- Email: [email protected]
NoGo task: ERP components and behaviour
International Journal of Psychophysiology 123 (2018) 74–79
Contents lists available at ScienceDirect
International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho
A processing schema for children in the auditory equiprobable Go/NoGo task: ERP components and behaviour
T
⁎
Robert J. Barry , Frances M. De Blasio, Jack S. Fogarty Brain & Behaviour Research Institute, and School of Psychology, University of Wollongong, Northfields Avenue, Wollongong, NSW 2522, Australia
A R T I C L E I N F O
A B S T R A C T
Keywords: Event-related potentials (ERPs) Equiprobable Go/NoGo task Principal components analyses (PCA) Misallocation of variance Sequential processing schema Children
A sequential processing model for adults in the auditory equiprobable Go/NoGo task has been developed in recent years. This used temporal principal components analysis (PCA) to decompose Go/NoGo event related potential (ERP) data into components that mark stages of perceptual and cognitive processing. The model has been found useful in frameworking several studies in young and older adults, and in children. Recently, it has been demonstrated that the common PCA approach of decomposing Go and NoGo ERP data together results in misallocation of variance between the conditions, distorting the timing, topography, and amplitudes of the resultant components in each condition. The present study thus reanalyses data from a child study, conducting separate PCAs on the data from each condition. Multiple regression was then used to seek links with behavioural measures from the task. In addition to confirming the previous NoGo N2b/inhibitory processing link, novel NoGo Negative Slow Wave/error evaluation and Go N1-1/RT variability links were obtained. Based on these outcomes, the recommended separate application of PCAs to Go and NoGo data was confirmed. The present data were used to develop a child-specific sequential processing schema for this paradigm, suggesting earlier separation of the Go and NoGo processing chains, and the need to include an additional inhibition and evaluation stage. The child schema should be useful in future studies involving this and other two-choice reaction tasks.
1. Introduction The unwarned equiprobable auditory Go/NoGo task is an auditory intermediary between the traditional active Oddball task, with rare targets (probability < 50%) amongst frequent standards, and the traditional Go/NoGo task, with frequent Go targets (probability > 50%) amongst rare NoGo stimuli. Apart from target/Go probability, these two-stimulus discrimination tasks share common stimulus evaluation processes at each stimulus presentation, leading to discrimination between one stimulus to be responded to and one that does not require a response. Subsequent differential processing chains involve response preparation and execution processes, or cessation/inhibition of further processing. These different control processes (Huster et al., 2013; Folstein and van Petten, 2008) may be subject to monitoring (van Veen and Carter, 2006) as the participant attempts to optimise performance through the task. We were initially interested in this task because it provides equal numbers of trials to form separate event related potentials (ERPs) to Go and NoGo stimuli with similar signal:noise ratio (SNR), facilitating exploration of their prestimulus electroencephalographic (EEG) determinants (Barry et al., 2010). Subsequently we continued brain
⁎
dynamics studies exploring prestimulus EEG amplitude effects on the ERP components in this task (De Blasio and Barry, 2013a, 2013b), as well as the impact of the EEG phase of narrow-band and traditional (delta, theta, alpha, and beta) bands at stimulus onset (Barry and De Blasio, 2012; Barry et al., 2014c, submitted). In parallel, we embarked on a program to obtain more understanding of the components constituting these ERPs, and employed temporal principal components analysis (PCA) to obtain data-driven decompositions of the Go and NoGo ERPs for that purpose. Barry and De Blasio (2013) reported our first sequential processing interpretation of the Go/NoGo processing stages using temporal PCA of the ERPs from a sample of young adults. In that schema, summarised in Fig. 1, we suggested that N1 (with its “true” subcomponents N1-3 and N1-1; and Processing Negativity, PN; Näätänen and Picton, 1987) marked the beginning of Go/NoGo differentiation, with complete differentiation or categorization marked by N2. This P2/N2 complex was followed by two separate processing chains. These chains lead to either the Go response (marked by P3b and the classic Slow Wave, SW) or the NoGo non-response (marked by the P3a and a novel Late Positivity, LP). LP appeared at the end of the NoGo epoch, apparently earlier in NoGo than Go, marking the earlier termination of stimulus processing in the
Corresponding author at: School of Psychology, University of Wollongong, Wollongong, NSW 2522, Australia. E-mail address: [email protected] (R.J. Barry).
https://doi.org/10.1016/j.ijpsycho.2017.10.014 Received 5 July 2017; Received in revised form 30 October 2017; Accepted 31 October 2017 Available online 06 November 2017 0167-8760/ © 2017 Elsevier B.V. All rights reserved.
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 1. Barry and De Blasio's (2013) early conceptualisation of the adult processing schema for the equiprobable Go/ NoGo task. PN: Processing Negativity; SW: Slow Wave; LP: Late Positivity.
We expect to confirm that stimulus-specific components are emphasised in the separate streams shown in the processing schema, and to confirm that the NoGo N2b is linked to response inhibition in children. We also take the opportunity to seek other component-function links clarifying the determinants of the behavioural measures characterising child performance in this paradigm.
absence of the active Go response. This schema was subsequently used to framework a number of studies in young adults exploring brain dynamics (Barry et al., 2014c, submitted), the effects of caffeine (Barry et al., 2014b), the impact of a varying interstimulus interval (Borchard et al., 2015), differences between young adults and children (Barry et al., 2014a), and the effects of ageing in young and older adults (Barry et al., 2016a). As well as helping to explore these different topics, the stability of the component structures obtained in the studies helped clarify and refine details of the processing schema as new data were linked to a number of components. Children show both similarities and differences relative to adults in their behaviour and ERPs across the range of two-choice active discrimination tasks, and this was explored in the context of the sequential processing schema in Barry et al. (2014a). Early (N1-1, and PN) and late components (P3, SW, LP) were similar between 18 children and 18 young adults, but the intermediate P2 and N2 differed substantially in their stimulus effects. Barry and De Blasio (2015) explored the child response pattern in a group of 40 children aged 8–13 years, focussing on whether the NoGo N2b was linked to inhibition, given our view that the equiprobable task was so simple that active inhibition of the Go response was not required. This view reflected an earlier ERP source study by Barry and Rushby (2006), which noted that electrodermal responses were smaller to the NoGo than Go stimuli in this paradigm, suggesting less effortful processing. However, the observation in Barry et al. (2014a) of child N2b enhancement, and children's difficulty with the task, suggested the need to revisit this question, at least in children. Barry and De Blasio (2015) found that larger NoGo N2b amplitudes correlated with fewer NoGo errors, confirming the inhibition function of the child N2b in the NoGo processing chain. All the above studies used a traditional approach to the PCA input — both the Go and NoGo mean ERPs were used together to define the components present during the task. However, recently Barry et al. (2016b) argued that Go/NoGo effects in the ERP components are better detected by PCAs separately implemented on the Go ERP set and the NoGo ERP set. That study demonstrated the occurrence of substantial misallocation of variance (Wood and McCarthy, 1984; Beauducel and Debener, 2003) between components obtained using a combined PCA of the two ERP streams together. Essentially, PCA components provide an optimal summary of the variance at hand, and these will change as the input variance changes from combined Go and NoGo data to separate Go or NoGo. For example, a component that occurs in only one condition may appear to occur in both Go and NoGo conditions if derived from a combined PCA including both sets of ERPs. Further, similar components that occur with different latencies in different conditions may appear to occur at the same latency if a combined PCA is used. Such errors may result in misleading interpretations of the processing supposedly reflected by the components, impeding conceptual development. We found that separate PCAs yielded components that better matched the latencies of the separate Go and NoGo ERPs, and larger components were found in the appropriate stream in the processing schema, as shown in Fig. 1. This methodological shift forms the investigative context for the present study of children's processing in the equiprobable Go/NoGo task. Here we revisit the child data of Barry and De Blasio (2015), and submit the Go and NoGo ERPs to separate PCAs.
2. Method The EEG and performance data from Barry and De Blasio (2015) were re-processed in this study. We provide a minimal outline here; further methodological details can be obtained from the original paper. 2.1. Participants Forty children aged 8–13 years participated; 15 were female; 32 were right-handed. All were recruited via newspaper advertisements, and screened for head injuries and conditions likely to affect the EEG. Participation was voluntary and a parent/guardian provided informed consent in line with the protocol approved by the local ethics committee. 2.2. Task The equiprobable auditory Go/NoGo task was presented in 4 blocks, each containing a random mix of 75 Go and 75 NoGo stimuli. Stimuli were 1000 and 1500 Hz 60 dB SPL tones of 50 ms + 5 ms rise/fall times. Targets were counterbalanced between children, and required a button-press response. 2.3. EEG recording and data quantification Continuous EEG was recorded from 19 sites using an electrode cap with linked-ear reference, and EOG from vertical (left eye) and horizontal placements. All electrodes were tin and all impedances were < 5 kΩ; care was taken to balance the ear impedances. A 16 bit A/D system (AMLAB II) recorded data from 0.3 to 35 Hz at 512 Hz for later analysis. EEG was low-pass filtered to 25 Hz, epoched from − 100 to + 800 ms relative to each stimulus onset, and baselined (− 100 to 0 ms) using Neuroscan software (Compumedics, v. 4.5.1). Single trials with incorrect responses, with RT > 600 ms, or with blinks, muscular or other artefact exceeding ± 100 μV were removed. Mean artefact-free ERPs to correct Go and NoGo trials were then formed. For each child, the mean RT for accepted trials, the standard deviation of these RTs as a measure of variability (RTvar), Go omissions, and NoGo commission errors were recorded as behavioural indices. ERPs from − 100 to + 800 ms relative to stimulus onset were downsampled to 256 Hz to reduce the number of variables/components for extraction and rotation, and hence processing time. As suggested in Barry et al. (2016b), these were subjected to separate temporal PCAs (Go, NoGo). Each PCA used software based on Dien's PCA toolkit (v. 2.23; Dien, 2010) in MATLAB® (The Mathworks, Version 8.0.0.783, R2012b), with covariance input. For the Go data set, the scree slope (Cattell, 1966) had “elbows” at 6 and 9 components, and the Parallel Test (Horn, 1965) 75
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
indicated that 10 components carried more than random variance; the NoGo data set had scree slope “elbows” at 7 and 9 components, and the Parallel Test suggested 12 components. Such uncertainty regarding how many components to extract and rotate has long been a bothersome characteristic of the PCA field. Rather than extracting and rotating any such numbers of components, we follow Kayser and Tenke (2003) in extracting and rotating all components in each PCA — here, 230. Such unrestricted extraction better spreads the noise variance than does selecting a small set, as was common with previous computing limitations. We used Kayser and Tenke's (2003) Varimax 4M function to complete the unrestricted orthogonal Varimax rotation (available at http://psychophysiology.cpmc.columbia.edu/software/). Our procedure ranks all components in variance order, and seeks to identify these in terms of the existing literature. All components carrying more than an arbitrary percentage of the variance are analysed further. In this study we chose 2.5% as our variance threshold, ensuring that no substantive components were omitted. 2.4. Statistical analysis The two sets of unscaled factor loadings (Go, NoGo PCAs) were each compared with the unscaled loadings from Barry and De Blasio (2015), using the Congruence Coefficient (rc; Tucker, 1951) to illuminate latency and amplitude similarities and differences; this is evaluated using a rule of thumb: equality is indicated by rc ≥ 0.95, and fair similarity by 0.85 ≤ rc ≤ 0.94 (Lorenzo-Seva and ten Berge, 2006). Component headmaps represent the amplitude at each scalp site, and the topographies of components were compared by correlating corresponding values from each of the 19 pairs of sites. Mean amplitudes of the current components (at or computed across their maximal site/s), together with age, were used in linear regression analyses to predict each of the behavioural variables separately. Forward step-wise regressions were used with SPSS (Version 21) default settings, and standardised beta coefficients are reported.
Fig. 2. Mean Go and NoGo ERPs at the midline sites. Anteriorisation of the NoGo P3a is clearly evident. Note the large frontal N2 that characterises the child's auditory ERP.
counterparts obtained in the combined PCA (rc ≥ 0.91; see values in the middle row of Fig. 5), except for P3a. 3.2. Topographic component consistency Topographies of the Go components (see Fig. 4) all correlated significantly with those from the combined PCA, r (17) ≥ 0.84, p < 0.001. However, N1-1 was more frontal, and N2c was more central. The LN appears to be in the timeslot of the negative Go version of the NoGo LP component identified previously (and so labelled in quotes here). The electrodes contributing to the mean Go component amplitudes for use in the later regressions were: F3, Fz, F4, and C4 for N1-1; F8, F4, and T8 for PN; Cz for N2c; P7, P3, and Pz for P3b; and Fz for the LN. Topographies of the separate NoGo components (see Fig. 5) all correlated significantly with those from the combined PCA, r (17) ≥ 0.72, p < 0.001, except for P3a, which was enhanced in hemispheric regions compared with the midline. The NegSW component appears to occupy the P3b timeslot from the combined PCA. The electrodes used as estimates for NoGo amplitudes were: F3, Fz, and F4 for N1-1; F8, F4, and T8 for PN; Fz, C3, Cz, and C4 for N2b; Fp1, Fp2, F3, and F4 for N2c; Fz and Cz for P3a; Fz for the NegSW; and C3, C4, and Pz for the LP.
3. Results Go ERPs included 58–270 trials (M = 161.0, SD = 44.3), while NoGo included 76–272 trials (M = 169.2, SD = 49.5), a non-significant difference. Fig. 2 shows the mean ERPs at the midline sites for each condition, with major peaks labelled at Fz. The large child N2 is apparent, as is anteriorisation of the NoGo P3a compared with the posterior Go P3b. The Go PCA yielded 5 identifiable components, carrying a total of 88% of the variance. By comparison with the components reported in Barry and De Blasio (2015), these were identifiable as occurring in the temporal order corresponding to N1-1, PN, N2c, P3b, and LP. The NoGo PCA yielded 7 identifiable components, also carrying 88% of the variance; these corresponded to the N1-1, PN, N2b, N2c, P3a, P3b, and LP reported in Barry and De Blasio (2015). 3.1. Temporal component consistency Fig. 3 displays the scaled factor loadings from the present PCAs below the corresponding components from the combined PCA of Barry and De Blasio (2015). In general there was close similarity in the pattern of factor loadings from the three PCAs. The present Go PCA did not yield a substantial N2b or P3a, and had a Late Negativity (LN) rather than LP. The Congruence Coefficients (based on the unscaled loadings; see values in the middle row of Fig. 4) indicated equivalence between the factors in the Go PCA and their counterparts obtained in the combined PCA (rc ≥ 0.96). The NoGo PCA did not yield a substantial P3b but rather a smaller Negative Slow Wave (NegSW), and NoGo P3a appeared later and larger than the combined P3a. The Congruence Coefficients indicated fair similarity to equivalence between the factors in the NoGo PCA and their
3.3. Regressions The Go omission errors, RT, and RT variability were each regressed on Age and the Go ERP component amplitudes in three separate stepwise regressions. The Go omission errors were inversely related to Age (β = −0.42, t = − 2.82, p = 0.008); i.e., as Age increased, omission errors decreased. Go RT was directly related to N1-1 (β = 0.43, t = 2.90, p = 0.006); that is, as N1-1 increased in amplitude (i.e., became more negative), RT was reduced. Go RT variability was inversely related to Age (β = − 0.33, t = −2.13, p = 0.040); in other words, as age increased, RT variability decreased. The NoGo commission errors were regressed on Age and the NoGo 76
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 3. Scaled factor loadings from the combined PCA of Barry and De Blasio (2015) above the separate Go and NoGo PCA loadings in the present study. Dashed lines are dropped from the top panel to mark the peak latencies of the combined components. Figure is shown in colour on the web.
Fig. 4. Go component outcomes from the combined PCA of Barry and De Blasio (2015) above those from the separate Go PCA. Temporal (rc) and topographic (r(17)) consistency data are shown between the headmaps for each corresponding component. Figure is shown in colour on the web.
77
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 5. NoGo component outcomes from the combined PCA of Barry and De Blasio (2015) above those from the separate NoGo PCA. Temporal (rc) and topographic (r (17)) consistency data are shown between the headmaps for each corresponding component. Figure is shown in colour on the web.
studies such as this, using separate PCAs for each condition to clarify the stimulus-specific components and their links with task performance. The present study indicates similar N1-1 in response to both Go and NoGo stimuli, compatible with similar levels of sensory processing, but the larger PN to Go indicates differential categorization and enhanced processing of the Go stimulus. Subsequently two separate processing streams are apparent: Go is associated with N2c, P3b, and LN, while NoGo is associated with N2b, N2c, P3a, NegSW, and LP. Comparison of these sequences with Fig. 1 suggests substantial similarities, but also some differences. In temporal order, the early P1 and N1-3 are not found consistently even in adults, being of small variance; the following N1-1 and PN fit well. The following P2/N2 timeslot has had varying P2 results over studies, but the present N2b and N2c fit well, as do the traditional Go N2c-P3b and NoGo N2b-P3a links to the P300/Late Positive Complex (LPC) timeslot. The present NoGo N2c in the same timeslot as the Go N2c was unexpected, and would appear to contradict the N2c-P3b/N2b-P3a pattern expected from the literature (Barry and De Blasio, 2015; Barry et al., 2014a). We note however, that the child NoGo N2c has a distinctly frontal topography compared with the central Go N2c, indicating different cortical sources, and suggesting that these might be different subcomponents of N2c subserving different functions. This needs further exploration in future child studies. The classic SW reported in our previous child studies at 544.9 ms carried < 3% of the variance there and was not noted here. However, the novel NoGo NegSW may relate to that aspect of the LPC. The final NoGo component previously identified, LP, was confirmed here, and appears to be complimented by the Go LN. These components have been inserted in a new sequential processing schema
ERP component amplitudes. Commission errors were directly related to N2b (β = 0.38, t = 2.63, p = 0.012) and inversely related to NegSW (β = − 0.41, t = − 2.91, p = 0.006). That is, larger N2b predicted fewer commission errors, and larger NegSW was associated with more commission errors.
4. Discussion Barry et al. (2016b) demonstrated that misallocation of variance between components occurs with the usual PCA over ERPs combined from the Go and NoGo conditions, and this provided the impetus for the present study. The impact of this issue can clearly be seen when comparing the loading structures in Fig. 3. Essentially, the top set of loadings (A) from the combined PCA of Barry and De Blasio (2015) reflects an approximate average of the loadings from the separate PCAs (B and C) shown below. For example, the combined PN in A roughly averages the later PN in B and the earlier PN in C; N2b and P3a appear in C, but not B, while P3b is large in B and corresponding activity is largely absent in C. While these thus appear to be large in one condition only, the combined PCA output implies they occur in both due to their variance being (at least partly) misallocated from the condition in which they dominate to the other. Despite this misallocation of variance in the combined PCA, the broad similarity between the combined PCA components in Barry and De Blasio (2015) and the present results, together with the prior adult combined/separate PCA findings in Barry et al. (2016b) supports the broad outline of the sequential processing schema shown in Fig. 1. However, the adult schema would clearly benefit from being refined by
Fig. 6. Updated schema for children's sequential processing in the equiprobable auditory Go/NoGo paradigm.
78
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
for children (Fig. 6). In this figure, components that appear robust and consistent in their relation to stimulus processing stages are shown in bold font, while those needing further investigation are shown using fainter font. The regression analyses attempting to link behavioural performance measures to these components and the children's age have provided some interesting outcomes. The percentage of Go omission errors, and RT variability, both decreased with increasing age. That is, older children were more consistent in their RT and made fewer omission errors, but these behaviours were not related to Go component amplitudes. Interestingly, RT reduced in children with larger Go N1-1. This novel finding appears compatible with a general attention effect (Näätänen and Picton, 1987) rather than Go-specific processing. The NoGo component links with commission errors were more informative. Children with larger N2b amplitudes showed fewer commission errors, while those with larger NegSW amplitudes displayed more commission errors. The N2b finding is compatible with that found in Barry and De Blasio (2015), and confirms their link of the child NoGo N2b with inhibitory processing in the equiprobable Go/NoGo task, as has long been associated with this component in adult studies of traditional Go/NoGo paradigms (see Huster et al., 2013 for a review). The NegSW finding is novel, and its opposite direction to the N2b link suggests a different role. Indeed, this is demanded by the large difference in peak latency (441 ms cf. 219 ms). The child mean Go RT was 359.4 ms, so the NoGo NegSW is likely to reflect a late evaluative process (van Veen and Carter, 2006), in which the child evaluates the adequacy of their NoGo performance (i.e., non-response). Although not related to age, the results suggest that such evaluation is more prominent in children with more NoGo errors, suggesting an individual-difference performance variable. The present NegSW component was obtained from correct NoGo trials, and it would be interesting to investigate links between this component and the error-related negativity (ERN) in response-locked NoGo error trials. The mean commission error rate found here was 7.6% (range 1.0–29.8%), which suggests that a more-difficult paradigm would be necessary to obtain sufficient trials for a good SNR for the ERN. In the light of these findings for NoGo N2b and NegSW, the model has been extended in children — the final NoGo “Terminate Processing” has been overlapped by a new “Inhibition and Evaluation” stage, reflecting important cognitive control processes. As the role of the P3a in inhibitory processing is still not resolved (e.g., Smith et al., 2006), these processes are not separated in the model. The present study used separate PCAs to decompose the Go and NoGo ERPs from a child sample. We aimed to optimise the data-driven decomposition of the electrophysiological markers of the cognitive processes involved in error-free execution of the task. This avoided misallocation of variance between components in one condition and the other, as occurs when the ERPs from different conditions are analysed jointly. This methodology is recommended for wider application in the field. We then explored potential linkages between age and these components with their corresponding Go or NoGo performance variables. The results confirmed the main stages of a sequential processing schema developed using adult data, and extended this in relation to children's ERPs and behavioural performance, including the addition of new insights into cognitive control processes such as inhibition and performance evaluation. The extended schema should prove useful in understanding child ERPs, cognitive processing, and performance in a range of two-stimulus discrimination tasks, and points to potential extensions in relation to the adult schema.
Go/NoGo task. Int. J. Psychophysiol. 86, 251–261. http://dx.doi.org/10.1016/j. ijpsycho.2012.10.001. Barry, R.J., De Blasio, F.M., 2013. Sequential processing in the equiprobable auditory Go/ NoGo task: a temporal PCA study. Int. J. Psychophysiol. 89, 123–127. http://dx.doi. org/10.1016/j.ijpsycho.2013.06.012. Barry, R.J., De Blasio, F.M., 2015. ERP components and performance in the equiprobable Go/NoGo task: inhibition in children. Psychophysiology 52, 1228–1237. http://dx. doi.org/10.1111/psyp.12447. Barry, R.J., Rushby, J.A., 2006. An orienting reflex perspective on anteriorisation of the P3 of the event-related potential. Exp. Brain Res. 173, 539–545. http://dx.doi.org/ 10.1007/s00221-006-0590-8. Barry, R.J., De Blasio, F.M., Rushby, J.A., Clarke, A.R., 2010. Brain dynamics in the auditory Go/NoGo task as a function of EEG frequency. Int. J. Psychophysiol. 78, 115–128. http://dx.doi.org/10.1016/j.ijpsycho.2010.06.357. Barry, R.J., De Blasio, F.M., Borchard, J.P., 2014a. Sequential processing in the equiprobable auditory Go/NoGo task: children vs. adults. Clin. Neurophysiol. 125, 1995–2006. http://dx.doi.org/10.1016/j.clinph.2014.02.018. Barry, R.J., De Blasio, F.M., Cave, A.E., 2014b. Caffeine effects on ERP components and performance in an equiprobable auditory Go/NoGo task. J. Caff. Res. 4, 83–92. http://dx.doi.org/10.1089/jcr.2014.0011. Barry, R.J., De Blasio, F.M., De Pascalis, V., Karamacoska, D., 2014c. Preferred EEG brain states at stimulus onset in a fixed interstimulus interval equiprobable auditory Go/ NoGo task: a definitive study. Int. J. Psychophysiol. 94, 42–58. http://dx.doi.org/10. 1016/j.ijpsycho.2014.07.005. Barry, R.J., De Blasio, F.M., Cave, A.E., 2016a. Sequential processing in young and older adults in the equiprobable auditory Go/NoGo task. Clin. Neurophysiol. 127, 2273–2285. http://dx.doi.org/10.1016/j.clinph.2016.02.010. Barry, R.J., De Blasio, F.M., Fogarty, J.S., Karamacoska, D., 2016b. ERP Go/NoGo condition effects are better detected with separate PCAs. Int. J. Psychophysiol. 50–64. http://dx.doi.org/10.1016/j.ijpsycho.2016.06.003. Beauducel, A., Debener, S., 2003. Misallocation of variance in event-related potentials: simulation studies on the effects of test power, topography, and baseline-to-peak versus principal component quantifications. J. Neurosci. Methods 124, 103–112. http://dx.doi.org/10.1016/S0165-0270(02)00381-3. Borchard, J.P., Barry, R.J., De Blasio, F.M., 2015. Sequential processing in an auditory equiprobable Go/NoGo task with variable interstimulus interval. Int. J. Psychophysiol. 97, 145–152. Cattell, R.B., 1966. The scree test for the number of factors. Multivar. Behav. Res. 1, 245–276. http://dx.doi.org/10.1207/s15327906mbr0102_10. De Blasio, F.M., Barry, R.J., 2013a. Prestimulus delta and theta determinants of ERP responses in the Go/NoGo task. Int. J. Psychophysiol. 87, 279–288. http://dx.doi. org/10.1016/j.ijpsycho.2012.09.016. De Blasio, F.M., Barry, R.J., 2013b. Prestimulus alpha and beta determinants of ERP responses in the Go/NoGo task. Int. J. Psychophysiol. 89, 9–17. http://dx.doi.org/10. 1016/j.ijpsycho.2013.04.018. Dien, J., 2010. The ERP PCA Toolkit: an open source program for advanced statistical analysis of event-related potential data. J. Neurosci. Methods 187, 138–145. http:// dx.doi.org/10.1016/j.jneumeth.2009.12.009. Folstein, J.R., van Petten, C., 2008. Influence of cognitive control and mismatch on the N2 component of the ERP: a review. Psychophysiology 45, 152–170. http://dx.doi.org/ 10.1111/j.1469-8986.2007.00602.x. Horn, J.L., 1965. A rationale and test for the number of factors in factor analysis. Psychometrika 30, 179–185. http://dx.doi.org/10.1007/BF02289447. Huster, R.J., Enriquez-Geppert, S., Lavallee, C.F., Falkenstein, M., Herrmann, C.S., 2013. Electroencephalography of response inhibition tasks: functional networks and cognitive contributions. Int. J. Psychophysiol. 87, 217–233. http://dx.doi.org/10.1016/ j.ijpsycho.2012.08.001. Kayser, J., Tenke, C.E., 2003. Optimizing PCA methodology for ERP component identification and measurement: theoretical rationale and empirical evaluation. Clin. Neurophysiol. 114, 2307–2325. http://dx.doi.org/10.1016/S1388-2457(03) 00241-4. Lorenzo-Seva, U., ten Berge, J.M.F., 2006. Tucker's congruence coefficient as a meaningful index of factor similarity. Methodology 2, 57–64. http://dx.doi.org/10.1027/ 1614-1881.2.2.57. Näätänen, R., Picton, T., 1987. The N1 wave of the human electric and magnetic response to sound: a review and an analysis of the component structure. Psychophysiology 24, 375–425. http://dx.doi.org/10.1111/j.1469-8986.1987.tb00311.x. Smith, J.L., Johnstone, S.J., Barry, R.J., 2006. Response priming in the Go/NoGo task: the N2 reflects neither inhibition nor conflict. Clin. Neurophysiol. 118, 343–355. http:// dx.doi.org/10.1016/j.clinph.2006.09.027. Tucker, L.R., 1951. A Method for Synthesis of Factor Analysis Studies (Personnel Research Section Report No. 984). Department of the Army, Washington, D.C. van Veen, V., Carter, C.S., 2006. Conflict and cognitive control in the brain. Curr. Dir. Psychol. Sci. 15, 237–240. http://dx.doi.org/10.1111/j.1467-8721.2006.00443.x. Wood, C.C., McCarthy, G., 1984. Principal component analysis of event-related potentials: simulation studies demonstrate misallocation of variance across components. Electroencephalogr. Clin. Neurophysiol. 59, 249–260.
References Barry, R.J., De Blasio, F.M., 2012. EEG-ERP phase dynamics of children in the auditory
79
Contents lists available at ScienceDirect
International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho
A processing schema for children in the auditory equiprobable Go/NoGo task: ERP components and behaviour
T
⁎
Robert J. Barry , Frances M. De Blasio, Jack S. Fogarty Brain & Behaviour Research Institute, and School of Psychology, University of Wollongong, Northfields Avenue, Wollongong, NSW 2522, Australia
A R T I C L E I N F O
A B S T R A C T
Keywords: Event-related potentials (ERPs) Equiprobable Go/NoGo task Principal components analyses (PCA) Misallocation of variance Sequential processing schema Children
A sequential processing model for adults in the auditory equiprobable Go/NoGo task has been developed in recent years. This used temporal principal components analysis (PCA) to decompose Go/NoGo event related potential (ERP) data into components that mark stages of perceptual and cognitive processing. The model has been found useful in frameworking several studies in young and older adults, and in children. Recently, it has been demonstrated that the common PCA approach of decomposing Go and NoGo ERP data together results in misallocation of variance between the conditions, distorting the timing, topography, and amplitudes of the resultant components in each condition. The present study thus reanalyses data from a child study, conducting separate PCAs on the data from each condition. Multiple regression was then used to seek links with behavioural measures from the task. In addition to confirming the previous NoGo N2b/inhibitory processing link, novel NoGo Negative Slow Wave/error evaluation and Go N1-1/RT variability links were obtained. Based on these outcomes, the recommended separate application of PCAs to Go and NoGo data was confirmed. The present data were used to develop a child-specific sequential processing schema for this paradigm, suggesting earlier separation of the Go and NoGo processing chains, and the need to include an additional inhibition and evaluation stage. The child schema should be useful in future studies involving this and other two-choice reaction tasks.
1. Introduction The unwarned equiprobable auditory Go/NoGo task is an auditory intermediary between the traditional active Oddball task, with rare targets (probability < 50%) amongst frequent standards, and the traditional Go/NoGo task, with frequent Go targets (probability > 50%) amongst rare NoGo stimuli. Apart from target/Go probability, these two-stimulus discrimination tasks share common stimulus evaluation processes at each stimulus presentation, leading to discrimination between one stimulus to be responded to and one that does not require a response. Subsequent differential processing chains involve response preparation and execution processes, or cessation/inhibition of further processing. These different control processes (Huster et al., 2013; Folstein and van Petten, 2008) may be subject to monitoring (van Veen and Carter, 2006) as the participant attempts to optimise performance through the task. We were initially interested in this task because it provides equal numbers of trials to form separate event related potentials (ERPs) to Go and NoGo stimuli with similar signal:noise ratio (SNR), facilitating exploration of their prestimulus electroencephalographic (EEG) determinants (Barry et al., 2010). Subsequently we continued brain
⁎
dynamics studies exploring prestimulus EEG amplitude effects on the ERP components in this task (De Blasio and Barry, 2013a, 2013b), as well as the impact of the EEG phase of narrow-band and traditional (delta, theta, alpha, and beta) bands at stimulus onset (Barry and De Blasio, 2012; Barry et al., 2014c, submitted). In parallel, we embarked on a program to obtain more understanding of the components constituting these ERPs, and employed temporal principal components analysis (PCA) to obtain data-driven decompositions of the Go and NoGo ERPs for that purpose. Barry and De Blasio (2013) reported our first sequential processing interpretation of the Go/NoGo processing stages using temporal PCA of the ERPs from a sample of young adults. In that schema, summarised in Fig. 1, we suggested that N1 (with its “true” subcomponents N1-3 and N1-1; and Processing Negativity, PN; Näätänen and Picton, 1987) marked the beginning of Go/NoGo differentiation, with complete differentiation or categorization marked by N2. This P2/N2 complex was followed by two separate processing chains. These chains lead to either the Go response (marked by P3b and the classic Slow Wave, SW) or the NoGo non-response (marked by the P3a and a novel Late Positivity, LP). LP appeared at the end of the NoGo epoch, apparently earlier in NoGo than Go, marking the earlier termination of stimulus processing in the
Corresponding author at: School of Psychology, University of Wollongong, Wollongong, NSW 2522, Australia. E-mail address: [email protected] (R.J. Barry).
https://doi.org/10.1016/j.ijpsycho.2017.10.014 Received 5 July 2017; Received in revised form 30 October 2017; Accepted 31 October 2017 Available online 06 November 2017 0167-8760/ © 2017 Elsevier B.V. All rights reserved.
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 1. Barry and De Blasio's (2013) early conceptualisation of the adult processing schema for the equiprobable Go/ NoGo task. PN: Processing Negativity; SW: Slow Wave; LP: Late Positivity.
We expect to confirm that stimulus-specific components are emphasised in the separate streams shown in the processing schema, and to confirm that the NoGo N2b is linked to response inhibition in children. We also take the opportunity to seek other component-function links clarifying the determinants of the behavioural measures characterising child performance in this paradigm.
absence of the active Go response. This schema was subsequently used to framework a number of studies in young adults exploring brain dynamics (Barry et al., 2014c, submitted), the effects of caffeine (Barry et al., 2014b), the impact of a varying interstimulus interval (Borchard et al., 2015), differences between young adults and children (Barry et al., 2014a), and the effects of ageing in young and older adults (Barry et al., 2016a). As well as helping to explore these different topics, the stability of the component structures obtained in the studies helped clarify and refine details of the processing schema as new data were linked to a number of components. Children show both similarities and differences relative to adults in their behaviour and ERPs across the range of two-choice active discrimination tasks, and this was explored in the context of the sequential processing schema in Barry et al. (2014a). Early (N1-1, and PN) and late components (P3, SW, LP) were similar between 18 children and 18 young adults, but the intermediate P2 and N2 differed substantially in their stimulus effects. Barry and De Blasio (2015) explored the child response pattern in a group of 40 children aged 8–13 years, focussing on whether the NoGo N2b was linked to inhibition, given our view that the equiprobable task was so simple that active inhibition of the Go response was not required. This view reflected an earlier ERP source study by Barry and Rushby (2006), which noted that electrodermal responses were smaller to the NoGo than Go stimuli in this paradigm, suggesting less effortful processing. However, the observation in Barry et al. (2014a) of child N2b enhancement, and children's difficulty with the task, suggested the need to revisit this question, at least in children. Barry and De Blasio (2015) found that larger NoGo N2b amplitudes correlated with fewer NoGo errors, confirming the inhibition function of the child N2b in the NoGo processing chain. All the above studies used a traditional approach to the PCA input — both the Go and NoGo mean ERPs were used together to define the components present during the task. However, recently Barry et al. (2016b) argued that Go/NoGo effects in the ERP components are better detected by PCAs separately implemented on the Go ERP set and the NoGo ERP set. That study demonstrated the occurrence of substantial misallocation of variance (Wood and McCarthy, 1984; Beauducel and Debener, 2003) between components obtained using a combined PCA of the two ERP streams together. Essentially, PCA components provide an optimal summary of the variance at hand, and these will change as the input variance changes from combined Go and NoGo data to separate Go or NoGo. For example, a component that occurs in only one condition may appear to occur in both Go and NoGo conditions if derived from a combined PCA including both sets of ERPs. Further, similar components that occur with different latencies in different conditions may appear to occur at the same latency if a combined PCA is used. Such errors may result in misleading interpretations of the processing supposedly reflected by the components, impeding conceptual development. We found that separate PCAs yielded components that better matched the latencies of the separate Go and NoGo ERPs, and larger components were found in the appropriate stream in the processing schema, as shown in Fig. 1. This methodological shift forms the investigative context for the present study of children's processing in the equiprobable Go/NoGo task. Here we revisit the child data of Barry and De Blasio (2015), and submit the Go and NoGo ERPs to separate PCAs.
2. Method The EEG and performance data from Barry and De Blasio (2015) were re-processed in this study. We provide a minimal outline here; further methodological details can be obtained from the original paper. 2.1. Participants Forty children aged 8–13 years participated; 15 were female; 32 were right-handed. All were recruited via newspaper advertisements, and screened for head injuries and conditions likely to affect the EEG. Participation was voluntary and a parent/guardian provided informed consent in line with the protocol approved by the local ethics committee. 2.2. Task The equiprobable auditory Go/NoGo task was presented in 4 blocks, each containing a random mix of 75 Go and 75 NoGo stimuli. Stimuli were 1000 and 1500 Hz 60 dB SPL tones of 50 ms + 5 ms rise/fall times. Targets were counterbalanced between children, and required a button-press response. 2.3. EEG recording and data quantification Continuous EEG was recorded from 19 sites using an electrode cap with linked-ear reference, and EOG from vertical (left eye) and horizontal placements. All electrodes were tin and all impedances were < 5 kΩ; care was taken to balance the ear impedances. A 16 bit A/D system (AMLAB II) recorded data from 0.3 to 35 Hz at 512 Hz for later analysis. EEG was low-pass filtered to 25 Hz, epoched from − 100 to + 800 ms relative to each stimulus onset, and baselined (− 100 to 0 ms) using Neuroscan software (Compumedics, v. 4.5.1). Single trials with incorrect responses, with RT > 600 ms, or with blinks, muscular or other artefact exceeding ± 100 μV were removed. Mean artefact-free ERPs to correct Go and NoGo trials were then formed. For each child, the mean RT for accepted trials, the standard deviation of these RTs as a measure of variability (RTvar), Go omissions, and NoGo commission errors were recorded as behavioural indices. ERPs from − 100 to + 800 ms relative to stimulus onset were downsampled to 256 Hz to reduce the number of variables/components for extraction and rotation, and hence processing time. As suggested in Barry et al. (2016b), these were subjected to separate temporal PCAs (Go, NoGo). Each PCA used software based on Dien's PCA toolkit (v. 2.23; Dien, 2010) in MATLAB® (The Mathworks, Version 8.0.0.783, R2012b), with covariance input. For the Go data set, the scree slope (Cattell, 1966) had “elbows” at 6 and 9 components, and the Parallel Test (Horn, 1965) 75
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
indicated that 10 components carried more than random variance; the NoGo data set had scree slope “elbows” at 7 and 9 components, and the Parallel Test suggested 12 components. Such uncertainty regarding how many components to extract and rotate has long been a bothersome characteristic of the PCA field. Rather than extracting and rotating any such numbers of components, we follow Kayser and Tenke (2003) in extracting and rotating all components in each PCA — here, 230. Such unrestricted extraction better spreads the noise variance than does selecting a small set, as was common with previous computing limitations. We used Kayser and Tenke's (2003) Varimax 4M function to complete the unrestricted orthogonal Varimax rotation (available at http://psychophysiology.cpmc.columbia.edu/software/). Our procedure ranks all components in variance order, and seeks to identify these in terms of the existing literature. All components carrying more than an arbitrary percentage of the variance are analysed further. In this study we chose 2.5% as our variance threshold, ensuring that no substantive components were omitted. 2.4. Statistical analysis The two sets of unscaled factor loadings (Go, NoGo PCAs) were each compared with the unscaled loadings from Barry and De Blasio (2015), using the Congruence Coefficient (rc; Tucker, 1951) to illuminate latency and amplitude similarities and differences; this is evaluated using a rule of thumb: equality is indicated by rc ≥ 0.95, and fair similarity by 0.85 ≤ rc ≤ 0.94 (Lorenzo-Seva and ten Berge, 2006). Component headmaps represent the amplitude at each scalp site, and the topographies of components were compared by correlating corresponding values from each of the 19 pairs of sites. Mean amplitudes of the current components (at or computed across their maximal site/s), together with age, were used in linear regression analyses to predict each of the behavioural variables separately. Forward step-wise regressions were used with SPSS (Version 21) default settings, and standardised beta coefficients are reported.
Fig. 2. Mean Go and NoGo ERPs at the midline sites. Anteriorisation of the NoGo P3a is clearly evident. Note the large frontal N2 that characterises the child's auditory ERP.
counterparts obtained in the combined PCA (rc ≥ 0.91; see values in the middle row of Fig. 5), except for P3a. 3.2. Topographic component consistency Topographies of the Go components (see Fig. 4) all correlated significantly with those from the combined PCA, r (17) ≥ 0.84, p < 0.001. However, N1-1 was more frontal, and N2c was more central. The LN appears to be in the timeslot of the negative Go version of the NoGo LP component identified previously (and so labelled in quotes here). The electrodes contributing to the mean Go component amplitudes for use in the later regressions were: F3, Fz, F4, and C4 for N1-1; F8, F4, and T8 for PN; Cz for N2c; P7, P3, and Pz for P3b; and Fz for the LN. Topographies of the separate NoGo components (see Fig. 5) all correlated significantly with those from the combined PCA, r (17) ≥ 0.72, p < 0.001, except for P3a, which was enhanced in hemispheric regions compared with the midline. The NegSW component appears to occupy the P3b timeslot from the combined PCA. The electrodes used as estimates for NoGo amplitudes were: F3, Fz, and F4 for N1-1; F8, F4, and T8 for PN; Fz, C3, Cz, and C4 for N2b; Fp1, Fp2, F3, and F4 for N2c; Fz and Cz for P3a; Fz for the NegSW; and C3, C4, and Pz for the LP.
3. Results Go ERPs included 58–270 trials (M = 161.0, SD = 44.3), while NoGo included 76–272 trials (M = 169.2, SD = 49.5), a non-significant difference. Fig. 2 shows the mean ERPs at the midline sites for each condition, with major peaks labelled at Fz. The large child N2 is apparent, as is anteriorisation of the NoGo P3a compared with the posterior Go P3b. The Go PCA yielded 5 identifiable components, carrying a total of 88% of the variance. By comparison with the components reported in Barry and De Blasio (2015), these were identifiable as occurring in the temporal order corresponding to N1-1, PN, N2c, P3b, and LP. The NoGo PCA yielded 7 identifiable components, also carrying 88% of the variance; these corresponded to the N1-1, PN, N2b, N2c, P3a, P3b, and LP reported in Barry and De Blasio (2015). 3.1. Temporal component consistency Fig. 3 displays the scaled factor loadings from the present PCAs below the corresponding components from the combined PCA of Barry and De Blasio (2015). In general there was close similarity in the pattern of factor loadings from the three PCAs. The present Go PCA did not yield a substantial N2b or P3a, and had a Late Negativity (LN) rather than LP. The Congruence Coefficients (based on the unscaled loadings; see values in the middle row of Fig. 4) indicated equivalence between the factors in the Go PCA and their counterparts obtained in the combined PCA (rc ≥ 0.96). The NoGo PCA did not yield a substantial P3b but rather a smaller Negative Slow Wave (NegSW), and NoGo P3a appeared later and larger than the combined P3a. The Congruence Coefficients indicated fair similarity to equivalence between the factors in the NoGo PCA and their
3.3. Regressions The Go omission errors, RT, and RT variability were each regressed on Age and the Go ERP component amplitudes in three separate stepwise regressions. The Go omission errors were inversely related to Age (β = −0.42, t = − 2.82, p = 0.008); i.e., as Age increased, omission errors decreased. Go RT was directly related to N1-1 (β = 0.43, t = 2.90, p = 0.006); that is, as N1-1 increased in amplitude (i.e., became more negative), RT was reduced. Go RT variability was inversely related to Age (β = − 0.33, t = −2.13, p = 0.040); in other words, as age increased, RT variability decreased. The NoGo commission errors were regressed on Age and the NoGo 76
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 3. Scaled factor loadings from the combined PCA of Barry and De Blasio (2015) above the separate Go and NoGo PCA loadings in the present study. Dashed lines are dropped from the top panel to mark the peak latencies of the combined components. Figure is shown in colour on the web.
Fig. 4. Go component outcomes from the combined PCA of Barry and De Blasio (2015) above those from the separate Go PCA. Temporal (rc) and topographic (r(17)) consistency data are shown between the headmaps for each corresponding component. Figure is shown in colour on the web.
77
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
Fig. 5. NoGo component outcomes from the combined PCA of Barry and De Blasio (2015) above those from the separate NoGo PCA. Temporal (rc) and topographic (r (17)) consistency data are shown between the headmaps for each corresponding component. Figure is shown in colour on the web.
studies such as this, using separate PCAs for each condition to clarify the stimulus-specific components and their links with task performance. The present study indicates similar N1-1 in response to both Go and NoGo stimuli, compatible with similar levels of sensory processing, but the larger PN to Go indicates differential categorization and enhanced processing of the Go stimulus. Subsequently two separate processing streams are apparent: Go is associated with N2c, P3b, and LN, while NoGo is associated with N2b, N2c, P3a, NegSW, and LP. Comparison of these sequences with Fig. 1 suggests substantial similarities, but also some differences. In temporal order, the early P1 and N1-3 are not found consistently even in adults, being of small variance; the following N1-1 and PN fit well. The following P2/N2 timeslot has had varying P2 results over studies, but the present N2b and N2c fit well, as do the traditional Go N2c-P3b and NoGo N2b-P3a links to the P300/Late Positive Complex (LPC) timeslot. The present NoGo N2c in the same timeslot as the Go N2c was unexpected, and would appear to contradict the N2c-P3b/N2b-P3a pattern expected from the literature (Barry and De Blasio, 2015; Barry et al., 2014a). We note however, that the child NoGo N2c has a distinctly frontal topography compared with the central Go N2c, indicating different cortical sources, and suggesting that these might be different subcomponents of N2c subserving different functions. This needs further exploration in future child studies. The classic SW reported in our previous child studies at 544.9 ms carried < 3% of the variance there and was not noted here. However, the novel NoGo NegSW may relate to that aspect of the LPC. The final NoGo component previously identified, LP, was confirmed here, and appears to be complimented by the Go LN. These components have been inserted in a new sequential processing schema
ERP component amplitudes. Commission errors were directly related to N2b (β = 0.38, t = 2.63, p = 0.012) and inversely related to NegSW (β = − 0.41, t = − 2.91, p = 0.006). That is, larger N2b predicted fewer commission errors, and larger NegSW was associated with more commission errors.
4. Discussion Barry et al. (2016b) demonstrated that misallocation of variance between components occurs with the usual PCA over ERPs combined from the Go and NoGo conditions, and this provided the impetus for the present study. The impact of this issue can clearly be seen when comparing the loading structures in Fig. 3. Essentially, the top set of loadings (A) from the combined PCA of Barry and De Blasio (2015) reflects an approximate average of the loadings from the separate PCAs (B and C) shown below. For example, the combined PN in A roughly averages the later PN in B and the earlier PN in C; N2b and P3a appear in C, but not B, while P3b is large in B and corresponding activity is largely absent in C. While these thus appear to be large in one condition only, the combined PCA output implies they occur in both due to their variance being (at least partly) misallocated from the condition in which they dominate to the other. Despite this misallocation of variance in the combined PCA, the broad similarity between the combined PCA components in Barry and De Blasio (2015) and the present results, together with the prior adult combined/separate PCA findings in Barry et al. (2016b) supports the broad outline of the sequential processing schema shown in Fig. 1. However, the adult schema would clearly benefit from being refined by
Fig. 6. Updated schema for children's sequential processing in the equiprobable auditory Go/NoGo paradigm.
78
International Journal of Psychophysiology 123 (2018) 74–79
R.J. Barry et al.
for children (Fig. 6). In this figure, components that appear robust and consistent in their relation to stimulus processing stages are shown in bold font, while those needing further investigation are shown using fainter font. The regression analyses attempting to link behavioural performance measures to these components and the children's age have provided some interesting outcomes. The percentage of Go omission errors, and RT variability, both decreased with increasing age. That is, older children were more consistent in their RT and made fewer omission errors, but these behaviours were not related to Go component amplitudes. Interestingly, RT reduced in children with larger Go N1-1. This novel finding appears compatible with a general attention effect (Näätänen and Picton, 1987) rather than Go-specific processing. The NoGo component links with commission errors were more informative. Children with larger N2b amplitudes showed fewer commission errors, while those with larger NegSW amplitudes displayed more commission errors. The N2b finding is compatible with that found in Barry and De Blasio (2015), and confirms their link of the child NoGo N2b with inhibitory processing in the equiprobable Go/NoGo task, as has long been associated with this component in adult studies of traditional Go/NoGo paradigms (see Huster et al., 2013 for a review). The NegSW finding is novel, and its opposite direction to the N2b link suggests a different role. Indeed, this is demanded by the large difference in peak latency (441 ms cf. 219 ms). The child mean Go RT was 359.4 ms, so the NoGo NegSW is likely to reflect a late evaluative process (van Veen and Carter, 2006), in which the child evaluates the adequacy of their NoGo performance (i.e., non-response). Although not related to age, the results suggest that such evaluation is more prominent in children with more NoGo errors, suggesting an individual-difference performance variable. The present NegSW component was obtained from correct NoGo trials, and it would be interesting to investigate links between this component and the error-related negativity (ERN) in response-locked NoGo error trials. The mean commission error rate found here was 7.6% (range 1.0–29.8%), which suggests that a more-difficult paradigm would be necessary to obtain sufficient trials for a good SNR for the ERN. In the light of these findings for NoGo N2b and NegSW, the model has been extended in children — the final NoGo “Terminate Processing” has been overlapped by a new “Inhibition and Evaluation” stage, reflecting important cognitive control processes. As the role of the P3a in inhibitory processing is still not resolved (e.g., Smith et al., 2006), these processes are not separated in the model. The present study used separate PCAs to decompose the Go and NoGo ERPs from a child sample. We aimed to optimise the data-driven decomposition of the electrophysiological markers of the cognitive processes involved in error-free execution of the task. This avoided misallocation of variance between components in one condition and the other, as occurs when the ERPs from different conditions are analysed jointly. This methodology is recommended for wider application in the field. We then explored potential linkages between age and these components with their corresponding Go or NoGo performance variables. The results confirmed the main stages of a sequential processing schema developed using adult data, and extended this in relation to children's ERPs and behavioural performance, including the addition of new insights into cognitive control processes such as inhibition and performance evaluation. The extended schema should prove useful in understanding child ERPs, cognitive processing, and performance in a range of two-stimulus discrimination tasks, and points to potential extensions in relation to the adult schema.
Go/NoGo task. Int. J. Psychophysiol. 86, 251–261. http://dx.doi.org/10.1016/j. ijpsycho.2012.10.001. Barry, R.J., De Blasio, F.M., 2013. Sequential processing in the equiprobable auditory Go/ NoGo task: a temporal PCA study. Int. J. Psychophysiol. 89, 123–127. http://dx.doi. org/10.1016/j.ijpsycho.2013.06.012. Barry, R.J., De Blasio, F.M., 2015. ERP components and performance in the equiprobable Go/NoGo task: inhibition in children. Psychophysiology 52, 1228–1237. http://dx. doi.org/10.1111/psyp.12447. Barry, R.J., Rushby, J.A., 2006. An orienting reflex perspective on anteriorisation of the P3 of the event-related potential. Exp. Brain Res. 173, 539–545. http://dx.doi.org/ 10.1007/s00221-006-0590-8. Barry, R.J., De Blasio, F.M., Rushby, J.A., Clarke, A.R., 2010. Brain dynamics in the auditory Go/NoGo task as a function of EEG frequency. Int. J. Psychophysiol. 78, 115–128. http://dx.doi.org/10.1016/j.ijpsycho.2010.06.357. Barry, R.J., De Blasio, F.M., Borchard, J.P., 2014a. Sequential processing in the equiprobable auditory Go/NoGo task: children vs. adults. Clin. Neurophysiol. 125, 1995–2006. http://dx.doi.org/10.1016/j.clinph.2014.02.018. Barry, R.J., De Blasio, F.M., Cave, A.E., 2014b. Caffeine effects on ERP components and performance in an equiprobable auditory Go/NoGo task. J. Caff. Res. 4, 83–92. http://dx.doi.org/10.1089/jcr.2014.0011. Barry, R.J., De Blasio, F.M., De Pascalis, V., Karamacoska, D., 2014c. Preferred EEG brain states at stimulus onset in a fixed interstimulus interval equiprobable auditory Go/ NoGo task: a definitive study. Int. J. Psychophysiol. 94, 42–58. http://dx.doi.org/10. 1016/j.ijpsycho.2014.07.005. Barry, R.J., De Blasio, F.M., Cave, A.E., 2016a. Sequential processing in young and older adults in the equiprobable auditory Go/NoGo task. Clin. Neurophysiol. 127, 2273–2285. http://dx.doi.org/10.1016/j.clinph.2016.02.010. Barry, R.J., De Blasio, F.M., Fogarty, J.S., Karamacoska, D., 2016b. ERP Go/NoGo condition effects are better detected with separate PCAs. Int. J. Psychophysiol. 50–64. http://dx.doi.org/10.1016/j.ijpsycho.2016.06.003. Beauducel, A., Debener, S., 2003. Misallocation of variance in event-related potentials: simulation studies on the effects of test power, topography, and baseline-to-peak versus principal component quantifications. J. Neurosci. Methods 124, 103–112. http://dx.doi.org/10.1016/S0165-0270(02)00381-3. Borchard, J.P., Barry, R.J., De Blasio, F.M., 2015. Sequential processing in an auditory equiprobable Go/NoGo task with variable interstimulus interval. Int. J. Psychophysiol. 97, 145–152. Cattell, R.B., 1966. The scree test for the number of factors. Multivar. Behav. Res. 1, 245–276. http://dx.doi.org/10.1207/s15327906mbr0102_10. De Blasio, F.M., Barry, R.J., 2013a. Prestimulus delta and theta determinants of ERP responses in the Go/NoGo task. Int. J. Psychophysiol. 87, 279–288. http://dx.doi. org/10.1016/j.ijpsycho.2012.09.016. De Blasio, F.M., Barry, R.J., 2013b. Prestimulus alpha and beta determinants of ERP responses in the Go/NoGo task. Int. J. Psychophysiol. 89, 9–17. http://dx.doi.org/10. 1016/j.ijpsycho.2013.04.018. Dien, J., 2010. The ERP PCA Toolkit: an open source program for advanced statistical analysis of event-related potential data. J. Neurosci. Methods 187, 138–145. http:// dx.doi.org/10.1016/j.jneumeth.2009.12.009. Folstein, J.R., van Petten, C., 2008. Influence of cognitive control and mismatch on the N2 component of the ERP: a review. Psychophysiology 45, 152–170. http://dx.doi.org/ 10.1111/j.1469-8986.2007.00602.x. Horn, J.L., 1965. A rationale and test for the number of factors in factor analysis. Psychometrika 30, 179–185. http://dx.doi.org/10.1007/BF02289447. Huster, R.J., Enriquez-Geppert, S., Lavallee, C.F., Falkenstein, M., Herrmann, C.S., 2013. Electroencephalography of response inhibition tasks: functional networks and cognitive contributions. Int. J. Psychophysiol. 87, 217–233. http://dx.doi.org/10.1016/ j.ijpsycho.2012.08.001. Kayser, J., Tenke, C.E., 2003. Optimizing PCA methodology for ERP component identification and measurement: theoretical rationale and empirical evaluation. Clin. Neurophysiol. 114, 2307–2325. http://dx.doi.org/10.1016/S1388-2457(03) 00241-4. Lorenzo-Seva, U., ten Berge, J.M.F., 2006. Tucker's congruence coefficient as a meaningful index of factor similarity. Methodology 2, 57–64. http://dx.doi.org/10.1027/ 1614-1881.2.2.57. Näätänen, R., Picton, T., 1987. The N1 wave of the human electric and magnetic response to sound: a review and an analysis of the component structure. Psychophysiology 24, 375–425. http://dx.doi.org/10.1111/j.1469-8986.1987.tb00311.x. Smith, J.L., Johnstone, S.J., Barry, R.J., 2006. Response priming in the Go/NoGo task: the N2 reflects neither inhibition nor conflict. Clin. Neurophysiol. 118, 343–355. http:// dx.doi.org/10.1016/j.clinph.2006.09.027. Tucker, L.R., 1951. A Method for Synthesis of Factor Analysis Studies (Personnel Research Section Report No. 984). Department of the Army, Washington, D.C. van Veen, V., Carter, C.S., 2006. Conflict and cognitive control in the brain. Curr. Dir. Psychol. Sci. 15, 237–240. http://dx.doi.org/10.1111/j.1467-8721.2006.00443.x. Wood, C.C., McCarthy, G., 1984. Principal component analysis of event-related potentials: simulation studies demonstrate misallocation of variance across components. Electroencephalogr. Clin. Neurophysiol. 59, 249–260.
References Barry, R.J., De Blasio, F.M., 2012. EEG-ERP phase dynamics of children in the auditory
79