Distinguishing the influence of task difficulty on error-related ERPs using surface Laplacian transformation
Introduction
In order to behave adaptively to the requirements of the environment, it is necessary to monitor signals that point out the need for adjustment. Although various signals are able to indicate suboptimal performance that requires cognitive adjustments, the detection of an error is probably the most important signal. Electrophysiological investigations have demonstrated a negative brain potential at frontocentral electrode sites, peaking between 0 and 100 ms, after error commission (error-related negativity, ERN, Falkenstein, Hohnsbein, Hoormann, & Blanke, 1991; Gehring, Goss, & Coles, 1993). The origin of the ERN has been linked to the posterior medial frontal cortex (for an overview see Ullsperger, Danielmeier, & Jocham, 2014). Using Laplacian transformation, which allows spatial deblurring of EEG (Babiloni, Cincotti, Carducci, Rossini, & Babiloni, 2001), previous research has demonstrated that a similar negativity (albeit of a smaller amplitude which is often referred to as CRN), can be discerned on correct responses (Allain, Carbonnell, Falkenstein, Burle, & Vidal, 2004; Vidal, Hasbroucq, Grapperon, & Bonnet, 2000). This challenges the idea that the ERN is specific to errors but rather reflects a more general response evaluation (Bonini et al., 2014). In line with this idea, CRN amplitude increases with higher uncertainty (Pailing & Segalowitz, 2004). Although it has been argued that ERN and CRN might reflect different mechanisms (Coles, Scheffers, & Holroyd, 2001; Endrass, Klawohn, Gruetzmann, Ischebeck, & Kathmann, 2012), there is now strong evidence that they reflect the same modulated underlying processes (Bonini et al., 2014; Roger, Bénar, Vidal, Hasbroucq, & Burle, 2010).
Following the ERN, a slow positive wave with a maximum amplitude between 200 and 400 ms and a more diffuse scalp distribution is observed (error positivity, Pe, Falkenstein et al., 1991), which has been attributed to error recognition or error awareness (Endrass, Reuter, & Kathmann, 2007; Nieuwenhuis, Ridderinkhof, Blom, Band, & Kok, 2001; O’Connell et al., 2007; Shalgi, Barkan, & Deouell, 2009; Wessel, Danielmeier, & Ullsperger, 2011). Interestingly, the Pe shares many characteristics with the P300, a positive stimulus-locked slow wave appearing between 200 and 400 ms after stimulus onset. The P300 has generally been associated with the processing of unexpected and motivationally significant events (for a review, see Nieuwenhuis, Aston-Jones, & Cohen, 2005) and has been divided into two subcomponents. The P3a occurs first with a frontocentral scalp distribution (Polich & Comerchero, 2003) and is mainly sensitive to the novelty of events. In contrast, the P3b is a later-occurring component with a parietal scalp distribution and is sensitive to the amount of attentional resources allocated to a stimulus (Polich and Comerchero, 2003, Polich, 2007). Similarly, there is support that the Pe consists of two subcomponents, an early and a late one (Arbel & Donchin, 2009; Endrass, Klawohn, Preuss, & Kathmann, 2012; Endrass et al., 2007; Van Veen & Carter, 2002), where only the late Pe is seemingly related to error awareness (Endrass, Klawohn, Preuss et al., 2012). The early frontocentral Pe has been proposed to be generated by the same generators as the ERN (Debener et al., 2005, Van Veen and Carter, 2002) while the late posterior Pe is attributed to the parietal cortex and rostral ACC (Herrmann, Römmler, Ehlis, Heidrich, & Fallgatter, 2004; Van Veen & Carter, 2002; for an overview, see Ullsperger et al., 2014).
Consistent with the idea that the ERN reflects an outcome evaluation, a relationship between the ERN and task difficulty (indexed by increased perceptual difficulty) has been reported, which show a decreased ERN magnitude when task difficulty increases (Endrass, Klawohn, Gruetzmann et al., 2012; Hoffmann and Falkenstein, 2010, Pailing and Segalowitz, 2004). Comparisons across different cognitive tasks, such as the Stroop, flanker and go/no-go task also showed a smaller ERN-amplitude for the Stroop-task, which was associated with the highest error rate (Riesel, Weinberg, Endrass, Meyer, & Hajcak, 2013). However, when task difficulty in conflict tasks is manipulated by increasing the number of stimuli, as well as their associated responses, no influence on the ERN or CRN has been reported (Compton, Bissey, & Worby-Selim, 2014; Pailing & Segalowitz, 2004).
A small number of studies that have investigated the relationship between the early and late Pe and task difficulty suggest that the late Pe is sensitive to error saliency. Arbel and Donchin (2009) showed that only the posterior positive deflection was sensitive to the accuracy instruction, and was larger when accuracy was stressed compared to a neutral condition. Similarly, Endrass, Klawohn, Gruetzmann et al. (2012), using spatio-temporal Principal Component Analysis, showed that a centroparietal component in the time range of the Pe varied significantly with perceptual difficulty while a frontocentral component within the same time range did not. Both studies also observed that the ERN was affected by task instruction/difficulty, with more pronounced ERN for the easier condition. Such findings are in contrast to the hypothesis that the early Pe is a continuation of the ERN (Wessel, 2012).
In the present study, we manipulated task difficulty by means of the complexity of the mapping rule and investigated its effect on the ERN/CRN, early Pe and late Pe. As we manipulated task difficulty by increasing stimulus-response mappings, we did not expect to find a modulation of the ERN or the CRN (Compton et al., 2014, Pailing and Segalowitz, 2004). Based on previous findings and the idea that the early Pe is a continuation of the ERN, no modulation of this component was expected. Our difficulty modulation, however, should result in a difference in error saliency, which decreases with higher error rates and more difficult task requirements. We therefore expected to find a larger late Pe component in the easy condition. Based on the orienting account (Notebaert et al., 2009) this difference in error saliency should also be reflected in the amount of post-error slowing since infrequent or salient events trigger a larger orienting response that interferes with subsequent processing.
An overview of the literature revealed that while some studies reported an early and late Pe (Endrass et al., 2007, Van Veen and Carter, 2002), this differentiation was not always observed (Hajcak, McDonald, & Simons, 2003; Nieuwenhuis et al., 2001, Shalgi et al., 2009, Wessel et al., 2011), largely because the two components overlap in time and space, further amplified by volume conduction (Burle et al., 2015). Statistical procedures, such as independent component analysis (ICA—Debener et al., 2005) or principal component analysis (PCA—Arbel & Donchin, 2009; Endrass, Klawohn, Gruetzmann et al., 2012; Endrass, Klawohn, Preuss et al., 2012), have successfully differentiated both components. Such statistical approaches, however, are not based on any physical or physiological assumptions and can lead to nonphysiologically-plausible outcomes (e.g., Delorme, Palmer, Onton, Oostenveld, & Makeig, 2012). As an alternative decomposition approach, the estimation of Current Source Density through the computation of the Surface Laplacian (SL) of the scalp potential, provides data that are interpretable both physically and physiologically. From a physical point of view, the SL is proportional to the radial component of the current density flowing through the skull. It is able to reduce the current diffusion induced by the skull (Carvalhaes & de Barros, 2015), leading to a significant improvement of the spatial resolution of EEG signals (Tenke & Kayser, 2012). From a physiological point of view, the SL provides a fair approximation of the corticogram (Gevins, 1989), defined as the activity one would record if the electrodes were positioned over the dura. Importantly, SL computation requires no assumptions, except that the scalp is locally isotropic. Furthermore, it can be easily computed for each participants and does not require any selection of components (for comparisons between ICA and SL, see e.g., Foffani et al., 2004, Roger et al., 2010, Tenke and Kayser, 2012). Since SL proved to be a very powerful tool in revealing the presence of the CRN, we anticipate that it would also help dissociate the early and late Pe, which would allow us to assess the impact of task difficulty on all response-evaluation related activities (CRN, ERN, early and late Pe) using the same methodology.
Section snippets
Participants
Sixteen participants participated in the experiment. Each participant gave written informed consent. The study was approved by the ethical committee of the Faculty of Psychology and Educational Sciences at Ghent University. All participants had normal or corrected to normal vision and were neurologically and psychiatrically healthy. Participants were paid 15€ per hour.
Material and procedure
A classic flanker task was modified in order to create two conditions that varied in difficulty. In the easy condition there
Behavioral data
Both error rates and correct reaction times were analyzed with repeated measures ANOVAs. The within-subject factors were condition and previous accuracy. As a measure of effect size, partial-eta2 (ηp2) is reported.
For accuracy, there was a significant difference between conditions, F(1,15) = 26.34, p < 0.001, ηp2 = 0.64, showing that participants made more errors in the hard (29%) than in the easy condition (20%). There was no significant effect of previous accuracy, i.e., no post-error accuracy
Discussion
In this ERP-study, task difficulty was manipulated in order to investigate the neural correlates of error processing using Laplacian transformation to dissociate the various brain activities related to performance monitoring (ERN, CRN, early and late Pe). The difficulty manipulation was established by creating an easy condition with a 2:1 stimulus response mapping and a difficult condition with a 4:1 stimulus response mapping. As expected, the hard condition revealed more errors and longer
Acknowledgments
The work of Liesbet Van der Borght and Wim Notebaert is supported by Research Foundation—Flanders (grant 3G076911). The work of Femke Houtman is supported by a doctoral grant from Ghent University (BOF Grant B/09928/02). The work of Boris Burle is supported by a European Research Council under the European Community’s Seventh Framework Program (FP/2007-2013 Grant Agreement no. 241077).
References (46)
- et al.
The modulation of the Ne-like wave on correct responses foreshadows errors
Neuroscience Letters
(2004) - et al.
Spatial enhancement of EEG data by surface Laplacian estimation: the use of magnetic resonance imaging-based head models
Clinical Neurophysiology
(2001) - et al.
The surface Laplacian technique in EEG: theory and methods
International Journal of Psychophysiology
(2015) - et al.
Why is there an ERN/Ne on correct trials? Response representations, stimulus-related components, and the theory of error-processing
Biological Psychology
(2001) - et al.
Effects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks
Electroencephalography and Clinical Neurophysiology
(1991) - et al.
Oops!.. I did it again: an ERP and behavioral study of double-errors
Brain Cognition
(2008) - et al.
Source localization (LORETA) of the error-related-negativity (ERN/Ne) and positivity (Pe)
Cognitive Brain Research
(2004) - et al.
Principal components analysis of Laplacian waveforms as a generic method for identifying ERP generator patterns: I. evaluation with auditory oddball tasks
Clinical Neurophysiology
(2006) - et al.
Post-error slowing: an orienting account
Cognition
(2009) - et al.
The effects of uncertainty in error monitoring on associated ERPs
Brain Cognition
(2004)
Spherical splines for scalp potential and current density mapping
Electroencephalography and Clinical Neurophysiology
Updating P300: an integrative theory of P3a and P3b
Clinical Neurophysiology
The ERN is the ERN is the ERN? Convergent validity of error-related brain activity across different tasks
Biological Psychology
Rostral cingulate zone and correct response monitoring: ICA and source localization evidences for the unicity of correct- and error-negativities
Neuroimage
Generator localization by current source density (CSD): implications of volume conduction and field closure at intracranial and scalp resolutions
Clinical Neurophysiology
Is the error negativity specific to errors?
Biological Psychology
Error negativity on correct trials: a reexamination of available data
Biological Psychology
Parsing the componential structure of post-error ERPs: a principal component analysis of ERPs following errors
Psychophysiology
Action monitoring and medial frontal cortex: leading role of supplementary motor area
Science (New York, N.Y.)
Spatial and temporal resolutions of EEG: is it really black and white? A scalp current density view
International Journal of Psychophysiology
Task motivation influences alpha suppression following errors
Psychophysiology
Trial-by-trial coupling of concurrent electroencephalogram and functional magnetic resonance imaging identifies the dynamics of performance monitoring
The Journal of Neuroscience
Independent EEG sources are dipolar
PLoS One
Cited by (11)
Neurocognitive efficiency in breast cancer survivorship: A performance monitoring ERP study
2021, International Journal of PsychophysiologyError-monitoring across social and affective processing contexts
2020, International Journal of PsychophysiologyEvaluation of performance monitoring ERPs through difficulty manipulation in a response-feedback paradigm
2019, Brain ResearchCitation Excerpt :More generally, this technique improves spatial, as well as temporal, EEG resolution (Burle et al., 2015). Moreover, surface Laplacian is physiologically relevant (Roger et al., 2010; Van der Borght et al., 2016) as opposed to other component isolating analyzes, like independent or principal-component analysis (ICA or PCA), which are only statistical and can lead to results that are not physiologically plausible (Delorme et al., 2012). The use of this technique in the present study may improve the ERP analysis, since it limits the impact of components on each other, particularly for components such as the Pe or P300, which are very broad and diffusing ERPs that can therefore influence recordings quite far from their sources.