Developmental trajectories of event related potentials related to working memory

Neuropsychologia 95 (2017) 215–226

Contents lists available at ScienceDirect

Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia

Developmental trajectories of event related potentials related to working memory

MARK

⁎

Catarina I. Barriga-Paulino , Elena I. Rodríguez-Martínez, Antonio Arjona, Manuel Morales, Carlos M. Gómez Human Psychobiology Laboratory, Department of Experimental Psychology, University of Seville, Spain

A R T I C L E I N F O

A BS T RAC T

Keywords: Working memory Maturation Individual maturation Event related potentials Delayed Match-To-Sample Principal components analysis

Working memory is an important cognitive function, and it is crucial to better understand its neurophysiological mechanisms. The developmental trajectories of the Event Related Potentials related to this important function have hardly been studied. However, these ERPs may provide some clues about the individual state of maturation, as has been demonstrated for anatomical brain images. The present study aims to determine the behavioral and neurophysiological development of Working Memory (WM) processes. For this purpose, 170 subjects with ages ranging from 6 to 26 years performed a visual Delayed Match-to-Sample task (DMTS). The RTs, total errors, and Event Related Potentials (ERPs) in the phases of encoding, retention, and matching were obtained. Results revealed a decrease in the amplitude of ERPs with age, paralleled by improved performance on the DMTS task (i.e., shorter RTs and fewer errors). None of these variables were affected by gender. To determine whether memory performance was influenced by the individual pattern of maturation beyond age, the amplitude of the different ERP components was correlated with RT and errors on the WM task after removing the effect of age. Frontal N2 and posterior P1 and the Late Positive Component were the only ERPs that presented significant correlations with behavioral errors. Behavioral performance was predicted by age and by the scores on the first component extracted from Principal Component Analysis (PCA) of the ERPs. Age (under 17 years old) explained 85.04% and the PCA component explained 14.96% of the variance explained by the bivariate model predicting behavioral errors (1/ age + scores of 1st PCA component). From the age of 17 on, the principal PCA component ceases to be an independent component predicting error performance. The results suggest that the individual maturation of ERP components seems to be of particular importance in controlling behavioral errors in WM, as measured by the DMTS.

1. Introduction In the past few years, several studies have focused on the neural activation patterns underlying WM processes (Ghetti et al., 2012; Shing et al., 2016; Constantinidis and Klinberg, 2016). The ERP technique applied to the investigation of the cerebral mechanisms of information storage and maintenance is useful for making inferences about the timing and anatomical location of these WM processes. In DMTS paradigms, there are distinctive ERP components related to information processing in each of three different phases of the task: the encoding of the stimulus, its retention (interval between S1-S2), and recovery (matching process between the stored and the presented stimulus). Regarding the encoding phase, some studies using visual tasks have found that P2 consists of an index related to WM (Wolach

⁎

and Pratt, 2001). This ERP appears approximately 200 ms after visual stimulation (Mecklinger and Pfeifer, 1996), and it seems to reflect a mechanism of attentional selection of the stimulus’ features. In addition, in the verbal domain, differences in P2 peak amplitude suggest that anterior and posterior distributional differences are elicited during the encoding of words for rote and elaborative memory tasks (Dunn et al., 1998). During the stimulus’ retention period on WM tasks, using S1-S2 paradigms, the typical ERP that arises is a Negative Slow Wave (Ruchkin et al., 1990, 1992). This ERP presents higher amplitude in the left hemisphere during phonological memory operations (Barrett and Rugg, 1990; Rugg, 1984a, 1984b) and higher amplitude in the right hemisphere during visual memory operations (Barrett et al., 1988; Barrett and Rugg, 1989). Moreover, this component is sensitive to task difficulty (Ruchkin et al., 1992) and seems to

Correspondence to: Human Psychobiology Laboratory, Department of Experimental Psychology, University of Seville, Calle Camilo José Cela, S/N, 41018 Seville, Spain. E-mail address: [email protected] (C.I. Barriga-Paulino).

http://dx.doi.org/10.1016/j.neuropsychologia.2016.12.026 Received 10 July 2016; Received in revised form 21 December 2016; Accepted 22 December 2016 Available online 23 December 2016 0028-3932/ © 2016 Elsevier Ltd. All rights reserved.

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

cognitive functions (Rojas-Benjumea et al., 2013; Roalf et al., 2014), it is possible that ERPs would help to establish the cognitive maturational status of subjects during development. For this purpose, two different approaches will be followed: (i) After eliminating the effect of age on the behavioral variables and the ERPs, the age residuals of ERPs and the behavioral variables would be correlated, indicating which ERPs are more related to individual behavioral performance during development; (ii) the ability of age and an ERP maturational component extracted from the PCA of ERPs to predict the WM parameters will be assessed. If ERPs contain information independent from age regarding the maturational status of subjects, the prediction of behavioral performance based on age will improve when the ERP maturational principal component is included as a regressor of behavioral performance. Thus, the present report will provide information about the maturational progress of each ERP during human development and about the value of ERPs in assessing the individual maturational cognitive status of children.

present a different topographical distribution for spatial WM tasks and object memorization (Ruchkin et al., 1997). Regarding the matching phase of the memorized stimulus, P3 consists of the ERP that has been shown to index processes related to WM updating. This positivity occurs between 300 and 800 ms after stimulus onset and reflects the activation of a processing mechanism responsible for updating the stimulus’ representations in the WM (Donchin and Coles, 1988). Its maximum amplitude is located in parietal regions (Johnson, 1993). Although P2 has been related to the encoding phase, as described above, some investigations seem to point in the opposite way, associating this ERP with the matching of the actual visual input with an expected form (Evans and Federmeier, 2009). In addition, Kaan and Carlisle (2014) found higher P2 amplitudes for predictive sequences of stimuli where participants generate an expectation of the form of the next letters, versus non-predictive sequences (random sequences). Although these studies did not use the same type of paradigm as ours (DMTS task), they showed a mechanism related to memory that allows subjects to anticipate, during the delay period, the possible appearance of the previously seen stimulus; thus, it is possible that the ERP amplitudes that appeared were similar to when a classical DMTS is used. Some studies have described that, during development, a synaptic pruning occurs (Huttenlocher, 1990; Whitford et al., 2007; Giedd et al., 2009), which would produce a reduction in neural sources. This reduction is clearly observed in the decline in absolute power of brain rhythms from delta to beta (Segalowitz et al., 2010). It has been proposed that this decrease in power is related to the reduction in the synaptic connections as age increases, which would be macroscopically observed in the reduction in cortical thickness with age. For ERPs, although not for all components, it is typical to observe a reduction in the amplitude of ERPs with age, for instance, in the visual P1 (Segalowitz et al., 2010) and the visual P3b (Courchesne, 1978; Stige et al., 2007; Flores et al., 2009). The relationship of brain rhythm maturation and ERPs with cognitive improvement has been described, for instance, as a parallel course of theta power and WM maturation (Rodríguez-Martínez et al., 2013). The possible relationship between individual ERP status and maturation, beyond age-related maturation, has also been studied. Segalowitz showed that a higher amplitude in early Contingent Negative Variation was related to performance on executive tasks such as planning and set-shifting (Segalowitz et al., 1992). On the other hand, the possible association between the individual level of brain maturation and individual cognitive performance has been described. Fast changes in cortical thickness during development are associated with a higher IQ (Shaw et al., 2006) and a multivariate parameter extracted from fractional anisotropy correlated with Working Memory and numerical abilities at school entry ages (Ullman and Klingberg, 2016). However, this relationship is lost in adolescence, suggesting a greater contribution of biological factors in early childhood and an increasing contribution of social factors as age increases. This individual rate of maturation is also perceived as an increase in performance variability on cognitive tests in early childhood (Rojas-Benjumea et al., 2013; Roalf et al., 2014). All these results suggest that an individual brain-age is present during child development. The goal of this manuscript is to describe the developmental trajectories of all the ERPs observed on a DMTS task performed by a wide range of subjects between 6 and 26 years old. No previous study offers such a complete description of the development of the complete series of ERPs on DMTS tasks across age. A reduction in the amplitude of ERPs with age is expected, given the possible reduction in neural sources due to synaptic pruning. From a functional point of view, the possibility will also be explored that the individual maturational state of ERPs, super-imposed on the age-related maturational general trend, would influence WM performance. This possibility is explored in model 2, described in the methodological section. As children have been shown to present a broad temporal window for the maturation of

2. Materials and method 2.1. Subjects One-hundred and seventy subjects between 6 and 26 years old (15.89 years ± 6.12) participated in this study. For each age, 8 subjects were recorded and analyzed (4 males and 4 females), with a total sample of 85 males and 85 females. However, five subjects were excluded due to excessive EEG artifacts. The final sample was composed of 165 subjects, 81 males and 84 females (16.00 years ± 6.07). Subjects did not report any neurological diseases or psychological impairments, and they were extracted from middle-class socioeconomic backgrounds. The children had good academic records and were recruited from public schools, and the young adults were college students recruited through advertisements on the University Campus. Experiments were conducted with the informed and written consent of each participant (parents or tutors in the case of children), following the Helsinki protocol. The study was approved by the Ethical Committee of the University of Seville. 2.2. Stimuli and task procedure Visual stimuli were Pokémon and Digimon-type cartoons. The size of all stimuli was adapted in Picassa to equal dimensions of 142×228 pixels. Uncommon stimuli were used to avoid verbal strategies and to ensure that the memorization processing was mainly visual. The stimulus presentation program used was E-Prime version 2.0, and an SRBOX Cedrus was used to record the subjects’ responses. The paradigm used was a DMTS task composed of a total of 128 trials organized in 4 experimental blocks with 32 trials each. The trials were counterbalanced; i.e., in half of them the target stimulus appeared on the left visual field, and in the other half the target stimulus appeared on the right visual field. The order of presentation was totally random, so that each subject performed a unique sequence of trials. The task was kept relatively simple in order to facilitate the testing of the youngest children. It is possible that subjects might have certain knowledge of the items presented; however, the experiments were performed in 2010– 2012, before the ‘Pokémon-GO’ game was launched. In any case, 256 figures were presented in the experiment, which is a very large number to remember. Moreover, the experiment was still a working memory experiment, regardless of the type of figures used. The task started with the appearance of the first stimulus (S1) at the center of the screen. The stimulus covered a visual angle of 4.56° on the horizontal meridian. Fig. 1 shows an example of a trial during the task. The subjects were situated 60 cm from the screen. The stimulus was presented for 1000 ms and had to be memorized by the subject. Then, a 216

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

2.4. Data analysis 2.4.1. Behavioral data Means of RT and the percentage of total errors (a combination of commissions, omissions and anticipations) were calculated. Anticipations were defined as responses to S1, responses less than 200 ms after the S2 appearance, and/or responses in the interval between S1 and S2. ANOVA with gender and age groups was computed. The total sample of subjects was split into 5 age subgroups for the ANOVA: 6–9, 10–13, 14–17, 18–21, 22–26 years old. All the groups were composed of 32 subjects, except the first one, which had 31 subjects, and the last one, which had 38 subjects. For the age group factor, no post-hoc analyses were computed; instead, the developmental trajectories of the behavioral variables were obtained by means of regressions between age and the two behavioral parameters. The linear and inverse models were applied, and the most explanatory model was reported. Fig. 1. Example of a trial of the Delayed Match-To-Sample paradigm used.

2.4.2. Electrophysiological data EEG pre-processing and ERP analysis were performed with EEGLAB 10.0.0.0b using the 2010 version of Matlab (MathWorks Inc., MA, USA). A low-pass filter of 25 Hz was applied to all the recordings to eliminate the higher frequencies, and in the recordings of five subjects we also had to apply a high pass filter of 0.5 Hz because the slow waves introduced noise in the recordings. An artifact rejection protocol was applied: discarding from the average the recorded voltages that exceeded ± 100 µV in the recordings of subjects from 16 years old on, and ± 150 µV in the recordings of subjects up to 15 years old in any channel, in order to eliminate any extra-cerebral contamination. The application of distinct voltage values was due to the known difference observed in the spectral power of children's and adults’ recordings, as children present a higher spectral power than adults do (see, for example, Barriga-Paulino, Flores, and Gómez, 2011). Only correct trials were considered for analysis. The baseline was adjusted for each phase of the DMTS task: from −100 to 0 ms, before S1 for the encoding phase; for the delay period, the baseline was set at-100–0 ms before the initiation of the delay period (just before S1 offset), and for the matching phase, the baseline was between −100 and 0 ms before the onset of the S2 stimulus. Averages were obtained independently for each phase of the DMTS task. The polarity, latency, and topographical distribution of the different brain waves associated with each phase of the DMTS task were the main criteria used to determine the different ERP components and electrodes that would be introduced in future analyses. This approach made it possible to differentiate C1, N1, P1, N2, P2 and the Late Positive Component (LPC) in the encoding and matching phases. Similarly, the offset P1 and offset P2 and SW from the retention interval were obtained after switching off the S1 stimulus. The peak amplitude procedure to extract the amplitude of each component (extract the peak amplitude in a given latency) was not possible because some components disappeared or highly reduced their amplitude with age. Some components such as the slow wave did not present any peaks. Therefore, the ERPs’ grand-average peak amplitudes were selected to determine the time window for the analysis of each ERP component. The latencies for each of the components after stimulus presentation (S1 or S2) were: C1 (40–60 ms), P1 (100– 130 ms), N1 (130–160 ms), N2 (160–230 ms), P2 (160–230 ms), LPC (290–400 ms), P1 offset (1100–1130 ms), P2 offset (1160–1230 ms) and Slow Wave (2100–2400 ms, post-S1). The latencies of the components in the matching phase were the same as those in the encoding phase. The developmental trajectories of the ERP amplitudes were obtained by regressing ERP amplitudes vs. age. Linear and inverse models were applied, and the best model for each component was

blank screen with a fixation point in the center appeared for 1500 ms. During this delay period, subjects had to maintain the S1 they had previously seen in memory. After that, two stimuli (S2) appeared for 2000 ms (one the same as S1, and the other different), one on the left, and the other on the right side of the screen. Subjects were instructed to press the left button with their left hand or the right button with their right hand if the previously presented S1 appeared on the left or right side of the screen, respectively. The experimental subjects had 2000 ms from the S2 onset to respond. Finally, the trial ended with auditory feedback, providing the subject with a different sound depending on whether a correct or incorrect response was produced. Then, 2000 ms passed until the start of a new trial. A practice session of 10 trials was carried out before the experimental task. Subjects were instructed to relax their facial muscles, keep their eyes focused on the fixation point presented at the center of the screen, and blink as little as possible. In the case of some young children, it was necessary to stay with them in the Faraday box and respond together during the practice trials to make sure that they understood the instructions. The instructions to the subjects were: “In the center of the screen a cartoon will appear that you should memorize. Next, two cartoons will appear, one on the left and the other on the right side of the screen. One of them is the same one you memorized before, and the other is different. You must press the left button or the right button depending on the side where the memorized cartoon appears. Then, you will hear a sound telling you whether you did ok.” The duration of the task was approximately 17 min, and rest periods were allowed between blocks.

2.3. EEG recording Recordings were made from 32 scalp sites (Fp1, Fp2, F7, F3, Fz, F4, F8, FC5, FC1, FC2, FC6, M1, T7, C3, Cz, C4, T8, M2, CP5, CP1, CP2, CP6, P7, P3, Pz, P4, P8, POz, O1, Oz, O2), organized according to the 10–20 system and using an Electro-Cap, with 4 additional electrodes to record ocular movements. The horizontal right and horizontal left electrodes were located at the outer canthus of each eye to record horizontal eye movements, and the vertical superior and vertical inferior electrodes were located above and under the left eye to record vertical eye movements. All the scalp electrodes were compared to an average reference, and impedance was maintained below 10KΏ. Data were recorded in DC at 512 Hz, with a 20,000 amplification gain using a commercial AD acquisition and analysis board (ANT). EEG was recorded in a dimly lit and electrically isolated Faraday box. 217

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Fig. 2. Inverse regressions between reaction times and total errors with age.

Fig. 3. ERPs observed in the three phases of the paradigm used in the experiment (encoding, retention, matching) for the five age groups. For each phase, three electrodes where the ERPs can be observed are displayed. Time is expressed from S1 onset.

218

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Fig. 4. Topographical maps of ERPs for encoding, retention, and matching phases in the five age groups.

Fig. 5. Regressions between age and the ERP representative electrodes in the encoding phase.

location for the ERP component; and (iii) the determination coefficient with regard to age (explained variance) has to be the highest of all the surrounding electrodes. During the delay period, 3 electrodes were selected, given the opposite polarity in centro-parietal and occipital electrodes for the slow wave. The selected electrodes for the encoding phase were: O1 and O2 (c1, P1, P2 and LPC components), F4 and C4

selected. For subsequent analysis, a selection of two electrodes for each ERP component was made, based on the following hierarchical criteria: (i) The selected electrode has to be clearly defined in terms of polarity and latency, and it should present high amplitude for the grand average of the ERP component; (ii) it must correspond to a topographical location previously described in the scientific literature as a typical 219

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Fig. 6. Regressions between age and the ERP representative electrodes in the encoding phase (Slow Wave and offset potentials).

sources of variability in the ERPs. For this purpose, the PCA for the ERP list in all subjects was computed, using the eigenvalue > 1 criterion (Guttman, 1954) to decide the number of significant sources of variability (see Supplemental Table 2 for the explained variance of principal components). Six components were extracted, making it possible to consider 12 independent correlations (6 principal components x 2 behavioral variables) and a threshold p value of p=0.05/ 12=0.0041. The correlation approach was also used to understand whether the individual behavioral variability was related to the individual ERP variability. For that, the residuals of the behavioral variable vs. age and the residuals of the ERPs vs. age were obtained from the regressions in Figs. 2 and 5–7, respectively. In this case, the PCA of the ERP residuals versus age detected 7 components with eigenvalues > 1; therefore, the p value threshold was fixed at p=0.05/ 14=0.0036. However, for the latter case, the Pearson correlation was used, instead of the Spearman correlation, given that the residuals are normally distributed. In order to determine whether the ERPs contained predictive information about the subjects’ maturation beyond age, the PCA analysis of the residuals of the ERPs’ amplitude vs. age, described above, was used to predict individual behavioral performance. In several instances of behavioral and EEG data (Barriga-Paulino et al., 2011, 2016; RodriguezMartinez et al., 2012), we showed that when the PCA is computed without rotation, the first factor accounts for a global factor. If the ability of age to predict the behavioral variables improves when including the scores on the first component of the PCA computed from the residuals of ERPs vs. age (which are independent of age), then the conclusion can be proposed that the PCA component, extracted from ERPs’ age-residuals, could be interpreted as a code for the maturational state of a given subject. Expressed in mathematical terms:

(N1) and F4 and F8 (N2 component). For the retention phase: O1 and O2 (P1 offset), P3 and P4 (P2 offset), Cz, Pz and O2 (SW). For the matching phase, the same electrodes were selected as in the matching encoding phase, except in the N1 component, where the electrodes with the highest determination coefficients in the regression with age were F3 and F4. A sub-sample of the selected electrodes for statistical analysis in each ERP component is presented in Figs. 5–8. To analyze statistical differences in gender and group with regard to the ERPs’ amplitude, repeated-measure ANOVAs were independently performed for each ERP component. To do so, the electrode factor with the representative electrodes of each ERP as the within-subjects factor and the group and gender variables as the between-subjects factors were used to compute the ANOVAs. The total sample of subjects was split into the 5 age subgroups previously indicated. The selected electrodes for each ERP component ANOVA are described above. No post-hoc comparisons were computed for the age-group factor's significant results, but the computed developmental trajectories would express the maturational process of the different ERP components. Amplitude comparisons of similar components obtained in the matching and encoding phases would not be computed for the early components because S1 and S2 had different numbers of cartoons displayed (1 and 2, respectively), and amplitude differences could be attributed to the different number of displays rather than to the phase: encoding or matching. However, for the LPC, an ANOVA was conducted to compare the LPC amplitude in the encoding and matching phases (factors: phases, electrodes, age group). To appreciate the possible parallel maturational dynamics of ERPs and the two behavioral variables recorded, Spearman's correlation was obtained for the ERPs and the two behavioral variables. The correlation matrix was thresholded, taking in account the number of significant 220

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Fig. 7. Regressions between age and the ERP representative electrodes in the matching phase.

differences in gender on the performance of the behavioral data. However, there were statistically significant differences for group for both the mean RT [F (1, 165)=97.946, p < 0.001)] and total errors [(1,165)=28.954, p < 0.001)]. Fig. 2 shows the developmental trajectory of RTs and total errors, where an inverse relationship with age can be observed.

Model 1: BV=β0 + β1*1/age Model 2: BV=β0 + β1*1/age + β2*CS1st Model 2 should be able to predict the behavioral performance better than model 1 if the ERPs are coding individual performance beyond age, where BV refers to the behavioral variable, β0 represents a constant, and β1 and β2 are weighting parameters. The CS1st refers to scores on the first PCA component computed on the residuals obtained from a regression analysis between the ERP amplitude and age. For the comparison of the predictive value of the two models, the Rv.3.2.3 software was used. The lm function was used for the estimation of the regression model parameters; the ANOVA function made it possible to compare the predictive value of the models; boot (from the car library) was used to validate the model parameters; boot was used for F statistic bootstrap validation with 500 replications; the confit function made it possible to obtain the confidence intervals for the validation of regression parameters, and boot.ci was used to compute the confidence intervals for the F statistic of the ANOVA function, by means of the corrected percentile method (Bca).

3.2. ERPs Fig. 3 shows the ERPs in three representative electrodes observed in the three phases of the experimental paradigm (encoding, retention, and matching) in the five age groups. Fig. 4 displays the topographical maps for all the components observed in each phase and in each age group. Figs. 5–7 show the developmental trajectory of the components in selected electrodes in which the components were more evident and the explained variance by age was the highest (see methods section for the electrode selection criteria). During the encoding phase, in posterior electrodes POz and Oz, the C1, P1, P2 and LPC components are observed (Fig. 3). Their posterior topographies can be observed in Fig. 3. The negative frontal components N1 and N2 are observed in the same figures. All these components decreased in amplitude as age increased. Fig. 5 quantifies, using linear and inverse regressions, the decrease in amplitude of positive and negative components during the encoding phase. In the retention phase, a negative slow wave that decreases its amplitude with age is obtained (Fig. 3). The negativity is present in

3. Results 3.1. Behavioral data The ANOVA showed that there were no statistically significant 221

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Fig. 8. Correlations between behavioral variables and ERP amplitudes (A) and correlations of residuals of behavioral variables with age and ERP residuals with age (B). Correlation values were thresholded as indicated in the methods section. The ERP residuals vs. total errors residuals are represented in Fig. 1 in the Supplemental material.

phases in the different age groups showed a significant difference in all the groups: p=0.005 for the groups of children, preadolescents, and adolescents, and p=0.025 for the young adult and adult groups. In all cases, the matching phase presented higher amplitudes than the encoding phase. The correlation matrix for behavioral variables vs. ERP amplitudes showed a high number of significant correlations between ERPs and behavior, indicating a parallel maturational course between these two sets of data (Fig. 8A). The ERPs that presented correlations with behavioral data are represented in Fig. 8A. The correlation matrix for residuals of ERPs and behavioral variables vs. age reveals a more limited pattern, showing positive correlations between the age residuals of the N2 component (during the encoding phase) and the age residuals of total errors (Fig. 8B), and a negative correlation between the age residuals of the P1 and LPC components (during the encoding and matching phases) and the age residuals for total errors (Fig. 8B). The PCA was computed on the residuals of ERP amplitude vs. age for all the components represented in Figs. 5–7. Seven components were extracted with the criterion of having an eigenvalue > 1 (Guttman, 1954) (see Supplemental Table 2 for explained variance). Only the first component presented a high correlation with all components, making it possible to consider it as a global factor for all the ERP residuals vs. age. Model 1 was applied to predict the behavioral variables, RTs, and Total errors. For RTs, model 1, including only the inverse of age as a predictor of RTs, was statistically significant (p < 0.001, R2=0.499). Model 2, including the scores on the first component as a regressor, was significant only for the age regressor, and the same statistical significance and explained variance as in model 1 were obtained. Model 1 applied to predict total errors based on age was statistically significant (p < 0.001, R2=0.358, adjusted R2=0.354). When the scores of the first component were included as a regressor, the explained variance was higher than in model 1 (p < 0.001, R2=0.412, adjusted R2=0.405). The inverse of age (p < 0.001) and the scores of the first component were independent regressors for predicting total errors (both regressors at p < 0.001); in fact, they were completely uncorrelated (p=0.97).

posterior electrodes, whereas in anterior electrodes, a positivity appears in children that decreases in amplitude with age. Just after S1 stimulus offset, P1 and offset P2 components are displayed in posterior electrodes. The topographies of these components can be observed in Fig. 4. The developmental trajectories of these components show the reduction in amplitude of the positive and negative phases of the slow wave (Fig. 6). The developmental trajectories of the offset P1 and offset P2 components, which show a reduction in amplitude with age, are displayed in Fig. 6. The ERPs during the matching phase were homologous to those of the encoding phase, and they also presented similar topographies and developmental trajectories (Figs. 4 and 7, respectively). In order to analyze whether gender and age were related to ERP amplitudes during development, ANOVAs with gender, age groups, and electrodes as factors were computed independently for each component. There were no gender effects on any ERP. All ERPs presented significant group effects (see Supplemental Table 1). Post-hoc age effects would not be computed, given that the ERP amplitude dependency on age had already been analyzed by means of developmental trajectories. Gender was not significant as a main factor in any of the ERP components for the three WM processing phases. There were two significant interactions that included gender: electrodes x gender in offset P2 (p < 0.038; F: 4.358; df: 1,155) and gender x group in C1 of the matching phase (p < 0.006; F: 3.721; df: 4,155). For offset P2, the post-hoc analysis revealed that only the P3 electrode showed statistically significant differences between genders (p < 0.028), with females presenting higher amplitudes than males. For C1, only the 6–9 year old group presented statistically significant differences between genders for both electrodes (p < 0.005 for O1, p < 0.014 for O2), with males presenting higher amplitudes than females. The ANOVA to compare the LPC in the encoding and matching phases showed a significant effect of the age group [F (4, 160) =33.858, p < 0.001], of the phase factor [F (1,160) =64.14, p < 0.001], due to higher amplitudes in the matching phase than in the encoding phase, and of the interaction phase for the group [F (4,160) =3.230, p < 0.014]. The Bonferroni comparison of the encoding and matching 222

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

the effects of age, improve the prediction produced by age of the number of errors on the DMTS task, suggesting that the individual brain maturational state of the subject influences cognitive performance in the 6–17 years’ period. There were no gender effects as a main factor in any of the 15 analyzed ERP components, and only a very small proportion of gendereffect interactions with the other analyzed factors (one electrode for P2 offset and the youngest group for C1), indicating that gender did not influence the maturation of the analyzed ERPs. The results of the development of all ERP components, from childhood to adulthood, will be described for each phase of the DMTS task used in this study.

However, when age-residuals for total errors were represented, three clear outliers (SD > 3) were observed that would have a great influence in the regression. Therefore, the PCA and the two models were re-computed without the data for these three subjects. The results for model 1 and 2 were improved by eliminating these three subjects (model 1: p < 0.001, R2=0.442, adjusted R2=0.438; model 2: p < 0.001, R2=0.488, adjusted R2=0.481; both regressors were significant at p < 0.001; and uncorrelated (p < 0.650). The ratio between the R2 of model 1/R2 of model 2 was 90.5%, indicating that the proportion of variance explained by the first component scores was 9.5% of the total variance explained by model 2. In order to demonstrate that model 2 was a better predictor of behavioral errors than model 1, a comparison of the two models was performed using the F statistics and a bootstrap method. The F statistics showed that model 2 produced a significant reduction in error (F (1,159) =14.35, p < 0.001) compared to model 1. Additionally, the bootstrap statistics obtained an estimated value for the F statistics of F=14.15 (IC95%=[2.49,39.36]). These results indicate that model 2 produces a better prediction of behavioral errors than model 1. The validation of the parameters of model 1 and model 2 appears in the Supplemental material. An additional effort was made to compare models 1 and 2 in predicting total errors in the age period where behavioral ERPs presented the most pronounced changes ( < 17 years, eliminating the three outliers). For this purpose, the residuals of ERPs for age were recomputed for subjects under 17. The linear model explained the regression of ERPs with age. PCA was re-computed for the residuals obtained from linear regressions between ERPs vs age. Model 1 and model 2 were reassessed for the under 17-year-old sub-sample (only for the prediction of behavioral errors). The proportions of explained variance were R2=0.398 (p < 0.001; adjusted R2=0.390) and R2=0.468(p < 0.001; adjusted R2=0.454) for models 1 and 2, respectively. In model 2, both regressors were statistically significant (1/age, p < 0.001; CS1st, p < 0.003, respectively). The ratio between the R2 of model 1/R2 of model 2, when only subjects younger than 17 years old were included in the analysis, was 85.04%, indicating that in model 2, the proportion of variance explained by the first component scores was 14.96%, which represents an increase in the proportion of variance explained by the first component compared to the total sample. In order to demonstrate that model 2 was a better predictor of behavioral errors than model 1 for the < 17-year-old sub-sample, a comparison of the two models was performed using the F statistics and a bootstrap method. The F statistics showed that model 2 produced a significant reduction in error (F (1,73)=9.69, p=0.003). Additionally, the bootstrap statistics obtained an estimated value for the F statistics of F=9.54 (CI95% =[1.67,25.61]. These results indicate that model 2 produces a better prediction of behavioral errors than model 1 for the < 17-year-old sub-sample. The validation of the parameters of model 1 and model 2 appears in the Supplemental material. When the complementary analysis for the population of subjects > =17 years old was computed, models 1 and 2 did not reach statistical significance, indicating that neither age nor ERP residuals with age were predictors of behavioral performance in the post-adolescent and young adulthood periods.

4.1. Encoding phase In the posterior regions during the encoding phase, components C1, P1, P2 and LPC were obtained, and in anterior regions, the N1 and N2 components were induced by the S1 stimulus. C1 is the earliest component of visual ERP and reflects the activation of the striate and extrastriate visual cortices (V1) induced by visual stimuli (Gómez et al., 1994; Clark et al., 1995). Regarding this ERP, there is no evidence in the literature about its developmental trajectory with age. In this study, the amplitude of C1 decreased as age increased. The decrease in the C1 with age suggests that the primary visual cortex continues pruning beyond early childhood, although because brain localization techniques have not been applied in the present report, the contribution of extrastriate sources cannot be ruled out. The N1 component presented a fronto-central topography and a linear decrease with age. The visual frontal N1 is a component related to prefrontal cortex activation through the dorsal pathway (Foxe and Simpson, 2002), which corresponds to an alerting initial response and would make it possible to start frontal feedback to posterior areas analyzing the incoming stimulus. There are no reports on the developmental trajectory of the frontal N1 component. The results obtained revealed a linear developmental trajectory, suggesting a slower maturation than ERP components showing an inverse developmental trajectory. This delayed maturation compared to other ERP components would be related to the delayed frontal maturation with regard to other brain areas (Giedd et al., 2009). Analyzing changes during development related to cerebral mechanisms of visuo-spatial WM, Farber and Beteleva (2011) compared two groups of children (7–8 and 9–10 years old) during the performance of a task consisting of stimuli with a delay between them. These authors found differences related to age in the involvement of several cortical areas in the formation and retention of the short-term memory trace of the reference stimulus, and during comparison of the short-term trace with the test stimulus presented. In both groups, WM was related to an increase in N1 amplitude in posterior visual areas, but not to frontal N1 (categorized as N2 by these authors). P1 is the earliest endogenous component modulated by selective attention (Mangun and Hillyard, 1988). This component is a positive occipital wave that occurs between 80 and 120 ms after stimulus onset. This component presents higher amplitudes in children and decreases slowly through adolescence (Segalowitz et al., 2010), a result also obtained in our study. The developmental change in P1, with the amplitude decreasing with age, suggests increased efficiency in early visual processing throughout childhood and could reflect pruning and/ or mielinization increment in visual cortical areas. In a WM developmental study, Beteleva et al. (2009), comparing adults to 7–8 years old children, found that children presented a longer duration and amplitude in P1 compared to adults, and this result is replicated for amplitude in the present report. The fact that, eliminating the age effect, P1 presents correlations with errors, indicates that deviation from the P1 amplitude age trend is related to an increased number of errors. Therefore, P1 amplitude would not only reflect the population's

4. Discussion The results indicate that there is a generalized reduction in the amplitude of all the ERPs with age that appears during the DMTS task, the most widely used experimental task to assess WM. The reduction follows an inverse or linear regression during development. Once the effects of age are eliminated, the total error residuals are correlated with the age residuals for frontal N2 and posterior P1 and LPC components, suggesting that frontal and posterior maturation are critical for error control on the WM task. Individual variability in RT was not affected by any ERP residual amplitude. ERPs, after controlling 223

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

The developmental trajectory of the slow wave related to WM during DMTS tasks in children has been previously described (Barrigapaulino et al., 2014). In addition, a Contralateral Delay Activity (CDA) has been documented in adolescents between 12 and 16 years old on a visuo-spatial WM task (Spronk et al., 2013). The authors observed an amplitude increase in this ERP when the number of items increased in the memory display in both groups, adolescents and adults, with a higher CDA amplitude in adolescents than in adults. CDA was also described in a group of children between 10 and 13 years old, and it was related to a top-down attentional control necessary to sustain the items temporarily in WM (Sander et al., 2011). Studying the development of this component with age, Farber and Beteleva (2011) also found that, as in adults, 9–10 year-old children presented great differences in late components in several phases of WM functioning: stimulus formation and retention in WM accompanied by an amplitude increase in negativity in the interval between 300– 500 ms, more accentuated in the fronto-central areas; whereas the comparison with the trace stored in memory was related to a slow positivity. The authors argued that the increase in negativity was associated with the encoding of the information about the object. The developmental trajectories of the slow wave observed in the present study, and analyzed in the later temporal window of the retention period (1800–2400 ms), showed an amplitude decrease in the positivity in parietal areas (Pz) as age increased, as well as an amplitude decrease in the negative slow wave in the occipital region (O2). The negativity decrease in the posterior region could be due to synaptic pruning, as it is well known that cerebral waves’ amplitudes decrease with the elimination of synaptic connections between neurons as the child develops (Whitford et al., 2007; Capilla et al., 2004), whereas the accentuated decrease in positivity in fronto-central areas could be due to a maturation of anterior areas, not only due to synaptic pruning, but also as a reorganization of the cortical layers as age increases (Bender et al., 2005), given the change in polarity from positive to negative. However, the positivity found in the anterior region in the youngest subjects could also be explained, at least in part, as the positive side of the posterior dipoles that generate the negative slow wave. We have previously analyzed this aspect, which can be consulted in detail in Barriga-Paulino et al. (2014). Regarding offset ERPs, as far as we know, there is no evidence in the literature that describes the developmental trajectories of these components. In this study, the P1 and P2 offsets appeared in the parieto-occipital region and presented higher amplitudes in children than in adults.

maturational trend, but also the individual maturational state. Regarding P2, in the present investigation we found that its topography moved to more parietal electrodes with increasing age, and amplitude decreased with age. In the study conducted by Coch et al. (2005), the authors did not find developmental effects in P2 amplitude and latency, although they found changes in the topographical distribution of P2 with age. The results found by Coch et al. (2005) suggested that this ERP indexes a processing related to some aspect of the visual system, which matures early during the developmental process. The present topographical results suggest a shifting of P2 from occipital to parietal sites with age, as reported by Coch et al. (2005), whereas a decrease in P2 amplitude with age was obtained. The most parsimonious explanation for this anterior shifting of the P2 component would be the reduction in amplitude of the simultaneous N2 component, which would allow a more anterior topographical expression of the P2 positivity, given the principle of superposition of electric fields generated by simultaneously active dipoles. The fronto-central N2 component obtained in the present results showed a decrease in amplitude with age, and once the effects of age were eliminated, it presented correlation with the age residuals of behavioral errors. The N2 component in channel selection attentional paradigms has been related to the orienting reaction needed to facilitate stimuli processing (Ridderinkhof and van der Stelt, 2000). The selection mechanism indexed by N2 seems to be active from 7 years old on (van der Stelt, 1998). Although in the encoding phase there were no competing distractors, the need for engagement when processing current stimuli for encoding in the WM process is obvious. The fact that, once age is ruled out, a relationship remains with the error residuals indicates that the individual maturation of the neural network indexed by the N2 component, beyond biographical age, is important for good WM performance. This result indicates that the control of task performance is strongly associated with frontal areas, where the executive functions are processed. Moreover, as the frontal area is the last cerebral region to mature, it can be argued that there were more errors committed at younger ages due to frontal immaturity (Giedd et al., 2009). In the present results, an LPC was obtained that decreased in amplitude with increasing age. This LPC presented a similar latency to the P3b. The P3b component has been related to working memory operations in the "context updating" hypothesis, where the P3b indexes the renewal of the WM contents (Duncan-Johnson and Donchin, 1982), or in the "neuro-inhibition" hypothesis, which suggests an inhibitory role in order to facilitate memory operations on currently processed items (Polich, 2007). A decrease in visual P3b amplitude with age has been reported (Courchesne, 1978; Stige et al., 2007; Flores et al., 2009; Rojas-Benjumea et al., 2015). The present results showing a decrease with age suggest that the posterior areas, where most of the LPC were obtained, are also maturing during childhood and adolescence. The prominent presence of the LPC component clearly indicates the importance of this component in the encoding phase of WM. The fact that, eliminating the age effect, LPC presents correlations with errors, indicates that deviation from the LPC amplitude age trend is related to an increased number of errors.

4.3. Matching phase The ERP developmental trajectories in the matching phase were very similar to those observed in the stimulus encoding phase. N1 and N2 appeared in the fronto-central region, and their amplitudes were higher in children than in adults. In the parietal and occipital regions, C1, P1, P2 and LPC appeared and also decreased their amplitudes with age. Pinal et al. (2014) also observed that both phases (encoding and recovery) originated a similar cerebral electric activity. Memory recovery has been conceptualized as involving re-activation of neural conjuncts that were initially used to encode the event. Thus, Friedman and Johnson (2000) used electrophysiological evidence to show activation in sensory cortical regions during memory recovery that reflected active processing also during encoding. In Farber and Beteleva's (2011) investigation comparing 2 groups of children (7–8 versus 9–10 years old) on a visual stimuli matching task, large differences were observed between the 2 groups in the late ERPs corresponding to the cognitive processes. In the 7–8-year-old group, the presentation of both stimuli (S1, S2) leads to an amplitude increase in LPC in the posterior area, with a maximum increase in the 300–800 ms period in the parietal zone. At the age of 9–10 years, the

4.2. Retention phase During the retention phase of the stimulus, all age groups showed a slow wave. However, the youngest children (6–9 years) showed a positive slow wave in the central region with decreasing amplitude, switching to negativity as age increased. In this age group, the posterior region showed an inverse trajectory, as a negative slow wave appeared and became less negative as age increased. Thus, we can conclude that the slow wave, in agreement with what Ruchkin et al. (1997) described, is associated with the active maintenance of information in WM, and consists of a neurophysiological signal present in children from 6 years of age. 224

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

pattern of maturation also influences performance beyond age. It is important to note that the ERP residuals vs. age are not an additional predictor beyond the age of 17, indicating that it is not the amplitude of the component per se, but rather the way the components’ amplitude deviated from the developmental trajectory at young ages, or how mature a given subject is, as indicated by the ERP amplitude criteria. The existence of a global factor explaining age plus individual variability related to maturation has been described in behavioral results on WM tasks (Barriga-Paulino et al., 2016), including DMTS, Oddball (working memory updating), and the three components of the Baddeley and Hitch WM model (Baddeley and Hitch, 1974). Now this individual behavioral variability component can be visualized in the individual maturation of ERPs. The idea of an individual brain maturational age beyond the biographical age, and also related to the maturational cognitive status, has been stressed recently. Shaw et al. (2006) showed a relationship between the cortical thickness developmental trajectory and IQ. A multivariate parameter extracted from fractional anisotropy correlated with Working Memory and numerical abilities at school entry ages (Ullman and Klingberg, 2016), suggesting that the brain's maturational state can be inferred from brain connectivity, and that this individual maturational state has functional behavioral implications. Erus et al. (2014) demonstrated that an estimated Brain Developmental Index (BDI), derived from a support vector regression procedure of MRI multimodal images, was able to separate subjects with low, normal or high cognitive status. The predictive power of BDI was related to the speed of responses rather than to accuracy. The present results, by improving the prediction of WM performance beyond biographical age, when the first PCA component scores extracted from the age-residuals of ERPs are included in the predictive model, suggest that ERPs can provide information about the individual maturational state. One main difference from the Erus et al. (2014) results is that in our case, the predictive value of the PCA component scores was related to accuracy rather than to speed. One possible reason would be that, in the ERP analysis, the wrong responses are excluded, whereas in BDI, obtained from static neuroanatomical images, this is, of course, not possible. Including errors in the ERP analysis could have slightly changed the ERPs and produced a different outcome in the prediction of RTs based on age and the first PCA score. The very different nature of functional ERPs and neuroanatomical MRI images is also another possible source for these differences. The present study is cross-sectional, which means it is not possible to disentangle developmental and cohort effects. However, as Pokémons and Digimons have been present on the market from 1998 and 1999, respectively, the influence of these images could have been present in most of the experimental subjects, reducing the cohort effect, which, as mentioned above, cannot be disentangled from true maturational effects. On the other hand, given the different difficulty of the task for children, adolescents and adults, it is possible that part of the obtained ERPs amplitude modulation could be due to the different engagement in the task, as a function of age, rather than to the working memory operation.

increase in this complex was only found for S2. During the time period that corresponded to the LPC, classified by Beteleva et al. (2009) as the P3b component, two activation peaks were found: an early peak with a latency between 250–450 ms and a late peak in the time period between 450–850 ms. According to the authors, the early peak would be related to the identification of the object, whereas the late peak would be related to memory recovery. These peaks were not identified in the present experiment. In the present results, the LPC in the encoding phase presented lower amplitude than the LPC in the matching phase for all the age groups. This could be considered a WM side of the so-called old-new effect in memory recognition experiments, where stimuli previously presented show higher LPC amplitude than those presented the first time (Rugg, 1995; Rugg and Doyle, 1992). As the S2 always displays the S1 stimulus, it corresponds to a repetition of a presented stimulus, and the increase in LPC in the matching phase would be homologous to this effect. The increase in LPC amplitude has been related to the process of recovery in memory, which certainly has to occur during the matching phase. Finally, it is important to indicate that the memory matching process must co-occur with the selection of the recognized item in order to produce the required motor response. We have already described the presence of a parietal contralateral to the recognized items that would be related to the recognized selected item (BarrigaPaulino et al., 2015). The latency of this component is around the P2 latencies and would correspond to the initial phases of recognition that would end at the LPC component. 4.4. General trend of maturation of ERP components As discussed previously, all the recorded components showed a decrease in amplitude, probably produced by the pruning process, which would produce a decrease in ERP amplitude, due to a reduction in the number of active synapses to produce local field potentials; in fact, a relationship between cortical thickness and slow brain rhythms has been obtained (Whitford et al., 2007). It must be noted that the decrease occurs not only in frontal sites, but also in occipital ones. It has been proposed that, although the most prominent gradient of brain maturation is antero-posterior, it also occurs from low order association areas to high order association areas inside lobules (Giedd et al., 2009). This inside-lobule maturation trend would explain the reduction in amplitude with age of the posterior components. Hence, it must be noted that posterior components such as the C1, P1, LPC and Slow Negativity presented an inverse relationship with age, whereas anterior components such as N1 and N2, but also posterior P2, presented a linear relationship. These results suggest that posterior components presented earlier maturation, given the steeper slope of inverse curves at lower ages than in the anterior components, where the slope is constant across the analyzed ages. This faster maturation of posterior sites corresponds to the posterior-anterior gradient of maturation described by neuroanatomical methods (Giedd et al., 2009). Although part of the generalized reduction in ERP amplitude could be due to the increase in skull thickness and mineralization, which would produce a reduction in electrical conductivity, magnetoencephalographic recordings recently showed that a reduction in neural sources occurs during maturation (Gómez et al., 2016). Moreover, the value of ERPs in improving the prediction of error performance (model 2) on the DMTS task suggests that the reduction in ERP generators is not only due to passive skull properties.

Acknowledgments This work was supported by the Spanish Ministry of Science and Innovation, grant PSI2013-47506-R, funded by the FEDER program of the UE, and Junta de Andalucía, grant number CTS-153. Catarina Isabel Barriga-Paulino was supported by a PhD fellowship (SFRH/BD/72469/2010) funded by Fundação para a Ciência e a Tecnologia, Portuguese Ministry of Education and Science.

4.5. Performance predictive model As expected, age was the most important predictor of performance on both errors and RTs. However, for ages under 17, the PCA global component extracted from the ERP age residual becomes an independent predictor of error performance, indicating that the individual

Appendix A. Supporting information Supplementary data associated with this article can be found in the 225

Neuropsychologia 95 (2017) 215–226

C.I. Barriga-Paulino et al.

Johnson, R., Jr., 1993. On the neural generators of the P300 component of the eventrelated potential. Psychophysiology 30, 90–97. Kaan, E., Carlisle, E., 2014. ERP indices of stimulus prediction in letter sequences. Brain Sci. 4, 509–531. Mangun, G.R., Hillyard, S.A., 1988. Spatial gradients of visual attention: behavioral and electrophysiological evidence. Electroencephalogr. Clin. Neurophysiol. 70, 417–428. Mecklinger, A., Pfeifer, E., 1996. Event-related potentials reveal topographical and temporal distinct neuronal activation patterns for spatial and object working memory. Cogn. Brain Res. 4, 211–224. Pinal, D., Zurrón, M., Díaz, F., 2014. Effects of load and maintenance duration on the time course of information encoding and retrieval in working memory: from perceptual analysis to post-categorization processes. Front. Hum. Neurosci. 8, 165. Polich, J., 2007. Updating P300: an integrative theory of P3a and P3b. Clin. Neurophysiol. 118, 2128–2148. Ridderinkhof, K.R., van der Stelt, O., 2000. Attention and selection in the growing child: views derived from developmental psychophysiology. Biol. Psychol. 54 (1–3), 55–106. Roalf, D.R., Gur, R.E., Ruparel, K., Calkins, M.E., Satterthwaite, T.D., Bilker, W.B., Hakonarson, H., Harris, L.J., Gur, R.C., 2014. Within-individual variability in neurocognitive performance: age and sex-related differences in children and youths from ages 8 to 21. Neuropsychology 28 (4), 506–518. Rodriguez-Martinez, E.I., Barriga-Paulino, C.I., Zapata, M.I., Chinchilla, C., LópezJiménez, A.M., Gómez, C.M., 2012. Narrow band quantitative and multivariate electroencephalogram analysis of peri-adolescent period. BMC Neurosci. 13, 104–126. Rodriguez-Martinez, E.I., Barriga-Paulino, C.I., Rojas-Benjumea, M.A., 2013. Comaturation between the Spontaneous Electroencephalogram and Working Memory. Neurosci. Lett. 550, 134–138. Rojas-Benjumea, M.A., Quintero-Gallego, E., Zozaya, L., Barriga-Paulino, C.I., Gómez, C.M., 2013. Children cautious strategy and variable maturation time window for responding in a visual search task. Psychology 4 (1), 19–32. Rojas-Benjumea, M.A., Sauqué-Poggio, A., Barriga-Paulino, C.I., Rodríguez-Martínez, E.I., Gómez, C.M., 2015. Development of behavioral parameters and ERPs in a novel-target visual detection paradigm in children, adolescents and young adults. Behav. Brain Funct. 11, 22. Ruchkin, D.S., Johnson, R., Jr., Canoune, H., Ritter, W., 1990. Short-term memory storage and retention: an event-related brain potential study. Electroencephalogr. Clin. Neurophysiol. 76, 419–439. Ruchkin, D.S., Johnson, R., Jr., Grafman, J., Canoune, H., Ritter, W., 1992. Distinctions and similarities among working memory processes: an event-related potential study. Cogn. Brain Res. 1, 53–66. Ruchkin, D.S., Johnson, R., Jr., Grafman, J., Canoune, H., Ritter, W., 1997. Multiple visuospatial working memory buffers: evidence from spatiotemporal patterns of brain activity. Neuropsychologia 35 (2), 195–209. Rugg, M.D., 1984a. Event-related potentials and the phonological processing of words and non-words. Neuropsychologia 22 (4), 435–443. Rugg, M.D., 1984b. Event-related potentials in phonological matching tasks. Brain Lang. 23, 225–240. Rugg, M.D., Doyle, M.C., 1992. Event-related potentials and recognition memory for lowfrequency and high-frequency words. J. Cogn. Neurosci. 4, 69–79. Rugg, M.D., 1995. ERP studies of memory. In: Rugg, In.M.D., Coles, M.G.H. (Eds.), Electrophysiology of mind. Oxford University Press, New York, 132–170. Sander, M.C., Werkle-Bergner, M., Lindenberger, U., 2011. Contralateral delay activity reveals life-span age differences in top-down modulation of working memory contents. Cereb. Cortex 21, 2809–2819. Segalowitz, S.J., Unsal, A., Dywan, J., 1992. Cleverness and wisdom in 12-yearolds: electrophysiological evidence for late maturation of the frontal lobe. Dev. Neuropsychol. 8, 279–298. Segalowitz, S.J., Santesso, D.L., Jetha, M.K., 2010. Electrophysiological changes during adolescence: a review. Brain Cogn. 72, 86–100. Shaw, P., Greenstein, D., Lerch, J., Clasen, L., Lenroot, R., Gogtay, N., Evans, A., Rapoport, J., Giedd, J., 2006. Intellectual ability and cortical development in children and adolescents. Nature 440, 676–679. Shing, Y.L., Brehmer, Y., Heekeren, H.R., Bäckman, L., Lindenberger, U., 2016. Neural activation patterns of successful episodic encoding: reorganization during childhood, maintenance in old age. Dev. Cogn. Neurosci. 20, 59–69. Spronk, M., Vogel, E.K., Jonkman, L.M., 2013. No Behavioral or ERP Evidence for a Developmental Lag in Visual Working Memory Capacity or Filtering in Adolescents and Adults with ADHD. PLoS ONE 8 (5), e62673. Stige, S., Fjell, A.M., Smith, L., Lindgren, M., Walhovd, K.B., 2007. The development of visual P3a and P3b. Dev. Neuropsycholy 32 (1), 563–584. Ullman, H., Klingberg, T., 2016. Timing of White Matter Development Determines Cognitive Abilities at School Entry but Not in Late Adolescence. Cereb. Cortex, 1–7. Van der Stelt, O., Kok, A., Smulders, F.T.Y., Snel, J., Gunning, W.B., 1998. Cerebral event-related potentials associated with selective attention to color: developmental changes from childhood to adulthood. Psychophysiology 35, 227–239. Whitford, T.J., Rennie, C.J., Grieve, S.M., Clark, C.R., Gordon, E., Williams, L.M., 2007. Brain maturation in adolescence: concurrent changes in neuroanatomy and neurophysiology. Hum. Brain Mapp. 28 (3), 228–237. Wolach, I., Pratt, H., 2001. The mode of short-term memory encoding as indicated by event-related potentials in a memory scanning task with distractions. Clin. Neurophysiol. 112, 186–197.

online version at doi:10.1016/j.neuropsychologia.2016.12.026. References Baddeley, A.D., Hitch, G., 1974. Working memory. In: Bower, G.H. (Ed.), The Psychology of Learning and Motivation: Advances in Research and Theory (8. Academic Press, New York, 47–89. Barrett, S.E., Rugg, M.D., Perrett, D.I., 1988. Event-related potentials and the matching of familiar and unfamiliar faces. Neuropsychologia 26 (1), 105–117. Barrett, S.E., Rugg, M.D., 1989. Event-related potentials and the semantic matching of faces. Neuropsychologia 27 (7), 913–922. Barrett, S.E., Rugg, M.D., 1990. Event-related potentials and the phonological matching of picture names. Brain Lang. 38, 424–437. Barriga-Paulino, C.I., Flores, A.B., Gómez, C.M., 2011. Developmental Changes in the EEG Rhythms of Children and Young Adults Analyzed by Means of Correlational, Brain Topography and Principal Component Analysis. J. Psychophysiol. 25 (3), 143–158. Barriga-Paulino, C.I., Rodríguez-Martínez, E.I., Rojas-Benjumea, M.A., Gómez, C.M., 2014. Slow wave maturation on a visual working memory task. Brain Cogn. 88, 43–54. Barriga-Paulino, C.I., Rodríguez-Martínez, E.I., Rojas-Benjumea, M.A., Gómez, C.M., 2015. Electrophysiological evidence of a delay in the visual selection process in youngest children. Front. Hum. Neurosci. 9, 622. Barriga-Paulino, C.I., Rodríguez-Martínez, E.I., Rojas-Benjumea, M.A., Gómez, C.M., 2016. Principal component analysis of working memory variables during child and adolescent development. Spanish J. Psychol. 19, e62. Bender, S., Weisbrod, M., Bornfleth, H., Resch, F., Oelkers-Ax, R., 2005. How do children prepare to react? Imaging maturation of motor preparation and stimulus anticipation by late contingent negative variation. NeuroImage 27, 737–752. Beteleva, T.G., Sinitsyn, S.V., Farber, D.A., 2009. Age-Related Specificity of the Processing of Visual Information in the System of Working Memory. Hum. Physiol. 35 (6), 672–683. Capilla, A., Romero, D., Maestú, F., Campo, P., Fernández, S., González-Marqués, J., Fernández, A., Ortiz, T., 2004. Ekmergencia y desarrollo cerebral de las funciones ejecutivas. Actas Esp. De. Psiquiatr. 32 (2), 377–386. Clark, V.P., Fan, S., Hillyard, S.A., 1995. Identification of early visually evoked potential generators by retinotopic and topographic analyses. Hum. Brain Mapp. 2, 170–187. Coch, D., Skendzel, W., Grossi, G., Neville, H., 2005. Motion and color processing in school-age children and adults: an ERP study. Dev. Sci. 8 (4), 372–386. Constantinidis, C., Klingberg, T., 2016. The neuroscience of working memory capacity and training. Nat. Rev. Neurosci. 17, 438–449. Courchesne, E., 1978. Neurophysiological correlates of cognitive development: changes in long latency event-related potentials from childhood to adulthood. Electroencephalogr. Clin. Neurophysiol. 45, 468–482. Donchin, E., Coles, M.G.H., 1988. Is the P300 component a manifestation of context updating? Behav. Brain Sci. 11, 357–374. Duncan-Johnson, C.C., Donchin, E., 1982. The P300 Component of the Event-Related Brain Potential as an Index of Information Processing. Biol. Psychol. 14, 1–52. Dunn, B.R., Dunn, D.A., Languis, M., Andrews, D., 1998. The relation of ERP components to complex memory processing. Brain Cogn. 36, 355–376. Erus, G., Battapady, H., Satterthwaite, T.D., Hakonarson, H., Gur, R.E., Davatzikos, C., Gur, R.C., 2014. Imaging Patterns of Brain Development and their Relationship to Cognition. Cereb. Cortex 25 (6), 1676–1684. Evans, K.M., Federmeier, K.D., 2009. Left and right memory revisited: electrophysiological investigations of hemispheric asymmetries at retrieval. Neuropsychologia 47, 303–313. Farber, D.A., Beteleva, T.G., 2011. Development of the Brain's Organization of Working Memory in Young School children. Hum. Physiol. 37 (1), 1–13. Flores, A., Digiacomo, M., Meneres, S., Trigo, E., Gómez, C.M., 2009. Development of preparatory activity indexed by the Contingent Negative Variation in Children. Brain Cogn. 71, 129–140. Foxe, J.J., Simpson, G.V., 2002. Flow of activation from V1 to frontal cortex in humans – A framework for defining “early” visual processing. Exp. Brain Res. 142, 139–150. Friedman, D., Johnson, R. Jr, 2000. Event-Related Potential (ERP) Studies of Memory Encoding and Retrieval: a Selective Review. Microsc. Res. Tech. 51, 6–28. Ghetti, S., Bunge, S.A., 2012. Neural changes underlying the development of episodic memory during middle childhood. Dev. Cogn. Neurosci. 2, 381–395. Giedd, J.N., Lalonde, F.M., Celano, M.J., White, S.L., Wallace, G.L., Lee, N.R., Lenroot, R.K., 2009. Anatomical brain magnetic resonance imaging of typically developing children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry 48 (5), 465–470. Gómez, C.M., Clark, V.P., Fan, S., Luck, S.J., Hillyard, S.A., 1994. Sources of AttentionSensitive Visual Event-Related Potentials. Brain Topogr. 7 (1), 41–51. Gómez, C.M., Rodríguez-Martínez, E.I., Fernández, A., Maestú, J., Poza, J., Gómez, C., 2016. Absolute Power Spectral Density changes in the magnetoencephalographic activity during the transition from childhood to adulthood. Brain Topography. In press. Guttman, L., 1954. A new approach to factor analysis: the radex. In: Lazarsfeld, P.F. (Ed.), Mathematical Thinking in the Social Sciences. Columbia University Press, New York, 258–348. Huttenlocher, P.R., 1990. Morphometric study of human cerebral cortex development. Neuropsychologia 28 (6), 517–527.

226