Delayed development of proactive response preparation in adolescents: ERP and EMG evidence

Highlights

* First examination of response-related processing in adolescents; combined ERP+EMG. * Deficient response preparation during adolescence indicated by attenuated CNV. * Inefficient response execution during adolescence (incorrect hand EMG). * Real time evidence that adolescents use reactive vs proactive control.

Abstract

The transition from late adolescence to young adulthood is often overlooked in the cognitive neuroscience literature. However this is an important developmental period as even older adolescents have not yet reached adult level ability on many cognitive tasks. Adolescents (16–17-year olds) and young adults (23–30-year olds) were tested on a cued task switching paradigm specifically designed to isolate response preparation from response execution. A combined ERP and eletromyographic (EMG) investigation revealed that adolescents have attenuated contingent negative variation (CNV) activity during response preparation followed by larger P3b amplitude and EMG activity in the incorrect response hand during response execution. This is consistent with deficient response preparation and a reactive control strategy. Conversely young adults engaged increased response preparation followed by attenuated P3b activity and early EMG activity in the correct response hand during response execution which indicates a proactive control strategy. Through real time tracking of response-related processing we provide direct evidence of a developmental dissociation between reactive and proactive control. We assert that adoption of a proactive control strategy by adolescents is an important step in the transition to adulthood.

1. Introduction

Impulsivity, lack of foresight, and poor decision-making are trademarks of adolescent behaviour (Crone, 2009, Paus, 2005, Steinberg, 2005). Nevertheless as adolescents transition into adulthood they will need to engage appropriate goal directed behaviour despite distracting complex environments. Currently the neural processes responsible for the transition from immaturity in adolescence to goal-directed behaviour in young adulthood have not been clearly established (Andrews-Hanna et al., 2011). Some research suggests that many of the immature behavioural characteristics of adolescence result from lack of cognitive control i.e. ‘the inability to regulate thoughts and actions in accordance with internally represented behavioural goals’ (Andrews-Hanna et al., 2011, Braver, 2012, Manzi et al., 2011).

Recently cognitive control has been dissociated into two components; proactive and reactive control (Braver and Gray, 2007, Jacoby, 1999). According to Braver and Gray's Dual Mechanisms of Control model (DMC) (2007) proactive control refers to a preparatory process that can be sustained over the course of the task whereas reactive control is a transient control process that is implemented directly following the perception of a stimulus. Research suggests that adolescents may use a reactive control strategy for completing complex cognitive tasks, whereas young adults have developed a proactive control strategy (Andrews-Hanna et al., 2011, Manzi et al., 2011). The aim of the current study is to determine the neuro-cognitive mechanisms underlying the developmental proactive control from adolescence to young adulthood.

Importantly, the key difference between reactive and proactive control could lie in differential management of response preparation. According to Aron (2011) the criteria for proactive control have two key elements (1) advance preparation and (2) selective control for a particular response tendency (Aron, 2011). Chen et al. (2010) theorized that the proactive control system affects behaviour by adjusting the threshold for response initiation. For example increased proactive control may be obtained by carefully amending response initiation (e.g. slowing responses). The pre-supplementary motor area (preSMA), the right inferior frontal circuit and the subthalamic nucleus (STN) are found to be involved in both reactive and proactive stopping however, importantly, in proactive stopping this stopping network is pre-activated (Aron, 2011). Therefore the key differences between proactive and reactive control could lie in the temporal activation of preparatory response related processing.

Additionally the developmental course for sustained response control extends into late adolescence (Hämmerer et al., 2010, Ladouceur et al., 2004, Luna et al., 2004a, Luna et al., 2004b, Shing et al., 2010, Williams et al., 1999). An fMRI investigation found that transient (reactive) activation of neural areas supporting inhibitory control decreased from childhood to adolescence whereas sustained (proactive) activation increased in adulthood (Velanova et al., 2009). This perhaps is due to the development of a proactive control strategy. In particular Ordaz et al. (2010) concluded that limitations in adolescents’ ability to inhibit a response may be related to fundamental differences invoked to prepare a response.

In order to examine developmental differences in proactive preparation we designed a conditional task switching paradigm that can separate response preparation from response execution. In this paradigm participants are firstly presented with a circle or square visual cue that has been designated at the start of the experiment to indicate that the subsequent trial will most likely be a ‘go’ trial (press on the same side as the stimulus) or a switch trial (press on the opposite side to the stimulus). Secondly after the circle or square (shape) cue a tone is heard that will indicate either go, switch, or stop responses (Fig. 1). The blocks also included trials of GO cues followed by switch tones (GO/sw) and SWITCH cues followed by go tones (SW/go). This was to ensure that the participants actively engaged control throughout the task and did not come to expect certain stimulus-response patterns. Stop trials were also included to ensure that participants attend to the stimuli and not just the cue. The time between the shape cue and the tone is considered a response preparation phase. Neural differences in proactive response preparation in conditions of low control (GO cue followed by go tone) and high control (Switch cue followed by switch tone) will be compared in this response preparation phase. It is thought that increased proactive control will be engaged during the condition of high control (Switch cue followed by switch tone). The primary aim of this study is to compare the neural activity of adolescents and young adults during the response preparation phase to identify whether or not they use a similar proactive control strategy.

Fig. 1.

Schematic of several trials. Participants were told that the blue circle indicates that a go tone will likely follow whereas the red square indicates that a switch tone will likely follow. This was counterbalanced. Response preparation was analysed between 0 and 1150ms and response execution was analysed between 1150 and 2250ms after the tone.

To examine preparatory neural activity two ERP components are commonly examined: the lateralized readiness potential (LRP) and the contingent negative variation (CNV). The LRP is thought to represent the initiation of a motor response as it measures the differential activation of electrodes over the left and right motor cortex (C3 and C4 respectively) (Gratton et al., 1988). The LRP can give precise temporal information about the activation of the motor cortex in the context of response hand preparation. The developmental progression of the LRP has been identified in children (Bryce et al., 2011, Ridderinkhof and van der Molen, 1995, Szucs et al., 2009a) however not in adolescents. In terms of childhood development it is found that correct response hand preparation becomes increasingly faster with age (Ridderinkhof and van der Molen, 1995) while early incorrect hand activity during interference decreases with age (Szucs et al., 2009a).

Similarly neural correlates of preparation can be measured using the CNV (Weerts and Lang, 1973). Response activity is normally preceded by an increasingly negative wave over frontal and central electrode cites. This negativity is found to reflect motor preparation for the response (Loveless and Sanford, 1975), sensory anticipation (Gómez et al., 2003) and activation of attentional networks (Fan et al., 2007). Although the CNV is not a direct measure of response-related processing it provides information on preparatory processing. Few studies have examined CNV preparatory activity across development, particularly in a cued task-switching context. However studies involving the No-Go task and a cue-probe task have found either the complete absence of the CNV in children (Flores et al., 2009, Perchet and Garcia-Larrea, 2005) or increasing amplitude with age (Jonkman et al., 2003, Jonkman, 2006). We predict that if adolescents have immature proactive control during the response preparatory phase they will exhibit decreased LRP and CNV amplitude compared with young adults.

As the focus is on response-related aspects of preparation, electromyography (EMG) provides a robust direct measure of motor activity at the level of effectors. Electrodes are placed on each hand and this can record parallel correct and incorrect activity over the course of a task (Szucs and Soltész, 2010a, Szucs et al., 2009b). This can provide direct evidence of correct or incorrect response preparation before the overt reaction time response (Szucs and Soltész, 2010a, Szucs et al., 2009b). Although previously unexplored in adolescents EMG has been examined in childhood development. Recently van de Laar et al. (2012) found that motor activity during a preparatory period (pre motor time) was most sensitive to age related changes in 8–12-year-old children and continues to develop beyond 12 years of age. In terms of reactive control, the robust EMG signal can be used to detect minute reactive motor activity to a cue followed by ‘just in time’ correction from incorrect to correct hand activity. Boulinguez et al. (2008) examined EMG activation in response to a warning signal in a simple RT task. From examining the distribution of errors they found that warning signals trigger transient automatic EMG activation. They suggest that activity triggered by the warning stimulus requires ‘proactive volitional inhibition’ to prevent premature responding.

Although the main aim of the study was to examine response related preparatory processing, stimulus level processing will also be examined to index early developmental differences. The latency and amplitude of the P3b are commonly thought to index the speed of stimulus categorization (Donchin, 1981). The P3b is commonly used to separate developmental change at the stimulus level from change at the response level as the P3b is thought to represent stimulus processing independently of response level processing (Bashore et al., 1989, Szucs et al., 2009b). If developmental differences are found in the amplitude and latency of the P3b this could indicate differences in perceptual and cognitive processing. Additionally another marker of early stimulus processing is the P3a which is a frontal activity that occurs around 300ms after the stimulus. In adults this activity is thought to be related to general stimulus selection (Dien et al., 2004, Polich, 2007). However in a study examining response inhibition and preparation, Jonkman et al. (2003) concluded that the absence of frontal P3a activity was related to developmental delay in response inhibition. It is currently unclear what developmental changes in the P3a indicate. Therefore another important aim of this study was to correlate neural physiological findings with real world behaviours. We examined the relationship between behavioural measures of short term memory and working memory (digit span forwards and backwards; WAIS III) and stimulus processing (P3b and P3a) to further explain the functional significance of developmental differences. If adolescents use a reactive control strategy this may be driven by their inability to effectively engage stimulus processing as evidenced by lack of P3b and P3a activity. By examining the relationship between neural correlates of reactive control and skills such as working memory we can ascertain if reactive control in adolescents has widespread effects in the classroom and everyday life.

Here our aim was to isolate neural–physiological differences in adolescents and young adults as they engaged either proactive or reactive control during the response preparation phase of a cued-task switching paradigm. We expected that young adults would show improved proactive control when compared with adolescents. This would manifest in increased negativity in the CNV amplitude and the LRP amplitude indicating early response preparation as well as increased EMG amplitude in the correct responding hand. We expected that adolescents would show very little or no CNV amplitude or LRP indicating lack of response preparation and increased incorrect hand preparation in EMG activity as indicative of a reactive control strategy. We also examined reaction time activity to identify the affect of proactive or reactive control strategy on behavioural performance. Finally in order to explore the relationship between proactive response preparation and other cognitive skills short term and working memory were compared between the two groups.

2. Method

2.1. Subjects

Initially 40 participants were examined. However due to EEG artefacts 5 participants from the adolescent group and 5 participants from the young adult group were rejected from the analysis. The remainder composed two age groups: adolescents (n=15, mean age 16.55 years, 1 left handed, 10 females), young adults (n=15, mean age 25.83, 4 left handed, 11 females). Handedness was determined according to dominant writing hand. All subjects were fluent in English, had normal or corrected to normal vision and were without a history of psychiatric disorder. Informed consent was obtained from each participant and from the parent or guardian of the adolescent participants. The adults were from the Cambridgeshire area (UK). Adolescents were students from local 6th form colleges around Cambridge, UK. The study received approval from the University of Cambridge Psychology Ethics Board.

2.2. Task and stimuli

On each trial a shape cue was presented followed by a tone stimulus. The shape cue was either a blue circle or a red square. Before the experiment instructions designated the blue circle or the red square as either a go cue or a switch cue. For example when the blue circle represented a go cue then participants could expect a go tone to follow in most of the trials; when the red square represented a switch cue then participants could expect a switch tone to follow in most of the trials. This was counterbalanced across participants. The tone stimulus could be 200Hz, 500Hz or 1800Hz. Instructions also designated the meaning of the tones in terms of go, stop or switch responses. This was also counterbalanced across participants. Participants were instructed to watch the computer screen and hold a button press using their left and right thumbs. They were told to respond as quickly as possible once they had heard the tone. As the aim was to examine LRP, vertical spatial orientation (up and down as opposed to horizontal; left and right) of the stimuli was used to avoid confounding and simultaneous lateralized activation of the occipital cortex. Therefore the different combinations of cue stimuli and tones created 6 different conditions (Table 1).

Table 1.

Stimuli proportions of the cued task switching experiment.

Cue	Tone	% of total trials
GO cue	GO	35%
	SWITCH	10%
	STOP	5%
SWITCH cue	GO	10%
	SWITCH	35%
	STOP	5%

For the purpose of this investigation we were interested in two conditions: the GO cue followed by the go tone (GO/go) and the SWITCH cue followed by the switch tone (SW/sw). This allows for the comparison of a condition where low control is required (GO/go) with a condition of high control (SW/sw). Both of these conditions occurred with 35% probability across the experiment.

Participants performed a total of 8 blocks with 80 trials per block. The stimuli were pseudo-randomized whereby each subject had a different random order of stimuli presentation. This was to ensure that there would be no random effects due to one particular stimuli randomization. Subjects were seated in a small room facing a 19in. computer screen. Stimuli were presented on a white background. Each trial started with a fixation point (picture of an eye) for 300ms. This was followed by a blank screen for 1000ms. The cue stimulus then appeared for 1050ms followed by blank screen. The tone, duration 100ms, therefore appeared 1150ms after the cue stimuli onset. A response period of 1000ms followed. Participants were instructed to blink when they saw the fixation eye. Before the experimental blocks one practice block was completed with 15 stimuli. Stimuli were presented using the Neurobehavioral Systems Presentation 11 program. For the purpose of this investigation we were interested in examining a preparation phase between the cue and the tone stimulus (0–1050ms relative to cue onset) and a response execution phase after the tone stimulus (1250–2250ms relative to cue onset) (Fig. 1). In the statistical analysis the variable of ‘pre/post tone’ describes a comparison of neural activity before the tone is presented during the response preparation phase and after the tone is presented during the response execution phase.

2.3. Data recording and analysis

In both the behavioural and physiological analysis post hoc Tukey tests were used to examine the contrasts. The EEG analysis and behavioural analysis included only correctly responded trials.

2.3.1. Behavioural data

A repeated measures ANOVA was used to examine reaction time and accuracy. Group (adolescents, young adults) was the between subjects factor while condition (GO/go, SW/sw) was the within subjects factor.

Additionally behavioural measures of working memory and short term memory were administered from the WAIS III digit span forward and digit span backward assessments. An independent t-test was used to compare the raw scores of the young adult and adolescent groups. Pearson's parametric correlation was performed to compare the raw scores of working memory or short term memory with CNV mean amplitude in the GO/go condition and the SW/sw condition during preparation.

2.3.2. EEG Data

EEG data were recorded in an electrically and acoustically shielded booth using 129-channel Hydro-Cell Net from an Electrical Geodesics system. A sampling rate of 500Hz was used. An on-line band pass filter of 0.01–70Hz was used. Offline the data were band-pass filtered between 0.01 and 30Hz and recomputed to an average reference. Epochs were from −100 to 2250ms relative to the onset of shape cue presentation. Data was average baseline corrected from −100 to 2250ms before stimulus presentation. Individual channels were spine interpolated if required using the parameters described by Perrin (1990); an order of 4 and degree 10, lambda 1E−05 and we interpolated a maximum of 10% of the total electrodes following the recommendation of Electrical Geodesics (EGI, Oregon, USA). Participants were excluded during preprocessing before any data analysis occurred. Epochs were excluded from the analysis if the following artefact rejection criteria were violated; voltage deviations exceeding ±120μV relative to baseline, maximum gradient exceeding 50μV, and the lowest activity below 0.5μV. If more than 50% of trials were rejected for a participant this participant was then excluded from analysis. After artefact rejection, for adolescents 68.2% of trials were retained for the GO/go condition and 69.1% for the SW/sw condition. For young adults 69.4% of trials were retained for the GO/go condition and 71.0% for the SW/sw condition. EEG data were processed using Brain Vision Analyzer (Brain Products, Munich), Matlab 7.9, SPSS 17.0 and Statistica 9.

For the P3b component a pool of centro-parietal electrodes was examined (electrodes 54, 55, 61, 62, 67, 72, 77, 78, 79). The electrode locations are shown in the schematic topography in Fig. 2. These electrodes were chosen based on previous literature that has found the P3b to be well defined at centro-parietal sites (Donchin, 1981, Luck, 2005, Szucs and Soltész, 2010b). The P3b peak latency and amplitude were examined at two time points; the most positive peak in the preparatory phase (250–600ms) and response execution phase (1500–2000ms). P3b peak amplitudes and latencies were entered into a repeated measures ANOVA of pre/post tone stimulus (2)×group (2)×condition (2).

Fig. 2.

P3b activity at centro-parietal electrodes. Vertical line represents the tone stimulus. The peak amplitude of the P3b is significantly greater in the adolescent group during response execution (i.e. after the tone is presented) (F(1,28) 11.326, p=0.0022). The schematic head displays the P3b centro-parietal electrode pool. Electrodes 54, 55, 61, 62, 67, 72, 77, 78, 79 (CpZ, P1, Pz, P2, Oz, O1, O2) are represented by the black squares.

The P3a effect was observed in a topographical comparison of adolescents and young adults during the SW/sw condition for the response execution phase. This frontal effect refers to an increased positivity in the difference potential between the SW/sw minus GO/go conditions. As this was only observed during response execution the P3a effect was identified as the positive mean amplitude in a frontal pool of electrodes between 1500 and 1600ms for each age group (frontal electrodes 14, 15, 17, 21) during the SW/sw condition. The mean amplitude of the average ERPs in the SW/sw condition between 1500 and 1600ms was examined in a one way ANOVA (group (2)×condition (1))

The LRP was calculated according to convention (Coles, 1989): [(ER–EL) left hand response+(EL–ER) right hand response]/2. ER represents the amplitude of the ERP at the electrode over the right motor cortex, whereas EL represents the amplitude of the ERP at the electrode over the left. The left and right motor cortex electrodes were electrodes 36 and 104 respectively. These have the equivalent of positions C3 and C4 in the traditional 10–20 electrode system. The raw LRP waveforms were smoothed using a 50ms moving average window to improve signal to noise ratio. An additional baseline correction of 100ms before the cue was also applied. The peak latency and amplitude of stimulus locked LRPs was calculated from 400 to 800ms relative to cue onset. The mean amplitude of LRP was examined between 600 and 800ms relative to cue onset. Response-locked LRP was not examined as we were interested in preparatory LRP activity during the response preparation phase (0–1050ms). A repeated measures ANOVA of group (2)×condition (2) was performed on the peak latencies and amplitudes as well as the mean amplitude during the specified time points.

The mean amplitude of the CNV was examined at frontal (FcZ) and central (Cz) electrodes during the preparatory period of 750–1050ms (Lutcke et al., 2008, Fan et al., 2007, Flores et al., 2009). An additional baseline correction of 100 prior to the cue was performed. Mean amplitude during preparation (750–1050ms) was entered into a repeated measures ANOVA group (2)×condition (2) separately for the FcZ and Cz data.

2.3.3. EMG data

Muscle activity in both response fingers was recorded using an MP150 data acquisition unit (Biopac Inc.) EMG was measured by EMG110C amplifiers. 110 S shielded touch-proof leads where connected to two disposable cloth-based hypoallergenic Ag–AgCl EL504 recording disc electrodes. The electrodes were placed along the left and right flexors of the thumb (flexor pollicis brevis). An electrode on the left elbow was used as a ground. Before the electrodes were applied the skin was washed with soap and cleaned with alcohol wipes. The electrodes were attached by adhesive solid gel. EMG was sampled at 2000Hz and band pass filtered between 10 and 500Hz. The data was then rectified and scaled relative to the maximum amplitude in each individual as measured from continuous data. EMG was baseline corrected between −100 and 0ms relative to cue presentation and is displayed as a percentage of the maximum value measured. Epochs extended from −100 to 2250ms relative to cue presentation. The grand-averaged EMG waves were calculated for each condition and smoothed with a moving average of 50ms to improve the signal to noise ratio. Peak latency and peak amplitude values were calculated for correct and incorrect hand activity during preparation (100–500ms) and response execution (1500–2000ms). For the peak latency and amplitude a repeated measures ANOVA of group (2)×condition (2) were performed separately for correct and incorrect hand activity as well as together group (2)×hand (2)×condition (2).

The mean amplitude of incorrect hand activity was also examined during preparation between 900 and 1100ms and during execution between 1700 and 2000ms. The mean amplitude of correct hand activity was examined between 200 and 300ms in preparation and between 1500 and 2000ms in execution. Repeated measures ANOVA of group (2)×condition (2) was performed separately for correct and incorrect hand activity. A t-test of all means against zero was performed on incorrect hand activity in each group to ensure a significant deviation from baseline. A dependent samples t-test was performed to compare the peak latency, peak amplitude and mean amplitude of the GO/go and SW/sw conditions in the adolescents’ incorrect hand activity.

3. Results

3.1. Behavioural

Accuracy and RT values are shown in Table 2. RT and accuracy values were obtained after excluding potential fast guesses (responses faster than 150ms). RT during SW/sw trials (1782ms) was 28ms slower than GO/go trials (1754ms) (F(1,28) 6.995, p=0.0133). Additionally there was a marginal group×condition interaction (F(1,28) 2.981, p=0.095) whereby young adults were slightly faster during the GO/go trials compared to the SW/sw, however adolescents did not show the same improvement in RT during the GO/go trials. Error rates between the two groups did not significantly differ and there were no interactions.

Table 2.

Reaction time (ms), accuracy (%), standard error values in brackets.

Group	Measure	GO/go	SW/sw
Adolescents	RT	1789 (±26.9)	1799 (±32.5)
	Accuracy	90.1 (±2.47)	85.1 (±3.34)
Young adults	RT	1718 (±33.23)	1765 (±33.9)
	Accuracy	94.0 (±2.10)	91.4 (±1.83)

3.2. P3b

Grand averaged ERPs of representative centro-parietal electrodes (54, 55, 61, 62, 67, 72, 77, 78, 79) are shown in Fig. 2. In the repeated measures ANOVA pre/post tone (2)×group (2)×condition (2) a significant pre/post tone stimulus×group interaction was found for the peak amplitude of the P3b (F(1,28) 11.326, p=0.0022). During the response execution time window (post cue) the P3b peak amplitude was significantly greater in adolescents (6.2μV) when compared to young adults (3.4μV) (p=0.0053). There were no significant effects for the P3b peak latency.

3.3. P3a effect

The topographical differences between the SW/sw and GO/go condition (SW/sw minus GO/go difference topographies) at a grouping of 4 frontal electrodes (14, 15, 17, 21) are shown in Fig. 3. For the one way ANOVA (group (2)×condition (1)) a marginal group effect was found (F(1,28) 3.907, p=0.0580) between 1460 and 1550ms. The mean amplitude of the P3a difference potential was significantly more positive in young adult group (0.67μV) than in the adolescent group (−0.40μV) during the SW/sw condition.

Fig. 3.

P3a effect SW–sw minus GO–go difference topographies. (A) Adult P3a positivity is visible between 1460 and 1550ms. (B) Adolescent P3a is not present. The P3a effect is clearly dissociable from P3b activity.

3.4. CNV

Grand averaged ERPs of representative FcZ electrode where CNV showed the maximum amplitude are shown in Fig. 4. For the repeated measures ANOVA group (2)×condition (2) a significant effect of group was found (F(1,28)=5.03, p=0.0329) whereby young adults had an enhanced CNV negativity (−2.68μV) compared to adolescents (−0.99μV). The mean amplitude of the CNV was also examined at Cz however no significant group or condition effects were found.

Fig. 4.

CNV FcZ. Vertical line represents the presentation of the tone stimulus. A significant effect of group was found (F(1,28)=5.03, p=0.0329) whereby young adults had an enhanced CNV negativity (−2.68μV) compared to adolescents (−0.99μV). The schematic head displays where an electrode was measured frontally at 6 (Fcz) represented by a small circle.

There were no significant group differences between young adults and adolescents on the working memory backward digit span raw scores (t(28)=1.00, p=0.325). However as seen in Fig. 5 in adolescents a significant negative correlation was found when the CNV mean amplitude during the GO/go condition was compared with raw working memory scores on the WAIS III backward digit span test (r=−0.6214, p=0.013). This was not found for adults. In adults an additional correlational analysis was preformed without the outlying data point (x=8, y=−9 in the young adult graph). This confirmed our previous results and therefore did not have any undue influence on this correlation. Without this outlying data point there was still a non-significant correlation between CNV mean amplitude and working memory score in the young adult data (r=−.1174, p=0.689).

Fig. 5.

Pearson's parametric correlation between FcZ CNV mean amplitude between 500 and 1050ms and raw working memory scores on WAIS Digit span backward. (A) A significant negative correlation is found for adolescents r=−0.6214, p=0.013. (B) No significant correlation was found for adults for CNV mean amplitude at working memory score r=−0.0911, p=0.747.

3.5. LRP

The peak latency and amplitude as well the mean amplitude of the LRP at several different time points during preparation and execution were examined. No significant group differences were found.

3.6. EMG

Grand averaged EMG activity in the correct and incorrect responding hands is shown in Fig. 6. EMG activity during execution phase (1250–2000ms): Correct hand activity. A repeated measures ANOVA of group (2)×condition (2) of the peak latency between 1500 and 2000 revealed a significant main effect of group (F(1,28) 5.022, p=0.0331). The EMG peak occurred significantly earlier in young adults (1695ms) than in adolescents (1781ms). The repeated measures ANOVA of group (2)×condition (2) on the mean amplitude of correct hand activity between 1500 and 2000ms revealed a significant group×condition interaction (F(1,28)=7.398, p=0.0111). The mean amplitude in young adults significantly differed between the GO/go vs SW/sw conditions (1.25μV vs 1.35μV) whereas there was no significant difference between these conditions in adolescents (0.93μV vs 0.91μV). There is a trend in the peak latency of the EMG data for increased latency during the SW/sw condition (e.g. GO/go 1733.95ms versus SW/sw 1767.5ms) however it was not significant (F(1,28)=0.209, p=0.6510).

Fig. 6.

EMG activity. (A) Correct hand activity. The peak latency between 1500 and 2000 revealed a significant main effect of group (F(1,28) 5.022, p=0.0331). Young adults were significantly earlier than adolescents. (B) Incorrect hand activity. In adolescents incorrect hand activity significantly deviated from zero (t(14)=2.56, p=0.0224). Vertical line represents tone stimulus.

3.6.1. Incorrect hand activity

The t-test on the mean amplitude between 1700 and 2000ms confirmed that the adolescent incorrect hand activity significantly differed from zero (t(14)=2.56, p=0.0224) whereas the young adult hand activity was only marginally different from zero and this was not significant (t(14)=1.93, p=0.0739). The activity surrounding an ‘initial dip’ in young adult mean amplitude was also examined between 1150 and 1800ms (see Fig. 6). Firstly the mean amplitude was examined from 1150 to 1350ms and no significant group effects were found (F(1,28)=0.209, p=0.6510). Secondly the peak amplitude was examined between 1400 and 1500ms. A marginal group effect was found (F(1,28)=3.638, p=0.0667). Finally in the period after the dip from 1500 to 1800ms the mean amplitude was examined and there was no significant difference between the groups (F(1,28)=0.833, p=0.3691).

4. Discussion

The neural mechanisms responsible for the transformation from the immature adolescent brain to the goal-directed young adult brain are elusive. A recent theory of cognitive control, the dual mechanisms of control model (Braver and Gray, 2007) suggests that variability in the ability to perform complex cognitive tasks may be due to the use of two different types of control; proactive and reactive. It has been suggested that adolescents use a reactive as opposed to proactive control strategy and this may underlie many of the behavioural differences observed between adolescents and young adults (Andrews-Hanna et al., 2011). Here we sought to pinpoint developmental differences in the neural properties of proactive and reactive control during the response preparation phase of a cued-task switching paradigm.

We found that adolescents did not use a proactive response preparation strategy. The lack of CNV and LRP amplitude during the preparatory phase indicates less efficient response preparation. EMG revealed delayed correct hand activity and increased incorrect hand activity. Interestingly increased P3b amplitude in the execution phase in adolescents but not in young adults suggests the lack of ability to translate cue information into efficient response execution. This is discussed in terms of different proactive neuro-cognitive mechanisms.

The behavioural results followed the expected pattern. The reaction time in the SW/sw trials was significantly longer than the GO/go trials. In previous studies a hallmark of a proactive strategy has been slower reaction times during more difficult trials (Aron, 2007, Aron, 2011). This confirms that the SW/sw condition in this task provides the platform for engaging proactive control. Additionally in the young adult group the GO/go condition and SW/sw condition differed by 50ms whereas in the adolescents this difference was only 10ms. This could indicate that adolescents did not use different strategies for these two conditions and instead used a slower reactive strategy for both low and high conflict conditions. This is a non-significant trend in RT but significant in EMG activity.

Arguably it is beneficial to have similar behavioural performance between two age groups as this allows for the analysis of neural activity without behavioural confounds (Luna, 2009). However in future studies differences at the behavioural and neural level could be correlated with real world measures such as decision making, impulsivity, and risk taking as this would provide greater insight into the breadth of neuro-cognitive change during adolescence (see Geier and Luna, 2010, Andrews-Hanna et al., 2011). For example individual differences in adolescence performance and neural activity may help to explain differences in real world behaviours such as impulsivity and lack of foresight.

4.1. Neural correlates of response preparation: CNV and LRP

We found no significant condition or group differences in the latency and amplitude of incorrect hand LRP activity. Even though the LRP is found to be a useful measure of response preparation there are some difficulties. One problem is that the LRP is calculated by examining the sum of activity over both cortices (correct and incorrect activity) (Gratton et al., 1988). Therefore incorrect hand activity is undetectable if it occurs at the same time as correct hand activity. Szucs et al. have only been able to detect incorrect hand activity in the LRP in one previous study. In this case incorrect hand activity onset very quickly whereas correct activity occurred much later due to the nature of the task (Szucs et al., 2009a). Perhaps the design of this current task permitted simultaneous correct and incorrect hand activity therefore rendering differences in the LRP undetectable.

CNV activity confirms the use of a proactive control strategy in young adults. Young adults displayed an augmentation in CNV negativity during the preparatory phase thereby indicating increased neural preparation. This was not present in adolescents. Several authors have found that during development the CNV is either absent (Perchet and Garcia-Larrea, 2005) or of significantly smaller amplitude than in young adults (Jonkman et al., 2003, Jonkman, 2006). This lack of CNV activity in adolescents could be due to the extended development of the frontal lobes (Adleman et al., 2002, Rubia et al., 2000, Sowell et al., 2003, Velanova et al., 2009). Jonkman (2006) concluded that the reduced CNV in adolescents is due to an immature frontal parietal network involved in the regulation of motor control.

Additionally the link between the CNV, working memory and prefrontal cortex activity provides an interesting framework for interpreting the current results. The relationship between CNV activity and behavioural measures of working memory and short term memory was also examined. Only in adolescents was there a significant negative correlation between working memory score and CNV mean amplitude; as working memory improved CNV preparatory negativity increased. As this was not found in young adults, it appears that adolescents are still using working memory to mediate and improve preparatory CNV activity. Fuster (2002) suggested that the prefrontal cortex uses two different skills for the temporal organization of behaviour; working memory and preparatory set. He suggests that working memory is a ‘retrospective action’ to collate sensory information whereas preparatory set is a ‘prospective action’ which depends on the relationship between the cue and the response. Perhaps delayed prefrontal development in adolescence results in an increase in working memory as a retrospective/reactive as opposed to prospective/proactive strategy as indicated by lack of CNV activity.

4.2. EMG: correct and incorrect hand activity

Overall the RT condition differences previously mentioned and EMG activity validate the paradigm. The SW/sw condition activated the correct hand to a greater degree than the GO/go condition during response execution. This confirms that the more difficult condition required greater hand activation and response control. EMG has been used to examine motor level engagement of response control during conditions of differing degrees of difficulty or conflict (Burle et al., 2002, Szucs and Soltész, 2010a, Szucs et al., 2009b). Increased correct EMG activity is often found during the more difficult condition (incongruent condition) (Szucs et al., 2009b, Fig. 8; Szucs and Soltész, 2010a, Fig. 2).

EMG activity reveals a performance benefit in young adults at the motor level potentially due to earlier proactive preparation; this is absent in adolescents. First of all young adults had earlier peak latency in EMG activity during response execution than adolescents across all the conditions. This potentially indicates greater certainty in responding. Subsequently during the response execution phase young adults successfully engaged additional processing resources during the SW/sw condition as indicated by increased mean amplitude compared to the GO/go condition. Adolescents did not differentially engage motor resources during these two conditions. Additionally incorrect hand activity was significantly greater than baseline in adolescents indicating a lack of response control during response execution. Interestingly in young adults an initial dip (decreased activity) in incorrect hand activity between 1400 and 1500ms may indicate inhibition of both responding hands before the correct hand was engaged. This was not present in adolescents. Overall due to their delayed peak latency and increased incorrect hand activity this indicates an inefficient response processing strategy for adolescents.

Only two previous studies have examined EMG activity developmentally (Ridderinkhof and van der Molen, 1995, van de Laar et al., 2012). The current results confirm a delayed developmental course for response control that extends beyond childhood. van de Laar et al. (2012) examined developmental changes in a choice RT task using a similar methodology (LRP and EMG analysis) and found changes to the central cortical as well as peripheral motor structures underlying response speed. They found that young children (mean age 7.7) experienced a greater delay during the pre-response selection time period coupled with ‘less steep slope of EMG onset and variability in the motor command’. They conclude that younger children are late to engage a central motor command due to slow stimulus processing and this results in inefficient response preparation. Older children (mean age 11.9), however, were found to have an extended interval between EMG onset and response. They suggest that in young childhood stimulus level difficulty extends to motor level processing whereas in older childhood the locus of developmental change lies in the continued maturation of peripheral motor processes. Earlier development of stimulus level processing in childhood followed by the late development of the motor system has been previously noted (Hogan et al., 2005, Shaw et al., 2008) however the exact functional and structural changes have not been clearly documented. The late development of the motor system is an interesting area of research that requires further investigation as it has implications for better understanding behavioural characteristics such as impulsivity (Aron, 2011).

4.3. P3b and P3a effect

Although we confirm that the main locus of developmental transformation from adolescence to young adulthood lies in response related aspects of control, developmental differences in the P3b during the execution phase provide an interesting neuro-cognitive framework of proactive control. The P3b during the response execution phase was significantly greater in adolescents than in young adults. Additionally the lack of P3a effect in adolescents suggests that fast updating of stimulus properties may be underdeveloped in adolescents (Dien et al., 2004). Overall these results suggest that protracted stimulus categorization in adolescents leads to more inefficient response execution.

Jahfari et al. (2010) identified three different neuro-cognitive mechanisms that may underlie the slower RT found in proactive control. Adolescents and young adults may use different proactive strategies that drive the differences in P3b amplitude throughout the preparation and execution phases. First active braking is described as the process of proactively suppressing an initiated response without completely cancelling it. This would manifest in reduced excitability of neural representations associated with the response before the stimulus has been presented; for example suppressed (delayed and decreased) P3b amplitude whereby subjects will suppress their responses proactively. Second, the prolonged decision stage hypothesis suggests that the increased cognitive load required during high conflict trials may affect information processing. This suggests that delay rests at the decision making stage of processing; for example delayed P3b latency. Thirdly the slower response facilitation hypothesis suggests that delayed RT results from prolonged hesitancy or caution at the response level. This does not suggest inhibition (as the active braking hypothesis suggests) but instead a delayed processing between the initiation of a response and button press. In this case the P3b would reflect normal stimulus categorization processing however a delay or modification in response level processing would be expected. In terms of young adults it appears that the current data fits the ‘slower response facilitation’ hypothesis. The early and effective P3b activity followed by increased amplitude EMG activity for the SW/sw trials suggests successful stimulus level processing facilitates response related processing. Conversely the adolescent pattern suggests a prolonged decision stage. The increase P3b activity after the tone is heard could be representative of delayed stimulus categorization. It is likely that adolescents use a reactive control strategy to rapidly update stimulus categorization after the tone during response execution.

In conclusion we confirm that adolescents use a reactive control strategy. This is marked by significant neural and physiological differences in how adolescents and young adults prepare and execute a response during a cued task switching paradigm. In adolescents lack of CNV, and P3a preparatory activity followed by delayed and incorrect EMG activity and increased P3b activity during response execution reveals a more transient, reactive and less efficient cognitive strategy. Further research into the precise nature of the transition from impulsive and inefficient adolescent behaviour to goal-directed young adult behaviour is important for advancing developmental theory and establishing models of neural change.

Conflict of interest

We confirm that this manuscript has not been published elsewhere and is not under consideration by another journal. All authors have approved the manuscript and agree with submission to the Journal of Developmental Cognitive Neuroscience. We have read and have abided by the statement of ethical standards for manuscripts. The authors have no conflicts of interest to declare.