Uncertain world: How children's curiosity and intolerance of uncertainty relate to their behaviour and emotion under uncertainty

The study investigates how individual differences in children’s curiosity and intolerance of uncertainty (IU) relate to their behavioural (information seeking) and affective (emotion and worry) responses under uncertainty. Curiosity is conceptualised as a preference for uncertainty that drives exploratory behaviour, potentially linked to positive affect and learning. IU is a dispositional tendency to find uncertainty aversive, often linked to anxiety and worry. The authors aimed to examine, in children aged 8–12, whether parent-reported curiosity and IU predict information seeking and emotional responses when uncertainty is experimentally manipulated. Hypotheses: (1) Higher IU and higher curiosity would each be associated with more information seeking (more button presses) under higher uncertainty; (2) Higher curiosity would be associated with more positive affect under high uncertainty; (3) Higher IU would be associated with more negative affect and greater worry under high uncertainty.

Curiosity is multifaceted and has been linked to enhanced learning and wellbeing. It can be distinguished into interest-type (seeking new knowledge with positive affect) and deprivation-type (reducing an information gap with potential negative affect until resolved). The only validated parent-report measure for children, the I/D-YC, captures these subtypes. IU reflects difficulty enduring uncertainty and is associated with anxiety and worry; in adults, IU relates to safety behaviours and increased information seeking even at a cost. In children, IU is linked to anxiety and worry primarily via questionnaires; behavioural links are less clear. Prior child work (e.g., Beads task adaptations) showed increased information seeking and worry as task uncertainty rose but limited association with parent-reported IU, suggesting parent versus child reports may capture different facets (observable behaviour vs internal affect). Adult studies examining curiosity and IU together show curiosity relates more to arbitrary information seeking, whereas IU relates to outcome-relevant information seeking; IU correlates positively with deprivation-type curiosity and negatively with interest-type in some adult samples. Experimental work in adults shows uncertainty increases information seeking despite potential negative outcomes. However, individual differences in such tasks have been underexplored in children, motivating the current study combining curiosity and IU in a single child sample with both behavioural and affective measures under manipulated uncertainty.

Preregistration: Design, sample size, hypotheses, and analysis plan were preregistered on OSF. Task materials and code were made available. Participants: 133 children aged 8–12.96 years (M=9.71, SD=1.30); 68 boys, 64 girls, 1 preferred not to say. Predominantly White ethnicity; most parents had higher-education degrees. Inclusion: typical development, UK-based, normal/corrected hearing/vision. Ethical approval from University of Reading (2020-072-HD). Exclusions: Additional respondents failing inclusion or flagged as suspicious were not invited; 36 eligible not in final dataset due to non-completion/technical issues/withdrawal. Parent-report measures: (a) Curiosity via the 10-item parent-report Interest/Deprivation–Young Children (I/D-YC) scale (Interest- and Deprivation-type subscales; 4-point Likert). Internal consistency α=.82 (total). Subscales moderately correlated (r=.47). Main analyses used total score; subscales explored. (b) Intolerance of Uncertainty via the 17-item parent-report RULES (5-point Likert), α=.90. Task and procedure: An online jsPsych-based game (asynchronous). After calibration and a practice phase teaching button labels (neutral, aversive, uncertain), children completed four test trials (two high-uncertainty, two low-uncertainty) in counterbalanced order. Each trial included: camera check; a 10 s anticipation period showing 48 buttons while recording facial video; self-reports of emotional valence (5-point SAM), worry (4-point), and certainty (4-point; reversed to uncertainty); then a 1-minute button-press phase where children could press as many buttons as desired. Trial design: Four total trials. Each used one neutral and one aversive sound (from IADS-2) assigned across buttons. High uncertainty: 44 uncertain buttons and 4 certain buttons (2 neutral, 2 aversive). Low uncertainty: 4 uncertain buttons and 44 certain buttons (22 neutral, 22 aversive). Uncertain buttons had a fixed sound mapping (half neutral, half aversive). Sounds played for 2 s; maximum possible presses per trial = 30. Measures: (1) Behaviour: total number of button presses per trial; exploratory proportions of unique, certain, and uncertain buttons pressed. (2) Self-reports: emotional valence (SAM), worry (4-point), and uncertainty (4-point reversed). (3) Facial affect: During anticipation, coders rated subjective affect (−2 to +2) and coded objective smile/frown durations to derive a −1 to +1 score; a scaled composite (best ICC=.69) was used as the main facial affect outcome. Manipulation checks: Post-task sound ratings (SAM) for each sound; and trial-level uncertainty ratings. Data handling and analysis: Video missingness minimal; Winsorisation applied to outliers (RULES, button presses, facial affect) per preregistration. Self-report variables were skewed; robustness checks used binarised versions (results consistent). Linear mixed-effects models (lme4 in R) with participant random intercepts tested effects of trial uncertainty (effect coded: low=-1, high=+1), z-scored RULES (IU) and I/D-YC (curiosity), and their interactions with trial uncertainty for four DVs: button presses, self-reported emotion, self-reported worry, and facial affect. Model formulas: DV ~ (I/D-YC + RULES) * trial_uncertainty + (1|participant). Exploratory models: (a) Proportion of minority buttons pressed as DV to assess asymmetry: prop_minority_buttons_pressed ~ (I/D-YC + RULES) * trial_uncertainty + (1|participant). (b) Inclusion of age as a moderator with two- and three-way interactions. (c) Separate models using Interest- and Deprivation-type curiosity subscales.

Manipulation checks: Children reported higher uncertainty on high-uncertainty trials than low (b=0.04, 95% CI [0.01, 0.07], p=.010; trend when binarised). Neutral sounds were rated more positively than aversive (b=0.22, 95% CI [0.18, 0.27], p<.001). Behaviour (button presses): Children pressed slightly more buttons in high- vs low-uncertainty trials (b=0.25, 95% CI [0.01, 0.49], p=.039); means: High 18.79 (SD 3.30) vs Low 18.28 (SD 3.33). IU did not predict button pressing (b=−0.30, p=.268) and curiosity did not predict button pressing (b=0.10, p=.716). No interactions of IU or curiosity with trial uncertainty on button presses. Exploratory behaviour (type of buttons): Proportion of minority buttons pressed showed a main effect of trial uncertainty (b=0.04, 95% CI [0.03, 0.06], p<.001) and an interaction with IU (b=0.02, 95% CI [0.00, 0.03], p=.032). Children, especially those higher in IU, pressed proportionally more certain buttons in high-uncertainty trials than uncertain buttons in low-uncertainty trials, consistent with checking/safety-seeking. Facial affect: Trend toward less positive facial affect in high-uncertainty trials (b=−0.06, 95% CI [−0.13, 0.00], p=.066). IU (b=−0.05, p=.404) and curiosity (b=0.06, p=.385) did not predict facial affect; no interactions with trial uncertainty. Self-reported emotion: Trend to feel less happy in high uncertainty (b=−0.06, 95% CI [−0.13, 0.00], p=.055; not robust when binarised). Curiosity predicted higher happiness overall (b=0.20, 95% CI [0.07, 0.32], p=.002). IU did not predict happiness (b=−0.07, p=.266). No interactions with trial uncertainty. Subscales: Both Interest- (b=0.18, p=.006) and Deprivation-type (b=0.17, p=.009) curiosity related to higher happiness. Self-reported worry: No main effects of trial uncertainty (b=0.04, p=.145), IU (b=0.06, p=.176), or curiosity (b=−0.05, p=.269). No interactions with trial uncertainty. Age-related exploratory effects: Children reported less worry with age (b=−0.08, 95% CI [−0.15, −0.01], p=.020). Three-way interactions suggested that IU may relate to lower happiness/higher worry primarily in high-uncertainty trials for older children (8–12 years), though confidence intervals overlapped and effects were exploratory. Correlations among affect measures (trial level): Emotion and worry were negatively related (r≈−.48); emotion weakly positively related to facial affect (r≈.13); worry not related to facial affect.

Findings indicate that children generally seek more information under higher uncertainty, consistent with a basic drive to resolve uncertainty, but this behaviour was not modulated by parent-reported IU or curiosity in terms of quantity of button presses. Contrary to hypotheses, IU did not predict more information seeking nor more negative affect or worry under uncertainty. Curiosity was associated with greater self-reported happiness overall during the task, but this effect did not specifically increase under high uncertainty and was not reflected in facial expressions. Exploratory analyses suggest a qualitative behavioural signature of IU: children higher in IU showed greater asymmetry, pressing proportionally more certain buttons in predominantly uncertain contexts, consistent with checking or safety-seeking strategies rather than broad increases in information seeking. Age-related exploratory patterns hint that IU’s association with negative affect in uncertainty may strengthen in older children within this age range, but these effects require cautious interpretation and replication. Overall, the results suggest that in middle childhood, trait IU (parent-report) may manifest more in the type of information sought under uncertainty than in overall quantity or immediate affective reports, while trait curiosity aligns with more positive subjective experience during exploratory tasks.

IU did not predict children’s emotional responses or the amount of information seeking during an uncertain task. Exploratory evidence suggests IU may relate to qualitative differences in information seeking (e.g., checking/safety-seeking) rather than quantity. There may be developmental changes such that IU is more strongly related to worry under uncertainty in older children within the sampled age range. Children higher in curiosity reported feeling happier overall during the task, but greater uncertainty did not further increase their happiness. Generally, children sought slightly more information under higher uncertainty, independent of IU or curiosity. Future research should clarify behavioural manifestations of IU in children, consider developmental trajectories and levels of perceived threat and control, and investigate how curiosity can be leveraged to support motivation and learning.

• The uncertainty manipulation produced only a small difference in perceived uncertainty; children often reported feeling quite sure, possibly reflecting developmental overconfidence or limited variability in stimuli (few sounds per trial). Increasing the diversity of stimuli could enhance perceived uncertainty.
• The task may not have included sufficient threat to elicit IU-related negative affect; aversive sounds were mildly negative and children had control over exposure, potentially dampening worry/affect.
• Parent-report IU may not capture children’s internal affective states under uncertainty as effectively as child self-report, limiting observed associations with worry/affect.
• Facial affect measurement challenges: modest inter-rater reliability, weak correspondence with self-reported affect, and potentially limited sensitivity of facial expressions as indicators of emotion in this context.
• Online/home testing environment limited experimental control (e.g., sound volume, distractions), which may have affected affective responses and video quality.
• Sample not highly diverse socioeconomically or ethnically, potentially limiting generalisability.
• Limited availability of validated child measures for constructs like IU and curiosity may constrain measurement precision; newer measures (e.g., YIU-PR) may be more developmentally sensitive.
• Cross-sectional design precludes developmental inferences beyond exploratory age interactions.