2

3

4

5

M

SD

M

SD

M

SD

M

SD

M

SD

.87

.12

.82

.14

.71

.18

.62

.18

.64

.19

.92 .95 .94 .93 .95

.02 .02 .02 .03 .02

.98 .97 .95 .97 .98

.01 .01 .02 .01 .01

.95 .95 .97 .98 .96

.02 .02 .01 .01 .02

.97 .95 .97 .95 .98

.02 .02 .01 .02 .01

.97 .93 .93 .97 .95

.01 .03 .03 .02 .02

46

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

maintenance of verbal information in working memory, namely rehearsal and attentional refreshing. In three experiments, we demonstrated that presenting a blank delay interval directly after the presentation of the memory item and before the onset of the processing activity produced enhanced memory performance relative to when the same delay interval was presented after participants had completed the processing activity. The pattern of performance observed across experiments suggests that postencoding processes have a beneficial effect on working memory span performance, and is consistent with the process of short-term consolidation described by Jolicoeur and Dell’Acqua (1998; see also Nieuwenstein & Wyble, 2014; Ricker & Cowan, 2014). Importantly, the finding of an effect of processing position in these experiments is novel and provides a challenge to a number of models of working memory. A second novel finding from this study is that whatever post-encoding processes are operating during the delay interval appear to be separable from the processes involved in articulatory rehearsal. When the opportunity for participants to engage in rehearsal of the memory items was blocked by the imposition of articulatory suppression, the beneficial effect of consolidation remained and was of a similar magnitude to the previous experiments that did not require articulatory suppression. These results are consistent with the findings of Stevanovski and Jolicoeur (2007) showing that the consolidation functions evident in Jolicoeur and Dell’Acqua (1998) remain even when participants engage in an articulatory suppression task to minimise any verbal recoding and rehearsal, and provide support for the claim of Jolicoeur and Dell’Acqua (1998) that short-term consolidation is not synonymous with rehearsal. Another intriguing finding from this study is that the post-encoding processes operating during the delay interval also appear to be separable from attentional refreshing. There were effects of processing difficulty (Experiment 1) and processing pace (Experiment 2), indicating that the manipulation of cognitive load, which in turn varies the opportunity for attentional refreshing, was successful in affecting performance. However, in Experiments 1–3, the beneficial effect of providing an opportunity for consolidation was unaffected by the manipulation of cognitive load. Moreover, providing an unfilled delay interval immediately following the presentation of the storage item led to better memory performance than providing the same unfilled delay interval after the processing activity, even though the cognitive load of these two conditions was identical. Of course, it could be argued that maintenance activities performed prior to the processing activity may be more beneficial than maintenance activities performed after the processing activity because memory traces will have deteriorated more in the latter case. In relation to articulatory rehearsal, we would argue that the findings of Experiment 3 provide strong evidence against this suggestion because when rehearsal was blocked in this experiment, the same pattern of superior recall in the delayed processing position condition was observed. Furthermore, although the findings from these experiments do not allow us rule out this possibility with respect to attentional

refreshing, even if participants were able to engage in some form of maintenance activities during the delay intervals, whether that be attentional refreshing or articulatory rehearsal, we believe the pattern of results is not consistent with such an argument. For example, let us assume that in the delayed processing position condition, participants were able to maintain the memory item, either through refreshing or rehearsal, with a high level of fidelity until the onset of the processing activity, at which point the memory trace underwent a certain amount of forgetting while the participant was engaged in the processing activity. This would be the final activation level for this item as the next memory item would then be presented. In the immediate condition, memory traces would begin from the same starting level of activation and would suffer the same degree of loss during the processing activity, however, in this condition participants would then have the delay interval to refresh the memory trace before the presentation of the next memory item. Consequently, and paradoxically, this argument would have to predict either no difference between the two processing position conditions (in the case when the memory trace could not be reactivated after the processing activity because it was too degraded), or better performance in the immediate condition (in the case where the memory trace could be reactivated). The results from the current experiments are clearly not consistent with such a view. Together, the results of these experiments therefore suggest that there may be three mechanisms involved in the maintenance of verbal information in working memory: articulatory rehearsal, attentional refreshing, and a separate mechanism, which we believe may be responsible for the initial consolidation of information into working memory. While we have interpreted the beneficial effect of providing a blank delay interval after the presentation of each memory item as evidence of a consolidation process, there may be other potential explanations of this effect. For example, it could be argued that providing a blank delay interval directly after the memory item allowed more time for sensory encoding. Saults and Cowan (2007) showed that interrupting sensory encoding through the use of postperceptual masks led to poorer performance on visual and auditory array comparison tasks. However, Turvey (1973) showed that the identification of trigrams was no longer affected by postperceptual masks presented approximately 200 ms post-stimulus onset. Thus, we assume that the stimulus presentation times used in the current series of experiments would have provided ample time for the sensory encoding of single letters. Moreover, Nieuwenstein and Wyble (2014) recently showed that whereas the consolidation of a visual stimulus was disrupted by the presentation of an attentionally demanding secondary task (i.e., a 2-alternative forced choice task; 2-AFC), it was unaffected by the presentation of a mask. Taken together, these studies provide converging evidence that the effect of processing position evident in the present study is most likely due to the disruption of a later stage of processing, such as consolidation. However, this in itself raises the question of whether the stimulus presentation times that we have used are so long that any consolidation would be complete prior

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

to the onset of the processing activity. Contrary to this suggestion, Experiment 4 clearly demonstrated that RT functions comparable to those used by Jolicoeur and Dell’Acqua (1998) to argue for a process of short-term consolidation were evident with the stimulus presentation times used in the current experiments (see also, Nieuwenstein and Wyble (2014) who showed that the process of consolidation was still susceptible to disruption by a 2-AFC task for several hundred milliseconds after the stimulus was masked). This suggests that the consolidation of a single item may take longer than originally thought (Jolicoeur & Dell’Acqua, 1998; see also, Stevanovski & Jolicoeur, 2007, Experiment 3). Thus, although consolidation is almost certain to be underway by the time the processing activity is presented in our immediate conditions, the findings from Experiment 4 equally provide strong evidence to suggest that it would not be complete by this time. Another potential alternative explanation of the effect of processing position evident in Experiments 1–3 is that the positioning of the blank delay interval in the delayed condition allowed more time for individuals to switch from encoding (and perhaps also consolidating) the memory items to completing the processing activity in this condition relative to the immediate condition. Oberauer and Lewandowsky (2014) showed that response times to the first of a series of sequential processing items were significantly longer than response times to subsequent processing items (see also, Jarrold et al., 2011; Vergauwe et al., 2014), which they attributed to a task-switch cost. They then argued that if performing a task-switch demands attentional resources, this would lead to a momentary surge in the cognitive load of the task at the point when individuals were required to initiate the processing task. Thus, the better memory performance evident in the delayed conditions of our experiments could, potentially, be explained by a temporary increase in the cognitive load of the task when participants are required to switch to the processing activity, which is more easily accommodated in the delayed condition than the immediate condition. However, there are a number of reasons to believe that taskswitching may not be an issue of concern in the current experiments. First, in all experiments, participants were required to name the memory items, which were letters, and then name the processing items, which were digits. Arguably, any task-switch involved in switching from naming letters to naming digits would be minimal. Indeed, in their seminal paper on the dynamics of task-switching, Allport, Styles, and Hsieh (1994, Experiment 4) showed that after the first few trials, alternating between naming the attributes of two different stimuli (i.e., words and digits) did not incur a task-switch cost. Thus, any task-switch costs occurring in the current experiments are likely to be small at best. Second, in the tasks used in Experiments 1–3, participants were required not only to switch from encoding to processing, but also from processing to encoding the next memory item. In terms of memory performance, if performing a task switch is detrimental, it is reasonable to suggest that the most damaging task switch is likely to be from processing to encoding the subsequent memory item in the delayed condition, rather than from encoding to processing

47

in the immediate condition, and so, we should see poorer performance in the former case. The pattern of results does not support this suggestion. Third, it could be argued that making a task-switch to a more difficult task should be more demanding (cf. Arbuthnott, 2008) and so, we might expect to see a larger effect of delay in the difficult as opposed to the easy processing condition of Experiment 1. This would be evident as a measurable interaction between processing difficulty and processing position, but this was not observed. Finally, it could be argued that participants were required to complete additional task-switching in Experiment 3 as they were required to switch to performing the articulatory suppression task and then to completing the processing task (or vice versa depending on the processing position condition). This means that even in the delayed processing position condition of Experiment 3, participants would have been required to perform a task-switch immediately following the presentation of the memory item, and yet, this did not affect the magnitude of the processing position effect in Experiment 3 or indeed, across Experiments 2 and 3. Thus, it appears that a number of aspects of the data are not consistent with an explanation in terms of task-switching. Importantly, however, even if we accept the possibility of a task-switch cost on the first processing item, this does not impact on our explanation of the beneficial effect of providing a delay interval after the presentation of each memory item in terms of a process of consolidation. Assuming that there is a momentary surge in cognitive load at the start of each processing activity, then this would presumably have its effect on the previous memory item by drawing attentional resources away from, or interfering with, some post-encoding process operating immediately after stimulus offset. We have already demonstrated that these post-encoding processes are unlikely to be articulatory rehearsal or attentional refreshing and so, we would argue that the most plausible candidate is a process of consolidation. Implications for models of working memory The results of the current experiments have important implications for a number of models of working memory. Although Barrouillet et al. (2013) acknowledged a possible role for consolidation, they have yet to explicitly incorporate this process into their TBRS model or the mathematical function governing the trade-off between processing and storage in working memory (cf. Barrouillet et al., 2011). However, given that the description of the process of consolidation given by Jolicoeur and Dell’Acqua (1998) shares many similarities with the attentional refreshing mechanism described by Barrouillet et al. (2004, 2007, 2011) in that both are thought to engage a central attentional bottleneck and both processes act to strengthen the memory trace, it would be more parsimonious if the current results could be explained in terms of a single process responsible for both. Indeed, in a recent computational instantiation of the TBRS, Oberauer and Lewandowsky (2011) included a processing rate variable that governed the speed of both encoding and refreshing, with differences in the overall duration of the two pro-

48

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

cesses achieved by setting different criterion levels for required learning strength. Thus, in this model, a single parameter was used for both consolidation and refreshing. One other difference between encoding and refreshing in this model is that refreshing is contingent on the retrieval of the to-be-refreshed item, whereas encoding is not. To the extent that retrieval for refreshing is error-prone, this could potentially explain the advantage of an early over a late free-time period evident in the present results.2 However, we feel that this suggestion is subject to the same line of reasoning that we have used above to argue against the suggestion that that attentional refreshing performed prior to the processing activity may be more beneficial than attentional refreshing performed after the processing activity. As Oberauer and Lewandowsky (2011) did not simulate the effect of varying encoding strength (which is presumably the way in which the model would simulate an interruption to the process of consolidation), or the effect of varying the position of the processing activity within the inter-item interval, it remains to be seen whether this model would be able to produce effects similar to the experimental data presented here. Nonetheless, even if a single attentional activation process is responsible for both consolidation and refreshing, it is clear that varying the opportunity for this activation to occur in relation to the initial encoding of memory items affects memory independently of varying the opportunity for this activation to occur in relation to refreshing the memory items once they have been encoded. That is, the exact time point at which an opportunity for activation is provided within the working memory span paradigm is important for memory performance. These findings indicate that in order to provide a full account of working memory span performance, the TBRS model of Barrouillet et al. (2004, 2007, 2011) needs to be modified to take into account the temporal parameters associated with the initial encoding of memory items as well as those associated with the maintenance of items. More recently, Oberauer et al. (2012) put forward an alternative computational model (SOB-CS) that explains working memory span performance without reference to decay and/or attentional refreshing mechanisms. SOB-CS instead explains working memory capacity limits in terms of interference that occurs due to the superposition of associations in a distributed neural-network. In this model, distractors create interference by being encoded into working memory and being associated with the memory item preceding the distractor. The strength of encoding of both memory items and distractors is determined by both the amount of time that attention is devoted to the item/ distractor, and the novelty of the item/distractor. To prevent working memory from becoming cluttered, SOB-CS also incorporates a mechanism responsible for the removal of distractors from memory. Distractor removal occurs during any free time that is available following encoding of the distractor. Thus, longer time intervals following the presentation of a distractor item will enable that distractor to be removed more effectively, resulting in less

2

We would like to thank Klaus Oberauer for this suggestion.

interference with the preceding memory item. This would lead to the prediction that performance should be better in the immediate conditions of Experiments 1–3 in the current study as these conditions would provide maximal time for distractor removal after presentation of the distractors. This was not the pattern of performance observed. Moreover, while the strength of memory encoding in this model is a time-dependent process, and therefore consistent with the notion of consolidation described in the current study, the instantiation reported in Oberauer et al. (2012) assumed that encoding strength would be maximal after approximately 500 ms, and so, it remains to be seen whether SOB-CS can reproduce the pattern of effects reported in Experiments 1–3 that operate over a longer time scale. Of course, it is possible that mechanisms associated with both consolidation of the memory traces and distractor removal may be involved in working memory span performance. The extent to which evidence consistent with the operation of either mechanism is apparent in behavioural results may depend on the balance of these processes within any given task. Notably, evidence consistent with the importance of consolidation is clearly demonstrated in the current experiments. The current results also appear not to be fully consistent with a temporal distinctiveness account of memory (e.g. Brown, Neath, & Chater, 2007). According to temporal distinctiveness models, memory items are partly represented in terms of their locations along a temporal dimension in psychological space. Memory traces of items that are presented close in time to one another are crowded along this temporal dimension and, consequently, are more difficult to discriminate from one another at the point of retrieval than temporally isolated items, resulting in poorer memory performance (Brown et al., 2007). A number of studies have shown evidence of temporal isolation effects (e.g. Brown, Morin, & Lewandowsky, 2006; Geiger & Lewandowsky, 2008; Neath & Crowder, 1996), however, these effects are typically reduced and often non-significant in paradigms that require forward serial recall (Morin, Brown, & Lewandowsky, 2010). In the current study, a delay was presented either immediately after the presentation of each memory item (delayed condition) or immediately before the presentation of the following memory item (immediate condition), both of which should have led to a similar degree of temporal isolation for the memory items. However, a memory advantage was found in the condition where the delay was presented immediately following the presentation of the memory item, suggesting that the post-item interval is more important for memory performance than the pre-item interval. This finding is difficult to accommodate in terms of a temporal distinctiveness account. However, having said that, it must be acknowledged that the current experiments did use a closed pool of memory items and used the repetition of a single distractor to block rehearsal in Experiment 3, two conditions that have been suggested to mask temporal isolation effects in serial recall tasks (Morin et al., 2010). Nevertheless, this argument does not discount the fact that a significant effect of post-item interval was found in the current experiments, consistent with a process of consolidation.

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

Implications for models of short-term consolidation These findings also have implications for understanding the nature of the consolidation process itself and the attentional constraints on performance. Jolicoeur and Dell’Acqua (1998) have argued that short-term consolidation requires central mechanisms and that while consolidation processes are engaged, any concurrent processing that also requires central mechanisms will either be postponed or slowed due to some form of capacity-sharing. In the current study, memory performance was impaired when the processing activity was presented immediately following the presentation of the memory item (see also, Nieuwenstein & Wyble, 2014). It is difficult to see how this could be explained by an attentional bottleneck model in which the processing of the second task is postponed until the first task is completed (cf. Jolicoeur & Dell’Acqua, 1998). According to these models, manipulating the onset of Task 2 should not affect Task 1 performance, as Task 1 should have access to the attentional bottleneck first (Tombu & Jolicœur, 2003). The fact that we see effects of distractor processing (Task 2) on memory performance (Task 1) that are moderated by the timing of the onset of the distractor processing activity suggests that these two components of the task are not processed in a strict serial fashion. An alternative explanation is that the allocation of attention to each component of the task is under strategic control such that individuals may switch their attention away from consolidation to focus on the processing activity. This would effectively interrupt the consolidation process resulting in a weaker memory trace that is susceptible to interference from incoming information, and consequently, poorer memory performance. This explanation would seem to fit well with the TBRS model discussed above. Alternatively, the current results could also, potentially, be accounted for by a central capacity sharing model such as that proposed by Tombu and Jolicœur (2003). In this model, the processing required for both Task 1 and Task 2 proceeds in parallel, but any central processing required by each task must share the available processing capacity, which is limited. When the delay between the two tasks is short, there is an increase in the duration during which the two tasks must share the available resources, which leads to costs in performance on both tasks (Tombu & Jolicœur, 2003). Importantly, the proportion of the available resources allocated to each task can vary and may be influenced by task demands such that virtually all of the shared capacity could be allocated to one of the tasks if the task demands encouraged this. This form of graded capacity-sharing could account for the fact that effects of processing position on processing accuracy were seen in Experiment 3, but not Experiment 2, as well as the effects of processing position on memory performance evident across all experiments. That is, when the distractor activity was presented immediately following the presentation of the memory item, participants may have shared their available resources between consolidating the memory item and performing the distractor task. This is likely to have interfered with the consolidation of the memory items, either in terms of the fidelity of the representation that was formed or the strength with which it was encoded,

49

and also performance on the processing activity. Although the present results cannot be used to distinguish between a graded capacity-sharing and task-switching account of central attentional processing, which is an important avenue for future research, they do provide a challenge to attentional bottleneck models that hold the assumption that Task 2 processing is postponed until Task 1 processing is complete (see also, Nieuwenstein & Wyble, 2014, for further evidence against a central attentional bottleneck model of this form). The present results also provide further evidence that attentional limits place important constraints on working memory capacity. Conclusions In conclusion, we have demonstrated that providing an opportunity for consolidation immediately following the presentation of each memory item in a working memory span task leads to better working memory performance. Moreover, the beneficial effect of providing an opportunity for consolidation was independent of manipulations of processing difficulty, processing pace, and articulatory suppression. These results suggest that the mechanisms underlying the consolidation of information into working memory are separable from the mechanisms underlying attentional refreshing and articulatory rehearsal and indicate that the temporal parameters associated with consolidating information into working memory need to be acknowledged as an important factor that contributes to working memory performance and incorporated into current models of working memory. Furthermore, the current results highlight the importance of manipulating or controlling the opportunity for consolidation when designing experiments involving working memory span paradigms as variation in the time allowed for consolidation can lead to differences in working memory performance that may mask the influence of other variables on working memory. Finally, the results of these experiments may have practical implications in terms of assisting those with working memory problems to improve their working memory performance, which would be of great benefit to educators and clinicians alike. Acknowledgments This research was supported by Australian Research Council (ARC) Discovery Project DP0988288 and a Small Research Grant from the University of Western Australia awarded to Donna Bayliss and Christopher Jarrold. References Allport, D. A, Styles, E. A, & Hsieh, S. (1994). Shifting intentional set: Exploring the dynamic control of tasks. In C. Umilta & M. Moscovitch, et al. (Eds.), Attention and performance 15: Conscious and nonconscious information processing. Cambridge, MA: Mit Press (pp. Attention and performance series. 421–452). Arbuthnott, K. D. (2008). Asymmetric switch cost and backward inhibition: Carryover activation and inhibition in switching between tasks of unequal difficulty. Canadian Journal of Experimental Psychology/Revue canadienne de psychologie expérimentale, 62, 91–100. http://dx.doi.org/10.1037/1196-1961.62.2.91. Awh, E., Jonides, J., Smith, E. E., Schumacher, E. H., Koeppe, R. A., & Katz, S. (1996). Dissociation of storage and rehearsal in verbal working

50

D.M. Bayliss et al. / Journal of Memory and Language 81 (2015) 34–50

memory: Evidence from positron emission tomography. Psychological Science, 7, 25–31. Baddeley, A. D. (1986). Working memory. Oxford: Clarendon Press. Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.). The psychology of learning and motivation: Advances in research and theory (Vol. 8, pp. 47–89). New York: Academic Press. Baddeley, A., Lewis, V., & Vallar, G. (1984). Exploring the articulatory loop. The Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 36, 233–252. http://dx.doi.org/10.1080/ 14640748408402157. Baddeley, A. D., Thomson, N., & Buchanan, M. (1975). Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575–589. http://dx.doi.org/10.1016/S00225371(75)80045-4. Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time constraints and resource sharing in adults’ working memory spans. Journal of Experimental Psychology: General, 133, 83–100. http://dx.doi.org/ 10.1037/0096-3445.133.1.83. Barrouillet, P., Bernardin, S., Portrat, S., Vergauwe, E., & Camos, V. (2007). Time and cognitive load in working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, 570–585. http:// dx.doi.org/10.1037/0278-7393.33.3.570. Barrouillet, P., Plancher, G., Guida, A., & Camos, V. (2013). Forgetting at short term: When do event-based interference and temporal factors have an effect? Acta Psychologica, 142, 155–167. http://dx.doi.org/ 10.1016/j.actpsy.2012.12.003. Barrouillet, P., Portrat, S., & Camos, V. (2011). On the law relating processing to storage in working memory. Psychological Review, 118, 175–192. http://dx.doi.org/10.1037/a0022324. Brown, G. A., Morin, C., & Lewandowsky, S. (2006). Evidence for timebased models of free recall. Psychonomic Bulletin & Review, 13, 717–723. http://dx.doi.org/10.3758/bf03193986. Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory. Psychological Review, 114, 539–576. http://dx.doi.org/ 10.1037/0033-295x.114.3.539. Camos, V., Lagner, P., & Barrouillet, P. (2009). Two maintenance mechanisms of verbal information in working memory. Journal of Memory and Language, 61, 457–469. http://dx.doi.org/10.1016/ j.jml.2009.06.002. Chun, M. M., & Potter, M. C. (1995). A two-stage model for multiple target detection in rapid serial visual presentation. Journal of Experimental Psychology: Human Perception and Performance, 21, 109–127. http:// dx.doi.org/10.1037/0096-1523.21.1.109. Cowan, N. (1988). Evolving conceptions of memory storage, selective attention, and their mutual constraints within the human information-processing system. Psychological Bulletin, 104, 163–191. http://dx.doi.org/10.1037/0033-2909.104.2.163. Cowan, N. (1995). Attention and memory: An integrated framework. London: Oxford University Press. Cowan, N. (1999). An embedded-processes model of working memory. In A. Miyake & P. Shah (Eds.), Models of working memory: Mechanisms of active maintenance and executive control (pp. 62–101). New York: Cambridge University Press. Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450–466. http://dx.doi.org/10.1016/S0022-5371(80)90312-6. Geiger, S., & Lewandowsky, S. (2008). Temporal isolation does not facilitate forward serial recall—Or does it? Memory & Cognition, 36, 957–967. http://dx.doi.org/10.3758/mc.36.5.957. Jarrold, C., Tam, H., Baddeley, A. D., & Harvey, C. E. (2011). How does processing affect storage in working memory tasks? Evidence for both domain-general and domain-specific effects. Journal of Experimental Psychology: Learning, Memory, and Cognition, 37(3), 688–705. http:// dx.doi.org/10.1037/a0022527 (705a). Johnson, M. K. (1992). MEM: Mechanisms of recollection. Journal of Cognitive Neuroscience, 4, 268–280. http://dx.doi.org/10.1162/ jocn.1992.4.3.268. Jolicoeur, P. (1999). Dual-task interference and visual encoding. Journal of Experimental Psychology: Human Perception and Performance, 25, 596–616. http://dx.doi.org/10.1037/0096-1523.25.3.596. Jolicoeur, P., & Dell’Acqua, R. (1998). The demonstration of short-term consolidation. Cognitive Psychology, 36, 138–202. http://dx.doi.org/ 10.1006/cogp.1998.0684. Jolicoeur, P., & Dell’Acqua, R. (1999). Attentional and structural constraints on visual encoding. Psychological Research/Psychologische Forschung, 62, 154–164. http://dx.doi.org/10.1007/s004260050048.

Lépine, R., Bernardin, S., & Barrouillet, P. (2005). Attention switching and working memory spans. European Journal of Cognitive Psychology, 17, 329–345. http://dx.doi.org/10.1080/09541440440000014. Massaro, D. W. (1975). Experimental psychology and information processing. Chicago: Rand McNally College. Morin, C., Brown, G. A., & Lewandowsky, S. (2010). Temporal isolation effects in recognition and serial recall. Memory & Cognition, 38, 849–859. http://dx.doi.org/10.3758/mc.38.7.849. Neath, I., & Crowder, R. G. (1996). Distinctiveness and very short-term serial position effects. Memory, 4, 225–242. http://dx.doi.org/10.1080/ 09658211.1996.9753032. Nieuwenstein, M., & Wyble, B. (2014). Beyond a mask and against the bottleneck: Retroactive dual-task interference during working memory consolidation of a masked visual target. Journal of Experimental Psychology: General, 143, 1409–1427. http://dx.doi.org/ 10.1037/a0035257. Oberauer, K., & Lewandowsky, S. (2011). Modelling working memory: A computational implementation of the Time-Based Resource-sharing theory. Psychonomic Society, 18, 10–45. http://dx.doi.org/10.3758/ s13423-010-0020-6. Oberauer, K., & Lewandowsky, S. (2014). Further evidence against decay in working memory. Journal of Memory and Language, 73, 15–30. http://dx.doi.org/10.1016/j.jml.2014.02.003. Oberauer, K., Lewandowsky, S., Farrell, S., Jarrold, C., & Greaves, M. (2012). Modeling working memory: An interference model of complex span. Psychonomic Bulletin & Review, 19, 779–819. http://dx.doi.org/ 10.3758/s13423-012-0272-4. Raye, C. L., Johnson, M. K., Mitchell, K. J., Greene, E. J., & Johnson, M. R. (2007). Refreshing: A minimal executive function. Cortex, 43, 135–145. http://dx.doi.org/10.1016/S0010-9452(08)70451-9. Raye, C. L., Johnson, M. K., Mitchell, K. J., Reeder, J. A., & Greene, E. J. (2002). Neuroimaging a single thought: Dorsolateral PFC activity associated with refreshing just-activated information. NeuroImage, 15, 447–453. http://dx.doi.org/10.1006/nimg.2001.0983. Ricker, T. J., & Cowan, N. (2014). Differences between presentation methods in working memory procedures: A matter of working memory consolidation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 417–428. http://dx.doi.org/10.1037/ a0034301. Saults, J. S., & Cowan, N. (2007). A central capacity limit to the simultaneous storage of visual and auditory arrays in working memory. Journal of Experimental Psychology: General, 136, 663–684. http://dx.doi.org/10.1037/0096-3445.136.4.663. Stevanovski, B., & Jolicoeur, P. (2007). Visual short-term memory: Central capacity limitations in short-term consolidation. Visual Cognition, 15, 532–563. http://dx.doi.org/10.1080/13506280600871917. Stevanovski, B., & Jolicoeur, P. (2011). Consolidation of multifeature items in visual working memory: Central capacity requirements for visual consolidation. Attention Perception & Psychophysics, 73, 1108–1119. http://dx.doi.org/10.3758/s13414-011-0099-5. Sun, H., Zimmer, H. D., & Fu, X. (2011). The influence of expertise and of physical complexity on visual short-term memory consolidation. The Quarterly Journal of Experimental Psychology, 64, 707–729. http:// dx.doi.org/10.1080/17470218.2010.511238. Tam, H., Jarrold, C., Baddeley, A. D., & Sabatos-DeVito, M. (2010). The development of memory maintenance: Children’s use of phonological rehearsal and attentional refreshment in working memory tasks. Journal of Experimental Child Psychology, 107, 306–324. http:// dx.doi.org/10.1016/j.jecp.2010.05.006. Tan, L., & Ward, G. (2008). Rehearsal in immediate serial recall. Psychonomic Bulletin & Review, 15, 535. http://dx.doi.org/10.3758/ PBR.15.3.535. Tombu, M., & Jolicœur, P. (2003). A central capacity sharing model of dual-task performance. Journal of Experimental Psychology: Human Perception and Performance, 29, 3–18. http://dx.doi.org/10.1037/00961523.29.1.3. Turvey, M. T. (1973). On peripheral and central processes in vision: Inferences from an information-processing analysis of masking with patterned stimuli. Psychological Review, 80, 1–52. http://dx.doi.org/ 10.1037/h0033872. Vergauwe, E., Camos, V., & Barrouillet, P. (2014). The impact of storage on processing: How is information maintained in working memory? Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 1072–1095. http://dx.doi.org/10.1037/a0035779.