Sensory Memory in Schizophrenia Spectrum Disorders

Sensory memory (SM) refers to the window over which the neural response to a sensory stimulus remains active after the stimulus is withdrawn. The duration of the memory trace is brief (e.g., (Sperling, 1960) and considered to be preattentive (Naatanen et al., 1982). Measures of SM are appealing as a clinical biomarker because they are unaffected by attentional focus - which is compromised in schizophrenia. The SM trace is an input for predictive coding and very sensitive to changes in sensory stimuli (Winkler, 2007; Winkler et al., 2009). It is also extremely adaptive to the current environment (May & Tiitinen, 2010; May & Tiitinen, 2007). Abnormal SM in those with schizophreniamay be domain general (Neuhaus et al., 2013; Saint-Amour et al., 2007). To date the most research has been conducted in audition and vision, although sensory memory in the somatosensory and olfactory domains is reduced in schizophrenia (Somatosensory: (Ludwig et al., 2016) Haigh et al., 2016; Olfactory: (Moberg et al., 2003; Wu et al., 1993)) and can be tracked using similar methods to vision and audition (Restuccia et al., 2009). Despite the numerous reports of oculomotor dysfunction in schizophrenia and SSD, relatively little has been reported in visual SM. We will now discuss each in turn.

Auditory Sensory Memory

Poor auditory SM impairs the encoding of simple parameters of sounds. Neural measures of auditory SM are a clinically useful biomarker. They are minimally affected by attention and can be measured before the participant is aware of the percept, thereby isolating early sensory processing from upstream cognitive processes.
In people with schizophrenia there are numerous examples of deficits in auditory SM, including impaired pitch discrimination (distinguishing between tones of different pitches; (McLachlan et al., 2013), tone-pair matching (Rabinowicz et al., 2000), sound localization (Donde et al., 2020; Donde, Martinez, et al., 2019), discrimination of intensity, duration and sequences (Schnakenberg Martin et al., 2018) and auditory scene segregation (Ramage et al., 2012; Weintraub et al., 2012). Moreover, there is evidence that mismatch negativity amplitude is lower in those with schizophrenia, reflective of a general auditory sensory memory deficit (Catts et al., 1995), lower educational achievement before the onset of schizophrenia, worse cognitive symptoms, poorer global function, and longer duration of illness (Friedman et al., 2012). There is some question over whether these abnormal auditory responses are due to sensory memory or if they are due to problems encoding the information in the first place. While it is difficult to disentangle how poor early encoding impacts the ability to encode short-term memory traces, there is evidence that both individuals with chronic schizophrenia and individuals at their first-episode of psychosis show deficits in encoding deviants in complex patterns (Haigh et al., 2016; Haigh, Coffman, et al., 2017; Salisbury et al., 2018), suggesting that memory impairments persist after encoding. However, other aspects of audition are preserved, including the time-course of auditory SM (March et al., 1999) and the ability to categorize syllables (Dale et al., 2010; Haigh et al., 2019). Categorization can also show the expected shifts when participants are habituated to one syllable (for example, adapting to one syllable category makes it easier to detect a syllable in another category) (Haigh et al., 2019) suggesting residual plasticity in the auditory system. Finally, more closely associated with specific symptoms of schizophrenia, auditory hallucinators reporting phenomenon of speech and voices exhibit worse auditory SM (McCleery et al., 2018), highlighting the link between auditory processing and symptomology.
There is evidence of similar, but smaller, SM abnormalities in schizotypy that are similar to those reported in schizophrenia. However, the number of studies focusing on schizotypy are relatively small and often identify potential biomarkers in isolation. For example, sensory cortices send a ‘corollary discharge’ indicating a self-generated response compared to external in origin (Sperry, 1950). Auditory corollary discharges are produced in the superior temporal gyri (Creutzfeldt et al., 1989; Muller-Preuss & Ploog, 1981) and are manifest by reduced auditory N1 ERP responses to self-generatedspeech compared to the N1 response to externally generated speech (Ford & Mathalon, 2005). Individuals with schizophrenia show equivalent N1 amplitudes for self-generated and externally generated speech, suggesting they cannot differentiate between conditions. Impaired corollary discharge could reflect a confusion between self-generated auditory hallucinations and externally generated speech (Ford & Mathalon, 2005). In schizophrenia there is a reliable reduction in N1 amplitude (Salisbury, 2010; Shelley et al., 1999); reviewed in: (Rosburg et al., 2008) that correlate with the perceptual deficit (Javitt et al., 2000). Similarly, individuals with high schizotypy show equally large N1 responses to external and self-generated speech, whereas neurotypical participants with low schizotypy show N1 suppression to self-generated speech (Oestreich et al., 2016). Those high in schizotypy show reduced precision in auditory sensory memory despite normal rates of acquisition and decay (Bates, 2005).
In addition, reductions in well-characterized cognitive-related ERPs in combination with sensory ERPs to simple auditory stimuli have shown some utility. For example, reductions in the P3 complex (Javitt et al., 1995; Pritchard et al., 1985), are evident early in the disease course (Devrim-Ucok et al., 2016; Salisbury et al., 2020), as are reductions in steady-state potentials (Light et al., 2006), and in the slow-wave potentials associated with auditory scene segmentation (Coffman et al., 2018; Coffman et al., 2017) that correlate with hypoactive midcingulate cortex (Coffman et al., 2018). Reduced amplitude of P50, mismatch negativity, P3, and antisaccades (making eye movements in the opposite direction to a cue) are reported in individuals with schizophrenia, and significant reductions in P50 and P3 in relatives, compared to controls. However, combining all four measures in a multivariate analysis permitted the model to accurately (80%) classify group assignment (Price et al., 2006). In short, throughout early auditory sensory processing there are atypical ERPs in the schizophrenia population. Further investigation in high schizotypy individuals to expressly investigate auditory SM would be valuable to understand the boundary conditions between intact and impaired function in SSD.

Auditory Mismatch Negativity

The most robust neural measure of SM is the mismatch negativity (MMN). In neurotypical individuals, fronto-central electrodes record larger (more negative) responses to an infrequent deviant stimulus than to the standard stimulus at around 100-200ms after stimulus-onset (Naatanen, 1985). The MMN is calculated by subtracting the EEG response to the standard tone from response to the deviant tone producing a difference waveform (Näätänen, 1985). In individuals with chronic schizophrenia, the MMN is smaller (Cohen’s d: 0.99) (Javitt & Sweet, 2015; Umbricht & Krljes, 2005), and magnetic activity in the planum temporale (secondary auditory cortex) is reduced during auditory SM tasks that generated a smaller (magnetic) MMN (Kircher et al., 2004). Reduced MMN generalizes across stimuli (Erickson et al., 2016; Umbricht & Krljes, 2005) and includes responses to pattern deviants based on grouping (Haigh et al., 2016), pitch (Haigh, Matteis, et al., 2017), rules (Haigh et al., 2019), and phonemes (Fisher et al., 2019).
In SSD the degree of MMN reduction correlates with symptom severity (Fisher et al., 2019; Light & Braff, 2005). Greater reductions are apparent in chronic schizophrenia compared to individuals at high-risk (Shin et al., 2012). Local and global change detection have both been found to be poor in schizophrenia suggesting evidence of abnormal predictive coding that cannot be due to abnormal adaptation to simple stimulus parameters (Kirihara et al., 2020). Together, these data highlight a pervasive deficit in preconscious auditory SM that occurs early in the auditory processing stream (Escera & Malmierca, 2014). MMN is sensitive to the balance between excitatory and inhibitory neurotransmitters in the auditory cortex (Holliday et al., 2018). There is some debate regarding neurotransmitter systems’ roles in the auditory MMN. Some evidence indicates a restored MMN with nicotinic agonists, implicating acetylcholinergic modulation (Baldeweg et al., 2006) but see (Knott et al., 2014), whereas other studies implicate acetylcholinergic (Kantrowitz et al., 2018) or serotonergic influence (Sehatpour et al., 2022).
In SSD more broadly, there is some evidence of reduced MMN. For example, MMN to pitch oddballs successfully discriminates between neurotypicals and people diagnosed with schizotypal personality disorder (Niznikiewicz et al., 2009). Furthermore, combining MMN responses to multiple deviants: duration, gap or silence instead of a tone, or location of tone, lead to identification accuracy up to 80.5% when using multivariate machine learning techniques. The combined MMN accuracy correlated with global assessment of functioning, highlighting auditory MMN as a biomarker of symptoms in chronic schizophrenia (Taylor et al., 2017). MMN deficits are evident in early course schizophrenia (within 2-years of the first episode of psychosis; (Salisbury et al., 2007)). There are inconsistent findings in those within 6-months of initial psychosis (Erickson et al., 2016; Haigh, Coffman, et al., 2017), and in those at-risk of developing schizophrenia (compare (Shin et al., 2012) to (Hsieh et al., 2019)), or who are unaffected first-degree relatives (Magno et al., 2008). Several accounts address why MMN deficits are greater in those with advanced disease. One theory is that MMN is tied to symptom severity (Kantrowitz et al., 2018), as MMN correlates with negative symptoms in chronic (>5 years) patients (Catts et al., 1995). Of relevance, the MMN correlates with negative symptoms in first-episode patients, and with measures of social functioning (Murphy et al., 2020). Certainly, individuals with chronic schizophrenia have the most severe symptoms and neurological consequences. For example, MMN correlates with reduced grey matter volume in primary auditory cortex and both MMN deficits and the great matter volume reductions increase with the duration of time past first-episode (Salisbury et al., 2020; Salisbury et al., 2007). A recent meta-analysis indicated that MMN shows progressively greater abnormality, with first degree relatives showing the smallest deficit, followed by individuals at high-risk of developing schizophrenia, and finally recently diagnosed (early-stage schizophrenia), compared to those with chronic schizophrenia (Erickson et al., 2016); Figure 3).