Introduction
Schizophrenia is characterized by enduring difficulties in adaptive functioning, including diminished engagement in productive, goal-directed activities (Barch & Dowd, Reference Barch and Dowd2010; Blanchard et al. Reference Blanchard, Kring, Horan and Gur2011). One crucial element of adaptive functioning is the accurate monitoring of responses and integration of feedback, which informs decision making and guides behavior based on the consequences of our actions. Sensitivity to favorable versus unfavorable actions and outcomes has been extensively investigated in healthy subjects through event-related potential (ERP) measures of neural activity. The aim of the current study was to assess two aspects of response monitoring in schizophrenia, namely sensitivity to the internal detection of errors compared to correct responses indexed by error-related negativity (ERN), and sensitivity to external feedback that indicates good versus bad outcomes indexed by feedback negativity (FN).
ERN
The ERN is a response-locked ERP component that reflects the activity of a neural system involved in monitoring actions and detecting errors (Falkenstein et al. Reference Falkenstein, Hohnsbein, Hoormann, Blanke, Brunia, Gaillard and Kok1990; Gehring et al. Reference Gehring, Goss, Coles, Meyer and Donchin1993; Simons, Reference Simons2010). The ERN, which is typically studied using simple choice reaction time (RT) tasks (e.g. flanker tasks), differs following erroneous versus correct responses; it is evident as a larger negative deflection at frontocentral sites approximately 50 ms following the commission of erroneous compared to correct responses. The size of the ERN has been shown to reflect the motivational significance of errors (Hajcak et al. Reference Hajcak, Moser, Yeung and Simons2005). Converging evidence from source localization, functional magnetic resonance imaging (fMRI) and single unit recording studies indicates that the ERN is generated within the anterior cingulate cortex (ACC), a structure centrally involved in response monitoring and error detection (Taylor et al. Reference Taylor, Stern and Gehring2007). Prevailing reinforcement learning theories propose that the ERN reflects dopaminergic disinhibition of neurons in the ACC when actions are evaluated as worse than anticipated (Holroyd & Coles, Reference Holroyd and Coles2002); this early error detection then recruits input from other brain regions (e.g. the dorsolateral prefrontal cortex) to enhance performance and facilitate learning.
Reductions in the ERN are consistently reported in people with schizophrenia across a variety of paradigms, including Eriksen-type flanker, Go-No/Go, Stroop color-word naming and probabilistic learning tasks (e.g. Mathalon et al. Reference Mathalon, Whitfield, Gray, Faustman, Glover and Ford2002b , Reference Mathalon, Jorgensen, Roach and Ford2009; Morris et al. Reference Morris, Yee and Nuechterlein2006, Reference Morris, Heerey, Gold and Holroyd2008). A reduced ERN was also recently reported in children with putative antecedent features of schizophrenia (Laurens et al. Reference Laurens, Hodgins, Mould, West, Schoenberg, Murray and Taylor2010), implicating reduced internal error monitoring as a trait-like feature associated with liability to this disorder. fMRI studies provide converging evidence of diminished ACC responses to errors in schizophrenia (Carter et al. Reference Carter, MacDonald, Ross and Stenger2001; Polli et al. Reference Polli, Barton, Thakkar, Greve, Goff, Rauch and Manoach2008; Koch et al. Reference Koch, Schachtzabel, Wagner, Schikora, Schultz, Reichenbach, Sauer and Schlosser2010).
The ERN deficit in schizophrenia does not simply reflect general reductions in response accuracy or a generalized decrease in neural activation during response monitoring. ERN impairments are present regardless of whether patients differ from controls in accuracy rates, and patients demonstrate a normal or even an enhanced correct response negativity (CRN) (Alain et al. Reference Alain, McNeely, He, Christensen and West2002; Mathalon et al. Reference Mathalon, Fedor, Faustman, Gray, Askari and Ford2002a ; Morris et al. Reference Morris, Heerey, Gold and Holroyd2008), a corresponding but smaller ERP component 50 ms following correct responses. Furthermore, a later response-locked ERP component, the error-related positivity (Pe), is consistently unaffected in schizophrenia (Alain et al. Reference Alain, McNeely, He, Christensen and West2002; Mathalon et al. Reference Mathalon, Fedor, Faustman, Gray, Askari and Ford2002a ; Morris et al. Reference Morris, Heerey, Gold and Holroyd2008). The Pe is a positive deflection in the waveform at more posterior midline sites that typically peaks at around 300 ms and is larger following erroneous responses than the corresponding ERP that follows correct responses (the ‘Pc’). This component is hypothesized to index conscious evaluation or a P3-like response to infrequent errors of commission (Overbeek et al. Reference Overbeek, Nieuwenhuis and Ridderinkhof2005; van Veen & Carter, Reference van Veen and Carter2006; Ridderinkhof et al. Reference Ridderinkhof, Ramautar and Wijnen2009). Thus, the diminished differentiation between ERN and CRN in schizophrenia has been interpreted to reflect impaired early self-monitoring and internal error processing, whereas later response evaluation (i.e. differentiation between Pe and Pc) seems to be intact.
FN
A related ERP component, the FN, is sensitive to favorable versus unfavorable external feedback. This FN has been studied extensively using simple gambling or guessing paradigms (Simons, Reference Simons2010). The FN is apparent as a relative negativity at frontocentral recording sites approximately 300 ms following outcomes indicating relatively unfavorable outcomes, such as monetary loss or negative performance feedback, compared to a favorable outcome, such as monetary gains or positive performance feedback. It has been interpreted to reflect an early binary evaluation of outcomes as either favorable or unfavorable, and is insensitive to reward magnitude (Yeung & Sanfey, Reference Yeung and Sanfey2004; Sato et al. Reference Sato, Yasuda, Ohira, Miyawaki, Nishikawa, Kumano and Kuboki2005; Hajcak et al. Reference Hajcak, Moser, Holroyd and Simons2006). Importantly, the FN is typically elicited in tasks or trials in which subjects must rely on external feedback to evaluate the veracity of their responses. The FN and the ERN have often been described as reflecting common error monitoring processes subserved by the ACC (Holroyd & Coles, Reference Holroyd and Coles2002). However, recent evidence suggests important functional and source localization differences between these components. For instance, some studies indicate that the FN reflects reward-related activity of the striatum rather than error-related activity in the ACC (Holroyd et al. Reference Holroyd, Pakzad-Vaezi and Krigolson2008; Carlson et al. Reference Carlson, Foti, Mujica-Parodi, Harmon-Jones and Hajcak2011; Foti et al. Reference Foti, Weinberg, Dien and Hajcak2011). Indeed, a decreased FN is found in psychiatric conditions associated with altered reward sensitivity, such as depression and anxiety (Simons, Reference Simons2010). The FN may also be useful in delineating the scope of reward-related feedback sensitivity impairments in schizophrenia, but has thus far received only limited attention in this disorder.
The literature on reward sensitivity and feedback-based learning in schizophrenia provides a mixed picture. On the one hand, individuals with schizophrenia consistently show normal levels of self-reported pleasure and physiological responses to pleasant or rewarding evocative stimuli (Kring & Moran, Reference Kring and Moran2008; Horan et al. Reference Horan, Wynn, Kring, Simons and Green2010). On the other hand, although patients usually show intact performance on relatively simple reinforcement learning tasks, they show substantial impairment on more complex tasks involving implicit probabilistic habit learning, reversal learning, or value computation (Gold et al. Reference Gold, Waltz, Prentice, Morris and Heerey2008; Barch & Dowd, Reference Barch and Dowd2010). In the one prior study of the FN in schizophrenia, patients showed diminished differentiation between correct versus incorrect feedback in one condition of a complex probabilistic learning task that manipulated the validity of feedback information (Morris et al. Reference Morris, Heerey, Gold and Holroyd2008), which could result from deficient learning, reward insensitivity, or both. It remains to be determined whether individuals with schizophrenia show a diminished FN in simpler paradigms that do not require learning and integration of feedback under varying task conditions.
The current research
Two ERP studies were conducted to clarify the scope of feedback processing impairments in schizophrenia. In study 1, we compared the ERN of patients and healthy controls during a flanker task. We expected to replicate findings of a diminished ERN in schizophrenia. Study 2 considered the unexplored area of the FN during a simple monetary gambling task using larger samples of patients and controls, including all participants from study 1. Existing literature did not support a clear directional hypothesis for this task, although one prior study (Morris et al. Reference Morris, Heerey, Gold and Holroyd2008) pointed toward diminished FN in schizophrenia.
Method
Participants
Thirty-five out-patients with schizophrenia and 33 healthy control subjects participated in this research. A subset of 16 patients and 14 controls completed study 1 and all participants completed study 2. Patients met criteria for schizophrenia based on the Structured Clinical Interview for DSM-IV Axis I Disorders – Patient Edition (SCID-I/P; First et al. Reference First, Gibbon, Spitzer and Williams1996). Patients with schizo-affective disorder were excluded and none of the patients were in a major depressive or manic episode at the time of testing. Additional exclusion criteria for patients included: substance abuse or dependence in the past 6 months; IQ <70 based on chart reviews; a history of loss of consciousness for more than 1 h; an identifiable neurological disorder; or insufficient fluency in English. Regarding substance use history diagnoses: three patients had alcohol abuse, 11 had drug dependence, five had other substance abuse, and 14 had other substance dependence. All patients were medicated at clinically determined dosages, with 30 receiving atypical antipsychotic medications, three receiving typical antipsychotic medications, and two receiving both types of medication. Medication dosages were converted to chlorpromazine equivalents (Andreasen et al. Reference Andreasen, Pressler, Nopoulos, Miller and Ho2010) for supplemental analyses. All patients were clinically stable, which was defined as follows: no hospitalizations in the past 3 months, no medication changes in the past 6 weeks, and no changes in living status in the past 2 months.
Healthy controls were recruited through flyers posted in the local newspapers, websites, and posted advertisements. An initial screening interview excluded potential controls with identifiable neurological disorder or head injury, psychotic disorder in a first-degree relative, or insufficient fluency in English. Potential controls were then screened with the SCID-I/P and excluded for history of psychotic disorder, bipolar disorder, recurrent depression, lifetime history of substance dependence, or substance abuse in the past 6 months. Controls were also administered portions of the Structured Clinical Interview for DSM-IV Axis II Disorders (SCID-II; First et al. Reference First, Spitzer, Gibbon, Williams and Benjamin1994) and excluded if they had avoidant, paranoid, schizoid or schizotypal personality disorder.
All participants had the capacity to give informed consent and provided written informed consent in accordance with Institutional Review Board (IRB)-approved procedures.
Symptom ratings
For all patients, psychiatric symptoms during the previous month were rated using the expanded 24-item University of California, Los Angeles (UCLA) version of the Brief Psychiatric Rating Scale (BPRS; Overall & Gorham, Reference Overall and Gorham1962; Lukoff et al. Reference Lukoff, Nuechterlein and Ventura1986). Ratings from the positive and negative symptom subscales, and also total scores, were examined (Kopelowicz et al. Reference Kopelowicz, Ventura, Liberman and Mintz2008). Data from two patients were missing because of scheduling conflicts. All SCID and BPRS interviewers were trained through the Treatment Unit of the Department of Veterans Affairs VISN 22 Mental Illness Research, Education, and Clinical Center based on established procedures (Ventura et al. Reference Ventura, Green, Shaner and Liberman1993, Reference Ventura, Liberman, Green and Shaner1998). All interviewers had a masters or doctoral-level degree. The process included formal didactics, achieving a minimum level of reliability (minimum κ=0.75) for key psychotic and mood items using an extensive library of videotaped interviews, and also live, co-rated interviews conducted with faculty members. After certification, all raters participated in a continuous quality assurance program that involved periodic reliability checks and co-rated live interviews with faculty.
ERP paradigms
Study 1: ERN flanker task
An arrow version of the flanker task (Eriksen & Eriksen, Reference Eriksen and Eriksen1974) was administered following procedures used by Hajcak et al. (Reference Hajcak, Moser, Yeung and Simons2005) . On each trial, five horizontally aligned arrowheads were presented. Half of all trials were compatible (<<<<< or >>>>>) and half were incompatible (<<><< or >><>>); the order of compatible and incompatible trials was random. All stimuli were presented for 200 ms followed by an inter-trial interval (ITI) that varied randomly from 2300 to 2800 ms.
Participants were instructed to press the right mouse button if the center arrow was facing to the right and to press the left mouse button if the center arrow was facing to the left. Participants performed a practice block containing 30 trials during which they were instructed to be both as accurate and as fast as possible. The actual task consisted of 11 blocks of 30 trials (330 trials total) with each block initiated by the participant. To encourage both fast and accurate responding, participants received feedback based on their performance at the end of each block. If performance was 75% correct or lower, the message ‘Please try to be more accurate’ was displayed; performance above 90% correct was followed by ‘Please try to respond faster’; otherwise, the message ‘You're doing a great job’ was displayed.
Study 2: FN gambling task
To assess processing of feedback indicating good versus bad outcomes, a simple gambling paradigm was used (Foti & Hajcak, Reference Foti and Hajcak2010; Foti et al. Reference Foti, Weinberg, Dien and Hajcak2011). On each trial, participants were shown a graphic displaying two doors horizontally adjacent and were told to choose which door they wanted to open. They were told to press the left mouse button to choose the left door or the right mouse button to choose the right door. Following each choice, a feedback stimulus appeared on the screen informing the participants whether they had won or lost money on that trial. A green upward arrow (↑) indicated a correct guess and a gain of US$0.80 whereas a red downward arrow (↓) indicated an incorrect guess and a loss of US$0.40. A fixation mark (+) was presented prior to the onset of each stimulus. At the end of each trial, participants were presented with the instruction to ‘Click for the next round’. The order and timing of all stimuli were as follows: (i) the graphic of two doors was presented indefinitely until a response was made, (ii) a fixation mark was presented for 1000 ms, (iii) a feedback arrow was presented for 2000 ms, (iv) a fixation mark was presented for 1500 ms, and (v) ‘Click for the next round’ was presented until a response was made. Participants were told that they would gain US$0.80 each time they opened a door that hid a prize and lose US$0.40 each time they opened a door without a prize, and that they would earn between US$0 and US$20 total. In actuality, participants completed 50 trials with exactly 25 wins and 25 losses, for a net sum of US$10.00; feedback order was randomized across participants.
EEG recording and processing
Participants had their EEG activity continuously recorded in studies 1 and 2 using the same procedure. The EEG was recorded using a custom cap (Cortech Solutions, USA) and the ActiveTwo BioSemi system (BioSemi, The Netherlands). The signal was pre-amplified at the electrode with a gain of one; the EEG was digitized at 24-bit resolution with a sampling rate of 512 Hz using a low-pass fifth-order sinc filter with a half-power cut-off of 104 Hz. Recordings were taken from 64 scalp electrodes based on the 10/20 system, and from two electrodes placed on the left and right mastoids. The electro-oculogram was recorded from four facial electrodes: two 1 cm above and below the left eye, one 1 cm to the left of the left eye, and one 1 cm to the right of the right eye. Each electrode was measured online with respect to a common mode sense electrode that formed a monopolar channel.
Off-line analysis was performed using Brain Vision Analyzer software (Brain Products, Germany). All EEG data were re-referenced to the average of the two mastoids. Filtering, segmenting and averaging parameters for each task are described separately below. For both tasks, each trial was corrected for blinks and eye movements using the method developed by Gratton et al. (Reference Gratton, Coles and Donchin1983) . Specific channels were rejected in each trial using a semi-automated procedure, with physiological artifacts identified by the following criteria: a step of more than 50 μV between sample points, a difference of 300 μV within a trial, and a maximum difference of less than 0.5 μV within 100-ms intervals. Additional physiological artifacts were identified using visual inspection.
For the flanker task, the data were band-pass filtered with cut-offs of 0.1 and 30 Hz. The EEG was segmented for each trial beginning 400 ms before each response onset and continuing for 1000 ms (i.e. until 600 ms after response onset). Response-locked ERPs were averaged separately for error and correct trials. The ERN was evaluated as the average activity on error trials from response onset to 100 ms (i.e. 0–100 ms) at pooling of FCz/Cz (where the effect was largest) and the CRN was evaluated in the same time window and electrodes on correct trials. The Pe and Pc were evaluated on error and correct trials respectively, as the average activity from 200 to 300 ms at Cz following response onset. A 200 ms window from 400 to 200 ms before response onset served as the baseline. ERP activity on correct trials has been associated with response monitoring (Simons, Reference Simons2010). Moreover, error-related brain activity may reflect processes common to both error and correct responses. Accordingly, it is particularly informative to examine the difference between error and correct trials so as to separate activity that is uniquely related to error processing from activity related to response monitoring in general (Burle et al. Reference Burle, Roger, Allain, Vidal and Hasbroucq2008). The difference wave approach can help to isolate ERP components that are more readily interpreted in terms of specific cognitive functions (Luck, Reference Luck2005). Difference scores for error minus correct trials were therefore calculated in the time windows of the ERN/CRN, in addition to the Pe/Pc; we refer to these as the ΔERN and ΔPe. Based on the literature (Olvet & Hajcak, Reference Olvet and Hajcak2009), participants who made less than six errors were excluded from all analyses (schizophrenia=3; controls=4), resulting in final sample sizes of 16 patients and 14 controls.
For the gambling task, the EEG data were band-pass filtered with cut-offs of 0.1 and 30 Hz. The EEG was segmented for each trial, beginning 200 ms before feedback onset and continuing for 800 ms following feedback onset. Stimulus-locked responses were averaged separately for non-rewards and rewards, and the activity in the 200-ms window before feedback onset served as the baseline. The FN was quantified as mean activity from 250 to 350 ms at a pooling of FCz/Cz for non-reward and reward trials, and also the difference between non-reward and reward trials (ΔFN). A difference wave approach is particularly relevant in studies of the FN, insofar as non-reward and reward are thought to elicit phasic decreases and increases in dopamine respectively (Holroyd & Coles, Reference Holroyd and Coles2002). One outlier in the control group (ERPs >3 standard deviations above the group mean) was excluded from all analyses for this task, resulting in final sample sizes of 35 patients and 32 controls.
Results
Sample characteristics
As shown in Table 1, the groups did not differ significantly in sex, age, ethnicity or marital status. The patients had lower personal education levels than controls but the groups did not differ in parental education, which was the variable intended to control for family socio-economic status, as opposed to personal education, which can be influenced by the illness itself. Patients in the schizophrenia group had a typical age of onset, were chronically ill, and showed mild to moderate levels of clinical symptoms at the time of testing. For antipsychotic medications, patients were taking an average of 305.50 chlorpromazine equivalents.
Table 1. Demographic and clinical data
BPRS, Brief Psychiatric Rating Scale; s.d., standard deviation.
* p<0.001.
There were no significant differences on any demographic or clinical variables (all p's >0.05) between the subgroups of schizophrenia (n=16) and control (n=14) participants included in study 1 versus those participants not in study 1.
Study 1: Flanker task
Behavioral data
Behavioral measures included both the number of error trials for each subject and the accuracy expressed as a percentage. Average RTs on error and correct trials were also calculated separately. To reduce the influence of outliers, trials were removed from the analysis of RTs that were faster than 200 ms or slower than 1200 ms. The number (s.d.) of trials removed for the schizophrenia [14.13 (27.49)] and control [3.50 (10.21)] groups did not differ significantly [t 28=0.18, p>0.05].
Accuracy and RT data are presented in Table 2. An independent-samples t test indicated that the schizophrenia and control groups made a comparable number of errors (t 28=−0.02, p>0.05) and had a comparable percentage correct [t 28=−0.01, p>0.05]. For RT, a 2 (trial type)×2 (group) mixed-model ANOVA revealed a significant trial type effect (F 1,28=24.44, p<0.001, ηp 2=0.446), indicating that participants were faster on error than on correct trials. There was also a significant group effect (F 1,28=10.51, p<0.005, ηp 2=0.273), reflecting the typical finding of generally slower RTs in schizophrenia, but the trial type×group interaction was not significant (F 1,28=0.05, p>0.05, ηp 2=0.002). Thus, the groups showed similar accuracy levels and patterns of RT differences across trial types.
Table 2. Behavioral and event-related potential (ERP) data for the flanker task
ERN, Error-related negativity; CRN, correct response negativity; Pe, error-related positivity; Pc, correct response positivity.
Values given as mean (standard deviation).
ERPs
For the ERN, grand average response-locked ERPs are presented in Fig. 1 and average ERP values are presented in Table 2. A 2 (trial type)×2 (group) mixed-model ANOVA revealed non-significant effects for trial type (F 1,28=1.66, p>0.05, ηp 2=0.056) and group (F 1,28=0.43, p>0.05, ηp 2=0.179). However, a significant trial type×group interaction indicated that the difference between the ERN and the CRN was smaller in the schizophrenia group than in the control group (F 1,28=6.10, p<0.05, ηp 2=0.015). A post-hoc interaction contrast comparing error minus correct trials (i.e. ΔERN) in the two groups confirmed that patients showed less discrimination between the ERN and the CRN than controls (t 28=2.47, p<0.05, d=0.90).
Fig. 1. Flanker task response-locked event-related potentials (ERPs) at FCz/Cz for error (ERN) and correct (CRN) trials for (a) control and (b) schizophrenia groups, and also the difference. Response onset occurred at 0 ms and negative is plotted up.
For the Pe, grand average response-locked ERPs are presented in Fig. 2 and average ERP values are presented in Table 2. A 2 (trial type)×2 (group) mixed-model ANOVA revealed a significant trial type effect (F 1,28=39.25, p<0.001, ηp 2=0.584), indicating that the Pe was significantly more positive than the Pc. There were no significant effects for group (F 1,28=0.38, p>0.05, ηp 2=0.013) or the trial type×group interaction (F 1,28=1.33, p>0.05, ηp 2=0.045). Consistent with these results, the ΔPe between correct versus error trials did not significantly differ across groups (t 28=1.16, p>0.05, d=0.43). In summary, the ERP data for the flanker task indicated a smaller ΔERN in the schizophrenia than the control group, but a comparable ΔPe for both groups.
Fig. 2. Flanker task response-locked event-related potentials (ERPs) at Cz for error (Pe) and correct (Pc) trials for (a) control and (b) schizophrenia groups, and also the difference. Response onset occurred at 0 ms and negative is plotted up.
Study 2: Gambling task
For the FN, grand average response-locked ERPs are presented in Fig. 3 and average ERP values are presented in Table 3. A 2 (trial type)×2 (group) mixed-model ANOVA revealed a significant trial type effect, indicating that the FN was significantly more negative for non-reward than for reward trials across groups (F 1,65=32.57, p<0.001, ηp 2=0.334). However, both the effect of group (F 1,65=2.47, p>0.05, ηp 2=0.037) and the group×trial type interaction (F 1,65=0.03, p>0.05, ηp 2=0.001) did not reach significance, indicating that the difference in the FN for reward versus non-reward trials was comparable across groups. Consistent with these results, the ΔFN between reward versus non-reward trials did not differ between groups (t 65=0.19, p>0.05, d=0.05).Footnote 1,2,3 Footnote †
Fig. 3. Gambling task feedback-locked feedback negativity (FN) at FCz/Cz for non-reward and reward trials for (a) control and (b) schizophrenia groups, and also the difference. Feedback onset occurred at 0 ms and negative is plotted up.
Table 3. Event-related potential (ERP) data for the gambling task
FN, Feedback negativity.
Values given as mean (standard deviation).
Supplemental analyses
Exploratory analyses examined Spearman rank-order correlations between the ΔERN, ΔPe and ΔFN difference wave scores and BPRS positive, negative and total symptoms, and also chlorpromazine equivalents, within the schizophrenia group. We wanted to determine whether the ΔFN during the gambling task related to individual differences in negative symptoms. Of the nine correlations computed, only one was marginally significant and the direction of the correlation was counterintuitive. For the gambling task, higher positive symptoms were associated with more negative ΔFN scores (i.e. greater FN differentiation between reward versus non-reward trials) (r=−0.35, p=0.05). There were no significant or trend-level correlations for negative symptoms. Finally, there were no significant or trend-level correlations between chlorpromazine equivalents and any of the ERP variables (all r's<0.20, p's>0.10).
Discussion
Individuals with schizophrenia showed a reduced ΔERN accompanied by an intact ΔPe, indicating deficient early error monitoring. By contrast, sensitivity to external feedback during a simple gambling task was intact; the ΔFN significantly differentiated feedback indicating monetary reward from non-reward to a comparable degree in patients and controls. It has been suggested that the ΔERN and ΔFN reflect common activity of an error monitoring system that is sensitive to internal and external feedback respectively (Holroyd & Coles, Reference Holroyd and Coles2002). Within this context, the differential pattern for the ΔERN and ΔFN suggests that error monitoring is not universally impaired in schizophrenia, and that the processing of external feedback may be relatively unaffected. Taken together, these findings help to clarify components of error monitoring and reward processing that are differentially impaired and intact in schizophrenia.
The intact sensitivity to external reward-related feedback (ΔFN) demonstrated by the schizophrenia group is broadly consistent with considerable evidence of normal self-reported and physiological responses to pleasant or rewarding stimuli in this population (Kring & Moran, Reference Kring and Moran2008; Horan et al. Reference Horan, Wynn, Kring, Simons and Green2010). Thus, there is converging evidence that, at a basic level, sensitivity to external reward feedback and pleasurable stimuli is essentially intact in schizophrenia. Of note, a similar pattern of normal FN accompanied by impaired ERN was recently reported in adolescents and young adults with autism (Larson et al. Reference Larson, South, Krauskopf, Clawson and Crowley2011), a neurodevelopmental disorder that shows significant behavioral, neural and genetic overlap with schizophrenia (Nylander et al. Reference Nylander, Lugnegard and Hallerback2008; Burbach & van der Zwaag, Reference Burbach and van der Zwaag2009). However, the feedback sensitivity profile shown by our patients differs from some other psychiatric conditions; people with obsessive–compulsive and generalized anxiety disorders show enhanced ΔERN and diminished ΔFN (Simons, Reference Simons2010), and diminished ΔFN is also associated with depressive symptoms (Foti & Hajcak, Reference Foti and Hajcak2009).
Although schizophrenia patients generally show intact sensitivity to simple external reward feedback, the translation of reward information into adaptive, goal-directed behavior involves coordinated activity among several additional reward processing subcomponents. As reviewed by Barch & Dowd (Reference Barch and Dowd2010), in the context of intact hedonic or ‘liking’ responses to rewarding stimuli, people with schizophrenia may experience difficulties integrating reward information in the context of learning, anticipation and/or decision making to guide current and future behavior. This framework may help to account for a previous report of impaired FN in schizophrenia. Morris et al. (Reference Morris, Heerey, Gold and Holroyd2008) examined FN during a complex probabilistic reward learning task, in which participants were required to learn the correct responses associated with a range of stimuli and were rewarded for accurate performance. Among controls, FN amplitude decreased as stimulus–response pairings were learned, whereas among patients this effect of learning on the FN was attenuated (see Koch et al. Reference Koch, Schachtzabel, Wagner, Schikora, Schultz, Reichenbach, Sauer and Schlosser2010 for comparable findings in an fMRI probabilistic learning task). Differences in the FN results between the current study and that of Morris et al. (Reference Morris, Heerey, Gold and Holroyd2008) may reflect differences in how rewards were delivered and incorporated in the paradigms used in these studies. The current study used a simple gambling task in which reward delivery was random, with no additional learning or performance demands. By contrast, rewards were contingent on effective learning and accurate performance under varying reinforcement conditions in Morris et al. (Reference Morris, Heerey, Gold and Holroyd2008) . Although schizophrenia patients may show intact basic reward liking on tasks with minimal integrative processing demands, impairments may emerge in the context of higher-level reward learning tasks that involve more complex reinforcement contingencies or value computations.
The schizophrenia patients' differential pattern of performance adds to growing evidence that the ERN and the FN do not necessarily reflect functionally identical neural activity related to a general error-detection network (e.g. Hajcak et al. Reference Hajcak, Moser, Yeung and Simons2005, Reference Hajcak, Moser, Holroyd and Simons2006). Instead, evidence consistently indicates that the ERN reflects error monitoring processes subserved by the ACC (Holroyd & Coles, Reference Holroyd and Coles2002). By contrast, recent evidence suggests that the FN reflects increased neural activity to favorable outcomes, a reward-related positivity, and that this response is generated in the striatum (Holroyd et al. Reference Holroyd, Pakzad-Vaezi and Krigolson2008; Carlson et al. Reference Carlson, Foti, Mujica-Parodi, Harmon-Jones and Hajcak2011; Foti et al. Reference Foti, Weinberg, Dien and Hajcak2011). This conceptualization of the FN is supported by functional differences between extensive animal and human research linking reward processing to the striatum and the medial prefrontal cortex (MPFC; see Foti et al. Reference Foti, Weinberg, Dien and Hajcak2011). The emerging distinction between these ERP components suggests a pattern in which schizophrenia is characterized by impaired error-related activity in the ACC, but intact reward-related activity in the striatum and MPFC, at least when rewards are presented randomly without additional integrative processing demands. However, differences in the complexity of the ERN and FN tasks used in this study should also be considered when interpreting these findings. For example, the ERN task requires a representation of the actual and intended response whereas the FN task does not require any representation of the response. It could be argued that errors/unfavorable outcomes are simply more obvious and easier for patients to detect in the FN task than the ERN task, an explanation that does not require group differences in distinct neural circuits. Studies that combine ERP and fMRI (e.g. Mathalon et al. Reference Mathalon, Jorgensen, Roach and Ford2009) can address this issue directly.
The current study should be interpreted in the context of some limitations. First, patients were taking antipsychotic medications and their effects on feedback and reward processing are uncertain. It is possible that the patients' normal FN reflects medication benefits, as the majority of patients were taking atypical antipsychotics, which have been found to improve some components of reward processing (Juckel et al. Reference Juckel, Schlagenhauf, Koslowski, Filonov, Wustenberg, Villringer, Knutson, Kienast, Gallinat, Wrase and Heinz2006a , Reference Juckel, Schlagenhauf, Koslowski, Wustenberg, Villringer, Knutson, Wrase and Heinz b ; Schlagenhauf et al. Reference Schlagenhauf, Juckel, Koslowski, Kahnt, Knutson, Dembler, Kienast, Gallinat, Wrase and Heinz2008). Alternatively, the impaired ERN could be a consequence of prolonged exposure to antipsychotics, although evidence that ERN impairment is detectable in children with putative antecedent features of schizophrenia (Laurens et al. Reference Laurens, Hodgins, Mould, West, Schoenberg, Murray and Taylor2010) argues against this possibility. If medications did impact performance, they did not have a uniform effect across tasks. Second, our use of the BPRS may have limited our ability to detect an association between negative symptoms and reward processing; the BPRS negative symptom subscale focuses on expressive symptoms (e.g. blunted affect) whereas experience-related symptoms (e.g. avolition, asociality) have a stronger theoretical link to feedback and reward processing (Blanchard et al. Reference Blanchard, Kring, Horan and Gur2011). Third, the patients were predominantly male and chronically ill, potentially limiting generalizability. Further efforts to specify impaired and intact reward processing subcomponents may help to shed light on the underlying causes of the diminished motivation and goal-seeking seen clinically in many people with schizophrenia.
Acknowledgements
Funding for this project came from a NARSAD Young Investigator Award (to W.P.H.) and NIMH grants MH043292 and MH065707 (to M.F.G.)
Declaration of Interest
None.
