|Home | About | Journals | Submit | Contact Us | Français|
Schizophrenia illness is characterized by significant impairments in long-term episodic memory, which are associated with hippocampal abnormalities. The current study assessed long-term memory for preference conditioning, which is believed to be more strongly based in the basolateral amygdala, in order to determine whether abnormalities in biological systems supporting long-term memory are specific to the hippocampus, or shared across brain regions involved in different types of memory.
18 schizophrenia (SC) and 24 healthy (HC) subjects, matched on age, sex, and years of education participated in the study. All subjects completed an implicit preference conditioning task which associated different patterns with different frequencies of reward. Subjects were then tested for their preference for the patterns both immediately after training, and following a 24 hour delay.
Both SC and HC subjects demonstrated a preference for the more frequently rewarded pattern immediately after training. Following a 24 hour delay, HC subjects continued to prefer the more rewarded pattern in contrast to the less rewarded pattern, but SC subjects did not maintain this differentiation.
These data suggest a significant deficit in the ability to maintain stimulus-reward relationships in memory over long delay periods (24 hours) in individuals with schizophrenia. This data is consistent with prior research indicating normal response to emotional stimuli during learning, but impaired long-term memory for the stimuli, and suggest that there may be a common abnormality in biological systems supporting consolidation of long-term memory across multiple types of memory in individuals with schizophrenia.
Impairments in emotional functions are central to the functional deficits associated with schizophrenia (1-3). Despite evidence of abnormalities in motivation, hedonic tone, and emotional expression, individuals with schizophrenia have been repeatedly shown to self-report normative emotional responses to a variety of emotional stimuli, including emotional pictures (4-5), movies (6-8), and foods (6-7). This inconsistency between reports of emotional response “in the moment of exposure” and both self-report and clinician ratings (9) indicating significant decreases in the experience of pleasure in schizophrenic individuals has required reconsideration of what particular aspect of emotional processing might be impaired in schizophrenia. Kring and colleagues (10) have suggested that individuals with schizophrenia may have intact consummatory pleasure, but impaired anticipatory pleasure. Consummatory pleasure refers to enjoyment of stimuli or activities at the moment of engagement, while anticipatory pleasure refers to the ability to effectively utilize cognitive representations of past pleasurable experiences to increase interest in pursuing similar activities in the future. Research on reward processing similarly suggests intact reward sensitivity in individuals with schizophrenia (11).
The construction of representations of past experience is dependent on the biological processes underlying memory. Further, there is clear evidence that the brain response to emotionally significant stimuli includes the initiation of neuroendocrine and/or neurotransmitter processes that effectively enhance late long-term potentiation, and thus support better memory for emotional in contrast to neutral experiences (12-13). Past research in this area indicates that modulatory effects of emotion particularly influence late long-term potentiation, and effects are most apparent after sleep periods (14-16).
The current study used a preference conditioning task to first assess initial learning of stimulus-reward relationships in individuals with schizophrenia, and second, to determine whether individuals with schizophrenia demonstrate impairment in long term retention of stimulus-reward relationships. By using an incidental learning preference conditioning paradigm, the study focuses specifically on the ability to learn stimulus-reward relationships, and limits the contribution of factors such as differences in ability to effectively recruit or use encoding strategies. This task specifically assesses whether subjects learn different emotional responses to different patterns based on their reinforcement history and can more specifically assess abnormalities in learning relationships between stimuli and rewards, and maintenance of reward contingencies over time, relying primarily on processes occurring in the medial temporal lobe (cf., 17-19).
Eighteen individuals meeting DSM-IV criteria for schizophrenia spectrum disorders (schizophrenia, schizoaffective disorder) (SC) were recruited at the University of Illinois at Chicago, and via ads in the community and special-interest websites (NAMI, NIH) and physician referral. Twenty-four healthy individuals (HC) were recruited via ads in the community. Potential participants were excluded if they reported histories of head trauma with loss of consciousness greater than 15 minutes, current substance abuse, or neurological abnormalities that might influence cognitive functioning. Diagnoses for all participants were established using the Structured Clinical Interview for DSM-IV diagnosis (20). Severity of symptoms was rated by clinicians blind to individuals' task performance using the Positive and Negative Syndrome Scale (PANSS: 21) and the Hamilton Depression Rating Scale (HDRS; 22). Premorbid IQ was estimated using the Wide Range Achievement Scale (WRAT: 23)
All SC participants were clinically stable outpatients. Fourteen of the SC participants were taking second generation antipsychotic medications, one was taking a first-generation antipsychotic medication, and four were not taking any antipsychotic medications. Medicated subjects were tested after a minimum of 4 weeks on a stable medication regimen. Subject characteristics are shown in Table 1. The study protocol was approved by the Institutional Review Board of the University of Illinois at Chicago. All participants had to demonstrate the ability to describe the study demands, and articulate how they could terminate participation, as part of the consent process. Written informed consent was obtained from all participants.
This task is a slight modification of a preference conditioning task that has been successfully used in studies of human subjects with brain disorders (cf. 24-25). It is designed to assess incidental preference conditioning, which is believed to be dependent on the amygdala, and closely replicates conditioned place preference tasks used in animal studies. The first two phases of the task were directly based on Johnsrude and colleagues' task (25), and included a preference formation phase and a preference testing phase completed immediately after learning. A third phase was added to the task to assess retention of conditioned preferences following a 24 hour delay. The task is illustrated in Figure 1. As a check on subject awareness of the conditioning, we included a final phase (based on the original Johnsrude et al paper), which required that subjects report the rationale for their preferences. Given our focus on indirect conditioning, if subjects were aware that their preferences were influenced by their reward history, their data was not included in statistical analyses. Data for one SC subject was excluded for this reason.
In the initial conditioning phase, subjects were presented with three boxes on a computer screen in each trial, and told to select a box to “open” (See Figure 1). Subjects were instructed that their goal was to accurately guess which box contained a red ball on each trial, and that they would receive reward points for every occasion on which they found the red ball. Subjects were also instructed that they should try to remember how often they found the red ball in the three different boxes, as they would be tested on this information. Subjects completed a total of 180 trials in the learning phase, and there was no time limit for responding.
Each time a box was selected and “opened” either a red or black ball would be shown in front of an un-namable symmetric black-and-white background for three seconds, and then the box closed and a new trial began. When a red ball was revealed, subjects heard a musical flourish and received feedback that they had “won 10 points”; when a black ball was revealed a buzzer would sound and the feedback was given “you did not win any points”.
Unbeknownst to subjects, the specific order of pattern and ball color presentations was pre-determined, such that a specific pattern and ball would appear regardless of what box was selected. The 180 learning trials were presented in three blocks of 60, and within each block of 60, the red ball was presented with pattern A 18 times, pattern B 10 times, and pattern C 2 times. The same pattern was never presented with the same color ball more than two times in succession. After every 60 trials, subjects were asked to indicate how many times they had found the red ball in each of the three box locations.
The red ball was differentially associated with the three different background patterns, such that 90% of the time that pattern 1 was shown it was associated with the red ball, 50% of the trials using pattern 2 paired it with the red ball, and 10% of the trials using pattern 3 paired it with the red ball. Subjects were distracted from conscious attention to the relationship between the red ball and the different patterns by being asked to focus on the number of times that they found the red ball in the three different locations.
At the end of the full 180 learning trials, preferences for the background patterns were assessed. In this immediate testing phase, the three patterns used in the study as well as three new patterns with similar physical characteristics were presented in pairs. Each pattern was presented with every other pattern two times in order to balance presentation on the left and right sides of the screen, resulting in a total of 30 preference trials. Subjects were instructed to indicate, for each pair of patterns, the pattern that they preferred, and to not think too much about it but go with their first impression.
On the second day of testing, subjects completed an additional preference test. This task again included the same six patterns used in the immediate preference test, but stimuli were presented in groups of three. Each pattern was presented with all other patterns on three occasions, in order to balance for presentation in each of the three visual positions, resulting in a total of 60 trials. Again, subjects were instructed to simply indicate the pattern that they preferred, and not to think about it too much but to go with their first impression. The delayed preference task was designed to differ from the immediate preference task in order to decrease the potential for subjects to attempt to use their memories from the immediate preference task to influence their responses to the delayed memory task. At the end of the task, subjects were shown the six different patterns and how often they had indicated they had preferred each. They were then asked to state why they preferred their top three selections. This information was used to assess whether subjects were aware of the relationship between patterns and reward histories. Notably, we found that pattern 2 was disliked by all subjects, and thus removed that background pattern from further analyses.
Preference rankings of background designs 1 and 3 at both the immediate and 24 hour delayed preference assessments were used in statistical analyses.
Because we used rank order data, nonparametric statistical approaches were used. Healthy and SC subjects did not significantly differ in their ranking of either of the stimuli immediately following learning (Z = -.345, p > .20, and Z -.345, p > .20) for differences in ratings of Pattern 1 and Pattern 3 respectively), and both rated pattern 1 as significantly more preferred than pattern 3 (HC: Z = -2.13, p < .05; SC: Z = -2.18, p < .05). (Figure 2, left half).
Change within each group was assessed using Wilcoxin Signed Ranks Test. There were no significant changes in ranking by the healthy controls over time (Pattern 1: Z = .12, p > .20; Pattern 3: Z = .29, p > .20), but there was a significant change for ranking of Pattern 1 by the SC subjects (Pattern 1: Z= 2.44, p < .05; Pattern 3: Z = 1.24, p > .20) (Figure 2, right half), indicating decreased ranking for the highly reinforced stimuli at the 24-hour delay.
There were no statistically significant relationships found between symptom measures and learning or retention performance. Similarly, there were no statistically significant relationships found between measures of duration of illness or medication dosage (in chlorpromazine equivalents) and memory performance.
Despite intact preference conditioning during the learning paradigm, individuals with schizophrenia did not maintain this preference following a 24 hour delay. These data suggest that biological systems underlying initial preference conditioning are intact, but that the systems supporting long-term memory for preference conditioning are compromised in individuals with schizophrenia. As noted above, past studies of responses to emotional stimuli in individuals with schizophrenia are consistent with the current result, specifically indicating normal emotional and learning responses to a variety of emotional stimuli such as pictures, movies, or foods (4-8). In a series of studies, Gold, Heerey and colleagues have also documented intact reward sensitivity, and normative subjective valuation of potential gains in individuals with schizophrenia (5, 11).
It has been suggested that, while individuals with schizophrenia experience normal consummatory pleasure, they have difficulty generating and/or effectively utilizing representations of the affective value of these positive experiences, which may account for deficits in anticipatory pleasure. The current data would suggest that one factor contributing to difficulty in the generation of such representations is a deficit in memory consolidation. Notably, there is now a significant literature on the impact of emotional salience on memory, and more particularly on biological processes which support late long-term potentiation (12). Current models note the importance of sleep periods for enhancement of memory for emotional stimuli (14-16). Notably, studies of episodic memory for emotional stimuli with schizophrenia samples have consistently demonstrated significant impairment following delays of 24 hours or longer (4, 26-29), while episodic memory for emotional stimuli following shorter delays (up to four hours) have shown inconsistent effects (cf., 7,26, 29 28-33). Holt and colleagues (34) recently documented a similar effect of long-term delay on extinction learning in individuals with schizophrenia in this journal. Overall, these data suggest that individuals with schizophrenia are significantly impaired in long-term retention of emotional information.
Meta-analyses of cognitive deficits in schizophrenia have repeatedly noted quite significant impairments in episodic memory (35-36), with the degree of impairment influenced by a variety of methodological factors, including type of stimuli (37), use of encoding strategies (38), task demands (recall versus recognition) (35-36), and delay period (4, 29). Although there have been relatively few studies assessing memory performance over periods of 24 hours, a recently published study by Skelley and colleagues (40) found significant impairment in initial learning and short delay memory for neutral words and images, but little additional deterioration of memory over a 24 hour delay period. However, Skelley and colleagues' study differs from the current work in a variety of ways, including use of intentional rather than incidental encoding, assessment of recall rather than recognition memory, and use of neutral in contrast to emotional images. Given that individuals with schizophrenia are clearly impaired in utilizing effective strategies at both encoding and retrieval, it may be that deficits on these aspects of the memory task significantly outweighed effects due solely to difficulties in memory consolidation. Further, there is a large body of literature indicating that biological processes supporting late long term potentiation for emotional information differs from that for non-emotional information; thus the current results may reflect abnormality with a more emotion-specific aspect of memory functioning.
In the current study, preference at the immediate and 24 hour delay testing was assessed using different methods. The difference in methods was introduced to decrease the possibility that subjects would base their preference decisions at the delayed testing on their memory of how they had responded at the immediate testing. Unfortunately, the use of different testing methods (i.e., presenting two versus three stimuli simultaneously during preference assessment), confounds method differences with delay period effects. Although the current findings are consistent with data from past studies of emotionally-modulated long-term memory in schizophrenia, it will be important to rule out potential effects of these methodological differences in future studies.
Difficulties in developing accurate cognitive representations of positive experiences have been posited as potential contributors to deficits in anticipatory pleasure. The current data suggests that impairment in creating such cognitive representations may be, in part, due to abnormalities in consolidation of emotional memories. By using a preference conditioning paradigm, the contribution of deficits in executive functioning (such as use of encoding and retrieval strategies) were minimized, allowing us to focus more specifically on basic learning and retention of stimulus-reward relationships. These data document compromise in processes underlying the maintenance of memories for such relationships over time, which could reasonably contribute to difficulties in using reward-related memories to guide behavior.
This study was supported by a NIMH grant (MH 67223) to the author. I thank Lindsay Termini, Sunil Shrestra, and Dr. Cherise Rosen for direct assistance with this study, and the Center for Cognitive Medicine for providing the infrastructure to support this work.
Financial Disclosures The author reports no biomedical financial interests or potential conflicts of interest.