Is the ‘error negativity’ specific to errors?

https://doi.org/10.1016/S0301-0511(99)00032-0Get rights and content

Abstract

When subjects make an erroneous response in a choice reaction time task, an error negativity, or error-related negativity (NE/ERN), peaking at about 100 ms after EMG onset, has been described. This wave is often considered to be absent on correct response trials. We report a small NE/ERN wave on correct response trials during a choice reaction time task in which surface Laplacians were estimated by the source derivation method. This wave is well focused at FCz, and its time course is the same for correct responses trials, incorrect sub-threshold EMG activation trials, and error trials. Current source density maps, also indicate a focus at FCz. A second experiment showed the existence of a NE at FCz on correct trials during a simple RT task. Rather than an error detection process per se, we propose that the NE/ERN reflects either a comparison process leading secondarily to error detection, or an emotional reaction.

Introduction

Information processing theories generally agree that perceptual-motor performance is mediated by elementary mental operations (for instance perception, decision, response elaboration and execution). Irrespective of whether or not there is temporal overlap between these operations (see Miller (1988) for a review), nowhere is it explicitly assumed that they are under the control of a supervisory process. Nor do these models specify any mechanism whereby errors may be detected, and/or corrected. Yet, when subjects make errors in choice reaction time tasks, there have recently been reports of a negative wave peaking at about 100 ms after EMG onset; this has been called error negativity (NE, Falkenstein et al., 1990, Falkenstein et al., 1991, Falkenstein et al., 1995, Falkenstein et al., 1996, Kaiser et al., 1997), or error-related negativity (ERN, Gehring et al., 1993, Dehaene et al., 1994, Bernstein et al., 1995, Gehring et al., 1995, Scheffers et al., 1996, Holroyd et al., 1998). With traditional monopolar recordings this wave has a fronto-central maximum, and is generally considered not to be present in the case of correct responses. (Coles et al. (1992) report that the NE was “absent on correct trials” (p. 787); Gehring et al. (1993) wrote that “…an error-related negativity (ERN) appears selectively on error trials” (p. 385); Dehaene et al. (1994) consider that the NE reflects the activation of a system which “comes into play only when an error is detected in time for a correction to be attempted” (p. 304); speaking of the NE Falkenstein et al. (1995) consider that “…in error trials at least one additional component is present compared to correct trials” (p. 287) and Kaiser et al. (1997) indicate that “This wave is absent for correct responses” (p. 216), although Sheffers et al. left open the possibility that a NE might occur in correct trials indicating that the NE is “virtually absent” (p. 42) on the average waveforms for correct trials, without referring to specific data.)

This led several authors to view it as an index of an error detection and/or error compensation mechanism, which would support earlier proposals by, for example, Rabbitt (1966) and Logan (1985), that a supervisory process can operate in parallel with the information processing stages that are currently assumed by most cognitive psychologists to underlie performance. In a recent study Scheffers et al. (1996) concluded: “We found a weak link between the ERN and error compensation… Taken together, our results are most readily explained by assuming that the ERN is associated with an error detection process.” (p. 52). This view is also advocated by Falkenstein et al. (Falkenstein et al., 1995, Falkenstein et al., 1996).

However, recently reported results by Vidal et al. (1995) (Fig. 7), although they did not discuss them in their article, might not fit this interpretation. In that study, subjects performed a timing task in which they were asked to produce a time interval between two brief button presses (with the left hand) as accurately as possible. When the produced interval was of an acceptable duration (i.e. ±22.5% of the target duration), a negative wave, peaking about 40–50 ms following response onset was observed at FCz, but not at C4 (contralateral to the left responding hand), nor over the right parietal area. Such specific localisation was possible in that study because of the use of the source derivation method (Hjörth, 1975) which enables one to approximate the surface Laplacian of recorded activities. Because, in that study, the time separating EMG onset from button presses was about 50 ms, the peak of this wave occurred about 90–100 ms following EMG onset. The latency, polarity, and topography of this wave are, therefore, very comparable to the NE wave reported by others, and discussed above. Moreover, the fact that it was obtained at FCz is congruent with its hypothesised origin in either the anterior cingulate cortex or in the supplementary motor areas (SMA) (Gehring et al., 1993, Dehaene et al., 1994, Holroyd et al., 1998).

This raises the question of why such a wave was observed on correct trials. At least two possibilities can be proposed. Either this NE-like wave is not the NE reported by others or this NE-like wave does correspond to an NE, but had not been noticed before because of certain weaknesses of conventional, monopolar recordings (Nuñez, 1981) as compared to methods estimating the surface Laplacian.

In this article we shall attempt to distinguish between these possibilities. However, we dropped the timing task because the nature of the error might be ambiguous: the difference between an ‘error’ and a ‘correct’ response is not discontinuous in nature and relies on the internal representation of time of each subject. Subjects might judge some correct responses as incorrect, and vice-versa, which could explain the existence of a NE on trials defined as ‘corrects’ by the experimenters. Therefore, in the experiments reported in this study, we have collected the EEG of subjects while they were performing a choice-reaction time task. In such situations, the nature of the error is discontinuous and there is no ambiguity on the nature of the given response.

We applied the source derivation method (Hjörth, 1975) to the EEG data in order to calculate an approximation of the surface Laplacian (more details will be given in Section 2.1). In order to examine the distribution of the recorded activities we will calculate current source densities (CSDs) maps by the method described by Perrin et al. (1987). Since surface Laplacians and CSDs are equivalent, a resistivity constant apart, we shall refer to ‘Laplacians’ for both data obtained with the source derivation method and for the CSDs mapping ones. Laplacian data are relatively free from activities originating from remote sources. Therefore, they yield fine, uncontaminated topographical distributions of the different components (Nuñez, 1981, Perrin et al., 1987, Nuñez et al., 1994), and are particularly well suited for analysing the time course of brain activities. As such they provide more reliable measures than conventional monopolar recordings of the latencies of peaks and troughs (Law et al., 1993b).

Dehaene et al. (1994) used a forward-search dipole localisation method, namely the brain electric source analysis (BESA), to identify the origin of the NE generator. Their method appeared to converge onto a very deep source for the NE that they identified as the anterior part of the cingulate cortex. However, considering on the one hand the strong assumptions made by BESA concerning head sphericity, conductance values of the different media, and the dipolar nature of the source, and on the other hand, the tight focus that they obtained with their recordings these authors expressed some doubts about their localisation conclusions, and considered that a more superficial source, in the SMA or the anterior cingulate cortex, was a more realistic option. Therefore, a secondary rationale for our study, was to examine whether or not the NE would be clearly observed after Laplacians transformation. Since surface Laplacian are relatively insensitive to deep sources (Perrin et al., 1987, Pernier et al., 1988), any NE activity obtained by source derivation would provide evidence favouring the interpretation presented by Dehaene et al. (1994) that the source of the NE lies rather superficially.

We used a variant of a go/no go task with Stroop-like stimuli with a high stimulus presentation rate in order to make the task more complex, i.e. more difficult to automate. In addition to key presses made with the left and right thumbs, we also recorded electromyographic (EMG) activity of the flexores pollicis brevis.

Section snippets

Subjects

There were 12 subjects (10 men, and two women, ranging in age from 23 to 48 years). Written informed consent was given before the experiment. They were all right handed as assessed by the Edinburgh inventory (Oldfield, 1971), they had normal or corrected to normal vision, and the absence of dyschromatopsia was verified with Ishihara’s test for colour blindness (one subject was discarded for this reason before the experiment). They were comfortably seated in an armchair placed in a Faraday cage,

Experiment II

The main purpose of this experiment was to verify whether the NE would be observed in a simple RT task where, presumably, errors would be extremely unlikely, and to attempt to replicate the results of experiment I in a different choice RT task. Our analysis focused on FCz. For other purposes that we shall not discuss, we also examined brain activities over the sensorimotor motor cortices.

In order to check whether the presence of an NE on correct trials could be generalised, or whether it

Localisation.

Gehring et al. (1993) speculated that the NE might originate either in the anterior cingulate cortex, and/or the SMA. Using BESA, Holroyd et al. (1998) found sources close to those obtained by Dehaene et al. (1994). Dehaene et al., on the basis of BESA, tentatively concluded that NE probably originated from very deep sources in the lowest part of the anterior cingulate cortex, probably deeper than the genu of the corpus callosum. However, they criticised their own BESA solution and associated

Acknowledgements

We are very grateful to Raphaël Apfeldorfer, Monique Chiambretto, Martial Dessemond, Sébastien Drot, Frédéric Chazelle and Dominique Reybaud for their technical contribution. We are also indebted to Sylvan Kornblum for his help with the English language and his helpful comments. Parts of this work were presented in poster form at the 3rd European Conference of Psychophysiology, Konstanz, July, 1997. This research was supported by a DRET research grant (No. 21/96).

References (35)

  • P.L. Nuñez et al.

    A theoretical and experimental study of high resolution EEG based on surface Laplacians and cortical imaging

    Electroencephalogr. Clin. Neurophysiol.

    (1994)
  • R.C. Oldfield

    The assessment and analysis of handedness. The Edinburgh inventory

    Neuropsychologia

    (1971)
  • J. Pernier et al.

    Scalp current density fields: concepts and properties

    Electroencephalogr. Clin. Neurophysiol.

    (1988)
  • H. Steinmetz et al.

    Craniocerebral topography within the international 10–20 system

    Electroencephalogr. Clin. Neurophysiol.

    (1989)
  • P.S. Bernstein et al.

    “Where did I go wrong?” A psychophysiological analysis of error detection

    J. Exp. Psychol.

    (1995)
  • M.G.H. Coles et al.

    Response activation and verification: a psychophysiological analysis

  • S. Dehaene et al.

    Localization of a neural system for error detection and compensation

    Psychol. Sci.

    (1994)
  • Cited by (0)

    View full text