ª Federation of European Neuroscience Societies

European Journal of Neuroscience, Vol. 20, pp. 1960–1968, 2004

Why do we produce errors of commission? An ERP study of stimulus deviance detection and error monitoring in a choice go⁄no-go task Martin Elton,1 Marcus Spaan1 and K. Richard Ridderinkhof1,2 1

Department of Psychology, University of Amsterdam, Amsterdam, The Netherlands Department of Psychology, Leiden University, Leiden, The Netherlands

2

Keywords: error negativity, error positivity, error-related negativity, mismatch negativity

Abstract The present study investigated event-related brain potential (ERP) reflections of stimulus deviance detection and error monitoring recorded during a go ⁄ no-go auditory oddball task. The mismatch negativity and error negativity were analysed as indices of these processes, respectively. We examined whether errors of commission occurred because of failures to detect stimulus deviance. On error trials the mismatch negativity amplitude did not differ between big and small deviant stimuli and was clearly reduced as compared with mismatch negativity amplitude on correct trials. Following an error response an error negativity was elicited, the amplitude of which was unaffected by stimulus deviancy. This pattern of ERP results was interpreted as supporting the defective deviance detection hypothesis.

Introduction It has been argued in the literature that the inclusion of psychophysiological measures may augment normal measures of response speed and accuracy of human information processing, and the term ‘chronopsychophysiology’ has been introduced to describe this approach (van der Molen et al., 1991). Event-related brain potentials (ERPs) also contribute to our understanding of human information processing in situations in which subjects are not required to make a response (e.g. amplitude modulations of ERPs associated with attended compared with unattended stimuli, or associated with deviant compared with standard stimuli). This paper reports a study of ERP components linked to specific aspects of information processing associated with errors of commission. Specifically, a choice go ⁄ no-go task was used in which subjects issued left or right discriminatory responses to tones presented to their left or right ear (respectively), but were to withhold their response on those infrequent occasions where stimulus pitch deviated from standard tones. Deviant tones were either easy or difficult to discriminate from the standard tones. This type of task typically elicits a considerable number of errors of commission on no-go trials. The number of errors of commission increases with the perceptual similarity between the deviant and standard tones. This paradigm was used to examine whether errors of commission are made because the perceptual system fails to efficiently detect stimulus deviance. To this end, attention was focused on ERP indices of stimulus deviance detection and error monitoring.

Correspondence: Dr M. Elton, as above. E-mail: [email protected] Received 12 December 2003, revised 14 July 2004, accepted 19 July 2004

doi:10.1111/j.1460-9568.2004.03631.x

Stimulus deviance detection When subjects are instructed to read a book and ignore a series of auditory stimuli of standard pitch intermixed infrequently with tones of a slightly deviant pitch, they typically report not to have noticed the presence of deviant tones. However, their perceptual system has detected the stimulus deviance nonetheless, as shown by the modulation of early exogenous ERP components. This modulation, labelled mismatch negativity (MMN), can be obtained by a change, either an increase or a decrease, in stimulus frequency, intensity, duration or change in location of presentation of stimuli in the auditory modality (see Na¨a¨ta¨nen, 1990, 1992 for review). The MMN is obtained by subtracting the ERP to the standard stimulus from that obtained to the deviant stimulus. The MMN is typically largest at Fz of the midline leads (Ritter et al., 1995), a result that has been replicated repeatedly in previous research in our laboratory (see, for example Winter et al., 1995). Although MMN amplitude can saturate, generally MMN amplitude increases as the degree of frequency deviation becomes greater. The MMN is typically recorded in passive conditions in which the subject is principally engaged in another task (such as reading a book) and is not necessarily consciously aware of the stimulus deviance. Thus, MMN is considered to reflect preconscious or preattentive processes of stimulus deviance detection. However, the MMN to deviant pitch can be obtained in conditions in which subjects are actively attending to the stimuli and, as such, is independent of attentional focus (Na¨a¨ta¨nen, 1992, p. 170; Alho et al., 1998). Modelling of both the electroencephalographic MMN (Giard et al., 1995) and the magnetoencephalographic MMN (Alho et al., 1998) has revealed that it is generated by a source in the superior temporal cortex. The sensitivity of the MMN to the degree of stimulus deviation suggests that the efficiency of early deviance detection processes can

ERPs to no-go stimuli 1961 be manipulated experimentally. Thus, when button-press responses are required to standard but not deviant tones, subjects can be expected to show more errors of commission when tones deviate in pitch only slightly from standard tones as compared with when tones deviate more markedly. In such go ⁄ no-go paradigms, a subject may successfully refrain from responding to a deviant stimulus. In these trials the preattentive processes of stimulus deviance detection either prevent the subthreshold activation of the go response from flowing into full response activation or prevent such subthreshold activation altogether. When a subject does produce a commission error, however, such failures may be due to a failure in stimulus discrimination or response selection (Rabbitt & Vyas, 1981). On the basis of reaction time (RT) data from a tone comparison experiment, comprising both easy and difficult conditions, Rabbitt & Vyas (1981) observed that in their easy discrimination condition subjects were more likely to be able to correct for errors than in their difficult condition. The authors attributed different sources to corrected and uncorrected errors in the two conditions (see also Fiehler et al., 2004). Uncorrected errors, which occurred more often in the difficult condition, were slow and were attributed to a failure in perceptual discrimination. The present authors are unaware of any previous experimental report that has examined the MMN in association with commission errors. In the present study the MMN was examined in separate trial blocks in passive as well as active conditions in which subjects were instructed to withhold their response to rare, deviant stimuli that differed slightly or markedly from the standard, go stimulus. We hypothesized that if error commission trials reflected defective deviance detection the MMN would be reduced in amplitude in comparison to correct no-go trials on which the response was successfully withheld. That is, if the deviant is not recognized as such then MMN amplitude will be reduced. We further expected that, on commission error trials, defective deviance detection would not permit the usual relationship in MMN amplitude between big and small deviants to be obtained; that is, no difference in MMN amplitude was expected between big and small deviants on error commission trials.

Error monitoring In addition to MMN, reflecting stimulus deviation detection, a second set of ERP components that may inform us about what goes wrong when people make errors of commission is related directly to error processing. Psychophysiological researchers have identified two ERP components that are strongly associated with making errors. The first was called error negativity (Ne, Falkenstein et al., 1991) or errorrelated negativity (Gehring et al., 1993): the term Ne will be used in the remainder of this article for reasons of parsimony. Ne appears robust in that it has been obtained in a variety of experimental paradigms including conflict tasks (e.g. Bernstein et al., 1995; Ridderinkhof et al., 2002), reinforcement learning tasks (Holroyd & Coles, 2002; Nieuwenhuis et al., 2002) and go ⁄ no-go tasks (Falkenstein et al., 1995; Scheffers et al., 1996). Also Ne appears independent of stimulus modality (Leuthold & Sommer, 1999; Falkenstein et al., 2000) and response modality (Holroyd et al., 1998; Nieuwenhuis et al., 2001; Van ‘t Ent & Apkarian, 1999). Ne is believed to originate from anterior cingulate cortex (ACC; for reviews of ERP and neuroimage evidence see, e.g. Ullsperger & von Cramon, 2001; van Veen et al., 2001; for a review of ACC function see Devinsky et al., 1995). Depending on the theoretical perspective, Ne is seen to reflect: (i) the activity of a comparison process that detects a mismatch between actual and appropriate responses

(Falkenstein et al., 2000; Coles et al., 2001); (ii) the impact on ACC of dopaminergic-based error-related signals upon arriving from the basal ganglia (Holroyd & Coles, 2002; Nieuwenhuis et al., 2002); or (iii) the detection of conflict between simultaneously active competing response channels (Carter et al., 1998; Botvinick et al., 2001). Regardless of the specific theoretical perspective, competition between incompatible response channels may arise from partial information on stimulus features being extracted from the eliciting stimulus (Bernstein et al., 1995; Coles et al., 1995). Thus, in go ⁄ no-go tasks, one aspect of the stimulus may activate the go response, while another aspect may activate the no-go response (cf. Nieuwenhuis et al., 2003). Ne upon errors of commission is thus elicited by continued processing of the no-go aspect of the stimulus, resulting in a post-response mismatch or conflict with the response elicited by the go aspect (cf. Holroyd & Yeung, 2003). A second error-related ERP component, error positivity (Pe; Falkenstein et al., 1991) has not been subjected to the same experimental scrutiny as Ne, due possibly to the controversy as to whether Pe represented a genuine component or was rather a delayed P300 elicited by the stimulus. Falkenstein et al. (2000) suggested that the functional nature of Pe may be related to conscious error recognition, adjustment of response strategy or emotional evaluation of error and, as such, was unrelated to the P300. Leuthold & Sommer (1999) and Vidal et al. (2000) obtained a Pe on error trials that was dissociable from the P300 relating to stimulus evaluation. In a study using an anti-saccade task, Nieuwenhuis et al. (2001) observed a dissociation between Pe and Ne: the amplitude of Pe but not Ne was reduced or eliminated on trials in which subjects did not perceive their error. Whereas subjects are normally unaware of making a mistake, slips may be amenable to awareness (Reason, 1990). Pe thus appears to be an independent component on error trials and depends on subjects consciously perceiving that an error (slip) had been committed. Ne and Pe may help us gain insight into the cause of commission errors in the following way. Under the deviance detection hypothesis, the failure to adequately detect stimulus deviance results in a failure to prevent the activation of the go response, regardless of the degree of stimulus deviance. It follows that the amplitude of both Ne and Pe should be unaffected by the degree of stimulus deviance. If the failure to preattentively detect stimulus deviance was to result in the absence of error awareness, then Pe should be eliminated during errors of commission. Some preliminary evidence seems to point in the direction of the deviance detection hypothesis. Leuthold & Sommer (1999) required spatially compatible or incompatible responses to lateral stimuli (presented to the left or right of fixation) and manipulated eccentricity of stimulus presentation, on the assumption that stimuli presented close to the midline would present more difficult perceptual judgements than stimuli at more lateral presentations. No significant effect of eccentricity of stimulus presentation was obtained either for Ne latency (Leuthold & Sommer, 1999) or for Ne amplitude (Leuthold, personal communication, 2002).

The present study A choice go ⁄ no-go RT task was administered in which rare deviant stimuli were either easy or difficult to discriminate (in separate trial blocks) from the standard tones. Two separate conditions of the task are reported here: a passive condition (in which subjects read a book and ignored the tones), serving to verify proper replication of the MMN and of the effects of degree of deviance on MMN amplitude;

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

1962 M. Elton et al. and an active version (in which subjects responded to the location of the tone but refrained from responding when the tone deviated in pitch). Based on our deviance detection hypothesis, we predicted that: (i) MMN to errors of commission should be smaller than MMN to correctly withheld responses for a given level of stimulus deviance; (ii) MMN to errors of commission, possibly reflecting change in the location attribute of the stimulus (see Materials and methods), should be of equally small amplitude regardless of the degree of stimulus deviance; (iii) the amplitude of both Ne and Pe should be unaffected by the degree of stimulus deviance; and (iv) Pe is absent if the failure to detect deviance results in lack of error awareness.

Materials and methods Subjects First year undergraduates at the University of Amsterdam were recruited to participate in the experiment. Students were rewarded with points for a compulsory part of their course. Before the experiment proper, subjects supplied answers to a health questionnaire aimed at revealing neurological and psychiatric disorders. One subject was excluded at this stage because of a neurological disorder, and the data of one further subject were discarded from the present study because of too much noise in the electroencephalogram (EEG). The final group comprised nine female and four male students (aged 18–28 years). All subjects gave their informed consent to participate in the experiment, which was also passed by the Ethics Committee of the Department of Psychology.

Procedure Having completed the health questionnaire, subjects were requested to sit in a dentist chair in a light-dimmed and quiet room for the recording of behavioural and psychophysiological data. During the experiment subjects fixated on a cross on a monitor, situated directly in front of them, so as to minimize eye movements. The experimental task implemented in this experiment consisted of a choice go ⁄ no-go RT task in which subjects were instructed to respond to standard stimuli and to withhold their response to deviant stimuli. Standard tones were 1100 Hz, small deviants were 1150 Hz and big deviants were 1350 Hz. Big and small deviants (each 15%) were presented in separate blocks with standard tones (85%). Each tone had a duration of 40 ms and was presented at an intensity of 75 dB sound pressure level. The inter-stimulus interval was 1 s. Stimuli were presented in blocks of 210 trials whereby the first 10 trials were excluded from the data analysis. The sequence of tones was varied in four stimulus lists for each deviant so as to eliminate any transfer of learning effects. The sequence of standards and deviants was kept identical between equivalent blocks of big and small deviants. Monaural tones were presented quasi-randomly to the two ears such that a deviant was always presented to a different ear, i.e. successive deviants were presented to different ears. This manipulation was designed to obtain a maximal MMN in that on such trials both frequency and location of the stimulus changed. At least three standard tones always preceded a deviant. Deviant tones were presented equally often to each ear. Data were recorded in two afternoon sessions lasting 4 h. Each session contained two passive blocks and 15 active blocks. Both sessions were begun with the passive condition. Big and small deviant blocks were alternated in both conditions. In the passive condition subjects received a total of four blocks of trials. In these blocks

subjects were instructed to read a book and to ignore the stimuli. In the 30 blocks, which comprised the go ⁄ no-go active conditions, subjects were instructed to respond as quickly and accurately as possible to the standard tone and to withhold their response to the deviants. Responses were given by depressing switches located in the arms of the chair. The side of the go response was indicated by the ear in which the standard tone had been presented.

Psychophysiological data acquisition and analysis All signals [EEG and electro-oculogram (EOG)] were recorded with a Nihon-Kohden 5210 polygraph using tin electrodes. The EEG was recorded from leads at Fz, Cz, Pz, C3 and C4, and referred to the left mastoid. Electrode impedance was measured using a 16 Hz square wave and maintained below 5 kW for all subjects. Vertical EOG was recorded from above and below the right eye, and the horizontal EOG obtained from electrodes located at the external canthus of each eye. The ground electrode was located in the centre of the subject’s forehead. The EEG signal was amplified by 100 000. The EEG and EOG were recorded with a low-pass filter of 35 Hz and a 5-s time constant. All leads were sampled at 100 Hz in a window from 100 ms prestimulus to 900 ms post-stimulus and saved on the hard disk of a Pentium computer for off-line analysis. Presentation of the tones was controlled by a second Pentium computer with a Soundblaster 16 sound card and delivered to the subject using Sony headphones. Trials containing clips (amplitudes greater than 250 lV) or flat lines (less than 0.06 lV change over 50 ms) were excluded before the EEG data were corrected for eye movement contamination (Woestenburg et al., 1983). Subsequently, trials containing movement artefact (greater than 50 lV change across any two samples) or drift (100 lV over the whole sweep) were excluded. Following these corrections, stimulus-locked and ⁄ or response-locked averages were obtained for the separate stimuli in the passive condition (standard, small deviant, big deviant) and for the type of response in the active condition (standard correct, deviant correct and deviant error). Stimulus-locked and response-locked averages were aligned to a 100-ms prestimulus baseline. Where appropriate, difference waves were determined, e.g. deviant minus standard in the passive condition or deviant error minus deviant correct in the active condition. Amplitude of the peaks was determined using a semiautomatic peak-picking computer programme and was used as input for subsequent anova analyses. MMN activity was detected in the stimulus-locked averages in a window 0–250 ms, and the Ne and Pe were scored in the response-locked averages 0–150 ms and 150–450 ms, respectively.

Results Behavioural data Average reaction times and number of errors for each type of response are presented in Table 1. Omission errors, i.e. nonresponding to standard stimuli, were rare and excluded from the table and from the analyses because the cell size was too small. Further, subjects responded with the incorrect hand to standard stimuli, on average, 10 times on small deviant blocks and 10 times on big deviant blocks: these trials were also excluded from the analysis. Finally, the first standard stimulus involving a change of ear of presentation was excluded from the analyses because change of location itself can evoke MMN activity and this form of contamination should be excluded from the analysis of the standard stimuli.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

ERPs to no-go stimuli 1963 Table 1. Reaction times in the active condition with the number of trials included in the analyses Small deviant

Big deviant

Response type

Reaction time (ms)

Trials (n)

Reaction time (ms)

Trials (n)

Active condition Standard correct Deviant error

265 ± 27 323 ± 42

25448 1305

255 ± 25 314 ± 57

24818 677

Data are presented as mean ± SD.

RT data were evaluated using anova with the within-subject factors of response type (standard correct, deviant error) and deviancy (big, small). RT was 10 ms faster across types of response to big deviants than to small deviants (F1,12 ¼ 4.8, P ¼ 0.05). Standard correct responses were significantly faster than incorrect no-go responses (F1,25 ¼ 42.2, P ¼ 0.001). The factor deviancy did not interact with the factor response type (F1,12 ¼ 0.04). Incorrect responses occurred less often on trials with a big deviant than on trials with a small deviant (11.8% compared with 23.6%, t ¼ 8.1, P ¼ 0.001). These results indicated that errors of commission to deviant stimuli were slower than correct responses to standard stimuli. Further, the magnitude of deviancy did not differ in its effect across the type of response.

MMN amplitude: overall analyses Figure 1 shows the difference waves between ERPs to standard and deviant stimuli at the three midline electrodes in the passive and active conditions. For passive conditions all trials were included; for active conditions correct responses to standard stimuli and correctly withheld responses to deviant stimuli were included. These time series were used to compute MMN amplitude to the big and small deviant in the separate blocks. An initial, overall analysis of variance was conducted over conditions (active, passive) for the two deviants (big, small) at the three midline leads (Fz, Cz and Pz). MMN amplitude was greater in the passive condition (F1,12 ¼ 5.1, P ¼ 0.05) and was greater for the big deviant (F1,12 ¼ 7.7, P ¼ 0.05). Amplitude differed across electrode locations (F1,12 ¼ 24.0, P ¼ 0.001). Helmert contrasts showed that amplitude at Fz was greater than at the other leads (F1,12 ¼ 19.2, P ¼ 0.001) and was greater at Cz than at Pz (F1,12 ¼ 34.5, P ¼ 0.001). In addition to these main effects, an interaction was obtained between condition and lead (F2,11 ¼ 7.7, P ¼ 0.01), which was interpreted as indicating that the increase in amplitude in the passive condition compared with the active condition was greatest at Fz. No significant interaction was obtained between condition and deviancy (F2,11 ¼ 0.5), deviancy and lead (F2,11 ¼ 1.3), or between condition, deviancy and lead (F2,11 ¼ 0.6). This pattern of results indicates that the same MMN component is measured across conditions, and that the MMN shows the typical fronto-central scalp distribution and the typical sensitivity to stimulus deviancy. These results thus provide a basis for separate analyses on the MMN in the passive and active conditions that are presented in the following sections. Data from the Pz lead are excluded from these analyses.

MMN amplitude: passive condition Difference waves are presented in Fig. 1A and the absolute amplitudes for both deviants are in Table 2. The results of an anova on the

within-subjects factors deviancy (big, small) and electrode (Fz, Cz) revealed significantly greater negativity to the big deviant in comparison to the small deviant (F1,12 ¼ 6.2, P ¼ 0.05); no significant differences were obtained between Fz and Cz (F1,12 ¼ 1.4), nor was there any interaction between deviancy and electrode (F1,12 ¼ 0.02). Thus, we observed the MMN enhancement typically associated with an increase in stimulus deviance.

MMN amplitude: active condition Stimulus-locked difference waveforms are presented in Fig. 1B for correct responses to deviants and in Fig. 2 for commission error responses to deviants after subtracting the ERPs obtained in response to the standard stimulus. The average ERPs that served as the basis for these difference waveforms are presented in Fig. 3. Two observations can be made based on visual inspection of these averages. Firstly, greater peak amplitude at about 100 ms post-stimulus was observed on correct no-go trials to big deviants than to small deviants. This negativity is termed MMN in this paper. Secondly, a second negative peak was seen on error trials at about 400 ms, followed immediately by a positive-going deflection. By contrast, waveforms of the two deviants on correct no-go trials showed a slower and less peaked negativity that maintained constant amplitude; that is, did not manifest the positive-going deflection observed on error trials. anova was applied to evaluate differences in MMN amplitudes, as obtained for errors (MMNincorrect) and correct no-go responses (MMNcorrect) to both deviants, respectively. MMN amplitude was greater on correct no-go trials (F1,12 ¼ 12.6, P ¼ 0.01). Amplitude was greatest at Fz (F1,12 ¼ 13.3, P ¼ 0.01). Overall, MMN amplitude in this analysis did not differentiate between the two deviants (F1,12 ¼ 0.1), but this absence is a direct consequence of the two-way interaction between response type (correct vs. error) and deviancy (F1,12 ¼ 8.7, P ¼ 0.01). Visual inspection of the waveforms (Fig. 3) suggested that this interaction consists of the reversal of the effect of deviancy from MMNcorrect to MMNincorrect amplitude. No other interaction obtained significance (response type · electrode,: F1,12 ¼ 0.6; deviancy · electrode, F1,12 ¼ 0.8; response type · deviancy · electrode: F1,12 ¼ 0.1). In order to focus specifically on the response type–deviancy interaction, subsequent analysis of variance was directed at the MMN in two separate difference waves: (i) deviant correct minus standard correct (referred to as the MMNcorrect); and (ii) deviant error minus standard correct (referred to as the MMNincorrect). MMN amplitudes are presented in Table 2 for both deviants. As in the passive condition, larger MMNcorrect amplitudes were obtained to the big deviant than the small deviant (F1,12 ¼ 8.4, P ¼ 0.01). MMNcorrect amplitude was greater at Fz than Cz (F1,12 ¼ 13.5, P ¼ 0.01). On those deviant trials in which subjects failed to withhold their response, MMNincorrect amplitude was apparently smaller for big deviants compared with small deviants. This difference failed to obtain statistical significance (F1,12 ¼ 2.0). MMNincorrect amplitude was greater at the Fz lead (F1,12 ¼ 11.4, P ¼ 0.01). Deviancy did not interact with electrode (F1,12 ¼ 0.7).

Error-related ERPs Analysis of the Ne and Pe was confined to the active condition because no response was made in the passive condition using response-locked averages (the more conventional approach to the analysis of error-related ERPs). Inspection of Fig. 3 showed no differences in amplitude between big and small deviants for the

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

400

600

800 msec

0

0

0

200

200

200

small deviant

400

400

400

big deviant

600

600

600

Difference waveform: deviant correct-standard good

B: Active

800

800

800

msec

msec

msec

Fig. 1. The stimulus-locked, difference waveforms obtained in the passive (left) and active (right) conditions. In the passive conditions, standards were subtracted from the deviants in the two blocks (small deviants shown in thin lines and big deviants in thick lines), and in the active condition only correct trials were included.

-7

-7 big deviant

-5

-5

small deviant

-3

-1

-1

-3

µV 1

µV 1 msec

3

3

800

5

5

600

7

7

400

Pz 9

Pz 9

200

-7

-7

0

-5

-5

-1

-1 -3

µV 1

µV 1

-3

3

3

msec

5

5

800

7

7

600

Cz 9

Cz 9

400

-7

-7

200

-5

0

-3

-5

-1

200

-3

µV 1

3

5

7

Fz 9

-1

0

Difference waveform: deviant - standard

µV 1

3

5

7

Fz 9

A: Passive

1964 M. Elton et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

ERPs to no-go stimuli 1965 Table 2. The MMN amplitude obtained in the passive and active conditions The MMN amplitude (lV) Condition and deviation

Fz

Cz

Pz

)5.0 ± 2.5 )6.3 ± 3.6

)3.5 ± 1.8 )4.1 ± 2.8

2.4 2.4

)3.6 ± 2.3 )4.5 ± 2.7

)1.3 ± 1.4 )1.7 ± 1.5

2.1 2.4

)3.1 ± 2.0 )2.6 ± 2.5

)0.9 ± 1.3 )0.5 ± 2.0

Passive Deviant correct ) standard correct Small deviation )5.5 ± 3.7 Big deviation )6.9 ± 4.0 Active Deviant correct ) standard correct Small deviation )5.2 ± Big deviation )5.9 ± Deviant error ) standard correct Small deviation )4.6 ± Big deviation )3.8 ±

Data are presented as mean ± SD. In addition to the conventional calculation of the MMN amplitude between waveforms to standard and deviant waveforms, the amplitude is also calculated in the active condition between correct and incorrect no-go response categories.

waveforms for correct responses in the latency range of the Ne. A clear Ne and Pe was observed on incorrect no-go trials (commission error responses) compared with correct responses to standard stimuli (Fig. 4); Ne and Pe amplitudes are reported in Table 3. Ne and Pe amplitudes were measured in difference waveforms for deviant error trials minus standard correct trials. Separate anovas on Ne amplitude included the factors deviancy (big, small) and electrode (Ne: Fz, Cz; Pe: Cz, Pz). No difference in Ne amplitude was obtained

between big and small deviants (F1,12 ¼ 0.1), or between Fz and Cz (F1,12 ¼ 1.0). The interaction between deviancy and electrode was significant (F1,12 ¼ 10.3, P ¼ 0.01) and consisted of greater negativity at Cz to the big deviant but more negativity at Fz to the small deviant. Similarly, the analysis of Pe amplitude revealed no effect of deviancy (F1,12 ¼ 0.01). Pe amplitude was greater at Pz than at Cz (F1,12 ¼ 18.7, P ¼ 0.001). No interaction effect was obtained in this analysis (F1,12 ¼ 2.3).

Discussion The present report examined ERP components that were obtained in an experiment in which subjects made a response to standard go stimuli or withheld their response to deviant no-go stimuli in an active condition. Task difficulty was manipulated by varying the degree of frequency difference between blocks of trials. The response hand was manipulated by varying the ear of presentation. In addition, the ERP was obtained in a passive condition, in which subjects were instructed to ignore the stimuli while reading a book. Specifically, attention was focused on the MMN, the Ne and the Pe. In the analysis of the responses in the active condition, incorrect responses were almost exclusively restricted to commission errors on no-go trials; errors of response omission or incorrect side of response to standard stimuli were too rare to permit analysis. Incorrect no-go responses to both big and small deviant stimuli were slower than responses to standard stimuli. This suggests that errors were not predominantly produced by fast guesses, and that these responses may have been slowed by erroneous stimulus–

Active Difference waveform: deviant error-standard correct Fz 9 7 5 3

µV 1 -1

0

200

400

600

800

msec

0

200

400

600

800

msec

-3 -5 -7

Cz 9 7 5 3

µV 1 -1 -3 -5 -7

small deviant

big deviant

Fig. 2. The stimulus-locked, difference waveforms obtained in the active condition on error trials. Correct standard trials were subtracted from the response on error trials in the two blocks (small deviants shown in thin lines and big deviants in thick lines). ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

1966 M. Elton et al. Fz

Raw ERP’s from active blocks with big deviants 12

Fz

Raw ERP’s from active blocks with small deviants 12

9

9

6

6 3

3

µV

µV 0

0

0

-3

200

400

600

800 msec

-3

0

200

400

600

800 msec

0

200

400

600

800 msec

-6

-6 -9

-9

-12

-12

Cz

Cz

12

12

9

9

6

6

3

3

µV

µV 0

0

0

-3

200

400

600

800 msec

-6

-3 -6

-9

-9

-12

-12

standard correct

deviant correct

deviant error

standard correct

deviant correct

deviant error

Fig. 3. The stimulus-locked, grand average event-related brain potentials (ERPs) for correct responses to standard trials (thin line), correct no-go trials (thick line) and error no-go trials (broken line) for big deviant blocks (left) and small deviant blocks (right).

Fz 8

Table 3. The error negativity positivity amplitudes for the response-locked average following subtraction of standard correct response trials from incorrect deviant responses for both big and small deviant blocks

4

Error negativity and positivity amplitudes (lV)

Responce locked difference waveforms Deviant error-standard correct

µV

Condition and deviation

0 R

100

300

500 msec

-4

-8

Fz

Error negativity Deviant error ) standard correct Small deviation )6.6 ± 3.5 Big deviation )6.0 ± 3.1 Error positivity Deviant error ) standard correct Small deviation – Big deviation –

Cz 8

4

µV

Cz

Pz

)4.9 ± 3.8 )5.7 ± 4.1

– –

6.5 ± 5.7 6.9 ± 5.8

9.6 ± 4.5 9.2 ± 4.4

Data are presented as mean ± SD. 0 R

100

300

500 msec

-4

-8

Pz 8

4

µV 0 R

100

300

500 msec

-4

-8

small deviant

big deviant

Fig. 4. The response-locked difference waveforms at the midline electrodes on error trials for small deviants (thin lines) and big deviants (thick lines). The ERP on correct standard trials was subtracted from the ERP on error trials.

response mapping and subsequent response activation processes (Bernstein et al., 1995). In addition, as expected, errors of commission occurred more frequently to the small deviant. This finding replicated the experimental report of Rabbitt & Vyas (1981). However, in contrast to their finding that subjects corrected the majority of their errors in their easy condition, our subjects persisted in making a large number of uncorrected errors in our easy condition with a big deviant. This difference may be due to the faster rate of trial presentation in our experiment. Our deviance detection hypothesis predicted that a failure or insufficiency in preattentive deviance detection processes results in a failure to prevent the activation of the go response. On the basis of this hypothesis we derived the following set of predictions: (i) MMNincorrect is smaller than MMNcorrect; (ii) the degree of stimulus deviance does not influence MMNincorrect amplitude; (iii) the degree of stimulus deviance does not influence the amplitude of Ne and Pe; and (iv) if the failure to detect deviance results in lack of error awareness,

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

ERPs to no-go stimuli 1967 Pe is absent altogether. The results confirmed the hypotheses on almost all counts. (i) MMNincorrect was observed to be significantly smaller than MMNcorrect, providing a first indication that deviance detection was inadequate on trials that resulted in producing a commission error to a no-go stimulus. (ii) On those deviant trials in which subjects failed to withhold their response, MMNincorrect amplitude was slightly but non-significantly smaller for big deviants compared with small deviants. This pattern, which runs opposite to the effect of stimulus deviance on MMNcorrect, strongly suggests that errors of commission are produced because deviance detection failed; had deviance detection been in operation, then MMNincorrect should have displayed the same sensitivity to the degree of deviance as MMNcorrect. (iii) & (iv) Ne and Pe amplitudes were not modulated by the degree of stimulus deviance in any of the analyses. Stimulus deviance detection Our confidence in having recorded MMN activity was based on its characteristics of being greatest to the big deviant and at the Fz lead in the passive condition (Na¨a¨ta¨nen, 1992). In the active condition, MMN was obtained on correct no-go trials that also demonstrated these characteristics; i.e. greater on big deviant blocks than to small deviants and greatest amplitude at the Fz lead. By contrast, in the active condition this relationship was no longer obtained on incorrect no-go trials. In addition, MMN amplitude was significantly less for incorrect no-go trials, in which subjects produced a response, than for correct no-go trials. This suggests that the process of feature extraction on incorrect no-go trials was less complete than on correct no-go trials and may reflect primarily detection of the change in the location of the stimulus presentation as manifested in the reduced but still pronounced MMN on error trials. Of possible relevance to the difference between correct and incorrect no-go trials, Cowan (1988) argued that two different processes contribute to the MMN in different phases. The detection of deviance generates the MMN and probably marks the first phase of sensory storage. Stimulus comparison, involving a further phase, requires that sensory memory be vivid. It may be noted that it is uncommon, as in the present experiment, to record MMN activity triggered by tones of deviant frequency to which subjects incorrectly respond and the present authors are unaware of any research report on this point.

results of a separate analysis of the lateralized readiness potential (Elton et al., 2004), which suggested that big and small deviants did not differentially affect the degree of activation of the incorrect response. The lack of difference in Ne amplitudes to deviant stimuli may be surprising bearing in mind the difference in probability of committing these two errors. Falkenstein et al. (2000, pp. 93–94) observed that Ne was larger for conditions with fewer errors, i.e. larger for conditions with moderate compared with severe time pressure. Similar results have been reported by Gehring et al. (1993): Ne was larger when accuracy was emphasized than when speed of response was emphasized. In order to elucidate whether the Ne differences relate to differences in accuracy or to other such factors as, e.g. differences in RT or time pressure per se, Falkenstein et al. (2000) performed two additional analyses. First they averaged errors separately for fast and slow errors using a median split of RT. Ne amplitude did not differ between fast and slow errors, which indicated that differences in RT were not responsible for differences in Ne amplitude. Second, they examined the influence of individual error rates on Ne amplitude in a condition with constant time pressure. Half of their subjects had error rates of about 6% and the other half had error rates of about 20%. However, no differences in Ne amplitude were observed. This invariance of Ne amplitude with individual accuracy levels suggests that error probability does not determine Ne amplitude directly. Ne may, thus, co-vary with error rates only through indirect effects like time pressure. Support for our lack of difference in error negativity amplitude to deviant stimuli, which differed markedly in pitch, may be found in the experimental findings of Leuthold & Sommer (1999; Leuthold, personal communication, 2002) where the degree of eccentricity in stimulus presentation did not affect Ne amplitude or latency. Errors of commission may have been due to attentional lapses that occurred to some deviant stimuli against a background of monotonous standard go stimuli (cf. Ridderinkhof et al., 2003). In a study by Scheffers et al. (1999), both performance change and Ne reduction were attributed to a deterioration in early perceptual processes as a function of time on task. Our finding of no difference in MMN amplitude to deviant stimuli may provide indirect confirmation of their interpretation and suggest that in studies into the error negativity attention should also be focused on early components in the ERP that reflect the quality of stimulus processing (cf. Holroyd & Yeung, 2003).

Acknowledgements Error processing Ne and Pe were determined from response-locked averages between incorrect no-go responses to deviant stimuli and correct go responses to standard stimuli: it may be noted that this is the traditional manner of determining Ne from the difference waveform. Ne amplitude was equal at Cz and Fz leads and did not differ between the two deviants. Additionally, no differences in Pe amplitude, maximal at Pz, were obtained. If subjects did detect stimulus deviance on errors of commission but for some reason failed to suppress the go response activation elicited by the location aspect of the tone, then continued processing of the readily discriminable no-go aspect of the big deviant would have resulted in stronger post-error mismatch or conflict with the response elicited by the go aspect, and thus an increase in the amplitude of Ne (Holroyd & Yeung, 2003) compared with small deviants. Our finding of a lack of difference in Ne amplitude between big and small deviants in the predicted direction thus argues against intact deviance detection on commission error trials. This conclusion was corroborated by the

The authors thank Bert Molenkamp and his team for technical support during this experiment, Inge Schoutsen for data collection as a part of her Master’s degree requirements and two anonymous referees for their constructive reviews.

Abbreviations ACC, anterior cingulate cortex; EEG, electroencephalogram; EOG, electrooculogram; ERP, event-related brain potential; MMN, mismatch negativity; Ne, error negativity; Pe, error positivity; RT, reaction time.

References Alho, K., Winkler, I., Escera, C., Huotilainen, M., Virtanen, J., Ja¨a¨sela¨inen, I.P., Pekkonen, E. & Ilmoniemi, R.J. (1998) Processing of novel sounds and frequency changes in the human auditory cortex: magnetoencephalographic recordings. Psychophysiology, 35, 211–224. Bernstein, P.S., Scheffers, M.K. & Coles, M.G.H. (1995) ‘Where did I go wrong?’ A psychophysiological analysis of error detection. J. Exp. Psychol: Human Perception Performance, 21, 1312–1322. Botvinick, M.M., Braver, T.S., Barch, D.M., Carter, C.S. & Cohen, J.D. (2001) Conflict monitoring and conflict control. Psychol. Rev., 108, 624–652.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968

1968 M. Elton et al. Carter, C.S., Braver, T.S., Barch, D.M., Botvinick, M.M., Noll, D. & Cohen, J.D. (1998) Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 280, 747–749. Coles, M.G.H., Scheffers, M.K. & Holroyd, C.B. (2001) Why is there an ERN ⁄ Ne on correct trials? Response representations, stimulus-related components, and the theory of error processing. Biol. Psychol., 56, 173–189. Coles, M.G.H., Smid, H.G.O.M., Scheffers, M.K. & Otten, L.J. (1995) Mental chronometry and the study of human information processing. In Rugg, M.D. & Coles, M.G.H. (Eds), Electrophysiology of Mind: Event-Related Brain Potentials and Cognition. Oxford University Press, Oxford, pp. 86–131. Cowan, N. (1988) Evolving conceptions of memory storage, selective attention, and their mutual constraints within the human information-processing system. Psychol. Bull., 104, 163–191. Devinsky, O., Morrell, M.J. & Vogt, B.A. (1995) Contributions of anterior cingulate gyrus to behaviour. Brain, 118, 279–306. Elton, M., Ridderinkhof, K.R., Spaan, M. & Schoutsen, I. (2004) Mismatch negativity, lateralized readiness potential, and error negativity and positivity in a go ⁄ no-go task. In Ullsperger, M. & Falkenstein, M. (Eds), Errors, Conflicts, and the Brain. Current Opinions on Performance Monitoring. Max Planck Institute for Cognition and Neurosciences, Leipzig, pp. 111–117. Falkenstein, M., Hohnsbein, J. & Hoormann, J. (1995) Event-related potential correlates of errors in reaction tasks. In Karmos, G., Molna`r, M., Cse`pe, V., Czigler, I. & Desmedt, J.E. (Eds), Perspectives of Event-Related Potentials Research. Elsevier, Shannon, pp. 287–296. Falkenstein, M., Hohnsbein, J., Hoormann, J. & Blanke, L. (1991) Effects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks. Electroencephalogr. Clin. Neurophysiol., 78, 447–455. Falkenstein, M., Hoormann, J., Christ, S. & Hohnsbein, J. (2000) ERP components on reaction errors and their functional significance: a tutorial. Biol. Psychol., 51, 87–108. Fiehler, K., Ullsperger, M. & von Cramon, D.Y. (2004) Neural correlates of error detection and error correction: Is there a common neuroanatomical substrate? Eur. J. Neurosci., 19, 3081–3087. Gehring, W.J., Goss, B., Coles, M.G.H., Meyer, D.E. & Donchin, E. (1993) A neural system for error detection and compensation. Psychol. Sci., 4, 385–390. Giard, M.H., Lavikainen, J., Reinikainen, K., Perrin, F., Bertrand, O., Thevenet, M., Pernier, J. & Na¨a¨ta¨nen, R. (1995) Separate representation of stimulus frequency, intensity, and duration in auditory sensory memory. J. Cogn. Neurosci., 7, 133–143. Holroyd, C.B. & Coles, M.G.H. (2002) The neural basis of human error processing: reinforcement learning, dopamine, and the error-related negativity. Psychol. Rev., 109, 679–709. Holroyd, C.B., Dien, J. & Coles, M.G.H. (1998) Error-related scalp potentials elicited by hand and foot movements: evidence for an output independent error-processing system in humans. Neurosci. Lett., 242, 65–68. Holroyd, C.B. & Yeung, N. (2003) Alcohol and error processing. Trends Neurosci., 26, 402–404. Leuthold, H. & Sommer, W. (1999) ERP correlates of error processing in spatial S-R compatibility tasks. Clin. Neurophysiol., 110, 342–357. van der Molen, M., Bashore, T.R., Halliday, R. & Callaway, E. (1991) Chronopsychophysiology: mental chronometry augmented by psychophysiological time markers. In Jennings, R. & Coles, M.G.H. (Eds), Handbook of Cognitive Psychophysiology: Central and Autonomous Nervous System Approaches. Wiley, New York, pp. 9–178.

Na¨a¨ta¨nen, R. (1990) The role of attention in auditory information processing as revealed by event-related potentials and other brain measures of cognitive function. Behav. Brain Sci., 13, 201–288. Na¨a¨ta¨nen, R. (1992) Attention and Brain Function. Lawrence Erlbaum, Hillsdale. Nieuwenhuis, S., Ridderinkhof, K.R., Blom, J., Band, G.P.H. & Kok, A. (2001) Error-related brain potentials are differently related to awareness of response errors: evidence from an anti-saccade task. Psychophysiology, 38, 752– 760. Nieuwenhuis, S., Ridderinkhof, K.R., Talsma, D., Coles, M.G.H., Holroyd, C.B., Kok, A. & van der Molen, M.W. (2002) A computational account of altered error processing in older age: dopamine and the error-related negativity. Cogn., Affect. Behav. Neurosci., 2, 19–36. Nieuwenhuis, S., Yeung, N., van den Wildenberg, W. & Ridderinkhof, K.R. (2003) Electrophysiological correlates of anterior cingulate function in a go ⁄ no-go task: effects of response conflict and trial-type frequency. Cogn., Affect. Behav. Neurosci., 3, 17–26. Rabbitt, P. & Vyas, S. (1981) Processing a display even after you make a response to it. How perceptual errors can be corrected. Q. J. Exp. Psychol., 33A, 223–239. Reason, J. (1990) Human Error. Cambridge University Press, Cambridge. Ridderinkhof, K.R., de Vlugt, Y., Bramlage, A., Spaan, M., Elton, M., Snel, J. & Band, G.P.H. (2002) Alcohol consumption impairs detection of performance errors in mediofrontal cortex. Science, 298, 2209–2211. Ridderinkhof, K.R., Nieuwenhuis, S. & Bashore, T.R. (2003) Errors are foreshadowed in brain potentials associated with action monitoring in cingulate cortex. Neurosci. Lett., 348, 1–4. Ritter, W., Deacon, D., Gomes, H., Javitt, D.C. & Vaughan, H.G. Jr (1995) The mismatch negativity of event-related potentials as a probe of transient auditory memory: a review. Ear Hearing, 16, 52–67. Scheffers, M.K., Coles, M.G.H., Bernstein, P., Gehring, W.J. & Donchin, E. (1996) Event-related brain potentials and error-related processing: an analysis of incorrect responses to go and no-go stimuli. Psychophysiology, 33, 42–53. Scheffers, M.K., Humphrey, D.G., Stanny, R.R., Kramer, A.F. & Coles, M.G.H. (1999) Error-related processing during a period of extended wakefulness. Psychophysiology, 36, 149–157. Ullsperger, M. & von Cramon, D.Y. (2001) Subprocesses of performance monitoring: a dissociation of error processing and response competition revealed by event-related fMRI and ERPs. Neuroimage, 14, 1387–1401. Van ‘t Ent, D. & Apkarian, P. (1999) Motor response inhibition in finger movement and saccadic eye movement: a comparative study. Clin. Neurophysiol., 110, 1058–1072. van Veen, V., Cohen, J.D., Botvinick, M.M., Stenger, V.A. & Carter, C.S. (2001) Anterior cingulate cortex, conflict monitoring, and levels of processing. Neuroimage, 14, 1302–1308. Vidal, F., Hasbroucq, T., Grapperon, J. & Bonnet, M. (2000) Is the ‘error negativity’ specific to errors? Biol. Psychol., 51, 109–128. Winter, O., Kok, A., Kenemans, L. & Elton, M. (1995) Auditory event-related potentials (AEPs) to deviant stimuli during drowsiness and sleep. Electroencephalogr. Clin. Neurophysiol. ⁄ Evoked Potentials Section, 96, 398–412. Woestenburg, J.C., Verbaten, M.N. & Slangen, J.L. (1983) The removal of eyemovement artefact from the EEG by regression analysis in the frequency domain. Biol. Psychol., 16, 127–147.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1960–1968