Disrupted habituation in the early stage of psychosis

Abstract

Background

Learning and memory are impaired in schizophrenia. Some theories have proposed that one form of memory, habituation, is particularly impaired. Preliminary evidence suggests memory impairment is associated with failed hippocampal habituation in chronic schizophrenia patients. Here, we studied how abnormal habituation of the hippocampus is related to relational memory deficits in the early stage of psychosis.

Methods

We measured hippocampal activity in 62 patients with early psychosis and 70 healthy individuals using fMRI. Habituation was defined as the slope of fMRI signal change to repeated presentations of faces and objects. Relational memory ability was measured as the slope of preferential viewing during a face-scene pair eye movement task outside the scanner.

Results

Early psychosis patients showed impaired relational memory (p < 0.001) and less hippocampal habituation to objects (p = 0.01) than healthy control subjects. In the healthy control group, better relational memory was associated with faster anterior hippocampal habituation (faces, r = −0.28, p = 0.03). In contrast, early psychosis patients showed no brain-behavior relationship (r = 0.12, p = 0.40).

Conclusions

We found evidence for disrupted hippocampal habituation in the early stage of psychosis, along with an altered association between hippocampal habituation and relational memory ability. These results suggest neural habituation may provide a novel target for early cognitive interventions in psychosis.

INTRODUCTION

Repetition is one of the most familiar memory tools. Repetition is often intentional (e.g., memorizing words by repeating them), but also shapes memory through automatic processes. As information is repeatedly encountered, a stimulus-specific reduction in neural activity—i.e., habituation—can be measured at both the behavioral and neural level (1–3). Habituation is a fundamental, highly conserved process through which repeated information is incorporated into memory (1; 4–8). Sustained processing, in contrast, may reflect continued attempts to incorporate information (9) or the continued processing of repeated information as if it were novel. Despite its ubiquity across the nervous system, substantial individual differences in habituation exist as early as infancy (10) and have been hypothesized to contribute to psychopathology (11).

Habituation deficits have long been observed in schizophrenia. As early as the 1960’s (12), numerous behavioral and electrophysiological studies have pointed to persistent habituation deficits in chronic schizophrenia patients (13–20). Early habituation studies hypothesized habituation deficits may underlie cognitive deficits in schizophrenia (12). One of the largest cognitive deficits is memory dysfunction (21–23); however, few studies have investigated whether habituation deficits are related to memory dysfunction in schizophrenia. Holt and colleagues (24) initially identified a deficit in hippocampal habituation to repeated faces in chronic schizophrenia patients. A recent study replicated and extended these findings, showing hippocampal habituation deficits were sustained and associated with poorer memory scores (25).

Memory impairments in schizophrenia are present early in the illness (22; 26–28), progress with illness duration (22; 28–30) and are strong predictors of functional outcome (31–33). Particularly striking are selective deficits in relational memory (34), a type of memory linked to hippocampal impairments (30; 35; 36). Although convergent evidence indicates habituation deficits exist in chronic schizophrenia, we do not know at what phase of illness they occur and whether they are linked to hippocampal-dependent relational memory performance.

Here we examined habituation and relational memory in the early stage of psychosis. Habituation was studied with functional magnetic resonance imaging (fMRI) signal change during passive viewing of faces and objects, while relational memory was studied with a face-scene viewing task. We defined habituation as the slope of fMRI signal change, and relational memory as the slope of change in preferential viewing of the matching face-scene pair (Figure 1). We hypothesized that neural habituation deficits exist in the early stage of psychosis, consistent with findings of existing memory deficits at the time of diagnosis (27; 28; 37). Because early hippocampal pathology is localized to the anterior hippocampus, we also hypothesized that habituation deficits are specific to the anterior hippocampus.

Figure 1.

We studied habituation and relational memory slopes (i.e., change over time). A. Habituation of the fMRI signal. The negative slope of the fMRI signal, during an a priori defined window of observation, was defined as habituation (green line). A sustained fMRI signal indicated impaired habituation (blue line). To accurately characterize habituation independent of novelty response differences, fMRI habituation slopes were adjusted for individual novelty response (b’). B. Preferential viewing of faces. The positive slope of preferential viewing (of a face previously paired with a background scene) was defined as relational memory (green line). A slope of zero indicated no preferential viewing and was defined as poor relational memory (blue line).

METHODS AND MATERIAL

Participants

We studied 62 patients in the early stage of psychosis, who were defined as: patients with a non-affective psychotic disorder (schizophreniform disorder (n=43), schizophrenia (n=17), and schizoaffective disorder (n=2)) within the first 2 years following their index episode of psychosis (Table 1). To specifically target early pathology (38), the majority of patients were recruited during the initial months of illness (i.e., the average duration of psychosis was 7 months, ranging from less than 1 month to no more than 24 months). More than 80% of the patients were in the first episode of psychosis (i.e., meeting criteria A for schizophrenia at the time of study inclusion), and half of the sample were studied after their first hospitalization for psychosis. On average, patients reported prodromal symptoms for 1.5 years. Patients were recruited from the inpatient units and outpatient clinics of the Vanderbilt Psychiatric Hospital. More than 80% of the patient sample were treated with antipsychotic medication at the time of the study.

Table 1.

Participant characteristics.

	fMRI sample					Relational memory sample
Demographic	Healthy Control	Early Psychosis	Statistic	df	p	Healthy Control	Early Psychosis	Statistic	df	p
Age, years	22 ± 2.9	22 ± 3.4	0.18	130	0.86	22 ± 2.9	22 ± 3.4	0.46	112	0.65
Sex (% male)	73%	79%	0.68	1	0.41	73%	78%	0.30	1	0.58
Race (white/black/other)	54/12/4	49/12/1	1.56	2	0.46	46/11/3	44/9/1	0.93	2	0.63
Handedness (% right)	91%	97%	1.54	1	0.22	90%	96%	1.59	1	0.21
Participant education, years	14.6 ± 1.9	13.6 ± 2.2	2.86	128	0.005*	14.6 ± 1.9	13.6 ± 2.3	2.66	110	0.009*
Parental education, years	15.0 ± 2.4	15.7 ± 2.6	−1.59	127	0.11	14.8 ± 2.3	15.8 ± 2.7	−1.36	109	0.23
IQ, WTAR	112 ± 10.8	105 ± 14.6	3.15	128	0.002*	112 ± 10.5	105 ± 14.9	2.74	110	0.007*
Clinical	Mean		Range			Mean		Range
Ham-D	11.5 ± 8.5		0 – 39			11.6 ± 8.7		0 – 39
YMRS	2.6 ± 4.7		0 – 20			2.7 ± 5.0		0 – 20
PANSS – total	68.0 ± 19.4		32 – 114			67.5 ± 20.5		32 – 114
PANSS – positive	17.9 ± 7.8		7 – 37			17.4 ± 7.9		7 – 37
PANSS – negative	17.0 ± 7.1		7 – 36			17.0 ± 7.4		7 – 36
PANSS – general	33.0 ± 8.8		16 – 59			33.1 ± 9.4		16 – 59
CPZ	299 ± 198		0 – 900			294 ± 203		0 – 900
Duration of illness, months	7.4 ± 5.8		0.9 – 24			7.3 ± 5.9		0.9 – 24
Duration of prodrome, months	18 ± 20.0		0 – 77			17 ± 18.3		0 – 77
First episode, %	84%		-			83%		-
First hospitalization, %	50%		-			50%		-
Antipsychotic treatment	84%		-			82%		-

Patients with early psychosis were compared to a group of 70 healthy control participants recruited from the surrounding community. All participants were assessed by a trained rater using the Structured Clinical Interview for the DSM-IV (SCID I-P) (39), and diagnoses were confirmed by a senior clinician (S.H.). Patients with early psychosis were assessed for current mood, anxiety, psychotic symptom severity, and intellectual function (see Supplementary Methods). There were no significant between-group differences in age, sex, race, handedness, or years of parental education (Table 1).

Exclusion criteria were age less than 16 or greater than 35, a history of significant head injury, major medical (i.e., HIV, cancer) or neurological illness, any contraindication for MRI scanning (e.g., pregnancy, metal implants, claustrophobia), current alcohol or substance abuse within the past month, estimated IQ < 75, and uncorrected vision deficits. Healthy control subjects were excluded for history of major mood or psychotic disorders, a first-degree relative with a psychotic illness, current substance abuse or dependence, or current psychotropic medication use.

This study was conducted in accordance with the Vanderbilt Human Research Protection Program and all participants provided written informed consent prior to study procedures. Participants received financial compensation for their time.

Experimental paradigm

Repetition task

Patients with early psychosis and healthy control participants completed a task designed to assess habituation to faces and objects. Participants viewed four 2-minute runs of a repeated neutral face (run 1 & 2) and a repeated neutral object (run 3 & 4). Images were presented for 500 ms followed by a 500 ms black screen. To promote and assess attention, a target detection task was included. Targets were small versions of the face or object images (25% of original size (40)) presented on 10% of trials. No targets were presented during the first 10 s of stimulus repetitions. Subjects were asked to press a button during each small image presentation. Target detection was high across both groups (detection mean > 93%; ps > 0.62; see Supplementary Results).

Relational memory

Outside the scanner, eye movements were recorded to assess memory performance during a face-scene pair task (41; 42) (Supplementary Figure 1). During training, participants viewed 36 face-scene pairs, with each pair presented once per block over three training blocks. Participants were instructed to memorize face-scene pairs, as they would be tested on the pairings later. During testing, participants were shown three previously-seen faces superimposed over a previously-seen background scene for 10 s. All faces and backgrounds presented during testing were equally familiar. During Match trials, one of the three overlaid faces (Match face) had been previously paired with the background scene. During Non-Match trials, none of the three overlaid faces (Non-Match faces) had been previously paired with the background scene. Participants were instructed to try to remember which of the three faces had been paired with the background scene during training and to focus their eyes on it as quickly as possible, or to keep their eyes focused on the computer screen if no matching face was detected. Fixation durations for each element (three faces + background) were summed in 250 ms time bins (e.g., 0–250, 250–500, 500–750 ms, ... 1750–2000) and averaged across test trials. Because hippocampally-driven preferential viewing occurs rapidly (43), viewing was examined over the first 2 s of each trial. Relational memory performance was characterized as the change in the proportion of time spent viewing the Match face (slope). Slopes were corrected for between-subject differences in initial eye position (first 250 ms) following face presentation (44; 45). For further task details see Supplementary Methods. Sixty healthy control subjects and 54 patients with early psychosis completed the relational memory task.

Imaging data acquisition and processing

Imaging data were collected on a 3T Philips Intera Achieva magnetic resonance imaging (MRI) scanner located in the Vanderbilt University Institute for Imaging Science. A high-resolution T1-weighted fast field echo (FFE) structural scan and four 2.5-minute functional echo planar images (EPI) were acquired for each subject (see Supplementary Methods for details). Functional data were processed in SPM12 (http://www.fil.ion.ucl.ac.uk/spm) and assessed for motion (46) and outliers (ART; Neuroimaging Informatics Tools and Resources Clearinghouse (NITRC)). Motion and outliers were similar between early psychosis and healthy control groups (ps > 0.16). Small images, outliers, and motion (rotation, translation, mean displacement) were entered into the first-level general linear model (GLM; 47) as regressors of no interest.

Data analysis

Regions of interest (ROIs)

Our goal was to test for hippocampal differences in early psychosis patients. Subject-specific ROIs were created using in-house automated multiatlas segmentation techniques (48; 49). Hippocampal multiatlas segmentations were manually divided into an anterior and posterior mask at the uncal apex (50; 51) by a trained rater (S.A.). To determine if an expanded network of regions implicated in the early course of psychotic illness (52–54) also showed habituation differences, we conducted secondary exploratory analyses in the striatum (caudate and putamen), visual cortex (calcarine cortex and occipital pole), fusiform face area (FFA), medial temporal lobe cortex (entorhinal cortex and parahippocampal gyrus (MTLC)), ventromedial prefrontal cortex (vmPFC), and amygdala. The FFA was defined in each subject using a localizer task (55) (Supplementary Figure 2; Supplementary Methods).

Habituation

Habituation was defined across two timescales: habituation to the stimulus visual properties (rapid); and habituation to stimulus category (slow). For rapid habituation (Figure 1), the habituation window was defined as the peak signal magnitude following the initial presentation of a stimulus and the volume in which the mean residual signal returned to baseline in the healthy control group. Healthy control signal peaked 8 s following stimulus onset, and returned to baseline by 18 s following stimulus onset, resulting in a 10 s habituation window. This window is consistent with prior fMRI studies of habituation in healthy subjects (56); additionally, this window did not include any small image presentations to minimize the effects of intervening images on habituation results (2).

To analyze slow changes in signal related to stimulus category, 10 s time bins (n = 12) were modeled using the first-level GLM, and residuals from the first-level GLM were entered into a second-level analysis. Novelty response was defined as signal in the first 10 s time bin, and habituation was defined as the slope of change in signal from the 1^st to the 12^th time bin for each run.

For each analysis, average timeseries for each ROI was extracted from participants’ residual data using MarsBar (57). Residual time courses were plotted for each subject and time course calculations were conducted using Matlab R2017b (The MathWorks Inc., Natick, MA). Signal in the left and right hemispheres were highly correlated across regions; to increase statistical power and minimize type I error, data were averaged across hemispheres.

Habituation slope

Habituation is highly dependent on novelty response—there is more opportunity for signal to attenuate over time if signal is initially high. Because we were interested in examining differences in rate of habituation independent of differences in novelty response, we calculated a normalized habituation slope (b′), corrected for novelty response differences, for each participant (44; 45; 58). Novelty response was defined as the peak signal magnitude following the initial presentation of a stimulus in the healthy control group (Figure 1). Habituation slope (b′) values were calculated for each participant using linear regression analysis (see Supplementary Methods for details). Adjusted habituation slopes were calculated separately for faces and objects.

Habituation over consecutive runs

Because faces were presented prior to objects, we explored effects of time on habituation differences by examining habituation over the course of the experiment (see Supplementary Methods and Results).

Statistical analysis

Linear mixed effects models tested for group differences in neural response to stimuli with stimulus type (face, object), region (anterior, posterior hippocampus), and group as fixed factors, and participant as a random factor. For consistency with prior analyses (25) the first run of each condition was used for novelty and habituation analyses, as neural response is minimal following the initial habituation run (Figure 2). Regional habituation greater than zero and novelty response greater than baseline was examined using one-sample t-tests (p = 0.05). Between-group differences in relational memory were tested by analysis of variance (ANOVA). Spearman correlations tested for associations between novelty response or habituation and relational memory function (p = .05). Spearman correlations test for monotonic rather than linear relationships between variables. Correlation coefficients were compared using z-scores (59). Statistical analyses were performed using SAS software v9.4 (SAS Institute Inc., Cary, NC).

Figure 2.

Hippocampal habituation to repeated faces and objects. Mean fMRI signal for the first 20 s of stimulus repetitions in each run is shown for the anterior (A) and posterior (B) hippocampus. Shaded area indicates the 95% confidence interval. Patients with early psychosis (red) and healthy control subjects (blue) show an initial novelty response to visual stimuli followed by a decrease in mean fMRI signal with stimulus repetition. Habituation slopes were calculated for a 10 s habituation analysis window, indicated by black bars below the fMRI signal (runs 1, 3). Slopes were corrected for differences in initial novelty response (b′). Dot plots of habituation slopes (b′) are shown for the anterior (C) and posterior (D) hippocampus. Black bars indicate mean habituation slope ± 95% confidence interval. Healthy control subjects had larger negative b′ slope values to repeated objects compared to early psychosis patients, indicating greater habituation of fMRI signal. Asterisks indicate between-group comparisons significance for habituation slopes. ** p = 0.01.

RESULTS

Novelty response

We first tested for novelty responses in the hippocampus. All hippocampal regions showed a novelty response greater than baseline for both faces and objects (Supplementary Results). Novelty responses were similar across groups (no main effect of group, p = 0.34).

Habituation

Rapid habituation

We examined the ability of the hippocampus to rapidly habituate over a 10 s window following the peak novelty response. Participants showed evidence for habituation to both faces and objects across hippocampal regions (one-sample t-test, ps ≤ 0.002, Supplementary Table 2, Supplementary Results). Habituation was faster for faces than objects (main effect of stimulus type, F _1,130 = 11.38, p = 0.001) and for anterior hippocampus compared to posterior hippocampus (main effect of region, F _1,130 = 8.54, p = 0.004). Overall, healthy control subjects showed greater habituation than early psychosis patients across hippocampal regions (main effect of group, F _1,130 = 3.77, p = 0.05; no group by region interaction, p = 0.18). However, groups differed in their habituation to faces and objects (group by stimulus type interaction, F _1,130 = 9.38, p = 0.003). Posthoc linear mixed analyses of faces and objects separately showed that group effects were driven by differences in habituation to objects—habituation to objects was greater in healthy control subjects compared to early psychosis patients (object, main effect of group, F _1,130 = 6.48, p = 0.01), while habituation to faces was similar across groups (face, main effect of group, F _1,130 = 0.01, p = 0.94).

Habituation over consecutive runs

To explore whether stimulus type effects may be related to presentation order (faces before objects), we tested for linear changes in habituation rate across the four consecutive runs. Hippocampal habituation rate decreased over consecutive runs (main effect of time, F _3,390 = 19.37, p < 0.001; no time by region interaction, p = 0.18), with early psychosis patients showing less habituation over time than healthy control subjects (group by time interaction, F _3,390 = 4.02, p = 0.006). Posthoc analysis of variance revealed the group by time effect was driven specifically by greater habituation in the healthy control group to the initial objects run (run 3) compared to early psychosis patients (F _1,130 = 6.10, p = 0.02, Supplementary Figure 3), while habituation was similar between groups across the remaining runs (ps ≥ 0.83).

Slow habituation

We next examined habituation over the full two minutes of stimulus exposure. Hippocampal habituation slopes were not significantly different from zero (Supplementary Table 2, Supplementary Results), indicating no habituation over two minutes. There was also little evidence for a hippocampal novelty response in the first time bin (Supplementary Table 1), suggesting a floor effect when averaging signal over 10 s time bins. Habituation did not differ by stimulus type, region, or group (ps ≥ 0.35).

Potential moderators

Hippocampal habituation in early psychosis patients was not predicted by medication status (chlorpromazine equivalent units), current psychosis symptoms, or state anxiety (ps ≥ 0.12, see Supplementary Results for further details).

Memory correlations

Healthy control subjects were better at identifying face-scene relational pairs than early psychosis patients (F _1,113 = 22.31, p < 0.001; Figure 3). In healthy control subjects, greater anterior hippocampal habituation to faces was correlated with better relational memory (r = −0.28, p = 0.03; Figure 3). Removal of an outlier healthy control subject did not significantly decrease the correlation (r = −0.25, p = 0.05). In contrast, early psychosis patients did not show the normal association between anterior hippocampal habituation and relational memory (r = 0.12, p = 0.40; significant between-group difference in correlational coefficients, z = −2.11, p = 0.03). Relational memory was not correlated with posterior hippocampal habituation or habituation to objects in either group (ps ≥ 0.09). Novelty response was not correlated with relational memory (ps > 0.18).

Figure 3.

Relational memory and habituation in the anterior hippocampus. A. The slope of preferential viewing of the correct face-scene pairing is shown for patients with early psychosis(red) and healthy control subjects (blue). Black bars indicate mean relational memory slope ± 95% confidence interval. Faster preferential viewing (positive slopes) indicate better relational memory. Patients with early psychosis had worse relational memory for trained face-scene pairs compared to healthy control subjects. B. In healthy control subjects, greater habituation in the anterior hippocampus was correlated with better relational memory performance. There was no relationship between habituation and relational memory in early psychosis patients. Asterisks indicate between-group comparisons significance for habituation slopes. * p ≤ 0.01.

Exploratory analysis

To determine whether habituation deficits in early psychosis were specific to the hippocampus, or rather, representative of neural processing deficits across a broader set of brain regions, we examined neural response in a set of brain regions commonly implicated in schizophrenia, including the striatum, visual cortex, FFA, MTLC, vmPFC, and amygdala. Exploratory regions significantly differed in their pattern of habituation and were analyzed separately (main effect of region, F _8,1031 = 37.31, p < 0.001; region by group interaction, F _8,1031 = 2.91, p = 0.003; Supplementary Results). Over the first 10 s of stimulus exposures, healthy control subjects showed greater habituation than early psychosis patients in two regions—the occipital pole and FFA (occipital pole, F _1,130 = 12.21, p < 0.001; FFA, F _1,130 = 6.44, p = 0.01). For both regions, group patterns were similar for faces and objects (no group by stimulus type interaction, ps ≥ 0.24). Habituation to faces in the occipital pole tended to be associated with better relational memory in healthy control subjects (r = −0.25, trend p = 0.06), but not early psychosis patients (r = 0.04, p = 0.76; Supplementary Figure 4).

DISCUSSION

We examined hippocampal habituation, or decrease in response, and its behavioral correlates in a group of early psychosis patients. We found two notable differences when comparing patients with healthy control subjects. First, early psychosis patients had less hippocampal habituation in response to repeated objects relative to healthy control subjects. Second, despite similar hippocampal habituation to faces across groups, the rate of face habituation was associated with differential behavioral performance on a relational memory task between groups. In healthy control subjects, greater anterior hippocampal habituation to faces was associated with better relational memory. In contrast, in early psychosis patients, there was no relationship between anterior hippocampal habituation to faces and relational memory performance. Brain-behavior associations were specific to the anterior hippocampus, consistent with evidence of preferential pathology in anterior hippocampus during the early phase of schizophrenia (60–62), and were not associated with current positive, negative, or general clinical symptoms, suggesting a trait-like relationship. Together, these findings suggest reduced habituation in the anterior hippocampus is associated with relational memory deficits in the early stage of psychosis.

Early psychosis patients showed failed habituation to repeated objects, but not faces. During an initial block of repeated faces, groups habituated similarly; however, in a later block of repeated objects, healthy control subjects continued to show the expected pattern of rapid hippocampal habituation (63–65) while early psychosis patients showed a sustained hippocampal response characteristic of failed habituation. One explanation may be that early disruptions in hippocampal processing are specific to non-social, but not social, information. However, a consistent literature has shown deficits in social cognition in schizophrenia (66; 67), making this explanation less likely. Alternatively, habituation deficits may be related to task order. For consistency with our previous studies (25), faces were always presented first. When modeling changes in the rate of habituation across the four consecutive runs, we found that early psychosis patients did not show the same pattern as healthy control subjects over time. Habituation rate is exquisitely sensitive to time and interval effects, making this explanation plausible. Habituation deficits in early psychosis patients may be associated with an inability to sustain rapid habituation over time and/or following changes in stimuli. However, further studies stringently controlling for time and stimulus type are needed.

To determine whether hippocampal habituation was associated with behavior, we examined the relationship between habituation patterns and performance on hippocampal-based relational memory task (68). Hippocampal habituation to faces was associated with better relational memory in healthy control subjects, but not early psychosis patients, suggesting a disrupted brain-behavior relationship. These findings are consistent with our prior findings of a disrupted relationship between habituation to faces and memory ability in chronic schizophrenia patients (25). We now extend prior findings to show that disruptions can be detected early in the illness at a younger age as deficits in object processing and are specific to the anterior hippocampus.

Deficits in hippocampal function are among the most consistent and replicable findings in schizophrenia. Hippocampal neurobiology is altered at the earliest stages of illness (52) with strong evidence for an imbalance in inhibitory/excitatory signaling originating in the anterior CA1 region during early illness (62; 69). In a previous study we found that chronic schizophrenia patients also show deficient hippocampal habituation to faces, and that hippocampal habituation to faces was associated with memory ability (25). Here, we extend these findings to show a specific relationship between anterior hippocampal habituation to faces and relational memory, a form of memory dependent on hippocampal integrity. In contrast, novelty response was not associated with relational memory, suggesting habituation may be a selective marker of hippocampal pathology.

We conducted an exploratory analysis to determine whether habituation deficits were specific to the hippocampus, or rather, reflected a broader neural processing deficit in patients. Of the nine additional regions investigated, two regions—the fusiform face area and occipital pole—also showed habituation deficits to both faces and objects in early psychosis patients. The occipital pole showed a similar brain-behavior relationship across groups as the anterior hippocampus. Because the hippocampus has strong reciprocal connections with the visual cortex, findings could result from either feedforward or feedback mechanisms. However, our data do not support a widespread feedforward effect originating with the hippocampus or visual cortex, as we did not find habituation deficits in other brain regions with strong hippocampal and visual processing connections (e.g., amygdala, parahippocampal cortex).

Although habituation is a simple form of learning (70–72), ubiquitous across the nervous system, and highly conserved across species (73; 74), little is known about the underlying cellular and molecular processes (73; 75–77). An influential early model proposed that, as a stimulus is repeated, feedback inhibition promotes strong inhibition of continued novelty response, thus yielding habituation of neural response and a memory trace (78). The hippocampus may be uniquely organized to facilitate this type of response—the human hippocampus shows a strong novelty response (79–81) and strong interneuron-mediated suppression of firing to repeated information (82–84). In schizophrenia, there is compelling evidence for interneuron dysfunction in the hippocampus (85). Habituation differences may also result from interactions across neural systems (71). Schizophrenia has been described as a disorder of widespread connectivity deficits (89; 90), with evidence for degraded facial visual processing (86) and altered salience system activity (63; 87; 88), which may converge to increase hippocampal novelty response. Our current findings do not support altered salience interactions, as we did not find fMRI differences in the striatum. However, we do find habituation deficits in the ventral visual system, consistent with prior findings in chronic patients (25). Future work should investigate connectivity of hippocampal connections to disentangle within-hippocampus deficits from broader network dysfunction.

There are a variety of factors that influence habituation, including attention, anxiety, arousal, and saccadic suppression (3; 45; 91–93). It is unlikely our current findings are due to differences in attention, as target detection performance was high and did not differ between groups. Similarly, measures of trait anxiety were not correlated with habituation. Because the experiment did not include a measure of arousal at the time of the scan, it is possible that arousal during the scanning session differed between groups. However, we did not detect habituation differences in the striatum (94), rendering arousal differences less plausible. Eye movements during the habituation task may also alter habituation, as visual processing is suppressed during saccades (93); future studies should monitor eye movements during scanning.

Memory impairments in schizophrenia are substantial (95), progress with illness duration (28), and are strong predictors of functional outcome (31–33). We previously found that hippocampal habituation was disrupted in chronic schizophrenia and associated with memory impairment (25). Here we find that habituation is also disrupted in patients with early psychosis and associated with relational memory ability. Together, these findings further our understanding of the pathophysiology of schizophrenia and suggest that habituation may be useful as a marker of neurocognitive function and illness progression, although longitudinal studies are necessary.