Acta Psychologica 148 (2014) 123–129

Contents lists available at ScienceDirect

Acta Psychologica journal homepage: www.elsevier.com/ locate/actpsy

When working memory updating requires updating: Analysis of serial position in a running memory task Marta Botto a,⁎, Demis Basso b,c,1, Marcella Ferrari a,2, Paola Palladino a,3 a b c

Department of Brain and Behavioral Science, University of Pavia, Piazza Botta 6, 27100 Pavia, Italy Faculty of Education, Free University of Bozen, Viale Ratisbona 16, 39042 Bressanone, Italy CeNCA, Centro di Neuroscienze Cognitive Applicate Rome, Italy

a r t i c l e

i n f o

Article history: Received 1 January 2013 Received in revised form 18 December 2013 Accepted 20 January 2014 Available online 10 February 2014 PsycINFO classifications: 2300 2340 2343 Keywords: Updating Working memory Prospective memory Serial position curve

a b s t r a c t This study aimed to investigate updating in working memory (WM), analyzing the effects of task demand and memory resources on serial position curve (SPC), in a running memory task with slow pace presentation and a probed recognition procedure. These task conditions were supposed to produce an easier WM updating task, which may allow evidencing whether the task is performed through an active or a passive updating. Serial position curves were compared in conditions of high or low memory load, and with or without interference of a secondary (prospective memory, PM) task. With either a high WM load, or a high PM load, results showed a SPC with both primacy and recency effects, indicating the use of an active strategy. When resources were taken up by both PM task and high WM demand the usual pattern with only recency effect was obtained. Taken together, these findings support the ideas that 1 — people can effectively update WM, and 2 — the performance is dependent on both memory and executive resource availability. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Efficient use of working memory (WM) implies selectively focusing on goal-relevant information. From this perspective, the role of working memory updating becomes crucial: this is a specific mechanism of continuous monitoring, selection of incoming information and replacement of no-longer-relevant information with new, more relevant material (Morris & Jones, 1990). Traditionally, WM updating is investigated with a running memory task procedure (Pollack, Johnson, & Knaff, 1959), asking participants to recall the last few items of lists of uncertain lengths. This kind of task was used by Morris and Jones (1990) to study updating process within the WM model of Baddeley (1986). These authors conceptualized updating as a continuous all-or-nothing mechanism of maintenance-substitution, with the maintenance function carried out by the phonological loop and substitution by the central executive (Morris & Jones, 1990). The running memory task has also been considered a suitable procedure for investigating executive functioning (Miyake, Friedman, Emerson, Witzki, & Howerter, 2000), in connection with either a general ⁎ Corresponding author. Tel.: +39 0382 986273; fax: +39 0382 986132. E-mail addresses: [email protected] (M. Botto), [email protected] (D. Basso), [email protected] (M. Ferrari), [email protected] (P. Palladino). 1 Tel.: +39 0472 014294. 2 Tel.: +39 0382 986273. 3 Tel.: +39 0382 986271. 0001-6918/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.actpsy.2014.01.012

intelligence factor (Friedman et al., 2006), or other complex cognitive processes (such as reading comprehension; e.g., Palladino, Cornoldi, De Beni, & Pazzaglia, 2001). It has been argued (Postle, 2003) that five discrete mental operations in short-term memory (STM) are necessary to execute this task: adding items to STM (encoding), discarding items from STM, repositioning, storing and, finally, rehearsing items in STM. However, it is considered as an updating task due to the discarding and repositioning operations only, which may themselves require executive control processes (D'Esposito & Postle, 1999, 2000). Following the pioneering work of Morris and Jones (1990), running memory tasks are often described by use of serial position curves (SPCs). In this instance, SPCs were characterized by a marked recency effect, without primacy. This pattern in running memory tasks was validated by several other studies (Bunting, Cowan, & Saults, 2006; Fiore, Borella, Mammarella, & De Beni, 2011; Palladino & Jarrold, 2008; Ruiz, Elosùa, & Lechuga, 2005) and is therefore considered a robust result. Based on this dissociation, Bunting et al. (2006) hypothesized that two different strategies may be used to perform the task. The first consists of an active strategy of continuous updating of memory content (i.e., consistent with Postle, 2003), while the second consists of a passive “wait” until the end of the list, and a subsequent recall of the most recent items. The use of one of these two strategies was thought to depend on both task demand and memory availability, in turn, enabling use of executive resources. For example, with either low task demands and/or low memory load, participants may have had more executive

124

M. Botto et al. / Acta Psychologica 148 (2014) 123–129

resource available. Thus, they were likely to continuously update information and rehearse the updated sequence. Under these conditions an active strategy is used, otherwise a passive strategy is preferred. To explore this further, Palladino and Jarrold (2008) examined the strategies employed in a running memory updating task by comparing SPCs in updating tasks with those in standard serial recall. The results showed a clear “uncertainty effect”, with worse performance for updating trials compared with serial recall trials of identical length, and an overall lack of primacy in the updating curves. Taken together, this evidence suggested that participants may not be engaged in continuous active maintenance/updating. Moreover, it suggested that serial position analysis could be an effective procedure to investigate the mechanism underlying updating tasks. A possible alternative explanation of recency effects in the serial curve is provided by the SIMPLE model (Brown, Neath, & Chater, 2007; but see also Brown, Vousden, McCormack, & Hulme, 1999), which assumes that people represent items according to their position in a multidimensional space. Among several dimensions (such as ordinal list position or phonological similarity), a central role for recall is played by the relative temporal distances, 1) between items in the list (i.e., if two items are temporally distant each other, then they are isolated in memory and likelihood of recall increases), and 2) between the items and the moment of recall (i.e., given a logarithmic transformation of the elapsed time, the further items are from the moment of recall, the more mistakable they are). According to the temporal distinctiveness (TD) model, a recency effect is clearly due to the latter point. Whilst, primacy effects could appear only when the distinctiveness of the first items due to the first issue (i.e.: inter-items distance) is strong enough to contrast the hindrance due to the second issue (i.e.: temporal distance form recall). Unlike the standard free recall task, in a running memory task, the first to-be-recalled item is typically preceded by interfering items, thus the distinctiveness of this item due to the first issue is lower than that of the first item of the list. In this situation, a primacy effect is not expected, unless other processes needed in a running memory procedure (e.g., discarding and/or repositioning) might also influence the relevance of the items. However, according to the TD approach, the SPC performance does not depend on other executive processes, but only on characteristics of the stimuli. Geiger and Lewandowsky (2008) used a running memory task procedure, and provided support for the TD account. They showed that both temporal and nontemporal information were maintained in memory until the point of cueing. In particular, Geiger and Lewandowsky's experiments described a clear recency effect when list length was higher than the number of to-be-recalled items (four), and a flat function with list length of four items. When item numerosity is below the individual span, and no discarding/ repositioning operations were needed, all the items have the same likelihood of being recalled. Otherwise, when task demands are higher than individual resources, the first to-be-remembered items is confused among the other items, and other processes, specific for the running memory task, would not influence its retrieval likelihood. Several studies (i.e., Fiore et al., 2011; Palladino & Jarrold, 2008; Ruiz et al., 2005) have presented memoranda at a relatively rapid pace, and this choice may have precluded them from finding a primacy effect. Executive processes, such as discarding and repositioning an item, require time and resources to take place. Therefore, a rapid presentation of stimuli is less likely to allow WM updating occurring in a running memory task. That said, only Bunting et al. (2006) have compared slow and fast presentation pace directly. Their data suggest that a reduction in pace leads to an increase in the primacy portion of the SPC. This notwithstanding, their slow pace presentation (1000 ms) could still be considered fairly rapid and, in both slow and fast pace conditions, they failed to find a clear primacy effect. The authors themselves acknowledge that a slower presentation pace than that used is likely to allow primacy effects to occur in a running memory task (p. 1694). Support for this suggestion was reported in Postle (2003). In his first two experiments, Postle employed a running memory task with presentation pace

ranging from 2.5 to 3.5 s, and a subsequent probe recognition task. The probe consisted of one letter that might (or might not) match an item in the memory set. Unfortunately, using a recognition procedure, Postle did not analyze his results in respect of serial position; for example, in Experiment 3, where he used complete recall instead of probe recognition. Although flat SPC and the absence of primacy were probably due to a ceiling effect here, these experiments provide sound evidence that updating is occurring at this slow presentation rate. The mixed methodology (i.e. recognition and serial recall procedures) used in Postle's study represents a good starting point for further investigations, capable of replicating and extending his results. The present study represents an extension of these data on the running memory task, by means of manipulating memory demands (via memory load and task type), at a slow presentation rate, a probe recognition procedure and the analysis of the serial position curves. Memory load demands were compared via the number of items to be maintained in WM. Participants had to remember either the last 3 or 5 items of the lists (for low and high loads, respectively): these numbers were selected in order to represent quantities sub- and supra-span (Cowan, 2001), which are known to influence performance. A sub-span quantity is not expected to produce a serial curve, but a nearly perfect recall with a flat serial position function, because sufficient memory resources are available for the task. However, supra-span quantities need resources to be optimized. The continuous update of items is expected to ‘push’ people into using memory strategies that would, in turn, increase the likelihood of items retrieval, but also have higher costs in terms of resource demand. With supra-span quantities, if participants are able to optimize resources and actively update items in a continuous stream, rehearsing new updated sequences, a serial curve with both primacy and recency effects would be produced. Otherwise, if the task demands do not allow implementation of such a strategy, the primacy effect would not appear. Consistently with the TD account, the recency effect is not thought to be affected by manipulation of memory load demands, since it is based on retrieval from passive storage at the time of recall (Cowan et al., 2005). However, the primacy effect, which is related to the amount of rehearsal (Tan & Ward, 2000), or to highly focused encoding due to top-down attention (Sederberg, Howard, & Kahana, 2008) during study of early list items, is thought to appear only under conditions with low resource demands and the participants' ability to process incoming stimuli proficiently. Therefore, if the primacy effect appears in running memory task, it may be considered a marker of an effective updating (e.g., Palladino & Jarrold, 2008; Postle, 2003; Sederberg et al., 2008). To manipulate resource availability through task demands, a secondary prospective memory (PM) task was selected to be performed simultaneously with the primary WM updating task. Prospective memory (i.e. possessing a behavioral intention, to be performed at a certain moment in the future) might compete for memory resource with a WM task, as both PM and WM are considered executive processes (Kliegel, Martin, McDaniel, & Einstein, 2002; Mäntylä, 2003; Okuda et al., 1998). In particular, two independent processes have been indicated to be resource-demanding in a PM task (Guynn, 2003; 2008): a retrieval mode and a target checking mechanism. Retrieval mode consists of a continuous monitoring process that occupies the memory central executive by maintaining representation of the PM task. Target checking is a transient process needed to check the environment continuously, detecting PM cues and discarding distracters (e.g.: Bisiacchi, Cona, Schiff, & Basso, 2011). Accordingly, the PM task is a good candidate for engaging the central executive in a resource-consuming activity, being both a continuous secondary task and disrupting on-going activity whenever a cue (or distracting cue) appears. In this vein, Basso, Ferrari, and Palladino (2010) showed that PM demand affected performance in a verbal updating WM task, but only at high WM loads. Conversely, no effects were found with low WM loads, and overall, this effect was enhanced for higher PM demands. These data show that PM and WM compete (at least, partially) for the

M. Botto et al. / Acta Psychologica 148 (2014) 123–129

same available resources, and it seems reasonable that a highly demanding PM task could influence a WM updating task, in a similar fashion to other, more classical tasks (Baddeley, 2000). Therefore, we hypothesized that a concurrent PM task would reduce the resources available to perform a WM updating task. A first experiment was conducted using a WM updating task, to test the effects of high and low WM loads on SPC with slow pace presentation and probe recognition. In a second experiment, a PM task was incorporated, in order to add a resource demanding task and evaluate its effect on serial curve performance. Therefore, task demands were manipulated through the comparison of absence/presence of the concurrent PM task (in the first and second experiment, respectively). Consistent with Postle (2003), we used a running memory task with a slow pace presentation and a recognition procedure, because these conditions are thought to generate an effective updating process. The recognition procedure is known to increase accuracy generally (Oberauer, 2003); since the present task required participants to exploit their memory resources in an unconstrained manner, they were likely to comply with both accuracy and speed requirements. Thus, unlike the common SPC analysis, both accuracy and/or reaction times would represent performance effectively, as recently demonstrated by Ecker, Lewandowsky, and Oberauer (2014). Analysis of the serial position of the probe in study lists will show whether primacy and recency effects have occurred. The recognition (probed) procedure was used in order to produce either positive (presence of the target) or negative (absence) responses. While positive responses enable production of complete SPC, negative ones could support interpretation of the results, and facilitate a refined examination of underlying mechanism. For positive responses, we predicted that a SPC with marked recency effect, but without primacy, would be expected when participants did not actively updated items, either guided by the items, or due to a lack of available memory resource. In other words, participants are likely to wait until the end of the list and, when the probe appears, try to look for (and retrieve) the item. However, according to the TD hypothesis, this strategy will be successful only when the probe is related to the last items in the list, thus generating the recency effect alone. Conversely, if participants also show the primacy effect, it will mean that they have been able to use resources efficiently, updating information and rehearsing the most recently updated list. This will be evidence in favor of an active strategy. According to this scenario, participants are expected to perform effective updating, characterized by actively discarding and repositioning items, and relying on a rehearsal mechanism to refresh the target list. Considering negative responses (i.e.; responses to items in the list, which are no longer among those items to be remembered), effects were expected according to the strategies adopted. With more effective updating, responses to items near the target list onset are expected to be confused with targets more often than earlier responses, thus producing lower rate of correct rejections and RTs. This effect would depend on inhibitory processes (e.g., Hasher & Zacks, 1988; Jonides, Smith, Marshuetz, Koeppe, & Reuter-Lorenz, 1998) thought to exert stronger influence on the items discarded earlier: this would be possible only if effective updating took place. However with a passive wait until the list end, we expect participants to produce no difference in correct rejections between early and late negative items, since these items have been processed weakly. According to the TD model, the distinctiveness of the first items of a list should be lower than that of the following items, leading to worse performance (higher RTs, lower correct rejections). 2. Method 2.1. Participants Two groups of thirty-two (males: 4) and thirty-six (males: 7) students at the University of Pavia (aged 21–35), were tested. All participants were native Italian speakers, right-handed, without reading,

125

comprehension or memory disorders, and had normal (or corrected to normal) visual acuity. Students received course credit in return for their participation in the study. 2.2. Materials and procedure Participants were presented with a WM updating task comprising lists of words (see for details: Basso et al., 2010). Within each list, participants had to remember a certain number of items, dependent on the condition: in the low WM load, they were instructed to remember the last 3 words of the list, while in the high WM load this was the last 5 words. In order to keep the number of updating constant between conditions, list length for each WM condition varied from 3 to 6 words (low WM condition), and from 5 to 8 words (high WM condition), producing a potential range of 0 to 3 updating steps. WM conditions were separated into 2 blocks, with 8 lists for each of the 4 list lengths, and a total of 32 lists in each block. Stimuli were presented centrally on a PC monitor, in light gray on a black background, sized 40 point font on a screen with 800 × 600 screen resolution. Each word within a list was presented on screen for 700 ms, at a rate of one word every 3 s. A cue-signal, represented by a cross (presented for 1 s), indicated the end of the list and, after 1 s of blank screen, a word (the probe) appeared and remained on the screen until response (with a maximum of 4 s to provide a response). Participants were asked to judge whether the probe presented was an item to be remembered or not, by pressing one of two different keys (“yes” or “no”) on the computer keyboard. Positive responses (“yes”) were probes presented/updated in the target list only; negative responses (“no”) could be either a) an item that was in the list, but no longer belonged to the target subset, or b) a word that was not presented before (i.e. a new item). Since the latter were considered catch trials, only the former kind of negative probe were analyzed. The words used in the present tasks were all concrete trisyllabic words, controlled for frequency (high frequency, accordingly to the Italian norms: De Mauro, Mancini, Vedovelli, & Voghera, 1993), length (6, 7 or 8 letters, equally balanced between the three categories) and rate of imagery (medium/high). Participants were tested via a within subject design, in which block presentation was randomized and fully counterbalanced between participants. The experiment was created and administered using the software Presentation (Neurobehavioral Systems, San Francisco, CA). This same UWM task was used for both experiments. However, in the second, participants undertook an additional condition with a PM task. Three “prospective” words were presented before each block; participants were told that, whenever they noticed one of these in the block lists, they had to press the space bar. Participants were free to study the three words as long as they wanted to before starting the task. Thereafter, prospective words appeared 8 times within the word lists overall, at the rate of about one every 2 min. Prospective words were never in a target position and were changed between the two blocks. 2.2.1. Data analysis The data of the two experiments were combined in a mixed analysis, which examined positive and negative responses separately. The analyses considered the presence of the PM task as a between subject factor (i.e. PM load), and two within subject factors: probe position and WM load. Probe positions were created by establishing an imaginary zeropoint between the end of the pre-target list and the beginning of the target positions, thus attributing negative values to positions before, and positive values for positions after the zero-point. Therefore, negative positions (ranging from − 1 to − 3) represented items that were updated and no longer in the target subset. Positive positions (ranging from +1 to +3 in the low WM condition, and from +1 to +5 in the high WM condition) represented items that actually belonged to the target subset.

126

M. Botto et al. / Acta Psychologica 148 (2014) 123–129

For the negative positions, two levels of probe position were created (−1 and −2), by collapsing position −2 and −3 into a single position (−2). This choice was made to balance the probability of occurrence of the probe position in the whole task; as the lists were of different lengths, positions did not have the same frequency. It should be noted here that negative responses were not errors, but correctly given (i.e., “no” responses to probe in a pre-target position). In the positive positions, the number of targets varied according to WM load (3 vs 5). In order to obtain the same number of levels for comparison in both high and low WM load conditions, probe positions were combined in the high WM load condition. Thus, in 5 WM load condition, position +1 included the first target position only; position +2 included position +2 and +3; and position +3 included the two last position (+4 and +5). In summary, we considered the following factors: probe position for negative responses (2 levels: −2 and −1), probe position for positive responses (3 levels: + 1, + 2 and + 3), WM load (2 levels: 3 or 5 words to be maintained in the target list) and PM load (2 levels between subject: with or without PM task). A series of ANOVAs was conducted individually for negative and positive responses, with accuracy (measured as percentage of correct responses) and reaction times in milliseconds (RTs) as dependent variables. Post-hoc analyses were performed with Bonferroni tests (at an alpha level of 0.05), correcting for multiple comparisons. Partial eta-square values were reported as an index of effect size. 3. Results 3.1. Positive responses Results are shown in Fig. 1. 3.1.1. Accuracy A significant main effect was found for PM load, F(1,66) = 6.405; p = .014; η2 = .088, WM load, F(1,66) = 100.195; p b 001; η2 = .603, and probe position, F(2,132) = 20.060; p b .001; η2 = .233. The interaction between WM load and PM load was also significant, F(1,66) = 25.493; p b .001; η2 = .279, as well as the interaction between WM load and probe position F(2,132) = 10.171, p b .001, η2 = .134. This was also highlighted by a significant three way interaction, F(2,132) = 3.686, p = .028, η2 = .073. Post hoc comparisons run for probe position showed that, in high WM, there was better performance at the last position (i.e.: a recency effect) in both high, F(2,65) = 32.112 p b .001, η2 =.497, and low, F(2,65) = 5.033 p = .009, η2 = .134, PM conditions. In the low WM load, no significant differences were found in either PM load conditions. Moreover, the difference between PM

load conditions was significant in the high WM load condition only, at pos +1, F(1,66) = 5.096 p = .027 η2 = .072, and pos +2, F(1,66) = 16.124 p b .001 η2 = .196. No other effects were found (PM load by probe position interaction: F(2,132) = 2.245 p = .110 η2 = .033). 3.1.2. Reaction times A significant effect was found for WM load, F(1,66) = 42.836; p b .001; η2 = .394, and probe position, F(2,132) = 10.752; p b .001; η2 = .140, but no difference due to PM load, F(1,66) = .147 p = .703, η2 = .002. The interaction between WM load and probe position was significant, F(2,132) = 4.370, p = .015, η2 = .062, as well as between PM load and probe position, F(2,132) = 4.034, p = .020, η2 = .058. Here, the three way interaction was also significant, F(2,132) =11.195, p b .001, η2 = .145. Post hoc comparisons run for probe position in low WM load showed slower RTs at the middle position, with no PM demand, F(2,65) = 6.535, p = .003, η2 = .167, and faster RTs at the first position, with high PM demand, F(2,65) = 8.153, p = .001, η2 = .201. High WM load showed slower RTs in position +2 in comparison to position + 1, in the no PM demand condition, F(2,65) = 4.861, p = .011, η2 = .130, while RTs at position + 3 were faster than the other two positions in the high PM demand condition, F(2,65) = 11.518, p b .001, η2 = .262. No other effects were found (WM load by PM load interaction: F(1,66) = .071 p = .791, η2 = .001). 3.2. Negative responses A significant main effect of PM load was found on accuracy, F(1,66) = 14.403, p b .001, η2 = .179, with better performance in conditions without a PM task. A significant interaction between PM load and probe position was also found on accuracy, F(1,66) = 11.747, p = .001, η2 = .151. As seen in Table 1, post hoc comparisons showed no difference between positions in the high PM condition (p = .102, η2 = .040), but an advantage for position − 2 in the low one, F(1,66) = 9.917, p = .002, η2 = .131. No other effects were found on accuracy (all F values b 2.7; all p values N .1). The main effect of WM load was significant for RTs, F(1,66) = 4.536, p = .037, η2 = .074, with faster RTs in the low WM load condition (WM-3: 1254 ± 38.7 ms; WM-5: 1338 ± 47.1 ms), while no other significant effects were noted (all F values b 2.8; all p values N .098). 4. Discussion These results demonstrated that an updating WM task can be effectively manipulated to obtain recall performances varying in serial position curves. When participants were tested with a probe recognition task presented at a slow item presentation, the SPC pattern showed

Fig. 1. Positions (separated by levels of WM load and PM load), against percentage of accurate responses (panel a) and mean reaction times (panel b). Error bars show the standard error means.

M. Botto et al. / Acta Psychologica 148 (2014) 123–129 Table 1 Mean accuracy rate of negative responses for each WM load-probe position combinations (columns), by PM task demand (in rows). Standard error means are presented in brackets. WM-3

PM0 PM3

WM-5

Pos −2

Pos −1

Pos −2

Pos −1

88.54 (±3.6)% 70.83 (±3.4)%

76.46 (±4.9)% 70.83 (±4.7)%

86.25 (±3.9)% 58.33 (±3.7)%

77.60 (±4.5)% 68.61 (±4.2)%

primacy and recency effects dependent on memory resources availability. When executive resources were available (i.e.: no PM task condition), a flat SPC was obtained with low WM demands (i.e.: 3 items to be maintained) in the accuracy data. This result could be explained by several models, since this task could be managed within the individual memory span, and it is not possible to distinguish between an active or passive behavior. When increasing task demand, by asking people to encode 5 elements in WM, accuracy data showed a recency effect, but RTs showed SPC with both primacy and recency effects (consistently with Baddeley, 1986, 2000; Jahnke, 1963; McElree & Dosher, 1989). According to the TD model, a primacy effect would be possible only when the distinctiveness of the first items is high. However, intervals between items were kept constant, and no temporal markers, that could distinguish items, were present. It can be argued that the time between items was distinctive per se, but in this case, the likelihood of retrieval of the first to-beremembered item would have been the same as the second and third items. However, the speed of retrieval showed a significant difference, and that the initial to-be-remembered item was more recognizable than the second one. Moreover, the initial to-be-remembered item was in the first position of the list in only 25% of lists, while in the others, it was preceded by 1, 2, or 3 other items. Therefore, the TD hypothesis can explain only part of the present data, obtained through a running memory task. According to the literature on WM updating, a primacy effect would represent the signature of an active approach to performing the task (e.g., Bunting et al., 2006; Palladino & Jarrold, 2008). Therefore, a logical interpretation of these results is that an effective approach, with executive processes such as discarding/removal and repositioning must be at work, providing further support to the claims of Ecker et al. (2014). Their paper examined the components of an updating memory task, and demonstrated the independence of an active removal process, among the other executive processes involved in the task. In general, a decrease of performance (lower accuracy, slower RTs) is observed for high compared with low WM conditions. This pattern suggests that participants were faced with a task demanding an amount of resources beyond their capacity, and some information was not available in the moment of retrieval. To challenge the suggestion that the SPC with both primacy and recency effects could be explained on the basis of higher WM availability (i.e., compared with the flat SPC obtained with low task demands), we introduced an additional resource competing task. Such competing task involved PM, which is known to absorb both memory and executive resources (Basso et al., 2010). When WM load was low, the addition of the PM task produced a flat curve in accuracy, but relevant primacy and recency effects in RTs. When resources are subtracted, the residual ones must be strategically optimized. However, an active strategy, is feasible only when sufficient resources remain for fulfilling the task. Thus, since a primacy effect may be obtained by absorbing resources via either a WM or a PM load, these results corroborate those obtained in the high WM — no PM condition, supporting the interpretation of participants' adoption of an active task approach. When resources were taken up by the concurrent PM task and the WM demand was high, the primacy effect disappeared, replicating results from the existing SPC literature (e.g. Morris & Jones, 1990; Palladino & Jarrold, 2008; Ruiz et al., 2005). Both accuracy and RTs measures showed noticeable recency effects in the absence of primacy. This

127

result is consistent with the TD hypothesis of an isolation advantage, specific to the last positions. Another explanation for the absence of the primacy effect, contrasting the passive interpretation, could be the simple notion that under high task demands, less resources are used for those items (i.e., presented in the first positions) that are least likely to be tested. However, participants were not aware of list length. It is true that first items are likely to be updated and discarded, but these items were also those likely to be rehearsed by participants. There is similar probability for both cases (i.e., discarding or maintaining), for both high and low WM demands: lists may have between 0 and 3 negative items. Therefore, a higher probability of discarding the first element cannot account for the presence/absence of the primacy effect, since it varies, depending on other conditions, and is prevented by the uncertain length of the list. The result obtained with high WM load also fits with adoption of a passive approach. Participants had insufficient resources available, and may be constrained to adopting a recency-based strategy (e.g. Bunting et al., 2006; Palladino & Jarrold, 2008), such as waiting until the end of the list and than trying to recall the last items. When resource availability is low, this strategy may be the only possibility, since it is less resource-consuming than active processing, despite being more prone to errors. Target items in the first position were weakly encoded and generated conflict at moment of recall, thus producing an increase in time needed to decide if the item is among those to-be-recalled or not. Moreover, items in first and second positions alone showed a reduction in performance with respect to conditions of higher memory resource availability (whereas items at in last positions showed the same accuracy rate in both PM conditions). Therefore, we can explain these results based on the strategy employed by participants to perform the task. This account appears parsimonious and, hence, preferable. Negative responses may help in supporting this assertion. In experimental conditions where resources were available (i.e. no PM task), higher accuracy level for “pos − 2” items in comparison to “pos − 1” items was found, and this pattern is consistent with adoption of an active strategy. That is, earliest items in the list may have received selective rehearsal, and participants were more likely to actively remove them when they exceed the length of the target list. Conversely, in the high load condition, in which more errors occurred in the stored target list, there was no effect for negative responses. This may suggest difficulty in interference control, and potentially, less clear identificationactivation of relevant/irrelevant items. Both these possibilities are related to an ineffective updating process, and could be consistent with TD and a passive strategy interpretation, since the removal process is resource-demanding. Similarly to item storage and rehearsal processes, the more resources are employed to maintain information, the less effective the removal of no-longer useful information (Lewandowsky, Oberauer, & Brown, 2009). Taken together, the pattern of results obtained with introduction of the PM task confirmed that participants' approach to the task depended on both task demands and memory resource availability. A primacy effect in RTs was obtained with low WM load, even though a PM task was absorbing executive resources. Therefore, the presence of primacy effects with low WM task demands, and their absence under high WM task demands, could not be attributed to the PM task only, nor do they depend on recognition procedure. It is more likely that this result was obtained by a reduced amount of available memory resources with high WM demands (showed by the general increase of RTs), rather than by interference due to the additional PM-related memory items. Since this study is among the first in which primacy effects have emerged from a running memory task (see also Postle, 2003), methodological comment is merited. Our probe recognition procedure represents an important difference compared with the recall used in previous studies (e.g., Morris & Jones, 1990). As noted by Oberauer (2003), recall and recognition differ in at least three aspects: retrieval demands (i.e., higher in recall than in recognition); number of retrieval events (i.e., all list items in recall and only a single retrieval event in

128

M. Botto et al. / Acta Psychologica 148 (2014) 123–129

recognition) and the kind of information being retrieved (i.e., only which items were in the list in recognition, but this, and also order information, in recall). In our version of the updating task, the absolute position of the item is not important per se, but acts as an index of identification of the item in one of the two categories (i.e.: included on the last n items, or not). Therefore, the position of the word in the list is only useful in the encoding phase, while in the retrieval phase, the test is likely to be made only on the basis of a categorical choice. Moreover, the single response should be given to a probe, and this may decrease the complexity of the response and the probability of producing errors. From this perspective, recognition seems to require management of less item information, and therefore, to be a less demanding procedure that may have promoted the primacy effect. However, Bunting et al. (2006) also used a “probed recall” procedure, in which the context (i.e., the non-target part of the list) was cued and participants were required to recall the target item. Nevertheless, they also failed to find a primacy effect, in which case, the recognition procedure itself may not be the only factor involved. As suggested by the TD hypothesis, the more items in a list are temporally distant from each other, the more they are isolated in memory. The use of a slow pace, increasing the temporal distance between items, could have increased their distinctiveness, and the first words may have received a benefit such that their likelihood of retrieval increased. However, Jarrold suggested (personal communication) an alternative explanation: maintaining information in working memory might have “protected” items from interference, thereby affecting temporal distinctiveness (TD) effects (Bunting et al., 2006; Jarrold, Tam, Baddeley, & Harvey, 2010; Unsworth, Heitz, & Parks, 2008). Under this view, one might expect that TD previsions could be applied only to tasks that do not require updating. Our pattern of results seems to support this view and, since the TD account could explain the recency effect, it is likely that both item's distinctiveness and task demands should be considered among the factors involved in the running memory performance. Therefore, an interaction between TD accounts and executive processing may represent a most suitable model. A reconsideration of the SIMPLE model by Brown, Chater, and Neath (2008) is in line with these suggestions, since the authors acknowledged TD was not sufficient to explain performances in serial and free recall, and stated that “rehearsal must be included in the model” (p. 782). Which sexecutive processes are also needed, in order to provide a description of the primacy effect too? The mechanism advocated for primacy effects is likely to encompass both rehearsal (Tan & Ward, 2008) and additional executive-based processes, such as top-down attention (Sederberg et al., 2008) and/or removal process (Ecker et al., 2014; see also Kessler & Meiran, 2008; Artuso & Palladino, 2011, 2013). These executive-based processes are likely to (at least, partially) correspond to those required for the PM task; it follows that running memory task and PM task were in competition for the available resources. In conclusion, our study was able to extend existing knowledge on working memory updating and its measurement. Recent research on WM updating has raised insidious questions such as: ‘do updating tasks require updating?’ (Palladino & Jarrold, 2008). However, these data support the idea that updating task performance is sensitive to both task demands and memory resource availability. In fact, we have demonstrated that participants are likely to update incoming information when task demands and memory resource availability allow them to do this. Moreover, the probe recognition procedure appears particularly suitable for highlighting both serial position and WM load effects in updating tasks Acknowledgments This research was funded by grants from the Italian PRIN (2008YFTC3C_003) to P.P. and by the Free University of Bozen (BW5093) to D.B.

References Artuso, C., & Palladino, P. (2011). Content–context binding in verbal working memory updating: On-line and off-line effects. Acta Psychologica, 136, 363–369. http://dx.doi.org/10.1016/j.actpsy.2011.01.001. Artuso, C., & Palladino, P. (2013). Binding and content updating in working memory tasks. British Journal of Psychology. http://dx.doi.org/10.1111/bjop.12024. Baddeley, A.D. (1986). Working memory. Oxford, UK: Oxford University Press. Baddeley, A.D. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 4(11), 417–423. http://dx.doi.org/10.1016/S13646613(00)01538-2. Basso, D., Ferrari, M., & Palladino, P. (2010). Prospective memory and working memory: Asymmetrical effects during TMS frontal lobe stimulation. Neuropsychologia, 49, 3282–3290. http://dx.doi.org/10.1016/j.neuropsychologia.2010.07.011. Bisiacchi, P.S., Cona, G., Schiff, S., & Basso, D. (2011). Modulation of a fronto-parietal network in event-based prospective memory: An rTMS study. Neuropsychologia, 49(8), 2225–2232. http://dx.doi.org/10.1016/j.neuropsychologia.2011.05.007. Brown, G. D. A., Chater, N., & Neath, I. (2008). Serial and free recall: Common effects and common mechanisms? A reply to Murdock (2008). Psychological Review, 115(3), 781–785. http://dx.doi.org/10.1037/a0012563. Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory. Psychological Review, 114(3), 539–576. http://dx.doi.org/10.1037/0033-295X.114.3.539. Brown, G. D. A., Vousden, J. I., McCormack, T., & Hulme, C. (1999). The development of memory for serial order: A temporal–contextual distinctiveness model. International Journal of Psychology, 34, 389–402. http://dx.doi.org/ 10.1080/002075999399747. Bunting, M., Cowan, N., & Saults, S. J. (2006). How does running memory span work? The Quarterly Journal of Experimental Psychology, 59(10), 1691–1700. http://dx.doi.org/10. 1080/17470210600848402. Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–185. http://dx.doi.org/10.1017/S0140525X01593929. Cowan, N., Elliott, E. M., Saults, S., Morey, C. C., Mattox, S., Hismjatullina, A., et al. (2005). On the capacity of attention: Its estimation and its role in working memory and cognitive aptitudes. Cognitive Psychology, 51, 42–100. http://dx.doi.org/10.1016/j. cogpsych.2004.12.001. D'Esposito, M., & Postle, B. R. (1999). The dependence of span and delayedresponse performance on prefrontal cortex. Neuropsychologia, 37, 1303–1315. http://dx.doi.org/10.1016/S0028-3932(99)00021-4. D'Esposito, M., & Postle, B. R. (2000). Neural correlates of processes contributing to working memory function: Evidence from neuropsychological and pharmacological studies. In S. Monsell, & J. Driver (Eds.), Control of cognitive processes: Attention and performance XVIII (pp. 579–602). Cambridge, MA: MIT Press. De Mauro, T., Mancini, F., Vedovelli, M., & Voghera, M. (1993). Lessico di frequenza dell'italiano parlato (Lexical frequency dictionary of spoken Italian). Roma: Etas Libri. Ecker, U. K. H., Lewandowsky, S., & Oberauer, K. (2014). Removal of information from working memory: A specific updating process. Journal of Memory and Language. http://dx.doi.org/10.1016/j.jml.2013.09.003 (in press). Fiore, F., Borella, E., Mammarella, I. C., & De Beni, R. (2011). Age differences in verbal and visuo-spatial working memory updating: Evidence from analysis of serial position curves. Memory, 20, 14–27. http://dx.doi.org/10.1080/09658211.2011.628320. Friedman, N., Miyake, A., Corley, R. P., Young, S. E., De Fries, J. C., & Hewitt, J. K. (2006). Not all executive function are related to intelligence. Psychological Science, 17, 172–179. http://dx.doi.org/10.1111/j.1467-9280.2006.01681.x. Geiger, S. M., & Lewandowsky, S. (2008). Temporal isolation does not facilitate forward serial recall — Or does it? Memory & Cognition, 36(5), 957–967. http://dx.doi.org/ 10.3758/MC.36.5.957. Guynn, M. J. (2003). A two-process model of strategic monitoring in event-based prospective memory: Activation/retrieval mode and checking. International Journal of Psycology, 38(4), 245–256. http://dx.doi.org/10.1080/00207590344000178. Guynn, M. J. (2008). Theory of monitoring in prospective memory: Instantiating a retrieval mode and periodic target checking. In M. Kliegel, M. A. McDaniel, & G. O. Einstein (Eds.), Prospective memory: Cognitive, neuroscience, developmental, and applied perspectives (pp. 53–76). New York, NY: Taylor & Francis Group/Lawrence Erlbaum Associates. Hasher, L., & Zacks, R. T. (1988). Working memory, comprehension, and aging: A review and a new view. In G. H. Bower (Ed.), The psychology of learning and motivation, Vol. 22. (pp. 193–225). San Diego: Academic Press. Jahnke, J. C. (1963). Serial position effects in immediate serial recall. Journal of Verbal Learning and Verbal Behavior, 2, 284–287. http://dx.doi.org/10.1016/S0022-5371(63) 80095-X. Jarrold, C., Tam, H. H. Y., Baddeley, A.D., & Harvey, C. E. (2010). The nature and the position of processing determines why forgetting occurs in working memory tasks. Psychonomic Bulletin and Review, 17, 772–777. http://dx.doi.org/10.3758/PBR.17.6.772. Jonides, J., Smith, E. E., Marshuetz, C., Koeppe, R. A., & Reuter-Lorenz, P. A. (1998). Inhibition in verbal working memory revealed by brain activation. Proceedings of the National Academy of Sciences, 95, 8410–8413. Kessler, Y., & Meiran, N. (2008). Two dissociable updating processes in working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34, 1339–1348. http://dx.doi.org/10.1037/a0013078. Kliegel, M., Martin, M., McDaniel, M.A., & Einstein, G. O. (2002). Complex prospective memory and executive control of working memory: A process model. Psychologische Beiträge, 44, 303–318. http://dx.doi.org/10.1037/ 0012-1649.44.2.612. Lewandowsky, S., Oberauer, K., & Brown, G. D. A. (2009). No temporal decay in verbal short-term memory. Trends in Cognitive Sciences, 13(3), 120–126. http://dx.doi.org/10. 1016/j.tics.2008.12.003.

M. Botto et al. / Acta Psychologica 148 (2014) 123–129 Mäntylä, T. (2003). Assessing absent mindedness: Prospective memory complaint and impairment in middle-aged adults. Memory & Cognition, 31(1), 15–25. http://dx.doi.org/ 10.3758/BF03196078. McElree, B., & Dosher, B.A. (1989). Serial position and set size in short-term memory: The time course of recognition. Journal of Experimental Psychology: General, 118(4), 346–373. http://dx.doi.org/10.1037/0096-3445.118.4.346. Miyake, A., Friedman, N.P., Emerson, M. J., Witzki, A. H., & Howerter, A. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100. http://dx.doi.org/ 10.1016/j.bbr.2011.03.031. Morris, N., & Jones, D.M. (1990). Memory updating in working memory: The role of central executive. British Journal of Psychology, 81, 111–121. http://dx.doi.org/10. 1111/j.2044-8295.1990.tb02349.x. Oberauer, K. (2003). Understanding serial position curves in short-term recognition and recall. Journal of Memory and Language, 49, 469–483. http://dx.doi.org/10. 1016/S0749-596-X. Okuda, J., Fujii, T., Yamadori, A., Kawashima, R., Tsukiura, T., Fukatsu, R., et al. (1998). Participation of the prefrontal cortex in prospective memory: Evidence from a PET study in humans. Neuroscience Letters, 253(2), 127–130. http://dx.doi.org/10. 1111/j.2044-8295.1990.tb02349.x. Palladino, P., Cornoldi, C., De Beni, R., & Pazzaglia, F. (2001). Working memory and updating processes in reading comprehension. Memory & Cognition, 29, 344–354. http://dx.doi.org/10.3758/BF03194929.

129

Palladino, P., & Jarrold, C. (2008). Do updating tasks involve updating? Evidence from comparisons with immediate serial recall. The Quarterly Journal of Experimental Psychology, 61, 392–399. http://dx.doi.org/10.1080/ 17470210701664989. Pollack, I., Johnson, L., & Knaff, P. (1959). Running memory span. Journal of Experimental Psychology, 57, 137–146. http://dx.doi.org/10.1037/h0046137. Postle, B. R. (2003). Context in verbal short-term memory. Memory & Cognition, 31(8), 1198–1270. http://dx.doi.org/10.3758/BF03195803. Ruiz, M., Elosùa, M. R., & Lechuga, M. T. (2005). Old-fashioned responses in an updating memory task. The Quarterly Journal of Experimental Psychology, 58A(5), 887–908. http://dx.doi.org/10.1080/02724980443000395. Sederberg, P. B., Howard, M. W., & Kahana, M. J. (2008). A context-based theory of recency and contiguity in free recall. Psychological Review, 115, 893–912. http://dx.doi.org/ 10.1037/a0013396. Tan, L., & Ward, G. (2000). A recency-based account of the primacy effect in free recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(6), 1589–1625. http://dx.doi.org/10.1037/0278-7393.26.6.1589. Tan, L., & Ward, G. (2008). Rehearsal in immediate serial recall.Psychonomic. Bulletin and Review, 15(3), 535–542. http://dx.doi.org/10.3758/PBR.15.3.535. Unsworth, N., Heitz, R. P., & Parks, N. A. (2008). The importance of temporal distinctiveness for forgetting over the short-term. Psychological Science, 19, 1078–1081. http://dx.doi.org/10.1111/j.1467-9280.2008.02203.x.