Oops!.. I did it again: An ERP and behavioral study of double-errors
Introduction
The ability to detect errors and prevent their future occurrence is crucial for successful response monitoring, and recent studies have begun to integrate behavioral and neurobiological data in order to understand action monitoring. Researchers are increasingly focusing on activity of the anterior cingulate cortex (ACC) in studies of response monitoring. Most notably, when subjects commit errors in speeded response tasks, the response-locked ERP is characterized by a distinctive negative deflection at fronto-central recording sites that begins around the time of the mistake and peaks approximately 50 ms later (Falkenstein et al., 1991, Gehring et al., 1990, Gehring et al., 1993, Falkenstein et al., 2000, Nieuwenhuis et al., 2001). This error-related negativity or error negativity (ERN/Ne) has been observed both when subjects make an incorrect response choice, and when subjects commit errors of action (e.g., responding when a response should be withheld; Scheffers, Bernstein, & Donchin, 1996); additionally, the ERN appears independent of stimulus and response modality (Holroyd et al., 1998, Nieuwenhuis et al., 2001, Van’t Ent and Apkarian, 1999). In this way, the ERN appears to be an electrophysiological index of the activity of a generic error detection system (Falkenstein et al., 1991, Falkenstein et al., 2000). Using whole head recording systems, the ERN has been source-localized to the anterior cingulate cortex (ACC; Dehaene et al., 1994, Holroyd et al., 1998).
Following the ERN, the error positivity (Pe) also appears specific to error trials. The Pe is a positive deflection that occurs 200–400 ms after the commission of an error (Falkenstein et al., 2003, Nieuwenhuis et al., 2001) and has a more posterior midline scalp distribution than the ERN (Falkenstein et al., 2000). Van Veen and Carter (2002) noted that the Pe consisted of two subcomponents—a Pe closely linked to the ERN and a later Pe with a more posterior scalp distribution. There are several ideas as to what the Pe represents functionally, such as error salience or an orienting response to errors (Hajcak et al., 2003, Nieuwenhuis et al., 2001).
In addition to the error-specific ERN and Pe, a number of studies have focused on an ERN-like component that is evident in the response-locked ERP on correct trials. This correct response negativity (CRN; Ford, 1999) has a similar morphology and scalp topography as the ERN, and may index the response monitoring activity of the ACC on correct trials (Bartholow et al., 2005, Vidal et al., 2000). In fact, several studies have reported a reduction in the CRN on correct trials that precede errors. Ridderinkhof and colleagues first reported that error-preceding trials were associated with an increased positivity in the response-locked ERP in a similar time window as the ERN (Ridderinkhof, Nieuwenhuis, & Bashore, 2003). Subsequent studies replicated this pattern of findings (Allain et al., 2004, Hajcak et al., 2005), and have suggested that this error-preceding positivity reflects a reduction in the CRN. Thus, the available data suggests that some errors may result from transient deficiencies in the response monitoring system on preceding correct trials, and that these impending failures of action monitoring can be indexed by a reduction in the CRN.
Behaviorally, errors tend to be characterized by relatively fast reaction times (RTs) compared to correct responses. However, trials immediately following errors have relatively long RTs—an effect loosely referred to as post-error slowing. Because error trials tend to be especially fast, some increase in RT following errors is likely a simple regression back toward the mean. Reaction time on correct trials that follow errors, however, actually tends to be slower than the average RT on all correct trials and, therefore, must reflect more than simple regression. This true post-error slowing has been interpreted as a compensatory adjustment that minimizes the risk of subsequent errors; support for this possibility comes from studies that report increased performance accuracy following errors (Laming, 1979, Rabbitt, 1966, Rabbitt and Rodgers, 1977). Just as the ERN, CRN, and Pe have been attributed to the response monitoring activity of the ACC, multiple functional neuroimaging studies have found that post-error RT adjustments relate to activation in the dorsolateral prefrontal cortex (PFC; Garavan et al., 2002, Kerns et al., 2004).
These data raise the possibility that in addition to ‘monitoring’ errors, some errors might also result from a failure to implement compensatory adjustments once an error has occurred. Thus, two possibilities present themselves with regard to why errors might occur following errors. First, double-errors may result from transient deficits in action monitoring, much like data on errors that follow correct trials (Allain et al., 2004, Hajcak et al., 2005, Ridderinkhof et al., 2003). That is, the first error in a double-error sequence might not be properly detected. Based on the work described above on the ERN and Pe, this failure of error detection should be associated with a reduced ERN and/or Pe elicited by the first error. Alternatively, double-errors may not be due to failures of error detection, but rather, might result from a failure to implement control processes following the initial error. If this were the case, post-error compensatory processes, such as post-error RT slowing might be reduced or absent following the initial error. It is unclear why double-errors occur, and no study to date has examined behavioral and ERP indices surrounding this unique type of action monitoring failure.
The primary aim of the present study is to identify double-errors and to explore the possibility that errors might result either from action monitoring failures or from the failure to implement post-error RT adjustments. The ERN, Pe, and post-error slowing were measured in the context of both single and double-errors during a speeded response task. If double-errors result from deficient error monitoring activity, the first error should be characterized by a reduced or absent ERN and/or Pe. On the other hand, double-errors may occur because of a failure to implement post-error behavioral adjustments—in which case, post-error slowing should be reduced or absent following the initial error.
Section snippets
Subjects
ERP data from forty low-anxious (control) subjects who participated in previous studies were re-analyzed in the present investigation. These subjects were all students in an introductory psychology course. All subjects received course credit for their participation, and all subjects provided informed consent prior to the experiment.
Task
The Stroop task was administered on a Pentium I class computer, using Presentation software (Neurobehavioral Systems, Inc.) to control the presentation and timing of
Accuracy
Overall, subjects made an average of 84.2 mistakes (SD = 39.8); expressed as a percentage of valid trials, accuracy was 92.4% (SD = 3.6%). Consistent with previous studies, RTs on correct trials that followed errors (M = 422 ms; SD = 52) were reliably slower than overall correct trial RT (M = 406 ms; SD = 52; t(39) = 3.05, p < .01—participants evinced the typical pattern of post-error slowing. Subjects made 5.2 (SD = 5.2) double-errors on average, and accuracy following all errors was 93.2% (SD = 4.1%). The increase
Discussion
The present study evaluated ERPs and RT in single- and double-error sequences to evaluate possible mechanisms by which errors might be produced. It was hypothesized that errors might result from a failure in response monitoring or a failure to implement cognitive control and behavioral adjustment. In the present study, errors that were followed by subsequent errors (i.e., cEe sequences) were associated with ERN and Pe amplitudes that were virtually identical to single-error trials (i.e., cEc
Acknowledgment
This research was supported in part by National Institutes of Mental Health (NIMH) predoctoral fellowship MH069047 (G.H.).
References (34)
- et al.
The modulation of the Ne-like wave on correct responses foreshadows errors
Neuroscience Letters
(2004) - et al.
Effects of cross- modal divided attention on late ERP components: II. Error processing in choice reaction tasks
Electroencephalography and Clinical Neurophysiology
(1991) - et al.
ERP components on reaction errors and their functional significance: A tutorial
Biological Psychology
(2000) - et al.
Dissociable executive functions in the dynamic control of behavior: Inhibition, error detection, and correction
NeuroImage
(2002) - et al.
Error-preceding brain activity: Robustness, temporal dynamics, and boundary conditions
Biological Psychology
(2005) - et al.
Error-related scalp potentials elicited by hand and foot movements: Evidence for an output-independent error-processing system in humans
Neuroscience Letters
(1998) Choice reaction performance following an error
Acta Psychologica
(1979)- et al.
Errors are foreshadowed in brain potentials associated with action monitoring in cingulate cortex in humans
Neuroscience Letters
(2003) - et al.
Motoric response inhibition in finger movement and saccadic eye movement: A comparative study
Clinical Neurophysiology
(1999) - et al.
Is the ‘error negativity’ specific to errors?
Biological Psychology
(2000)
Neural signals for the detection of unintentional race bias
Psychological Science
Strategic control and medial frontal negativity: Beyond errors and response conflict
Psychophysiology
Digital filtering: Background and tutorial for psychophysiologists
Psychophysiology
Trial-by-trial coupling of concurrent electroencephalogram and functional magnetic resonance imaging identifies the dynamics of performance monitoring
Journal of Neuroscience
Localization of a neural system for error detection and compensation
Psychological Science
Short-term mobilization of processing resources is revealed in the event-related potential
Psychophysiology
Cited by (72)
Evidence for aversive withdrawal response to own errors
2017, Acta PsychologicaError detection and behavioural inhibition in young heavy drinkers
2017, Drug and Alcohol DependenceCitation Excerpt :This minimum represents the generally accepted lower limit of trials for reliable measurement of the ERN (Olvet and Hajcak, 2009; Pontifex et al., 2010; Steele et al., 2016), although the medians were higher at around 10 epochs. In studies of error processing generally, it is a common methodological difficulty to achieve high enough error rates to allow reliable statistical comparisons (Wessel, 2012), and it is also common to exclude participants who do not make sufficient errors (e.g., Bailey et al., 2015; Hajcak and Simons, 2008; Logan et al., 2015; O’Connell et al., 2009). Based on error and awareness rates in previous research (e.g., Hester et al., 2005), we expected around 23 Unaware error trials for controls, and an even greater number for heavy drinkers; however, participants in this study made fewer NoGo errors (24%) and were aware of more of these errors (82%) than expected, meaning that there were far fewer trials available for averaging for Unaware errors.
Post-error adjustments and ADHD symptoms in adults: The effect of laterality and state regulation
2016, Brain and CognitionCitation Excerpt :Of more importance is that negative feedback after errors loads on the emotional system, which in turn may enhance the right-hemisphere ability to compensate for errors (Kaplan & Zaidel, 2001; Luu, Collins, & Tucker, 2000). The present study contributes to the issue whether post-error slowing and increased post-error accuracy reflect the same neural mechanism (Fiehler et al., 2005; Hajcak & Simons, 2008; Hajcak et al., 2003; King et al., 2010). In our study post-error slowing occurred together with improved post-error accuracy in the RVF.