Expectancy-Value Theory in Persistence of Learning Effects in Schizophrenia: Role of Task Value and Perceived Competency

Abstract

Expectancy-value theory, a widely accepted model of motivation, posits that expectations of success on a learning task and the individual value placed on the task are central determinants of motivation to learn. This is supported by research in healthy controls suggesting that beliefs of self-and-content mastery can be so influential they can predict the degree of improvement on challenging cognitive tasks even more so than general cognitive ability. We examined components of expectancy-value theory (perceived competency and task value), along with baseline arithmetic performance and neuropsychological performance, as possible predictors of learning outcome in a sample of 70 outpatients with schizophrenia randomized to 1 of 2 different arithmetic learning conditions and followed up after 3 months. Results indicated that as with nonpsychiatric samples, perceived self-competency for the learning task was significantly related to perceptions of task value attributed to the learning task. Baseline expectations of success predicted persistence of learning on the task at 3-month follow-up, even after accounting for variance attributable to different arithmetic instruction, baseline arithmetic ability, attention, and self-reports of task interest and task value. We also found that expectation of success is a malleable construct, with posttraining improvements persisting at follow-up. These findings support the notion that expectancy-value theory is operative in schizophrenia. Thus, similar to the nonpsychiatric population, treatment benefits may be enhanced and better maintained if remediation programs also focus on perceptions of self-competency for the training tasks. Treatment issues related to instilling self-efficacy in cognitive recovery programs are discussed.

Introduction

While there is much evidence for the direct effects of neurocognition on functional outcomes in schizophrenia,^1–5 new research suggests that this relationship may be partially mediated by negative symptoms,⁶ and in particular, amotivation.⁷ Consequently, both neurocognition and motivation have received increased scrutiny as potential treatment targets to enhance functional outcomes.⁷ The nature of the overlap between neurocognition and motivation has begun to be explored more intensely, and specific hypotheses have been made about psychological factors that may be causally related to these 2 variables. One such hypothesis concerns the role of dysfunctional performance beliefs. There is evidence that dysfunctional or defeatist performance beliefs are related to the presence of negative symptoms in schizophrenia and that defeatist beliefs may partially mediate the relationship between neurocognition and negative symptoms such as amotivation as well as the relationship between neurocognition, learning, and functioning.^8–10

From a cognitive behavioral perspective, it has been suggested that task disengagement and withdrawal from effortful activities may serve as a defense against anticipated failure and critical evaluation¹¹ and that this leads to attenuated verbal behavior (alogia), diminished drive (anergia), and hopelessness for change (apathy). This is parallel to social learning theories of motivation for achievement in normals. Perception of self-competency has been a central constituent of motivation for learning and a strong predictor of high levels of motivation in educational and treatment settings.^12–14 Bandura^15,16 proposed a social cognitive model that focused on the role of perceptions of self-efficacy to engage in effortful learning tasks (ie, confidence in the ability to solve a problem). He concluded that efficacy expectations are the major determinant of (a) a person’s choice to participate in a learning activity, (b) a person’s willingness to expend effort if the learning becomes challenging, and (c) possibly the degree of learning retention from the activity. Eccles and her colleagues¹⁷ have elaborated on this social learning perspective and developed and tested a model called “expectancy-value” where they link learning, achievement performance, and task engagement directly to perceptions of self-efficacy and task value. In this model, the choice to engage and continue in a learning activity is influenced by task-specific beliefs, such as perceptions of self-competency and task difficulty as well as the relative value of the task to the individual. This theory emphasizes the importance of assessing and instilling self-competency beliefs and assessing the interaction of self-competency and the motivation or volition to diligently pursue and complete a task.¹⁸

In terms of psychopathology, this expectancy-value theory of motivation may inform how people with schizophrenia can be more intrinsically motivated to learn if they value the tasks as meaningful and useful to reaching their goals (utility value). When a learning exercise is seen as having utility, in essence if it is perceived as helpful in achieving personally relevant goals, the person may be more intrinsically motivated to learn and possibly benefit more from the learning task.¹⁹ Taken together, the emerging empirical literature on dysfunctional performance beliefs and expectancy-value theory both center on how improved self-efficacy may lead to better task engagement, higher motivation for treatment, and greater duration of learning effects. In an earlier report, we detailed how changing learning task characteristics through the use of personalization, contextualization, and choice led to increased intrinsic motivation (IM) for the learning task and how this motivational manipulation resulted in greater task learning as assessed at the end of training.²⁰ We also reported that baseline perceptions of self-competency were significant predictors of task learning at the end of training. In this follow-up report, we sought to examine the role self-competency in the “persistence” of learning effects 3 months after training to examine how important self-competency is to maintaining gains. In the nonpsychiatric population, self-efficacy for learning is a powerful predictor of retaining the lessons taught despite the challenging nature of the task.²¹ When confronted with a difficult learning task that does not present any obvious benefit to the person (eg, complex arithmetic), beliefs about whether the material can be learned are more important to learning outcome than intellectual abilities or even prior knowledge of the material. In the same manner, when patients with schizophrenia are faced with demanding cognitive tasks in a cognitive training program where remediation goals and rewards are not readily apparent or internally represented,²² it is important to determine if self-competency is a factor in treatment outcome beyond that explained by elementary cognitive functioning.

We hypothesized that (1) perceptions of self-competency would be positively correlated with task value attributed by the individual at all 3 assessment points (expectancy-value theory), (2) baseline perceptions of self-competency would be related to durability of learning effects at 3 month follow-up, and (3) self-competency would be malleable over time, with posttraining changes in self-competence persisting at follow-up. We also sought to examine the specific contribution of these variables, along with premorbid function and baseline neurocognitive performance, to persistence of learning effects.

Methods

Participants

Participants in the study were 70 outpatients, ages 21–55, diagnosed with schizophrenia or schizoaffective disorder, enrolled in a computer-based arithmetic learning program. A total of 92 subjects were recruited and consented, 88 were screened and enrolled, 82 completed training and posttesting, and 70 returned for 3-month follow-up. Participants were recruited at the New York State Psychiatric Institute in New York City and the VA Connecticut Healthcare System in West Haven, CT. Prior to participation in the study, participants had to be psychiatrically stable for at least 30 days, with no changes in psychotropic regimen. Diagnosis was confirmed by the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition Axis I Disorders.²³ Any participants with significant auditory/visual impairment, lack of fluency in English or medical illness known to impair brain function, other than schizophrenia, were excluded. Participants who met criteria for current substance abuse/dependence were also excluded, including those with active substance abuse 30 days prior to intake. A description of the study was given to all participants who provided written informed consent in accordance with each respective hospital Institutional Review Board.

Measures

Estimated Intelligence

Estimated premorbid intelligence for each participant was established at study intake with the vocabulary subtest scaled score from the Wechsler Adult Intelligence Scale-Revised²⁴ and the reading subtest standard score from the Wide Range Achievement Test-Third Edition (WRAT-3²⁵).

Attention

The Continuous Performance Test-Identical Pairs Version (CPT-IP²⁶) was used to measure the capacity to sustain focus on critical information in an attention-demanding visual environment. The CPT-IP serially presents 2 types of stimuli (numbers and nonsense shapes), which appear on the screen at a rate of 1 stimulus per second. The participant is instructed to respond as quickly as possible by pushing a control button whenever 2 identical stimuli are presented simultaneously. Major performance indices include number of correct responses to target, false positives, random errors, and d-prime. Attention was assessed at baseline, end of training, and 3-month follow-up.

Direct Learning

Arithmetic skill was assessed at baseline, end of training, and 3-month follow-up using a 48-item paper-and-pencil arithmetic test used by Columbia University Teacher’s College to assess general arithmetic ability in young adults. The test comes in 4 alternative forms to address possible “training to test” effects and evaluates the participant’s knowledge and calculation skills in addition, subtraction, division, multiplication, use of parentheses, and order of operations (eg, “(4 + 5) × 5 = __”). Item responses were open-ended, with the total number correct at each assessment ranging from 0 to 60. Alternate versions of the arithmetic test were used at each assessment point.

Task Interest

IM for the learning task was measured using the Enjoyment/Interest subscale score on the Intrinsic Motivation Inventory for Schizophrenia Research (IMI-SR^27,28), a Likert-type self-report scale designed to assess a participant’s subjective experience of an activity specifically in an experimental setting. IM is generally considered the impetus for an activity due to associations with positive feelings or enjoyment for that activity.²⁹ The subscale assesses the participant’s interest/enjoyment for a given learning activity (ie, “I enjoyed doing this activity very much”), with higher scores indicative of greater IM for the task. The subscale is highly associated with germane constructs of motivation for health-related behaviors, including perceived competency for attempting challenging tasks and autonomous treatment engagement. The subscale has 7 items and a previous report show that the subscale possesses good internal consistency (alpha = .95) and test retest reliability (intraclass correlation coefficient [ICC] = 0.74).²⁸ Task interest was assessed at baseline, end of training, and 3 months. For baseline assessment, the subscale was given at the halfway point during the first learning session, once the participants were familiar with the learning tasks, they would be doing for the remainder of the training. For posttreatment, the task interest subscale was given immediately following the last learning session. For 3 month follow-up, the subscale was given in the same manner the original scale developers assessed IM for a past learning task or an activity in normals.^30,31 Participants were briefly shown the respective learning exercise and instructed to complete the questionnaire in the context of how they recall feeling about attempting the mathematics lesson.

Task Value

Reported value attributed to the learning task was measured using the Task Value/Usefulness subscale score on the IMI-SR. The subscale measures the perceived significance of the task to the participant (ie, “I think that doing this activity is useful,” “I would be willing to do this again because it has some value to me”), with higher scores indicative of greater perceptions of task utility. The subscale has 7 items, and a previous report shows that the subscale possesses good internal consistency (alpha = .91) and test retest reliability (ICC = 0.70).²⁸ Task value was assessed at baseline, end of training, and 3 months, concurrently with the task interest assessment and in the same manner.

Self-competency

The Perceived Competency Scale (PCS³²) was used to measure the participant’s self-competency about completing and mastering the learning exercise. The PCS consists of 4 items on a 7-point Likert-type scale ranging from “not at all true” to “very true” (ie, “I feel confident in my ability to learn the computer program,” “I am able to achieve my goals in this program”), with higher scores indicative of greater feelings of self-competency for the task. The questionnaire has shown good validity and consistency in repeated studies examining its factor loadings related to internalized motivation and interest and possesses excellent internal consistency (alpha = .80–.94).³² Self-competency was assessed at baseline, end of training, and 3 months, concurrently with the task interest and task value assessments and in the same manner.

Symptoms

Psychiatric symptomatology was measured by the expanded Brief Psychiatric Rating Scale (BPRS), an updated version of the original BPRS.^33,34 The expanded BPRS is a 24-item, self-report measure which quantifies the level and presence of psychopathology on a 7-point Likert-type scale ranging from “not present” to “extremely severe.” We parsed the BPRS into the standard 4-factor solution of the 24-item BPRS³⁵ (negative, positive, agitation–mania, and depression–anxiety). ICC between raters at the same site and between sites ranged from 0.82 to 0.87. The BPRS was administered at baseline, posttraining, and 3-month follow-up.

Procedures

Participants were randomly assigned to 1 of 2 instructional methods of computer-based arithmetic learning.²⁰ Because the emphasis of the current report is on the persistence of learning effects (without regard to instructional methods that may impact the degree of learning gains), we collapsed the 2 groups into a single sample for purposes of the present analyses. We do, however, provide a brief description of the 2 groups here to allow interpretation of the results in the context of the original study. Participants were paid for participating in baseline, posttraining, and follow-up assessments but not for their participation in the arithmetic learning sessions themselves.

Arithmetic Learning Program.

Arithmetic provides a gage of direct domain-specific learning that allows the researcher to quantify the degree of material absorption from a specific lesson or intervention. Learning arithmetic (parentheses and order of operation) was also specifically chosen as a learning exercise because it is challenging for many people and it does not present itself as having any obvious benefit or intrinsic reward (similar to cognitive training tasks in remediation programs). The arithmetic learning game board consists of a numbered line from 1 to 60, and the first player to reach 60 is declared the winner. In each trial, working memory and executive ability are needed to combine 3 numbers generated by the computer into a valid arithmetic expression. Participants employ arithmetic knowledge and mental flexibility by using addition, subtraction, division, multiplication, and parenthetic operations to create a correct expression. The resulting value of the created expression is the number of spaces the participant can advance on the screen. If the participant cannot provide a valid arithmetic expression using the generated numbers, the computer provides instructional feedback in the form of a brief arithmetic lesson and offers the correct answer. However, the participant is not allowed to advance unless a correct answer is provided. Two equivalent versions of this learning program were created for the original study, one with and another without instructional variables selectively manipulated into the learning process to target task interest. Both arithmetic learning game versions had identical feedback and lesson content and involved the same fundamental procedure noted above. The program used was a single player option where the participant challenged the computer at varying levels of difficulty.

Following baseline testing on all measures, participants were randomly assigned to 1 of the 2 learning programs for 10 thirty-minute sessions, to be completed over a 4-week period. Importantly, participants could work on the lessons at their pace. A research assistant conducted 4 sessions a week where participants were allowed to come in anytime during those sessions to work on the lessons. Cognitive and symptom posttesting were completed within 2 days of the last lesson by a research assistant blind to the randomization while 3-month follow-up testing was completed on average 92 days after posttesting. As mentioned in the instrumentation section, the questionnaires (task interest, task value, and self-competency) were administered during the middle of the first session to obtain baseline and then immediately following the last session for postassessment. At follow-up, participants were briefly shown their respective math learning program to remind them of the task they had completed during the intervention in order to fill out the questionnaires on their perceptions of the task.

Data Analysis

Distributions of scores were inspected for normality and compared with relevant comparison groups for homogeneity of variance. The comparison of the efficacy of the different learning interventions at posttesting is reported elsewhere.²⁰ Here, we focus on self-competency for the arithmetic task and the durability of learning effects after targeted training for the entire sample. To assess hypotheses 1—the relationship between task value and perceived self-competency—we conducted a set of cross-sectional correlations between these variables at baseline, posttreatment, and follow-up. To assess hypothesis 2—the relationship of self-competency to learning persistence—we examined baseline self-competency as an independent variable for predicting arithmetic ability at 3-month follow-up, also taking into account other potential baseline predictor variables. In order to determine whether premorbid cognitive function was related to learning persistence and whether it should be included as a variable in the analysis examining predictors of learning persistence, we performed partial correlations between baseline Vocabulary and WRAT-3 Reading scores and 3-month arithmetic ability, controlling for baseline arithmetic ability. Next, we conducted a hierarchical multiple regression, with forced entry at each block based on an order of entry previously reported between each construct and arithmetic learning.²⁰ For the analysis of predictors of 3-month arithmetic ability, learning condition was forced in first, baseline arithmetic scores second, followed by baseline attention in the third block, then baseline task interest, baseline task value, and finally baseline perceived self-competency. To assess hypothesis 3—that self-competency is malleable with continued effects evident at 3-month follow-up—we compared baseline, posttraining, and 3-month performance scores for PCS. We also examined baseline, posttraining, and 3-month follow-up scores for all the other variables to better characterize the pattern of changes in learning, attention, symptoms, task interest, and task value. All statistical tests were 2-tailed, and alpha was set at .05.

Results

Attrition from baseline to 3-month follow-up was 20%, leaving 70 subjects who completed all procedures including 3-month follow-up (table 1). The sample took 15.17 (4.64) days on average to complete the 10 arithmetic sessions. As shown in the correlation matrix (table 2), there was a significant, positive cross-sectional correlation between perception of self-competency for learning arithmetic and task value at all 3 time points (baseline, posttesting, and 3 month). To explain the relationship of baseline variables to persistence in arithmetic ability 3 months following the end of training, we entered group assignment and baseline performances on arithmetic, CPT-IP d-prime, task interest, and value subscales, and PCS into a hierarchical regression with forced, blocked entry as noted above. The model, which partitions equally exclusive components of the overall variance for each variable, explained a significant portion of the variance in the total number correct on the arithmetic test at 3-month follow-up (R² = .51, F[4,66] = 5.37; P = .001). The results of the regression involving learning condition (step 1), baseline arithmetic ability (step 2), baseline attention (step 3), baseline task interest (step 4), baseline task value (step 5), and baseline perceived self-competency (step 6) are presented in table 3. In the first 5 steps, 50% of the variance is explained. With the addition of PCS in step 6, an additional 9% of the total variance was explained, which was a significant increase in R² from the previous step (R² change = .09; significance of R² change = .00). In this final step, only perceived self-competency (PCS) accounted for the unique variance in 3-month arithmetic ability, presenting evidence of the association between perceptions of self-competency and learning persistence, even when different arithmetic instruction methods and baseline arithmetic ability, attention, and task interest and value were controlled. Partial correlations between premorbid intelligence measures (Vocabulary and WRAT-3 Reading) and 3-month arithmetic score, controlling for baseline arithmetic score, were not significant (r’s = −.07 to .10, P’s > .27); therefore, these variables were not included in this analysis examining predictors of learning persistence.

Table 1.

Demographic and Clinical Characteristics of the Whole Learning Sample

	Arithmetic Learning Sample, (N = 70)
Age (y)	38.54 (5.67)
Education (y)	11.42 (4.16)
Gender, male (%)	62
Duration of illness (y)	12.11 (6.98)
Percentage on atypicals	85
Percentage on anticholinergics	15
Percentage diagnosed disorganized type	6
Percentage diagnosed with schizoaffective disorders	49
Premorbid IQ estimate
WRAT-3 Reading	91.19 (6.17)
WAIS Vocabulary	8.79 (2.05)
BPRS
Positive factor	27.34 (7.71)
Negative factor	15.57 (5.56)
Agitation–mania factor	13.08 (5.63)
Depression–anxiety factor	16.08 (7.21)
Total	71.08 (17.05)

Note: BPRS, Brief Psychiatric Rating Scale-Expanded Version total score and 4-factor solution; WRAT-3, Wide Range Achievement Test-Third Edition; WAIS, Wechsler Adult Intelligence Scale.

Table 2.

Cross-Sectional Correlations between Measures for the Learning Sample

	Task Value	Task Interest	Arithmetic Scores	BPRS Total	BPRS Neg	BPRS Pos	CPT-IP	WRAT3	Vocabulary
Task value
Baseline	—	17	0.19	0.04	0.15	0.12	0.11	0.11	0.03
Post	—	0.16	0.22*	0.12	0.19	0.10	0.11	—	—
Follow-up	—	0.13	0.20	0.02	0.16	0.18	0.13	—	—
Task interest
Baseline	0.17	—	0.12	0.11	−0.19	0.10	0.21*	0.05	0.12
Post	0.20	—	0.19	0.14	−0.20	0.08	0.15	—	—
Follow-up	0.13	—	0.14	0.14	−0.20	0.06	0.02	—	—
PCS
Baseline	0.46**	0.28*	0.20	0.08	0.16	0.16	0.22*	0.06	0.16
Post	0.41**	0.25*	0.27*	0.06	0.20	0.10	0.16	—	—
Follow-up	0.29*	0.18	0.21	0.12	0.17	0.04	0.14	—	—

Note: PCS, Perceived Competence Scale; BPRS Total, Brief Psychiatric Rating Scale-Expanded Version total score; BPRS Neg, Brief Psychiatric Rating Scale-Negative Symptoms Factor; BPRS Pos, Brief Psychiatric Rating Scale-Positive Symptoms Factor; CPT-IP, Continuous Performance Test-Identical Pairs Version, d′; WRAT3 Reading, Wide Range Achievement Test-Third Version, Reading subtest standard score; WAIS-R Vocabulary, Wechsler Adult Intelligence Scale-Revised, Vocabulary subtest scaled score.

*P < .05.

**P < .01.

Table 3.

Results of the Hierarchical Block Regression Predicting Postarithmetic Ability at 3 Months

	β	t Value	P Value
Step 1
Learning condition	.28	2.24	.05
Step 2
Learning condition	.21	1.43	.07
Baseline arithmetic ability	.36	2.33	.03
Step 3
Learning condition	.19	1.05	.11
Baseline arithmetic ability	.31	2.15	.04
CPT-IP d-prime	−.12	−0.52	.48
Step 4
Learning condition	.18	1.00	.10
Baseline arithmetic ability	.29	1.98	.05
CPT-IP d-prime	−.14	−0.51	.45
Task interest	.19	1.04	.10
Step 5
Learning condition	.13	0.67	.29
Baseline arithmetic ability	.19	1.02	.12
CPT-IP d-prime	−.10	−0.32	.53
Task interest	.11	0.55	.21
Task value	.28	1.81	.06
Step 6
Learning condition	.12	0.55	.29
Baseline arithmetic ability	.17	0.42	.12
CPT-IP d-prime	−.07	−0.27	.60
Task interest	.18	0.68	.17
Task value	.24	1.52	.08
Perceived competency	.38	2.39	.03

Note: Learning conditions (motivational instruction or control learning instruction); CPT-IP, Continuous Performance Test-Identical Pairs; Task Interest and Task Value subscales of the Intrinsic Motivation Inventory for Schizophrenia Research.

Step 1: R² = .41, F = 16.57, P < .05.

Step 2: R² = .44, F = 4.48, P < .05; R² change = .03, significance of R² change = .09.

Step 3: R² = .44, F = 4.12, P < .05; R² change = .00, significance of R² change = .75.

Step 4: R² = .46, F = 3.68, P < .05; R² change = .02, significance of R² change = .11.

Step 5: R² = .50, F = 3.25, P < .05; R² change = .04, significance of R² change = .07.

Step 6: R² = .59, F = 5.31, P < .05; R² change = .09, significance of R² change = .00.

To examine whether self-competency was malleable and whether any changes in this variable persisted at 3-month follow-up, t-tests were conducted comparing baseline with posttraining, baseline to 3-month follow-up, and posttraining to 3-month follow-up PCS scores. As can be seen in table 4, there was a significant increase in PCS scores from baseline to posttraining and from baseline to 3-month follow-up (t’s = 2.85–3.14, P = .02–.03). PCS scores did not significantly decrease from posttraining to 3-month follow-up (t = 0.82, P = .23). We also compared baseline with posttraining 3-month follow-up scores for other measures to examine PCS changes in the context of other variables. As can be seen in table 4, significant improvements in arithmetic ability from baseline to posttraining (t = 3.45, P = .01) did not significantly decrease at 3 months (t = 1.12, P = .10). Only self-competency followed this same pattern of improvement and persistence in learning following treatment. Task interest showed improvement at posttreatment (t = 2.34, P = .04), but the effects were lost at 3 months (t = 2.17, P = .04). Changes in task value from baseline to posttraining or baseline to 3 month were nearly significant (P’s = .06–.07), with the slight gains at post maintained at 3 months (t = 1.04, P = .19). There were no significant changes in symptoms or any other variables from baseline to post to 3-month follow-up (P’s >.19).

Table 4.

Scores at Baseline, Posttesting, and 3-Month Follow-up for the Whole Learning Sample

	Arithmetic Learning Sample, (N = 70)
Arithmetic ability (total correct 0–60)
Pretesting	31.54 (5.67)
Posttesting	49.35 (7.85)a
3 months	45.78 (5.12)a,b
Perceived self-competency (4–28)
Pretesting	9.35 (2.13)
Posttesting	17.98 (3.73)a
3 months	16.90 (3.45)a,b
Task interest (7–49)
Pretesting	21.34 (7.12)
Posttesting	34.54 (4.29)a
3 months	22.43 (6.43)
Task value (7–49)
Pretesting	18.86 (4.11)
Posttesting	27.89 (6.14)
3 months	24.68 (4.81)b
CPT-IP (d-prime)
Pretesting	1.27 (0.50)
Posttesting	0.99 (0.31)
3 months	1.32 (0.62)b
BPRS (total)
Pretesting	71.08 (17.05)
Posttesting	62.73 (12.54)
3 months	70.42 (10.26)b

Note: Abbreviations are explained in the first footnote to table 2.

^aChange in mean differences from baseline is significant at P < .05.

^bNo significant mean differences from posttesting to 3 month-follow-up at P > .05.

Discussion

We found that self-competency for a learning task was significantly related to the value the individual attributed to the task. This was true at baseline prior to any intervention and at posttraining as well as 3-month follow-up. In other words, if the task was perceived as useful or worthwhile to the person, there were greater expectations of success for the task. Greater expectations of success on a task, in turn, were related to greater persistence of learning effects at follow-up, even after accounting for baseline task performance, neurocognition, and self-reports of task interest and task value. These findings are in accordance with expectancy-value theory, which posits that expectations of success and the value placed on a learning task are central determinants of motivation to learn and indicate that this theory can be expanded to individuals with schizophrenia. These findings also advance the emerging evidence that performance beliefs play a vital role in the relationship between neurocognition, motivation, and learning in schizophrenia. This is especially relevant when the learning task does not present any obvious benefit to the participant. The arithmetic task in this study was specifically chosen because it is a difficult learning task where goals and rewards are not readily apparent—similar to training trials in cognitive remediation tasks. Participants were paid for the assessments but not for the training sessions, thus minimizing any influence of external reward for the exercises. Descriptively, participants reported little interest in doing the task at baseline, which is expected given the mundane content of the exercises (use of parentheses, order of operations). However, even in the context of low interest and no external reward, we found that baseline expectation of success—the perception that they could do the tasks—was the most important factor in explaining how much was retained 3 months after the end of training. This becomes particularly important when one considers that these findings hold true when prior arithmetic skill and elementary cognitive function were accounted for. Our results help explain the relationship between learning and motivation by highlighting the importance of assessing and instilling self-efficacy for challenging cognitive tasks in order to influence how well patients learn and how much they retain.³⁶ Similar to the nonpsychiatric population, people with schizophrenia must believe their actions can lead to positive outcomes or else they may have little incentive to take on challenging treatment tasks.^37,38

Our results also indicate that expectations of success (perceived self-competency) improve following training and that these improvements are sustained at 3-month follow-up. This pattern of changes mirrors that observed for learning arithmetic, highlighting the important relationship between perceived self-competency and learning. Our findings not only show the interrelatedness of learning effects and self-competency but also suggest a promising potential target for intervention in cognitive remediation programs—by structuring treatments in a way that increases self-competency, learning effects may be enhanced. In our earlier work, we reported on one potential method of increasing self-competency and subsequent learning effects. Compared with a standard arithmetic learning task without learning cue manipulations, increasing IM for the learning task by manipulating social contexts—linking the task to personal goals and providing experiences of enjoyment, control, and mastery—led to improvements in self-competency and learning at the end of training.²⁰ Helping patients understand the value of a training activity by linking it to their personal goals and proceeding with the training in a way that provides them with experiences of enjoyment, control, and mastery is both consistent with the philosophy of recovery-based approaches³⁹ and supported by the data presented here.

Although this study examined the impact and malleability of self-efficacy solely in the context of a cognitive learning exercise, the findings have potential implications for other learning-based therapeutic interventions. Psychosocial programs for people with schizophrenia are typically predicated on the notion that skills can be taught to help people reach their functional goals.⁴⁰ Thus, much of the therapy in psychosocial programs involves learning-based activities.^41–43 For example, psychosocial programs teach social skills, wellness, symptom management, and vocational skills with the expectation that patients will learn the material and apply it in everyday life.^44–46 The findings presented here emphasize the importance of patients valuing the skill and believing that they will be competent at learning the skill. We believe that promoting expectations of success in traditional psychosocial rehabilitation programs and other learning-based therapeutic programs may lead to higher perceived self-competency and, subsequently, greater learning effects and greater overall benefit from these programs.

There are several limitations to the current study. While respectable, our sample size remains too small to run theta estimates, which would have allowed us to investigate item response vectors in our self-report measures. The use of self-report measures in schizophrenia to tap into psychological constructs has its disadvantages compared with objective or clinician-rated measures^47–49; however, there is also evidence that self-reports are equally valid instruments to measure abstract constructs in schizophrenia.⁵⁰ Additionally, given the nature of the psychological constructs studied, interview-based instruments may not accurately gauge a person’s perspective of task value or enjoyment.

Future research is needed to investigate whether perceptions of task value can be enhanced through psychotherapy techniques as well as examine what components of task value are related to self-competency. In addition to self-competency, related psychological variables such as self-esteem have also been found to be related to learning, with interventions focusing on training in cognitive strategies associated with improvements in self-esteem. Unlike our current results regarding self-competency, these self-esteem improvements appear to only occur during participation in the training task itself and are attenuated at follow-up.^51,52 Nevertheless, these findings underscore the intermingled nature of self-psychological constructs that influence learning and the potential to profit from cognitive remediation training or other therapeutic programs. In order to maximize treatment engagement and benefits, we must continue exploring how to address these psychological factors, as they seem to be directly related to functional outcome.⁷

Funding

National Institute of Mental Health (grant number 1 R03 MH071733-01A2 to J.C. and A.M.).