خطای منفی و مثبت همانگونه که به دیگر شاخص های ERP کنترل توجه و پردازش محرک مربوط می شود

Abstract We compared individual differences in the ERP associated with incorrect responses in a discrimination task with other ERP components associated with attentional control and stimulus discrimination (N2, P3, CNV). Trials with errors that are detected by the subject normally produce a negativity (NE) immediately following the response followed by a positivity (PE). The morphology of the NE and the PE is similar to that of the standard N2–P3 complex on correct discrimination trials. Our findings suggest that the PE is a P3 response to the internal detection of errors. The NE, however, appears to be distinct from the N2. Finally, even though both the contingent negative variation (CNV) and the NE are associated with prefrontal cortex and the allocation of attention to response accuracy, the NE and CNV did not relate to one another.

Introduction The error-negativity (NE, or error-related negativity, ERN) is a recently documented component of the event-related potential (ERP) and has been associated with acknowledged incorrect responses that occur in target discrimination tasks (e.g. Falkenstein et al., 1991, Gehring et al., 1993, Dehaene et al., 1994 and Scheffers et al., 1996). The error waveform is time-locked to the behavioral response, as opposed to stimulus onset, and consists of a negative deflection (NE) followed by a positive deflection (PE). The NE has engendered much interest because examining the parameters within which it occurs may enrich our understanding of the processes involved in the monitoring and evaluation of response tendencies ( Gehring et al., 1993 and Luu et al., 2000). However, both the NE and the PE share functional and morphological similarities with other ERP components. The purpose of this paper is to compare individual differences in the NE–PE with other ERP components associated with attentional control and stimulus discrimination, specifically the contingent negative variation (CNV) and the N2–P3 complex. 1.1. Error-negativity (NE) The NE is time-locked to the execution of an incorrect response, is absent for trials on which the correct response is made when the subject is certain of the correctness of that response ( Coles et al., 2001), and does not seem to be dependent on the type of error made. Scheffers et al. (1996), using a simple Go/NoGo task, found that both errors of choice (incorrect responses on go-trials) and errors of action (uninhibited responses on NoGo trials) were associated with an NE of similar morphology, latency, and scalp distribution. Tasks used to elicit the NE are usually not difficult (e.g. the flanker task of Eriksen and Eriksen, 1974; the letter discrimination task of Falkenstein et al., 1991) so that errors are usually caused by quick or impulsive responding rather than by the inability of the individual to discriminate the stimuli or choose the correct response. On tasks in which individuals are unable to determine correct or incorrect responses on their own, an NE occurs only when error-feedback is provided. Miltner et al. (1997), presented feedback 600 ms following behavioral response and found that an NE was produced only when feedback indicated that an incorrect response had been made. These data indicate that the elicitation of the NE is not dependent on whether the detection of the error is internally driven or signalled by external cues, typically as long as there is awareness that an error has occurred ( Miltner et al., 1997). In addition to its error detection role, NE has been associated with the magnitude of individuals’ response to their own error, as well as with error correction and compensation mechanisms ( Gehring et al., 1993). In fact, we have found the NE to be related to individual differences in the impulsivity of response style on the task ( Pailing et al., 1999). However, there is still some question as to whether the NE is more directly related to the processes involved in the generation of an error signal or to the processes following it, such as emotional or remedial reactions ( Bernstein et al., 1995 and Stemmer et al., 2000, for evidence that the NE follows error detection rather than being generated simultaneously with it ). It is, nonetheless, consistently observed that the NE occurs when a mismatch results from the comparison between that which is anticipated and that which actually occurs. This observation led Falkenstein et al. (1991) to suggest that the NE might resemble other similar negative potentials, such as the mismatch negativity (MMN; Näätänen, 1992) which reflects an automatic mismatched auditory stimulus. However, Bernstein et al. (1995) found that the NE was influenced by response conditions and participants’ response strategies, which highlight the endogenous nature of the NE and distinguish it from the mismatch negativity, which is typically dependent on the physical characteristics of the stimuli ( Coles and Rugg, 1995). It might also resemble the N400 which reflects an endogenous mismatch in the semantic ( Byrne et al., 1995) or perceptual domain ( Bobes et al., 2000), although this has not been examined directly. 1.2. Comparing NE with CNV The first comparison we wanted to make was between the NE and CNV. The CNV is elicited by providing the individual with a warning stimulus followed at some fixed interval such as 2000 ms by a second ‘imperative’ stimulus ( Walter et al., 1964). In Go–NoGo versions of this task, the participant is cued as to whether or not the second stimulus requires a response. In those cases in which a response is required, a large negative potential is observed in the interval between the warning and the imperative stimuli. Behavioral measures associated with frontal lobe processing have been shown to correlate with the initial portion of the CNV ( Segalowitz et al., 1992 and Dywan and Segalowitz, 1996) and the CNVs of patients with unilateral prefrontal lesions are reduced ( Rosahl and Knight, 1995). As well, presenting stimuli in conjunction with a high amplitude spontaneous CNV increases the likelihood of a correct behavioral response ( Stamm, 1987). Similarly, the size of the CNV has been shown to correlate with performance accuracy (e.g. Stamm, 1987 and Hohnsbein et al., 1998). The more negative the deflection preceding the target stimulus, the less likely the subject is to make an incorrect response. Thus, the CNV is associated with response anticipation and the NE is associated with response monitoring, and thus both are sensitive to the production of accurate responses. In other words, the CNV and NE are endogenous components that jointly bracket behavioral responses on tasks which require attending to imperative stimuli. Error monitoring has been associated with the anterior cingulate, the dorsolateral prefrontal cortex, and the left premotor cortex (Carter et al., 1998). Both anterior cingulate and dorsolateral prefrontal cortex have also been associated with the control of attention (Mesulam, 1981, Posner and Petersen, 1990, Corbetta et al., 1991, Chow and Cummings, 1999 and Mesulam, 1999). Physiologically, the NE and CNV are both associated with these brain structures in the prefrontal cortex. Whereas there is some controversy concerning the specific generator site of the CNV, both magnetoencephalogram (MEG) ( Basile et al., 1997 and Tarkka and Basile, 1998) and intracellular recordings ( Fuster, 1987) support the view that it is associated with the dorsolateral and medial prefrontal cortex. Topographical EEG mapping has shown that the response begins prefrontally and spreads posteriorly where it reaches maximum scalp potential at the vertex ( Yamamoto et al., 1986). There are however, lesion data suggesting that the CNV is associated with the cingulate gyrus and other subcortical structures ( Rosahl and Knight, 1995). There is less controversy surrounding the generator site of the NE, which has been associated with the anterior cingulate cortex ( Dehaene et al., 1994, Miltner et al., 1997, Dikman and Allen, 2000 and Stemmer et al., 2000). Given this similarity in cognitive functions and anatomical source associated with error monitoring and attentional control, our goal was to examine the NE and CNV as they occur in the same individuals during the performance of attention-demanding tasks. We hypothesized that if the amplitude of both CNV and NE could serve as an index of attentional capacity or attentional allocation to response accuracy, then these seemingly separate components of the ERP should be related to one another. 1.3. Comparing NE–PE with the N2–P3 component complex The second focus of this study is based on the observation that the NE and the PE have morphological features similar to other well-researched ERP components, the N2 and P3. In fact, the morphology and topography is so similar that there has been some discussion as to whether they might represent the same phenomenon ( Miltner et al., 1997, Falkenstein et al., 1999 and Leuthold and Sommer, 1999). The N2 is a negative deflection that occurs at about 200 ms following stimulus onset and is maximal over the centro-frontal scalp region ( Rugg et al., 1988). It is associated with stimulus discrimination ( Ritter et al., 1979) and categorization ( Rugg, et al., 1988). Whereas the inhibitory N2 has been compared with and shown to be independent of the NE ( Falkenstein et al., 1999), the target (non-inhibitory) N2 has not been similarly compared. Overwhelming interest in the significance in the NE has overshadowed the investigation of the PE deflection. Nonetheless, the topography and polarity of the PE has led some investigators to view it as a late P3 (e.g. Miltner et al., 1997). The P3 is a positive deflection with a latency from 300 ms following a simple stimulus to as much as 800 ms in response to complex tasks. It is characterized by a parietally maximal scalp distribution ( Coles, Smid, et al., 1995) and can be elicited by several paradigms but especially when a rare target event occurs in the context of more frequent non-target stimuli ( Coles and Rugg, 1995; Coles, Smid, et al., 1995). The cognitive correlates of P3 include orientation, attention, stimulus evaluation and memory, but there is a traditional controversy regarding these functional components ( Ritter et al., 1979, Picton et al., 1984, Donchin and Coles, 1988 and Hoffman, 1990; Coles, Smid, et al., 1995). The NE–PE has both overall morphological and functional similarities to the N2–P3 complex. In this context, we wished to examine whether the NE–PE might be an N2–P3 response to the internal error-detection event. The N2–P3 component complex itself is a response to a salient stimulus (e.g. Donchin and Coles, 1988 and Polich, 1993) and the detection of an incipient erroneous response could be such a salient event. In effect, we wished to learn more about the NE and PE by examining them in the context of other more traditional components of the ERP responses. We accomplish this by measuring CNVs in a Go/NoGo task, and NE/PE in an Eriksen task and comparing these components using a correlational design.

. Results 3.1. Correct and incorrect responses: NE and PE The mean error rate was 8.4% (range, 3–17%) of the 480 trials. An average of 31.3 errors were made on the 320 incongruent trials and an average of 9.0 errors were made on the 160 congruent trials. The accuracy rate was better for the trials with congruent flankers than for those with incongruent trials (t (13)=3.53, P<0.005). In addition, responses were faster for correct congruent flanker trials (M=433 ms) than for the correct trials with incongruent flankers (M=463 ms), t (13)=10.28, P<0.0001, suggesting the incongruent flankers led to response interference. A negative deflection, maximal at CZ, followed incorrect but not correct responses (see Fig. 1). The NE and PE reported in this paper will be computed from the ERPs to incongruent error trials time-locked to response unless otherwise stated. These waveforms were essentially identical to those which included congruent error trials. The negative deflection (NE) relative to a −600 to −400 ms preresponse baseline (M=−10.0 μV at CZ) peaked at 66.6 ms on average following the response for incorrect trials and was followed by a positive peak, PE, (M=11.7 μV at CZ) occurring about 250 ms following the response. In contrast, correct trials produced a positive peak about the time of the response followed by a negative drift peaking about 200 ms later. The dramatic distinction between correct and incorrect trial ERPs is illustrated in the difference waveform but is also seen in the original ERPs (see Fig. 1 and Fig. 2). Averaged ERP waveforms for correct and incorrect trials time-locked to the ... Fig. 1. Averaged ERP waveforms for correct and incorrect trials time-locked to the response (n=14). The group average peak amplitudes are attenuated compared with single scores due to inter-subject latency jitter. The hash marks represent the time of the response. Figure options The difference waveform time-locked to response (incorrect minus correct ... Fig. 2. The difference waveform time-locked to response (incorrect minus correct responses). The hash marks represent the time of the response. Figure options 3.2. Contingent negative variation (CNV) The CNV elicited by the Go trials was maximal at CZ (M=−10.34 μV, see Fig. 3). The CNV was scored time-locked to the imperative stimulus with a baseline of −4200 to −4000 prior to the imperative stimulus (i.e. the 200 ms before the first stimulus). As expected, the negative deflection was greater for the Go relative to the NoGo trials, F (1, 13)=11.70, P=0.005, with an increased deflection leading up to the imperative stimuli (when we divide the 1400 ms CNV period into four equal epochs, there is a systematic increase in the negative deflection across epochs, F (3, 11)=10.21, P=0.002). In order to derive a more pure reflection of response expectation, we used the CNV from the Go trials with the CNV from the NoGo trials partialed out by regression, producing a residualized CNV score. Averaged waveforms for CNV. The imperative stimulus onset is indicated as zero ... Fig. 3. Averaged waveforms for CNV. The imperative stimulus onset is indicated as zero ms. The warning stimulus onset is at −2000 ms. Go trials are indicated by the thick line and NoGo trials are indicated by the thin line. Figure options 3.3. Relationship between NE and CNV Correlation analyses were used to relate the amplitude of the NE to the residualized CNV Go trials. The NE measured at FZ and CZ did not correlate significantly with the CNV measured at any of the three sites. However, the NE measured at PZ did correlate with the CNV at CZ and at PZ (see Table 1). However, since the degree of negative deflection of the NE with respect to the early baseline can be influenced by the size of the abortive P3 preceding it (see Fig. 1), we recalculated these correlations using linear regression with the amplitude at −200 to 0 ms (with respect to the −600 to −400 ms baseline) partialed out. The correlations remained significant (for CNV at CZ: t=2.46, P=0.032; for CNV at PZ, t=3.41, P=0.006). However, the NE is not normally measured at PZ and indeed was only +0.6 μV (with respect to the early baseline). Thus, this correlation at PZ does not represent strong support for a relationship between the CNV and the NE, although this finding remains intriguing. Table 1. Correlation of NE amplitudes with each epoch of CNV amplitude CNV NE FZ CZ PZ FZ −0.21 0.25 0.24 CZ 0.12 0.53 0.65* PZ 0.37 0.52 0.68** The NE was elicited during the incorrect incongruent trials time-locked to response and scored with a baseline correction of −600 to −400. * P<0.05; **P<0.01. Table options 3.4. ERPs to correct responses To examine the NE in the context of other ERP components, the EEG waveforms elicited by trials correctly responded to in the flanker paradigm were time-locked to stimulus onset and scored using a 100 ms prestimulus baseline. This produced six scorable deflections: N1, P2, N2i, P3i, N2ii, and P3ii (see Fig. 4) 1, as has been found before with complex stimuli (e.g. Segalowitz et al., 1997). Evidently, the flanker paradigm used to elicit the NE in this study does not produce the classic N2–P3 complex that is normally obtained in less complex oddball paradigms. Consequently, we compare the NE to the N2i and N2ii and the PE to the P3i and P3ii, respectively. The NE was not significantly correlated with the N2 components at any of the fronto-central sites (see Table 2). The one significant correlation between the NE at PZ and the N2ii at PZ appears to be an artifact of the P3 leading up to the NE since it disappears when the P3 is partialed out. The PE amplitude correlated with the P3 components elicited by correct trials at CZ and PZ (see Table 3). ERP waveforms to correct responses time-locked to the stimulus onset. The hash ... Fig. 4. ERP waveforms to correct responses time-locked to the stimulus onset. The hash marks represent the time of the stimulus onset. Figure options Table 2. Correlation of NE amplitude with ERP components measured at frontal (FZ), central (CZ) and posterior (PZ) midline scalp electrode sites Site ERP components from correct trials Congruent trials Incongruent trials N2i N2ii N2i N2ii FZ −0.08 0.20 0.19 0.28 CZ 0.27 0.46 0.19 0.48 PZ 0.28 0.55* 0.48 0.50 ERP components were elicited by correct congruent trials on the visual flanker task and time-locked to stimulus onset. The NE was elicited during the incorrect incongruent trials time-locked to response and scored with a baseline of −600 to −400 ms. The significance of the relationship between N2ii (congruent trials) and NE at PZ disappears when the P3 is partialed out (see text). *P<0.05 level. Table options Table 3. Correlation of the PE with the amplitudes of ERP components at frontal (FZ), central (CZ), and posterior (PZ) midline electrode sites Site ERP components from correct trials Congruent trials Incongruent trials P3i P3ii P3i P3ii FZ 0.33 0.46 0.31 0.24 CZ 0.66* 0.75* 0.68** 0.77** PZ 0.46 0.73** 0.72** 0.70** ERP components were elicited during correct congruent trials of the visual flanker task and were time-locked to stimulus onset. The PE component was elicited during the incorrect incongruent trials time-locked to response and scored with a baseline correction of −600 to −400 ms. *P<0.05; **P<0.01. Table options 3.5. Topographical maps Topographical maps of the NE, the PE, and the P3ii of the correct trials are displayed in Fig. 5 (based on the algorithms of Junghofer et al., 1997). These maps represent the data from the 11 of the 14 subjects for whom we had a full montage. As can be seen in these maps, the NE has a strong negativity in the centromedial frontal scalp region replicating the results of Luu et al. (2000). The PE based on incorrect trials time-locked to response and the P3ii based on correct trials time-locked to the stimulus both display a strong positivity over the centro-parietal regions of the scalp verifying that these two components are similar (see Fig. 5). Topographical maps (n=11) of the NE, PE and the P3ii from the correct trials ... Fig. 5. Topographical maps (n=11) of the NE, PE and the P3ii from the correct trials time-locked to stimulus onset.