Research into the physiological mechanisms of schizophrenia has traditionally focused almost exclusively on how the brain processes and interprets information from the outside world. However, there is growing evidence that internal signals from the body can influence a wide range of cognitive and emotional functions (Makowski et al., 2020; Pramme et al., 2014, 2016; Tsakiris and De Preester, 2019). The processing of bodily signals by the central nervous system is called interoception. Disturbances in interoception may underlie a wide range of schizophrenia symptoms (see Yao and Thakkar, 2022 for review).

Interoception is closely linked to allostasis (Schulkin and Sterling, 2019; Sterling, 2012): a predictive regulation process wherein the brain adjusts and prepares the body based on predicted metabolic needs (e.g., drink before we become dehydrated). According to the interoceptive inference theory (Barrett and Simmons, 2015; Seth et al., 2012; Seth and Friston, 2016), the brain maintains an internal model of the body and is constantly trying to minimize discrepancies between the predicted and actual signals from the body (i.e., prediction error signals). To minimize prediction errors, the brain will either shift the prediction to match the incoming sensory data (i.e., update the model of the body), or shift the sensory data towards the prediction (i.e., active inference; Seth, 2015). The latter can be achieved by changing the sensory input through hormonal, visceral, immunological, autonomic, and behavioral mechanisms (e.g., taking a break during exercise when heart rate is too fast; Pezzulo et al., 2015).

Emotion has been theorized as a crucial component of the interoceptive inference process (Barrett, 2017; Seth, 2013). It is a conscious “label” that helps us better understand our everchanging bodily state, so that we can prepare for incoming bodily sensations and take adaptive actions to resolve interoceptive prediction errors when needed. For example, we may interpret the same bodily sensations experienced while riding a rollercoaster as either “afraid” or “excited” and decide to rest or ride again as a result of those attributions. We may assign different emotion labels to similar patterns of bodily signals because emotion results from dynamically incorporating interoception, past experience, current context, and attentional focus.

According to the Embodied Predictive Interoceptive Coding (EPIC) model, the computational model of interoceptive inference discussed above is instantiated by a specific brain network. Namely, the visceromotor cortices generate predictions regarding the state of the body depending on past experience and the external environment, while the body sends signals of actual physiological states to mid-posterior insula, the interoceptive sensory cortex (Barrett and Simmons, 2015). Tract-tracing studies in monkeys and neuroimaging studies in humans provide support for this allostatic-interoceptive network (Kleckner et al., 2017). Moreover, individuals with stronger functional connectivity within this network exhibit higher interoceptive accuracy (Kleckner et al., 2017).

Although few studies have directly investigated interoception in schizophrenia (Ardizzi et al., 2016; Critchley et al., 2019; Koreki et al., 2021; Torregrossa et al., 2022), a larger body of work investigating various bodily systems further supports altered interoception in schizophrenia (Yao and Thakkar, 2022). Furthermore, there is a robust evidence base for structural and functional abnormalities in a key node in the aforementioned interoception network–the insula–that are associated with symptom severity and emotion recognition (Sheffield et al., 2020; Takahashi et al., 2020; Wylie and Tregellas, 2010).

The current study aimed to extend this previous neuroimaging work by investigating the structural connectivity of a putative interoception network and its relationship with emotional functioning and clinical symptoms in schizophrenia. White matter microstructural abnormalities are widespread in schizophrenia (Kelly et al., 2018). Such abnormalities may lead to disrupted communications within the interoception network, resulting in altered allostasis and/or emotional functioning. Therefore, we examined the white matter microstructural integrity within this network in the current study. We predicted that relative to healthy controls, participants with schizophrenia would have reduced structural connectivity. We also predicted that structural connectivity within this network would be correlated with emotional functioning across participants and with clinical symptom severity in schizophrenia.