We conducted a priori power analysis using PowerAnalysisIL designed for EMA [55] using data provided by Li et al [50] (Table S1 and Figure S2 in Multimedia Appendix 1). Analysis suggested that 80% power could be achieved with a sample size of 50, consistent with estimates from multilevel logistic [56] and linear models [57] using our design.

Young people (N=95), aged 16‐25 years with varying depression symptoms, were recruited from local schools and universities.

Based on feedback from piloting sessions with young people, we revised the activity categories and allocated time for app troubleshooting in our study briefing sessions. We also assessed completion time for questionnaires and participant feedback. We found the daily burden manageable, supporting our decision to retain the questionnaire structure.

Participants completed the Mood and Feelings Questionnaire (MFQ; ≥27 cut off for clinical depression) [58]. Higher scores indicate higher depression symptoms. The 33-item questionnaire has been shown to have excellent internal reliability in adolescents (Cronbach α=0.91-0.93) [59]. It is a widely used and validated questionnaire to examine depressive symptoms in young people.

Similar to an EMA study in schizophrenia [54] at each assessment (beep), participants rated their mood (negative affect and positive affect) and then selected their current activity from categories and rated their enjoyment. They then selected current company from a selection of categories and rated their enjoyment. Next, they selected their future activity from the same categories and rated their anticipatory pleasure, expectation (likelihood of activity happening), and motivation (interest and preference). They then selected future company from categories and rated their anticipatory pleasure. The motivation questions were posed indirectly so as not to intervene or influence, and this style of questioning limits the potential for the participants to change their behavior in response to the question [54] (Table 1).

The experimenter met with each participant to brief them on the app. Participants were then required to log on and fill in their age, gender, ethnicity, and the MFQ. EMA assessments began the next day. Participants were asked to respond to each assessment as soon as possible; otherwise, they would expire.

We collected data 7 times a day, between 8:30 AM and 10 PM, for 6 days, during the period July 2022 to October 2023. There was at least a 45-minute delay between each semirandom assessment, which took approximately 1 minute to complete and expired after 20 minutes. This sampling frequency and questionnaire design has been shown to encourage compliance and reduce burden in young people and those with mood disorders [56,57,60,61]. We contacted participants on days 2 and 5 to check that they were receiving notifications and to troubleshoot any problems.

At the end of the study, we collected app user experiences (Table S2 in Multimedia Appendix 1), and participants were debriefed and advised to contact their general practitioner or the mental health charity, the Samaritans, if they were concerned about their mood.

Fisher exact tests, ANOVAs, and chi-square statistics were used to examine group differences in the demographic data. We prepared the EMA data using esmpack in R (version 4.4.1; R Core Team, 2024) [62]. Activity and company enjoyment and engagement were calculated across participants and ranked from highest to lowest.

To check for potential confounds, we ran multilevel regression models using the lme4 package in R with mood (negative affect and positive affect), assessment time (1-7), and assessment day (1-6) as predictors of anticipation, motivation, and enjoyment. Predictors were subject-mean centered and represented at level 1 and symptoms were level 2 fixed effects. All multilevel models were represented by a subject-level random intercept. As assessment time and mood covaried, they were added to all multilevel analyses as covariates (Table S3 in Multimedia Appendix 1).

We next ran multilevel models with depression as a continuous variable (MFQ) as a predictor of anticipation, motivation, and enjoyment across all activity, then specifically leisure (relaxing, other leisure, and exercise) and functional activity (work or school, studying, chores, shopping, hygiene, eating or drinking, and traveling). We also ran multilevel models with depression as a continuous variable as a predictor of anticipatory pleasure and enjoyment across all company and then specifically social (partner, friends, and family) and nonsocial company (nobody). We then examined the effects of depression symptoms on enjoyment and engagement for each specific activity and specific company, using separate standard linear regression models.

The Temporal Experience of Pleasure cycle shows that anticipation and motivation are temporally predictive of enjoyment and engagement in activities [28,29]. This is the basis of our theoretical framework (Figure S1 in Multimedia Appendix 1) for examining the temporal dynamics using time-lagged analyses. Using anticipation and motivation as predictors at (t−1), we modeled activity enjoyment and engagement (0 = functional, 1 = leisure) as outcomes at (t) using multilevel linear and logistic models, respectively. Similarly, using anticipation at (t−1), we modeled company enjoyment and engagement (0 = nonsocial, 1 = social) as outcomes at (t). We reported data for the whole sample (WS) and participants by depression symptom severity; controls (C, score of ≤16 on MFQ), moderate (MD, score of >16 and <27 on MFQ), and high depression symptoms (HD, score of ≥27 on MFQ). Splitting the sample into groups, similar to Li et al [50], allows us to compare the 4 main predictors (anticipatory pleasure, interest, preference, and expectation) in the same model while controlling for each other, across each of the 4 groups.

As participants may anticipate doing something but then do something else, we also examined the data when the predictors (anticipation and motivation) and outcomes (enjoyment and engagement) were matched for the same events, for example, anticipatory pleasure for exercise predicting enjoyment of exercise. However, matched data can mean much fewer assessments and possible model overparameterization.

The study adhered to ethical standards (Revised Declaration of Helsinki 2008) and was approved by the University of Reading Psychology Department ethics committee (REC number 2021‐120-CM). All participants provided written informed consent after reviewing the information sheets. To preserve participant privacy, no identifiable features were included in the manuscript and supplementary materials. After completion, participants were compensated by SONA credits [63].

A total of 2426 questionnaires were collected from 95 participants. We excluded participants who had <30% threshold of assessments (N=15) [64,65], and the mean compliance rate was 69%, similar to previous EMA studies [66]. About 1.5% of assessments were incomplete and therefore removed from analysis.

The final sample consisted of 2316 questionnaires. Participants self-identified as White (63/80, 78.8%), Black (4/80, 5%), Asian (9/80, 11%), Mixed (2/80, 2.5%), and Other (2/80, 2.5%). The depression groups differed on symptoms as expected and also for age and assessment period but not for gender, ethnicity, compliance, or assessment delay (Table 2).

The leisure activities “exercising,” “relaxing,” and “other leisure” were rated most enjoyable (Table S4 in Multimedia Appendix 1). Young people engaged most frequently in “relaxing” and “other leisure” activities, followed by “eating or drinking” and then “work or school” and “studying” (Multimedia Appendix 2).

Being with “friends,” “family,” and “partner” was rated highest for enjoyment (Table S5 in Multimedia Appendix 1). Although young people spent more time alone than with others, they were mostly with “friends,” “family,” and “partner” when socializing (Multimedia Appendix 2).

To our knowledge, this is the first study to demonstrate in real-world settings that anticipatory pleasure reliably predicts future enjoyment across various physical activities and social contexts in young people. Anticipatory pleasure and motivation (preference) predicted leisure engagement, and anticipatory pleasure predicted social engagement in the WS and controls but not in individuals with depression symptoms. However, when measured for the same specific event, anticipatory pleasure does predict leisure engagement in those with higher depression symptoms, indicating a potential target for intervention.

While prior research has established that depression blunts reward-related processes (Ma X, Sahni A, McCabe C, unpublished data, 2024), most studies rely on retrospective reports or laboratory tasks that fail to capture the temporal dynamics of reward processing in daily life. By leveraging EMA, this study provides novel insights into how anticipation and motivation dynamically shape behavior in real-world contexts. Consistent with prior work [50,67], we found that individuals with higher depression symptoms reported less enjoyment across various activities. However, our study extends these findings by showing that depression is also linked to a broader dampening of anticipation and motivation across functional, leisure, and social experiences. This supports the idea that depression is associated not only with reduced momentary pleasure but also with a diminished ability to anticipate and seek out rewarding experiences.

Our study also provides a nuanced perspective on real-world activity deficits in depression. Individuals with higher depression symptoms reported lower enjoyment in activities such as studying, eating or drinking, and other leisure activities, as well as reduced engagement in work or school and hygiene. Importantly, depression was associated with doing more “nothing,” a behavioral pattern consistent with wearable-monitor studies showing reduced activity in depression [68,69]. However, our results go beyond previous findings by contextualizing these activity deficits within specific daily-life behaviors. The reduction in work or school and hygiene aligns with known risk factors for depression, such as academic underachievement [70] and self-neglect [71]. These findings suggest that ecological momentary interventions (EMIs) could be developed to support goal setting, progress tracking, and real-time encouragement for maintaining daily routines [47,72].

Crucially, we demonstrate for the first time that anticipatory pleasure significantly predicts future enjoyment across all physical activities and social contexts. Anticipatory pleasure and motivation (preference) predicted leisure engagement, and anticipatory pleasure predicted social engagement in the WS and controls but not in individuals with depression symptoms. However, anticipatory pleasure did predict leisure engagement in those with high depression symptoms, when assessed for the same specific event. This suggests that while individuals with depression may struggle to generalize anticipation across contexts, if they can be encouraged to anticipate specific leisure activities, they may be more likely to engage in them, which could strengthen the anticipation-engagement link over time.

Our findings support the development of interventions that could exploit the time-lagged relationships such as episodic future thinking (EFT) [73], a cognitive exercise that uses mental imagery to anticipate the feelings associated with future positive activities. Preliminary data suggest that EFT can increase anticipatory pleasure using both traditional scales [74-76] and EMA [77] in healthy controls and those with major depressive disorder. Therefore, we propose that an EFT-based EMI could be developed and tested to see whether it could increase engagement in leisure and social activities in young people with depression symptoms.

A striking finding was that higher expectation (likelihood of activity happening) was associated with lower enjoyment, especially in those with high depression symptoms. This aligns with our finding that the least enjoyable activities were work or school and studying, activities that are usually scheduled and therefore highly expected. Consistent with this, we found that enjoyment for activities such as studying decreased as depression increased and that higher expectation predicted more engagement with functional versus leisure activities across groups. These findings are supported by previous reports that show that activity scheduling reduces enjoyment, potentially due to diminished spontaneity [78].

Our findings could be explained by Self-Determination Theory [79], whereby activities that are externally imposed (such as school or studying) and extrinsically motivated (to pass examinations) tend to be less enjoyable, especially if autonomy is low, contrasted with activities that are intrinsically motivated such as leisure (inherently enjoyable). In depressed individuals, the lack of intrinsic reward and reduced agency may amplify the negative experience of expected, scheduled tasks. As the least enjoyable activities in this study are the ones most likely scheduled—this aligns with the Self-Determination Theory that autonomy is key to enjoyment and well-being. Moreover, as individuals with depression tend to hold negative core beliefs and cognitive distortions [80], negative biases could lead to negative perceptions of predictable activities, such as school or work, sustaining low enjoyment.

Taken together, highly expected activities are often externally imposed, reducing autonomy and intrinsic motivation. For youth with depression, these activities—especially when perceived as unpleasant—could confirm negative schemas about being powerless or trapped [80], which could lead to lower enjoyment and possibly deeper anhedonia. Future research should explore whether EMIs that promote spontaneous enjoyable activities, while addressing negative cognitive biases, could enhance activity engagement in depression.

While our study offers significant contributions, several limitations must be acknowledged. First, although EMA captures real-time experiences, self-report measures remain subject to bias. Future studies could integrate passive-sensing data (eg, activity trackers and voice recordings) to objectively assess engagement [81]. Second, our sample consisted of young adults, limiting generalizability to older populations. In future studies, time-lagged relationships should be investigated in a broader demographic. Third, while we demonstrated the predictive role of anticipation and motivation, causality remains uncertain. Experimental studies using reward-training interventions could further elucidate these mechanisms.

Our results support anticipation and intrinsic motivation as novel intervention targets for real-time, technology-driven approaches to managing depression. By leveraging digital tools to enhance anticipation and motivation, future interventions could empower individuals to engage with rewarding experiences, ultimately improving mental health outcomes.