Twenty seven (21 males) individuals who met the DSM-IV criteria (American Psychiatric Association, 2000) for schizophrenia were recruited from several outpatient clinics from Northumberland, Tyne and Wear NHS foundation trust and Tees, Esk and Wear Valleys NHS Foundation Trust. The psychiatric diagnosis was confirmed by an independent psychiatrist. All patients were taking antipsychotic drugs (see Table 1). An experienced psychiatrist examined the patients to ensure they did not meet any of the following exclusion criteria: current presence and/or history of co-morbidities with axis I disorders of the DSM or existence of neurological condition including head trauma and long periods of loss of consciousness. Additionally, 16 healthy participants (11 males) were recruited via advertisement in the local post office. They were also screened for history of psychiatric and neurological illness and drug use using the Schedules for Clinical Diagnosis in Neuropsychiatry (SCAN, WHO) to exclude psychopathology. Exclusion criteria for controls were the presence (current or history) of any axis I and II disorders of the DSM.

Additional interviews in patients were conducted by a qualified clinical psychologist, using the Comprehensive Assessment of Symptoms and History (CASH) (Andreasen et al., 1992). This semi-structured interview includes the Scale for the Assessment of Positive Symptoms (SAPS) (Andreasen, 1984) and the Scale for the Assessment of Negative Symptoms (SANS) (Andreasen, 1983). Both diagnostic scales require participants to report symptoms during the week prior to the assessment. Twelve patients who were not currently experiencing hallucinations in any modality (as defined by a score of 1 or below in SAPS hallucination global score) were allocated to the NAVH group. The NAVH group consisted of four patients who had never experienced hallucinations, as well as eight patients with a history of hallucinations but currently not experiencing them as measured by the SAPS, as we aimed to study the impact hallucinations exert on the experimental task as a state, not as a trait. Patients reporting hallucinations (scoring at least 3 on the SAPS hallucinations global score) were allocated to the AVH group. None of the patients scored between 1 and 3 in this scale. A cut-off score was chosen based on previous studies (Allen et al., 2004; Shea et al., 2007). The AVH group subsequently completed the auditory hallucination subscale corresponding to the PSYRATS (Haddock et al., 1999). Finally, all subjects were assessed with the National Adult Reading Test (NART) (Nelson and Willison, 1991) to estimate premorbid intellectual performance (IQ) with a high test–retest reliability in schizophrenia (Morrison et al., 2000) (see Table 2).

There were no between-group differences in age [F_{(2, 42)} = 0.72, ns], education as measured by highest qualification achieved [1 = primary school, 2 = secondary school, 3 = 0 levels, 4 = A levels, 5 = university degree; F_{(2, 42)} = 0.35, ns], premorbid verbal IQ [F_{(2, 42)}= 0.51, ns] and handedness measured by the Edinburgh Handedness Questionnaire [F_{(2, 42)} = 1.10, ns]. AVH patients had significantly higher total scores (U = 21.00, Z = −3.37, p < 0.05), hallucinations global subscale scores (U = 1.50, Z = −4.37, p < 0.001), and delusions global subscale scores (U = 26.50, Z = −3.18, p < 0.05) of the SAPS than NAVH patients. None of the other psychopathology measures revealed significant differences between patient groups (all U < 70.50, ns).

Participants' overall accuracy (across all groups and conditions) in the sex discrimination task was significantly above chance [M = 73% ± SD = 0.05, t₍₄₀₎ = 20.1, p < 0.001, one-sample t-test, chance level was 50%]. Accuracy scores (%) were subjected to a 2 × 3 × 3 mixed ANOVA with Ear attended (left, right) and Trial-type [neutral binaurally (baseline), emotion on the left ear (LEE), emotion on the right ear (REE)] as within-participants factors and Group (AVH, NAVH, HC) as between-participants factor. The ANOVA revealed a main effect of Trial type [F_{(2, 76)} = 3.34, p < 0.05]. Alpha adjusted pairwise comparisons indicate that the accuracy during baseline (75.00% ± 0.01) is higher than during the REE condition (72.00% ± 0.01, p < 0.05). The LEE condition (73.40% ± 0.01) did not differ from REE and baseline. The main effect of Ear attended did not approach significance [F_{(1, 38)} = 0.42, ns]. Moreover, there was a significant interaction between Ear attended and Trial type [F_{(2, 76)} = 7.94, p < 0.05]. Alpha-adjusted post hoc t-tests revealed that in the REE condition, participants obtained lower accuracies when the right ear was attended (69.40% ± 1.27) than when the left ear was attended [75.00% ± 1.07, t₍₄₁₎ = −3.09, p < 0.05], suggesting that if the emotion is presented to the attended ear, there is an increased difficulty to ignore the affective distractor while performing the sex discrimination task (see Figure 1A). No other post hoc test approached significance (all t ≤ 2.45, ns). No other main effect or interaction approached significance, (all F ≤ 1.49, ns).

Identical to the accuracy data analysis, a 2 × 3 × 3 ANOVA was calculated for RTs. As predicted, we found a significant main effect of Ear attended [F_{(1, 38)} = 9.33, p < 0.05] with significantly longer RT when attending to the right (1269 ms ± 39) than the left (1205 ms ± 43) ear, suggesting that, across the whole sample, sex discrimination by voice depends more the right hemisphere. Moreover, in agreement to our hypothesis that emotionally neutral trials are more efficiently processed than emotional trials there was significant main effect of Trial type [F_{(2, 76)} = 9.23, p < 0.001]; alpha-adjusted pairwise comparisons showed lower RT in the baseline (1187 ms ± 42) than in the LEE (1260 ms ± 38; p < 0.05) or REE (1256 ms ± 40; p < 0.05) conditions, again suggesting that, across the whole sample, EP interferes with the gender prosody discrimination task. The LEE condition did not differ from REE. Additionally, the main effect of Group was significant [F_{(1, 38)} = 5.45, p < 0.05]; alpha-adjusted pairwise comparisons revealed only lower RTs for AVH (1354 ms ± 65) than for HC (1062 ms ± 65, p < 0.05). Other group differences did not approach significance.

We predicted interference, if both processes that are gender discrimination (explicit task) and EP (implicit) information converge on the same side. The interference between both processes should be even stronger after stimulus presentation to the right ear, corresponding to the left hemisphere (inferior for prosody processing), as the non-specialized hemisphere cannot efficiently deal with both processes simultaneously. Accordingly, there was a significant interaction between Ear attended and Trial type [F_{(2, 76)} = 8.44, p < 0.001]; alpha-adjusted post hoc t-tests revealed a significantly slower RT for the REE condition when the right ear was attended [t₍₄₀₎ = 4.35, p < 0.001], whereas LEE did not differ from the baseline condition (see Figure 1B), suggesting that the left hemisphere (inferior for prosody processing) has difficulties to perform the explicit task when the emotional distractor is simultaneously processed by the same (left) hemisphere.

Finally, there was a significant three-way interaction [F_{(4, 76)} = 6.74, p < 0.001]. This interaction comprised the factors Ear attended, Trial type and Group. To investigate this effect, three separate ANOVAs (one for each group) were performed. There was no main effect of Ear attended in any of the groups (all F ≤ 3.82, ns). The main effect of Trial type was only significant in HC [F_{(2, 28)} = 5.50, p < 0.05], and in the AVH group [F_{(2, 28)} = 8.74, p < 0.001]. In the control group, alpha-adjusted post hoc test showed faster responses in the baseline (1026 ms ± 68) than in the REE condition [1085 ms ± 68, t₍₁₄₎ = 3.09, p < 0.05] but the baseline condition did not differ from LEE [1076 ms ± 68, t₍₁₄₎ = 1.36, ns]. The REE and LEE conditions did also not differ significantly (see Figure 2). In the AVH group, the baseline condition (1309 ms ± 73) revealed faster RTs than REE [1395 ms ± 71, t₍₁₄₎ = 3.01, p < 0.05] and LEE [1384 ms ± 77, t₍₁₄₎ = 3.52, p < 0.05]. REE did not differ from LEE [t₍₁₄₎ = 1.62, ns]. The main effect of Trial type in the NAVH group was not significant [F_{(2, 20)} = 0.37, ns]. Most importantly, the interaction between Ear attended and Trial type was significant in HC [F_{(2, 28)} = 28.07, p < 0.001, η² = 0.513] and NAVH [F_{(2, 20)} = 15.45, p < 0.001, η² = 0.480] but not in the AVH group [F_{(2, 28)} = 2.98, ns, η² = 0.131]. The missing interaction between Ear attended and Trial type in AVH patients indicates that automatic processing of EP interfered with performing the explicit task in this group, regardless which ear was stimulated, probably as a result of aberrant lateralization in processing prosodic information.

To explain the significant interaction between Ear attended and Trial type in HC and NAVH, alpha-adjusted post hoc t-tests showed that RTs were slower when the Ear attended and the emotion coincided on the right [healthy controls: t₍₁₄₎ = 3.70 p < 0.05; NAVH: t₍₁₀₎ = 3.34, p < 0.05], not when the left ear was attended in the REE condition (all t ≤ 2.55, ns). The comparison between attended ears was not significant in LEE and the baseline condition (all t ≤ 0.85, ns). To compare convergent (Ear attended and side of emotion co-occurring at the same hemispace) and divergent condition (Ear attended and the side of emotion occurring at opposite hemispace), additional alpha-adjusted post hoc t-tests were performed. The analyses revealed that both control groups showed a significant difference between LEE and REE, when the left ear was attended [healthy controls: t₍₁₄₎ = 4.88, p < 0.001; NAVH: t₍₁₁₎ = −2.72, p < 0.05], not when the right ear was attended.

All participants had faster responses in the sex discrimination task when attending to the left (as opposed to the right) ear, in agreement with previous reports (Lattner et al., 2005; Sokhi et al., 2005). Likewise, all participants had faster responses on the emotionally neutral baseline condition than in emotional trials, also in agreement to previous findings (Aue et al., 2011) suggesting that interference with this task by auditory emotional stimuli is due to involuntary capture of attention. Also consistently with others (Grandjean et al., 2008), such interference was independent from emotional valence stimuli. The right hemisphere superiority for sex discrimination (Lattner et al., 2005; Sokhi et al., 2005) and EP (Ross and Mesulam, 1979; Ross et al., 1997; Friederici and Alter, 2004; Ross, 2010) may explain the longer RTs when the right ear was attended to, corresponding to the inferior-for-the-task left hemisphere, as an interhemispheric transfer is required.

We also found lower accuracies and slower responses when emotion was presented to the attended right ear, possibly due to interference caused by the convergence of explicit task and the (implicit) emotion presented in the same hemispace; perhaps because one cerebral hemisphere does not have the capacity to process both tasks in parallel. Interference was less pronounced when the explicit and implicit tasks occurred in opposite sides. Thus, it may be more difficult to allocate voluntary attention when the emotional distractor drives involuntary attention toward irrelevant but salient features of the target stimuli (Grandjean et al., 2005; Sander et al., 2005). Orienting attention away from the task-relevant hemispace probably makes emotional content of the prosody stimulus and its automatic processing less salient, consequently causing less interference.

The convergence of the emotional distractor and the attentional focus on the same side disrupted the performance significantly only when it happened on the right ear side, suggesting that the (inferior-for-the-task) left hemisphere is particularly sensitive to this effect. Sander and colleagues (Sander et al., 2005) observed the same convergence effect, but only in the right hemisphere; however, their sample was much smaller (15 subjects) and did not include patients with schizophrenia, perhaps reducing their ability to identify subtle behavioral change. Indeed, they did not find a significant behavioral asymmetry in the sex-discrimination task (Sander et al., 2005).

It still remains open whether this abnormal lateralization is a trait or a state in hallucinators. A previous study assessed phonological processing in two groups of schizophrenia patients, one with on-going AVH and another with history of AVH as well as healthy controls (Løberg et al., 2004). They implemented a dichotic listening paradigm in which participants either focused attention to the left or right ear, or performed a non-forced condition without specific attentional instruction (Løberg et al., 2004). They found that since abnormal dichotic listening asymmetry for language stimuli is a state marker for AVH which is coincident with the perception of AH, the modulation of dichotic listening performance by means of voluntary attention is a trait marker, seen both in patients with ongoing and with only a history of AVH (Løberg et al., 2004). This voluntary attention difficulty in dichotic listening might also relate to schizophrenia in general instead of a AVH trait in particular, however, our data shows that NAVH patients do not present such difficulty suggesting this deficit is an AVH trait.

Importantly, there is vast literature applying lexico-phonological dichotic listening tasks in schizophrenia patients (Green et al., 1994; Bruder et al., 1999; Levitan et al., 1999; Hugdahl et al., 2007; Force et al., 2008). Summarizing the cited studies, we can conclude that schizophrenia patients do not show the normal right ear advantage for lexico-phonological stimulus due to electrophysiological abnormalities in the left temporal lobe (Bruder et al., 1999) and reduced event related potential component of early auditory processing (Force et al., 2008). In relation to hallucinations, AVH patients show lack of ear advantage for linguistic stimulus in comparison to NAVH patient and controls (Green et al., 1994; Hugdahl et al., 2007; Ocklenburg et al., 2013), which are highlighted by gray matter reduction in the left temporal lobe (Hugdahl et al., 2007). Our study complements the literature by suggesting the existence of right hemisphere abnormalities in AVH patients reflected in abnormal lateralization of EP. It has previously been proposed that attentional control on dichotic listening reflects an interaction between frontal executive processing and the left and right temporal lobe speech processing, underlied by frontotemporal connections (Løberg et al., 2006; Hugdahl, 2009), which seems disrupted in AVH patients (Lawrie et al., 2002).

AVH patients could be more symptomatic and thence more cognitively impaired than NAVH patients, but in our sample there were no differences in premorbid verbal IQ. Unfortunately, we cannot fully rule out that a greater general cognitive impairment in the AVH group might have caused top-down difficulties in hallucinators. However, it should be noted that the NAVH and AVH groups did not differ in the baseline condition. Moreover, it is negative, rather than positive, symptoms that are associated with neurocognitive impairment (Lewandowski et al., 2007; Ventura et al., 2009). AVH patients rated higher on the delusion subscale, and therefore we cannot exclude delusions contributing toward the interference of implicit EP. However, impaired performance in dichotic listening in psychotic patients relates to hallucinations particularly (Løberg et al., 2002). We found no differences in hearing thresholds, thus a hearing deficit cannot explain the observed effects. Importantly, we do not disregard the impact that audio perceptual abnormalities such as pitch perception can exert on prosodic processing, which was demonstrated in previous research (Leitman et al., 2005, 2007, 2011). We did not test pitch perception specifically, as it would have exceeded the scope of the study. Both strand of research (pitch perception abnormalities and EP deficits in AVH) do not invalidate each other, but instead they are regarded as complementary. Importantly, in a previous study (Alba-Ferrara et al., 2011) we demonstrated that the brain representation of emotional processing reminds unchanged when statistically controlled for pitch differences.

AVH patients are not typically right-lateralized for EP because they do not benefit from attending to either ear during baseline, and convergent emotion does not deteriorate performance preferentially on either side. In addition, AVH patients' inability to filter emotion trials when the attended ear diverges from the ear in which EP is presented may support lack of hemispheric specialization for implicit EP processing. Although our study included two potentially right lateralized processes (EP and sex discrimination), we believe that the atypical lateralization rather refers to EP because of the significant main effect of ear attended. The left ear advantage for the explicit task was present for all groups. Atypical asymmetries in EP processing in AVH patients might not only help to understand the formation of hallucination and its cognitive impairments. They might also be useful as a diagnostic marker and risk factor for AVH formation in schizophrenic patients.

Several caveats are needed. Since all our patients took antipsychotic medications, we cannot fully ascertain whether the differences between groups are not influenced by them. The ability to ignore irrelevant stimuli, which is compromised in schizophrenia (Leumann et al., 2002), improves with atypical antipsychotics (Green et al., 2001). The three groups did not differ in premorbid IQ, education, years of illness, age, or handedness; thus, these factors are unlikely to explain the findings. Lastly, a mood disorder screening was performed ensuring that our patients were free from such confound. It should, however, be noted that patients had in general longer latencies compared to controls. Slow RT in general have been long recognized as one of the behavioral deficits of schizophrenia (Cromwell and Held, 1969). However, since the NAVH and the control group show the same pattern performance (accuracy and RT across conditions), the AVH group behaves differently in that they do not show the convergence-dependent interference effect.

The present findings have implications for neuropsychological models of hallucinations. According to Woodruff (Woodruff, 2004), an attentional bias toward emotional stimuli in schizophrenia patients has been previously observed (Green et al., 2001; Waters et al., 2006) and it may underlie AVH. The present results extend the finding of abnormal attentional bias toward prosodic emotional stimuli presented to the non-dominant side regardless of the focus of voluntary attention in AVH patients specifically. Emotional salience of prosodic stimuli may capture attention even when it should be oriented away from the distractor, resulting in emotional inputs accessing processing in detriment of non-emotional competitors. Such attentional bias toward irrelevant EP stimuli might impair the ability to focus on relevant aspects of the acoustical environment (Javitt, 2009). A breakdown in selective attention may cause an overwhelming sensorial influx of irrelevant data resulting in abnormal perceptions (Alba-Ferrara et al., 2012). Hallucinations are aberrant perceptual processes which usually convey emotional salience. Such bottom-up saliency may diminish top-down control as AVH may shift attention toward the perceived voices (Hugdahl et al., 2007).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.