Older adults show impaired working memory (WM), a function necessary for high-level cognition [1]. Working memory is theorized to be a multi-component, limited-capacity, information processing system comprising interrelated attention and memory subsystems that actively maintain and transform information for goal-relevant behavior over short periods of time [2,3]. Age-related performance deficits have been well-documented in tasks requiring a strong WM component such as task-switching [4], n-back performance [5], delayed match-to-sample [6], and a variety of complex tasks assessing WM capacity (e.g., [7,8,9,10]. Neuroimaging research indicates that cortical regions associated with WM [11,12,13] are also implicated in age-related differences in the activation of fronto-parietal networks [14,15,16,17,18,19,20,21,22] and in white matter degeneration in the frontal and temporal lobes [23,24]. In this study, we used electroencephalography to determine whether WM capacity mediates the effects of age on two different eventrelated potentials (ERPs) known to be sensitive to WM demands.
WM is characterized as comprising both storage and executive operations (e.g., [25]. The executive operations, in turn, comprise shifting attentional control between tasks, strategies, or contexts, updating the contents of storage, and inhibiting interference (e.g., [26]. Complex span tasks are currently thought to be one of the best ways to assess WM because they involve both types of operations [27]. They might be described as simply measuring WM storage capacity, but this characterization is incomplete. To perform these tasks, the individual must maintain strategies or sets of rules, one for the principal task (e.g., determining the result of an arithmetic operation) and one for the secondary task of storing the letter for later recall. Within each episode, it is necessary to shift from one rule set to the other. Further, the stored letter list must be updated by adding the new letter. In addition, the items in the stored list must be protected from interference by, e.g., phonetically similar letters. So, pure storage is embedded in a complex web of executive processing.
Two classic tasks used in ERP-eliciting paradigms also rely on WM functions that require both information manipulation and storage, but in different ways: the visual oddball task and the word-pair judgment task. Both of these tasks show age-related changes in neural response with age which may be associated with WM processing [28].
The visual oddball task might be characterized simply as a visual attention task, but it has also been shown to involve WM processes and, more specifically, context updating processes [21]. In the visual oddball task, as implemented in Kappenman et al. [29] and in this study, working memory is targeted as it involves the presentation of five letters, one of which is specified as the target for each of the five blocks. For each block, the target letter is different and must be placed in temporary storage while the target from the previous block is suppressed. On each trial, the target must be called up and compared to the present letter. There is interference because the target letter will have also served as a nontarget in separate trial blocks. This interference must be suppressed.
Similarly, the word-pair judgment task might be characterized as simply a semantic long-term memory task, but WM processes are also involved in its performance. On each trial, the semantic network of the first word must be activated and held in short-term storage. Then the network for the second work must be activated and overlap between the two networks assessed. After the response, the two semantic networks must be suppressed to prevent interference with the next pair. In sum, complex span tasks, the visual oddball task, and the word-pair judgment task all involve both the temporary storage and executive operation aspects of WM, though in different combinations and to different extents. The question is whether WM processes assessed by the complex span tests are also used in the visual oddball task as well as the word-pair judgment task.
Several studies using encephalography (EEG) and event-related potentials (ERPs) have demonstrated that WM demands influence posterior P3 (also known as the P300 or P3b) and N400 ERPs. The P3 is a large, positive ERP that is associated with the processing of relevant but unexpected information and that occurs 200-500 ms following stimulus onset in centroparietal regions [21,30,31]. Typically increased cognitive demands reduce P3 component size [32,33], making it plausible that P3 amplitudes may be modulated by individual differences in WM capacity. If P3 amplitudes reflect limits on cognitive resources available for current mental operations, then smaller P3 amplitudes may indicate fewer available WM resources [34,35]. Lefebvre et al. [36] proposed that the P3 is a potential neurophysiological marker for WM capacity. Indeed, Reed et al. [37] found that individual differences in WM performance influence the P3.
Further, tasks such as the n-back or visual oddball task that load on WM modulate the P3. In the n-back task participants must update the current item while holding the last n (e.g., 1 to 4) items in WM to determine whether the current item matches the item presented n items back. Dong et al. [38] found that the P3 decreased as n increased. Further, individuals high and low in WM capacity measured by a digit span task differed in the P3 amplitudes elicited by an n-back (1-4 item) task, regardless of the level of n-back difficulty. WM demands of a visual oddball task have also been found to affect P3. In the visual oddball task, a relatively rare target occurs within a series of relatively frequent non-targets and participants must hold the target in WM and respond only if the current stimulus matches the target [39,40]. Larger P3 amplitudes are observed for targets compared to non-targets [41][42][43][44]. Nittono et al. [45] assessed WM capacity using a reading span test [46] and elicited the P3 using a multiple-choice reaction time task. They found larger P3 amplitudes for high-span compared to low-span individuals, but only in the more demanding task conditions. Consistent with Polich [20] who proposed that the P3 plays a role in attentional resource allocation among concurrent operations, these findings suggest that those with greater WM capacity make more efficient use of neural resources that keep attention focused on task-relevant information.
The N400 ERP is also affected by semantic processing tasks that involve WM (e.g., [47,48]. The N400 amplitude is thought to be an index of the priming of terminal words for both sentences and word-pairs. It is largest for unrelated words and gets smaller as the association increases. Thus, prior words must be held in WM to compare them to current words. For example, N400 amplitudes are sensitive to the context in which a word occurs. When processing and attending to words in lists, the N400 amplitude varies with semantic relationships [49][50][51]. The presentation of a prior word that is associatively related or semantically related to a current word reduces the N400 amplitude. The N400, then, is thought to reflect contextual integration, or rather the lack of contextual fit, between the previous item held in WM and the current item or items [51-53,]. In a study comparing neural responses to sentences with different end wordings, Federmeier and Kutas [54] contrasted local, context-based relations (i.e., items expected or unexpected) with long-term memory organization (i.e., items within or between categories). They found that the immediate language context held in WM and the context-independent relationships between items extracted from long-term semantic memory both affected the neural processes reflected in the N400 simultaneously and independently.
How might age affect these ERPs that are associated with WM? Age-related differences in the amplitudes and latencies of the P3 and N400 ERPs have been attributed to differences in cognitive processing, especially context updating indicated by the P3 and semantic processing indicated by the N400 [55,56]. Older adults tend to produce smaller ERP amplitudes and delayed latencies relative to younger adults (e.g., [28]). Specifically, the P3 effect, reflecting differences to infrequent compared to frequent events, increases in latency and decreases in amplitude with increasing age [57]. The N400 effect, reflecting amplitude differences between semantically related and unrelated word pairs, decreases with age [58]. Denaro et al. [28] investigated the effects of age on the N170 (face processing), P3 (visual oddball), N400 (word-pair judgement), and error-related negativity (ERN, visual flanker task), and found the strongest age-related differences in ERPs from those tasks requiring WM.
There is some evidence of age-related changes in brain activity with variations in WM performance efficiency. Using EEG/ERP, Saliasi et al. [59] examined older and younger adults in an n-back task, with 0-and 1-back loads, and measured WM efficiency as the difference in response speed and accuracy between the two load conditions. P3 amplitudes varied with WM efficiency in both age groups, but age interacted with WM efficiency in terms of the P3 scalp distribution where higher efficiency was related to higher P3 amplitudes at parietal electrodes for younger adults but not older adults. To relate WM capacity to age and the N400, Gunter et al. [47] conducted a language recognition task in which sentences either had high or low WM requirements and participants varied in age (young vs. middle aged) and WM span performance (e.g., digit span). They reported an interaction between congruency and WM load which reflected smaller effects of congruency under greater WM load. Also, despite the overall decrease in WM and a smaller, delayed N400 observed for middle-aged adults, the ERP effects appeared to depend on WM span. Although young participants with high WM capacity showed similar congruency effects in both high and low WM-load conditions, those with low WM capacity showed congruency effects only in the low WM-load condition. Middle-aged participants with high WM capacity showed congruency effects only in the low WM-load condition and those with low WM capacity showed no congruency effects in either WM-load condition. Thus, it appears that WM capacity interacts with age for semantic processing indicated by the N400.
To summarize, demands on WM are reflected in ERPs. Our central goal in this study was to examine whether WM capacity mediated the effect of age on both P3 and N400 ERPs. We used complex span tasks that are well established as reliable and valid measures of WM capacity, where capacity depends on both storage and manipulation components [60]. These tasks focus on the ability to store information while concurrently manipulating it. Performance on these tasks is related to higher-level cognitive function as well as agerelated declines of WM capacity [61,62]. We used two tasks known to elicit changes in P3 and N400 in response to WM demands. Our ancillary question was whether the ERPs in the two tasks were related to WM in the same way or in different ways, which would be indicated by an age-related dissociation.
Eighty-five participants were recruited from the Claremont Colleges and the surrounding Claremont, CA (USA) community (Table 1). Based on a G*power analysis [63], we determined the sample size necessary for an a priori power of 0.80 with a moderate effect size (0.15), alpha set at 0.05, for a regression with two predictors (one categorical variable, age group, and one continuous variable, WM capacity). The required sample size is 68. Our sample size was 78-82, depending on the analysis. Younger adults (YA) received either financial compensation or partial course credit; older adults (OA) received financial compensation. Compensated participants were paid $10 per hour. All participants had at least 20:30 vision, measured using a vision contrast test (Vistech Consultants, Inc.). No participant reported a history of psychological disorders, neurological injury or disease, loss of consciousness for more than two minutes, or stroke. The study was approved by the Claremont McKenna College Institutional Review Board. All participants provided informed consent.
We assessed general levels of intellectual functioning with vocabulary and abstraction subscales of the Shipley Institute of Living Scale (SILS; [64,65]). YAs and OAs did not differ in the combined score without scaling for age, t(83) = 0.53, p = 0.60. OAs had higher Vocabulary scores than YAs, t(83) = -6.54, p < 0.001, and YAs had higher Abstraction scores than OAs, t(83) = 5.24, p < 0.001. No OAs scored below the cut-off score of 21 in the vocabulary assessment, indicating participants did not have impaired cognitive function [66]. OAs reported more years of education than YAs, t(76) = -9.19, p < 0.001 (see Table 1). Finally, participants were excluded from the reported analyses for less than 70 % accuracy on ERP task performance or poor data quality. For the visual oddball/P3 task, two YAs and five OAs were excluded, for a total of 78 participants (42 YA, 36 OA). For the word-pair judgment/N400 task, three YAs were excluded, for a total of 82 participants (41 YA, 41 OA).
Participants completed a demographic survey (age, gender, education, and self-reported health information) and measures of visual contrast and cognitive ability, followed by three WM capacity tasks and two ERP tasks.
WM capacity was assessed via shortened, computerized versions of the operation, reading and symmetry complex span tasks (https://englelab.gatech.edu/shortenedtasks.html; [27]). The validity and reliability of these tasks for measuring WM capacity has been verified in both long and shortened versions [67,68]. These tasks are routinely used to measure individual differences in WM capacity and are correlated with higher level cognitive functions (e.g., [9,[69][70][71][72][73]). Each task contains a storage and a processing component and each follows the same basic structure. A list of stimuli is presented in a certain order to store in WM, and the participant is asked to recall the items in the same order. Interspersed between each item is a processing task that requires active cognitive engagement. Participants completed a series of practice trials for each task. Participants practiced each component of the task alone before completing a combined practice with the processing component followed by the storage component.
For the operation span task, participants received a set of mathematical operations and were asked to judge if each equation was true or false (approximately half were true). After each mathematical operation, participants were presented with a randomly chosen letter. They recalled the letters at the end of each block; set sizes ranged from 3 to 7 with each set size appearing three times (75 total operation-storage pairs).
For the reading span task, participants received a set of sentences ranging from 10 to 15 words in length and they determined whether the sentence made sense (approximately half were sensible). After each sentence, participants were presented with a randomly chosen letter. They recalled the letters at the end of each block; set sizes ranged from 3 to 7 with each set size appearing three times (75 total sentence-storage pairs).
For the symmetry span task, participants received an 8 × 8 matrix of black and white squares and determined if the matrix was symmetrical around the vertical axis. After each matrix, participants were presented with a randomly chosen letter which they recalled at the end of each block; set sizes ranged from 2 to 5 with each set size appearing three times (52 total sentence-storage pairs).
WM capacity for each task was assessed using the edit distance scoring method (the number of stimuli in a trial minus the number of errors made by the participant). It is similar to the partial-credit scoring method, but takes into account errors which are transpositions, intrusions, or omissions made earlier in the recalled sequence [74,75] (https://englelab.shinyapps.io/taskscoring/).
Participants were seated 75 cm away from the monitor, with eyes level with the center of the screen, legs uncrossed, and feet flat on the floor. Stimuli were presented on a ViewPixx monitor (35.5 cm x 28 cm, 43.5 cm diagonal, 1280 by 1024 resolution, Vision Science Solutions). EEG/ERP tasks were presented using E-Prime3 software (Pstnet.com) and participants responded using a Chronos response box (Pstnet.com). EEG data were collected during active visual oddball and word-pair judgement tasks, modified from Kappenman et al. [29], which were counterbalanced by order across participants.
For the active visual oddball/P3 task, participants viewed a sequence of letter stimuli and classified each stimulus as a target or nontarget. On each trial, a letter (A, B, C, D, E, in uppercase, Geneva font, subtending 2.5 • x 2.5 • of visual angle) was presented for 200 ms in the center of the screen over a continuously visible central white fixation point (0.15 • visual angle), with a jittered SOA of 1200-1400 ms (rectangular distribution, average of 1300 ms). Participants pressed one button for targets and another button for nontargets using the index and middle fingers of their dominant hand. Before each block of trials, one letter was designated as the target stimulus and the other four letters were non-targets. Each of the five letters served as a target in one block of the experiment and as a non-target in the other four blocks, with the order of blocks randomized across participants. The stimulus-response mapping was counterbalanced across participants (50% of participants used the index finger for targets; 50% of participants used the index finger for non-targets). In each block, the probability of the target was 20%, and the probability of the non-targets was 80%, for a total of 8 target trials and 32 non-target trials per block. The probability of each of the non-targets and targets within a block was 20%. Participants completed five blocks with 40 trials each and a break between blocks, for a total of 200 trials.
For the word-pair judgment/N400 task, participants determined if two words were semantically related. The two words were presented in different colors of ink to allow participants to track which word required a response. Both colors were equally distant from the gray background in the CIE (1976) color space. Words were presented in uppercase, Geneva font, with each letter in a word subtending 1 • × 1 • of visual angle. Each word was presented over a continuously visible white central fixation point (0.15 • visual angle). The first word on each trial (prime) was presented in red (100, 0, 0) for 200 ms, followed by an ISI of 900-1100 ms (rectangular distribution, average of 1000 ms). The second word (target) was then presented in green (0, 90, 0) for 200 ms, followed by an ITI of 1400 to 1600 ms (rectangular distribution, average of 1500 ms). Participants pressed one button for related word pairs and another button for unrelated word pairs using the index and middle fingers of the dominant hand. The stimulus response mapping was counterbalanced across participants (50% of participants pressed with the index finger for related word pairs; 50% of participants pressed with the index finger for unrelated word pairs). Each target word was presented once in a related word pair and once in an unrelated word pair. Word pairs were randomly intermixed. Participants completed 120 trials, with a break provided every 20 trials.
Continuous scalp EEG was recorded from 64 active Ag/AgCl electrodes (actiCAP slim, Brain Products GmbH, Gilching, Germany) using the Brain Vision actiCHamp system (actiCHamp, Brain Products GmbH, Gilching, Germany). Impedances were kept below 50 kΩ throughout the experiment. Electrodes were placed according to the international 10/10 system (https://www.brainproducts.com/ downloads/cap-montages/). Data were sampled at 500 Hz and the on-line reference was FCz.
Data were exported and analyzed offline using EEGLAB [76] and ERPLAB toolboxes (https://www.erpinfo.org/erplab). EEGs were adjusted for DC offset by removing the mean value across the EEG and then filtered using an IIR Butterworth bandpass filter from 0.1 to 30 Hz (half amplitude cut off, 12 db/oct and 40db/dec roll off). Data were re-referenced off-line to linked mastoids (TP9/TP10). Continuous data were segmented into epochs (200 ms pre-stimulus to 800 ms post-stimulus) for rare target and frequent non-target conditions. Only data from correct trials were analyzed. Epochs were baseline corrected using the mean of the 200 ms pre-stimulus period. Artifacts in the data were addressed in two ways. First, we removed trials if they contained significant ocular artifacts (+/-100 μvolts at bipolar VEOG channels) during stimulus presentation (+/-150 ms surrounding stimulus presentation). Second, we identified artifactual signals (e.g., eye blinks, eye movements, noisy channels, line noise, muscle movement) using independent component analysis via second-order blind identification, which "unmixes" the EEG signal into independent components that can be classified as task-related or artifactual [77]. Initial artifact classification was performed using automated routines from the SASICA [78] and ICLabel [79] toolboxes, but all classifications underwent follow-up visual inspection to ensure that meaningful cognitive components were not mistakenly flagged for removal. For each participant's cleaned EEG data set, the trials were averaged for each condition.
Mean amplitudes and 50% fractional area latencies obtained from difference waves were calculated for P3 (rare-frequent) and N400 (unrelated-related) effects. ERP time windows were established for each age group and task at Pz based on the grand averaged data and the literature (e.g., [20,29]). The P3 window was 275 to 575 ms for YAs and 300-600 ms for OAs. The N400 window was 275 to 475 ms for YAs and 345 to 545 ms for OAs.
Complex span scores (i.e., edited distance performance measure) were calculated for each participant. An Age (OA, YA) x Span Task (Operation, Reading, Symmetry) mixed-model ANOVA indicated that the effects of Age, F( 1 Prior studies have shown that multiple span tasks create a more reliable measure of WM capacity (e.g., [68]). For YAs, the Operation Span was significantly correlated (p's < 0.05) with the Reading Span (r = 0.37) and Symmetry Span (r = 0.49) and the Reading Span was marginally correlated with Symmetry Span (r = 0.28, p = 0.07). For OAs, the Operation Span was significantly correlated with the Reading Span (r = 0.79) and Symmetry Span (r = 0.68) and the Reading Span was significantly correlated with Symmetry Span (r = 0.72).
When principal component analysis (PCA) was conducted for OA and YA together, the three span tasks loaded on a single component accounting for 73% of the raw variance. To address the possibility that OAs and YAs might not have the same underlying structure for the WM factor, we conducted PCA on YAs and OAs separately for the three WM complex span scores. Both groups showed the same structure; only a single significant component had an eigenvalue significantly above 1, indicating that they have the same structure. Nonetheless, they did differ to some extent on the factor loadings. Consequently, we calculated a regression-weighted factor score separately for each age group so that the scores reflected the different loadings on the single WM component. This WM factor score was used in subsequent regressions. 1 2 and 3; Fig. 1a) For behavioral data, proportion correct (accuracy) and median correct response times (RTs) averaged over rare and frequent conditions were analyzed using regressions with age group as a dummy variable and WM capacity as a continuous variable. For accuracy, the regression model was only marginally significant. WM was significant, indicating that greater accuracy was associated with greater WM capacity. However, there was no effect of age and no interaction of WM and age. For RTs, the regression model was significant. Age was significant such that OAs responded more slowly than YAs, but there was no effect of WM and no interaction of WM and age. The regression model with age group as a dummy variable and WM capacity as a continuous variable for P3 amplitudes was significant. Age was significant such that YAs had larger P3 amplitudes than OAs, but neither WM nor the interaction was significant. The same regression for 50% fractional area latency was also significant. Age was significant such that YAs had shorter P3 latencies than OAs, but neither WM nor the interaction was significant. 1 Regression analyses produced similar results for the WM factor when it was calculated from combined OA and YA scores. 3 and 4; Fig. 1b) Proportion correct (accuracy) and median correct response time (RT) averaged over related and unrelated conditions were analyzed using regressions with age group as a dummy variable and WM capacity as a continuous variable. For accuracy, the regression model was only marginally significant. Age was significant indicating that OAs had greater accuracy than YAs. However, there was no effect of WM and no interaction of WM by age. For RTs, the regression model was significant. Age was significant such that OAs responded more slowly than YAs, but there was no effect of WM and no interaction of WM and age. The regression model with age group as a dummy variable and WM capacity as a continuous variable for N400 amplitudes was not significant. Age was significant, however, such that YAs had larger N400 amplitudes than OAs, but neither WM nor the interaction was significant. The same regression for 50% fractional area latency was significant. Age was significant such that YAs had faster N400 latencies than OAs, but neither WM nor the interaction was significant.foot_0
Age-related declines are evident in both WM (e.g., [81]) and ERP measures (e.g., [28]). In this study, we investigated whether a composite of widely used measures of WM capacity, namely complex span tasks, mediated overall effects of age on P3/visual oddball and N400/word-pair semantic judgment effects elicited from tasks with WM requirements. Our study confirmed age-related decreases in WM capacity, especially for visuospatial tasks (c.f., [80]). We also confirmed that the span measures loaded on a single principal component [68] for both age groups. In addition, we replicated the reduced P3 and N400 amplitudes and longer ERP latencies in older adults [28]. But, contrary to our predictions, WM capacity did not affect either ERP amplitudes or latencies. Consequently, we failed to find that WM capacity influences age-related effects on the ERP measures, regardless of task.
One possible explanation is that the P3 in the oddball task and N400 in the semantic comparison task do not involve WM. To the contrary, Reed et al. [37] found that tasks loading on WM (task switching and digit symbol substitution) were predictive of individual differences in P3. Also, the involvement of WM in semantic judgement and the N400 has been observed in tasks which interleaved semantic judgement and n-back trials [82]. Semantic judgements were slowed as n increased, suggesting semantic judgments engaged WM. Salisbury [83] found that performance on the Trail Making Test and the digit symbol substitution tests, both engaging WM, affected the N400.
A second and more likely explanation is that the WM operations elicited by complex span tasks are different from those elicited by the oddball and semantic comparison tasks. WM is multidimensional [60] and plausibly, not all tasks rely on the same WM processes. The construct of WM comprises information storage as well as attentional control operations on stored information-updating, set shifting, and inhibition. For example, Reed et al. found previously that only measures related to context updating and attentional control are predictive of the P3 as measured by a visual oddball task; by contrast, Rmeasures of simple storage (visual short-term memory and spatial short-term memory) were unrelated to P3 [37]. Unsworth et al. [84] found strong correlations between memory span and measures of short-term storage. (They also found correlations with tasks intended to measure attentional control, but close examination shows they, too, largely involved simple short-term storage.) Arguably, the present ERP tasks had very limited storage demands and reflected more the operation of attentional control. Despite their manipulation component, complex span tasks have high storage demands but lower demands for attentional control. This would explain how there could be age-related differences in span measures and in ERP measures, but complex span and ERP measures could be unrelated. The version of the visual oddball task that we used relied more on context updating processes and the word-pair task relied on comparing current information with semantic relations in long-term memory.
Dong et al. [38] found that individuals with high scores on a hybrid memory span task had significantly higher P3 amplitudes than those with low scores. P3 was measured in an n-back task. P3 amplitude decreased monotonically as n was increased from one to four, but this decrease in the P3 was similar for low and high WM capacity groups. The n-back task has been called the gold standard of WM measures [85]. However, it correlates only weakly with span measures of capacity [86]. Arguably, n-back performance reflects strong but somewhat independent loadings on both storage capacity and attentional control. Following this reasoning, the difference found by Dong et al. between the low and high WM groups in the P3 may reflect the storage aspect of the n-back task whereas the change with increasing load may reflect independently the increasing demands for updating and set shifting. Although the n-back task has been related to P3 amplitudes, Kane et al. [86] have argued that the n-back task is not a construct valid measure of WM function, principally because it does not show convergent validity with complex span measures. A more agnostic way to view this is that n-back tasks and complex span tasks both challenge WM but that they call on, or emphasize, different aspects of WM function. That is, neither is an invalid measure of WM. The prior finding that n-back performance is related to P3 taken together with our finding that complex span is not-given that both are valid measures of WM-is consistent with the argument that WM involves a variety of component operations, not all of which are elicited in the same way by a particular task. Thus, not all WM tasks elicit the same WM processes. In other words, the same operations may not be used in the same ways in different tasks, as our results suggest. Here we tried to take a measure of WM capacity established as reliable and valid and show how it was related to two different tasks that require WM. We found that it did not relate to P3 and N400. So, we can infer that storage and concurrent processing operations in WM are not the processes differentiating the age-related effects for P3 and N400. Instead, we suggest that tasks relying heavily on WM component processes other than or in addition to storage (e.g., updating and interference suppression), such as the n-back task, may better relate to the WM processes indexed by the P3 and N400.
In conclusion, we replicated prior findings of age-related differences in WM capacity and in ERP measures. Additionally, however, we found no relationship between the span measures of WM and either P3 or N400 ERPs, regardless of age. The most likely explanation for the absence of a relationship is that the WM operations required for span tasks are different from those elicited by oddball and semantic comparison tasks. Perhaps the attentional control components of WM influence oddball and semantic judgement task performance and their corresponding ERPs in a different manner from storage components, on which complex span tasks load more strongly. Future research using a choice of tasks that allowed independent manipulation of storage and attention control would permit us to explore the fractionation of the effects of WM load on P3 and N400. In sum, our results converge with the conclusion that WM is a multifaceted construct and that the subcomponents are not assessed by all WM measures and are not elicited by all tasks.
The Corresponding Author, Catherine Reed, declares that this work has not been published previously except in the form of abstracts. It is not under consideration for publication elsewhere, and that its publication has been approved by all authors and tacitly or explicitly by the responsible authorities where the work was carried out, that, if accepted, it will not be published elsewhere in the same form, in English or in any other language, including electronically without the written consent of the copyright-holder.
Dichotomizing WM gives essentially equivalent results.