Reduced auditory evoked gamma band response and cognitive processing deficits in first episode schizophrenia.

Abstract Objectives. Gamma-band oscillations (e.g., the early auditory evoked gamma-band response, aeGBR) have been suggested to mediate cognitive and perceptual processes by driving the synchronization of local neuronal populations. Reduced aeGBR is a consistent finding in patients with schizophrenia and high-risk subjects, and has been proposed to represent an endophenotype for the illness. However, it is still unclear whether this reduction represents a deficit in sensory or cognitive processes, or a combination of the two. The present study investigated this question by manipulating the difficulty of an auditory reaction task in patients with first-episode schizophrenia and healthy controls. Methods. A 64-channel EEG was recorded in 23 patients with first-episode schizophrenia and 22 healthy controls during two conditions of an auditory reaction task: an easy condition that merely required low-level vigilance, and a difficult condition that placed significant demands on attention and working memory. Results. In contrast to healthy controls, patients failed to increase aeGBR power and phase-locking in the difficult condition. In patients, aeGBR power and phase-locking indices were associated with working memory deficits. Conclusions. The observed results confirm the applicability of aeGBR disturbances as a stable endophenotype of schizophrenia, and suggest a cognitive, rather than sensory, deficit at their origin.


Introduction
Current pathophysiological theories assume that the cognitive defi cits appearing in schizophrenia are a correlate of the disturbed coordination of distributed processes involving multiple brain areas (Friston 1999;Phillips and Silverstein 2003). Abnormalities in the synchronization of neural oscillatory activity are an important mechanism, through which this disturbed coordination is thought to occur (Uhlhaas and Singer 2010). In this context, high-frequency activity in the gamma-band range (30 -100 Hz) has received increasing attention, for two reasons: fi rst, gamma-band oscillations have been suggested to mediate cognitive and perceptual processes by driving the synchronization of local neuronal populations (Singer 1999;Engel et al. 2001;Canolty et al. 2006;Fries et al. 2007). Second, their generation depends on a closed microcircuit involving parvalbuminpositive GABAergic interneurons and glutamatergic pyramidal cells (Bartos et al. 2007;Sohal et al. 2009), which is disrupted in patients with schizophrenia and in pharmacological or genetic models of the illness (Gandal et al. 2012;Uhlhaas and Singer 2013).
The largest body of research on gamma-band oscillations in schizophrenia concerns evoked gamma-band responses associated with sensory stimuli. Patients with schizophrenia have been reported to display abnormalities in visually evoked gamma response (Spencer et al. 2003(Spencer et al. , 2004, or in visually evoked steady state potentials (Krishnan et al. 2005), indicating disturbed visual feature-binding processes. In the auditory domain, alterations have been detected during auditory oddball paradigms (Haig et al. 2000;Lee et al. 2001;Gallinat et al. 2004; patients with schizophrenia and controls have been reported both by investigating aeGBR to both non-target (standard) (Roach and Mathalon 2008;Taylor et al. 2013) and target (deviant) tones ). However, attentional and workingmemory mechanisms are presumably relevant for both of these stimulus types, as all incoming tones need to be attended to and matched against a template for the task to be performed correctly. The present study aimed to elucidate the differential contributions of sensory and cognitive processes to aeGBR defi cits in patients with fi rst-episode schizophrenia. This was achieved by manipulating the diffi culty of an auditory reaction task. Two conditions were applied, an easy condition that merely required a motoric reaction whenever a sensory stimulus was perceived, and a diffi cult condition consisting in a three-tone oddball paradigm that placed signifi cant demands on attention and working memory. On one hand, if reduced aeGBR in patients results from sensory processing defi cits, then it should be apparent in both conditions; on the other hand, if cognitive factors (also) contribute to reduced aeGBR, then the effect should (additionally) be more pronounced in the diffi cult than in the easy condition.

Ethics statement
The present study was part of a larger project investigating resting-state and task-related brain connectivity in schizophrenia by means of EEG, MEG, and simultaneous EEG-fMRI, within the context of the Collaborative Research Centre 936 ( " multi-site communication in the brain " , www.sfb936.net). The study was approved by the Ethics Committee of the Medical Association Hamburg the investigation was and carried out in accordance with the latest version of the Declaration of Helsinki. Written informed consent was obtained from all participants after the nature of the procedures had been fully explained.

Participants
Twenty-three patients with a fi rst-episode of schizophrenia and 22 healthy controls participated in the study. First-episode status was defi ned as having received the fi rst diagnosis and psychiatric treatment less than a year prior to study participation, and presence of psychotic symptoms in any form for no more than 5 years. Patients were recruited through the Psychosis Center of the Department of Psychiatry of the University Medical Center Hamburg-Eppendorf. Diagnosis of schizophrenia in patients was established with the Mini International Neuropsychiatric Symond et al. 2005;Roach and Mathalon 2008;Spencer et al. 2008a) and during auditory steadystate-stimulation (Kwon et al. 1999;Light et al. 2006;Spencer et al. 2008b;Teale et al. 2008;Vierling-Claassen et al. 2008;Wilson et al. 2008) in the gamma frequency range.
A particularly interesting feature of evoked gamma-band responses is that they do not only refl ect sensory processes, but are also affected by cognitive functions such as attention or memory (Gurtubay et al. 2004;Tallon-Baudry et al. 2005;Cho et al. 2006;Herrmann et al. 2010). These top-down infl uences on even early stages of information processing are exemplifi ed in the case of the early auditory evoked gamma-band response (aeGBR), which appears 25 -100 ms upon the presentation of an auditory stimulus (Tiitinen et al. 1993). Although the aeGBR originates in the primary auditory cortex (Pantev et al. 1991) and is modulated by sensory properties of the stimulus (Schadow et al. 2007), it also has an undisputable cognitive component, as its magnitude is strongly infl uenced by task diffi culty (Mulert et al. 2007;Herrmann et al. 2010), and more specifi cally by attentional (Tiitinen et al. 1997) and (working-) memory processes (Herrmann et al. 2010). Indeed, an additional aeGBR generator has been localized in the medial prefrontal cortex and the dorsal anterior cingulate cortex (dACC), using both EEG-based source localization (Mulert et al. 2007;Leicht et al. 2010) and single trial EEG-fMRI coupling (Mulert et al. 2010).
Similarly to other evoked gamma-band responses, the aeGBR is reduced in schizophrenia. This defi cit has been reliably replicated across all stages of the illness: fi rst-episode (Taylor et al. 2013) and chronic patients (Roach and Mathalon 2008;Leicht et al. 2010), high-risk subjects (Perez et al. 2013), and symptom-free fi rst-degree relatives of patients (Hall et al. 2011a;Leicht et al. 2011). This feature, in conjunction with their association with a biologically plausible mechanism of the illness (Tsuang et al. 1993), has established reduced aeGBR as a probable endophenotype for schizophrenia which could be of potential interest for future studies investigating glutamatergic treatment strategies, genetic mechanisms, or the prediction of transition to psychosis in high-risk individuals. However, it is still unclear whether this reduction represents a defi cit in sensory (bottom-up), or cognitive (topdown) processes, or a combination of the two. Previous fi ndings have not provided suffi cient evidence regarding this point. Typically, the aeGBR is investigated in the context of auditory oddball tasks, in which subjects are asked to respond by button press only to non-frequent, deviant tones intermingled in a sequence of standard tones. Differences between literature, i.e., working memory and attention tasks (see Introduction), were included in analyses. This was done in order to minimize the number of variables included in correlational analyses reported below, and thereby reduce the risk of Type I errors.
The majority of patients were on antipsychotic medications at the time of EEG recording (atypical antipsychotics: n ϭ 18; typical antipsychotics: n ϭ 1; no psychotropic medication: n ϭ 2). Moreover, seven patients were currently in treatment with antidepressants (either escitalopram or venlafaxine). No subjects were receiving benzodiazepines or anticholinergic agents. Demographic characteristics of the two groups, and clinical characteristics of patients, are presented in Table I. The groups were matched with respect to age, sex and educational level. All subjects had hearing better than 30 dB at a pitch of 1000 Hz.

Paradigm
We used two different diffi culty levels of an auditory reaction task (Mulert et al. 2001) that had been earlier shown to increase aeGBR amplitude according to the level of diffi culty (Mulert et al. 2007). Accordingly, the experiment consisted of two different runs during which tones varying in pitch (duration: 250 ms, generated using the Presentation software version 16.1) were presented via earphones at 85 dB SPL with pseudo-randomized interstimulus intervals (ISI: 2.5 -7.5 s; mean 5.0 s). In the easy condition (EC), 80 tones at a pitch of 800 Hz had to be responded to per button-press with the left index fi nger. In the diffi cult condition (DC), 120 tones of different pitch (33% 800 Hz, 33% 1000 Hz and 33% 1200 Hz) had to be differentially responded to, by pressing a button with the left index fi nger following the low tone and with the right index fi nger following the high tone. No response was required for the 1000 Hz (middle) tone. Prior to each run, subjects were instructed to respond as fast and accurately as possible. Before the beginning of the measurement, a short test run was carried out. Reaction times (from stimulus onset until button press) and errors (incorrect response or no response within 2000 ms after stimulus presentation) were registered during the experimental run.

EEG recording
Recording took place in a sound-attenuated and electrically shielded room. Subjects were seated with their eyes open in a slightly reclined chair with a head rest and were asked to keep the eyes open and look at a fi xation cross presented at a 19" computer monitor 1 m in front of them. The EEG was recorded at a sampling rate of 1000 Hz and an analog band-pass Interview (Sheehan et al. 1998). Two patients and one healthy control had to be excluded from further analyses due to poor EEG data quality resulting in an insuffi cient number of trials suitable for analysis.
Exclusion criteria for all participants were current substance abuse or dependence, and presence of major somatic or neurological disorders. For healthy control subjects, additional exclusion criteria were any previous psychiatric disorder or treatment, and a family history of psychotic disorders. The presence of inclusion/exclusion criteria was assessed by means of a semi-structured interview conducted by a clinical psychiatrist or trainee with at least 4 years of clinical experience. Healthy controls were recruited from the community through advertisement and word-of-mouth.
The severity of clinical symptomatology was assessed with the Positive and Negative Syndrome Scale (PANSS; Kay et al. 1987); subscores for positive, negative, disorganization, excitement and distress symptoms were created according to a fi vefactor model of the PANSS (van der Gaag et al. 2006). Because participation in the original project included three to fi ve neurophysiological testing sessions (EEG, MEG and simultaneous EEG-fMRI), it was not always possible to conduct clinical assessments close to the EEG session. Therefore, based on reported trajectories of antipsychotic treatment response Stauffer et al. 2011), clinical severity ratings were used for analyses only if they were separated from EEG analyses by no more than a week for acutely ill patients, or 2 months for stable patients (i.e., clinical stability as per medical record and no change in medication for at least 2 months prior to study participation). Thus, appropriate clinical ratings were available for 19 patients.
Participants of both groups also underwent neuropsychological testing with an extensive battery that included tests of: memory [Logical Memory subtests from the Wechsler Memory scale, revised (Wechsler 1987); Verbal Learning and Memory Test (Helmstaedter et al. 2001)]; attention [Digit span forward and Digit-Symbol-Coding of the Wechsler Adult Intelligence Scale-III (Wechsler 1997)]; working memory [Digit span backward and Letter-Number Sequencing of the Wechsler Adult Intelligence Scale (Wechsler 1997)]; visuomotor sequencing [Trail Making Test Parts A and B (Reitan and Wolfson 1985)]; letter fl uency (Aschenbrenner et al. 2001). Neurocognitive performance data were available for 20 healthy controls and 17 patients. In all patients, neurocognitive data were collected within a week or within 2 months from EEG for acutely ill and stable patients, respectively. Given the relatively small sample size, only those tasks, for which an association with aeGBR indices could be expected based on previous window starting 210-ms pre-stimulus in any channel were automatically rejected. After re-referencing to common average reference and baseline correction (using an interval of 210 -10 ms pre-stimulus), averaged event-related potential (ERP) wave-shapes were computed. Only wave-shapes based on at least 35 segments were accepted.

Evoked gamma power and PLF
Using the BVA Software, evoked gamma power and phase-locking factor (PLF) were computed using a wavelet transformation [complex Morlet wavelet with the formula w ( t ) ϭ A exp ( -t ² /2) exp ( i 2 π ct ), Morlet parameter c ϭ 5, Instantaneous Amplitude (Gabor) Normalization], as used previously by other groups (Herrmann et al. 1999;Senkowski and Herrmann 2002) and our group (Mulert et al. 2007).
In order to reveal the phase-locked evoked gamma power, wavelet transformation was performed on averaged ERP wave-shapes. Layer-wise baseline correction was applied using a timeframe of 200 ms starting 210 ms prior to stimulus presentation. The frequency range from 20 to 80 Hz was divided into 30 frequency steps (distributed on a logarithmic scale) for each subject. For aeGBR peak detection, the wavelet layer with the central frequency of 40 Hz (frequency range 32 -48 Hz) was extracted. Based on prior knowledge (Mulert et al. 2007(Mulert et al. , 2010Leicht et al. 2010) the aeGBR-peak was defi ned as the highest value within the timeframe 30 -100 ms poststimulus at the electrode Cz.
PLFs were calculated by performing wavelet transformation (without layer-wise baseline correction) and extracting complex-phase information with all fi lter (0.1 -1000 Hz) with 66 active electrodes mounted on an elastic cap (ActiCaps, Brain Products, Munich, Germany) using the Brain Vision Recorder software Version 1.10 (Brain Products, Munich, Germany). Electrodes were arranged according to a modifi ed 10/10 system without electrodes at the positions FPz, F9, F10, T9, T10, CP3, CP4, P9, P10, PO7, PO8 and with additional electrodes at positions PO9 and PO10. Eye movements were recorded through four EOG channels (positioned at the outer canthi bilaterally and infra-and supraorbitally on the right). An electrode at the FCz position was used as the reference, the electrode at position AFz served as ground. Impedances were always kept below 5 k Ω .

EEG pre-processing
Data analysis was carried out using Brain Vision Analyzer (BVA) Version 2.0 (Brain Products, Munich, Germany). The channels PO9 and PO10 were excluded from further analysis due to persistent muscle artefact contamination in most subjects. After band-pass fi ltering (1 -100 Hz), topographic interpolation (spherical splines) of up to six channels was performed (mean number of interpolated channels: EC 0.52 Ϯ 1.31; DC 0.60 Ϯ 1.23; no signifi cant differences between groups or conditions). Channels were selected for interpolation if more than 5% of data in the respective channel was affected by technical artifacts or muscle artifacts exceeding amplitudes of Ϯ 70 μ V. The continuous EEG was segmented into epochs of 1400-ms starting 400 ms prior to the auditory stimulus. Segments including incorrect responses or amplitudes exceeding Ϯ 70 μ V within a 410-ms applied to calculate robust confi dence intervals of correlation coeffi cients. Bonferroni correction was applied to correct for multiple comparisons; however, due to the exploratory nature of these analyses, uncorrected results are also reported.

Behavioural performance
With regard to reaction times, there was a signifi cant main effect of group ( F ϭ 9.56, P ϭ 0.003, η 2 partial ϭ 0.199), while the condition ϫ group interaction did not achieve signifi cance ( F ϭ 2.9, P ϭ 0.098, η 2 partial ϭ 0.067). Follow-up t -tests revealed signifi cantly longer reaction times in SZ compared to HC in both conditions, and in DC compared to EC in both groups (see Table II) -although the latter effect was tendentially more pronounced in patients.
With regard to error rates, the ANOVA revealed a signifi cant condition ϫ group effect ( F ϭ 5.9, P ϭ 0.019, η 2 partial ϭ 0.129). Follow-up t -tests revealed significantly higher error rates in SZ compared to HC in DC but not in EC, and signifi cantly higher error rates in DC compared to EC in both groups (see Table II).

Evoked GBR power and PLF
Around 50 ms after stimulus presentation in both groups, an increase of evoked gamma activity was observed (see Figure 1) at electrode Cz. For scalp topographies please see Supplementary Figure 1 (Supplementary material to be found online at http:// informahealthcare.com/doi/abs/10.3109/15622975. 2015.1017605). With respect to the peaks of this evoked GBR power, a signifi cant condition ϫ group interaction vector lengths normalized to the unit circle ( " Phase locking factor " and " Complex Data Measures " solution, BVA software) before averaging of the phase information. Gamma PLF peaks were defi ned as the highest value of the wavelet layer centred around 40 Hz (same as above) within the timeframe 30 -100 ms post-stimulus at the electrode Cz.

Statistical analyses
All statistical analyses were performed using the SPSS software package (21.0). In order to describe signifi cant group ϫ condition effects, repeatedmeasures ANOVAs with condition as the withinsubject factor and group as the between-subject factor were conducted on the variables of interest. Signifi cant results were followed-up by exploratory pairwise t -tests, using paired-tests to assess differences between conditions, and t -tests for independent samples to check for signifi cant group differences; because of the small sample size, bootstrapping was performed in t -test analyses to adjust signifi cance values based on estimates of the properties of the sampling distribution. For the comparison of the two groups on educational level and gender differences, the chi-square test was used. For ANOVAs, the partial eta squared ( η 2 partial ) is provided as an estimate of effect size (small: 0.01 -0.06; moderate: 0.06 -0.14; large: Ͼ 0.14); for t -tests, Pearson ' s r is provided (small: 0.1 -0.3; moderate: 0.3 -0.5; large: Ͼ 0.5).
In patients, exploratory correlational analyses (Spearman ' s rho) were conducted between electrophysiological parameters on one hand and severity of clinical symptomatology, attention/working memory performance and medication dose (in chlorpromazine equivalents) on the other. Bootstrapping was

Correlations with clinical and neuropsychological variables in the patient group
In the easy condition, there were negative correlations between PLF (rho ϭ 0.510, P ϭ 0.05, CI ϭ -0.035 -0.877) and evoked power (rho ϭ 0.600, P ϭ 0.02, CI ϭ 0.0186 -0.876) and the negative factor score, i.e., PLF and evoked power were lower in patients with higher negative symptom load. Moreover, a trend-wise signifi cant correlation emerged between aeGBR evoked power and disorganized factor scores (rho ϭ 0.504, P ϭ 0.06, CI ϭ 0.004 -0.835). These correlations were not signifi cant after Bonferroni correction.
No signifi cant correlations were noted between chlorpromazine equivalents and either evoked aeGBR power or gamma PLF (all P Ͼ 0.2). emerged ( F ϭ 4.3, P ϭ 0.045, η 2 partial ϭ 0.096). t -Tests revealed signifi cantly diminished aeGBR power peaks in SZ compared to HC in DC but not in EC, and marginally increased aeGBR power peaks in DC compared to EC in HC but not in SZ (see Table III). There was no signifi cant condition ϫ group effect concerning the latency of the aeGBR peaks at Cz (mean latencies: HC-DC: 73.7 ms, HC-EC: 74.4 ms, SZ-DC: 69.1 ms, SZ-EC: 69.6 ms).
Regarding the PLF, we observed a signifi cant condition ϫ group interaction ( F ϭ 8.46, P ϭ 0.006, η 2 partial ϭ 0.175). Compared to HC, SZ showed signifi cantly reduced PLF values in DC but not in EC. Follow-up t -tests revealed a signifi cant increase of PLF values in HC during DC compared to EC; this effect was not present in SZ (see Figure 2).  (DC, upper row) and the easy condition (EC, lower row). Scaling was uniform for both groups. The auditory evoked gamma-band response (aeGBR) can be seen as an increased activity at about 50 ms after stimulus presentation (dashed line) and in the frequency range around 40 Hz. In contrast to SZ, healthy subjects showed a signifi cant increase of the aeGBR power in DC compared to EC. studies that have suggested a role of the aeGBR as an endophenotype for the illness (Hall et al. 2011b;Leicht et al. 2011;Perez et al. 2013). Importantly, aeGBR was reduced in patients compared to controls only in the cognitively more demanding condition, while it was not affected in the easy condition. Consistent with the latter fi nding, studies that assessed the aeGBR during passive P50 paradigms (reviewed in Taylor et al. 2013) have failed to observe abnormalities in patients with schizophrenia (although note that these paradigms use clicks instead of tones as auditory stimuli, which might lead to different gamma-band responses, cf. Taylor et al. 2013). The above suggest that the aeGBR reductions observed in patients with schizophrenia refl ect defi cits in cognitive, rather than sensory processes. In this sense, our fi ndings are comparable to those of previous studies reporting a failure of patients with schizophrenia to enhance gamma-band activity in response to increased task demands in working memory paradigms (reviewed in Gandal et al. 2012). In line with this conclusion, a previous study by our group ) demonstrated decreased activity not only in the auditory cortex in patients with schizophrenia, but also in the dorsal ACC -a region suggested to be involved in functional interactions with the auditory cortex, as its activation correlates with task diffi culty in auditory choice reaction paradigms (Mulert et al. 2007(Mulert et al. , 2008. In combination with the fi ndings of post mortem studies pointing to parvalbuminpositive GABAergic interneuron abnormalities in the ACC in patients with schizophrenia (Kalus et al. 1997;Woo et al. 2004), the results of the present Of the neuropsychological test scores, only Letter-Number-Span (a demanding test of working memory) was signifi cantly correlated with aeGBR evoked power in the EC (rho ϭ 0.711, P ϭ 0.003, CI ϭ 0.324 -0.898). This correlation remained signifi cant ( P ϭ 0.04) after correcting for multiple comparisons. A correlation between the same variables emerged at trend-level in the DC (rho ϭ 0.487, P ϭ 0.066, CI ϭ 0.046 -0.832); this trend disappeared after Bonferroni correction.

Discussion
The present study used two auditory reaction tasks differing in diffi culty, in order to investigate the relative contributions of sensory and cognitive processes to aeGBR defi cits in patients with fi rst-episode schizophrenia. There were no differences between patients and healthy controls in an easy condition dependent only on sensory processing and low-level vigilance. In contrast to healthy controls, though, patients failed to increase early evoked gamma power and phase-locking in response to a more cognitively demanding task. Power and phase-locking indices of the aeGBR in the easy condition showed weak associations with negative and disorganized symptoms in patients. There was also a signifi cant correlation of aeGBR power in the easy condition with one measure of working memory. This is the second study to report decreased aeGBR in fi rst-episode patients, confi rming that reported defi cits in schizophrenia are not simply a result of chronicity and/or long-term effects of antipsychotic medication, which is in line with several previous stimulus-related evoked responses ( " signal " ), but may additionally be affected by pre-stimulus gamma-band activity ( " noise " ), which has been consistently reported to be increased in patients with schizophrenia Hong et al. 2008;Spencer 2011). In fact, in a previous auditory steady-state stimulation study, total gamma-band power was correlated with baseline (pre-stimulus) gamma-band activity, whereas this was not the case for gamma-band phase-locking, suggesting the two measures are dissociable (Spencer 2011). Moreover, total gamma-band aeGBR is affected in chronic  but not fi rst-episode patients with schizophrenia and prodromal subjects (Perez et al. 2013;Taylor et al. 2013) -in contrast to evoked aeGBR, which is consistently reduced across all stages of the illness (see Introduction). Thus, it is possible that different aspects of the gamma-band study could indicate a disturbed " gamma-modulated " functional interaction between the ACC and the auditory cortex in schizophrenia. It is more diffi cult to reconcile the present results with reported reductions in the gamma-band auditory steady-state response (ASSR) in patients with schizophrenia (Kwon et al. 1999;Light et al. 2006), which does not depend on cognitive factors, but rather is assumed to refl ect bottom-up neural synchronization. In a recent study , gamma-band ASSR and aeGBR phase-locking (and, to a lesser extent, total power) indices were significantly correlated in patients with schizophrenia as well as in healthy controls, suggesting that the two gamma-band responses are, at least partly, dependent on the same mechanisms. Contrastingly, the ASSR involves a sustained elevation in total gammaband power that does not necessarily refl ect only Figure 2. Auditory evoked gamma-band response (aeGBR) phase-locking. Analysis of the phase locking factor within the timeframe 200 ms prior to the stimulus and 300 ms post-stimulus averaged over all subjects of healthy controls (HC, left column) and patients with schizophrenia (SZ, right column) for the diffi cult condition (DC, upper row) and the easy condition (EC, lower row). Scaling was uniform for both groups. An increased gamma phase-locking can be observed at about 50 ms after stimulus presentation (dashed line) and in the frequency range around 40 Hz. In contrast to SZ, healthy subjects showed a signifi cant increase of gamma phase-locking in DC compared to EC. encoding in patients (Light et al. 2006). An alternative interpretation in the present study could be that aeGBR defi cits refl ected the poor performance of patients. However, this is rather unlikely, as the correlation of aeGBR with Digit-Number Span was observed in the easy rather than the diffi cult condition.
A limitation of the present study is that most patients were medicated at the time of testing. However, there was no signifi cant correlation between antipsychotic medication dose and the aeGBR in the patients group, and medication status cannot generally explain similarities and differences in the fi ndings of previous studies (see Hamm et al. 2012). Only one study (Hong et al. 2004) reported an association between antipsychotic medication type and the magnitude of the gamma-band response, which was higher in patients treated with second-generation antipsychotics compared to those treated with fi rst-generation antipsychotics. As the majority of patients in the present study were on treatment with second-generation antipsychotics, it can be assumed that the effect of antipsychotic medication, if any, would have been to reduce differences in aeGBR between patients and healthy controls. A possible infl uence of the antidepressive treatment of seven patients with escitalopram or venlafaxin cannot be ruled out, although there are no correspondent reports and a serotonergic or noradrenergic infl uence on the generation of gamma-oscillations seems to be unlikely.
The DC compared to the EC involves both an increase of stimulus type variations and task demands (more than one response is required). Therefore, based on the present data, it is not possible to distinguish between these two possible reasons for the increase in aeGBR measures. However, we have explicitly investigated this question in a previous study with healthy subjects using the same paradigm as in the present study (Mulert et al. 2007): both an increase of stimulus type variation without increasing the number of different responses and an increase of the number of response options led to an increase of reaction times, error rates and participant ' s subjective self-ratings of task diffi culty and mental effort demands. According to that, the increase of cognitive demands in the DC compared to the EC at the bottom of the stronger gamma response is supposed to result from both an increase of stimulus type variation and an increase of response options.
In conclusion, evoked power and phase-locking of the early auditory gamma-band response were signifi cantly reduced in a sample of fi rst-episode patients with schizophrenia compared to healthy controls in a cognitively demanding auditory choice-reaction paradigm, whereas there were no differences between patients and controls in an easy version of the same generator circuit are involved in phase-locked (e.g., aeGBR evoked power) and non-phase-locked (e.g., total ASSR power) gamma-band responses and their abnormalities in schizophrenia. This conclusion is tentative, given that the present study was not designed to address this question. However, it is of particular interest, since it is consistent with the fact that pharmacological (Gandal et al. 2012) and computational (Spencer 2009) models of schizophrenia predict opposite changes (decrease or increase) in gamma-band activity, depending on which part of the gamma-generator circuit is affected (GABAergic interneurons or glutamatergic NMDA-receptors mediated signalling, respectively). Accordingly, it is conceivable that different dysfunctions across this circuit are responsible for different neurophysiological abnormalities.
According to the above, experimental design differences might also explain the rather inconsistent fi ndings of previous studies regarding the correlations of gamma-band responses with symptoms. In the present study, similar to previous ones by our group and others (Hall et al. 2011a;Leicht et al. 2010;Perez et al. 2013;Taylor et al. 2013), we observed no signifi cant correlation of evoked aeGBR power or phase-locking with positive symptoms. The state-independency of these markers is consistent with the notion of an endophenotype (Gottesman and Gould 2003), although our results indicate a relationship between aeGBR measures and negative symptoms of schizophrenia. However, negative symptoms display much greater longitudinal stability than positive symptoms, and are often regarded as " trait " features of schizophrenia (Harvey et al. 2006). Interestingly, all of the studies that failed to observe correlations between gamma-band responses and positive symptoms used auditory oddball paradigms that engage attentional and working memory mechanisms. In contrast, several other studies that reported signifi cant positive correlations between gamma-band responses and positive symptoms all used auditory steady-state stimulation paradigms.
Our fi nding of a signifi cant correlation between aeGBR and working memory should be interpreted with caution, given the small sample size and the fact that it only applied to one of the two measures used to assess working memory. However, it is a plausible fi nding, given recent studies that suggest a relationship between working memory capacity and synchronized activity in the theta-and gamma-band range (reviewed in Haenschel 2011). A similar correlation was observed in a much larger patient sample in a previous study (Light et al. 2006). As the study in question used a passive auditory steady-state paradigm, the authors concluded that defi cits in the early sensory processing of stimuli might lead to impaired working memory task. aeGBR indices were not associated with symptomatology, but a possible correlation with working memory defi cits was noted. This pattern of results confi rms the applicability of aeGBR disturbances as a stable endophenotype of schizophrenia, and suggests a cognitive, rather than sensory, processing defi cit at their origin.