Fifteen patients with schizophrenia or SSD and 15 healthy controls were recruited via postings of in- and outpatients from the Department of Psychiatry and Psychotherapy at Philipps-University Marburg, Germany. DSM-IV criteria of the diagnosis were confirmed by a psychiatrist via a structural clinical interview. All patients had a diagnosis of schizophrenia or SSD (F20–F29) and received constant antipsychotic medication (for medication and diagnostic details, see Supplementary Table A1 in the Supplementary Material). Healthy participants were free of psychiatric diseases in their medical history. The inclusion criteria were monolingual German as the mother language, right-handedness, and legal age. According to the declaration of Helsinki, all subjects gave written informed consent to participate in the study. The local ethics committee also approved the study (ethics proposal no. 37/10), and all participants were paid for their participation.

The schizophrenia group (SZ) consisted of 15 participants (13% women, n = 2) with a mean age of 38.8 years (SD = 14.33) and an average of 10.5 educational years (SD = 1.59). The healthy control group (HC) comprised 15 participants (20% women, n = 3) with a mean age of 38.8 years (SD = 10.57) and an average of 10.7 educational years (SD = 1.16). Participants in both groups were matched for age, sex, years of education, and level of education, with the latter dichotomized based on the presence of the university entrance qualification (for detailed information, see Table 1 ).

The functional imaging acquisition was performed by two experienced clinical psychologists. After scanning, speech was transcribed by a clinical linguist. An experienced psychiatrist, blind to the diagnosis, assessed and evaluated psychopathological phenomena for all participants using the Thought and Language Index (TLI) (16). In the clinical group, the sum value of the TLI was 0.52 (SD = 0.57) for positive FTD and 2.45 (SD = 2.88) for negative FTD. In the healthy control group, the sum value of the TLI was 1.7 (SD = 3.49) for positive FTD and 1.97 (SD = 3.74) for negative FTD. Within 48 h of scanning, psychopathological symptoms were also rated in patients by one of the clinical psychologists using the scales for the assessment of positive and negative symptoms (SAPS/SANS) (17, 18). We collected total values for SAPS (11.2, SD = 11.97), global positive FTD (1.33, SD = 1.25), SANS (5.87, SD = 6.13), and global alogia (0.87, SD = 0.88) for the clinical group (for detailed information regarding psychopathological symptom severity, see Table 1 and Supplementary Table A1 in the Supplementary Material).

All participants were instructed to describe social situations in the MRI scanner, where at least two persons were pictured using eight pictures from the Thematic Apperception Test (TAT) (19). The exact instruction for the participants was as follows: “Please describe in as much detail as possible what you see in the picture and what could be going on there.” The pictures were presented on a monitor (aspect ratio: 16:9) behind the scanner and mirrored via a display approximately 15 cm from the participants as they lay in the MRI scanner. The sequence of the pictures was randomized to avoid sequence effects.

At the beginning of each trial, a fixation cross was presented for 5 ms. After every picture depiction, a rhombus was presented on the screen for a duration of exactly 15 s. This duration of 15 s, which was equal for all participants and was defined as experimental rest, was the contrast of interest. The overall duration of the spontaneous speech paradigm was 10 min (eight pictures of 60 s + eight experimental rest periods of 15 s). Furthermore, every participant displayed an individual behavior regarding rests during speech blocks, defined as self-chosen rests, which were considered a contrast of no interest. Regarding previous evidence for the German language, we considered self-chosen rests during speech blocks of at least 100 ms (20).

Participants’ speech was recorded during the entire scanning session using a noise-reducing microphone system (FOMRI-II, Optoacoustics Ltd., Or-Yehuda, Israel) including a dual adaptive filter system (21) that subtracted the reference input (background MRI noise) from the source input (speech signal). The microphone was fixed at the head coil and wired to the sound filter box. The output port was directly wired to the audio in-line plug of the notebook sound card.

The digitally recorded audio files (Audacity, version 1.2.6, Softonic International S.L.) were saved in 16-bit WAV format and transcribed into text files using the F4 plus transcription software (version 5.10.x, Dr. Dresing & Pehl GmbH, Marburg, Germany). Contrary to the F4 transcription guidelines, hesitations (like “ehm,” “uh”), breakup of words, and self-chosen rests were transcribed for the correct presentation of FTD. Regarding previous evidence for the German language, we considered self-chosen rests of at least 100 ms (20). We considered every realized word and did not distinguish between word types and tokens (22).

Initially, independent of the activation direction, significant neural activation differences were found during spontaneous speech production of healthy control participants (HC) compared to that of patients with schizophrenia (SZ) relative to experimental rests ([HC speech > HC exp. rest] > [SZ speech > SZ exp. rest]). In the clinical group, right-hemispheric BOLD signal changes were found in the superior frontal gyrus, medial frontal gyrus, and IFG (particularly the Rolandic operculum and orbital part); medial temporal pole; STG; hippocampus; and cerebellum. Furthermore, relative to the patients, the healthy control participants showed left-hemispheric BOLD signal changes in the postcentral gyrus and left hippocampus (see Table 2 ; for details, see Supplementary Table A2 ). These differences in activity were not observed in the opposite contrast ([SZ speech > SZ exp. rest] > [HC speech > HC exp. rest]).

In the second step, we considered the direction of activation differences between the groups in speech-relevant areas, including the right STG and right IFG (Rolandic operculum). Particularly in the region of the right STG, patients showed increased neural activation during experimental rests relative to the healthy control participants (see Figure 1 ). In the right IFG, particularly in the Rolandic operculum, patients showed opposite neural activation patterns to the healthy control participants: During experimental rests, the control group showed reduced neural activity, whereas the clinical group showed increased neural activity (see Figure 2 ). For both areas, no significant effect with antipsychotic medication, operationalized as chlorpromazine equivalent, was observed (for detailed information, see Supplementary Table A3 ).

As mentioned above, we observed increased activation in the right STG and right Rolandic operculum in the patients during experimental rests. To show that the experimental activation was actually stronger in the right-hemispheric network than in the left one, we performed a repeated measure ANOVA as a post-hoc analysis with the extracted contrast-estimates (eigenvariates) of homolog regions from the left and right hemispheres in the area of the Rolandic operculum and STG for both groups. We found significant group, task, and laterality interactions. As shown in Figure 3 , the effect of lateralization was mainly driven by activation during experimental rests in the areas of interest, with significant differences between the clinical groups (STG: F = 4.56, p <.042; Rol. operc.: F = 10.373, p <.003).

In our correlation analysis, we extracted contrast estimates from speech-relevant areas for the patients, particularly peak voxels from the right STG and IFG (area of the right Rolandic operculum) with a sphere radius of 4 mm. The neural activity during speech production, which was correlated with the number of words per block (included as a parametrical regressor in the model), was then correlated with psychopathological assessments. The amount of activity in the right STG (4-mm sphere), explained by the number of spoken words, correlated negatively with the SAPS sum scale “global PFTD” (rho: −0.521, p = .046). However, this effect was not robust to multiple testing. No correlation was found between the amount of activity in the right STG (4-mm sphere) and antipsychotic medication (rho: −0.317, p = .25; for detailed information, see Supplementary Table A3 ).

The amount of activity in the right Rolandic operculum (4-mm sphere), explained by the number of spoken words, correlated negatively with the TLI subscale “PFTD sum” (rho: −0.604, p = .017; see Table 3 ). This effect was stable after multiple testing. No significant correlation was found between the amount of activity in the right Rolandic operculum (4-mm sphere) and antipsychotic medication (rho: −0.97, p = .731; see Supplementary Table A3 ).

The whole-brain analysis revealed significant neural activation differences between the groups. During spontaneous speech production, patients showed BOLD signal changes in a right-hemispheric network in comparison to healthy control participants, particularly in the superior frontal gyrus, medial frontal gyrus, and IFG (especially the orbital part and Rolandic operculum); STG; medial temporal pole; hippocampus; and cerebellum. Furthermore, in contrast to the control participants, the patients showed left-hemispheric BOLD signal changes in the postcentral gyrus and hippocampus. Therefore, the clinical group showed a reversed lateralization of areas within and outside of the traditional language-related network. These findings aligned with previous fMRI research in patients with schizophrenia (3, 5, 7, 9, 26).

One reason for the predominantly right-hemispheric lateralization in the IFG and STG could be structural volume reductions in the homologous left hemisphere areas (3–6, 8, 27, 28). Another reason could be the task of spontaneous speech elicitation on which this work is based: The TAT (19) partially depicts ambiguous everyday situations. In other words, the neuronal activation during the elicitation of spontaneous speech by the TAT can also be influenced by the effects of social cognition (29).

Moreover, patients also presented activation differences in areas outside the classical language-related network, such as the right cerebellum, which plays an important role in auditory self-monitoring as well as articulatory speech planning (3). These functions are often impaired in schizophrenia patients (30–32).

Patients showed an increase in neural activity during rest conditions relative to speech in the STG and IFG lateralized to the right hemisphere, whereas the healthy participants showed clear activation differences between rest and speech. A post-hoc analysis of this unexpected finding showed that the effect of right-hemispheric lateralization was mainly driven by activation during experimental rests in schizophrenia patients. This finding suggested a general overactivation of the right-hemispheric language areas in schizophrenia patients.

Models of semantic priming or mental lexicon (10, 33, 34) assume a main lexicon in the left STG with rapid, targeted activation of semantically closely related terms, as well as an additional lexicon in the right STG, which ensures slower and more widespread activation of semantically more distant terms. The latter is necessary for the reception of sentences and decoding of ambiguity (35). In the context of schizophrenia, ambiguity means that patients with FTD experience challenges with the pragmatic level of speech. These problems may manifest in interpreting daily situations and their social implications incorrectly due to a lack of transferring the intended meaning or abstract semantic content of metaphoric expression (36). In an event-related fMRI study, Rapp and colleagues examined 15 female patients with schizophrenia in comparison to healthy control participants. The participants had to identify whether target sentences had ironic, literal, or meaningless content. BOLD signal change was decreased in patients during irony comprehension, particularly in the right temporal and posterior medial prefrontal regions, as well as the left insular regions, whereas the left parahippocampal gyrus showed an increased activation relative to healthy participants (36). In the TAT used for spontaneous speech elicitation in the current study, some pictures depicted ambiguous social situations. For this reason, the pragmatic level, which is often disturbed in patients with schizophrenia, may have influenced lateralization.

The lack of downregulation in the right STG during experimental rests could be explained by word retrieval, which is more unprecise in schizophrenia patients due to extended, increased activation of the additional lexicon (9). Regarding the right IFG, the Rolandic operculum is both an element of the opercular part of Broca’s area and a component of the precentral gyrus (37). Therefore, other functions in this area may influence experimental rests, such as processing emotions or sensory information (38, 39), which only had been relevant for the clinical group but were not depicted by the present paradigm.

Various studies have shown that schizophrenia patients with persisting symptoms have heightened emotional recognition deficits (40, 41), particularly in recognizing facial emotions correlating with negative symptoms (42). As a component of the precentral gyrus, the Rolandic operculum plays an important role in emotional processing through the fronto-limbic structures (38). Potvin and colleagues investigated 39 patients with schizophrenia as well as healthy control participants who rated emotionally positive, negative, and neutral images during fMRI scanning (40). They found that patients with schizophrenia gave emotionally neutral situations more emotional meaning than controls. Both groups showed significant activations in the dorso-medial prefrontal cortex (dmPFC) and the bilateral amygdala. However, in contrast to the healthy control participants, the Granger connectivity from the right amygdala to the dmPFC was reduced in patients during the neutral and negative conditions. Furthermore, the Granger connectivity from the left amygdala to the dmPFC was reduced in patients during the positive condition. This finding of reduced connectivity between the left and right hemisphere amygdala to the dorso-medial prefrontal gyrus may explain why patients with schizophrenia attach emotional meaning to emotionally irrelevant (i.e., neutral) stimuli (40). Since all images in the TAT used in the present study depicted people with different facial expressions, the influence of misinterpretation of emotional stimuli by the patient group must also be considered.

For right STG and right IFG, the observed lack of downregulation during experimental rests in patients may also have been due to dysfunctional executive functions. Selecting convenient thoughts in a special context and suppressing non-convenient thoughts are fundamental for coherent speech. However, this ability is often reduced in schizophrenia patients and may lead to an ongoing neural activation during experimental rests (11). Therefore, this should be considered in future studies on spontaneous speech in patients with schizophrenia; test procedures to assess executive functions, such as the Trail Making Test (TMT) (43), should be used.

Regarding the lateralization of language, the gender distribution of the patient population should also be considered. Our clinical group consisted of 13 male and two female patients. Gender differences are a frequently replicated finding in schizophrenia research (44). While women present the first symptoms of the disease an average of 5 five years later than men (45), the severity of these symptoms and the response to pharmacotherapy also differ (44). For example, women often show positive symptoms, while men tend to show negative symptoms (46).

In an fMRI study, Sommer and colleagues examined the language lateralization of female patients with schizophrenia compared to healthy subjects (47). In the clinical group, language lateralization in left hemisphere areas was lower compared to the control group. However, activation of the right hemisphere language areas was increased. The group of female schizophrenia patients was then compared with male schizophrenia patients (48). The male group also showed less left-sided language lateralization compared to healthy individuals, as well as increased activation on the right side. Therefore, no significant group difference in language lateralization was found between the clinical groups (47). Despite known gender-specific peculiarities in patients with schizophrenia, we do not assume any influences of gender-specific differences in language lateralization, based upon these findings from Sommer.

To achieve the second aim of this fMRI study, the variance in BOLD signal accounted for by the number of words per 20-s speech block was correlated with measures of psychopathology (TLI, SAPS/SANS). The amount of activity in the right Rolandic operculum, explained by the number of spoken words, correlated negatively with the severity of positive FTD measured with the TLI subscale “PFTD sum”.

Although the patients did not differ from the healthy control subjects in terms of the severity of FTD at the group level, there was a significant group difference at the neuronal level. At this point, the existing pharmacotherapy of the patients must be considered: Atypical antipsychotics, such as clozapine or quetiapine, bind to and block the dopamine receptor (D2-R) with a lower affinity than typical antipsychotics, such as haloperidol, which cause a strong D2-R blockade. In a behavioral design, deBoer and colleagues compared the spontaneous speech production of patients receiving antipsychotics with high and low affinity for the D2-R (49). They found that patients who used antipsychotics with high D2-R occupancy showed more pronounced negative speech disorders, such as increased pauses and shortened utterance length, compared to patients who used antipsychotics with low D2-R occupancy (49).

Regarding the influence of antipsychotic medication on the lateralization of language, Hwang and colleagues investigated 21 therapy-naive patients with psychotic symptoms before and after 6 weeks of treatment with aripiprazole (50). During fMRI scanning, patients and healthy control participants completed a semantic decision paradigm. This task was repeated after 6 weeks of treatment with the atypical antipsychotic. After treatment, the patients showed an increased left-hemispheric lateralization index, particularly in the IFG, which was due to decreased activity in the right-hemispheric IFG. This improved lateralization was positively correlated with verbal fluency. The authors explained the improved semantic performance by the increased left-hemisphere lateralization caused by aripiprazole (50). Regarding the stability of functional language lateralization over time in schizophrenia patients, Razafimandimby and colleagues investigated 10 patients with schizophrenia in comparison to healthy control participants in two fMRI sessions at an interval of 21 months (51). All patients were in a stable chronic phase of their illness with no psychotic exacerbation during the study period and were regularly treated with either typical or atypical antipsychotics. The authors found significantly reduced left hemisphere lateralization in the clinical group compared to the control group, especially in the so-called semantic areas of interest. Since this effect was not influenced by the study period, the specific task, or the presence of psychotic symptoms, the authors advocated a special organization of language in patients with schizophrenia (51).

In the present study, all patients were also in the chronic phase of their disease and received different antipsychotic medications, as well as co-medication, comparable to the abovementioned study. Only three of the patients received aripiprazole, and only one received it in monotherapy. Nevertheless, an effect of antipsychotic medication on the lateralization of language seems possible and should be re-examined in future studies on spontaneous speech in patients with schizophrenia using a homogeneous antipsychotic medication. However, regarding chlorpromazine equivalent doses (see Supplementary Table A3 in the Supplementary Material), medication did not influence the within-patient correlational analyses. Comparable to the study by Razafimandimby and colleagues, these results suggest an abnormal neural organization of language in patients with schizophrenia that appears to be independent of the duration of the illness (51).

Because of its topography in the human brain, the right Rolandic operculum is not only relevant in speech processing but also in processing emotion (39) and sensory information (38, 52). Due to this role, this area might influence symptoms like delusions (47, 48), depression, or anxiety (39). Future studies should consider the influence of these additional psychopathological symptoms. Our aim was to reflect a broader range of schizophrenia patients, so we avoided highly selected participants regarding the single phenomenon of FTD.