Sensitivity to intrinsic rewards is domain general and related to mental health

Humans spend much of their time engaging in activities that are pleasurable in their own right. These activities are undertaken even when they do not lead to external outcomes: they are intrinsically rewarding. Watching the sunset, reading, solving crossword puzzles, playing, exploring nature and observing works of art are a few such examples. On the surface these different activities do not have common features or goals. This contrasts with activities that lead to primary rewards (for example, eating, fornicating), which all have clear and direct survival benefits, and secondary rewards (for example, money), which in turn are associated with primary rewards. It is possible, however, that different intrinsic rewards do share core characteristics, mechanisms and goals not readily transparent. If so, such common features should elicit similar types of behavioural responses, and individual differences in these responses should be partially domain general. We test this hypothesis in this work, namely, that despite diverse intrinsic rewards seeming vastly different from each other, sensitivity to them is partially domain general and may be shared with secondary rewards. Engagement with specific intrinsically rewarding stimuli has been associated with happiness, mental health and professional achievement. Here we pose that these past findings can in fact be explained by a core association between mental health and domain-general sensitivity to intrinsic rewards. That is, if an individual finds a specific stimulus rewarding (for example, observing landscapes), they may be more likely to find other stimuli (reading, playing and so on) rewarding due to a domain-general sensitivity to (intrinsic) rewards, which may be associated with mental health. Individuals with high sensitivity to intrinsic rewards will be inclined to engage with a variety of seemingly diverse intrinsically rewarding activities, which in turn will increase the likelihood that they will eventually find rewarding activities that they also excel at. Low sensitivity to intrinsic rewards, on the other hand, will produce a general disinterest in a large variety of activities, which will lead to low mood and lack of motivation. Thus, a domain-general sensitivity to intrinsic rewards will contribute to flourishing and its absence to suffering. We focus here on a core aspect of mental health that we will refer to as affective health. We define affective health as a range of characteristics that are related to positive mood, high motivation, feelings of pleasure, interest and happiness. To investigate whether sensitivity to intrinsic rewards is domain general and related to affective aspects of mental health, we selected three putative intrinsically rewarding stimuli from the visuals, cognitive and social domains and also compared responses to monetary rewards. We measured liking, wanting and reinforcement of all stimuli to assess individuals’ reward sensitivity. We also asked participants to fill a range of questionnaires related to affective aspects of mental health and implemented a dimensionality approach, allowing an investigation that cuts through classic clinical boundaries. Together, the data allowed us to examine if within-individual responses to intrinsic rewards are domain general and linked to mental health.

The study builds on literature proposing three characteristic components of reward—liking, wanting and reinforcement—and prior work indicating humans find consuming information and social confirmation rewarding, with neural responses in reward circuitry. Engagement with intrinsically rewarding activities has been linked to happiness, mental health and achievement. Prior studies have shown that confirmation bias leads individuals to seek confirmatory information, and such confirmation activates reward systems. Research on primary and secondary (monetary) rewards has yielded mixed associations with mental health features, suggesting a need to examine intrinsic rewards. Neuroimaging literature indicates common neural representations for different material rewards and overlaps between primary and intrinsic rewards in ventromedial prefrontal cortex and ventral striatum, motivating the investigation of domain-general intrinsic reward sensitivity and its mental health relevance.

Design: Three online cross-sectional experiments were conducted with independent samples (experiment 1: N=132; experiment 2: N=171; experiment 3: N=180). Experiments 1 and 2 had a fixed measure order (liking, wanting, reinforcement); experiment 3 counterbalanced measure order across participants and reward types and used a choice-based wanting task to address potential scale-related confounds. Participants: Recruited via Prolific. Inclusion required passing comprehension/attention checks (scale-movement prompts, association checks between letters and stimuli, instructed-response items in questionnaires, and repetition detection in experiment 3). Demographics: Experiment 1 analyzed 132 participants (34% female, mean age approx. 31); Experiment 2 analyzed 171 participants (51% female, mean age approx. 34–35); Experiment 3 analyzed 180 participants (41% female, mean age approx. 38–39). Compensation: £7.50/hour plus a £0.50 bonus. Ethical approval from UCL Research Ethics Committee (project no. 3990/003); informed consent obtained. Stimuli: Five categories—three intrinsic rewards (visual, cognitive, social), one secondary reward (monetary), and neutral stimuli. Visual reward: landscapes (OASIS image set); control: walls. Cognitive reward: informative trivia sentences (Encyclopedia Britannica); control: scrambled letters versions of the same sentences. Social reward: social similarity (another ostensible participant matched 4/5 of participant’s earlier stated preferences, shown on-screen before item pairs like coffee/tea); control: disagreement (1/5). Monetary reward: bonus money signaled by a £1 coin image; control: crossed-out coin (no bonus). Neutral: vertical vs. horizontal lines. In experiments 1–2, letters cued stimuli (L/W; I/O; names Addison/Daryl indicated social source; R/Z; V/H). In experiment 3, words replaced letters and coin images varied trial-to-trial. Procedure: 1) Preference collection: Participants first indicated their preferred item from 44 pairs; these preferences informed social similarity/disagreement stimuli. 2) Liking: For each reward type, participants completed two 5-trial blocks of the rewarding stimulus and two blocks of its alternative (order alternated). After each block, participants rated liking on a 0–100 scale. Stimulus durations: cognitive 5,000 ms; social 2,000 ms; visual 2,000 ms; neutral 2,000 ms; monetary 1,500 ms. Liking score per domain was the difference between reward and control block ratings. 3) Wanting: Experiments 1–2 used a continuous scale from definitely 10 trials of control to definitely 10 trials of reward, then implemented the chosen option (10 trials). Experiment 3 used 15 choice trials per domain between reward word and control word; the proportion of reward choices quantified wanting. 4) Reinforcement learning: For each domain, participants completed a 24-trial probabilistic instrumental learning task choosing between two abstract cues with 0.75/0.25 contingencies to reward vs. control stimuli. The percentage of choosing the cue more often leading to the reward quantified reinforcing strength. Order of reward types was counterbalanced. Questionnaires and mental health measures: At the end, participants completed PHQ-9 (depression), AES (apathy), SHAPS (anhedonia), DOPS (experienced pleasure subdomains), and life satisfaction and current happiness ratings. These formed inputs for a dimensional mental health factor. Statistical analysis: Non-parametric Wilcoxon signed-rank tests assessed whether liking (reward minus control) exceeded zero, wanting exceeded indifference (50) or was chosen above chance, and reinforcement exceeded chance (50%). Factor analyses (R psych::fa, maximum likelihood, bi-factor rotation) were performed on behavioral measures (liking, wanting, reinforcement across visual, cognitive, social, monetary, neutral) after verifying factorability (overall MSA > 0.5). The first factor’s loadings (strong on rewards, not neutral) defined a general reward sensitivity score; explained variance gauged domain-generality. A separate factor analysis on questionnaire and affective measures yielded a mental health score (positive loadings for happiness/life satisfaction, negative for depression/apathy/anhedonia). Linear models (R lm) regressed standardized mental health score on standardized reward sensitivity, with and without controls (age, gender, education/qualifications, income, marital status, IQ). Influential points were screened via DFFITS. Model comparisons used BIC; family-wise comparisons summed BIC across matched model families (intrinsic-only vs. intrinsic+monetary pairs).

- Intrinsic rewards are rewarding: Across all three experiments, visual (landscapes), cognitive (facts), and social agreement stimuli were liked more than their alternatives, wanted above indifference or chosen more often than alternatives, and were reinforcing (cues leading to them were chosen above chance). Monetary rewards showed the same qualitative pattern; neutral line stimuli did not (no significant liking differences, wanting not above indifference, not reinforcing). Statistical details summarized in Supplementary Tables 1–3; key omnibus results report P<0.001 by Wilcoxon signed-rank in many cases. - Domain-general reward sensitivity: Exploratory factor analyses (bi-factor rotation) across all behavioral responses and stimulus categories revealed a robust first factor loading on all rewards but not on neutral stimuli, explaining roughly 40% of variance (experiment 1: 40%; experiment 2: 45%; experiment 3: 34%). Loadings tended to be higher for intrinsic rewards than monetary reward in experiments 2–3, suggesting the general factor is dominated by intrinsic rewards, with other factors capturing additional variance (including monetary-specific sensitivity). - Mental health factor: Factor analysis on affective measures produced a first factor explaining most variance (experiment 1: 70%; experiment 2: 65%; experiment 3: 77%), with positive loadings for happiness and life satisfaction and negative loadings for PHQ-9, AES, SHAPS, and DOPS subscales—interpreted as an affective mental health score. - Reward sensitivity predicts mental health: Linear regressions of mental health score on reward sensitivity, controlling for demographics and IQ, were significant: exp. 1 β=0.24±0.09, t(112)=2.8, P=0.005; exp. 2 β=0.18±0.07, t(149)=2.4, P=0.016; exp. 3 β=0.26±0.08, t(145)=3.4, P=0.001. Results were similar without controls. - Intrinsic vs. monetary sensitivity: When separating factors, intrinsic reward sensitivity significantly predicted mental health (e.g., Table 1: exp. 1 P=0.020 controlled; exp. 2 P=0.0010 controlled; exp. 3 P=0.039 controlled), whereas monetary reward sensitivity did not (e.g., exp. 2 P=0.711 controlled). BIC favored models with intrinsic reward sensitivity over those with monetary reward sensitivity (Tables 1–2). Pairwise balanced model family comparisons also favored intrinsic-only models (lower summed BIC across experiments).

The findings support the hypothesis that sensitivity to intrinsically rewarding stimuli is, in part, domain general. Visual, cognitive, and social rewards elicited the canonical reward signatures—liking, wanting, and reinforcement—akin to secondary (monetary) rewards, while neutral stimuli did not. A general factor captured a substantial portion of variability across individuals’ responses to all rewards but not to neutral stimuli, indicating shared mechanisms or features among diverse intrinsic rewards. Importantly, this domain-general reward sensitivity was associated with better affective mental health: higher sensitivity related to higher happiness and life satisfaction and lower depression, apathy, and anhedonia. The association was driven primarily by intrinsic reward sensitivity rather than monetary reward sensitivity, implying that intrinsic reward processing may be more relevant for well-being. This may help explain flourishing versus suffering: individuals with greater intrinsic reward sensitivity may explore and persist in a broad array of pleasurable activities, increasing the likelihood of identifying pursuits they excel at and maintaining positive affect; those with lower sensitivity may disengage, contributing to lower mood and motivation. Potential common features underlying intrinsic rewards include self-efficacy and curiosity, both linked to mental health. Neural commonalities with material and primary rewards have been observed in ventromedial prefrontal cortex and ventral striatum, suggesting partially shared neural substrates alongside systems that distinguish reward types and values, consistent with the non-shared variance observed across reward categories.

This work demonstrates that diverse intrinsic rewards elicit consistent reward responses and share a domain-general sensitivity factor that explains a substantial portion of behavioral variance. This intrinsic reward sensitivity is closely linked to affective aspects of mental health, more so than monetary reward sensitivity. The results underscore the importance of studying intrinsic rewards to understand human thriving and suffering. Future research should examine causality (e.g., interventions to modulate intrinsic reward sensitivity), test additional intrinsic (e.g., music, sports) and primary (e.g., food) rewards for generalizability, and investigate neural mechanisms separating intrinsic from material reward processing and disentangling shared versus domain-specific components.

- Order and measurement confounds: In experiments 1 and 2, liking, wanting, and reinforcement were measured in a fixed order, and both liking and wanting used rating scales, potentially inflating associations. Experiment 3 addressed these by counterbalancing measures and using choice-based wanting, replicating results. - Cross-sectional design: The studies are correlational; causal direction between intrinsic reward sensitivity and mental health cannot be inferred. A third variable may underlie both. - Stimulus scope and generalizability: Only a subset of intrinsic rewards (visual, cognitive, social) and one secondary reward (money) were tested. Additional intrinsic (e.g., music, sports) and primary rewards (e.g., food) should be examined to assess generalizability. - Online sample and task constraints: Tasks were conducted online with simplified proxies (e.g., coin images signaling small bonuses); although attention checks were implemented, ecological validity and sensitivity to context may limit generalization.