This study investigated the effects of self-report (TAS-20-DDF) and observer-rated (TSIA-DDF) facets of alexithymia on the labeling and neural processing of facial emotions presented for a rather long time (1 or 3 seconds). Our analysis of the main contrasts revealed significant clusters of brain activation in the fusiform gyrus, inferior and middle occipital gyrus (all conditions), in the middle temporal gyrus (fearful faces), inferior (fearful) and orbital and medial (happy) frontal gyrus as well as the cerebellum. All of these regions have been reported to be implicated in facial emotion processing (e.g.: [7, 8, 40–42]). Thus, we can assume that our experimental design is suitable for eliciting brain activation related to facial emotion recognition. Considering the specific effects of alexithymia, we found that high TSIA-DDF scores were related to increased reaction times when labeling angry and fearful faces and to increased brain activation in SMA and right S1 during the recognition of these negative faces. A post-hoc exploratory analysis suggests that activity in brain areas that are important in the affective components of facial emotion processing (AMG, vmPFC, striatum) does not show a particular relationship with alexithymia in the current task.
Their increased reaction times indicate that alexithymic individuals were slower in labeling negative emotions. Highly alexithymic individuals appear to need more time to reach a labeling accuracy level similar to subjects with low alexithymia. In contrast to previous studies describing a relationship between accuracy and degree of alexithymia [12, 13], we used relatively long stimulus presentation times and response windows and could not reveal interrelationships between alexithymia and recognition accuracy. Thus, it seems that alexithymic individuals have difficulties in recognizing facial expressions, which are reflected in decreased accuracy when presentation times and response windows are short (see also ). Prolonging presentation times and response windows could improve recognition accuracy, however, at the cost of increases in response time.
SMA is part of a brain network that is involved in the processing of motor-related information and motor preparation and has been shown to be involved in the production of facial emotions . Moreover, it has been argued that (especially pre-) SMA is involved in the recognition of facial emotions  by playing an important role in the motor components of simulation (see also ). Additionally, a cluster in S1 was revealed, which seems to reflect somatosensory aspects of facial emotion processing [3, 7, 45]. According to Adolphs et al. , recognizing emotions from facial expressions requires right primary somatosensory areas. The authors argue that recognition of another individual’s emotional state is mediated by internally generated somatosensory representations that simulate how the other individual would feel when displaying a certain facial expression. Taken together, this mediation could mean that highly alexithymic individuals have difficulties in automatically reenacting the negative facial emotion of others when these are presented briefly . When the presentation time is increased, highly alexithymic individuals can reach a similar performance as less alexithymic individuals, which seem to require an increased activation of motor and somatosensory areas. Interestingly, it has been found that highly (as compared to less) alexithymic individuals also show increased activation in motor-related brain areas when interpreting the directed actions of others in a classical mirror-neuron task and show no differences in interpreting these actions . Thus, highly alexithymic individuals may be more inclined to imitate the actions of others via (covert) motor simulation than are non-alexithymics. A recent meta-analysis by van der Velde et al.  reported that high levels of alexithymia are related to decreased activity in the SMA when participants are confronted with negative stimuli. However, this meta-analysis included all types of emotional paradigms and tasks (not only facial emotion recognition), so the published results may not necessarily reflect processes related specifically to facial emotion recognition.
There seems to be no particular relationship between activity in the amygdala, vmPFC and ventral striatum and alexithymia in the task studied here. This finding is very interesting because earlier studies on brain function [49–52] and structure  reported alterations in highly alexithymic individuals in these regions. In particular, functional studies on automatic processing of emotional faces (affective priming) [49–51] have revealed decreased activations in these brain areas. The lack of involvement in the current task may be the case because the emotional faces were presented for a rather long time in the current study. The amygdala and the ventral striatum, however, are thought to operate in a fast and automatic fashion and may be less relevant when the participants are fully aware of the emotional nature of the faces (e.g., [54, 55]), as in the current study. Thus, it seems that alexithymic individuals show less automatic activation in brain regions particularly involved in the affective components of face processing (AMG, ventral striatum, vmPFC), which most likely leads to alterations in the processing of and difficulties in the labeling of briefly presented faces. However, alexithymic individuals seem to be able to simulate the bodily aspects of facial expressions when the presentation times and response windows are long enough, which makes the correct recognition of faces possible in this case.
Our study points to deficits limited to the recognition of negative faces in alexithymia. Neither behavioral nor neurobiological differences were revealed for happy faces. This finding suggests that alexithymics have fewer problems interpreting positive compared to negative facial expressions. A recent review on alexithymia and the processing of emotional facial expressions concluded that the difficulties of alexithymic individuals in processing facial emotions are not specific to certain emotions . The work of Sonnby-Borgström  shows that the imitation of facial expressions (measured with facial EMG) in highly alexithymic individuals was only decreased for corrugator activity related to negative emotions, but not for zygomaticus activity related to happy faces. Against this background, alexithymic individuals may display fewer deficits in automatically simulating happy faces compared to neutral ones, which possibly renders the recognition of happy faces easier.
It is important to note that in our study, the objective measure of alexithymia (TSIA), but not the self-report measure (TAS-20), was predictive for recognition performance. Because some alexithymic individuals may not be aware of their own deficits, self-report tests could be less suitable for measuring difficulties in describing feelings compared to objective tests such as the TSIA.
It has been argued that the TAS-20 and the TSIA only measure cognitive aspects of alexithymia and neglect affective parts of the alexithymia construct . A questionnaire that possibly captures these affective components is the Bermond-Vorst-Alexithymia Questionnaire (, but see also ). It is possible that additionally applying this measure of alexithymia may have the potential to discover relationships between the brain areas involved in the affective components of emotional face processing. Future studies need to be conducted to determine whether the results of the current study are only related to cognitive alexithymia or whether they generalize to affective alexithymia as well.