Background
A diverse range of longitudinal courses has been observed in schizophrenia and related psychotic disorders (Heilbronner, Samara, Leucht, Falkai, & Schulze, Reference Heilbronner, Samara, Leucht, Falkai and Schulze2016). For positive symptoms, the course can vary between remission, persistent symptoms, and relapses with variable residual symptoms. An initial ‘full remission’ course may later evolve into a ‘residual symptom’ course, often after relapses (Emsley, Chiliza, & Asmal, Reference Emsley, Chiliza and Asmal2013; Emsley, Chiliza, Asmal, & Harvey, Reference Emsley, Chiliza, Asmal and Harvey2013; Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen and Chen2018; Taipale, Tanskanen, Correll, & Tiihonen, Reference Taipale, Tanskanen, Correll and Tiihonen2022; Wiersma, Nienhuis, Slooff, & Giel, Reference Wiersma, Nienhuis, Slooff and Giel1998). Compared with the many studies on factors leading to the onset of psychosis, relatively few have addressed illness courses after onset. A coherent account of psychosis should explain not only the onset of psychotic disorders but also the subsequent relapses and development of refractory symptoms, which can become increasingly independent of dopamine. The observation that symptoms during relapse often repeat contents similar to those in the previous episodes (Chaturvedi & Sinha, Reference Chaturvedi and Sinha1990; Grunfeld et al., Reference Grunfeld, Lemonde, Gold, Paquin, Iyer, Lepage and Shah2024; Palaniyappan, Reference Palaniyappan2019; Sinha & Chaturvedi, Reference Sinha and Chaturvedi1990) suggests that memory processes may be involved.
It is increasingly recognized that memory processes are modulated by dopamine (Sayegh et al., Reference Sayegh, Mouledous, Macri, Pi Macedo, Lejards, Rampon and Dahan2024; Shohamy & Adcock, Reference Shohamy and Adcock2010), which has already been observed to play a key role in producing psychotic symptoms (Howes, Bukala, & Beck, Reference Howes, Bukala and Beck2024; Howes & Kapur, Reference Howes and Kapur2009; Wong et al., Reference Wong, Suen, Wong, Chan, Hui, Chang and Chen2022) through an excessive sense of salience (Howes & Nour, Reference Howes and Nour2016; Kapur, Reference Kapur2003). We have previously argued broadly that known neuroplasticity processes interacting with dopamine may contribute to accounting for the course of psychotic symptoms (Chen et al., Reference Chen, Wong, Tang, Lei, Suen and Hui2023). In the current narrative review, we elaborate on how a dopamine-gated autobiographical memory (AM) account coherently brings forth specific hypotheses relevant to the longitudinal evolution of psychotic symptoms.
Memory gating roles of dopamine
Dopamine salience accounts
In everyday life, the brain monitors our environment by continuously predicting the next environmental state. The discrepancy between the predicted and the actual events generates a prediction error signal indicating the presence of novel information and preparing the brain for neuroplasticity (Heinz et al., Reference Heinz, Murray, Schlagenhauf, Sterzer, Grace and Waltz2019; Shohamy & Adcock, Reference Shohamy and Adcock2010). Such prediction error signals are associated with increased dopamine activities (Starkweather, Babayan, Uchida, & Gershman, Reference Starkweather, Babayan, Uchida and Gershman2017). When the dopamine system is active, correlated information in the environment is more readily encoded (Shohamy & Adcock, Reference Shohamy and Adcock2010). Notably, prediction error signals are supervened by a sense of salience: a subjective feeling that significant information has emerged in the environment (Howes et al., Reference Howes, Bukala and Beck2024; Howes & Kapur, Reference Howes and Kapur2009; Wong et al., Reference Wong, Suen, Wong, Chan, Hui, Chang and Chen2022). Dopamine activation is also associated with increased functional connectivity in the salience network in the brain (Conio et al., Reference Conio, Martino, Magioncalda, Escelsior, Inglese, Amore and Northoff2020). During acute psychosis, the aberrant increase in dopamine activity therefore leads to a ‘spurious’ sense of salience and overinterpretation of neutral environmental information, an account that is consistent with narratives of people with psychotic disorders (Heinz et al., Reference Heinz, Murray, Schlagenhauf, Sterzer, Grace and Waltz2019; Howes et al., Reference Howes, Bukala and Beck2024; Kapur, Reference Kapur2003; McCutcheon, Krystal, & Howes, Reference McCutcheon, Krystal and Howes2020).
While the prediction error-salience theory explains psychotic symptoms at illness onset, it does not fully explain how these symptoms evolve over time. Among the few accounts, the persistence of delusions has been explored using the prediction error model (Corlett, Krystal, Taylor, & Fletcher, Reference Corlett, Krystal, Taylor and Fletcher2009). The model posits that representations of psychotic and nonpsychotic experiences compete for dominance. Chronically, increased prediction error signals (presumably related to a sustained elevation in dopamine activity) are suggested to contribute to the persistence of delusions. However, the underlying assumption of a chronic state of dopamine overactivity remains contentious (Avram et al., Reference Avram, Brandl, Cabello, Leucht, Scherr, Mustafa and Sorg2019). The model also does not adequately account for the progressive increase in psychotic symptoms with each relapse.
Memory-based accounts
Importantly, while new learning is facilitated by dopamine, once encoding has taken place, ongoing dopamine activity is not required to maintain the memory representation. This is in contrast to the persistent prediction error-salience models above. The subsequent course of the spurious memories formed during a psychotic episode can be understood with reference to the known natural history of memory traces in the brain.
Studies of memory in psychosis have largely focused on deficits based on failure to recall in standardized memory tests, rather than on aberrant memories (Danion,Huron, Vidailhet, & Berna, Reference Danion, Huron, Vidailhet and Berna2007; Harvey et al., Reference Harvey, Bosia, Cavallaro, Howes, Kahn, Leucht and Vita2022; Ranganath, Minzenberg, & Ragland, Reference Ranganath, Minzenberg and Ragland2008). Previous studies addressing aberrant memory in psychosis were initiated with attempts to understand psychotic symptom formation using associative memory network models (Chen et al., Reference Chen, Wong, Hui, Tang, Chiu, Lam and Sham2009; Chen & Berrios, Reference Chen, Berrios, Stein and Ludik1998; Hoffman et al., Reference Hoffman, Rapaport, Ameli, McGlashan, Harcherik and Servan-Schreiber1995; Hoffman & McGlashan, Reference Hoffman and McGlashan1997; Rolls, Reference Rolls2021; Rolls, Loh, Deco, & Winterer, Reference Rolls, Loh, Deco and Winterer2008). In line with this direction, it has been suggested that psychosis could be conceptualized as a ‘learning and memory disorder’ and could be understood in terms of known memory functions in the hippocampus (Tamminga, Reference Tamminga2013). However, there has been little discussion on how this approach can be linked to the emerging dopamine-prediction error-salience findings. Independently, the role of traumatic experiences in psychotic disorders has increasingly been considered from a memory perspective (Hardy, Reference Hardy2017). Aberrant memory has also been incorporated into an account of brain state homeostasis fluctuating between neural overactivity and stabilization (Palaniyappan, Reference Palaniyappan2019). In the context of these approaches, we seek to review whether known memory processes interacting with dopamine activity may be sufficient to offer a parsimonious account of the longitudinal course of psychosis symptoms that can bring forth new clinical insights and hypotheses. The core account is concisely outlined in the next section.
A dopamine-gated memory account: Spurious Autobiographical Memories of Psychosis (SAMP)
We propose that during an acute psychotic episode, the contents of psychotic experiences and related aberrant associations are encoded as SAMP based on Hebbian synaptic plasticity, facilitated by elevated dopamine levels (Dringenberg, Reference Dringenberg2020; Duszkiewicz, McNamara, Takeuchi, & Genzel, Reference Duszkiewicz, McNamara, Takeuchi and Genzel2019; Kamiński et al., Reference Kamiński, Mamelak, Birch, Mosher, Tagliati and Rutishauser2018; Shohamy & Adcock, Reference Shohamy and Adcock2010; Wittmann et al., Reference Wittmann, Schott, Guderian, Frey, Heinze and Düzel2005). After remission, the newly encoded SAMP interacts with premorbid nonpsychotic AM to construct a coherent overall autobiographical account (Figure 1). Suboptimal integration results in a SAMP segregated from nonpsychotic AM. Access to SAMP during remission is also postulated to be impeded due to contextual mismatch. This limited access to SAMP in remission compromises the ability of normalized experience to modify the SAMP. Residual SAMP may become reactivated associatively and involuntarily upon encountering external or internal cues, leading to an elevated propensity for relapse (Chen et al., Reference Chen, Wong, Tang, Lei, Suen and Hui2023).
Illustration of Spurious Autobiographical Memory of Psychosis (SAMP).
Dark backgrounds denote SAMP memories formed during psychosis. Each unit represents autobiographical memory (AM) encoded over a period of time. Arrows on the right of each row indicate current life experience. A light gray background indicates partially normalized AM. Arrows represent integrative processes between new and old memories. From the top, (a) SAMP associated with short duration of active psychosis (DAP); (b) SAMP associated with a longer DAP; (c) AM integration in remission: (ci) SAMP in remission; well-integrated SAMP enables extinction and reduction in SAMP (cii); (d) poorly integrated SAMP does not benefit effectively from normalization, instead, (di) relapses can add to SAMP; and (dii) increasing treatment resistance; (e) premorbid threat-prone schema generally facilitates SAMP reactivation and persistence.
The formation of SAMP
Dopamine-gated neuroplasticity in SAMP
Recent accounts of everyday memory have emphasized the synergistic interaction between memory subsystems (Ferbinteanu, Reference Ferbinteanu2020; Rubin, Reference Rubin2006), which include memories of personal episodic events (Shohamy & Adcock, Reference Shohamy and Adcock2010), semantic memory (Battaglia & Pennartz, Reference Battaglia and Pennartz2011), implicit associations (Ferbinteanu, Reference Ferbinteanu2020), and emotional memory (Luminet, Reference Luminet2022). Comparative brain anatomy suggested that although specialized memory systems have emerged in different phylogenetic stages to serve different adaptive functions, the basic function of dopamine gating of neuroplasticity has already appeared in early vertebrates. Interestingly, dopamine gating of memory appears to have been conserved during the evolution of specialized vertebrate memory systems, including the phylogenetically more recent human AM system (Allen & Fortin, Reference Allen and Fortin2013; Murray, Wise, & Graham, Reference Murray, Wise and Graham2017).
Gated by dopamine, novel information is encoded in synapses through long-term potentiation (LTP) (Duszkiewicz et al., Reference Duszkiewicz, McNamara, Takeuchi and Genzel2019; Kamiński et al., Reference Kamiński, Mamelak, Birch, Mosher, Tagliati and Rutishauser2018; Sayegh et al., Reference Sayegh, Mouledous, Macri, Pi Macedo, Lejards, Rampon and Dahan2024; Sheynikhovich, Otani, & Arleo, Reference Sheynikhovich, Otani and Arleo2013). LTP strengthens a specific synapse when its presynaptic signal is associated with a successful post-synaptic activation (the Hebbian rule) (Sayegh et al., Reference Sayegh, Mouledous, Macri, Pi Macedo, Lejards, Rampon and Dahan2024; Shohamy & Adcock, Reference Shohamy and Adcock2010). Dopamine facilitates LTP through postsynaptic mechanisms involving molecules such as the Calmodulin protein kinase II (CAMKII) and the cAMP response element-binding protein (CREB). Similar molecular pathways have been identified in the hippocampus (Prince, Bacon, Tigaret, & Mellor, Reference Prince, Bacon, Tigaret and Mellor2016; Sayegh et al., Reference Sayegh, Mouledous, Macri, Pi Macedo, Lejards, Rampon and Dahan2024), the striatum (Speranza, Di Porzio, Viggiano, De Donato, & Volpicelli, Reference Speranza, Di Porzio, Viggiano, De Donato and Volpicelli2021), and the amygdala (Allen & Fortin, Reference Allen and Fortin2013; Maren, Reference Maren2015; Markowitsch & Staniloiu, Reference Markowitsch and Staniloiu2011; Speranza et al., Reference Speranza, Di Porzio, Viggiano, De Donato and Volpicelli2021). Thus, through LTP-mediated memory facilitation, dopamine interacts not only with implicit memory and emotional memory but also with the more recently evolved AM (Dringenberg, Reference Dringenberg2020; Hannula, Minor, & Slabbekoorn, Reference Hannula, Minor and Slabbekoorn2023; Klein, Cosmides, Tooby, & Chance, Reference Klein, Cosmides, Tooby and Chance2002; Paré & Headley, Reference Paré and Headley2023; Rubin, Reference Rubin2006).
Temporal integration between new and old AM, as well as between AM and implicit, emotional and semantic memories, may depend on synchronized neural oscillation (Fries, Reference Fries2009; Fuentemilla, Palombo, & Levine, Reference Fuentemilla, Palombo and Levine2018; Steinvorth, Wang, Ulbert, Schomer, & Halgren, Reference Steinvorth, Wang, Ulbert, Schomer and Halgren2010). A reduced neural oscillation phase coherence is observed to be relatively specific for schizophrenia (Wolff & Northoff, Reference Wolff and Northoff2024). This abnormality is postulated to be linked to a general temporal disorganization in a spatiotemporal psychopathology model proposed as a bridge between subjective and objective phenomena (Northoff, Daub, & Hirjak, Reference Northoff, Daub and Hirjak2023; Northoff & Hirjak, Reference Northoff and Hirjak2023). AM provides another portal to explore the interface between first-person phenomenological experience and mechanistic brain processes in psychotic disorders.
The focus on AM in the present account of psychosis is relevant not only because AM content is directly related to the conscious experience of psychotic symptoms but also because it contains identifiable time stamps that enable hypothesis testing by distinguishing between memories laid down in the psychotic and nonpsychotic periods (McWilliams et al., Reference McWilliams, Kaplan, Eline, Kaminer, Zodrow, Petersen and Osipowicz2022).
The structure of AM
AM involves personal memories of events organized into event groups and schemata (Brown, Reference Brown2023; Conway, Reference Conway2005; Rubin & Umanath, Reference Rubin and Umanath2015; Zacks, Reference Zacks2020). Memories from different time periods are integrated with the self-schema to generate individual life narratives (Conway, Reference Conway2005). AM thus contains information on event memories (Bird, Reference Bird2020); and how these event memories are integrated through major transitions in life (Habermas, Reference Habermas2011; Habermas & Köber, Reference Habermas and Köber2015).
At the neurobiological level, event encoding is automatically segmented into epochs, each consisting of relatively continuous contextual information (Bird, Reference Bird2020; Zacks, Reference Zacks2010, Reference Zacks2020). Events are demarcated by a shift in contextual information as flagged by increased prediction error signals (Kumar et al., Reference Kumar, Goldstein, Michelmann, Zacks, Hasson and Norman2023). Besides the hippocampus’ involvement in AM encoding, AM retrieval is mediated by the posterior cingulate gyrus and the ventromedial prefrontal cortex (Summerfield, Hassabis, & Maguire, Reference Summerfield, Hassabis and Maguire2009; Svoboda, McKinnon, & Levine, Reference Svoboda, McKinnon and Levine2006). Dopamine activation is associated with a reduction in functional connectivity of the default mode network, involving the posterior cingulate gyrus (Conio et al., Reference Conio, Martino, Magioncalda, Escelsior, Inglese, Amore and Northoff2020).
Deficits in AM retrieval in schizophrenia can be revealed as a reduction in self-defining memories (SDMs) and an increase in over-general memory (OGM) (Allé et al., Reference Allé, Potheegadoo, Köber, Schneider, Coutelle, Habermas and Berna2015; Berna et al., Reference Berna, Potheegadoo, Aouadi, Ricarte, Allé, Coutelle and Danion2015; Nieto et al., Reference Nieto, Latorre, García-Rico, Hernández-Viadel, Ros and Ricarte2019). Notably, the temporal distribution of AM shows difficulty retrieving vivid memories around illness onset (Elvevåg, Kerbs, Malley, Seeley, & Goldberg, Reference Elvevåg, Kerbs, Malley, Seeley and Goldberg2003). OGM appears to be a transdiagnostic phenomenon that is observable in various disorders other than schizophrenia (Barry, Clark, & Maguire, Reference Barry, Clark and Maguire2021).
The role of involuntary AM in relapses
The retrieval of AM has been suggested to lie on a continuum between intentional and involuntary processes (Berntsen, Reference Berntsen2010). Involuntary AM (IAM) retrieval constitutes a significant proportion of spontaneous cognition (Berntsen, Reference Berntsen2023). IAM is considered an evolutionary earlier form of memory than AM and can arguably be observed in other primates (Allen & Fortin, Reference Allen and Fortin2013). Under the SAMP account, IAM retrieval is considered to be a key process involved in psychosis relapse and treatment resistance.
IAM retrieval involves external or internal cues triggering associative retrieval of past events based on the uniqueness of the cue memory, or ‘encoding-retrieval’, match (Berntsen, Reference Berntsen2010, Reference Berntsen2021, Reference Berntsen2023). Commonly observed in everyday life, IAM involves the default mode network as in AM but there is, as expected, less activation of the prefrontal cortex (Hall et al., Reference Hall, Rubin, Miles, Davis, Wing, Cabeza and Berntsen2014; Hall, Gjedde, & Kupers, Reference Hall, Gjedde and Kupers2008). While IAM is mostly adaptive and functional, it can become dysfunctional when the content of the retrieved memory is distressing or disruptive (Berntsen, Reference Berntsen2023). Patients with psychosis are more prone to experiencing IAM, particularly those triggered by internal rather than external cues (Allé, Berna, Danion, & Berntsen, Reference Allé, Berna, Danion and Berntsen2020). Although IAM contents during remission are generally mundane, they are rated with higher self-relevance but lower ‘belief in actual occurrence’ (Allé et al., Reference Allé, Berna, Danion and Berntsen2020). Not requiring intentional retrieval, the automatic involuntary characteristic makes IAM a prime candidate for mediating the reactivation of SAMP during psychosis relapse. So far, most AM/IAM studies have not specifically focused on SAMP. Future studies are required to distinguish between memories encoded during psychosis and nonpsychotic periods.
After consolidation, memory traces become relatively stable and are stored in the cortex independent of hippocampal involvement until they are reactivated (Goto, Reference Goto2022). We propose that residual SAMP can be associatively reactivated as IAM. The probability of retrieval depends on the uniqueness of the encoding-retrieval match (distinctiveness of the cue-SAMP association, or SAMP potency) (Berntsen, Reference Berntsen2023). The dynamics of memory retrieval have been described using associative network models (Rolls, Reference Rolls2021). One inherent feature of associative retrieval is that the retrieved memory can act as internal cues for further iterations of retrieval. In this way, an initial cue can trigger a cascade retrieval of related memories. Notably, this dynamics is consistent with the clinical observation of the rapid build-up of psychosis symptoms from their first appearance to full expression in a relapse (Emsley, Chiliza, Asmal, & Harvey, Reference Emsley, Chiliza, Asmal and Harvey2013). The threshold-crossing of such retrieval may depend on both the potency of SAMP and dopamine levels. In a normal dopaminergic state, higher potency in SAMP is required for retrieval. Meanwhile, in a high dopaminergic state, the congruent internal physiological and psychological context of the first psychotic episode is emulated, and retrieval of SAMP with lower potency may be facilitated.
During a relapse, new psychotic memories are added to existing SAMP to synthesize a more potent SAMP cluster. Subsequently, less dopamine elevation is required to trigger this extended SAMP cluster, which may account for the increasing antipsychotic resistance following relapses. From this perspective, refractory psychosis develops when the accrued SAMP becomes so extensive that it can be triggered by common everyday cues without dopamine elevation. In this state, psychosis can be driven by memory retrieval processes alone independent of dopamine activity and is therefore nonresponsive to dopamine-blocking medication. The resolution of psychotic symptoms would require nondopamine-related strategies such as normalization of the SAMP by extinction.
Accommodation of SAMP in AM
Newly encoded memory interacts with existing memories in the brain in processes that have been described as assimilation and accommodation (Armelin, Heinemann, & de Hoz, Reference Armelin, Heinemann and de Hoz2017; McKenzie, Robinson, Herrera, Churchill, & Eichenbaum, Reference McKenzie, Robinson, Herrera, Churchill and Eichenbaum2013; Preston & Eichenbaum, Reference Preston and Eichenbaum2013). New memories structurally similar to existing memory templates (schema) are assimilated as new exemplars in the existing schema (Takeuchi et al., Reference Takeuchi, Tamura, Tse, Kajii, Fernández and Morris2022). A new memory that does not fit easily into the existing schema may initiate a revision of the schema to accommodate the new information (McKenzie et al., Reference McKenzie, Robinson, Herrera, Churchill and Eichenbaum2013). This requires an active process that involves the hippocampus and the detection of novelty reflected in the prediction error signal (Duszkiewicz et al., Reference Duszkiewicz, McNamara, Takeuchi and Genzel2019; Goto, Reference Goto2022; Kamiński et al., Reference Kamiński, Mamelak, Birch, Mosher, Tagliati and Rutishauser2018; Shohamy & Adcock, Reference Shohamy and Adcock2010).
Recent high-resolution brain imaging studies have supported two convergent pathways in the hippocampus: (1) an entorhinal cortex-CA1 monosynaptic pathway and (2) a dentate-CA3-CA1 trisynaptic pathway (Lavenex & Amaral, Reference Lavenex and Amaral2000; Van Strien, Cappaert, & Witter, Reference Van Strien, Cappaert and Witter2009). The entorhinal cortex-CA1 pathway encodes new memories while the dentate-CA3-CA1 pathway enables associative retrieval of old memory (Bakker, Kirwan, Miller, & Stark, Reference Bakker, Kirwan, Miller and Stark2008; Leutgeb & Leutgeb, Reference Leutgeb and Leutgeb2007; Leutgeb, Leutgeb, Moser, & Moser, Reference Leutgeb, Leutgeb, Moser and Moser2007; Yassa & Stark, Reference Yassa and Stark2011). CA1 may act as a mismatch comparator between the new memory and the retrieved memory (Duncan, Ketz, Inati, & Davachi, Reference Duncan, Ketz, Inati and Davachi2012). Further, the encoding and retrieval pathways may switch between competitive and integrative modes, as regulated by prediction error, as well as acetylcholine, dopamine, and noradrenergic activities (Richter, Chanales, & Kuhl, Reference Richter, Chanales and Kuhl2016; Schlichting, Mumford, & Preston, Reference Schlichting, Mumford and Preston2015; Schlichting & Preston, Reference Schlichting and Preston2015). In the competitive mode, either encoding or retrieval is facilitated while the other is suppressed (Kesner & Rolls, Reference Kesner and Rolls2015; Neunuebel & Knierim, Reference Neunuebel and Knierim2014); in the integrative mode, cooperation between encoding and retrieval pathways facilitates interaction between new and old memories. These observations suggest that interactions between new and old memories in the brain are mediated through highly coordinated processes.
The roles of recovery style and nonbelieved memory
Clinical accounts after acute psychosis suggest that new event memories of psychosis are integrated with old nonpsychotic AM to different extents in people with different ‘recovery styles’ (Allé et al., Reference Allé, Potheegadoo, Köber, Schneider, Coutelle, Habermas and Berna2015; McGlashan, Reference McGlashan1987; Ridenour, Knauss, & Neal, Reference Ridenour, Knauss and Neal2021). In the ‘integrative’ recovery style, there is awareness of autobiographical continuity between the acute psychosis period and the nonpsychotic premorbid and remission periods. In contrast, in the ‘sealing over’ recovery style, patients ‘tend to isolate the psychotic experiences’ (McGlashan, Reference McGlashan1987).
Consistent with the SAMP hypothesis, the integrative recovery style was associated with better long-term functional outcomes (McGlashan, Reference McGlashan1987; Thompson, McGorry, & Harrigan, Reference Thompson, McGorry and Harrigan2003). The success of integration can be reflected in the level of coherence of AM, which is found to be reduced in schizophrenia patients (Allé et al., Reference Allé, Potheegadoo, Köber, Schneider, Coutelle, Habermas and Berna2015, Reference Allé, Gandolphe, Doba, Köber, Potheegadoo, Coutelle and Berna2016; Bisby, Horner, Bush, & Burgess, Reference Bisby, Horner, Bush and Burgess2018). However, the integration between SAMP and nonpsychotic AM has rarely been specifically studied.
A relevant area of approach in understanding the possible processes in memory integration is the experience of ‘non-believed memory’, which has been explored mostly in nonclinical populations. Nonbelieved memories are vivid AMs of events once believed to be veridical but the belief of which is subsequently withdrawn (Scoboria et al., Reference Scoboria, Jackson, Talarico, Hanczakowski, Wysman and Mazzoni2014; Scoboria, Nash, & Mazzoni, Reference Scoboria, Nash and Mazzoni2017). While previously thought to be rare and occur mostly in children (Otgaar, Wang, Fränken, & Howe, Reference Otgaar, Wang, Fränken and Howe2018), recent data suggest that nonbelieved memories are common in the general population (up to 35–50.6% in Li, Otgaar, Muris, & Chen, Reference Li, Otgaar, Muris and Chen2024). The withdrawal of a belief can occur as a result of social feedback; reappraisal of plausibility; attribution to a source other than memory; internal recollective characteristics; external details of the memory; general metacognitive belief about remembering; attributions about self or others, and personal motivation to alter belief (Scoboria et al., Reference Scoboria, Nash and Mazzoni2017). SAMP may undergo similar processes after remission.
Nonbelieved memories segregate into several subtypes based on how ‘non-believed’ and how ‘memory-like’ the representation is: (a) ‘high recollection with low belief’ characterizes a classical nonbelieved memory; (b) ‘high recollection and moderate belief’ characterizes a partial nonbelieved memory, which indicates that the belief is not completely relinquished by the weaker disconfirmatory evidence; (c) ‘moderate recollection and low belief profile’ characterizes a weaker recollection, which may be the result of strong disconfirmatory evidence (Scoboria et al., Reference Scoboria, Nash and Mazzoni2017).
Characterizing SAMP according to ‘recollection and belief’ profiling enables pragmatic subtyping of SAMP integration. These characterizations may mediate the relationship between SAMP and future relapse and treatment resistance. We hypothesize that a ‘high recollection-high belief’ (residual SAMP) would be associated with poorer outcomes (increased relapse propensity), with increasingly positive outcomes being indicated by ‘high recollection-moderate belief’ (partial integrated SAMP), ‘high recollection-low belief’ (well-integrated SAMP), and ‘low recollection-low belief’ (normalization of the well-integrated SAMP).
The roles of salience and context exposure in the extinction of SAMP
Memory traces for episodic events are consolidated after a period in which hippocampus connectivity is required. Afterwards, the role of the hippocampus diminishes as the memory traces are transferred to the neocortex (Goto, Reference Goto2022). Subsequent retrieval of the memory would require reactivation of the memory trace into a labile state with the possibility of reconsolidation (Lee, Nader, & Schiller, Reference Lee, Nader and Schiller2017). This reactivation process requires a state of salience involving a ‘prediction error’ (Lee et al., Reference Lee, Nader and Schiller2017). Through this process, newly acquired information reduces the tendency for the original response (including implicit associations and emotions). This normalizing effect is described as ‘extinction’ in animal studies across a wide range of vertebrate and invertebrate species.
Since psychosis and remission constitute different memory contexts, both psychologically and physiologically (Bouton, Reference Bouton2002), access to psychosis memory during remission may be compromised (Chen et al., Reference Chen, Wong, Tang, Lei, Suen and Hui2023). Integration between SAMP and AM enables access to AM as a nonbelieved memory. Through memory integration, extinction is facilitated by access to the well-integrated SAMP-AM representations (extinction requires reactivation of the original memory trace) (Bouton, Reference Bouton2002; Lee et al., Reference Lee, Nader and Schiller2017). In contrast, poorly integrated SAMP is more likely to be left isolated and barred from extinction processes.
Normalization of SAMP requires normal salience and broad context exposure
Extinction has been observed in implicit and explicit memory in humans (Exton-McGuinness, Lee, & Reichelt, Reference Exton-McGuinness, Lee and Reichelt2015; Lee, Reference Lee2008; Lee et al., Reference Lee, Nader and Schiller2017). It is a highly context-dependent process (Bouton, Reference Bouton2000; Eisenhardt & Menzel, Reference Eisenhardt and Menzel2007) where a newly acquired normalized response is linked only to a specific context and may not be effective in other contexts. Therefore, the acquisition of ‘extinction’ learning across many different contexts may be important for efficacy (Papalini, Beckers, & Vervliet, Reference Papalini, Beckers and Vervliet2020). Extinction requires new learning, which depends on the integrity of the dopamine-prediction error salience mechanisms and may be impeded by oversuppression of the dopamine e.g. by high-dose antipsychotic medication (Sumiyoshi, Reference Sumiyoshi2008). Antipsychotics at higher dosages (Malandain, Leygues, & Thibaut, Reference Malandain, Leygues and Thibaut2022; McEvoy, Hogarty, & Steingard, Reference McEvoy, Hogarty and Steingard1991) may cause subjective dysphoria with an ‘inability to feel or think’ (Awad, Reference Awad2019; Mizrahi et al., Reference Mizrahi, Rusjan, Agid, Graff, Mamo and Zipursky2007) and disable the normal salience response. Clinical improvements in treatment-resistant patients transiting from high-dose first- and second-generation antipsychotics to clozapine may be partly accounted for by relief from dopamine suppression, enabling normalization processes to take place.
Based on the knowledge from extinction processes, normalization of SAMP could be facilitated by promoting exposure to normal life experiences, tackling social withdrawal, avoiding overmedication with antipsychotics, treatment of comorbid depression, and confronting ‘safety behaviour’ (as in cognitive behavioral therapy for psychosis).
Clinical applications of SAMP in psychotic disorders and related conditions
The SAMP model has clinical implications for a spectrum of psychotic disorders, including but not limited to schizophrenia. Psychotic disorders manifest a substantial heterogeneity that is best handled by a multidimensional approach characterized by dimensions such as positive, negative, disorganization, affective and motor symptoms, each empirically associated with distinguishable brain substrates (Goghari, Sponheim, & MacDonald, Reference Goghari, Sponheim and MacDonald2010; Tandon, Nasrallah, & Keshavan, Reference Tandon, Nasrallah and Keshavan2009). The SAMP perspective primarily addresses the positive symptom dimension.
There is evidence that increased dopamine synthesis is involved in the first psychotic episodes (Cheng et al., Reference Cheng, Chang, Lo, Chan, Lee, Hui and Howes2020; Jauhar et al., Reference Jauhar, Nour, Veronese, Rogdaki, Bonoldi, Azis and Howes2017), during which SAMP formation is expected (Figure 2). The contents (cue-memory discriminability) and duration of active psychosis determine SAMP potency (see above; Berntsen, Reference Berntsen2023). Upon remission, new SAMP is no longer actively formed. Nevertheless, as aforementioned, patients with different recovery styles (‘integrative’ or ‘sealing over’) attain different AM integrations between SAMP and lifetime memories (McGlashan, Reference McGlashan1987; Thompson et al., Reference Thompson, McGorry and Harrigan2003). Well-integrated SAMP is more accessible to the moderating effects of new normalized experiences (extinction), resulting in reduced SAMP potency, and thereby more favorable outcomes. Poor integration, overmedication, and negative symptoms compromise this process.
Flow diagram showing the relationship between clinical processes and the postulated SAMP status.
Circumscribed psychotic symptoms not involving elevated dopamine activities do not lead to SAMP formation. Clinical high-risk states and schizotypal conditions may involve mild dopamine elevation leading to low-grade SAMP which may lead to persistent mild psychotic symptoms or decompensate into a frank psychotic episode. The first psychotic episode (FEP) results in SAMP formation. SAMP potency may depend on the content and duration of the FEP. Antipsychotic treatment results in remission except when SAMP potency is high, resulting in early treatment resistance. In remission, SAMP potency may reduce if integration with nonpsychotic autobiographical memory (AM) is good, and subsequent extinction is facilitated via exposure to normalized experiences. If SAMP-AM integration is poor, extinction is limited. If a relapse occurs, the new psychotic experience will strengthen the SAMP potency. Positive symptoms re-emergence depends on SAMP potency and dopamine state. A high-potency SAMP increases the risk of relapse and treatment resistance.
Furthermore, SAMP potency interacts with maintenance medication to determine relapse risks. With adequate maintenance, relapse can be minimized in most patients (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung and Honer2010; Zipursky, Odejayi, Agid, & Remington, Reference Zipursky, Odejayi, Agid and Remington2020) since context-dependent memory retrieval is inhibited by the changed dopamine psychophysiological context (Bouton, Reference Bouton2002). Although relapse takes place only in a small proportion of individuals with high SAMP potency (Rubio et al., Reference Rubio, Schoretsanitis, John, Tiihonen, Taipale, Guinart and Kane2020), we contend that the discontinuation of maintenance therapy would reinstate the dopamine salience psychophysiological context and lower the threshold for SAMP retrieval, leading to elevated risk of relapse even in patients with lower SAMP potency (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung and Honer2010; Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen and Chen2018; Kishi et al., Reference Kishi, Ikuta, Matsui, Inada, Matsuda, Mishima and Iwata2019). Notably, among those who remain on long-term maintenance medication and did not relapse, poor integration and extinction could also result in persistent SAMP, which might account for the lasting tendency for relapse even after long periods of remission (Chan et al., Reference Chan, Chan, Liao, Suen, Hui, Chang and Chen2022; Tiihonen, Tanskanen, & Taipale, Reference Tiihonen, Tanskanen and Taipale2018). The role of memory in relapse is suggested by the similarity in psychosis themes and contents between relapse and previous episode(s) (Grunfeld et al., Reference Grunfeld, Lemonde, Gold, Paquin, Iyer, Lepage and Shah2024). When a relapse occurs, new SAMP is appended onto the existing SAMP, resulting in an increase in SAMP potency and accounting for the observed increase in tendencies for further relapses and treatment refractoriness (Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen and Chen2018; Taipale et al., Reference Taipale, Tanskanen, Correll and Tiihonen2022).
The SAMP model is also relevant for those in a clinical high-risk (CHR) state. Some of them exhibit attenuated psychotic symptoms but do not cross the threshold into psychotic disorders. Individuals with CHR have been shown to demonstrate elevated dopamine synthesis capacities, particularly in those with higher symptom levels and those who eventually convert (Girgis et al., Reference Girgis, Slifstein, Brucato, Kegeles, Colibazzi, Lieberman and Abi-Dargham2021; Howes et al., Reference Howes, Bose, Turkheimer, Valli, Egerton, Valmaggia and McGuire2011, Reference Howes, Bonoldi, McCutcheon, Azis, Antoniades, Bossong and McGuire2020). For those who did not convert into psychotic disorders, continuous low-grade psychotic experiences are expected to result in low-potency SAMP which integrates more easily with nonpsychotic AM. Consistent with the prediction from SAMP, there is a tendency for the persistence of attenuated psychotic symptoms with an increased risk of developing psychotic disorder (Addington et al., Reference Addington, Cornblatt, Cadenhead, Cannon, McGlashan, Perkins and Heinssen2011; Woods et al., Reference Woods, Addington, Cadenhead, Cannon, Cornblatt, Heinssen and McGlashan2009). Similarly, the SAMP model predicts outcomes for schizotypal disorder, where long-standing attenuated psychotic symptoms are associated with evidence of dopamine dysfunction (Mohr & Ettinger, Reference Mohr and Ettinger2014) but less molecular imaging evidence of dopamine excess (Thompson et al., Reference Thompson, Rosell, Slifstein, Xu, Rothstein, Modiano and Abi-Dargham2020). Indeed, the expected persistence of attenuated psychotic symptoms and increased risk of conversion to psychosis in schizotypy have also been reported (Addington et al., Reference Addington, Cornblatt, Cadenhead, Cannon, McGlashan, Perkins and Heinssen2011; Albert et al., Reference Albert, Glenthøj, Melau, Jensen, Hjorthøj and Nordentoft2017; Woods et al., Reference Woods, Addington, Cadenhead, Cannon, Cornblatt, Heinssen and McGlashan2009).
Individuals in the nonclinical population may have isolated psychotic experiences (e.g., Morgellon disease, isolated hallucinations, as well as delusion-like ‘alien abduction’ and ‘past life’ experiences) without the typical decompensation seen in psychotic disorders (Clancy, McNally, Schacter, Lenzenweger, & Pitman, Reference Clancy, McNally, Schacter, Lenzenweger and Pitman2002; Meyersburg, Bogdan, Gallo, & McNally, Reference Meyersburg, Bogdan, Gallo and McNally2009; Nunziato, Egeland, Gurman, & Henry, Reference Nunziato, Egeland, Gurman and Henry2021). Studies have shown that dopamine elevation is not observed in people with isolated hallucinations (Howes et al., Reference Howes, Shotbolt, Bloomfield, Daalman, Demjaha, Diederen and Sommer2013). The lack of increased salience may explain why there is little cascading of the anomalous experience. Instead, these phenomena might be more related to metacognitive factors such as source memory weaknesses explained by a proneness to false memory (Clancy et al., Reference Clancy, McNally, Schacter, Lenzenweger and Pitman2002; Meyersburg et al., Reference Meyersburg, Bogdan, Gallo and McNally2009) than to increased dopamine salience pathways.
SAMP and PTSD symptoms
SAMP offers a parsimonious approach to understanding PTSD symptoms that are often also present in psychosis. Nevertheless, it should be highlighted that the phenomenological characteristics of psychotic and PTSD symptoms and the pathogenesis that underlies their emergence differ (Hardy, Reference Hardy2017; Samuelson, Reference Samuelson2011). Higher-level SAMP retrieval extending to schema levels (delusions) is usually only found in psychotic disorders (OConghaile & DeLisi, Reference OConghaile and DeLisi2015). Moreover, whereas psychotic hallucinations are predominantly verbal, more complex, and often linked with delusions, PTSD re-experiencing typically involves event memories with richer visual perceptual details (Bloomfield et al., Reference Bloomfield, Chang, Woodl, Lyons, Cheng, Bauer-Staeb and Lewis2021; Coughlan & Cannon, Reference Coughlan and Cannon2017; Morrison, Frame, & Larkin, Reference Morrison, Frame and Larkin2003). How memories of traumatic events are encoded in the AM, the processes underlying the involuntary retrieval of these memories, and their roles in the reactivation of PTSD symptoms might differ from the processes observed in SAMP for psychotic disorders should be an area for further research. This would offer important information to advance the understanding of the similar and distinct pathways underlying psychosis and PTSD, as well as inform clinical innovations.
Limitations and future directions
Given the lack of studies specifically addressing spurious AMs in psychosis, we opted for a narrative review of studies capturing the neurobiological and cognitive basis of psychotic disorders in the literature. The testable framework of SAMP offers directions for studying the long-term management of psychotic disorders in terms of minimization of residual SAMP by addressing minimizing the duration of active psychosis, facilitating integration between psychosis memory and nonpsychotic memory, and supporting normalization of psychosis memory through new learning across different life contexts and by avoiding overmedication.
While most of the work we reviewed is based on existing AM study methods in clinical and nonclinical populations, future studies should distinguish between AM encoded in the psychosis (SAMP) and other nonpsychotic life periods. A naturalistic longitudinal first episode study with AM measured at different time points may reveal further relationships between subtypes of SAMP (using profiling similar to that for nonbelieved memories described above) and clinical outcomes in terms of relapse, remission, recovery, and refractory psychosis. Brain imaging and electrophysiological measures targeting SAMP to extract individualized activity patterns in relevant brain areas during encoding, retrieval, and accommodation processes in AM tasks may add to phenomenological and cognitive observations. Importantly, focused investigations of AM integration and extinction processes (e.g., nonbelieved memory processes and context-dependent novel learning processes) via different methodological approaches can be critical for facilitating intervention development.
As explicated in this review, the SAMP approach could also be applied to a broader range of clinical conditions involving psychotic symptoms (e.g., substance-induced psychotic disorders and bipolar disorder with psychotic symptoms) in which dopamine-gated neuroplasticity could be involved. Notably, while the current approach focuses on AM, we consider AM to be integrated with implicit, semantic, and emotional memories. Spurious memory may also involve behavior, emotion, and semantic elements (e.g., Magioncalda et al., Reference Magioncalda, Martino, Conio, Lee, Ku, Chen and Northoff2020). Concerning developmental trauma, content-specific symptoms have so far only been reported in a small proportion of cases (Bendall, Jackson, & Hulbert, Reference Bendall, Jackson and Hulbert2010; Reiff, Castille, Muenzenmaier, & Link, Reference Reiff, Castille, Muenzenmaier and Link2012). Future studies are required to clarify its relationship with SAMP.
Further, current research suggests that dopamine encompasses multiple neurocognitive roles, including salience (Kutlu et al., Reference Kutlu, Zachry, Melugin, Cajigas, Chevee, Kelly and Calipari2021), prediction error (Millard, Bearden, Karlsgodt, & Sharpe, Reference Millard, Bearden, Karlsgodt and Sharpe2022), reward processing (Berridge, Reference Berridge2007), and neuroplasticity at the synaptic level (Speranza et al., Reference Speranza, Di Porzio, Viggiano, De Donato and Volpicelli2021). Their potential integration has yet to be fully understood (Kutlu et al., Reference Kutlu, Zachry, Melugin, Cajigas, Chevee, Kelly and Calipari2021; Richter, Reinhard, Kraemer, & Gruber, Reference Richter, Reinhard, Kraemer and Gruber2020). In addition to classical ‘rewards’, it is recognized that dopamine also drives behavior through nonhedonic ‘incentives’ (Ventura, Morrone, & Puglisi-Allegra, Reference Ventura, Morrone and Puglisi-Allegra2007). Interestingly, the potential role of ‘salient information’ as a form of ‘incentive’ is compatible with the view that humans are ‘informavores’ driven by the consumption of information (Pylyshyn, Reference Pylyshyn1984), a position aligned with the evolutionary social brain hypothesis (Dunbar, Reference Dunbar2009). Consistent with this perspective, partial reinforcement paradigms showed that unpredictable rewards provide stronger motivational drivers for behavior (Harris, Reference Harris2019). Integration between dopamine’s roles in handling ‘salient information’ and ‘rewards’ could be an important area for future exploration. The current SAMP model focused on the role of dopamine on neuroplasticity in psychosis. Future work can also explore the roles of other neurotransmitters (e.g., serotonin, noradrenaline, GABA, or acetylcholine) on SAMP.
Conclusion
Memories of psychotic experiences are a relatively neglected area in psychosis research. We addressed the question of what happens to the memory traces encoded during psychotic episodes using new findings in the interaction between dopamine and the memory systems, as well as the emergent knowledge about the natural history of memory traces. We argue that this parsimonious account may be sufficient to address some key features in the longitudinal evolution of positive symptoms. The SAMP memory framework provides novel and pivotal conceptual tools that can facilitate the understanding of clinically relevant findings concerning the course of psychotic disorders, including the incremental accruement of dopamine-independent refractory psychotic symptoms after relapse, the need for active cross-contexts extinction in rehabilitation, and the relevance of the recovery style.
Open access
