The concept of alexithymia has been an important advance in our understanding of how dysregulated emotions may adversely affect physical health. The term “alexithymia” was coined by Sifneos [
1] with the intention of describing a deficit in the ability to identify and describe emotions that included a limitation in the ability to put emotions into words. Although there were forerunners of this concept in Ruesch’s [
2] “infantile personality,” Alexander’s [
3] “visceral neurosis,” and Marty and d’Muzan’s [
4] “pensée opératoire,” all of which involved limitations or impairments in the ability to symbolically represent emotions, the concept of alexithymia has come to dominate the field. Much credit for this goes not only to Sifneos and Nemiah for their clinical descriptions and early research but also to Graeme Taylor and his colleagues for creating what is now the leading measure of alexithymia, called the Twenty-Item Toronto Alexithymia Scale (TAS-20) [
5,
6]. In fact, in a recent paper attempting to expand upon the concept of alexithymia (which we will refer to as the “source” document), we described a variant of alexithymia called “affective agnosia” [
7]. In a surprising turn, Taylor and his colleagues [
8] stated (in what we will refer to as the “rebuttal” document) that our contribution did not advance the theory, measurement or treatment of alexithymia, and that the phenomenon in question had already been described. The purpose of the present paper is to correct certain potentially misleading statements in the rebuttal document, reaffirm our view that affective agnosia is a useful and novel concept, and explain in greater detail why we believe that to be the case. This explanation requires detailed consideration of the various points made by Taylor and colleagues. After addressing these points we will present our view of the way forward in alexithymia research.
Measurement issues
Many of the conclusions regarding the differing perspectives described above arise from the findings with the LEAS, a performance measure, and the TAS-20, a self-report measure. It is notable that both measures have strong psychometric properties. The LEAS asks respondents to describe in writing how they and another person would feel in 20 different emotion-evoking scenarios involving two people. Scoring is based on the number and differentiation of the words used to describe feelings. The internal consistency (Cronbach’s alpha = 0.88), inter-rater reliability (intra-class correlation of 0.96) [
21], intra-rater reliability over 2 years (Pearson correlation of
r = .99) and test-retest reliability at 4 weeks (spearman brown correlation = 0.8) [
22] are very good to excellent. Many studies support the construct validity of the LEAS in healthy volunteers, psychiatric disorders and systemic medical conditions [
7]. Moreover, multiple brain imaging studies [
23‐
29] and psychophysiological evidence [
30] also support the validity of the LEAS as a measure of emotional awareness.
The TAS-20 asks respondents to rate their degree of agreement on a 5-point scale with 20 statements divided into three factors (difficulty identifying feelings, difficulty describing feelings and externally oriented thinking). The TAS-20 similarly has excellent psychometric properties in terms of internal consistency, test-retest reliability, factorial validity and convergent and discriminant validity in student samples and clinical populations [
31].
A meta-analysis of 21 studies that used the LEAS and TAS-20 in the same participants reveals a low but statistically significant correlation between them of
r = − 0.12 [
17]. As noted above, this low correlation is likely attributable both to differences in methodology (self-report vs. performance measures) and content (meta-cognitive vs. direct measures of ability). Cultural factors could also be playing a role as the correlation between the two measures in a sample of 344 Japanese university students was negligible (
r = − 0.03, NS) [
32]. Although this meta-analysis did not address the relation between TAS-20 factors and the LEAS, one of the largest studies using both measures that also stratified for age, sex and socio-economic status revealed a somewhat stronger correlation with factor 3 (externally-oriented thinking) than the other factors [
33]. As discussed further below, externally oriented thinking is most plausibly associated with attentional processes in our more recently expanded “three-process model” of emotional awareness – which is the final “gatekeeper” in allowing the use of emotion concepts in conscious thought. Thus, it appears sensible that this factor correlates more strongly with the LEAS, as – even if individuals possessed granular emotion concept knowledge – they would not self-report granular emotions on the LEAS if they did not attend to internal (e.g., bodily) sources of information about emotions and use this information to guide their verbal responses. Indeed, it was only TAS-20 factor 3 that correlated with the LEAS in the Japanese student sample [
32].
In the remainder of this section we review statements and positions held by Taylor and colleagues in relation to measurement with which we disagree. Below we explain why we disagree and provide evidence in support of our position.
1.
A key claim by Taylor and colleagues is that the TAS-20 is the best measure of alexithymia and the LEAS does not measure alexithymia. We take issue with these claims on two primary grounds. First, the TAS-20 is a meta-cognitive measure of alexithymia, in that it assesses the respondent’s beliefs about the presence or absence of their own psychological deficits. Although there are significant advantages to a simple measure of this sort (such as ease of use and short completion time), which are not to be discounted, self-assessment is not equivalent to a direct performance measure of the ability to put emotions into words, which is central to the alexithymia construct.
A second reason for disagreeing is based on findings from the only study to date known to include the TSIA (which may be taken to be the leading objective measure of alexithymia), the TAS-20, and the LEAS in 87 German university students [
33]. Results showed that the TSIA-LEAS correlation was r = -0.47. By comparison, the original report on the development of the TSIA [
18] observed TAS-20-TSIA correlations of 0.36 in a community sample of 136 participants and 0.68 in a clinical sample of 97 participants. Subsequent published studies of the TSIA-TAS-20 correlation range from 0.23 to 0.52 (0.23 [
34], 0.34 [
35], 0.44 [
36], 0.49 [
37] and 0.52 [
38]). Clearly, many more studies are needed, especially in clinical populations. Nevertheless, in light of the greater content similarity of the TSIA and TAS-20, the evidence indicates comparability of the two measures relative to the gold standard of the TSIA rather than superiority of one over the other. Although the LEAS was not created to be a measure of alexithymia per se, it was developed to capture the developmental line along which impairment would give rise to alexithymia [
39,
40] and as such is foundational to the alexithymia construct. Of course, it should be acknowledged that the TSIA is directly measuring yet another phenomenon: how a clinician perceives an individual’s ability to recognize and express emotions, and their observed preference to attend to and focus thought on internal sources of emotional information. While this addresses many of the disadvantages of self-report measures in tracking the TAS-20 factors, and could also be more informed by the spontaneous use of emotion words measured by the LEAS, this approach might also fail to capture other relevant internal neuro-cognitive dynamics not directly observable in externally observable behavior within the context of a clinic. All three measures may therefore carry some uniquely valuable information.
2.
Taylor and colleagues assert that the LEAS is not a measure of alexithymia because it does not explicitly include pensée opératoire, or externally oriented thinking (TAS-20 Factor 3). Recent evidence indicates that Factor 3 reflects a limitation in the ability to focus one’s attention on one’s own emotions rather than a general bias toward focusing attention externally [
41]. These findings mean that if someone is limited in their capacity to attend to and focus on internal sources of emotional information, they would more likely be exteroceptively focused and tend not to spontaneously use granular emotion terns (as measured by the LEAS). This might be relevant in collectivist vs. individualist cultures [
42] and consistent with the observation in Japanese students that only Factor 3 (inversely) correlated reliably with the LEAS [
32]. One may conclude from this that the LEAS measures the output produced by participants who (to variable degrees) attend to their feelings, conceptualize/identify them, and then describe them, thus constituting a performance measure of the end product of the interactions between factors 3, 1, and 2 of the TAS-20, respectively, that ultimately determine an individual’s tendency to exhibit conscious awareness of their emotions. As such, the LEAS is arguably measuring performance that depends on the three skills that the TAS-20 aims to measure (beliefs about) through self-assessment.
3.
Taylor and colleagues assert that the TAS-20 is a comprehensive measure of the alexithymia continuum, it covers the extremes quite adequately, and thus
an independent objective measure of a severe deficit is not needed. Moreover, they assert that Factor 1 (difficulty identifying feelings) is an adequate measure of the affective agnosia phenomenon. In their rebuttal, they offered no evidence that TAS-20 Total or Factor 1 can capture severe alexithymia defined by independent objective criteria. Notably, we recently reported on a case of severe alexithymia [
43] in which the TAS-20 score was very high (and the LEAS score was very low); yet, the patient involved was highly aware of her deficit, which seems to be a requirement for accurate self-ratings on the TAS-20. Since the TAS-20 is a meta-cognitive measure and the LEAS is a performance measure, it would seem potentially useful to have both types of measures available given the low correlation between the two (i.e., especially in cases of low awareness of deficit in the presence of severe deficit, which could result in both low TAS-20 scores and low LEAS scores).
Taylor and colleagues argued that our study using “theory of mind” measures in relation to the TAS-20 and LEAS in patients with somatoform disorders, conversion disorders, and medical disorders with somatic symptoms [
44] was not relevant to the current discussion because cognitive theory of mind measures are independent of theory of mind related to emotional states. First, this is a debatable point. For example, large-scale neural networks associated with conceptualization and internal simulation processes, such as the default network, appear to do so in a domain-general manner [
45] – which would suggest at least partially overlapping mechanisms contributing to mental representation of emotional states, self-related cognition, and cognitive theory of mind. Further, current distinctions in the literature between “affective” and “cognitive” theory of mind for emotions (roughly corresponding to empathic sharing of another’s emotions vs. being aware of their emotions, respectively) have conceptually distinct relations to emotional awareness – with the latter being more directly linked to the ability to report awareness of others’ emotions [
44].
We would also highlight that the “Reading the Mind in the Eyes Test” [
46] included in that study involves the ability to accurately identify affective states from photos restricted to the eyes. In that study, the LEAS correlated positively and significantly with the Eyes Test, even after controlling for self-reported positive and negative affect on the Positive and Negative Affect Schedule (PANAS) measure [
47]. By contrast, while the zero-order correlation of the TAS-20 with the Eyes Test was significant, this significant correlation disappeared when controlling for positive and negative affect. Controlling for self-reported affect is a way to isolate variance due to difficulty in identifying and describing emotional states. These findings support the conclusion that the LEAS is of value in measuring impairment in the ability to mentalize emotional states and suggests that the TAS-20 alone may not be sufficient.
Taylor and colleagues argue that because alexithymia is a continuum, there is no reason to designate a separate concept for individuals at the extreme end of severity. Here we introduce a concept that we will elaborate upon below, which is that we believe that alexithymia, conceived fundamentally as a deficit in naming and mentally representing emotion, is actually a broad clinical phenomenon that encompasses a variety of different possible mechanisms that can generate phenotypically similar clinical manifestations. If one considers the analogy of hyperglycemia as a continuum, patients can manifest the full range of fasting blood glucose levels, but this continuum is not inconsistent with the fact that categorically different mechanisms involving insufficient insulin activity, Type 1 and Type 2 diabetes, can be at fault. Similarly, we argue below for intensified efforts to understand the various neural/cognitive mechanisms that lead to the phenotypic expression of what is called alexithymia.
4.
As noted above, Taylor and colleagues claim that
the covariation of TAS-20 and self-reported negative affect is not a problem. They point out that alexithymia involves dysregulated affect and as such should be expected to be associated with indicators of affective distress such as anxiety and depression. As we noted above, this is a valid point, but it is problematic in the context of examining the association between alexithymia and medical disorders, where self-reported negative affect is already a known predictor of adverse medical outcome [
20]. The concept of affective agnosia specifically captures the phenomenon that affective arousal that is persistent and undifferentiated may not be recognized as an emotional state and thus may not be reported as such (which at the extreme could reduce self-reported negative affect). Thus, it considers that arousal can still be present even though low self-reported affect would suggest that such physiological arousal is absent.
In addition to the findings regarding the Reading the Mind in the Eyes Test noted above, to date there are four other studies comparing clinical and control groups demonstrating significant group differences in TAS-20 scores in the predicted direction, but these differences in TAS-20 scores disappeared when controlling for self-reported negative affect. By contrast, comparable significant differences in these groups were observed with the LEAS that did not change when controlling for negative affect. For example, patients with essential hypertension were compared to patients with secondary hypertension due to other medical causes such as renal disease; patients with essential hypertension had higher TAS-20 and lower LEAS scores than the comparison group; when these differences were controlled for negative affect, the group difference in TAS-20 scores disappeared whereas the group difference in LEAS remained statistically significant [
48]. The very same pattern was observed in three additional studies comparing patients to healthy volunteers in separate clinical contexts including skin-restricted lupus erythematosus [
49], eating disorders [
50] and those receiving treatment on a psychosomatic medicine inpatient treatment unit [
51]. These results suggest that the LEAS is capable of detecting the influence of an impairment in emotion conceptualization in medical contexts, whereas the ability of the TAS-20 to do so is more uncertain and variable. Put in terms of underlying constructs, the tendency to use emotion concepts in thought (LEAS) may capture something uniquely important in these medical contexts that is not captured by self-reported beliefs pertaining to the (conceptually related) three TAS-20 factors.
5.
We also disagree about the extent to which
the TAS-20 is a good predictor of medical outcome in prospective studies. In the source document we stated that “research to date has provided only very limited evidence for an association between alexithymia as measured by the TAS-20 and medical outcomes.” One review [
52] revealed that “no prospective studies of TAS-20 as a predictor of medical outcome have been published.” Taylor and colleagues pointed out studies of the TAS-20 as a short-term predictor of pain and physical symptoms, one of which controlled for affective distress. There are also numerous cross-sectional studies using the TAS-20 in medical contexts [
53], which are of considerable value as they highlight medical contexts where affective deficits may be present. To our knowledge, however, there are still no long-term prospective epidemiological studies of TAS-20 as a predictor of systemic medical disorders after correcting for emotional distress, despite the relative ease with which the TAS-20 could be included in such studies. This is significant given that we believe that a measure of severe alexithymia (affective agnosia) could be such a predictor, based on the claims put forward above. Of course, this will need to be confirmed by such prospective studies using the LEAS as well, which to our knowledge have not yet been attempted.
In their rebuttal, Taylor and colleagues highlighted results of a 20-year follow-up study that began in the late 1980s [
54], in 2321 men in Finland using the version of the TAS developed prior to TAS-20, the TAS-26. Thus, the results do not address the statements made above about the TAS-20. The study in question [
55] revealed that the risk of cardiovascular death increased 1.2% for every 1-point increase in TAS-26 scores after controlling for a variety of variables, including depression. While this is certainly an important finding, there are some limitations worth considering. For example, the prevalence of alexithymia in this sample was 34% compared to a 17% prevalence rate of alexithymia in Finnish men using the TAS-20 [
56] – leading to potential issues regarding generalization. It is also unclear whether the higher prevalence of alexithymia could be related to 1) the use of the older TAS-26 measure, 2) a possible cohort effect of elevated alexithymia involving trauma related to World War II in Finnish men, or 3) the ability of older age to itself account for the high rates of alexithymia (which could be a different generational cohort effect).
Further light has been shed on the issue in a larger study in France that began in the 1990s [
57]. In contrast to the Finnish study, no participants had heart disease at baseline. In this study, 5586 participants (2312 men) were followed for an average of 8.9 years. Results showed that TAS-26 scores did not predict cardiovascular events. Thus, the generalizability of the finding in the Finnish study is open to question. Of course, it should be acknowledged that many current studies using the LEAS also have potential limitations regarding generalizability, and a greater amount of cross-cultural and longitudinal studies are necessary in this field generally.
Future research on affective processing deficits in alexithymia
Since our source article on affective agnosia [
7], our own thinking on this issue has advanced. In particular, we put forward the “Three-Process Model” of affective processing [
58] that extends and elaborates on the concept of affective agnosia. Specifically, we proposed that there are three inter-related (but partially dissociable) neuro-cognitive processes that can contribute to individual differences in emotional awareness - affective response generation, internal state representation (conceptualization), and cognitive control processes that modulate conscious accessibility - and that each can break down in multiple ways resulting in low emotional awareness. The affective agnosia model highlighted a deficit in conceptualization, a memory-dependent process that is the basis for categorization, that in turn leads to deficits in recognition. The three-process model highlights that a lack of affective (viscero−/skeleto-motor) response generation or overly coarse-grained affective responses could also contribute to a phenotypic lack of awareness of emotional experience. The latter was highlighted in the work of Vorst and Bermond [
59] who emphasized the need to include a deficit in affective arousal as a possible factor contributing to alexithymia. The third process, cognitive control, includes externally-focused attention that could be associated with a deficit in internally focused attention to emotion [
41]. Deficits in this third process can also involve state-dependent constraints in the ability to hold information about emotions in working memory (i.e., constraining explicit deliberation and verbal reporting), which could occur as a result of high levels of arousal (the first process) and an associated decline in prefrontal cortex function [
60]. The third process could also include avoidance strategies such as distraction or psychological defenses, as well as other attention-related factors. As such, the three-process model embraces the content addressed by the different factors of the TAS-20 and the Bermond–Vorst Alexithymia Questionnaire but does so in a way that allows for objective rather than self-report assessments.
We recently published a detailed case report of a woman with severe alexithymia using this framework that suggested contributions of each of these three factors [
43]. While the LEAS can be influenced by differences in each process, other measures may allow them to be separated. For example, affective response generation can be objectively measured using peripheral physiological responses such as skin conductance or facial EMG responses as well as self-reported valence and arousal ratings to standard affect-eliciting stimuli such as International Affective System Pictures [
61] (although these self-reports also depend on subsequent representation). Conceptualization ability can be measured using scales of emotional understanding [
62] and emotion granularity [
63]. Cognitive control can be assessed using TAS-20 Factor 3 (externally oriented attention), measures of cognitive suppression [
64] and a measure of defensiveness [
65] consisting of the combination of the Marlowe-Crowne [
66] and Bendig Short Form of the Taylor Manifest Anxiety Scale [
67]. These are simply examples and other measures in each category are certainly possible. For example, in a recent computational modeling study, we simulated dynamics within the three-process model (discussed further below) and demonstrated ways in which measures of task reaction times and neuroimaging measures could further distinguish different abnormal mechanisms [
68].
An advantage of this approach is that these three domains of objective assessment can be mapped onto different brain networks. In our source paper we focused on communication between subcortical affective response generation structures, peripheral bodily states, insula-based body state representation structures (among other relevant regions such as somatosensory cortex and anterior cingulate), medial prefrontal/default mode network structures involved in conceptualization (including conceptualization of valenced bodily states), and frontal-parietal structures associated with top-down cognitive influences. Establishing such correspondences between psychological characteristics and their neuroanatomic substrates advances the opportunities to identify brain mechanisms that influence bodily states. At the same time, detailed studies of relevant psychological processes can be undertaken corresponding to each of the three processes, such as the role of episodic and semantic memories in determining how individuals consciously or unconsciously interpret/evaluate/appraise the situations they’re in, which then affect the nature and granularity of affective responses [
69], the role of language and memory in categorization and conceptualization [
70], and the ways that attention can be deployed or redirected to enhance or avoid emotion experiencing [
71]. Based on the principle that autonomic regulation is under hierarchical control [
72], the relative contribution of each level of deficit could influence the bodily expression, pathophysiological potential and disease risk for any given individual. As described in our source paper, this is consistent with a specific model whereby the mental representation deficit associated with affective agnosia involves impairment or lack of engagement of the medial prefrontal cortex; relative inactivity in this region is plausibly related to reductions in parasympathetic influence (as indexed by heart rate variability) which can in turn disinhibit inflammatory disease-promoting processes [
73]. This framework could also potentially make it possible to identify individuals independent of self-report who phenotypically manifest affective agnosia because of a deficit in one of these processes alone or a combination of two or more.
By extension to the neurological literature, one may hypothesize the existence of additional subtypes of affective agnosia. As we discussed in the source document, a distinction is drawn in neurology between ‘apperceptive’ and ‘associative’ agnosia [
74].
Apperceptive agnosia involves failure to formulate a distinct exteroceptive percept. The analog of this in our model would be a failure to have a clearly defined bodily sensation or percept mediated by the insula (and interconnected somatosensory/proprioceptive processing cortices). To our knowledge, research has not addressed variation in how well-formed bodily percepts are. One way of assessing this independent of language is to have people draw the location and contours of their bodily experiences or somatic symptoms using a cell phone-based app [
75] or other tools currently used to gather data on felt distributions of bodily sensations during emotion [
76]. One can also imagine provoking somatic sensations with drugs such as isoproterenol [
75] or exertion and comparing the differentiation and complexity of verbal descriptions of bodily sensations in the resting/placebo vs. activated states.
Associative agnosia is a disconnection disorder; it involves a normal percept “stripped of meaning.” The analog of this in our model would be a failure to connect the affective bodily state to stored conceptual knowledge about it. The mechanism here would involve a failure in the functional interaction between the aforementioned structures representing bodily percepts and the medial prefrontal cortex (and associated default network structures) where conceptual knowledge about emotional states may be integrated with this perceptual information. There could also be a deficiency of stored knowledge about emotional states – which could be due to a failure to acquire emotion concept knowledge during development. Depending on its origin, this might be loosely analogous to what neurologists call ‘semantic dementia.’ It would be interesting, for example, to undertake a study of patients with somatic symptom disorders who appear to be overly focused on their bodily sensations. One can imagine, using this framework, that any of these possibilities might be observed in different cases. Could it be that a common problem in this context is that patients have vivid bodily sensations, and may have quite adequate stored conceptual knowledge about emotions, but fail to connect the two? The treatment strategies might be quite different depending upon what was found. All of this illustrates the heuristic potential of the affective agnosia concept.
The element of cognitive control in the three-process model also raises another very important point touched upon by Taylor and colleagues, who refer to alexithymia as a deficit in the ability to cognitively “elaborate” on experience. It is difficult to tell exactly what this means in relation to emotional experience. One interpretation is that individuals with severe alexithymia have the same emotional experience (in terms of range, differentiation and complexity) as do healthy individuals but they simply have a problem attending to or reporting on it – they in some sense “fail to formulate the words” in particular contexts despite having the conceptual resources to do so. This could represent one possible mechanism leading to the anomia version of alexithymia discussed above (however, this mechanism would be attention-dependent and could also hinder the use of emotion concept knowledge). A second interpretation is that the emotional experience and conceptualization ability of individuals with severe alexithymia is actually quite impoverished and limited relative to that of healthy individuals. This corresponds to the alternative view of Nemiah and Sifneos [
10] that these patients “may lack the basic human ability to experience feelings.” Thus, in the second interpretation their impairment in elaboration is a consequence of deficient internal state representation.
This second interpretation is consistent with newer constructivist views of emotion and emotional experience that hold that concepts (and the language used to refer to them) aren’t mere markers of experience but are constitutive of it and able to modulate it in important ways [
70,
77]. The analogy to wine-tasting illustrates this point. Does a fine wine taste the same to a novice and an expert? Most people would say no. The expert not only has more words to describe the experience but actually recognizes more nuances in the experienced taste sensations and flavors than the novice [
78]. This distinction is very important in the emotion domain as it corresponds to the distinction between agnosia and anomia. Although Taylor and colleagues, and their psychoanalytic predecessors, pointed out that alexithymic individuals lacked words and had impairments in mental representation, the key issue is that one or both could be present in any given individual – one can have anomia with or without agnosia or mental representation deficits. Therefore, it is important to make these distinctions so that accurate assessments can be made of the specific deficits that a given individual may have, and treatment strategy can be adapted accordingly. Standard treatment approaches that promote mental representation of emotion by encouraging patients to describe their feelings in a routine manner will likely be insufficient if a severe deficit of the kind described here is present.
As briefly mentioned above, we recently published a computational model [
68,
79] and simulated how variations in the three processes comprising the three-process model can result in phenotypically similar deficits in emotional awareness due to different mechanisms. This basic premise is consistent with evidence that different mechanisms can lead to the same phenotype of high scores on the TAS-20 [
80]. The simulation revealed that the same low emotional awareness phenotype could be due to seven distinct mechanisms. In addition, we proposed that different clinical treatment strategies could be used to address each of these distinct mechanisms. These include: 1) maladaptive affective response generation, treated by a variety of methods including cognitive therapy [
81], exposure therapy [
82] or acceptance and commitment therapy [
83]; 2) somatically biased expectations when interpreting affective bodily sensations, treated by psychoeducation about the bodily expression of emotion or “focusing,” which involves deriving the emotional meaning of bodily sensations in context [
84]; 3) poor emotion concept acquisition, treated by teaching emotion concepts or emotion-focused psychotherapy [
85]; 4) biased attention (failing to attend to either context or bodily sensations), treated by mindfulness-based approaches [
86] or cognitive-behavioral therapy [
87]; 5) high emotional volatility, treated by dialectical behavioral therapy [
88] or emotion regulation therapy [
89]; 6) conscious inaccessibility of emotional information, treated by overcoming cognitive avoidance tendencies [
90] or psychodynamic psychotherapy (analysis of defense) [
91]; and 7) reduced reflective capacity, treated by mentalization-based therapy [
92,
93] or teaching emotion regulation strategies [
89]. Note, the therapies just mentioned are just examples. It is not meant to be an exhaustive list, and many interventions could help with more than one mechanistic problem. In general, however, given that assessment methods exist to identify these mechanisms, this new perspective provides new opportunities for treating alexithymia or low emotional awareness by targeting deficits in a personalized way that can improve clinical outcomes in this area. In our view, this provides the ultimate justification for proposing the concept of affective agnosia and other variations within the alexithymia spectrum.
These considerations illustrate that there is potential utility in identifying different subtypes of alexithymia or affective agnosia that manifest in specific individuals based on objective assessment of deficits. Perhaps the concept of alexithymia could be refined and expanded to include the different types of emotion processing deficits manifested across the three processes, not captured by other well-defined syndromes, that may be present to variable degrees and perhaps in different ways in different psychiatric disorders. This is consistent with the approach taken by van der Velde and colleagues [
94], in which different facets of alexithymia are thought to confer risk for different types of psychopathology. This same approach can be used to understand how different aspects of alexithymia could potentially heighten risk for various systemic medical disorders. In an intellectual environment such as this it would seem prudent to undertake large scale studies of the psychological, behavioral, and neural mechanisms in affective processing in a large number of healthy individuals and patients to develop a potentially new taxonomy of emotion processing deficits. This research could capitalize on the many advances that have been used in recent decades in the study of the mechanisms and functional consequences of alexithymia. These include, but are not limited to, structural and functional neuroimaging techniques, autonomic and other psychophysiological measures, basic cognitive functions including encoding and retrieval of episodic and semantic memories, priming and implicit processing, working memory, linguistic abilities, conceptualization and categorization skills, and executive functioning, as well as basic bodily and affective functions including interoceptive abilities, affect labeling, emotion recognition and emotion knowledge [
95,
96]. Such research would also provide a context for exploring variations as a function of age, sex, socio-economic status, years of education [
33], cultural/ethnic background as well as the contributions of genetics, the early childhood environment (e.g. neglect, abuse, attachment style) and their interaction to deficits of different kinds [
53].