Schizophrenia (SCZ) is a debilitating psychiatric disorder (Schultz and Andreasen, 1999) characterized by extensive cognitive deficits (Davidson and Keefe, 1995; Gold, 2004). These deficits are observed in multiple domains, including learning and memory (Diwadkar et al., 2008), and patients are noted for brain network anomalies originating in key brain regions such as the hippocampus and the prefrontal cortex (Kody and Diwadkar, 2022; Wadehra et al., 2013; Woodcock et al., 2016). Typically employed experimental learning paradigms are characterized by distinct phases relating to both a) task condition (Baajour et al., 2020) and b) time (Brambilla et al., 2011). The overt task conditions include not only “active” phases of memory Encoding and Retrieval, but notably, also covert phases of memory consolidation (Büchel et al., 1999). These instruction-free covert phases which are widely used in fMRI block-design studies (Amaro and Barker, 2006), are placed between periods of explicit encoding and retrieval (and are typically considered to be “baseline” conditions against which task active conditions are compared). However, extensive evidence suggests that covert periods are, in fact, “constructive” and characterized by substantial network connectivity (Meram et al., 2023; Ravishankar et al., 2019). These observable network signatures are unsurprising because covert periods are presumed to be central to memory rehearsal and/or recapitulation (Butler and James, 2013). Thus, covert periods during learning link processes across successive task-active periods (Deco et al., 2013; Diwadkar et al., 2017).

In schizophrenia, impaired covert processing related to memory consolidation has been recognized, primarily in the context of sleep (given the relationship between distributed network activity during sleep and memory) (Creery et al., 2022; Stickgold, 2005; Torres-Herraez et al., 2022). Disruption during critical periods of sleep disrupts longer-term consolidation of previously formed memories (Chee and Chuah, 2008), specifically when disturbances are applied to slow-wave sleep. In addition, sleep spindles are implicated in contributing to cognitive deficits in SCZ (Göder et al., 2015; Keshavan et al., 1990; Wamsley et al., 2012), and sleep-related consolidation is strongly mediated by the role of the medial temporal lobe (Buzsaki, 1998). While sleep is a constructive covert period, sleep-related memory consolidation is more distal from the initial phases of memory formation. By comparison, constructive covert neural activity during ongoing tasks is a ubiquitous property of neural systems, underpinning planning, motor behavior, learning, and memory (Diwadkar et al., 2017; van der Meer et al., 2017).

In-task covert periods are more proximate to initial memory formation, and may complement some of the overt functional commitments of the hippocampus and the prefrontal cortex (Ravishankar et al., 2019). The hippocampus’ role is most likely related to the slow and ongoing process of the redistribution of memory traces (Haist et al., 2001). By comparison, the prefrontal cortex's role may relate to the recapitulation and maintenance of encoded memory traces through processes of working memory-related maintenance (Fiebig and Lansner, 2017; Miller, 2013; Pu et al., 2016). Moreover, paired-associate learning is staged. Such learning is typically underpinned by non-linear increases in behavioral proficiency, where performance is typically well-approximated by negatively accelerated growth functions (Baajour et al., 2020; Stanley et al., 2017). This temporal staging of memory acquisition (where proficiency in the early stages increases linearly before reaching asymptote in later stages), may be mirrored by systematic changes in brain network profiles (Ravishankar et al., 2019). Therefore, network profiles during covert periods may also show evidence of such staging.

Do brain network profiles during covert periods of learning differ between patients and healthy controls and are these differences subject to memory staging? We investigated this question by estimating network profiles of each of the hippocampus and the dlPFC during the Early and Late covert periods of an established associative learning paradigm (Diwadkar et al., 2016; Ravishankar et al., 2019; Stanley et al., 2017). Psychophysiological Interaction (PPI) was the network model of choice because a) PPI is an established method for estimating directed (i.e., from seed to target) functional connectivity (Friston et al., 1997; O'Reilly et al., 2012; Silverstein et al., 2016), and b) PPIs specifically quantify contextually driven changes, where the context represents experimentally manipulated periods within any task (Friedman et al., 2017; Meram et al., 2021). To establish that covert periods are in fact “active” (thus constituting a viable task phase), initial analyses were conducted (Figure 3) to established evidence for significant activity of each of the hippocampus and the dlPFC. Next, as our results will show, SCZ patients evince altered network profiles of both the dlPFC and the hippocampus during covert processing, but where the temporal staging of alterations was complex. Aberrant network interactions during covert periods of learning may be culpable for poorly formed memory traces in schizophrenia; this poverty, in turn, may contribute to the learning deficits that are a core feature of the illness (Ragland et al., 2012).