Neural correlates of the core facets of empathy in schizophrenia

Abstract

Empathy is a multidimensional construct composed of several components such as emotion recognition, emotional perspective taking and affective responsiveness. Even though patients with schizophrenia demonstrate deficits in all core components of this basic social ability, the neural underpinnings of these dysfunctions are less clear.

Using fMRI, we analyzed data from 15 patients meeting the DSM-IV criteria for schizophrenia and 15 matched healthy volunteers performing three separate paradigms tapping the core components of empathy, i.e. emotion recognition, perspective taking and affective responsiveness.

Behavioral data analysis indicated a significant empathic deficit in patients, reflected in worse performance in all three domains. Analysis of functional data revealed hypoactivation in a fronto-temporo-parietal network including the amygdala in patients. Moreover, amygdala activation correlated negatively with severity of negative symptoms.

The results suggest that schizophrenia patients not only suffer from a broad range of emotional deficits but also show cortical and subcortical abnormalities, extending previous findings on fronto-temporal cortical dysfunctions. Since empathy is related to psychosocial functioning and hence of high clinical relevance in schizophrenia, a more detailed understanding of the exact nature of these impairments is mandatory.

1. Introduction

Empathy is a multidimensional construct and requires the ability to perceive, understand and feel the emotional states of others. Due to the complexity of the construct (see Preston and de Waal, 2002; Singer and Lamm, 2009) various definitions exist. However, according to most models empathy consists of at least three core components (Decety and Jackson, 2004): 1) The ability to recognize emotions in oneself and others via facial expressions, speech or behavior, 2) a cognitive component, also referred to as perspective taking or theory of mind, describing the competency to take over the perspective of another person, though maintaining the essential distinction between self and other, and 3) an affective component, i.e. sharing of emotional states with others or the ability to experience similar emotions as others.

For each of the three factors several studies reported significant impairments of schizophrenia patients on the behavioral level (emotion recognition: e.g., Schneider et al., 2006; Van’t Wout et al., 2007; Strauss et al., 2010; perspective taking: Brüne, 2005; Langdon et al., 2006; Montag et al., 2007; affective responsiveness: Shamay-Tsoory et al., 2007a,b; Bora et al., 2008). Recently, we investigated all three empathy components and demonstrated that patients indeed show a significant deficit in every single domain (controlling for the deficit in the two other domains), thus indicating an overarching deficit in empathic abilities (Derntl et al., 2009a). As these form an essential part of social functioning, such deficits may lead to impairments in social interaction, which characterize and stigmatize schizophrenia patients (Brüne, 2005).

Most previous neuroimaging studies addressing these deficits in schizophrenia patients focused on only a single component. For emotion recognition, studies mainly report hypoactivation of regions known to be involved in facial emotion processing, e.g. fusiform gyrus, insula, amygdala (e.g., Gur et al., 2007; Fakra et al., 2008; Habel et al., 2010), a pattern that has also been observed in subjects at risk of psychosis (Seiferth et al., 2009) and juvenile patients (Seiferth et al., 2008). Interestingly, during the processing of neutral faces some studies report hyperactivation of schizophrenia patients in emotion-related regions (e.g., Holt et al., 2006; Seiferth et al., 2008; Habel et al., 2010; Mier et al., 2010).

Functional imaging studies investigating the cognitive component of empathy demonstrate a much more heterogeneous pattern of findings, mostly reporting hypoactivation of prefrontal areas in patients (e.g., Russell et al., 2000; Brunet et al., 2003; Lee et al., 2006; Marjoram et al., 2006).

Recently, two neuroimaging studies have been published investigating the neural correlates of empathy in schizophrenia patients by applying almost identical versions of a cartoon task measuring theory of mind/cognitive empathy as well as affective/emotional empathy (Benedetti et al., 2009; Lee et al., 2010). Benedetti et al. (2009) observed greater response of the right superior temporal gyrus (STG) to affective empathy cartoons and stronger response of the right transverse as well as right posterior superior temporal gyrus to cognitive empathy cartoons in patients, while controls showed the opposite pattern. Partly supporting these results, Lee et al. (2010) also reported stronger response of the right STG to the cognitive empathy cartoons in controls. They, however, observed a stronger response of the left insula to the affective empathy cartoons in patients compared to controls. Additionally, Lee et al. investigated another subtype of empathy, the so-called inhibitory empathy, measured via presentation of conflict situations between two persons asking for the empathic solution, demonstrating stronger right middle/inferior frontal gyri activation in patients compared to controls. Thus, both studies suggest dissociable neural networks for affective and cognitive empathy. Interestingly, neither study observed significant correlations between psychopathological parameters and neural activation. However, both studies have certain shortcomings such as the use of a task with limited socio-ecological validity, where it might be doubted that affections were really shared or experienced. Moreover, they both did not investigate the ability to recognize facial expressions of emotions (neither outside nor inside the scanner), a necessary prerequisite of empathy, known to be deficient in schizophrenia. Hence, a distinction between emotion recognition deficits underlying the observed empathy deficits is difficult.

The aim of the present study therefore was to investigate the neural substrate of all three defining components of empathy in schizophrenia patients and matched healthy controls, enabling more detailed and exact analyses of these emotional competencies, their interactions and possible neural dysfunctions in patients. Based on previous data (Derntl et al., 2009a), we expected significant behavioral impairments in all three tasks. Regarding the neural correlates we hypothesized that patients will show hypoactivation of emotion-related regions during facial emotion recognition (cf. Seiferth et al., 2008; Habel et al., 2010; Mier et al., 2010). Based on previous data (e.g., Benedetti et al., 2009; Lee et al., 2010), we expected stronger response of the right superior temporal region during emotional perspective taking which might be accompanied by hypofunctions of the prefrontal cortex (e.g., Russell et al., 2000; Brunet et al., 2003; Lee et al., 2006; Marjoram et al., 2006). For the affective responsiveness task we hypothesized less activation of emotion-related brain regions such as the anterior cingulate cortex, amygdala, orbitofrontal and temporal regions in the schizophrenia patients as has been observed in studies investigating mood induction and emotional experience (e.g., Habel et al., 2004; Reske et al., 2007).

2. Methods and material

2.1. Sample

Initially, 18 patients had been included in the study. Due to excessive head movement 3 patients were excluded from further data analysis. Hence, 15 schizophrenia patients meeting the DSM-IV criteria for schizophrenia and 15 healthy controls matched for gender, age and parental education were included in the analysis. All subjects were Caucasian. Exclusion criteria included substance abuse for the last six months and (other) psychiatric or neurological illness based on the German version of the Structured Clinical Interview (SCID, Wittchen et al., 1998). Symptom severity in patients was assessed with the Positive and Negative Syndrome Scale (PANSS, Kay et al., 1987). All subjects were paid for their participation and gave written informed consent after a complete study description. The study was approved by the ethics committee of the RWTH Aachen University.

Patients (both inpatient and outpatient) were recruited from the Department of Psychiatry, Psychotherapy and Psychosomatics of the RWTH Aachen University, Germany. All patients received second-generation antipsychotics but no other psychopharmaceuticals.

Three questionnaires measuring cognitive and affective empathy were administered: the Questionnaire Measure of Emotional Empathy (QMEE, Mehrabian and Epstein, 1972), the German Questionnaire for Assessment of Empathy, Social Attitude and Aggression (FEEPA, Lukesch, 2006) and the German version of the Interpersonal Reactivity Index (Paulus, 1997).

All participants completed tests tapping crystallized verbal intelligence (MWT-B, Lehrl, 1996), executive functions, such as cognitive flexibility (TMT-A/-B, Reitan, 1958), working memory (digit span, WAIS III, Von Aster et al., 2006), and verbal fluency (Aschenbrenner et al., 2000).

Performance of patients and controls did not differ in any of the neurocognitive tasks (MWT-B: t=1.066, p=.299; TMT-A: t=−.874, p=.392; TMT-B: t=−1.187, p=.249; phonematic verbal fluency: t=−.030, p=.976; semantic verbal fluency: t=−.203, p=.845; working memory: t=.472, p=.642).

Demographic, clinical and neuropsychological characteristics are shown in Table 1.

Table 1.

Sociodemographic characteristics of schizophrenia patients and healthy controls, showing no significant differences in age, parental education, and intelligence estimation between groups.

	Patients (n = 15)	Controls (n = 15)	p-Value
Gender (F:M)	5:10	5:10	–
Age	34.2 (9.1)	30.4 (8.9)	.259
Parental education	10.8 (3.7)	12.2 (3.4)	.328
Intelligence estimation (MWT-B)	30.2 (3.1)	32.0 (4.1)	.247
PANSS positive score	12.3 (3.8)
PANSS negative score	14.6 (3.7)
PANSS general score	24.5 (4.4)
PANSS global score	52.0 (7.8)
CPZ-Equivalents (mg/day)	329.9 (212.2)
Duration of illness (years)	7.3 (5.3)
Mean age of illness onset	26.8 (10.5)

2.2. Functional tasks

The same tasks have been used in an fMRI study addressing gender differences recently and are described in more detail there (Derntl et al., 2010). Briefly, we applied three tasks tapping the three components of empathy separately.

2.2.1. Emotion recognition

We presented 60 colored photographs of Caucasian facial identities depicting five basic emotions (happiness, sadness, anger, fear, and disgust) and neutral expressions. Half of the stimuli were used for an emotion recognition task, the other half for an age discrimination control task and all actors appeared only once. All stimuli were selected from a standardized stimulus set (Gur et al., 2002) that has been used frequently as neurobehavioral probes (e.g., Fitzgerald et al., 2006; Habel et al., 2007; Derntl et al., 2008a,b; 2009b; Müller et al., 2011). Stimuli were comparable with respect to gender, age, intensity, valence, and brightness. Stimulus presentation was randomized with regard to task, emotion and gender but kept constant for all subjects. For emotion recognition subjects were instructed to choose the correct emotion label from two possibilities presented on the left and right sides of the image, by pressing the corresponding button of a response box using the index or middle finger of the right hand. In the control trials, subjects had to judge which of two age decades was closer to the poser’s age and then press the corresponding button. One of the options was correct and the other was selected at random from all other choices. Position of correct choice was balanced for left vs. right button press. Facial expressions were presented maximally for 5 s with a randomized, variable interstimulus interval (ISI) ranging from 8 s to 12 s (during which subjects viewed a blurred face with a central crosshair). Manual responses triggered the next ISI.

2.2.2. Emotional perspective taking

Participants viewed 70 items in total depicting scenes showing two Caucasians involved in social interaction thereby portraying five basic emotions. Half of the stimuli were used for emotional perspective taking, the other half for the control task. The face of one person in the scene was masked, and participants were asked to infer the corresponding emotional expression of the masked face. These scenes were shown for 4 s and immediately afterwards responses were made by selecting between two different emotional facial expressions or an emotional and a neutral expression. Facial alternatives were taken from the same pool of stimuli described above. Again, one option was correct and the alternative was selected at random from the remaining choices. In the control trials, participants had to indicate whether the left or the right face was masked by pressing the corresponding button on the response device after being presented with two response alternatives (‘left’, ‘right’). The locations of the response labels were counterbalanced across trials (e.g. ‘left’ on right side of the screen, thus requiring a right button press). Stimuli were presented in blocks of seven stimuli, each block was preceded by an instruction slide indicating the next task for the block (emotional or control).

2.2.3. Affective responsiveness

We presented 70 short written sentences describing real-life situations which are likely to induce the basic emotions described above. Again, half of the stimuli were used for the affective responsiveness task and the other half for the control task. For affective responsiveness, participants were asked to imagine how they would feel if they were experiencing those situations. Stimuli were presented for 4 s and response format was the same as for emotional perspective taking, presenting two facial expressions, one showing the correct emotion and the other was chosen randomly from the other expressions. For the control task, participants had to indicate the number of words forming the sentence, and were then presented with two options. Again, subjects were confronted with blocks of stimuli (as above) and informed prior whether an emotional or control block will appear next.

Response format was kept maximally similar between tasks allowing comparisons between tasks, so that differences could be traced back to different task requirements rather than different response formats. All stimuli were presented using goggles (VisuaStimDigital, Resonance Technology Inc., Los Angeles, CA). The presentation of images, recording of responses and synchronization with the scanner was achieved using the Presentation© software package (Neurobehavioral Systems, Inc., Albany, CA).

2.3. Behavioral data analysis

Statistical analyses were performed using SPSS 15.0, and level of significance was set at p=.05 corrected. The proportion of correct responses of each empathy task acquired during scanning were analyzed using repeated measures ANOVAs with task as within-subject factor and diagnosis as between-subjects factor. Similarly, for analysis of reaction times we used repeated-measures ANOVA with task as within-subject factor and diagnosis as between-subjects factor. For significant effects partial-eta squares are listed as estimates of effect size. Whenever the assumption of sphericity was violated Greenhouse-Geisser corrected degrees of freedom and p-values are reported. Post hoc comparisons were Bonferroni corrected.

Group differences in the empathy questionnaires were assessed using independent samples t-tests and estimates of effect size are reported for significant differences (Cohen’s d).

Correlations between accuracy measures of the empathy paradigms and self-report scores as well as clinical parameters (positive and negative symptom severity, medication, duration of illness, etc.) were computed using non-parametric methods (Spearman rank correlation) testing one-sided for negative, respective positive correlations.

To analyze the influence of current medication, we computed CPZ-equivalents according to the transformations given by Andreasen et al. (2010).

2.4. FMRI acquisition parameters and data processing

All subjects were examined with a 3 T whole-body scanner (Philips, Best, The Netherlands) at the Interdisciplinary Centre for Clinical Research, Medical Faculty of RWTH Aachen University. Functional imaging was performed using gradient-recalled EPI; 35 oblique axial slices were acquired using asymmetric k-space sampling (FOV=24×24 cm, matrix size 64×64, slice thickness 3 mm, slice gap 0.3 mm, TR=2200 ms, TE=31 ms). Each task was divided into two experimental runs of about 8 to 13 min acquiring between 220 (emotion recognition and age discrimination paradigm) and 360 images (emotional perspective taking and affective responsiveness tasks). Taken together, 6 runs were performed (2 per task), tasks were presented in a randomized order, and empathy and control items were counterbalanced across runs. Three dummy volumes preceded each run. Total measurement time was about 60 min. After functional neuroimaging, anatomical images were acquired using standard T1-weighted and T2-weighted 3D sequences.

A signal artifact correction algorithm was applied to the raw functional imaging data to reduce large, slice-wise mean signal fluctuations in the fMRI time series. The correction algorithm estimates slice-wise signal levels across the fMRI time series, and then dampens deviations from the overall mean signal level for each individual slice. For each time point and slice, the algorithm computes an adjustment factor such that small, and therefore possibly physiological, signal deviations are left almost completely unchanged, whereas larger and therefore most likely non-physiological deviations are adjusted more strongly. Functional data were then preprocessed using SPM5 (http://www.fil.ion.ucl.ac.uk/spm/spm5.html). Images were slice timing corrected, realigned to the first image and normalized into the standardized stereotactic space. To increase SNR and allow group level inferences, functional data sets were spatially smoothed using an isotropic Gaussian kernel with a full-width-at-half-maximum of 8 mm.

2.5. FMRI data analysis

For this event-related design, each stimulus type was modeled with a separate regressor using a delta function convolved with the canonical hemodynamic response function. For the emotional perspective and the affective responsiveness task the initial task stimulus (scene or sentence) and the response option stimulus were modeled with separate regressors allowing for the differentiation between task-related neural activity and neural activation due to the response format (emotional facial expressions vs. verbal response labels in the control tasks). For the present data analysis, response related activity was not further analyzed. First-level models also included motion parameters as regressors of no interest to bind variance associated with motion-induced signal changes.

For the analysis of functional data, we pooled all stimuli across emotions to assess brain responses to each empathy task leaving a minimum of 30 stimuli per task for statistical analysis (emotion recognition: 30 stimuli; emotional perspective taking: 35 stimuli; affective responsiveness: 35 stimuli). Statistical analysis was performed at the individual and group level. To detect group differences, contrast images from all subjects for each task ([emotion recognition, perspective taking, affective responsiveness] vs. rest) were included in a second level random effects analysis. To specifically analyze neural activation during empathy processing we performed a 2 (diagnosis)×3 (task) mixed effects ANOVA with diagnosis as between-subjects factor and task as within-subject factor. Since we were specifically interested in group differences in empathic abilities and their neural correlates, we explored effects of group and task as well as the possible interaction by applying F-contrasts and, if significant, by subsequent post-hoc t-contrasts disentangling any differences between schizophrenia patients and controls.

In order to account for multiple comparisons we applied a combined height and extent threshold technique based on Monte-Carlo simulations using AlphaSim (Cox, 1996). According to 1000 simulations based on a height threshold of p<.001 (uncorrected) and the spatial properties of the residual image an extent threshold of 55 contiguous voxels suffices to comply with a family wise error of p<.05. This correction for thresholding will be referred to as “height and extent threshold” (HET) and group results as well as direct comparisons between schizophrenia patients and controls are demonstrated at this threshold.

2.5.1. ROI analysis

We performed a ROI analysis for the amygdala region with the aim of maximizing the sensitivity for group effects as well as hemispheric lateralization differences in the amygdala. The amygdala has been chosen due to several reasons: it plays a major role in emotion processing, has also been implicated in empathy (e.g., Derntl et al., 2010), and several studies observed dysfunctional amygdala activation during emotion processing in schizophrenia patients (e.g., Gur et al., 2007; Seiferth et al., 2008). Values for amygdala ROIs were extracted drawing a sphere (10 mm) centered at [x,y,z: +/− 20, 0, −20] which were based on previous results (Amunts et al., 2005), thereby covering both amygdalae. Mean parameter estimates were extracted for left and right amygdala in each condition and Levene tests for homogeneity of variances indicated homoscedasticity for all parameter estimates of all tasks (emotion recognition left: p=.720, emotion recognition right: p=.506; emotional perspective taking left: p=.823, emotional perspective taking right: p=.518; affective responsiveness left: p=.839, affective responsiveness right: p=.768).

A three-way ANOVA was applied with group as between-subject factor, and task and laterality as repeated factors. Greenhouse-Geisser corrected p-values are presented.

2.5.2. Corollary analyses

Correlation analyses were performed for each task between performance (% correct and reaction time) and whole-brain analysis as well as amygdala parameter estimates (taken from the ROI analysis). Moreover, neural activation was also correlated with performance in the self-report empathy questionnaires and clinical parameters (e.g., symptom severity, medication). Again, whole-brain results will be presented at a threshold of p<.05 HET corr.

3. Results

3.1. Behavioral performance

Due to an error of the response device, behavioral data of one male patient were not included in the final analysis.

Analysis of percent correct demonstrated a significant task effect (F(2,54)=13.780, p<.001, partial-eta sq=.346), a significant group effect (F(1,27)=4.824, p=.037, partial-eta sq=.156), with better performance in the control group, and no significant task-by-group interaction (F(2,54)=0.969, p=.386, ns).

Bonferroni corrected post hoc tests of the main effect of task revealed significantly lower performance in emotional perspective taking compared to affective responsiveness (p<.001) and emotion recognition (p=.002) while performance in emotion recognition did not differ significantly from affective responsiveness (p=.832).

For reaction times, no significant group effect (F(1,27)=0.239, p=.629, ns) but a significant task effect (F(1.554,40.397)=78.071, p<.001, partial-eta sq=.750), and a significant task-by-group interaction (F(1.554,40.397)=12.740, p<.001, partial-eta sq=.329) emerged.

Post-hoc tests disentangling the significant task effect revealed significantly faster reaction times for affective responsiveness than for either of the two other tasks (both p<.001). Moreover, reaction times during emotional perspective taking were significantly faster than during emotion recognition (p<.001). Regarding the significant task-by-group interaction post-hoc tests revealed significantly faster responses of patients in the perspective taking task (p=.003), while for the other tasks no significant group difference occurred (emotion recognition: p=.272, affective responsiveness: p=.845).

Fig. 1 illustrates performance on the three empathy tasks across patients and controls.

Fig. 1.

Performance (% correct with a line indicating the group mean, a) and reaction times (seconds, a line indicating the group mean, b) in emotion recognition (ER), emotional perspective taking (EPT), and affective responsiveness (AR) in schizophrenia patients and healthy controls. Results of data analyses revealed a significant group effect (p=.037) in performance, with patients showing less accuracy, and a significant task-by-group interaction (p<.001) for reaction times, indicating faster response times in patients during EPT (p=.003). Note that equal performance of participants sometimes might be covered by only one data point.

Correlation analyses between the empathy tasks revealed significant positive associations between all three tasks (all p<.008) as shown before (Derntl et al., 2009a).

3.1.1. Empathy questionnaires

Direct comparison of self-report empathy scores demonstrated significantly lower scores in the IRI empathy scale (t=3.177, p=.004, d=1.207) as well as the subscale fantasy (t=2.535, p=.018, d=.973) and a trend for empathic concern (t=1.986, p=.058, d=.756) in patients indicating reduced empathic abilities. For the QMEE (t=0.821, p=.420, ns) and the FEEPA (t=1.008, p=.324, ns) no significant difference was observed.

3.1.2. Influence of psychopathology

Correlation analysis with PANSS scores revealed a significant association between emotion recognition and PANSS general psychopathology scores (r=−.562, p=.029) as well as with PANSS total scores (r=−.538, p=.039). Regarding any impact of medication or duration of illness, correlation analyses revealed neither a significant association between behavioral performance and CPZ-equivalents (all p>.109) nor with duration of illness (all p>.115). Also, for the empathy questionnaires no significant correlation with medication or duration of illness emerged (all p>.194).

3.2. Functional data

Separate group analyses for schizophrenia patients and healthy control participants showed activation of a widespread cortical–subcortical network including inferior and middle frontal regions, orbitofrontal cortex, temporal gyri, fusiform gyri, inferior parietal cortex, and cingular cortex across all tasks.

Analysis of the whole brain functional data applying a mixed effects ANOVA revealed significant group differences across all empathy tasks in several regions, e.g., left thalamus, inferior frontal gyri bilaterally, left anterior cingulate cortex, and precuneus bilaterally (see Fig. 2a). Moreover, a significant task effect, and a significant task-by-group interaction emerged (see Fig. 2b). For more details see Table 2. Thus, post-hoc analyses for each paradigm were performed.

Fig. 2.

(a) Illustration of the significant group effect (F=5.69, p<.05 HET corr.), depicting hypoactivation of the patients vs. healthy controls in several brain regions. (b) Illustration of the significant group-by-task interaction (F=7.21, p<.05 HET corr.) showing significant differences in activation of the right putamen (slice: z=−4) and the right middle cingulate (slice: y=−4). Note that equal activation of participants sometimes might be covered by only one data point.

Table 2.

Results from the mixed effects ANOVA with group and task as factors, showing a significant main effect of group (threshold: F=5.69, p<.05 HET corr.) and a significant group-by-task interaction (threshold: F=7.21, p<.05 HET corr.). MNI coordinates, cluster size, F-value, laterality and region are given.

Contrast	MNI coordinates			Cluster size	F-value	L/R	Region
	X	Y	Z
Group effect	−4	−8	8	6792	20.51	L	Thalamus
	52	30	8	1135	14.81	R	Inferior frontal gyrus
	−40	34	4	342	13.28	L	Inferior frontal gyrus
	−44	−30	18	416	11.98	L	Middle temporal gyrus
	−2	34	8	83	11.28	L	Anterior cingulate cortex
	−12	−60	8	201	9.89	L	Posterior cingulate cortex
	−4	−66	54	519	9.48	L	Precuneus
	−2	34	58	72	9.30	L	Paracentral lobule
	40	−22	−20	60	9.21	R	Fusiform gyrus
	2	−16	44	294	9.15	R	Middle cingulate cortex
	−2	−52	−10	121	8.52	L	Cerebellum (vermis)
	−14	50	44	118	8.37	L	Superior frontal gyrus
	38	42	28	178	8.28	R	Middle frontal gyrus
	18	−58	42	61	8.21	R	Precuneus
	−22	−22	−26	81	7.09	L	Parahippocampal gyrus
Group ×task interaction	16	8	−4	63	8.51	R	Putamen
	10	−4	28	113	7.94	R	Middle cingulate cortex

For emotion recognition, the post-hoc t-contrasts indicated that controls recruited the left thalamus, the inferior frontal gyri (trigeminalis) bilaterally, the anterior, middle, and posterior cingulated more strongly than patients. Patients, however, showed no stronger activation than controls (see Table 3 for details).

Table 3.

Results from the post-hoc analysis of the significant group-by-task interaction. For each task MNI coordinates, cluster size, t-value, laterality and region are given (threshold: t=3.14, p<.05 HET corr.).

Contrast	MNI coordinates			Cluster size	F-value	L/R	Region
	X	Y	Z
Emotion recognition—controls > patients
	2	−8	8	4440	6.04	R	Thalamus
	40	−22	−20	314	4.51	R	Fusiform gyrus
	52	30	8	159	4.41	R
	−36	32	8	70	4.01	L	Inferior frontal gyrus
	−4	−74	52	191	4.39	L
	8	−46	74	270	4.37	R	Precuneus
	4	−30	44	196	4.31	R	Middle cingulate cortex
	0	−50	−12	140	4.11	M	Cerebellum
	8	42	8	340	3.87	R	Anterior cingulate cortex
	−44	−82	20	63	3.86	L	Middle temporal gyrus
Emotion recognition—patients > controls
Emotional perspective taking - controls > patients
	14	6	−4	519	4.76	R	Thalamus
	32	40	−4	966	4.69	R	Orbitofrontal gyrus
	44	2	42	254	4.67	R	Middle frontal gyrus
	20	−58	42	230	4.50	R	Precuneus
	−16	52	42	203	4.37	L	Superior frontal gyrus
	52	30	8	169	4.21	L
	−42	34	2	172	4.17	R	Inferior frontal gyrus
	−40	−78	14	156	4.19	L
	26	−82	14	86	3.84	R	Middle occipital gyrus
	−46	−50	0	122	3.66	L	Middle temporal gyrus
Emotional perspective taking - controls > patients
–
Affective responsiveness—controls > patients
	−14	−8	40	2365	4.52	L
	18	8	42	895	4.27	R	Middle cingulate cortex
	−10	−62	52	263	4.23	L	Precuneus
	−38	14	52	96	4.15	L	Middle frontal gyrus
	−10	−66	14	146	4.01	L	Posterior cingulate cortex
	−2	34	58	63	3.98	L	Superior frontal gyrus
Affective responsiveness—patients > controls
–

For emotional perspective taking, post-hoc t-contrasts demonstrated that controls showed stronger neural response in a broad network including the right middle frontal gyrus extending to the anterior cingulate, the inferior frontal gyri (trigeminalis) bilaterally, the right precuneus, the left amygdala, the thalamus, the superior frontal gyrus, and the cerebellum. Again, patients demonstrated no significantly stronger activation (see Table 3 for details).

Analysis of group differences in the neural correlates of affective responsiveness also revealed only stronger activation in controls, particularly in the left superior medial frontal gyrus, the left precuneus, and the middle and posterior cingulate (see Table 3 for details).

Table 3 shows MNI coordinates for all regions of the post-hoc t-contrasts with group differences in the neural activation underlying the three empathy components.

Performing a conjunction analysis across the three empathy tasks in patients, we observed neural activation in the left middle occipital gyrus (−26, −92, 16, k=201, t=4.18), the right superior occipital gyrus (26, −74, 30, k=1638, t=4.93), the left inferior frontal gyrus (−40, 14, 26, k=367, t=4.48), the left lingual gyrus (−14, −92, −12, k=57, t=4.45), and the left superior frontal gyrus (−4, 6, 56, k=56, t=3.57).

For controls, the conjunction analysis revealed activation in the inferior frontal gyrus bilaterally (left: −46, 32, 2, k=2940, t=5.97; right: 40, 16, 22, k=335, t=4.13), the middle occipital gyri extending to the superior occipital gyri (left: −22, −92, 14, k=2455, t=6.03; right: 28, −84, 18, k=2692, t=6.58), the right medial frontal gyrus (2, 16, 48, k=1045, t=7.12), the left thalamus (−10, −16, 8, k=72, t=4.67), and the left middle temporal gyrus (−60, −50, 4, k=104, t=3.99). See Fig. 3 for illustration of the group conjunction analysis.

Fig. 3.

Illustration of the conjunction analysis across all three tasks (p<.05 FEW corr.) in controls (left) and patients (middle), revealing significantly less activation in a wide-spread network including frontal, temporal, occipital, and subcortical regions in patients (right).

Corollary analyses between performance parameters and the corresponding neural activation for each task revealed a significant association in patients only in the middle cingulate cortex (−10, −24, 44, k=67, t=7.94) for emotion recognition. Moreover, reaction time during affective responsiveness correlated significantly negative with activation of the right precuneus (right: 12, −44, 44, k=59, t=4.69). For controls, neither accuracy nor reaction time revealed a significant association at the applied threshold.

Regarding the impact of clinical parameters, we analyzed the influence of PANSS negative and positive scores on neural activation during the three tasks and observed a significant positive correlation of PANSS negative scores with neural activation during emotion recognition in the right inferior frontal gyrus (18, 16, −12, k=64, t=6.14) and the right parahippocampal gyrus (26, −38, −10, k=104, t=5.18). Moreover, a significant positive correlation emerged between neural activation during affective responsiveness and PANSS negative scores in the left lingual gyrus (−20, −60, −6, k=259, t=6.50), the right amygdala (30, −2, −12, k=267, t =5.90), the right middle cingulate cortex (6, −30, 50, k=210, t=5.02), the thalamus bilaterally (right: 12, −14, 4, k=77, t=5.77, left: −14, −20, 8, k=93, t=5.53), and the right putamen (32, −18, 6, k=175, t =5.02). Additionally, PANSS negative scores and activation of the right precuneus (6, −56, 36, k=61, 4.81) during emotional perspective taking correlated significantly negative. For the PANSS positive scores no correlation reached significance at the applied threshold.

3.3. ROI analysis

The ROI analysis demonstrated a significant task effect (F(2,56)=13.724, p<.001, partial-eta sq.=.337), a significant group effect (F(1,28)=10.901, p=.003, partial eta-sq.=.288) with lower amygdala activation in schizophrenia patients, and no significant effect of laterality (F(1,22)=0.125, p=.726, ns). Moreover, no interaction reached significance (all p>.214).

3.3.1. Corollary analyses

After Bonferroni correction for multiple correlations, data analysis demonstrated a significant correlation between amygdala activation during emotion recognition and the IRI empathy score (r=.553, p=.025) in controls, while no other correlation between amygdala activation and self-report data in controls or patients reached significance (all p>.061).

Correlation analysis between performance accuracy and task-specific amygdala activation revealed only a significant relation for affective responsiveness (r=0.616, p=.009) but not for perspective taking (r=0.443, p=.053) nor emotion recognition (r=0.210, p=.471) in controls. For patients no correlation reached significance (all p>.253). For reaction times no significant correlation emerged (patients: p>.348; controls: p>.144).

Regarding clinical parameters, a significant negative correlation emerged between PANSS negative scores and amygdala activation during emotion recognition (r=−0.453, p=.045), while no other association reached significance. Moreover, CPZ-equivalents (all p>.180) and duration of illness (all p>.078) did not correlate with amygdala activation. Fig. 4 illustrates left and right amygdala activation for the three tasks across schizophrenia patients and controls and the significant negative correlation between PANSS negative scores and amygdala activation during emotion recognition.

Fig. 4.

Result of ROI analysis for the amygdala revealed a significant group effect (p=.003) indicating stronger amygdala activation bilaterally in controls in each of the three paradigms (mean activation of each group is indicated with a line, a). Note that equal amygdala activation of different participants sometimes might be covered by only one data point. Moreover, a significant correlation between amygdala activation during emotion recognition (ER) and PANSS negative scores emerged (r=−0.453, p=.045).

4. Discussion

The aim of the current study was to explore the neural correlates of dysfunctional affective and cognitive components of empathic abilities in schizophrenia patients and to compare patients’ performance with results from well-matched healthy controls. In contrast to previous studies addressing empathic abilities and the neural correlates in schizophrenia (e.g., Benedetti et al., 2009; Lee et al., 2010), we not only presented three separate paradigms tapping the core components of empathy but also relied on more socially relevant stimuli, e.g. facial expressions and photographic scenes instead of cartoon stories.

As hypothesized, patients showed significantly impaired performance in all three paradigms supporting the assumption of a broader emotional deficit (Derntl et al., 2009a). Moreover, in comparison to age-, gender-, and education-matched controls, we observed significant impairments in empathic behavior in the patient sample that was accompanied by dysfunctional activation in a widespread neural network, particularly in regions known to be associated with emotion processing, such as the inferior frontal gyrus, the anterior and middle cingulate cortex, the precuneus, and the amygdala. These results have several implications and limitations.

4.1. Abnormalities in the neural empathy network in schizophrenia

Recently, we reported a generalized deficit in empathic components in schizophrenia patients (Derntl et al., 2009a). In the present study, we explored the neural correlates of this deficit and observed dysfunctions in a wide-spread neural network. Previous literature has emphasized fronto-temporo-parietal abnormalities during empathy or theory of mind tasks in schizophrenia (e.g., Brüne et al., 2008; Benedetti et al., 2009; Lee et al., 2010). Our study provides evidence that subcortical regions, such as amygdala, thalamus and putamen may also contribute to these deficits.

In healthy samples, activity in the inferior frontal gyrus (IFG) has been repeatedly observed during imitation tasks (e.g., Iacoboni et al., 1999; Rizzolatti and Craighero, 2004) and various emotional processes including empathy (Derntl et al., 2010), passive viewing of faces (Dapretto et al., 2006), emotional perspective taking (Schulte-Rüther et al., 2007, 2008), emotion recognition and evaluation (Carr et al., 2003; Seitz et al., 2008). Consequently, Shamay-Tsoory et al. (2009) proposed that the IFG is the core structure of emotional empathy. In our data, schizophrenia patients showed significantly less activation in this region, not only across the three paradigms, but particularly during affective responsiveness, the task we believe to be tapping the affective empathy component. Hypoactivation of the IFG has been reported in some neuroimaging studies addressing emotion processing in schizophrenia patients investigating affective mentalizing (Russell et al., 2000), and emotion matching (Fakra et al., 2008). Recently, Habel et al. (2010) reported lower IFG activation during recognition of happy, sad and angry faces in a sample of schizophrenia patients. Interestingly, Lee et al. (2010) observed hyperactivation of the right IFG (BA 44) during tasks measuring inhibitory empathy, a sub-component that particularly focuses on conflict and harm prevention which was not addressed in the current study. Besides functional differences, Suga and colleagues (2010) not only reported significantly reduced gray matter volume in this area in schizophrenia patients, they also reported significant associations with symptom severity. Interestingly, we observed a significant inverse correlation between neural response of the right IFG during emotion recognition and the PANSS negative score, indicating that the more severe the negative symptoms the less IFG activation emerged.

Given the functionality of sub-regions of the IFG for action observation, imitation, mentalizing, and emotion processing (mainly BA 45, see review Nishitani et al., 2005), we speculate that the observed hypoactivation of this area during various empathic abilities is one important substrate of an impaired capacity to spontaneously simulate another person’s subjective world. As apparent in the behavioral data and previous studies (e.g., Langdon et al., 2006; Derntl et al., 2009a) schizophrenia patients have difficulties with empathetically appreciating the likely content of another person’s mind in order to take appropriate account of that other person’s feelings. Thus, evidence accumulates that the IFG and particularly the pars triangularis (BA 45) as found in the current study, plays an important role in the pathophysiology of emotional deficits in schizophrenia (e.g., Wisco et al., 2007; Suga et al., 2010).

Another region where we observed hypoactivation of patients across all three tasks was the precuneus. Interestingly, most studies observed hyperactivation, particularly of the right precuneus, during facial affect processing (Fakra et al., 2008; Habel et al., 2010) and speculate that this finding may reflect compensatory processes triggered to counterbalance the deficits in emotion processing. In their meta-analysis, Cavanna and Trimble (2006) demonstrate that activation of the anterior precuneus has repeatedly been observed during mental and motor imagery and evidence has accumulated that this region also plays a particular role in social cognition, self-agency, and self-processing (e.g., Vogeley and Fink, 2003; Koenigsberg et al., 2009). Notably, we observed a significant negative correlation with neural activation of the right precuneus during emotional perspective taking and PANSS negative scores, indicating that the more severe the symptoms the less activation occurred. Perhaps this finding partly explains why we observed hypoactivation of this region during this task in patients compared to controls.

Additionally, we observed abnormalities in activation of the ventral anterior cingulate cortex (ACC), which plays a crucial role in detecting (Lane et al., 2000) and recalling (Phan et al., 2002) emotional stimuli, specifically during tasks involving cognitive demands like all three tasks presented here. Schizophrenia patients have frequently been characterized by emotion-related ACC hypoactivation (e.g., Habel et al., 2004; Reske et al., 2007), suggesting disturbances in the top–down modulation of emotional responses leading to deficits in assessing the salience of emotion and motivational information.

The group-by-task interaction revealed significant differences in neural response of the right middle cingulate cortex and the right putamen. Looking at the parameter estimates (see Fig. 2b), patients relied more strongly on these regions during affective responsiveness, whereas for the other tasks hypoactivation was visible. However, this effect was not significant in the direct comparison. Interestingly, we also observed significant positive correlations between activation of these two regions and negative symptoms indicating the higher the symptoms the stronger the activation. The affective responsiveness task is the only of the three tasks that is focusing on the self (“how would you feel in a certain situation”) and notably, the middle cingulate cortex (mCC) extending to the posterior cingulate cortex has been shown to be implicated in self-reflection and introspectively focused tasks (Buckner et al., 2008), processing of acute pain (Derbyshire, 2000), visuospatial simulation (e.g., Vogt et al., 2003) and autobiographical memory retrieval (Wagner et al., 2005; Moran et al., 2011) as well as anxiety related to aversive conditioning prompting avoidance behavior (Büchel et al., 1998, 1999). Moreover, it represents an important node within a larger network of regions demonstrating elevated activity during resting states – the default network (e.g., Buckner et al., 2008) which has been shown to be disturbed in schizophrenia patients (Camchong et al., 2011; Salgado-Pineda et al., 2011). Interestingly, Holt et al. (2011) observed increased activation of this region during a self-reflection task in schizophrenia patients and authors speculate that this hyperactivity in midline cortical function is due to a general deficit of schizophrenia patients in introspection or social cognition.

The putamen forms the lateral part of the dorsal striatum. Due to the fact that it receives direct dopaminergic efferents from the substantia nigra (Moore, 2003) it has been considered as predominantly relevant in the context of reinforcement processing (Bischoff-Grethe et al., 2009; Ino et al., 2009), in signaling discrepancies between reward expectations and outcomes (Tobler et al., 2006), and reward-based learning (Haruno and Kawato, 2006; Wachter et al., 2009). Recently, Koch and colleagues (2010) demonstrated putamen abnormalities in schizophrenia patients during reward-related trial-and-error learning potentially suggesting the basis for learning deficits. Here, we observed a significant negative association between right putamen activation during affective responsiveness and reaction times reflecting stronger activation when patients responded faster.

Given these functionalities and the abnormalities in fronto-temporo-cingulate regions, we speculate that abnormalities in cingulate and putamen activation support assumptions about compensatory mechanisms for emotional dysfunctions. In order to perform the tasks, patients might show a higher level of self-awareness and effort triggering learning effects, which might be partly reflected in faster response times. Thus, we believe that these abnormalities might also reflect a lack of insight and awareness of the extent of the emotional dysfunction which of course fosters problems in social interaction the patients are confronted with.

4.2. Empathic deficit is associated with amygdala dysfunction in schizophrenia

Regarding the results of the region of interest analysis, we observed significant hypoactivation of the amygdala across all three paradigms. Moreover, PANSS negative scores correlated significantly with amygdala activation during emotion recognition. Hence, patients with prominent negative symptomatology exhibited less amygdala activation. Thus, our results support most previous neuroimaging studies on facial emotion processing where patients also failed to activate the amygdala (emotion discrimination: e.g., Fakra et al., 2008; emotional experience: e.g., Dowd and Barch, 2010; emotional face memory: e.g., Satterthwaite et al., 2010). However, some studies reported amygdala hyperactivation to neutral facial expressions in patients (e.g., Holt et al., 2006; Pinkham et al., 2008; Seiferth et al., 2008; Mier et al., 2010). The amygdala is one essential node in the neural network underlying salience attribution (Zald, 2003). Previous results showing hypoactivation to emotional stimuli and hyperactivation to neutral faces might indicate that the amygdala – together with other structures such as the striatum (Dowd and Barch, 2010) – fails to mark the salience of the emotional stimuli while on the other hand attributing salience to non emotional stimuli. Taking into account findings from studies on sad mood induction, these functional deficits seem to be genetically modulated (Habel et al., 2004) and may be crucial for the emotional impairments and observed affective symptoms.

We also observed a significant correlation between severity of negative symptoms and emotion recognition performance, pointing to worse performance of those patients with more severe negative symptoms which is in accordance to the finding of reduced amygdala activation in those patients. Thus our results partly support recent findings (Habel et al., 2010; Satterthwaite et al., 2010) but stand in contrast to several other published studies on facial affect recognition (e.g., Van’t Wout et al., 2007). This heterogeneity of results is further demonstrated in a recent meta-analysis of Kohler and colleagues (2010), who showed that severity of positive, negative and global symptoms correlated in about half the studies reviewed.

Despite this inconsistency, Wölwer et al. (2005) reported a negative relationship between facial affect recognition after training and the amount of negative symptoms, i.e., improvement in facial affect recognition performance was positively related to clinical improvement regarding negative symptoms. Therefore, a thorough analysis of the association of behavioral performance in tasks tapping socio-emotional competencies and psychopathology is mandatory, thereby helping to predict the course of illness or treatment effects.

Moreover, we observed significantly reduced self-report empathy scores in patients which has been shown before (e.g., Shamay-Tsoory et al., 2007a,b). However, groups did not differ in their self-reported personal distress and perspective taking scores as in most previous studies (Montag et al., 2007; Haker and Rössler, 2009), further supporting our assumption that schizophrenia patients may not be fully aware of their empathic impairments.

4.3. Limitations

Several limitations of this study should be acknowledged. Our sample was heterogeneous with respect to gender, age of illness, chronicity, and medication. Gender differences in empathic abilities have been reported on the behavioral (e.g., Rueckert and Naybar, 2008) and the neural level (e.g., Schulte-Rüther et al., 2008; Derntl et al., 2010) in healthy individuals. Thus, future studies should explore whether these behavioral differences are accompanied by distinct neural responses as seen in healthy samples, further characterizing female and male schizophrenia patients. Moreover, patients varied in predominant symptomatology. Therefore, the exact influence of severity of positive and negative symptoms on the different emotional functions is far from being elucidated and the small number of patients per subgroup limits the conclusions that can be drawn from the correlation analyses. This should be further highlighted in future studies measuring larger samples with mixed symptomatology.

All patients were medicated and correlation analyses did not reveal any significant association. However, we cannot exclude an influence of medication in our sample of patients. Most previous neuroimaging studies on emotion processing and empathy in schizophrenia reported that results remained significant when controlling for medication as covariate (e.g., Benedetti et al., 2009; Satterthwaithe et al., 2010).This supports findings from a meta-analysis showing no significant effect of medication on emotional experience in schizophrenia patients (Cohen and Minor, 2010) and results from several longitudinal studies (e.g., Harvey et al., 2006), that also observed no significant effect of drug treatment on emotion deficits in schizophrenia patients. Hence, deficits in socio-emotional competencies appear to be quite independent from pharmacological treatment.

Our study revealed hypoactivation in several regions that were either not observed or even showed hyperactivation in previous studies investigating empathic abilities (e.g., Benedetti et al., 2009; Lee et al., 2010) or sub-components such as emotion recognition (e.g., Habel et al., 2010) in schizophrenia. While Benedetti et al. and Lee et al. relied on a similar cartoon task we relied on a much more encompassing approach to assess empathic abilities using facial expressions (emotion recognition), situational pictures (perspective taking) and emotional sentences (affective responsiveness), thus using more realistic and probably more salient stimuli on one sample to enable a comprehensive coverage of the main empathic abilities. Due to the limited amount of trials per emotion (approx. 6 per emotion/neutral), we were not able to perform an emotion-specific analysis of neural responses as presented in Habel et al. (2010).

5. Conclusion

In the current study schizophrenia patients were confronted with diverse socio-emotional demands, which require salience attribution (e.g., amygdala), top–down modulation of affective responsiveness and self-reflection (e.g., anterior cingulate and middle cingulate cortex) and the ability to spontaneously simulate another person’s world (e.g., inferior frontal gyri), which are impaired in these patients. Moreover, behavioral and neural performance correlated significantly with negative symptom severity supporting recent assumptions that negative symptoms are particularly associated with dysfunctions in emotion processing. However, the exact influence of psychopathology on the different emotional functions is far from being elucidated and needs further research.

Furthermore, our findings support the assumption of a much broader emotional deficit, which is also characterized by neural abnormalities. A remediation of these emotional dysfunctions might improve patient’s everyday life (by reducing personal distress), social interactions and increase their opportunities for a better socio-occupational life, especially as social impairments in schizophrenia frequently worsen over the course of the disorder and probably contribute to the rate of relapse (e.g., Pinkham et al., 2003). Therefore, a more detailed understanding of the exact nature of these impairments is mandatory. The current study aimed at characterizing the empathy deficit in greater detail by identifying the neural substrate of subcomponents of empathic abilities in schizophrenia, as a core psychopathological dimension and thus, provides a target for the assessment and monitoring of patients during treatment.