ارتباطات عصبی ضربه پاسخ در اختلال شخصیت مرزی: شواهد ERP برای کاهش نظارت عمل

Abstract Patients with borderline personality disorder (BPD) are characterized by marked impulsive behaviour. The impulsive response style of patients with BPD may be associated with diminished action monitoring, which can be investigated by measuring the error-related negativity (ERN). The ERN is an ERP component generated in the anterior cingulate cortex (ACC) following erroneous responses. Behavioural and ERP measurements were obtained during performance on a speeded two-choice reaction task in a group of patients with BPD (N = 12) and in a group of age-matched controls (N = 12). The ERP results showed that ERN amplitudes were reduced for patients with BPD, as were the P300 amplitudes after late feedback. The behavioural results confirmed a more impulsive response style for the BPD group, reflected in larger RT differences between correct and incorrect responses and in an increase in erroneous responses to the easy congruent stimuli. Additionally, analyses on post-error congruency effects demonstrated that controls adjusted their behaviour following errors, but patients with BPD did not. The attenuated ERNs indicate reduced action monitoring in patients with BPD. This suggests that the ACC, or the action-monitoring network it is part of, is not functioning optimally. Due to this reduced action monitoring, patients with BPD do not learn from their errors as well as controls. Consequently, they do not adjust their behaviour when necessary and thus maintain their impulsive response style.

1. Introduction Marked impulsivity is seen as one of the main characteristics of borderline personality disorder (BPD), together with rapidly changing mood states, aggressive behaviour, instability of interpersonal relationships, self-image, and affects. Impulsive behaviour may express itself in promiscuity, substance abuse, adverse financial behaviour, reckless driving, and binge eating. Suicidal behaviour and self-mutilation are also frequently related to impulsivity (American Psychiatric Association, 2000). Although impulsiveness is an important clinical feature, the number of studies investigating its neural correlates is still relatively small. This may be explained by the lack of a unified definition, which is due to the wide range of behaviours in which the trait is present (see e.g. Evenden, 1999). Moeller et al. (2001), in their review of the psychiatric aspects of impulsivity, concluded that a reliable definition should at least include the elements of the various behavioural models that have been developed based on findings from targeted laboratory tasks. These elements are: (1) rapid, unplanned reactions to stimuli before complete processing of information, (2) decreased sensitivity to negative consequences of behaviour, and (3) lack of regard for long-term consequences. With the current study we aimed at investigating the neural correlates of the first two elements in patients with BPD. Rapid, unplanned reactions and diminished sensitivity to resulting erroneous responses can be studied by means of electrophysiological measurements during a speeded forced-choice task. Especially the discovery of an event-related potential (ERP) component associated with error or conflict detection has given this type of action monitoring research an important impetus. This so-called error negativity (Falkenstein et al., 1991) or error-related negativity (ERN; Gehring et al., 1993) is characterized by a sharp negative deflection over frontocentrally located electrodes appearing within 100 ms after an error has been made. Source localization and fMRI studies have found the anterior cingulate cortex (ACC) as the most likely generator of the ERN (see e.g., Dehaene et al., 1994, Kiehl et al., 2000 and Ullsperger and von Cramon, 2001), a finding that is in line with earlier studies demonstrating error-related activity in unit recordings from the ACC in monkeys (see e.g., Niki and Watanabe, 1979). The ACC is a mediofrontal brain structure known for its rich innervation from and to other regions of the brain and its rich concentration of different types of neurotransmitters like dopamine. The area is highly interconnected to the motor system, the limbic system, and to prefrontal regions. Because of these characteristics, the ACC has been described as the interface between cognition, motor control, and the drive of the organism (Paus, 2001). Originally, the ERN was taken to be elicited by a mismatch, i.e., after the error detection system has failed to match a representation of the actual behaviour with a representation of the desired behaviour (see e.g. Falkenstein et al., 1991 and Gehring et al., 1993). More recently, Holroyd and Coles (2002) extended this original interpretation in their so-called reinforcement-learning theory of the ERN. According to the theory, predictive error signals indicating whether events turn out to be worse than expected are carried to various brain areas by the dopamine system. These error signals are used to improve performance in order to prevent future errors. When a predictive error signal arrives at the ACC, the ERN is elicited. Alternative accounts refer to the ERN as the reflection of conflict that arises when two incompatible response tendencies are simultaneously activated (Botvinick et al., 2001, Cohen et al., 2000 and Yeung et al., 2004). For the current study it is relevant that all three accounts agree that the ERN is generated in the ACC, that it reflects the outcome of an action-monitoring process, and that it is used to optimize performance in the future. Differences in ERN amplitude have been observed in a variety of personality traits. Individuals low on socialization tend to exhibit smaller ERN amplitudes (Dikman and Allen, 2000), whereas individuals with greater negative affect (Luu et al., 2000) or those with obsessive-compulsive personality traits (Hajcak and Simons, 2002) show larger ERN amplitudes. Pailing et al. (2002) specifically investigated the relation between the ERN and impulsivity. In their study, subjects with large reaction-time differences between correct and incorrect responses had smaller ERN amplitudes. Because (overly) fast reaction times generally lead to more erroneous responses, these larger reaction-time differences were taken to reflect a more impulsive response style. This motivated us to investigate whether this line of reasoning could also be applied to patients with BPD. PET and fMRI studies investigating neuropsychiatric disorders have demonstrated increased ACC activity in individuals with obsessive-compulsive disorder (OCD; see e.g., Adler et al., 2000 and Ursu et al., 2003) and decreased ACC activity in patients with schizophrenia (Carter et al., 2001 and Laurens et al., 2003). These differences in ACC activity were also reflected in action monitoring: enhanced ERN amplitudes were found in individuals with OCD (Gehring et al., 2000), whereas decreased ERN amplitudes were observed in patients with schizophrenia (Alain et al., 2002, Bates et al., 2002, Kopp and Rist, 1999 and Mathalon et al., 2002). With regard to BPD, a number of brain imaging studies examining patients have shown hypometabolism in prefrontal cortical areas (see e.g., De la Fuente et al., 1997, Goyer et al., 1994 and Soloff et al., 2003). Recently, Tebartz van Elst et al. (2003) demonstrated volume loss of the right ACC in their BPD sample. The authors suggested that specifically this volume loss might differentiate BPD from other neuropsychiatric disorders. In the present and to our knowledge the first such study in patients with BPD, we employed a speeded two-choice task while measuring ERN amplitudes. We predicted that patients with BPD would show increased impulsivity in different behavioural measures and reduced action monitoring as evidenced by smaller ERN amplitudes. In order to examine the entire process of action monitoring from stimulus onset to feedback processing, we also examined two other ERP components known to be involved in action monitoring, namely the stimulus-locked N2 and the feedback-locked P300. The amplitude of the N2 is thought to reflect the monitoring of response conflict that arises from simultaneously active response tendencies as it is enlarged after incongruent stimuli compared to congruent ones (see e.g. Yeung et al., 2004). Consequently, the N2 is a reflection of a relatively early process. As we specifically anticipated group differences for the later processes directly related to erroneous responses, we did not expect to find any differences with regard to N2 amplitudes. Previously, using the same task, we demonstrated that P300 amplitudes elicited by less predicted negative feedback were larger for subjects who attributed more meaning to the feedback information (De Bruijn et al., 2004b). When patients with BPD indeed display decreased sensitivity to negative consequences of behaviour, smaller P300 amplitudes to negative feedback are expected compared to controls.

3. Results 3.1. Behavioural analyses Fig. 1 shows the mean RTs of both groups. With regard to correct responses, the group RTs differed, with the BPD group overall responding more slowly (382 ms) than the control group [347 ms; F (1, 22) = 5.97, p = 0.023]. Usually, responses to congruent stimuli are given faster than responses to incongruent stimuli. This congruency effect was also present in the current data, as the RTs for correct responses were shorter following congruent (349 ms) than following incongruent stimuli [380 ms; F (1, 22) = 130.29, p < 0.001]. The interaction between congruency and group was not significant (F < 1). Mean reaction times for correct and incorrect responses to congruent and ... Fig. 1. Mean reaction times for correct and incorrect responses to congruent and incongruent trials for the control and the BPD group. Error bars represent standard deviations. Figure options As in similar studies, incorrect responses were faster (322 ms) than correct responses [365 ms; F (1, 22) = 96.26, p < 0.001]. However, the interaction with group revealed that this response effect was larger for the BPD group (56 ms) than for the control group [28 ms; F (1, 22) = 10.78, p = 0.003]. Table 1 comprises the percentages of correct, incorrect, and late responses. Overall, error rate did not differ between the two groups (F < 1). More errors were made to incongruent trials (7.0%) than to congruent ones [2.0%; F (1, 22) = 164.91, p < 0.001]. A significant interaction between group and congruency revealed that this effect differed between the two groups: the BPD group made more errors in response to congruent trials (2.6%) than the control group [1.5%; F (1, 22) = 8.02, p = 0.010]. The proportion of late responses did not differ between the groups [F (1, 22) = 2.99, p = 0.098]. Incongruent trials induced a larger number of late responses (9.0%) than the congruent ones [3.8%; F (1, 22) = 83.56, p < 0.001]. The interaction between group and congruency showed that this effect was similar for both groups (F < 1). Table 1. Mean percentages of correct, incorrect and too late responses to congruent and incongruent trials for the control and the BPD group Control group BPD group Congruent Incongruent Congruent Incongruent Correct 45.3 (1.7) 34.2 (3.6) 43.0 (3.1) 33.9 (3.3) Incorrect 1.5 (0.5) 7.5 (1.9) 2.6 (1.5) 6.5 (2.1) Too late 3.2 (1.8) 8.3 (2.9) 4.4 (1.4) 9.6 (2.5) Standard deviations are given in parentheses. Table options 3.2. ERP analyses 3.2.1. Response-related ERN As Fig. 2 shows, ERN amplitude was smaller for the BPD group (−6.18 μV) than for the control group [−10.30 μV; F (1, 22) = 4.91, p = 0.037]. There was no main effect of electrode site, indicating that the ERN was equally large at both electrodes (F < 1). The interaction between electrode site and group was not significant either (F < 1). In addition, no differences were found for the peak latency of the ERN (F < 1). For both groups, the ERN peaked around 74 ms after response onset. Importantly, ERN amplitude was also reduced in the BPD group (−4.65 μV) when patients who used medication were excluded from the analysis [F (1, 15) = 6.42, p = 0.023]. Grand average response-locked waveforms for correct and incorrect responses for ... Fig. 2. Grand average response-locked waveforms for correct and incorrect responses for the control and the BPD group. Electrodes FCz and Cz are depicted. Responses are given at t = 0 ms. Figure options 3.2.2. Stimulus-related N2 Though Fig. 3 may suggest an overall difference in N2 amplitudes between the two groups, analyses demonstrated that the main effect of group was not significant (F < 1). As expected from previous studies, N2 amplitude was more negative for incongruent stimuli (−5.65 μV) than for congruent ones [−4.67 μV; F (1, 22) = 4.83, p = 0.039]. More importantly, the interaction between congruency and group showed that this N2 congruency effect did not differ between the two groups (F < 1). Grand average stimulus-locked waveforms for correct congruent and correct ... Fig. 3. Grand average stimulus-locked waveforms for correct congruent and correct incongruent stimuli for the control and the BPD group. Electrodes Fz and FCz are depicted. Stimuli are presented at t = 0 ms. Figure options 3.2.3. Feedback-related P300 Fig. 4 shows the feedback-locked ERP waveforms. A main effect of feedback type was present, indicating that P300 amplitude was affected by feedback [F (2, 44) = 41.31, p < 0.001]. Simple contrasts referenced to ‘correct’ feedback (5.14 μV) showed that P300 amplitude was larger after ‘late’ feedback [13.79 μV; F (1, 22) = 54.82, p < 0.001], but not different after ‘incorrect’ feedback [4.84 μV; F < 1]. Although there was no main effect for group (F < 1), the interaction between type of feedback and group was significant [F (2, 44) = 10.04, p = 0.002]. Here, simple contrasts referenced to correct feedback revealed that this was caused by smaller P300 amplitudes after late feedback for the BPD group (10.69 μV) compared to the control group [16.88 μV; F (1, 22) = 10.73, p = 0.003]. These effects on P300 amplitude were also present when patients who used medication were excluded from the analysis. Importantly, in these restricted analyses P300 amplitude after late feedback was also found to be smaller for the BPD group [9.44 μV; F (1, 15) = 6.65, p = 0.021]. 1 Grand average feedback-locked waveforms for correct, incorrect, and too late ... Fig. 4. Grand average feedback-locked waveforms for correct, incorrect, and too late feedback stimuli for the control and the BPD group. Electrodes Cz and Pz are depicted. Feedback stimuli are presented at t = 0 ms. Figure options 3.3. Performance adjustments As important as the ability to detect an error is the ability to adjust performance following an incorrect response to prevent similar errors in the future. One way to accomplish a successful performance adjustment in the flankers task is to reduce the interference effect of the flanking letters by focusing on the central target in the letter string. This type of performance adjustment can be investigated in more detail by examining behavioural congruency effects following incorrect and correct responses (see e.g., De Bruijn et al., 2004a and Ridderinkhof et al., 2002). Typically, standard congruency effects caused by slower responses to incongruent stimuli than to congruent ones are reduced following errors. This reduction reflects a change in response strategy after an incorrect response. The congruency effect is computed by subtracting RTs on congruent stimuli from RTs on incongruent stimuli following correct and incorrect responses. Note that only correct responses that were preceded by incongruent trials were included in the analyses. As can be seen in Fig. 5, the analyses only revealed a significant interaction between post-correctness and group [F (1, 11) = 9.64, p = 0.005]. This interaction indicated that a reduction of the congruency effect following errors was present for the control group (18 ms), but not for the BPD group (−12 ms). Performance adjustments following erroneous responses. Reaction-time differences ... Fig. 5. Performance adjustments following erroneous responses. Reaction-time differences between correct congruent and correct incongruent trials (congruency effect) are smaller following erroneous (dark) responses than when following correct (light) responses for the control group, but not for the BPD group. Error bars represent standard errors.