Feelings of helplessness increase ERN amplitudes in healthyindividuals

Abstract

Experiencing feelings of helplessness has repeatedly been reported to contribute to depressive symptoms and negative affect. In turn, depression and negative affective states are associated, among others, with impairments in performance monitoring. Thus, the question arises whether performance monitoring is also affected by feelings of helplessness.

To this end, after the induction of feelings of helplessness via an unsolvable reasoning task, 37 participants (20 females) performed a modified version of a Flanker task. Based on a previously validated questionnaire, 17 participants were classified as helpless and 20 as not-helpless. Behavioral measures revealed no differences between helpless and not-helpless individuals. However, we observed enhanced Error-Related Negativity (ERN) amplitude differences between erroneous and correct responses in the helpless compared to the not-helpless group. Furthermore, correlational analysis revealed that higher scores of helplessness were associated with increased ERN difference scores. No influence of feelings of helplessness on later stages of performance monitoring was observed as indicated by Error-Positivity (Pe) amplitude.

The present study is the first to demonstrate that feelings of helplessness modulate the neuronal correlates of performance monitoring. Thus, even a short-lasting subjective state manipulation can lead to ERN amplitude variation, probably via modulation of mesencephalic dopamine activity.

Highlights

► Research on neural performance monitoring correlates after helplessness induction. ► Increased feelings of helpless induced ERN amplitude enhancement. ► Behavioral measures were not affected by feelings of helplessness. ► Subjective state manipulation manifests itself in increased ERN amplitudes.

1. Introduction

Monitoring one’s own actions, and in particular the detection of errors or unfavorable outcomes, is of outstanding importance for human beings. However, our perception of performance is not only influenced by objective external performance indicators, but also by internal affective states. Positively valenced internal states are considered to improve cognitive functions in general (Ashby, Valentin, & Turken, 2002; Isen, 2001; Isen, Daubman, & Nowicki, 1987). In contrast, the effect of negatively valenced internal states on cognitive performance is less well understood and difficult to predict (Ashby, Isen, & Turken, 1999; Mitchell & Phillips, 2007). Consequently, the question arises whether performance monitoring is affected by changes in internal states. Event-related potentials (ERPs) pose a useful investigation tool to address this question since they permit precision in the millisecond range. Thus, they allow uncovering the time course of cognitive and emotional processes associated with performance monitoring.

Two ERPs, the Error-Related Negativity – ERN or Ne – (Falkenstein, Hohnsbein, Hoormann, & Blanke, 1991; Gehring, Goss, Coles, Meyer, & Donchin, 1993) and the Error Positivity – Pe – (Falkenstein, et al., 1991; Falkenstein, Hoormann, Christ, & Hohnsbein, 2000) are most relevant in the context of internal performance monitoring. The ERN is a response-locked negative ERP deflection over fronto-central electrode sites, peaking between 50–100 ms after the commission of an erroneous response. The anterior medial cingulate cortex [aMCC; the anterior supracallosal subdivision of cingulate cortex formerly labeled as anterior cingulate cortex (Vogt, 2005)] is thought to be the neuronal generator of the ERN as found by source localization studies (Dehaene, Posner, & Tucker, 1994; Hoffmann & Falkenstein, 2010) as well as functional neuroimaging data (Debener et al., 2005; Ridderinkhof, Ullsperger, Crone, & Nieuwenhuis, 2004). This ERP component is hypothesized to index either response conflict (Botvinick, Carter, Braver, Barch, & Cohen, 2001) or a reinforcement learning signal of the basal ganglia indicating that events are worse than expected (Holroyd & Coles, 2002). A different theoretical account poses that the ERN and related ERPs such as the Feedback-Related Negativity (FRN; Miltner, Braun, & Coles, 1997) are sensitive to the subjective value of committed errors (Gehring & Willoughby, 2002; Pfabigan, Alexopoulos, Bauer, Lamm, & Sailer, 2011; Yeung, Holroyd, & Cohen, 2005). The ERN precedes the Pe, a positive ERP deflection over fronto-central electrode sites, peaking between 200 and 500 ms after conscious error commission (Falkenstein, et al., 1991, 2000). Thus, the Pe is assumed to reflect conscious error processing (Larson, Perlstein, Stigge-Kaufman, Kelly, & Dotson, 2006; Nieuwenhuis, Richard Ridderinkhof, Blom, Band, & Kok, 2001) or affective responses to conscious errors (Falkenstein, et al., 2000; Overbeek, Nieuwenhuis, & Ridderinkhof, 2005). The ERN has been frequently investigated in the context of psychopathology and performance monitoring. However, for instance depressed individuals have been reported to display both larger (Chiu & Deldin, 2007; Holmes & Pizzagalli, 2008) and smaller (Olvet, Klein, & Hajcak, 2010; Schrijvers, De Bruijn, Destoop, Hulstijn, & Sabbe, 2010; Schrijvers, et al., 2009) ERN amplitudes than healthy controls. Olvet et al. (2010) tried to resolve this discrepancy by proposing that mild to moderate depressive symptoms might be related to ERN amplitude enhancement, whereas severe depressive symptoms might be related to ERN amplitude reduction.

Investigating the influence of long-lasting affective states or traits on neuronal correlates of performance monitoring, previous studies reported that individuals scoring high on anxiety and negative affect scales display enhanced ERN amplitudes (Hajcak, McDonald, & Simons, 2003a, 2004) and Pe decrement after error commission (Luu, Collins, & Tucker, 2000). Moreover, ERN enhancement was observed in individuals scoring high on negative affect and negative emotionality scales (Luu et al., 2000).

Interestingly, the influence of short-lasting affective states is less well understood. Previous studies revealed inconsistent and contradictory results. Two studies found ERN amplitude modulation in relation to positive affect induction. Enhanced ERN amplitudes were reported in a flanker task for trials with superimposed pleasant pictures compared to trials with unpleasant and neutral ones (Larson et al., 2006). In contrast, decreased ERN amplitudes were reported after the presentation of pleasant compared to neutral movie clips prior to a continuous performance task (Van Wouwe, Band, & Ridderinkhof, 2011). Wiswede, Münte, Goschke, and Rüsseler (2009) found enhanced ERN amplitudes during a flanker task when presenting unpleasant pictures prior to the flanker stimuli. Furthermore, enhanced ERN amplitudes were reported in participants receiving derogatory feedback during a flanker task (Wiswede, Münte, & Rüsseler, 2009), and after the induction of self-relevant failure in a probabilistic learning task (Unger, Kray, & Mecklinger, 2012). However, the induction of sad feelings via movie clips prior to a flanker task did not alter ERN amplitudes directly. Instead, the correlation between ERN amplitudes and self-reported sadness was moderated by neuroticism (Olvet & Hajcak, 2012). Moreover, a study by Clayson, Clawson, & Larson, 2011 failed to report ERN modulation after derogatory feedback in a flanker task. These contradictory results raise two issues. Firstly, it might be that the described discrepancies are related to different presentation modes of the affective stimuli. In this regard, Van Wouwe et al. (2011) proposed that pre-task affect induction might lead to a milder and more tonic effect than the repetitive presentation of affective stimuli. Secondly, it might be that affect induction influences neuronal correlates of performance monitoring more strongly in cases where participants were exceedingly engaged in task performance because they received individual and personalized feedback by the experimenter. Thus, the present study combined both assumptions and investigated a tonic affect induction procedure with concurrent high task involvement and subjective salience. In particular, participants performed a cognitive reasoning task with unsolvable items, thereby possibly inducing subjective feelings of helplessness. Subsequently, a simple choice reaction task was administered to investigate the consequences of the helplessness induction on behavioral and neuronal correlates of performance monitoring.

Feelings of helplessness can be considered as a specific variant of affect modulation. Seligman (1975) was the first to introduce the concept of learned helplessness. He postulated lack of control over aversive events as its main characteristic. Seligman (1975) concluded that uncontrollability induces motivational (e.g., decreasing escape behavior), cognitive (e.g., learning deficits; Mikulincer, 1994), and emotional deficits (feelings of anxiety and depression). Moreover, learned helplessness is considered to contribute to psychopathological conditions such as depression (Overmier, 2002). Seligman (1975) also drew parallels between learned helplessness and depression.

For the present study, we employed a learned helplessness induction explicitly targeting motivational and affective components of helplessness prior to a choice reaction task. Based on Wiswede, Münte, and Rüsseler (2009) who observed enhanced ERN amplitudes after the presentation of unpleasant pictures, we hypothesized that the induction of feelings of helplessness would yield comparable effects on error monitoring. Compared to previous studies merely presenting affective stimulus material, we chose an experimental manipulation addressing the perception of individual skills of our participants to directly manipulate subjective saliency.

We expected enhanced ERN amplitudes after error commission in helpless compared to not-helpless participants indicating depression- or negative affect-like stimulus processing in these individuals. In particular, we assumed that the amplitude difference between correct and erroneous responses (ΔERN) would be enhanced in helpless participants. For the Pe amplitude, we expected larger amplitudes after error than correct responses (Falkenstein, et al., 1991). Additionally, we explored Pe amplitude variation of helpless and not-helpless participants and the potential effects of the helplessness induction on behavioral task measures such as reaction times, error rates, conflict adaptation, and post-error slowing. In particular conflict adaptation effects might be susceptible to the present helplessness manipulation. For instance, effects of mood induction on conflict-driven control have recently been observed (van Steenbergen, Band, & Hommel, 2010).

2. Material and Methods

2.1. Participants and measures

Initially, 50 volunteers (25 females) participated in our study. Thirteen participants had to be excluded from further analysis due to data acquisition artifacts (n=2), or due to committing less than five errors (n=7) or more than 200 errors (n=4). The remaining 37 participants (20 females) were aged between 19 and 34 years with a mean age of 25.27±3.89 years. All participants were right-handed as assessed via the Edinburgh Handedness Inventory (Oldfield, 1971), had normal or corrected-to-normal vision, reported no past psychiatric disorder, and did not suffer from a current psychiatric disorder as assessed with a SCID-I screening (Wittchen, Wunderlich, Gruschwitz, & Zaudig, 1996). All participants gave written informed consent prior to the experiment. The study was conducted in accordance with the Declaration of Helsinki (1981) and local guidelines of the University of Vienna. Each participant received a remuneration of 20 Euros at the end of the experiment.

The experiment consisted of a helplessness induction phase in which a cognitive reasoning task was administered, and the experimental phase in which a reaction time task was administered. The helplessness induction phase was a prerequisite to manipulate participants’ actual motivational and affective states. Prior and after the helplessness induction phase, participants were administered the German version of the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988) to assess whether the helplessness induction led to individual changes in positive and negative affect. Moreover, participants filled in a modified version of a previously validated helplessness questionnaire directly after the helplessness induction phase (Bauer, Pripfl, Lamm, Prainsack, & Taylor, 2003; Fretska, Bauer, Leodolter, & Leodolter, 1999) asking for participants’ general motivation, their experience of control when confronted with solvable reasoning tasks, and their experience of loss of control when confronted with unsolvable reasoning tasks. Ratings on eleven questions had to be given on a five-point scale (++, +, 0, −, −−), indicating levels of agreement. This procedure of increasing loss of control has been successfully established in two previous experiments to induce feelings of helplessness (Bauer, et al., 2003; Fretska, et al., 1999). We chose this questionnaire as an indirect measure instead of directly asking for subjective feelings of helplessness as done in previous studies (e.g., Diener, Kuehner, & Flor, 2010), which might have resulted in biased responses.

To clarify which items of this specific questionnaire were best suitable to investigate feelings of helplessness, a principal component analysis was performed. Helplessness questionnaire data of the current participants as well as questionnaire data of 120 individuals (60 females; mean age 24.73±2.53 years) who took part in a comparable helplessness induction (unpublished data) were subjected to PCA. Applying a varimax rotation the PCA yielded a three factor solution for the eleven items of the questionnaire. Five items asking for excitement, aggressiveness, despondency, importance to solve the difficult tasks as well as general importance of task solution were subsumed under the factor “negative affective state”. Four items asking for initial feelings of motivation, control, and demotivation as well as difficulty at the beginning of the session were subsumed under the factor “initial affective state”. The remaining two items asking for passivity and demotivation after the session were subsumed under the factor “feelings of helplessness”, thereby accounting for 22, 20, and 17% of the variance in the data. The factor “feelings of helplessness” comprised two items regarding passivity (If you were not able to solve any items for a longer period, did you become passive?) and demotivation (If you were not able to solve any items for a longer period, did you feel demotivated?). Individual scores of these items were added up and served as basis for participants’ helplessness classification. Ratings on the factor “initial affective state” were used to assess potential group differences prior to the helplessness induction phase. Ratings on the factor “negative affective state” were used to corroborate the PANAS scales concerning affectivestate.

2.2. Helplessness induction procedure

Participants were presented with a series of mathematical reasoning items that gradually became unsolvable over the course of the induction procedure. Stimulus presentation (Pentium IV, 3.00 GHz) was performed using E-Prime software (Psychology Software Tools, Inc., Pittsburgh, PA). Participants were seated 70 cm in front of a 19 in. cathode ray tube monitor (Sony GDM–F520; 75 Hz refresh rate) in a sound-attenuated room. EEG data collection started with the helplessness induction phase. Participants were presented with 60 items each consisting of a series of seven numbers, arranged according to a mathematical rule, and presented centered on the screen. The task was to identify a number which would have logically continued the sequence. To this end participants had to choose one out of four possible solutions presented below each item (e.g., series of numbers: 3 5 9 15 23 33 45; possible solutions: 55, 57, 58, 59; correct answer: 59). Participants were given 30 s to answer each item by pressing a button on a 5-button response pad corresponding to what they assumed to be the correct solution. During the first half of the helplessness induction procedure, six items were unsolvable, i.e., no correct solution was presented, and 24 items were solvable. During the second half of the induction procedure, 24 items were unsolvable, and 6were solvable. Prior and after the helplessness induction, participants filled in the PANAS to assess differences in mood states, and after the task the helplessness questionnaire.

2.3. Procedure to measure effects of helplessness on error processing

Subsequent to the helplessness induction, participants were administered an arrowhead version of the Eriksen Flanker task (Eriksen & Eriksen, 1974). Five-arrow strings were presented centered vertically and horizontally on the screen. Half of the trials were congruently aligned flanker arrays (>>>>>, <<<<<), and the other half showed incongruent arrays (>><>>, <<><<). The task was to indicate the right- or left-hand direction of the middle arrow by two distinct motor responses on a response pad (Psychology Software Tools, Inc., Pittsburgh, PA), as quickly and accurately as possible. The experiment started with 20 training trials to familiarize participants with the task. Each trial started with a white fixation cross on a black screen presented for 1000 ms. Subsequently, the four outer arrows of the five-arrow string were presented. After 100 ms, the middle arrow was blended into the string for another 35 ms. This sequential procedure was chosen to enhance interfering effects of the flanking stimuli on the target stimuli (Kopp, Rist, & Mattler, 1996) and to consequently increase error rates. Immediately afterwards, the screen turned black for 870 ms and participants were required to respond via button press. Button 1had to be pressed for left-hand arrows with the left index finger, and button 5for right-hand arrows with the right index finger. During the subsequent inter-stimulus interval, a fixation cross was presented for a random duration of 450–650 ms. Overall, participants completed 400 trials where congruent and incongruent flanker arrays were presented randomly. After blocks of 50 trials each, the participants were given a short period of rest. The helplessness induction (30 min), questionnaires (5 min), and the flanker task (25 min) lasted about 60 min.

2.4. Electroencephalographic recording

Electroencephalogram (EEG) was recorded from 61 Ag/AgCl electrodes equidistantly embedded in an elastic cap (model M10; EASYCAP GmbH, Herrsching, Germany). EEG recordings were referenced to a balanced sterno-vertebral site—above the seventh vertebra and both sternoclavicular joints (Stephenson & Gibbs, 1951). Bipolar recordings of vertical and horizontal electrooculogram (EOG) were performed for off-line eye movement correction. EOG electrodes were placed about 1 cm above and below the left eye, and on the outer canthi. Two pre-experimental eye-movement calibration tasks were performed to calculate subject- and channel-specific weighted parameters for artifact correction (Bauer & Lauber, 1979). By applying a skin-scratching procedure prior to EEG data collection (Picton & Hillyard, 1972), electrode impedances were kept below 2 kΩ as assessed with a manual impedance meter. Signals were collected using an AC amplifier set-up with a time constant of 10 s (Ing. Kurt Zickler GmbH, Pfaffstätten, Austria). All signals were recorded within a frequency range of 0.016–125 Hz and sampled at 250 Hz for digital storage.

2.5. Behavioral data analysis

Participants were divided into a helpless and a not-helpless group based on their scores on the factor “feeling of helplessness” which served as group factor in the subsequent analyses. Dependent t tests were calculated to assess whether helpless and not-helpless participants differed in their scores on the factors “initial affective state” and “negative affective state” of the helplessness questionnaire. Positive and negative affective ratings before and after the helplessness induction as assessed by the PANAS were subjected to a three-way repeated-measures analysis of variance (ANOVA) with the between-subject factor group (helpless vs. not-helpless) and the within-subject factors affect (positive and negative) and time of measurement (prior vs. after helplessness induction). Additionally, difference scores for positive and negative affective ratings (measurement prior to the experiment [T1] minus measurement after the helplessness induction [T2]) were used for correlation analysis.

Reaction times were defined as the interval from the onset of the middle flanker stimulus until button press. In line with recently proposed procedures (Hajcak, Moser, Yeung, & Simons, 2005; Wiswede, Münte, Goschke, & Russeler, 2009), trials with reaction times lower than 200 ms and higher than 800 ms were discarded from analysis. Individual mean reaction times were calculated participant-, condition-, and block-wise. Experimental trials were divided into two equal blocks with 200 trials each to determine whether feelings of helplessness decrease in the course of the experiment. Analysis of the mean reaction times was performed using two three-way repeated-measures ANOVAs with the between-subject factor group (helpless vs. not-helpless), the within-subject factor block (half 1vs. half 2), and either the within-subject factor response type (correct vs. error) or stimulus type (congruent vs. incongruent). Furthermore, error rates in percent were calculated per participant for congruent and incongruent trials and analyzed via a three-way repeated-measures ANOVA with the between-subject factor group (helpless vs. not-helpless) and the within-subject factors block (half 1vs. half 2) and stimulus type (congruent vs. incongruent). Conflict adaptation (Gratton, Coles, & Donchin, 1992) was assessed via extracting mean reaction times for correct congruent trials either preceded by congruent trials (congruent–congruent) or incongruent trials (incongruent–congruent), and mean reaction times for correct incongruent trials either preceded by congruent trials (congruent–incongruent) or incongruent trials (incongruent–incongruent). These four trial sequences were presented in an unconstrained random order with approximately equal trial numbers. The mean reaction times were subjected to a four-way repeated-measures ANOVA with the between-subject factor group (helpless vs. not-helpless) and the within-subject factors block (half 1vs. half 2), actual trial (congruent vs. incongruent), and previous trial (congruent vs. incongruent). To determine post-error slowing (Rabbitt, 1966), reaction times after erroneous and after correct trials were extracted participant- and block-wise. Since the post-correct trials (i.e., trials following correct trials) occurred more frequently than post-error trials (i.e., trials following error trials), only the post-correct trials with the fastest reaction times were chosen for comparison with the post-error trials.¹ The mean reaction times of the same number of post-error and post-correct trials were subjected to a three-way repeated-measures ANOVA with the between-subject factor group (helpless vs. not-helpless) and the within-subject factors block (half 1 vs. half 2) and response type (post-error trials vs. post-correct trials).

2.6. EEG data analysis

The weighted EOG signals were subtracted from each EEG channel off-line and prior to further analysis. Additionally, blink coefficients were calculated and subtracted from each EEG channel trial-by-trial using a template matching procedure (see Lamm, Fischmeister, & Bauer, 2005 for a similar application of this approach). Subsequent data analysis was carried out using EEGLAB 6.03b (Delorme & Makeig, 2004), implemented in Matlab 7.5.0 (The MathWorks, Inc., Natick, MA). EEG data were low-pass filtered with a cut-off frequency of 30 Hz (roll-off 6 dB/octave). Epochs were time-locked to participants’ responses, starting 100 ms prior to the button press and lasting for 650 ms. The 100 ms interval preceding the button-press served as baseline interval. A semi-automatic artifact correction was applied to the epoched data. Artifact-afflicted trials that met the following criteria were labeled automatically: voltage values exceeding±75 μV in any channel or a voltage drift of more than 50 μV. The respective trials were rejected eventually if visual inspection also indicated artifact affliction. Subsequently, artifact-free epochs were averaged separately for each participant for the following two conditions: (1)trials including correct responses after congruent and incongruent flanker stimuli were combined into the condition correct-response; (2)trials including incorrect responses after congruent and incongruent flanker stimuli were combined into the condition error–response. Only participants with a minimum of five error trials which yielded distinct ERN peaks were included in further data analysis. Other authors also found reliable ERN peaks when using a minimum of five trials in their analysis (Amodio et al., 2004; Hajcak & Simons, 2008). On average, 12.43 errors (SD=7.87) per participant were subjected to analysis.² ERN and Pe peak amplitudes were assessed at midline electrode sites Fz, Cz, and Pz where both components were found to be most pronounced, which is consistent with previous literature (Gehring, et al., 1993; Wiswede, Münte, & Goschke, et al., 2009). Analysis of ERN amplitudes was based on the identification of the most negative peak in relation to baseline within a time window of 10–120 ms after the response. Pe amplitude was identified as the most positive peak in relation to baseline within a time window of 200–400 ms after the response.

ERN and Pe amplitudes values were analyzed separately by means of three-way repeated measures ANOVAs with the between-subject factor group (helpless vs. not-helpless), and the within-subject factors electrode site (Fz, Cz, Pz) and response type (correct-response vs. error-response). Additionally, ΔERN values were calculated at each electrode site (amplitude difference between error and correct trials) and subjected to a two-way repeated measures ANOVA with the between subject factor group and the within-subject factor electrode site. Additionally, Spearman’s rank correlations (r_s) were calculated separately for each participant and condition to investigate the association between the not normally distributed helplessness score, PANAS scores, absolute ΔERN amplitudes, and Pe amplitudes.

If not stated otherwise, significant interaction effects were explored with Tukey’s HSD post hoc test. Partial eta-squared is reported for significant results to demonstrate effect sizes of the respective ANOVA model (Cohen, 1973). All statistical analyses were performed using PASW Statistics 18.0 (IBM SPSS Statistics, Somer, NY, USA) and Statistica 6.0 (StatSoft Inc., Tulsa, OK); the alpha level was set at p<.05. If necessary, degrees of freedom were adapted applying the Greenhouse–Geisser correction.

3. Results

3.1. Behavioral results

Based on the helplessness questionnaire, 17 participants (10 females) were classified as helpless (M=8.53, SE=0.15), whereas 20 (10 females) were classified as not-helpless (M=6.10, SE=0.26). Individual scores on the feelings of helplessness factor ranged from three to ten (out of a maximum of ten). Participants who rated both questionnaire items positively (i.e., reached at least eight points) were subjected to the helpless group, whereas participants who gave neutral or negative ratings (i.e., reached seven points or less) were subjected to the not-helpless group. Thus, the present study applied an even stricter helplessness classification than previous studies did (Bauer, et al., 2003; Fretska, et al., 1999). Gender was not associated with helplessness classification (χ₍₁₎=0.53, p=0.467). Helpless and not-helpless participants did not differ significantly from each other concerning the questionnaire factors “negative affective state” (t(35)=0.89, p=0.381) and “initial affective state” (t(35)=0.89, p=0.379). The PANAS ratings are depicted in Fig. 1.

Fig. 1.

Mean PANAS scores for positive affect (PA) and negative affect (NA), before (pre) and after (post) the helplessness induction, for not-helpless (gray) and helpless (dotted pattern) participants. Error bars indicateSEM.

Regarding the affective ratings of the PANAS, we observed a main effect for affect (F(1,35)=138.68, p<0.001, η_p²=0.79), no significant effect for time of measurement (F(1,35)=0.54, p=0.47), but a significant interaction effect for affect×time of measurement (F(1,35)=13.20, p=0.001, η_p²=0.27). Post hoc tests indicated that positive affect ratings significantly decreased after the helplessness induction (p=0.027), whereas no significant increase for negative affect ratings was observable (p=0.161).³ No significant main or interaction effect emerged for the group factor (all p-values>0.194).

Main effects of the behavioral data analysis are depicted in Table 1. Illustrations of reaction times, post-error slowing and number or errors are depicted in Fig. 2.

Table 1.

F-value, degrees of freedom, p-value and partial eta-squared (for significant results) of the main effects of the ANOVA models analyzing reaction times, error rates, conflict adaptation, and post-error slowing.

Contrast	F	d.f.	p-value	partialŋ2
RT congruency
group	0.96	1,35	0.333
block	0.76	1,35	0.390
stimulus type	904.44	1,35	<0.001	0.96
RT correctness
group	0.63	1,35	0.431
block	3.08	1,35	0.088
response type	117.04	1,35	<0.001	0.77
percent errors
group	1.72	1,35	0.198
block	0.94	1,35	0.338
stimulus type	117.31	1,35	<0.001	0.77
conflict adaptation
group	0.78	1,35	0.384
block	0.55	1,35	0.462
actual trial	1249.50	1,35	<0.001	0.97
previous trial	33117.00	1,35	0.003	0.22
post-error slowing
group	0.44	1,35	0.513
block	5.95	1,35	0.020	0.15
response type	20.00	1,35	<0.001	0.36

Fig. 2.

Upper panel: mean reaction times of congruent (CON) and incongruent (INCON) trials for the first and second half of the experiment on the left; mean reaction times for error (ERR) and correct (CORR) trials for both halves on the right. Lower panel: mean reaction times of post-error slowing after post-error (PE) and post-correct (PC) trials for both halves on the left; mean error rates in percent for both halves on the right. Not-helpless participants are depicted in gray, helpless participants in a dotted pattern. Error bars indicateSEM.

Mean reaction times were faster after congruent than incongruent trials (F(1,35)=904.44, p<0.001, η_p²=0.96) and after error than correct trials (F(1,35)=117.04, p<0.001, η_p²=0.77). No group effect emerged in both ANOVAs (all p’s>0.634), no other main effects or interactions were observed either (all p’s>0.061).

Participants committed more errors during incongruent than congruent trials (F(1,35)=117.31, p<0.001, η_p²=0.77). No significant group (F(1,35)=1.72, p=0.198), block (F(1,35)=0.94, p=0.338), or interaction effect (all p’s>0.172) emerged.

Conflict adaptation analysis revealed a significant interaction between actual trial and previous trial (F(1,35)=44.96, p<0.001, η_p²=0.56). Post hoc tests revealed that incongruent–congruent trials were slower than congruent–congruent trials (p<0.001) reflecting a reduction of facilitation of the present congruent flanker array. In contrast, congruent–incongruent trials were slower than incongruent–incongruent trials (p<0.001) reflecting a reduction of interference of the present incongruent flanker array. No group (p=0.384) or any other effects (all p’s >0.230) emerged.

Regarding post-error slowing, reaction times were shorter after post-correct trials compared to post-error trials (F(1,35)=20.00, p<0.001, η_p²=0.36). Overall, post-error and post-correct mean reaction times increased in the second half of the experiment (F(1,35)=5.95, p=0.020, η_p²=0.15). Group membership had no significant impact on post-error slowing (p=0.437), no interaction effect emerged either (all p’s>0.215).

3.2. ERP results

Grand-average wave forms of correct and error responses of helpless and not-helpless participants at midline electrodes Fz, Cz, and Pz as well as topographical maps of ERN and Pe scalp distribution are illustrated in Fig. 3.

Fig. 3.

Grand average waveforms and scalp topographies. Grand averages of error and correct responses of helpless (black) and not-helpless (gray) individuals depicted at electrode sites Fz, Cz, and Pz. The dotted line at time 0indicates the button press; negative is drawn upwards per convention. Scalp topographies of the ERN in the upper panel at component peak latency for helpless (39 ms post response) and not-helpless (57 ms post response) participants. The lower panel depicts scalp topographies of the Pe at component peak latency for helpless (263 ms post response) and not-helpless (264 ms post response) participants.

Regarding ERN amplitudes, main effects of electrode site (F(2,70)=25.13, p<0.001, η_p²=0.42) and response type (F(1,35)=116.15, p<0.001, η_p²=0.77) emerged. A significant interaction of electrode site and response type was observed (F(2,70)=20.57, p<0.001, η_p²=0.37) indicating that only the error amplitudes differed significantly over the three electrode locations (all p’s<0.001). Amplitudes after correct responses were comparable for Fz, Cz, and Pz (all p’s>0.94). The factor group yielded no significant main effect (F(1,35)<0.01, p=0.982), but a significant group×response type interaction (F(1,35)=4.43, p=0.043, η_p²=0.11) was observed. This interaction indicated that group differences might be driven by participant-wise differences between correct and error trials. Indeed, ΔERN analysis revealed main effects for electrode site (F(2,70)=20.57, p<0.001, η_p²=0.37) and group (F(1,35)=4.43, p=0.043, η_p²=0.11). ΔERN was largest at Cz (all p’s<0.008) and ΔERN was enhanced in helpless compared to not-helpless participants. The interaction did not reach significance (F(2,70)=2.37, p=0.101).

Regarding Pe amplitudes main effects of electrode site (F(2,70)=13.51, p<0.001, η_p²=0.28) and response type (F(1,35)=131.11, p<0.001, η_p²=0.79) emerged. The factor group was not significant (F(1,35)=0.65, p=0.426), but subsumed under a significant interaction with electrode site (F(2,70)=4.38, p=0.034, η_p²=0.11). Post hoc tests revealed that Pe amplitude was larger at Pz than Fz in helpless participants (p=0.006). Furthermore, an electrode site×response type interaction emerged (F(2,70)=18.80, p<0.001, η_p²=0.35) indicating that Pe amplitudes were most positive after error compared to correct trials at Cz (all p’s<0.004).

The feelings of helplessness score correlated significantly with ΔERN assessed at Cz where amplitude differences were largest (r_s=0.40, p=0.014) indicating that the more helpless the participants felt, the larger the amplitude difference between erroneous and correct responses became. No significant correlations emerged for ΔERN and the affective ratings of the PANAS (all p-values>0.148), or for Pe amplitudes and the ratings (all p-values>0.293). Additionally, the feelings of helplessness score correlated significantly with the difference score for positive affect (r_s=0.33, p=0.044). Further analysis indicated that this result was due to the individual ratings on the positive affect scales of the PANAS only after the helplessness induction (r_s=−0.44, p=0.006); thus, participants who felt more helpless after the helplessness induction showed lower ratings of positive affect. No significant correlations emerged for the negative affect scales of the PANAS and the helplessness score (all p-values>0.550) Fig. 4.

Fig. 4.

Scatter plot including a regression line of the correlation of helplessness scores and ERN at electrode location Cz. Two outliers are marked with triangles instead of dots. Note that the correlational analysis still yields a significant result without these two participants (r_s =0.390, p=0.021).

4. Discussion

The present study aimed at investigating the effect of induced feelings of helplessness on internal performance monitoring. To this end, we applied a flanker task to examine neuronal correlates of error processing immediately after a helplessness induction via unsolvable cognitive reasoning stimuli. This task successfully induced reduced positive affect in all participants; concurrently, feelings of passivity and demotivation were observed in a subset of participants. Subsequently, these individuals were classified as helpless and displayed enhanced amplitude differences between correct and erroneous responses (ΔERN) compared to the not-helpless participants. In contrast, later stages of error processing as indexed by Pe amplitudes were not affected by the helplessness induction. Additionally, no differences in behavioral measures of the flanker task were observed between these two groups either.

Our study is the first to probe affective state modulation via inducing feelings of helplessness prior to the investigated task. Our data suggest that ERN amplitudes are not only modulated by affective state but also by motivational factors.

Enhanced subjective error saliency in helpless compared to not-helpless participants might explain the current results. Indeed, ERN amplitudes are reported to be enhanced when task salience is high (Falkenstein, Hohnsbein, & Hoormann, 1995). Gehring and Willoughby (2002) and Luu et al. (2000) proposed that ERN enhancement is reflecting either the affective significance or the emotional valence of the observed stimuli or response. In line with this assumption, Wiswede, Münte, and Rüsseler (2009) observed ERN enhancement after inducing negative affect via derogatory feedback, thereby manipulating subjective error salience.

In addition, enhanced feelings of self-relevance caused by the helplessness induction which targeted individual cognitive abilities might also lead to larger affective significance, and thus to larger ERN amplitudes (Gehring & Willoughby, 2002; Luu, et al., 2000). For example, Unger et al. (2012) manipulated the attribution of error commission to either high or low self-relevance. The authors assumed that threatening participants’ self-worth after error commission is associated with a concurrent increase in motivation to avoid errors (Brunstein, 2000). Indeed, ERN amplitudes were reported to be enhanced after self-relevant errors in their study (Unger, et al., 2012).

Our results are in line with the prominent theoretical account of reinforcement learning (reinforcement-learning theory; RL-theory; Holroyd & Coles, 2002). This theory proposes the ERN as a neuronal signal generated by a monitoring system within the basal ganglia for outcomes that are worse than expected. Changes in subjective, affective, or motivational states might also alter the expectation for different outcomes. Thus, ERN amplitude modulation due to short- or long-term subjective state changes might be interpreted as a consequence of changes in tonic mesencephalic dopamine levels (Ashby & Casale, 2003; Wiswede, Münte, & Goschke, et al., 2009). Indeed, a review on mesencephalic dopamine levels and stress coping strategies recently illustrated the strong link between inhibition of mesencephalic dopamine release and the experience of uncontrollable and unavoidable stressful situations (Cabib & Puglisi-Allegra, 2012). Thus, our helplessness induction presumably caused a decrease of mesencephalic dopamine levels in those participants experiencing feelings of helplessness. This effect may be reflected in enhanced ERN amplitudes in the helpless compared to the not-helplessgroup.

The three theoretical assumptions of larger ERN amplitudes caused by enhanced error saliency, enhanced self-relevance of an error, and inhibition of mesencephalic dopamine release do not exclude each other. On the contrary, it might be possible that enhanced self-relevance caused by the helplessness induction yielded enhanced error saliency in helpless participants, which might have influenced mesencephalic dopamine levels explicitly in these individuals.

The present data contradict several studies failing to observe ERN amplitude modulation after subjective state modulation. For example, Moser, Hajcak, and Simons (2005) failed to find ERN modulation after inducing anxious feelings during a flanker task. Although the authors used a very potent form of threat induction (confronting spider phobics with a tarantula during the task), no transfer occurred from the threat induction to task completion. This null finding might be explainable by the fact that the induced feelings of threat were not task-related and had no motivational impact on task performance. The same explanation might hold true for missing effects on task performance when inducing sad affect via the presentation of sad film clips prior and sad music during a flanker task (Olvet & Hajcak, 2012). In contrast, Clayson et al. (2011) induced negative affect via derogatory feedback in a replication study of Wiswede, Münte, and Rüsseler (2009), but reported no ERN amplitude modulation, although enhanced levels of anxiety, anger, and fatigue, and decreased levels of vigilance were observed after task completion. This observation is rather surprising given the fact that high self-relevance and repeated presentations of the derogatory feedback stimuli were used in this study. However, this null finding might be attributable to the fact that Clayson et al. (2011) did not offer course credit or any monetary remuneration to their participants compared to Wiswede, Münte, and Rüsseler (2009). Recently, it has been shown that ERPs related to external performance monitoring are sensitive to monetary reward (Van den Berg, Shaul, Van der Veen, & Franken, 2012). It is possible that also ERPs related to internal performance monitoring are prone to incentives. Thus, course credit or the final pay-off might have increased the self-relevance of Wiswede’s experimental manipulation and thereby enhanced ERN amplitudes. Furthermore, error rates were increased in the derogatory feedback group in the beginning of Wiswede’s study (2009), whereas Clayson et al. (2011) reported no feedback group differences in the first half of his experiment. Overall, error rates were higher in the Wiswede study than in the Clayson study. Thus, Wiswede’s derogatory feedback might have been more salient to the derogatory feedback group. In contrast, Clayson reported that both feedback groups felt worse after the experiment, so no specific effect emerged for the derogatory group. These results again point towards the importance of subjective error saliency on ERN amplitudes.

The present study adds a new perspective to the on-going debate as to whether ERN amplitudes are state- or trait-dependent. Reviewing the literature on ERN and its relation to affective states, personality, and psychopathology yields still inconclusive results. Support for the trait-dependent account of ERN amplitudes (Clayson, et al., 2011; Moser, et al., 2005; Olvet & Hajcak, 2012; Tops, Boksem, Wester, Lorist, & Meijman, 2006), as well as for the state-dependent account (Larson, et al., 2006; Van Wouwe, et al., 2011; Wiswede, Münte, & Goschke, et al., 2009; Wiswede, Münte, & Rüsseler, 2009) has been reported. The distinction between state or trait dependency of ERN amplitudes is important for the theoretical framework of error-related brain activity. Recently, ERN enhancement has been claimed to be associated with increased levels of depression and anxiety (Olvet & Hajcak, 2008). The authors go to such lengths as to propose that the ERN might be an endophenotype for specific psychiatric disorders. However, this assumption is only applicable when the ERN is truly trait- but not state-dependent. The present data clearly support the state-dependent account of ERN amplitudes since healthy individuals were investigated with no heightened levels of depression or anxiety, as assessed with the SCID-I screening (Wittchen, et al., 1996).

Regarding questionnaire data, we assessed positive and negative affective states prior and after the helplessness induction. Interestingly, no significant increase of negative affect scores was observed after the helplessness induction, but a significant decrease of positive affect scores. Positive affect scores of the PANAS reflect the extent to which someone is feeling enthusiastic, active, and alert at the moment. On the contrary, negative affect scores might be rather related to high levels of arousal and activation (Russell, 1980) during task engagement. Indeed, we found a significant correlation between the difference score in negative affect and the “negative affective state” factor of the helplessness questionnaire (r_s=−0.46, p=0.004), but no such effect for the positive affect difference scores (r_s=0.17, p=0.302). Thus, a decrease in feelings associated with positive affect fits better to the assumption of heightened levels of passivity and demotivation after the helplessness task than an increase in feelings associated with negative affect. To support this notion, lower ratings of positive affect correlated significantly with higher scores of feelings of helplessness in the presentdata.

Our behavioral results replicated previous flanker task findings; in particular faster reaction times for error responses, lower error rates for congruent trials (Eriksen & Eriksen, 1974), conflict adaptation effects (Gratton, Coles, Sirevaag, Eriksen, & Donchin, 1988), as well as post-error slowing effects (Rabbitt, 1966) were observed. Additionally, the missing behavioral effect of the helplessness induction is also in line with previous research reporting that affect- or drug-induced changes of the ERN amplitude are not consequently accompanied by behavioral effects (Gehring, Himle, & Nisenson, 2000; Hajcak, McDonald, & Simons, 2003b; Hajcak, et al., 2004; Johannes et al., 2001; Riba, Rodriguez-Fornells, Morte, Münte, & Barbanoj, 2005; Wiswede, Münte, & Goschke, et al., 2009). Moreover, Weinberg, Riesel, and Hajcak (2012) conclude in their review that ERN enhancement is seldom accompanied by behavioral effects, whereas diminished ERN amplitudes are frequently accompanied by increased error rates and response slowing. Furthermore, comparable error rates in both groups strengthen the present results since ERN amplitudes are susceptible to error frequency. The more frequent an error is occurring, the more decrease in ERN amplitudes can be observed (Yeung, Botvinick, & Cohen, 2004). Thus, the present group effects cannot be explained by error frequency since both groups displayed comparable performances in the task. This emphasizes the sensitivity of ERN amplitudes for our subjective state manipulation. Furthermore, the present study avoided confounds by experimenter effects because group allocation took place after data collection.

In line with Clayson et al. (2011), the present study used a mixed sample of intermediate size to investigate the impact of sex on correlates of internal performance monitoring. Recently, it was observed that male compared to female participants displayed enhanced ERN and Pe amplitudes during a flanker task (Larson, South, & Clayson, 2011). However, no sex effects emerged in the present study (adding the between-subject factor gender to all comparisons revealed no significant effects, all p’s>0.069).

5. Conclusion

The present results explicitly demonstrate that even a short-lasting subjective state modulation like a decrease in motivation has considerable impact on error monitoring in a specific subset of individuals. Future research has to characterize these individuals more precisely to assess whether our finding can be integrated in therapeutic programs, e.g., for depressive individuals. Additionally, we emphasize the impact of personal relevance and motivation on affective states. An affect induction directly manipulating individuals’ feelings of self-relevance might have a stronger effect on ERN amplitudes in comparison to paradigms where only passive viewing of emotional stimuli is required.

Conflict of interest

All authors declare that this research project was conducted in the absence of any commercial or financial relationship that could be construed as potential conflict of interest.