Behavioral and neurophysiological evidence for increased cognitive flexibility in late childhood

Executive functions, like the capacity to control and organize thoughts and behavior, develop from childhood to young adulthood. Although task switching and working memory processes are known to undergo strong developmental changes from childhood to adulthood, it is currently unknown how task switching processes are modulated between childhood and adulthood given that working memory processes are central to task switching. The aim of the current study is therefore to examine this question using a combined cue- and memory-based task switching paradigm in children (N = 25) and young adults (N = 25) in combination with neurophysiological (EEG) methods. We obtained an unexpected paradoxical effect suggesting that memory-based task switching is better in late childhood than in young adulthood. No group differences were observed in cue-based task switching. The neurophysiological data suggest that this effect is not due to altered attentional selection (P1, N1) or processes related to the updating, organization, and implementation of the new task-set (P3). Instead, alterations were found in the resolution of task-set conflict and the selection of an appropriate response (N2) when a task has to be switched. Our observation contrasts findings showing that cognitive control mechanisms reach their optimal functioning in early adulthood.


Results
Behavioral Results. Switch costs. Switch costs (SC) were obtained for correct trials by subtracting response times (RTs) during task repetition from RTs during task switching (cue-based SC = RT cue-based-switching -RT cue-based-repetition; memory-based SC = RT memory-based-switching -RT memory-based-repetition).
Post hoc t-tests, analyzing both groups separately, revealed significant differences in switch costs between the two conditions in young adults (t[24] = 6.27, p < 0.001), indicating significantly higher switch costs during the memory-based block (133 ms ± 11) as compared to the cue-based block (53 ms ± 13). In contrast, children showed no significant effect (p = 0.877), indicating comparable switch costs between the memory-(74 ms ± 16) and the cue-based block (71 ms ± 10). Post hoc t-tests, analyzing between group effects, revealed significant switch costs differences between groups in the memory-based block (t [48] = − 3.01, p = 0.004), indicating significantly reduced switch costs in children (74 ms ± 16) as compared to young adults (133 ms ± 11). In contrast switch costs in the cue-based block did not differ significantly (p > 0.2) between children (71 ms ± 10) and young adults (53 ms ± 11).
Supporting these results further, we observed a significant positive correlation between memory-based switch costs and age (r = 0.34, p = 0.009), indicating significantly increased switch costs with increasing age. In contrast no significant correlation (p > 0.3) was observed for age and switch costs during the cue-based condition (refer Fig. 1B). Finally, we tested whether switch-costs during cue-and switch-costs during memory-based conditions are different, by testing both against each other. We obtained no significant effect (p = 0.76), indicating that the effects are different from each other.

Figure 2. Attentional ERPs.
Grand average event-related potentials (ERPs) and voltage maps at occipitotemporal electrodes (P7, P8), depicting P1 and N1 components for the two blocks (cue-and memory-based) and conditions (switching and repetition) in late childhood (left) and young adulthood (right).

Discussion
We investigated in what way neurophysiological processes underlying cognitive flexibility are affected by age and more specifically to what extent memory-and cue-based task switching are differently modulated between late childhood and young adulthood.
Parts of the behavioral results are well in line with the literature showing increased switch costs as well as generally increased RTs during memory-versus cue-based switching 20,34 . In addition we observed lower accuracies during memory-versus cue-based blocks 37,38 as well as during switching versus repetition trials 37,38 . However, the results on differences between children and young adults in the memory-based switching block were unexpected, since children performed better in memory-based task switching than young adults and there were no behavioral group differences in the cue-based task switching condition. We observed no group effects or interactions in the accuracy data, indicating that children show the same accuracy as young adults, but are able to perform the task significantly faster. This suggests that any group differences observed do not arise due to task difficulty 39 , but rather stem from development-related differences in the execution of the task. Interestingly and further corroborating the interpretation of a developmental effect, we observed a positive correlation between memory-based switch costs and participant age. This was absent for cue-based switch costs, indicating an increase of switch costs with increasing age under memory load. Accordingly, the behavioral data indicate that cognitive flexibility triggered by working memory processes seems to be highly affected by developmental aspects, while cue-based switching and thus processes of cognitive flexibility without working memory load seem to represent age-independent processes between late childhood and young adulthood 13 . Research on the task switching paradigm in childhood is still limited. However, although results are inconsistent regarding the exact age, most studies indicate similar switch costs in cue-triggered task switching paradigms between children and adults 13,16,18,19,37,40 , as also found in the current study. These studies show that cue-triggered switch costs decrease significantly in children from 7 to 11 to 15 years 16 and are similar low between the age of 13-17 years as it is the case in young adults 13,14,18,40 . It has been suggested that theoretical explanations of cognitive mechanisms, like the ability to maintain and manipulate two mental task sets which are responsible for switch costs in adults can also be applied to children 14 . Accordingly, the ability to switch between rules was suggested to reach mature level during late childhood 40 . Interestingly, one other study 41 , solely analyzing children in 5 consecutive age groups from 5 to 13 years, indicated that switch costs and errors decrease in children between 5 and 8 years of age, but are similar in children between 9 and 13 years. Moreover another study 42 comparing children (mean age 9.4 years) to young (mean age 21.5 years) and older adults (mean age 65.3 years) observed similar switch costs in a cue-triggered switching paradigm in all three age groups. This confirms the assumption that cognitive flexibility measured via cue-based switching seems to represent an executive function which is developed relatively early in childhood and remains stable from childhood to old adulthood 42 .
The neurophysiological data point to the mechanisms underlying the behavioral modulations and differential developmental effects. Neurophysiological parameters known to reflect perceptual or attentional processes (i.e. the P1 and the N1) 43,44 did not show interactions with condition, block and group, suggesting that these processes do not explain the observed behavioral effects. However, we observed an interaction between block, condition and electrode in the P1 in both age groups, suggesting that early perceptual processes during stimulus perception across participants seem lateralized to the left hemisphere. With respect to the factors block and condition, we further observed that attention to switch and repetition trials is differently modulated depending on block: with higher attention to switch trials in cue-based blocks and higher attention to repetition trials in memory-based blocks. This modulation points to distinct mechanisms underlying cue-and memory-based blocks.
Moreover, in line with other developmental studies 45,46 we observed a main effect of group for the N1, indicating larger N1 amplitudes in young adults as compared to children. This has been suggested to represent maturational changes from childhood to young adulthood 45 . In addition to the group effect and in line with P1 results, we observed an interaction with block and electrode in N1, indicating increased lateralized attentional allocation processes 47 during memory-vs cue-based blocks. Both interactions with electrode for P1 and N1 suggest a distinct and age-independent left hemispheric distribution of early attentional and perceptual processes during task switching.
Finally we observed interactions for condition and block as well as for group and condition for the P3, indicating increased P3 amplitudes during repetition versus switching trials, in both groups and in both blocks. This indicates that processes related to the updating, organization, and implementation of a new task-set 27,48-50 are especially pronounced during repetition trials, which indicates a facilitated access of repetitions independent of age or block.
Of particular importance are the findings on the N2. Interactive effects of age group and cue-vs. memory-based task switching are in line with the behavioral data. While there were no N2 modulations between memory-and cue-based conditions in children, in young adults, the N2 amplitude during memory-based repetition was more negative as compared to memory-based switching trials. The findings in young adults are contradictory to the literature, in which it is usually shown that the N2 is stronger during switching than during repetition trials 51,52 (at least when classical cue-based task switching paradigms are applied). During task switching, the N2 has been interpreted as an index of the resolution of task-set conflict and the selection of an appropriate response when a task has to be switched (ref. 53). However, it has been suggested that for these switching processes inhibitory control mechanisms are important 54,55 . It is possible that under memory-based task switching these inhibitory control mechanisms cannot fully or effectively unfold because the increased working memory load already demands restricted resources of the central executive. Inefficient inhibitory control mechanisms of a previous task set are known to lead to high switching costs (i.e. compromised cognitive flexibility) 53,54 . This is what is reflected in the behavioral data pattern in young adults. Concerning the child group and based on the above interpretation, it seems that the central executive is less strained in children than in young adults during memory-based task switching. This is an unexpected finding, because working memory performance usually peaks between 24 and 35 years of age 56 and cognitive abilities in general rise steeply from infancy to young adulthood with a peak of cognitive control performance in the late teens and early twenties 12 . These changes are often attributed to the effects of synaptic elimination during development. It has been suggested that the decrease of synaptic density leads to reorganization and development of specialized and more efficient networks 19 . There is a peak of synaptic density during early childhood which is followed by major synapse elimination processes throughout early and mid-adolescence 57 . This process, in combination with the late maturation of the PFC, has been suggested to underlie the finding that internally guided behavior, working memory and organizational skills do not reach full functional capacity until late adolescence, or young adulthood, respectively 12,57 . The obtained results are in sharp contrast to these assumptions, which have been supported by different lines of research. Conclusions from the current results have thus to be drawn carefully and require further studies. However, aside from findings which are in line with the synaptic pruning hypothesis, several results from longitudinal MRI studies 58,59 might account for the observed results. It has been observed that cortical gray matter volume increases in childhood and peaks at the age of 11-12 years, followed by a decline during post-adolescence 58 . It has also been shown that cortical gray matter volume was observed to be positively correlated with cognitive flexibility 60 ; i.e. the larger the grey matter volume is in prefrontal cortices, the better cognitive flexibility is developed. It may be speculated that similar processes underlie the observed paradoxical finding of increased memory-based task switching in children, compared to young adults. Further, the execution of memory vs cue -based switching might be associated with different subdivisions of the PFC, which in turn may be characterized by diverging maturational trajectories. It cannot be ruled out that a combination of processes, i.e. cognitive flexibility and working memory, induces a different dynamic that is currently not covered and explainable by existing theories of the development of cognitive functions from childhood to adulthood.
In summary, we obtained an unexpected, paradoxical effect suggesting that memory-based task switching performance is better in late childhood than young adulthood. No group differences were observed in cue-based task switching. The neurophysiological data suggest that this effect is not due to altered attentional selection or processes related to the updating, organization, and implementation of the new task-set, but to alterations in the resolution of task-set conflict and in the selection of an appropriate response when a task has to be switched. Our observation is in contrast to findings showing that cognitive control mechanisms reach their optimal functioning in early adulthood. Future research should focus on developmental processes of cognitive operations that are a combination of different executive control processes in order to examine in more detail if the combination of processes is related to different developmental trajectories than it is the case for one of these cognitive functions alone.

Participants.
A total of n = 25 healthy young adults (17 females) (M = 23.80 ± 0.66) as well as a group of n = 25 healthy children (13 females) (M = 13.88 ± 0.40) took part in the experiment. All participants were righthanded, had normal or corrected to normal vision, reported no psychiatric or neurological disorders and were free of medication. Written informed consent was obtained from all participants before the test protocol was conducted. The study was approved by the institutional review board of the medical faculty of the TU Dresden and Scientific RepoRts | 6:28954 | DOI: 10.1038/srep28954 realized in accordance with the Declaration of Helsinki. Participants performed a cue-based and a memory-based task switching block in a counterbalanced order.
Task switching paradigm & procedure. For stimulus presentation, response recording and EEG triggers "Presentation" software (Neurobehavioral systems, Inc., Version 14.9) was used. Stimuli consisted of digits from 1-9, excluding the number 5, which were presented in white font, centrally on a black computer screen. Digits were either presented in a small or large size (font size 50 vs font size 80) on the screen. The paradigm consisted of two settings: a cue-and a memory-based block. The task was similar to the paradigm applied by Gajewski et al. 20 . Here, the task switching paradigm (see Fig. 4) consisted of two sequences: a cue-and a memory-based block. Within the cue-based block, a cue indicates the valid rule out of the following three alternating rules: numeric rule (indicated by the cue "NUM", as shortcut for the German word "Numerisch", numeric), parity rule (indicated by the cue "GER", as shortcut for the German word "Geradzahligkeit", parity) and font-size rule (indicated by the cue "SCH" as shortcut for the German word "Schriftgröße", font size). In the numeric rule participants have to decide whether the presented target is greater or less than 5, in the parity rule, whether the presented target is odd or even and in the font-size rule, whether the target is large or small. The rules and thus the cues alternated randomly with an equal distributed probability of cue occurrence, which was set to 33.3% and a frequency of task switching, which was set to 50%. During the memory-based block, participants were told to switch the rule after every third trial out of memory: starting with the numeric task, which needed to be repeated three times. Afterwards, the parity task and then the font-size task were valid, with each task needing to be repeated three times. Then, participants were required to start afresh (e.g. NUM, NUM, NUM, GER, GER, GER, SCH, SCH, SCH, NUM, NUM, NUM, GER; GER, GER, SCH, SCH, SCH, NUM… , refer to Fig. 4). Similar to the cue-base block, the rules had to be applied with an equally distributed probability (33.3%) but with a frequency of task switching, which was set to 33.3%. During both blocks, participants were allowed to take a self-paced period of rest every 33 trials.
Participants were seated in front of a computer screen and performed one practice block (18 trials per block) and 2 experimental blocks (198 trials per block). Participants were instructed to answer as quickly and accurately as possible via button presses with their left index finger (depending on experimental version e.g. if the digit was smaller than five, had small font size or was uneven), or the right index finger (depending on experimental version e.g. if the digit was larger than five, had a large small size or was even). Response key assignment and the sequence of blocks were counterbalanced across participants.
Each block included an equal number of stimuli (digits) as well as responses. Trials were structured as follows: first, a fixation point in combination with the cue stimulus was presented for 1300 milliseconds (ms). The cue-stimulus remained visible until the digit (target) was presented. The digit appeared at the same position at which the fixation point was presented before. A response had to be given within 2500 ms after target-onset. If no response was given within the 2500 ms, the trial was rated as missed response. 500 ms after the response, a feedback was displayed for 500 ms, which was either a plus (in case of a correct response) or a minus sign (after a wrong response). After the feedback the next cue was shown. The response-cue interval (RCI) was set to ~1500 ms and included the response-feedback delay (500 ms), the feedback (500 ms) and the feedback-cue delay jittered between 400 and 600 ms (mean 500 ms). When during the memory-based block three consecutive errors were made, cues were consecutively presented for the next 3 trials helping participants to remember the correct order. Supplementary analysis 2 provides details on how often this occurred for each condition and group. A test session took about 1 hour including arrival, EEG preparation, task explanation, obtaining consent from both parents and children and task performance. Children and young adults received a financial compensation of 10,-EUR.
EEG recording and analysis. We recorded electroencephalography (EEG) using a 60-channel system (containing the following channels: Cz, FCz, FC1, CP1, Fz, F1, FC3, C3, CP3, P1, AFz, AF3, F5, FC5, C5, CP5, P3, PO1, FP1, FT7, T7, TP7, P7, O1, FT9, TP9, P9, O9, P11, FC2, CP2, CPz, F2, FC4, C4, CP4, P2, Pz, AF4, F6, FC6, C6, Cp6, P4, PO2, Fp2, AF8, FT8, T8, TP8, P8, O2, Oz, FT10, TP10, P10, O10, Iz, P12, AF7). EEG recordings were performed using the Brain Vision Recorder software (Brain Products) with a sampling rate of 500 Hz and electrode impedances under 5 kΩ. Ag/AgCl-electrodes were mounted in an elastic cap and were arranged in equidistant positions. Data analysis was performed using the Brain Vision Analyzer 2 software package. After recording, data were down-sampled to 256 Hz and filtered (band-pass filter from 0.5 to 20 Hz, with a slope of 48 dB/oct each). Raw data were inspected manually in order to reject non-ocular artifacts from the EEG. Afterwards, an independent component analysis (ICA; infomax algorithm) was conducted on the un-epoched data sets in order to remove recurring artifacts. ICA components revealing horizontal and vertical eye movements, blinks and pulse artifacts were manually chosen to be discarded. Afterwards, EEG data was segmented for switch and repeat trials and separately for cue-and memory-based blocks. Only trials with correct responses were analyzed further. The segments started 200 ms before target presentation of the respective trial and ended 1500 ms after its onset. Subsequently, an automated artifact rejection procedure was conducted for all segments. An activity below 0.5 μ V in a 100 ms period and a maximal value difference of 200 μ V in a 200 ms interval were used as rejection criteria. To eliminate the reference potential from the data and to re-reference the data, we applied a current source density (CSD) transformation 61 which results in values for amplitudes in μ V/m 2 . A baseline correction from − 200 ms to −1800 ms prior to target onset was applied; i. e, the baseline was set before the cue onset. Information about the number of trials included in the data analysis after artifact rejection and rejection of incorrect trials can be found in supplementary analysis 3. For statistical analysis we used SPSS (IBM SPSS Statistics, Version 23). In addition, post-hoc tests were Bonferroni-corrected and data were checked for the presence of normal distribution (Kolmogorov Smirnov test, all p > 0.05) as well as for the presence of variance homogeneity (Levene test, all p > 0.05). We analyzed ERP components at the electrodes of their respective maximum amplitudes relative to pre-stimulus baseline. Since it has been observed that that the topographic distribution of the visual P1 and N1 indicate specific lateralized generators in the occipitotemporal gyrus 33 , mean amplitudes were quantified for P1 and N1 at electrodes P7 and P8 (time window P1: 100-130 ms, time window N1: 150-170 ms). Mean amplitudes were also quantified for N2 at fronto-central electrodes (FCz) in the time window from 280 till 320 ms after target onset. Finally, mean P3 amplitudes were calculated at central electrodes (Pz) between 350 and 600 ms. This choice of electrodes was validated using statistical tests as described in 60 . This validation procedure revealed the same electrode locations.