Social cognition refers to the mental operations underpinning social interactions, including the perception, encoding, storage, retrieval and regulation of information about oneself and others (Barkl et al., Reference Barkl, Lah, Harris and Williams2014; Brothers, Reference Brothers1990; Gao et al., Reference Gao, Zhao, Liu, Liu, Yang and Xu2021; Green et al., Reference Green, Horan and Lee2015, Reference Green, Penn, Bentall, Carpenter, Gaebel, Gur, Kring, Parl, Sileverstien and Heinssen2008). It is a multifaceted construct entailing five main subdomains: theory of mind (the ability to infer other people’s intentions, inclinations and beliefs), social perception (the identification of social context, roles and rules), attributional bias (the tendency to attribute causes of events to external situations or other’s actions), social knowledge (the awareness of roles, rules and goals that characterise social situations and guide social interactions) and emotion processing (the ability to correctly perceive and use emotions) (Green et al., Reference Green, Penn, Bentall, Carpenter, Gaebel, Gur, Kring, Parl, Sileverstien and Heinssen2008). These processes require the ability to infer the emotions and thoughts of others (Green et al., Reference Green, Horan and Lee2015). To do so, critical information is provided by facial expressions (Barkl et al., Reference Barkl, Lah, Harris and Williams2014; Fusar-Poli et al., Reference Fusar-Poli, Placentino, Carletti, Landi, Allen, Surguladze, Benedetti, Abbamonte, Gasparotti, Barale, Perez, McGuire and Politi2009), which convey emotional states and influence the generation and regulation of emotions and behaviour in response to these signals; thus, the accurate reading of expressions is crucial for affective communication and emotional bonding (Ekman, Reference Ekman1993).

A link has been established between deficits in facial emotion recognition, social and emotional functioning (Yoo and Noyes, Reference Yoo and Noyes2016), which not only contribute to the presence of mood alterations (Oldehinkel et al., Reference Oldehinkel, Hartman, Van Oort and Nederhof2015; Vrijen et al., Reference Vrijen, Hartman and Oldehinkel2016) but also have negative implications for subsequent treatments (Shiroma et al., Reference Shiroma, Thuras, Johns and Lim2014). Indeed, several mental disorders, including schizophrenia (Addington et al., Reference Addington, Saeedi and Addington2006; Green et al., Reference Green, Horan and Lee2015) and psychotic disorders (Benito et al., Reference Benito, Lahera, Herrera, Muncharaz, Benito, Fernández-Liria and Montes2013; Ulusoy et al., Reference Ulusoy, Gülseren, Özkan and Bilen2020), are characterised by deficits in facial emotion recognition.

There is robust evidence that people with affective and non-affective psychoses show impairments in emotion perception (Edwards et al., Reference Edwards, Jackson and Pattison2002; Kohler et al., Reference Kohler, Walker, Martin, Healey and Moberg2010; Priyesh et al., Reference Priyesh, Suryavanshi, Sasidharan, Bhandary, Behere and Nayak2022; Rocca et al., Reference Rocca, Heuvel, Caetano and Lafer2009), experience reduced interpersonal skills (Pinkham et al., Reference Pinkham, Penn, Perkins, Graham and Siegel2007) and report social and work difficulties (Addington et al., Reference Addington, Saeedi and Addington2006; Kee et al., Reference Kee, Green, Mintz and Brekke2003). A recent meta-analysis revealed a specific distinction between the two patient groups. Indeed, individuals with affective psychosis demonstrated a greater ability to identify emotional facial expressions compared to non-affective psychosis patients, in particular for emotions of anger, fear and sadness (De Prisco et al., Reference De Prisco, Oliva, Fico, Montejo, Possidente, Bracco, Fortea, Anmella, Hidalgo-Mazzei, Fornaro, de Bartolomeis, Serretti, Murru, Vieta and Radua2023). Importantly, the severity of psychotic symptoms correlates with deficits in emotion processing (Kohler et al., Reference Kohler, Bilker, Hagendoorn, Gur and Gur2000; Schneider et al., Reference Schneider, Gur, Gur and Shtasel1995), which can be even present in healthy individuals at higher risk of developing schizophrenia based on risk factors such as schizotypal personality traits or genetic susceptibility (Kee et al., Reference Kee, Horan, Mintz and Green2004; Van’t Wout et al., Reference Van’t Wout, Aleman, Kessels, Larøi and Kahn2004). Moreover, first episode psychosis patients, in particular non-affective patients, already showed lower ability to label positive and negative emotional prosody, suggesting early disruption in their emotion recognition system (Caletti et al., Reference Caletti, Delvecchio, Andreella, Finos, Perlini, Tavano, Lasalvia, Bonetto, Cristofalo, Lamonaca, Ceccato, Pileggi, Mazzi, Santonastaso, Ruggeri, Bellani and Brambilla2018). Emotional dysfunction may, therefore, represent an index of early signs of the disease (Seiferth et al., Reference Seiferth, Pauly, Habel, Kellermann, Shah, Ruhrmann and Kircher2008), while strengthening the emotion recognition ability may improve prevention and intervention strategies (Comparelli et al., Reference Comparelli, Corigliano, De Carolis, Mancinelli, Trovini, Ottavi, Dehning, Tatarelli, Brugnoli and Girardi2013).

As facial emotion recognition skills can be enhanced by training (Combs et al., Reference Combs, Adams, Penn, Tiegreen and Stem2007; Wölwer et al., Reference Wölwer, Frommann, Halfmann, Piaszek, Streit and Gaebel2005), much work has been done to characterise the deficit, its relationship to symptoms and neural basis (Marwick and Hall, Reference Marwick and Hall2008). Structural and functional anomalies have been found in the insula and amygdala (Crespo-Facorro et al., Reference Crespo-Facorro, Kim, Andreasen, O’Leary, Bockholt and Magnotta2000; Honea et al., Reference Honea, Crow, Passingham and Mackay2005; Wright et al., Reference Wright, Rabe-Hesketh, Woodruff, David, Murray and Bullmore2000), with evidence of hypoactivation in patients with schizophrenia for fearful compared to neutral faces (Aleman and Kahn, Reference Aleman and Kahn2005; Delvecchio et al., Reference Delvecchio, Sugranyes and Frangou2012). This finding suggests that an undifferentiated amygdala response to fearful and neutral faces may hinder their discrimination and lead to misattribution of fear depending on the context (Marwick and Hall, Reference Marwick and Hall2008). In contrast, patients with affective psychosis show increased activation of the amygdala and hippocampus, consistent with the notion of greater arousal responses to emotional stimuli (Critchley et al., Reference Critchley, Rotshtein, Nagai, O’Doherty, Mathias and Dolan2005; Delvecchio et al., Reference Delvecchio, Sugranyes and Frangou2012; Santos et al., Reference Santos, Mier, Kirsch and Meyer-Lindenberg2010), but also better contextual appraisal compared to schizophrenia patients (Delvecchio et al., Reference Delvecchio, Sugranyes and Frangou2012; Gerdes et al., Reference Gerdes, Wieser, Mühlberger, Weyers, Alpers, Plichta, Breuer and Pauli2010). In addition, increased pulvinar activation in affective psychosis has been found, suggesting a greater focus on emotionally salient stimuli from early stages of visual processing (Pessoa and Adolphs, Reference Pessoa and Adolphs2010), whereas the emotional dysregulation may be related to deficit of grey matter volume in subgenual anterior cingulate cortex (Maggioni et al., Reference Maggioni, Crespo-Facorro, Nenadic, Benedetti, Gaser, Sauer, Roiz-Santianez, Poletti, Marinelli, Bellani, Perlini, Ruggeri, Altamura, Diwadkar and Brambilla2017). Taken together, these findings highlight the complex interplay between neural processes, emotional responses and contextual appraisal in individuals with different forms of psychosis. The role of the amygdala in discriminating and attributing emotions, as well as the increased activation observed in affective psychosis, highlights its potential role in the enhance patients’ experience of fear and arousal (Marwick and Hall, Reference Marwick and Hall2008; Wright et al., Reference Wright, Rabe-Hesketh, Woodruff, David, Murray and Bullmore2000). On the other hand, findings of reduce volume in limbic regions such as the insula and cingulate cortex may underlie difficulties in adaptive responses to emotional stimuli even at the onset of symptoms (Crespo-Facorro et al., Reference Crespo-Facorro, Kim, Andreasen, O’Leary, Bockholt and Magnotta2000; Maggioni et al., Reference Maggioni, Crespo-Facorro, Nenadic, Benedetti, Gaser, Sauer, Roiz-Santianez, Poletti, Marinelli, Bellani, Perlini, Ruggeri, Altamura, Diwadkar and Brambilla2017). A better understanding of regional variations and their relation to patients’ symptoms is relevant to shed light on the mechanisms underlying emotional deficits and to offer potential paths for tailored interventions in individuals with psychosis.

Notably, emotion processing involves several stages, which are measured by different tasks. Four main levels can be identified: i) unconscious processing refers to stimuli which are potentially accessible to consciousness but are processed in absence of awareness because they are below the threshold of perception (Dehaene et al., Reference Dehaene, Changeux, Naccache, Sackur and Sergent2006; Mattavelli et al., Reference Mattavelli, Pisoni, Romero Lauro, Marino, Bonomi, Rosanova and Papagno2019), ii) perceptual sensitivity is the individual threshold at which stimuli can be differentiated from noise or other stimuli (Pessoa et al., Reference Pessoa, Japee and Ungerleider2005), iii) the discrimination requires to distinguish between expressions and iv) recognition further requires to identify the target emotion (Adolphs et al., Reference Adolphs, Damasio, Tranel, Cooper and Damasio2000). The first two levels are tested with forced-choice tasks and using very brief presentation of stimuli (e.g. 10–30 ms target). Discrimination and recognition are instead assessed by identification tasks. While sensitivity depends on sensory and visuospatial processes, emotion recognition also requires the ability to label the correct emotion among several alternatives (Haxby et al., Reference Haxby, Hoffman and Gobbini2002). As the neural pathways underlying perceptual processing and emotion recognition are partially distinct and can be selectively impaired in different neuropsychiatric conditions (Mattavelli et al., Reference Mattavelli, Barvas, Longo, Zappini, Ottaviani, Malaguti and Papagno2021; Tamietto and De Gelder, Reference Tamietto and De Gelder2010), it is necessary to assess whether impairments occur at early perceptual processing or/and recognition of emotions.

This commentary aims at presenting an overview of the strengths and weaknesses of the available instruments assessing the different stages of facial emotion processing in patients with non-affective and affective psychoses. Since the deficits in social cognition domain are related to psychotic symptoms and quality of life, a deeper characterisation of emotion processing capacity is crucial for a better understanding of the pathologies and to promote personalised interventions.

Up to date, most studies have used emotion recognition tests in schizophrenia (Addington et al., Reference Addington, Penn, Woods, Addington and Perkins2008; Pinkham et al., Reference Pinkham, Penn, Perkins, Graham and Siegel2007) and bipolar disorder (Benito et al., Reference Benito, Lahera, Herrera, Muncharaz, Benito, Fernández-Liria and Montes2013; Ulusoy et al., Reference Ulusoy, Gülseren, Özkan and Bilen2020). The most reported are the Ekman 60-Faces Test (EK-60F; Young et al., Reference Young, Perrett, Calder, Sprengelmeyer and Ekman2002), the Facial Emotion Identification Test (FEIT; Kerr and Neale, Reference Kerr and Neale1993) and the Penn Emotion Recognition Test (ER-40; Kohler et al., Reference Kohler, Turner, Bilker, Brensinger, Siegel, Kanes and Gur2003). In the former, participants are shown a series of 60 black-and-white photographs of male and female faces expressing one of six basic emotions (surprise, happiness, fear, disgust, anger and sadness) and are required to associate the correct label to each emotion. The reliability and validity of this instrument have been demonstrated in several studies (Róza et al., Reference Róza, Kálmán, Ko, Ngay, Fiáth, Magi, Eisinger and Oláh2012). The FEIT similarly uses the stimuli developed by Ekman and Friesen (Reference Ekman and Friesen1976) and Izard (Reference Izard1971) but presents only 19 faces, 15 expressing negative (anger, sadness, fear and shame) and 4 expressing positive emotions (happiness and surprise). The FEIT has established psychometric properties (Dougherty et al., Reference Dougherty, Bartlett and Izard1974; Feinberg et al., Reference Feinberg, Rifkin, Schaffer and Walker1986; Zuroff and Colussy, Reference Zuroff and Colussy1986) and includes a control task to rule out the possibility that the poor performance of patients generally reflects cognitive impairments (Chapman and Chapman, Reference Chapman and Chapman1973, Reference Chapman and Chapman1978; Oltmanns and Neale, Reference Oltmanns and Neale1978). It has been validated for Korean (Bahk et al., Reference Bahk, Jang, Lee and Choi2015) and Chinese (Lo and Siu, Reference Lo and Siu2018) populations, but no normative data are currently available for the European and South American populations, which limits its clinical use. In contrast, the EK-60F has been validated in Korea (Kim et al., Reference Kim, Kwon, Jung, Kim, Cho, Kim and Choi2017) and Italy (Dodich et al., Reference Dodich, Cerami, Canessa, Crespi, Marcone, Arpone, Realmuto and Cappa2014). Finally, the ER-40 includes 40 colour-posed facial expressions of four emotions (anger, sadness, happiness or fear) with high and low intensity, as well as neutral (Kohler et al., Reference Kohler, Turner, Bilker, Brensinger, Siegel, Kanes and Gur2003), and participants are instructed to identify the expressed emotion among five possible choices.

Instead, the most reported test for assessing emotion discrimination ability is the Face Emotion Discrimination Test (FEDT), which was developed by Kerr and Neale (Reference Kerr and Neale1993) in parallel with the FEIT. It uses 30 pairs of stimuli from the Izard (1971) set, and subjects are asked to decide whether the same or different emotions are presented. Like the FEIT, the FEDT is validated with established psychometric properties (Dougherty et al., Reference Dougherty, Bartlett and Izard1974; Feinberg et al., Reference Feinberg, Rifkin, Schaffer and Walker1986; Kerr and Neale, Reference Kerr and Neale1993; Zuroff and Colussy, Reference Zuroff and Colussy1986), although normative data are currently available only for USA and Korean (Bahk et al., Reference Bahk, Jang, Lee and Choi2015) populations. Another test is the Emotion Differentiation Task (EMODIFF; Kohler et al., Reference Kohler, Bilker, Hagendoorn, Gur and Gur2000), in which participants are asked to differentiate the intensity of emotions shown in two side-by-side faces of the same person.

Few previous studies have combined these instruments and reported that both discrimination and recognition were impaired in patients and in high-risk individuals for psychosis (Addington et al., Reference Addington, Penn, Woods, Addington and Perkins2008; Benito et al., Reference Benito, Lahera, Herrera, Muncharaz, Benito, Fernández-Liria and Montes2013; Comparelli et al., Reference Comparelli, Corigliano, De Carolis, Mancinelli, Trovini, Ottavi, Dehning, Tatarelli, Brugnoli and Girardi2013; Ulusoy et al., Reference Ulusoy, Gülseren, Özkan and Bilen2020), suggesting a general impairment involving early stage of processing. On the other hand, studies with experimental paradigms assessing unconscious emotion processing, at behavioural and brain activity level, reported inconsistent data on early automatic stages, which resulted impaired or preserved in different samples of patients with psychosis (Brennan et al., Reference Brennan, Harris and Williams2014; Gruber et al., Reference Gruber, Siegel, Purcell, Earls, Cooper and Barrett2016; Williams et al., Reference Williams, Whitford, Nagy, Flynn, Harris, Silverstein and Gordon2009). The independency vs hierarchical dependency of the different stages including unconscious processing, early visual processing, recognition and labelling of emotions is a debated issue (Barrett et al., Reference Barrett, Lindquist and Gendron2007; Kring et al., Reference Kring, Siegel and Barrett2014). Hierarchical models propose a feed-forward propagation of signals from sensory to higher-level cortical areas, with the latter projecting feedback on early processing areas to support coherent representations throughout top-down modulation. Other models hypothesise the presence of independent pathways for conscious and unconscious processing, with signal diverging from early stages and involving distinct cortical and subcortical neural networks (Dehaene and Changeux, Reference Dehaene and Changeux2011; Tamietto and De Gelder, Reference Tamietto and De Gelder2010). Within this framework, refining patients’ assessment has significant theoretical and translational implications. Individual perceptual sensitivity and unconscious processing can be evaluated with backward masking (Williams et al., Reference Williams, Whitford, Nagy, Flynn, Harris, Silverstein and Gordon2009) or continuous flash suppression (CFS) paradigms (Gruber et al., Reference Gruber, Siegel, Purcell, Earls, Cooper and Barrett2016) to clarify whether emotional dysfunction is related to a deficit in early processing of salience and elucidate the role of anomalies in the recurrent interactions between visual and higher-level associative areas for access to consciousness (Dehaene and Changeux, Reference Dehaene and Changeux2011; Del Cul et al., Reference Del Cul, Dehaene and Leboyer2006).

In conclusion, emotion processing involves multiple stages, from unconscious processing to recognition and discrimination of emotions. Different tests, summarised in Table 1, have been developed to assess these stages. Although impairments have been observed in patients with non-affective and affective psychoses, the precise nature of these deficits remains a subject of debate. Therefore, refining patients’ assessment is crucial for a better characterisation of deficits in both affective and non-affective psychosis, which could be used to modulate new cognitive remediation and social skills training interventions aimed at enhancing emotion processing skills and improving effective interaction in social contexts.

Table 1. Characteristics of tasks used to assess the different stages of emotion processing in psychosis