Atypical social cognition has been acknowledged as a hallmark of schizophrenia (Green et al., 2015; Kahn et al., 2015), but emerging evidence has attached great importance to the separation of bottom-up and top-down cognitive processes during the investigation of social information processing in people with schizophrenia (PSZ). In fact, recent studies using simple social stimuli (e.g., single faces or gazes) have indicated that social challenges in PSZ might not be ascribed to the bottom-up perception, but to the higher-order cognitive activities. To measure bottom-up perception, the breaking continuous flash suppression (b-CFS) paradigm (Tsuchiya & Koch, 2005) has become an effective approach. The typical protocol of this paradigm uses CFS masks to temporarily render the current target invisible for several seconds, which will perceptually “break” through the masks and enter awareness as the opacity of the target gradually increases. Hence, shorter time needed by a target to overcome CFS masks and break into consciousness (i.e., breakthrough time) would indicate faster and advantageous bottom-up processing (Stein, 2019). Since the differences in breakthrough time between targeted social stimuli, such as gaze (Seymour et al., 2015), face (Caruana et al., 2019) and facial emotion (Caruana and Seymour, 2021; Grave et al., 2021), and baseline stimuli were similar between PSZ and healthy control subjects (HCS), PSZ seem to preserve the ability of bottom-up social perception. On the other hand, PSZ have demonstrated consistent deficits in various higher-order social tasks (Weinreb et al., 2022), including face recognition (Bortolon et al., 2015), emotion discrimination (Pena-Garijo et al., 2023), emotion-related visual search (Pinkham et al., 2014), Theory-of-Mind (ToM; van Neerven et al., 2021) and other well established social test batteries (Lee et al., 2013). These findings have revealed the intricacy of social cognition, and motivated further explorations for the bottom-up and top-down aspects of social information processing in schizophrenia.

Social interaction, as an ubiquitous social event in daily life, introduces an extra binding of social relations over simple social elements (Vestner et al., 2019), which has not been sufficiently understood in schizophrenia. In real-life settings, we sometimes have to be an outsider of interactive conversations, especially when we are new to an occupation, community, or simply trying to enter a gathering of people. In such circumstances, we tend to rapidly and automatically observe the groups of people and compute their basic social interactions, so that we could appropriately fit ourselves in by subtly analyzing more social information, such as topics of interest, status, deception, competition, kinship and intimacy between the interactive individuals (Costanzo and Archer, 1989). Hence, this allocentric perspective of social interaction processing reflects the encoding of human relationships unrelated to the third-party observer and provides important real-life implications, as it determines the first-pass computation of social interactions and lays the foundation for social fusion in real-life scenes and the ensuing egocentric social activities.

Since prosocial engagement has long been a concern for the well-being and longevity of PSZ (Weittenhiller et al., 2021), investigation of social interaction processing from the allocentric view will importantly reveal automatic social computation of PSZ and may shed lights on the interventions for their social life. In the general population, socially interactive stimuli are processed with overall advantages than non-interactive stimuli, including preferential access to visual awareness (Fu et al., 2024), physically closer spatial distance in working memory, and higher accuracy for long-term memory recall (Vestner et al., 2019). To our knowledge, only few studies have provided preliminary evidence for the allocentric processing of social interactions in PSZ, and they have primarily used top-down cognitive tasks (Lee et al., 2018; Okruszek et al., 2015, 2018, 2019), leaving the bottom-up perception of social interactions for further investigation. With the typically adopted biological motion paradigm, PSZ showed difficulty in discriminating communicative interactions from non-communicative actions (Okruszek et al., 2015), but they seemed to preserve the ability to predict the existence of an invisible agent when a visible agent displayed communicative gestures (Okruszek et al., 2018, 2019). One recent study (Lee et al., 2018) used the social remember-know paradigm with dynamic multi-modal videos, and reported impaired recollection memory but intact familiarity memory for social interactions in PSZ. In this case, existing findings on social interaction processing of PSZ seemed somewhat controversial, and some confounding moderators, such as lower-level visual perceptual deficits (Kim et al., 2005), could have contributed to these inconsistencies. For instance, PSZ showed moderate-to-large deficits in biological motion recognition (Okruszek and Pilecka, 2017), which might independently bias previous observations of how PSZ processed social interaction itself. Hence, convincing evidence concerning how PSZ process social interaction information still needs to be accumulated, and more straightforward stimuli and task requirements may appropriately help minimize potential confounding effects.

It is also worthwhile to investigate the correlations between schizophrenic symptoms (i.e., positive and negative) and social cognition, as their interplay has been complex and inconclusive. Positive symptoms (e.g., hallucination and delusion) have shown inconsistent relationships with social cognition (Nelson et al., 2007; Shean and Meyer, 2009), and they tended to be associated with excessive social attributions in PSZ, especially in ToM tasks (Peyroux et al., 2018). However, negative symptoms, including anhedonia, avolition, asociality, blunted affect and alogia, have been more consistently acknowledged as a closely related but still independent construct with particular processes of social cognition (Piskulic and Addington, 2011), as they are predictive of general deficits in social processing (Bell et al., 2013; Peyroux et al., 2018) irrespective of the severity of positive symptoms (Bliksted et al., 2017). It has been suggested that negative symptoms may represent the functional outcomes of altered social cognition as they become more prominent in social settings (Pelletier-Baldelli and Holt, 2020). For example, anhedonia is characterized by lower intensity or frequency in pleasurable experiences in larger social networks. Moreover, the severity of negative symptoms has also been linked to worse performance of social tasks (Bliksted et al., 2017; Lincoln et al., 2011) and less satisfactory real-life social outcomes (Foussias et al., 2014). Therefore, we hypothesize that there may exist significant correlations between negative symptoms and the processing of social interaction information in this study, and the correlations could be different between the bottom-up and top-down processes.

Considering that studies on the processing of social interaction in schizophrenia have focused on either bottom-up or top-down procedure, previous empirical evidence could not be directly compared, especially when various social stimuli were used in different studies. To reconcile extant inconsistent findings, this study aimed to systematically investigate the processing of social interaction in PSZ under a unified research framework, which covered both bottom-up perception and top-down memory. To ensure the comparability of findings between our two experiments, we manipulated the facingness between two human agents throughout to ecologically represent socially interactive (i.e., face-to-face dyads) and non-interactive (i.e., back-to-back dyads) relationships (Fu et al., 2024; Vestner et al., 2019; Zhou et al., 2019). In specific, Experiment 1 aimed to investigate whether PSZ demonstrated advantageous bottom-up perceptual sensitivity of social interactions using the b-CFS paradigm. Experiment 2 aimed to gauge whether PSZ demonstrated advantages in the top-down processing of social interactions using two memory tasks, including a spatial working memory task and a surprise long-term memory task.