WORKING memory is the workspace of the mind used

Transcription

1 Basak, C., & Verhaeghen, P. (2011). Aging and switching the focus of attention in working memory: age differences in item availability but not in item accessibility. Journal of Gerontology: Psychological Sciences, /geronb/gbr028 Aging and Switching the Focus of Attention in Working Memory: Age Differences in Item Availability But Not in Item Accessibility Chandramallika Basak 1 and Paul Verhaeghen 2 1 Department of Psychology, Rice University, Houston, Texas. 2 School of Psychology, Georgia Institute of Technology, Atlanta. Objectives. To investigate age differences in working memory processing, specifically the accuracy of retrieval of items stored outside the immediate focus of attention. Methods. Younger and older adults were tested on a modified N-Back task with probes presented in an unpredictable order (implying also that some trials necessitated a switch in the focus of attention and others that did not). Results. Older adults showed intact item accessibility, that is, after taking general slowing into account, older adults were as fast as younger adults in locating the item in working memory. We found age differences, however, in item availability: Older adults were less likely to correctly retrieve items stored outside the focus of attention. Smaller age differences in availability were also found for items stored inside the focus of attention. Discussion. These results strongly suggest that item availability is a cognitive primitive that is not reducible to more basic constructs such as item accessibility or simple speed of processing. Key Words: Cognition Executive function Memory Working memory. WORKING memory is the workspace of the mind used for stimulus manipulation and temporary storage. It is an essential system for cognitive processing, implicated in reasoning (e.g., Kyllonen & Christal, 1990), reading comprehension (e.g., Daneman & Carpenter, 1980), and general intelligence (e.g., Conway, Kane, & Engle, 2003). With advancing age, the capacity of the working memory system to hold and manipulate information declines. In a meta-analysis, Bopp and Verhaeghen (2005) noted a decline in all measures of short-term memory; the decline was appreciably larger for task of true working memory, that is, tasks where presentation of the to-be-remembered items were interleaved with bouts of processing. This age-related decline in working memory capacity has implications for higher level aspects of cognition: In cross-sectional research, working memory and/or shortterm memory span explains about 38% of the age-related variance in episodic memory, 24% of the age-related variance in spatial ability, and 34% of the age-related variance in reasoning (Verhaeghen & Salthouse, 1997). Some have speculated that working memory capacity may be the underlying cause for age differences in selected aspects of executive control, such as task switching and dual-task control (e.g., Verhaeghen & Cerella, 2002). The present study investigates one possible root cause for this decline in working memory capacity: A loss of item availability (indicated by a drop in accuracy) after an item has left the focus of attention. A crucial aspect of the working memory system is its severely limited capacity (e.g., Baddeley & Hitch, 1974). Current theories, most notably, for example, Cowan s (2001), posit a hierarchy within those limits: some items are more accessible than others. Cowan s model proposes a two-store structure for working memory, distinguishing a zone of immediate access, labeled the focus of attention, from a larger activated portion of long-term memory in which items are stored in a readily available but not immediately accessible state (which we have labeled the outer store, Verhaeghen, Cerella, & Basak, 2004). When the number of items to be retained in working memory is smaller than or equal to the capacity of the focus of attention, they will be contained within the focus. There they will be immediately retrievable, and access times will be fast. When the number of items to be retained exceeds the capacity of the focus, the excess items will be stored outside the focus of attention. In that case, accessing items for processing will necessitate a retrieval operation; this will slowdown response time. This is what we refer to as focus switching the process of shunting items in and out of the focus of attention when the focus gets overloaded. The effect has been noted in a number of paradigms: the N-Back task (McElree, 2001; Verhaeghen & Basak, 2005; Verhaeghen et al., 2004), a running count task in which participants keep multiple counters active (Basak & Verhaeghen, 2011; Garavan, 1998), and an arithmetic updating task (Oberauer, 2002). In each of these tasks, a step function in response time is observed when memory load exceeds the size of the focus of attention. The size of the focus in all those studies, it is important to note, is 1, indicating a severely size limitation in these kinds of serial attention tasks. (When the task allows for the parallel deployment of attention, the focus is closer to 4; Cowan, 2001.) The Author Published by Oxford University Press on behalf of The Gerontological Society of America. 1 All rights reserved. For permissions, please journals.permissions@oup.com. Received February 16, 2010; Accepted February 27, 2011 Decision Editor name: Robert West, PhD

2 2 BASAK AND VERHAEGHEN Much of our own work on focus switching and aging has used an N-Back task (Vaughan, Basak, Hartman, & Verhaeghen, 2008; Verhaeghen & Basak, 2005). Our version is self-paced. Participants press one of two keys to indicate whether the current stimulus matches the stimulus encountered N positions back; this key press then triggers the next stimulus. To help participants keep track of item positions, stimuli are presented on the computer screen in N virtual columns, one at a time, the first stimulus in column 1, the second in column 2, and so forth; after (multiples of) N stimuli have been presented, a new row starts at Column 1. What this means in practice is that the participant compares the current stimulus with the stimulus presented previously in the same column. Our paradigm thus allows us to measure the speed of access of an item (an item s accessibility; McElree, 2001) as well as the accuracy of retrieval (an item s availability) while minimizing the participant s effort to keep track of item positions. With regard to aging, our research with this paradigm has yielded three main results. First, we have consistently failed to find age differences in focus-switching costs in latency once age-related slowing in the comparison process itself is taken into account. The perhaps surprising conclusion is that older adults can access items in working memory as effectively as younger adults can: The dynamics of focus switching are perfectly preserved in old age. We found this to be true even for more difficult versions of the task for instance, when all items are displayed at the same location on the screen (Vaughan et al., 2008, Experiment 1) or when all stimuli are three-digit numbers rather than single digits (Vaughan et al., 2008, Experiment 2). The second finding concerns item availability after a focus switch. In all our work, a focus-switching requirement has led to an exacerbation of existing age differences in accuracy. Thus, once items leave the focus of attention, they become less available to older adults than to younger adults. We stress here that this result is truly remarkable: We find a clear age difference in retrieval accuracy in the absence of a specific age difference in retrieval dynamics a rare occurrence. We also stress that we have obtained the same result with a very different paradigm, namely continuous calculation (Verhaeghen & Hoyer, 2007). The third finding concerns the dynamics of search inside the outer store. In our N-Back experiments, this concerns the Reaction Time (RT) by N slope for values of N larger than 1 the time to retrieve an item once it has left the focus of attention as a function of the size of the memory set. In all our experiments, we have found this slope to be either zero or negligibly small, and we find little or no age differences in that slope. What this means is that both younger and older adults search a set of five items as fast as a 2-item set. This in term indicates that both age groups search the outer store with perfect efficiency, that is, without any effect of concurrent memory load. It is this latter finding we aim to explore further in the present study. The paradigms we have used in all studies described earlier (i.e., N-Back and continuous calculation) involve a predictable, sequential, feed-forward search: Items are retrieved in the order in which they were encoded. Recent work, however, has suggested that under such conditions, memory search might be special. Specifically, forward-order search is much faster than any other type of search, be it backward-order or random-order search, or memory access in a predictable prelearned pattern (Basak & Verhaeghen, 2011; Lange & Verhaeghen, 2009; Lange, Verhaeghen, & Cerella, 2010; Oberauer, 2003, 2006). The latter types of search also give rise to a reliable RT by set size slope, something forward-order search fails to do. One consequence of these findings is that the perfect efficiency we observed in both younger and older adults is most likely an artifact of the paradigm. If we break up the standard forward-order pattern of the N-Back task, we expect that complete efficiency will be lost, response time will increase with memory load, N, and age differences might appear. This consideration led us to design the present study one in which memory access no longer proceeds in a forward order but is randomized. The new task is essentially an N-Back task with random cueing of locations. Participants first learn a memory set of N items, presented at N locations on the screen; next, they are probed with N items, one at each of the N locations but in unpredictable order. The participant indicates whether the probe matches the item presented before in the same location and updates the representation associated with the particular location with the new item. Then, a next set of N probes is delivered and so on. The focus-switching requirement is now measured differently and more directly: When the location probed is also the location visited on the previous trial, no focus switching is required; when the location probed shifts from one trial to the next, switching of the focus is required (see Table 1 for an example). Predictions for the differences between switch and non switch trials are straightforward. If items held within the focus of attention are immediately accessible and available as stated by most working memory theoreticians, we expect responses for nonswitch trials to be fast and perfectly accurate and we expect no set size effects. From our previous work as well as Oberauer s (Basak & Verhaeghen, 2011; Lange & Verhaeghen, 2009; Lange et al., 2010; Oberauer, 2002, 2006), we expect that the focus-switch trials will yield slower and less accurate responses, with RT increasing and accuracy decreasing over set size. There is no theoretical a priori reason to expect that the type of search (random order or forward order) should influence the pattern of age effects observed previously; therefore, we fully expect to replicate our earlier work in the present study. That is, first, we expect no age differences in RT on nonswitch trials once age-related slowing as present in the N = 1 condition is partialled out (Vaughan et al., 2008; Verhaeghen & Basak, 2005;

3 AGE DIFFERENCES IN WORKING MEMORY PROCESSING 3 Table 1. Example of One 4-Back Sequence (20 stimuli) Stimulus Location 1 Location 2 Location 3 Location 4 Verhaeghen & Hoyer, 2007). Second, we expect older adults to show larger accuracy decrements than younger adults after a focus switch (Vaughan et al., 2008; Verhaeghen & Basak, 2005). Given that this is the first study on the topic, we have no hypotheses concerning the existence of age differences on focus-switch trials. Method Participants The sample consisted of 25 younger adults (mean age = 19 years, SD = 1.13; mean years of education = 13.20, SD = 3.53; 12 females) and 30 community-dwelling older adults (mean age = 72 years, SD = 9.3; mean years of education = 14.13, SD = 5.14; 23 females). Older adults performed significantly better on the Shipley vocabulary test than the younger adults, t(53) = 5.42; p <.001 (mean score = 34.17, SD = 2.18 and 29.92, SD = 2.58, respectively). All participants had at least 20/40 corrected vision. Procedure The baseline task is an identity-judgment task (McElree, 2001; Verhaeghen & Basak, 2005; we present an example of a 4-Back sequence in Table 1). Within each trial, a sequence of 20 digits was shown, one at a time, in N virtual columns on the screen (with N varying from 1 to 4) and 20/N rows. Each column was depicted in a different color to facilitate positional coding of the stimuli. For the first row, a new digit was presented every 2,000 ms, sequentially from left to right; from the second row on, participants pressed either of Answer Switch/nonswitch 4 None (Encode) 8 None (Encode) 7 None (Encode) 6 None (Encode) 5 Mismatch Nonswitch 8 Match Switch 4 Match Switch 3 Mismatch Switch 4 Mismatch Nonswitch 5 Match Switch 9 Mismatch Switch 4 Match Switch 3 Mismatch Nonswitch 9 Match Switch 5 Match Switch 6 Mismatch Switch 2 Mismatch Nonswitch 4 Mismatch Switch 1 Mismatch Switch 3 Match Switch Notes: Stimuli are shown in four vertical locations or columns on the screen, each colored differently. The first 4 stimuli are presented left to right, one at a time, for 2 s each; participants simply encode these. From the 5 th stimulus onwards, participants indicate whether or not the stimulus shown matches the one previously shown in the same vertical location. The key press indicating the participant s answer also advances the trial to the next stimulus. In ¼ of the trials, there is no location change; this is nonswitch trial. On ¾ of the trials, the location changes, this is a switch trial. two keys to indicate their answer whether or not the digit presented was identical to the digit most recently presented in the same colored column. The / key, masked with a piece of green tape, stood for yes (i.e., identical); the z key, masked with a piece of red tape, stood for no (i.e., different). Participants were instructed to be both fast and accurate. Presentation order from the second row on was randomized across columns within each row, that is, all items within a row were probed before stimulus presentation moved on to the next row. The within-row randomization only allowed for 1/N nonswitch trials, all other trials were by necessity switch trials. Likewise by necessity, each non switch item was the first item probed within each row (i.e., it always marked the transition from one row to the next). In the N = 1 condition, all items were by necessity nonswitch items. Response stimulus interval was zero. After each trial, the participant received feedback about both total accuracy and average RT of 20 items. A total of 11 trials were presented for each value of N, where N varied from 1 to 4, distributed as follows: first 6 trials each for N = 1, N = 2, N = 3, and N = 4 and then 5 trials each for N = 4, N = 3, N = 2, and N = 1. The first trial for each of the values of N in the first half of the experiment was considered practice and discarded from further analysis. Analysis All RT analyses were conducted on correct responses only. Within each condition within each individual, RT distributions were truncated at three interquartile ranges above or below the mean to remove outliers. The RTs of the

4 4 BASAK AND VERHAEGHEN second row, when the comparison items were the items initially presented by the experimenter, were likewise discarded (Verhaeghen & Basak, 2005; Verhaeghen et al., 2004). To control for overall age-related slowing effects, all analyses of variance (ANOVAs) on RT were repeated using individual z-transformed RTs (Faust, Balota, Spieler, & Ferraro, 1999). In this analysis, within each individual, we took each condition s mean RT, subtracted the individual s overall mean for all conditions, and divided this result by the standard deviation of the individual s condition means; these z-scores are then entered in the ANOVAs. This z- transformation rescales the differences between conditions relative to each individual s performance and eliminates mean differences in RT between individuals, including agerelated differences. The z-transformed results are reported here only if the results deviated from those obtained on raw RT. Alpha level for all statistical testing was set at.05. Results Response Times The results for RT are graphically presented in Figure 1a. The main analysis of variance tested for the effects of age group (younger vs. older), switch type (switch vs. non switch), and set size (2 4; set size 1 was not included because it by definition only yielded nonswitch trials). Older adults were slower, on average, than younger adults, F(1, 53) = 32.69, p <.001, partial h 2 =.38. Response times increased with set size, F(2, 106) = 8.27, p <.01, partial h 2 =.13. Focus-switch trials were slower than nonswitch trials, F(1, 53) = , p <.001, partial h 2 =.70. These main effects were modulated by a significant set size by switch type interaction, F(2, 106) = 19.48, p <.001, partial h 2 =.27 (we will unpack this interaction below), as well as by a significant age group by switch type interaction, F(1, 53) = 13.44, p <.001 (age differences are larger on switch trials than on nonswitch trials). The age by switch type interaction, however, did not survive the z-score transformation, F(1, 53) = 1.89, p =.18, partial h 2 =.03. The other interactions set size by age group and the three-way interaction, Fs < 1 did not reach significance. To follow up on the set size by switch type interaction, we analyzed switch and nonswitch trials separately within the ranges where these two trial types overlap, that is, for N between 2 and 4. For nonswitch trials, we found a significant effect of age group, F(1, 53) = 32.24, p <.001, partial h 2 =.38 older adults are unsurprisingly slower than younger adults. Neither the set size main effect nor the age group by set size interaction were significant, Fs < 1, partial h 2 s =.01. Thus, RT in nonswitch trials remains flat for N > 1, and this is true for both younger and older adults. For switch trials, we found a main effect of age group, F(1, 53) = 29.86, p <.001, partial h 2 =.36 older adults are slower and set size, F(2, 106) = 17.66, p <.001, partial h 2 =.25 RT (a) Average RT (b) Accuracy Y oung:non-switch Y oung:switch Old:Non-Switch Old:Switch increases with set size. The age group by set size interaction was not significant, F < 1, partial h 2 =.01, indicating that older adults showed the same set size effect as younger adults. We conducted one additional analysis on the switch trials. As stated in the introduction, an RT by set size slope suggests that some form of search process is taking place. Our within-row manipulation allows for an additional check on this conjecture. Within each row, the size of the set to be searched decreases with probe order (e.g., in the 4-Back version, after the first nonswitch item, the set to be searched consists of three items; after the first switch probe is responded to, only two items would need to be searched and so on). We would therefore expect that RTs decrease over N N Y oung:non-switch Y oung:switch Old:Non-Switch Old:Switch Figure 1. Response time (for correct responses only; a) and accuracy (b) for switch and nonswitch trials as a function of N for both age groups in the randomized (within-row) N-Back task. The error bars depict the SE.

5 AGE DIFFERENCES IN WORKING MEMORY PROCESSING 5 (a) 2100 Response Time (b) Accuracy Young: N=2 Young: N=3 Young: N=4 Old: N=2 Old: N=3 Old: N= Order of item presentation Young: N=2 Young: N=3 Young: N=4 Old: N=2 Old: N=3 Old: N= Order of item presentation Figure 2. Average response time (for correct responses only; a) and accuracy (b) of switch trials for each value of N > 1 as a function of order of item presentation within a row for both age groups. The error bars depict the SE. order of item presentation within a row. The data are presented in Figure 2a. RT indeed decreased reliably as a function of presentation order in both N = 3, F(1, 53) = 23.08, p <.001, partial h 2 =.30 and N = 4, F(2, 106) = 3.36, p =.039, partial h 2 =.28. Older adults are slower than younger adults, N = 3: F(1, 53) = 27.77, p <.001, partial h 2 =.34; N = 4: F(1, 53) = 23.19, p <.001, partial h 2 =.30, but no significant age group by presentation order interaction emerged for either N = 3 or N = 4, Fs < 1. This speedup is not due to a speed-accuracy trade-off: Accuracy does not significantly vary as a function of presentation order for any N, Fs < 1, and neither is the interaction between age and set size significant, N = 3: F < 1; N = 4: F(2, 106) = 2.53, p =.08, partial h 2 =.05 (see Figure 2b). Older adults are, on average, less accurate than younger adults as evidenced from the main effect of group, N = 3: F(1, 53) = 13.85, p <.001, partial h 2 =.21; N = 4: F(1, 53) = 16.91, p <.001, partial h 2 =.24. We do note that this search process is not perfectly efficient. If it were, only the number of remaining items (i.e., the number of items that have not been probed) would determine the speed of retrieval. This is not the case. Response times on switch trials were submitted to two separate ANOVAs, one for the case where the number of remaining items is 2 (these data can be culled from N = 4 and N = 3 conditions) and another where the number of remaining items is 1 (these data can be culled from N = 4, N = 3, and N = 2). RTs differed for the three conditions when only one item remained, F(2, 106) = 23.44, p <.001, partial h 2 =.31. Repeated contrasts show that this difference is significant between N = 3 and N = 4, F(1, 53) = 18.39, p <.001, partial h 2 =.26, as well as between N = 2 and N = 3, F(1, 53) = 14.63, p <.001, partial h 2 =.22, although the difference is greater as set size increases. RTs also differed for the two conditions when two items remained [between N = 3 and N = 4, F(1, 53) = 24.56, p <.001, partial h 2 =.32]. The effect sizes for the above-mentioned analyses are large. Order of item presentation did not interact with age, Fs < 1. Thus, even when the number of not-yet-searched items is held constant, RT is monotonic with N, that is, the total working memory load. Finally, from Figure 1a, it can be surmised that the nonswitch RT is faster for N = 1 than for other values of N. Pairwise comparisons showed that RT at N = 1 was significantly different from nonswitch RT at N = 2, F(1, 53) = , p <.001, partial h 2 =.67; but nonswitch RTs for N = 2 and N = 3 were not different from nonswitch RTs at N = 3 and N = 4, respectively, Fs < 1, partial h 2 <.01. Accuracy Older adults were less accurate, on average, than younger adults, F(1, 53) = 17.79, p <.001, partial h 2 =.25 (set size 2 4; Figure 1b). Accuracy declined with set size, F(2, 106) = 31.50, p <.001, partial h 2 =.37, and focus-switch trials yielded lower accuracy than nonswitch trials, F(1, 53) = , p <.001, partial h 2 =.87. These main effects were modulated by a significant set size by switch type interaction, F(2, 106) = 27.88, p <.001, partial h 2 =.34, as well as by a significant age group by switch type interaction, F(1, 53) = 9.05, p =.004, partial h 2 =.15; age differences are larger on switch trials than on non-switch trials. Neither of the other interactions reached significance, Fs < 1, largest partial h 2 =.02. To follow-up on the set size by switch type interaction, we analyzed switch and nonswitch trials separately, restricting the analyses to the N = 2 4 range. For nonswitch trials, older adults were less accurate, F(1, 53) = 6.75, p =.012, partial h 2 =.11. Neither the effect of set size nor its interaction with age group was significant, Fs < 1, largest partial h 2 =.02. This result was replicated when the full range of set sizes (1 4) was considered: The main effects of age group, F(1, 53) = 6.65, p =.013, partial h 2 =.11, and set size, F(3, 159) = 4.46, p =.004, partial h 2 =.08, were significant, but their interaction, F(3, 159) = 2.5, p =.061, was

6 6 BASAK AND VERHAEGHEN only marginally significant, with a small effect size, partial h 2 =.04. For switch trials, we obtained a significant main effect of age group, F(1, 53) = 19.24, p <.001, partial h 2 =.27 (older adults were less accurate) and a significant effect of set size, F(2, 106) = 22.16, p <.001, partial h 2 =.51 (accuracy decreased with set size), but the age group by set size interaction was not significant, F < 1, partial h 2 =.00. Discussion This experiment used an N-Back-like task with a randomized probe order to investigate age-related differences in item accessibility (i.e., how fast an item can be accessed and retrieved) and item availability (i.e., the accuracy of retrieval) after a focus switch. Our own previous work with a predictable forward-probing N-Back task has shown that (a) focus-switch costs are incurred in both RT and accuracy and (b) that older adults incur a larger cost than younger adults in accuracy but not RT (Vaughan et al., 2008; Verhaeghen & Basak, 2005; Verhaeghen et al., 2004). Additionally, our previous work offers evidence for perfect or near-perfect search efficiency in both younger and older adults as evidenced by a flat RT-by-N slope for N > 1, indicating that the speed of retrieval does not vary with working memory load. The conclusion is that after correcting for general slowing, older adults are just as fast at accessing the location or slot where an item resides as younger adults but that the item residing in that location or slot is less likely to still be available for processing an age difference in item availability predicated on age invariance in item accessibility. The present experiment was designed to test the limits of this proposition by using a random probe order. Probing memory in a random order is not just more challenging, it also induces a qualitative shift in processing as shown by Oberauer (2006) and Lange and colleagues (Lange & Verhaeghen, 2009; Lange et al., 2010, in press): In the randomized version, RT for switch trials clearly increases with set size, indicating that participants engage in a true search process. General Cognitive Findings With regard to general cognitive findings, we replicated three key results from Oberauer (2006). First, for set sizes larger than 2, we found response times for nonswitch trials to be fast (about 865 ms for younger and 1,276 ms for older adults) and constant over set size, coupled with very high (97% for younger adults and 94% for older adults) and constant accuracy. These results suggest a privileged status for the single item stored within the focus of attention: It is maintained in an immediately accessible state, at high accuracy, and suffers no interference from the items stored in the outer store (see also Verhaeghen et al., 2004). Second, we observed a substantial increase in nonswitch response times from N = 1 to N = 2, suggesting that processing within the focus of attention is less efficient when the outer store is occupied with a least a single item than when it is empty. Given that RT in nonswitch trials remained constant over set sizes 2 4, it is highly unlikely that this increased RT is due to working memory load per se. Rather, it is more likely that simultaneously maintaining and scheduling two (or more than two) active representations in working memory is responsible for the time cost. Third, the focus switch cost incurred in RT was not constant but increased as a function of working memory load. This lends further support to the hypothesis that unpredictability of probe order induces some form of search through the outer store, leading to a monotonic linear relationship between RT and set size for switch items. (We remain agnostic as to exactly what type of search process is engendered.) The search hypothesis was explored in more detail by looking at the effects of probe order within each row. The data show indeed that response times decreased when the effective search set decreased in size. That data also show, however, that this search mechanism was not completely efficient: Even when the number of items that remained to be searched was identical, latency increased with set size. One likely reason is that while searching, participants must also keep the other elements in working memory active; this might sap resources that evidence themselves in an RT cost. Alternatively, this set size effect might be due to inefficient backward inhibition: Part of the total search set is still active enough to interfere with the search processes applied to the remaining items. Finally, the search effect was also evident in accuracy: Accuracy decreases with set size for switch trials, suggesting that items in the outer store compete with each other for activation although, as stated earlier, they do not affect the activation of items held inside the focus of attention. Age-Related Findings Of particular interest in the present study were age differences in item accessibility or availability as exposed by the focus-switching process. The new and more challenging task notwithstanding, we replicated the main results obtained previously with our forward-order N-Back task. First, once age-related slowing was taken into account, we found no evidence for age differences in item accessibility (the focus-switch cost incurred in RT), suggesting that older adults have no specific deficit in accessing the slot where an item representation is stored. Second, we did find evidence for age differences in item availability: Existing age differences in accuracy were exacerbated after a focus switch. The new paradigm additionally allowed us to unconfound age-related differences in processing within the focus of attention from set size effects. In our previous studies, nonswitch trials always occurred for N = 1 and switch trials always for N > 1; with the random probe order, we have

7 AGE DIFFERENCES IN WORKING MEMORY PROCESSING 7 both switch and nonswitch trials available for N > 1. We found, first, that older adults are as efficient as younger adults in processing items inside the focus of attention: RT by set size slopes for nonswitch items was flat in both groups. The picture is different for accuracy, however: Even for nonswitch trials, when items are stored inside the focus of attention, accuracy is significantly lower for older adults than for younger adults. The reason for this age-related decline in accuracy is as yet unclear. It is not due to working memory load per se: Over set sizes 2 4, accuracy does not decline significantly for nonswitch trials. Maybe older adults experience more interference between items held in the focus of attention and those held in the outer store. It is also possible that the requirement to simultaneously store and process items makes older adults less effective even when the item is currently in focus (much like dual-task requirements and dual-task set maintenance increase existing age differences; Verhaeghen & Cerella, 2002). The results for items held outside the focus of attention replicated those of our previous studies: Older adults retrieve items held outside the focus of attention with lower accuracy than younger adults do, but there is no age difference in the speed of search processes, as evidenced by the lack of a significant interaction between age and RT over set sizes 2 4 once agerelated baseline slowing was partialled out. To summarize, it appears, then, that the results obtained with the present paradigm converge with the results from our previous studies with similar and dissimilar paradigms: Once items leave the focus of attention, older adults are as good as younger adults in locating the item in working memory, but once the location of the item is accessed, it is less likely to be retrieved. Thus, item availability after a focus switch is dissociable from item accessibility the former is age-related, the latter is not. This in turn might imply that item availability is a cognitive primitive, a basic aspect of the working memory system that is not reducible to another more basic construct (see Verhaeghen, in press, for a more through discussion of item availability as a primitive). More research is needed before this conclusion can be reached with certainty. It also remains to be seen if age differences in item availability will be successful in explaining age-related variance in everyday memory tasks that appear to implicate some form of focus switching, such as picking up the thread of a conversation after an interruption, remembering the referent of a pronoun, or forgetting to follow through on intended but delayed actions. Funding This research was supported by grants from the National Institute on Aging of the National Institutes of Health (RO1 AG16201) to P. Verhaeghen and a Beckman Institute Postdoctoral Fellowship to C. Basak. Acknowledgments The authors gratefully thank Korena Onyper for assistance with data collection and scoring and John Cerella for discussion on an earlier draft of the manuscript. Correspondence Correspondence should be addressed to Chandramallika Basak, PhD, Department of Psychology, Rice University, 472 Sewall Hall, MS-25, Houston, TX cbasak@rice.edu. References 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 ). New York: Academic Press. Basak, C., & Verhaeghen, P. (2011). Three layers of working memory: Focus-switch costs and retrieval dynamics as revealed by the N-count task. Journal of Cognitive Psychology, 23, doi: / Bopp, K. L., & Verhaeghen, P. (2005). Aging and verbal memory spans: A meta-analyses.. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 60, doi: /geronb/ 60.5.P223 Conway, A. R. A., Kane, M. J., & Engle, R. W. (2003). Working memory capacity and its relation to general intelligence. Trends in Cognitive Science, 7, doi: /j.tics Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24, doi: /s x Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, doi: /s (80) Faust, M. E., Balota, D. A., Spieler, D. H., & Ferraro, F. R. (1999). Individual differences in processing rate and amount: Implications for group differences in response latency. Psychological Bulletin, 125, doi: / Garavan, H. (1998). Serial attention within working memory. Memory and Cognition, 26, doi: /bf Kyllonen, P., & Christal, R. E. (1990). Reasoning ability is (little more than) working memory capacity. Intelligence, 14, doi: / S (05) Lange, E. B., Cerella, J., & Verhaeghen, P. (2011). Ease of access to list items in short-term memory depends on the order of the recognition probes. Eprint. doi: /a Lange, E. B., & Verhaeghen, P. (2009). No age differences in complex memory search: Older adults search as efficiently as younger adults. Psychology and Aging, 24, doi: / Lange, E. B., Verhaeghen, P., & Cerella, J. (2010). Dual representation of item positions in short-term memory: Evidence for two access modes. European Journal of Cognitive Psychology, 22, doi: /a McElree, B. (2001). Working memory and focal attention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27, doi: / Oberauer, K. (2002). Access to information in working memory: Exploring the focus of attention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, doi: / Oberauer, K. (2006). Is the focus of attention in working memory expanded through practice? Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, doi: / Vaughan, L., Basak, C., Hartman, M., & Verhaeghen, P. (2008). Aging and working memory inside and outside the focus of attention: Dissociations of availability and accessibility. Aging, Neuropsychology & Cognition, 15, doi: / Verhaeghen, P.. Working memory still working: Age-related differences in working memory and executive control. In N. Ohta & M. Naveh- Benjamin (Eds.), Memory and aging. Psychology Press in press. Verhaeghen, P., & Basak, C. (2005). Aging and switching of the focus of attention in working memory: Results from a modified N-Back task. Quarterly Journal of Experimental Psychology A, 58, doi: /

8 8 BASAK AND VERHAEGHEN Verhaeghen, P., & Cerella, J. (2002). Aging, executive control, and attention: A review of meta-analyses. Neuroscience and Biobehavioral Reviews, 26, doi: /s (02) Verhaeghen, P., Cerella, J., & Basak, C. (2004). A working-memory workout: How to expand the focus of serial attention from one to four items, in ten hours or less. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, doi: / Verhaeghen, P., & Hoyer, W. J. (2007). Aging, focus switching and task switching in a continuous calculation task: Evidence toward a new working memory control process. Aging, Neuropsychology, and Cognition, 14, doi: / Verhaeghen, P., & Salthouse, T. A. (1997). Meta-analyses of age-cognition relations in adulthood: Estimates of linear and non-linear age effects and structural models. Psychological Bulletin, 122, doi: /