The authors have declared that no competing interests exist.
Conceived and designed the experiments: FS MS PS TK. Performed the experiments: FS MS. Analyzed the data: EB JE CHA SG FS FB. Contributed reagents/materials/analysis tools: FS TK. Wrote the paper: EB JE CHA FB AH.
While hemispheric specialization of language processing is well established, lateralization of emotion processing is still under debate. Several conflicting hypotheses have been proposed, including right hemisphere hypothesis, valence asymmetry hypothesis and region-specific lateralization hypothesis. However, experimental evidence for these hypotheses remains inconclusive, partly because direct comparisons between hemispheres are scarce.
The present fMRI study systematically investigated functional lateralization during affective stimulus processing in 36 healthy participants. We normalized our functional data on a symmetrical template to avoid confounding effects of anatomical asymmetries. Direct comparison of BOLD responses between hemispheres was accomplished taking two approaches: a hypothesis-driven region of interest analysis focusing on brain areas most frequently reported in earlier neuroimaging studies of emotion; and an exploratory whole volume analysis contrasting non-flipped with flipped functional data using paired t-test.
The region of interest analysis revealed lateralization towards the left in the medial prefrontal cortex (BA 10) during positive stimulus processing; while negative stimulus processing was lateralized towards the right in the dorsolateral prefrontal cortex (BA 9 & 46) and towards the left in the amygdala and uncus. The whole brain analysis yielded similar results and, in addition, revealed lateralization towards the right in the premotor cortex (BA 6) and the temporo-occipital junction (BA 19 & 37) during positive stimulus processing; while negative stimulus processing showed lateralization towards the right in the temporo-parietal junction (BA 37,39,42) and towards the left in the middle temporal gyrus (BA 21).
Our data suggests region-specific functional lateralization of emotion processing. Findings show valence asymmetry for prefrontal cortical areas and left-lateralized negative stimulus processing in subcortical areas, in particular, amygdala and uncus.
The two hemispheres of the brain differ in structure and function. While hemispheric specialization of some cognitive domains such as language is well established
According to the valence asymmetry hypothesis
Discrepancy between neuroimaging findings may, in parts, be related to a methodological issue
Direct comparison between hemispheres has been provided by a pioneering meta-analysis on emotion processing
The aim of the present study was to investigate functional lateralization by directly comparing BOLD responses to affective stimuli between hemispheres. A homogenous sample of 36 healthy participants underwent a standard affective stimulus processing paradigm for fMRI (International Affective Picture System)
Thirty-six right-handed
During fMRI participants viewed standardized photographs taken from the
MR images were acquired on a 1.5-Tesla scanner (Magnetom VISION Siemens®) with an Echo Planar Imaging (EPI) sequence (TE = 40 ms, TR = 2,3 s, α = 90°, matrix = 64×64, voxel size = 4×4×3,3 mm3). A total of 290 T2*-weighted images were acquired per run. Additionally, a 3D Magnetization Prepared Rapid Gradient Echo (TR = 9.7 ms, TE = 4 ms, flip angle 12°, matrix = 256×256, voxel size = 1×1×1 mm3) image data set was acquired.
FMRI data were analyzed using SPM 8 (Wellcome Department of Imaging Neuroscience, London). The functional images were slice time corrected and realigned to the mean functional image
For first level analysis, flipped and non-flipped images were analyzed separately in the context of the general linear model approach, using the onset of each picture for a box-car function (2 s stimulus duration) to provide a stimulus function. The stimulus functions were convolved with the canonical hemodynamic response function as implemented in SPM 8
Non-flipped functional images were used for the region of interest analysis. For the whole volume analysis, flipped and non-flipped functional images were compared using a paired t-test.
Regions of interest (ROIs) were chosen based on the frequency of reported activations in earlier neuroimaging studies of emotion, as reviewed by Phan et al.
The VOI tool (SPM 8) was used to extract the mean contrasts of parameter estimates (‘positive > neutral’, ‘negative > neutral’) of each ROI for each participant separately in the right and the left hemisphere. ‘Neutral picture viewing’ was chosen as control condition because it allows subtracting the general effect of picture viewing thus isolating the specific effect of positive or negative affective picture viewing. A paired t-test was applied to compare effects between hemispheres (SPSS 18). For the ROI analysis, results with
Given the small size and the shape of some of the ROIs (e.g., the amygdala) as well as the voxel size of 4×4×3,3 mm3 and the smoothing kernel of 15 mm, it is acknowledged that our ROI analysis does not allow to completely isolate activation in small ROIs from activation in adjacent structures (e.g., hippocampus and parahippocampal gyrus).
For exploratory purposes, a whole volume second-level random effects analysis was carried out, comparing non-flipped with flipped functional data (paired t-test). This allowed the direct comparison of BOLD responses between hemispheres. Statistical parametric maps were estimated for the contrasts ‘positive > neutral picture viewing’, ‘negative > neutral picture viewing’, and ‘negative >positive picture viewing’, comparing right with left hemispheric functional data. Again, picture conditions were chosen as control conditions to subtract out general effects of picture viewing. For the whole-volume analysis, statistical significance threshold was set to
To explore the impact of the control condition on lateralization effects, additional whole volume analyses were carried out contrasting positive and negative picture conditions with baseline (‘positive picture viewing > baseline’, ‘negative picture viewing > baseline’).
Response times for button responses to pictures during fMRI were 911 ms (±377, S.D.), 936 ms (±385), and 912 ms (±357) for positive, negative, and neutral pictures, respectively. Response times showed no valence effect (
The mean
The mean
Results of the ROI analysis are shown in
Lateralization of BOLD responses in regions of interest. The bars represent hemispheric differences of contrasts of parameter estimates (red: ‘positive > neutral’; blue: ‘negative > neutral’) extracted from the respective ROI (mean values for the respective ROI, averaged across the 36 participants). Bars extending to the left indicate lateralization towards the left (left > right), and bars extending to the right indicate lateralization towards the right (right > left). *
For exploratory purposes, a SPM whole volume analysis was carried out to directly compare BOLD responses between the right and the left hemisphere. A paired t-test was used to compare non-flipped versus flipped functional images that were normalized on a symmetrical template. Comparing ‘positive > neutral pictures’ (
Whole volume analysis. Brain areas showing lateralization (‘left > right’, ‘right > left’) during positive (versus neutral) picture viewing (‘pos > neu’); SPM whole volume analysis (paired t-test) comparing non-flipped with flipped functional data. Because data represent comparisons between left and right hemispheres, sections views display half of the brain only:
Anatomical Region | Hemisphere | Coordinates (MNI) | Max T Value | Cluster Size (k) | ||
X | Y | Z | ||||
Temporo-occipital junction (inferior & middle temporal gyrus; middle occipital gyrus; BA 19, 37)* | Right > Left | 48 | −64 | 1 | 5.70 | 91 |
Premotor cortex (precentral gyrus, BA 6) | Right > Left | 48 | −7 | 37 | 4.17 | 24 |
Medial prefrontal cortex (superior & medial frontal gyrus; BA 9) | Left > Right | −12 | 53 | 22 | 3.51 | 5 |
Maximum t-values and peak voxel coordinates for activation clusters, uncorrected
BA = Brodmann Area, k = number of voxels.
In the comparison ‘negative > neutral pictures’ (
Brain areas showing lateralization (‘left > right’, ‘right > left’) during negative (versus neutral) picture viewing (‘pos > neu’); SPM whole volume analysis (paired t-test) comparing non-flipped with flipped functional data. Because data represent comparisons between left and right hemispheres, sections views display half of the brain only:
Anatomical Region | Hemisphere | Coordinates (MNI) | Max T Value | Cluster Size (k) | ||
X | Y | Z | ||||
Dorsolateral prefrontal & premotor cortex (inferior & middlefrontal gyrus, precentral gyrus; BA 9, 46, 6) | Right > Left | 48 | 17 | 31 | 4.93 | 103 |
Temporo-parietal junction (posterior superior & middletemporal gyrus, angular gyrus; BA 37, 39, 42) | Right > Left | 45 | −40 | 10 | 4.44 | 55 |
Middle temporal gyrus (BA 21) | Left > Right | −60 | −10 | −11 | 4.10 | 11 |
Amygdala, parahippocampal gyrus | Left > Right | −15 | −7 | −17 | 4.06 | 11 |
Maximum t-values and peak voxel coordinates for activation clusters, uncorrected
BA = Brodmann Area, k = number of voxels.
The comparison ‘negative >positive pictures’ (
Brain areas showing lateralization (‘right > left’) during negative versus positive picture viewing (‘neg >pos’); SPM whole volume analysis (paired t-test) comparing non-flipped (right-sided) with flipped (left-sided) functional data. Because data represent comparisons between left and right hemispheres, sections views display half of the brain only:
Anatomical Region | Hemisphere | Coordinates (MNI) | Max T Value | Cluster Size (k) | ||
X | Y | Z | ||||
Caudate nucleus | Right > Left | 3 | −1 | 1 | 3.80 | 12 |
Dorsolateral prefrontal cortex (middle & inferiorfrontal gyrus; BA 9, 46) | Right > Left | 48 | 17 | 28 | 3.72 | 20 |
Posterior superior temporal gyrus (BA 41) | Right > Left | 51 | −43 | 10 | 3.60 | 9 |
Maximum t-values and peak voxel coordinates for activation clusters, uncorrected
BA = Brodmann Area, k = number of voxels.
To explore how the control condition chosen impacts on lateralization of emotional picture processing, we performed an additional analysis using the baseline (instead of the neutral picture condition) as a control. In these baseline contrasts, both positive and negative picture viewing were generally associated with lateralization towards the right hemisphere. The contrast ‘positive picture viewing > baseline’ revealed lateralization towards the right in the lateral prefrontal cortex (BA 45, 46), the occipital lobe (BA 18, 19) and the lateral parietal lobe (BA 7, 39) extending to the temporo-parietal junction (BA 37, 39); and lateralization towards the left in the postcentral gyrus (BA 2, 3). The contrast ‘negative picture viewing > baseline’ showed lateralization towards the right in the dorsolateral prefrontal cortex (BA 9, 46), the occipital lobe (BA 19) and the lateral parietal lobe extending to the temporo-parietal junction (BA 7, 39, 40, 21, 22); and lateralization towards the left in the orbitofrontal cortex (BA 11, 47).
The present fMRI study provides evidence for hemispheric asymmetry of affective stimulus processing in healthy participants. However, our present data do not generally associate positive and negative emotions with one or the other hemisphere. Instead, the pattern of lateralization differs between brain regions. Specifically, positive stimulus processing is lateralized towards the left in the medial prefrontal cortex; and towards the right in the premotor cortex and temporo-occipital junction. Negative stimulus processing shows lateralization towards the left in the amygdala, uncus and middle temporal gyrus; and lateralization towards the right in the dorsolateral prefrontal cortex (extending to the premotor cortex) and temporo-parietal junction.
The present study adds to the literature in that it places particular emphasis on several issues relevant for the investigation of functional lateralization. First, we directly compare BOLD responses measured in one hemisphere with the corresponding responses in the other hemisphere. Such direct comparison is necessary to demonstrate statistically significant lateralization and is preferable to an approach that argues for asymmetry when voxels in one hemisphere exceed statistical threshold while homologous voxels in the opposite hemisphere do not
The valence asymmetry hypothesis of emotion posits that the left hemisphere is dominant for positive and the right for negative emotions. The hypothesis has found particular support from lesion, EEG, TMS and functional neuroimaging findings in the prefrontal cortex
Extending earlier prefrontal findings, our data suggest that lateralization of emotion processing might differ between the medial (BA 10) and the dorsolateral prefrontal (BA 9, 46) cortex: Positive stimuli induce left-lateralization in the medial prefrontal cortex, but no significant lateralization in the dorsolateral prefrontal cortex. In contrast, negative stimuli provoke right-lateralization in the dorsolateral prefrontal cortex, but no significant lateralization in the medial prefrontal cortex.
Not consistent with the valence asymmetry hypothesis (in the above version), the amygdala and uncus show left-lateralization during negative stimulus processing, rather compatible with earlier suggestions that negative emotions are lateralized towards the left in limbic brain areas
Our main analyses used neutral picture viewing as a control for positive and negative picture conditions. Strikingly different findings were obtained when we used the blank screen baseline condition as a control in the whole volume analysis. Specifically, both positive and negative picture viewing showed lateralization towards the right in several cortical areas (including the lateral prefrontal cortex, the temporo-parietal junction and the occipital lobe) when contrasted with the blank screen baseline. At first glance, these findings seem to support the right hemisphere theory of emotion
Due to its important role in emotion processing, several earlier fMRI studies have searched for lateralization effects in the amygdala. Findings have been inconsistent, even in studies directly testing for condition-by-hemisphere interaction: Some studies report lateralization towards the left
The present study investigated functional lateralization of emotion processing in healthy subjects. Methodologically, particular emphasis was placed on the direct comparison between left and right hemispheric functional data, the systematic and comprehensive investigation of relevant brain areas (through both hypothesis-driven ROI and exploratory whole brain analysis), and the normalization of functional data to a symmetrical template. This approach revealed region-specific lateralization during passive IAPS picture viewing. Specifically, our data suggest valence asymmetry in prefrontal cortical areas and left-lateralized negative stimulus processing in subcortical brain areas, in particular amygdala and uncus. However, the pattern of lateralization observed here during passive IAPS picture viewing may not generalize to all forms of emotion processing. The approach taken here to study lateralization could be useful for further investigation of factors that influence lateralization of emotion processing. Such factors might include stimulus material (e.g., visual, auditory, olfactory), stimulus duration and level of awareness (e.g., visible, invisible). Lateralization may also be influenced by induction method (e.g., perception versus imagery), cognitive demand (e.g., passive viewing versus judgment task), and social content (e.g., emotional stimuli with versus without social information). In addition, lateralization might differ between approach- and withdrawal-related emotions (as opposed to positive and negative)