Curiosity across the adult lifespan: Age-related differences in state and trait curiosity

The study investigates how curiosity relates to age across adulthood, distinguishing between trait curiosity (a stable tendency to seek knowledge) and state curiosity (momentary curiosity triggered by specific information). The authors hypothesize, based on prior mixed evidence and a pilot study, that trait curiosity decreases with age, whereas state curiosity—especially when elicited by materials tapping semantic knowledge (e.g., trivia)—increases with age. The research aims to clarify these divergent age associations and their implications for learning and engagement in older adulthood.

Curiosity is linked to beneficial outcomes in education, work performance, and well-being. Trait curiosity is generally positive and related to openness, novelty seeking, and love of learning, while state curiosity reflects motivation to resolve specific information gaps and is often behaviorally measured. Evidence on age and trait curiosity is mixed: some studies suggest increases or stability across adulthood, while others, supported by socioemotional selectivity theory (SST), indicate declines due to prioritized emotional goals and limited future time perspective. Epistemic curiosity has been observed to be lower in older adults. Conversely, some studies show older adults exhibit higher state curiosity for specific, semantically grounded information and environments. The selective engagement hypothesis suggests older adults allocate cognitive resources to tasks with salient benefits and lower perceived costs, potentially increasing curiosity when materials connect to prior knowledge. Prior knowledge is a recognized driver of curiosity and often accumulates with age, which may heighten state curiosity for domain-relevant information.

Design and preregistration: The study was preregistered on OSF. Data and materials are publicly available. Participants: 2,000 MTurk participants were recruited (March–August 2022) anticipating ~25% exclusions. Exclusion criteria included: looking up answers (n=69), bot-like responses to open-ended items (n=147), significant missing data or partial completion (n=12, n=123), reported task/internet problems (n=156), duplicates (n=82), prior completion of a similar task (n=53), and mismatched birthdate vs. age (>1 year; n=140). Final analyzed sample: N=1,218 U.S. adults, age 20–84 (M=44.4, SD=15.5). Demographics: 51% men, 48% women; majority White; education M≈16 years; household income M≈$60,656. Compensation: $7.25/hour. Measures: State curiosity was assessed using 63 randomly selected trivia items (from 244 total) normed to ~16% average guess accuracy, spanning general knowledge domains (e.g., history, science, food). Each item: 20s viewing with optional text-box guess; then participants rated curiosity (1–10) and confidence (1–10), viewed the answer (2s), with brief inter-trial screens; halfway break prompt; average completion time ~34.87 minutes. Items correctly guessed were excluded from analyses to focus on unknown information. Reliability of state curiosity (across 63 items): 0.988 (generalizability theory). Trait curiosity was assessed with the 10-item Epistemic Curiosity Scale (ECS; 1–4 scale), α=0.82. A preregistered three-item Intellectual Curiosity facet (NEO FFI) showed low reliability (α=0.15) and was omitted from analyses. Additional surveys (boredom proneness, scam susceptibility, subjective age) were collected but not central to hypotheses. Procedure: Participants completed demographics; half completed surveys first, half the trivia first; birthdate collected at the end to verify age. Analysis plan: Known items were filtered out; number of correct guesses analyzed. Primary analyses included correlations among variables and regression models. Trait curiosity: multiple linear regression predicting ECS from centered age, gender (dummy-coded, male as reference), race (dummy-coded, White as reference), centered education, and centered income. State curiosity: single-level multiple regression predicting mean curiosity from the same covariates; then a mixed-effects linear model with trial-level curiosity as outcome, predictors included age, gender, race, education, income, and average confidence; random intercepts for participants and items; random slopes of items for age (random item slope regression) to generalize across items and control Type I error inflation. Exploratory quadratic age terms were tested post hoc based on visual inspection.

• State and trait curiosity are positively correlated: r = .23, t(1216) = 8.37, p < .01. • Age vs. trait curiosity: Negative correlation r = −.18, t(1211) = 6.54, p < .001. Multiple regression: age b = −0.006, SE = 0.001, t(1139) = 6.12, p < .001, Cohen’s f = 0.032. Education positively predicted trait curiosity (b = 0.02, SE = 0.008, t = 2.68, p = .007); gender, race, income ns. • Age vs. state curiosity: Positive correlation r = .16, t(1211) = 5.72, p < .001. Single-level regression: age b = 0.02, SE = 0.003, t(1139) = 5.16, p < .001, Cohen’s f = .022. Females > males in state curiosity (b = 0.21, SE = 0.10, t = 2.16, p = .031). • Mixed-effects model (trial-level): age b = 0.02, SE = 0.003, t(1143) = 7.25, p < .001; ICC for the age (random slope) effect = 0.022. Average confidence positively predicted curiosity (b = 0.16, SE = 0.023, t = 7.01, p < .001). Education negatively related to state curiosity (b = −0.09, SE = 0.025, t = 3.82, p < .001). Females > males (b = 0.24, SE = 0.10, t = 2.47, p = .014). Racial differences: African American > White (b = 0.55, SE = 0.18, t = 3.08, p = .002); more than one race > White (b = 1.74, SE = 0.617, t = 2.82, p = .005). • Exploratory quadratic effect for state curiosity: significant quadratic age term b = 0.10, SE = 0.02, t(1141) = 3.05, p < .001; pattern suggests lowest state curiosity in middle age, with increases into older age. • Correct guesses: Age negatively related to number of correct guesses r = −.12, p < .001. More correct guesses associated with higher trait curiosity (r = .10, p < .001) and higher state curiosity (r = .07, p = .021). • Task order had no significant effect on trait curiosity (t(1216) = 1.54, p = .123) or state curiosity (t(1216) = 0.96, p = .335).

Findings support a nuanced, bidirectional relationship between age and curiosity types: trait curiosity declines with age, consistent with theories like socioemotional selectivity (shifts away from knowledge acquisition goals), whereas state curiosity in response to concrete, semantically rich materials increases with age. The positive association between trait and state curiosity indicates conceptual overlap yet distinct age trajectories. The exploratory quadratic trend suggests middle-aged adults may exhibit lower momentary curiosity than younger and older adults, potentially reflecting stress or happiness trends across the lifespan. Mechanistically, accumulated prior knowledge in older adults may both increase the motivational salience of related new information and reduce perceived cognitive costs (consistent with selective engagement), leading to greater situational curiosity. Confidence ratings also tracked with state curiosity, and education showed opposite associations with trait (positive) and state (negative) curiosity, underscoring the complexity of demographic covariates and measurement contexts.

This large, preregistered lifespan study demonstrates that age relates differently to trait and state curiosity: trait curiosity decreases, while state curiosity elicited by trivia increases, with evidence for a nonlinear pattern across adulthood. The work clarifies that curiosity is multifaceted and age effects depend on the construct and context. Practically, fostering domain-relevant, semantically grounded learning opportunities may engage older adults’ curiosity and support cognitive and social well-being. Future research should: examine diverse forms of state curiosity beyond trivia (e.g., magic tricks, real-world tasks), test learning outcomes and mechanisms linking prior knowledge to curiosity and engagement, assess populations with cognitive impairment, and longitudinally track changes to disentangle cohort effects. Investigating when and why middle adulthood shows a dip in state curiosity is also warranted.

• State curiosity assessed only via trivia questions; generalizability to other materials and contexts requires testing. • Learning outcomes were not measured; mechanisms linking curiosity to memory were inferred from prior work. • No screening for age-related cognitive impairment; online sample may not represent impaired populations. • Cross-sectional design; potential cohort effects and response biases cannot be fully ruled out, though overlap of trait and state measures mitigates concern. • Uneven distribution across covariate categories and the table-2 fallacy caution the interpretation of covariate effects. • MTurk sampling and extensive exclusions may affect generalizability; online task effort variability is possible. • Items participants already knew were excluded; older adults guessed fewer items correctly (r = −.12), which may interact with curiosity measures.