Curiosity shapes spatial exploration and cognitive map formation in humans

The study investigates how momentary states of curiosity influence spatial exploratory behavior and the formation of cognitive maps in humans. While hippocampus-dependent navigation research has established that exploration of novel environments is key to building cognitive maps, the motivational drivers of exploration—especially in the absence of external rewards—remain unclear. The authors posit that curiosity, defined as the desire to seek novel information, may be a primary driver of exploration. They hypothesize that higher pre-exploration curiosity will stimulate spatial exploration in novel virtual rooms and, in turn, improve the precision of cognitive maps. They also distinguish pre-room curiosity (anticipatory motivation to explore) from post-room interest (retrospective evaluation of engagement), predicting distinct effects on spatial versus visual exploration and subsequent spatial memory.

Prior work in animals and humans demonstrates that active exploration supports cognitive map formation and efficient navigation. Foundational theories (Tolman; O’Keefe & Nadel) and contemporary frameworks propose that curiosity drives exploratory behavior and may facilitate hippocampus-dependent memory via dopaminergic mechanisms. Human curiosity research has focused largely on epistemic curiosity and semantic learning, showing that curiosity enhances information seeking and memory. Perceptual curiosity has been linked to responses to novel or surprising stimuli. However, the role of curiosity in spatial exploration and cognitive map formation in humans is understudied. Recent human studies suggest curiosity motivates information seeking (e.g., online search) and anticipatory gaze, but direct tests of curiosity’s impact on spatial exploration patterns and memory for spatial-relational details are lacking. This work fills that gap by measuring exploration complexity (roaming entropy) and assessing spatial memory via sketch maps.

Design and stimuli: 18 distinct 16 m × 16 m virtual rooms were created (two for familiarisation/practice; 16 for exploration), designed with comparable complexity and object density. Rooms contained layout-defining objects (e.g., sofa, bookshelf) and smaller details, with occlusions and placements requiring movement to see all features, preventing full mapping from a single viewpoint. An outdoor pier connected to each room via a zigzag pathway; room labels were visible on the room’s front wall. The environment was built/presented in Unity 3D (v2019.4.15). Apparatus: Desktop PC with LCD monitor (1920×1080, 60 Hz). Movement via keyboard (W forward, S backward); mouse to change field-of-view/steering. Footstep audio via headphones; sideward movements restricted; movement speed ~3.4 m/s. Participants: Recruited from Cardiff University. Experiment 1: N=28 (after exclusions; 3 men, 25 women; age 18–25, M=19.79, SD=1.70). Experiment 2: N=60 (5 men, 55 women; age 18–25, M=19.6, SD=1.28). Inclusion: normal hearing and normal/corrected vision; naïve to study aims. Ethics approved by Cardiff University School of Psychology. Procedure: Three phases: familiarisation, exploration, and (Experiment 2 only) a sketch map memory test. In familiarisation, participants explored two example rooms. In exploration, participants completed 16 trials (room order randomized), starting at the pier where they rated pre-room curiosity (1–10). After walking the pathway, pressing E triggered a 5 s door animation; inside, participants explored freely with position and head-direction recorded at 60 Hz. Upon leaving (press B), they rated post-room interest (1–10). Experiment 2 additionally administered the Five-Dimensional Curiosity Revised (5DCR) scale before familiarisation and included an immediate sketch map test after a 5-min break, where participants drew room layouts on provided square outlines, including furniture, doors, and windows; drawing order randomized. Five participants were excluded from memory analyses due to minor instruction/order changes. Exploration quantification (Roaming Entropy): Path roaming entropy (Path RE) indexed spatial coverage, computed as Shannon entropy of time spent across accessible 0.5 m × 0.5 m grid cells within a 32×32 down-sampled room grid (occluded cells removed via AI/manual labeling). Entropy was normalized by log2(k), where k is the number of accessible cells, yielding 0–1 (sometimes expressed as %). Head-direction roaming entropy (Head-direction RE) indexed visual scanning breadth, computed over an 18×36 grid of head directions spanning horizontal −180° to 180° and vertical −90° to 90°, with 10°×10° cells; normalized by log2(k) with k=648. Sketch map scoring (Experiment 2): Two independent raters scored four dimensions on 1 (poor) to 5 (excellent) with 0.5 increments: Object Presence (OP), Spatial Distortion/Rotation of Features (SD), Relative Positioning (RP), and Spatial Proportion (SP). Scoring emphasized spatial relationships and layout accuracy (minimizing impact of drawing skill). Inter-rater reliability ranged 0.70–0.78; internal consistency Cronbach’s alpha posterior mean 0.93 (93% HPDI [0.92, 0.94]). Composite cognitive map precision scores were computed by averaging the four dimensions per room. Statistical analysis: Bayesian multilevel (and multivariate/bivariate) models using brms in R (4 chains, 4800 iterations per chain), with weakly informative normal priors centered at 0. Predictors were centered at individual means to assess intra-individual relationships, with grand-mean-centered averages to capture inter-individual effects. Models included duration spent in each room as a covariate (Exp1 M=30.35 s, SD=19.19; Exp2 M=33.10 s, SD=22.82). Outcomes included Path RE, Head-direction RE, and composite cognitive map scores. Posterior means and 93% HPDIs summarized effects; residual correlations between RE measures were estimated in bivariate models. Posterior predictive checks and sensitivity analyses were conducted. Study not preregistered.

Exploration measures and ratings varied across rooms and participants. Path RE and Head-direction RE were positively correlated (residual correlations after accounting for predictors: Experiment 1 posterior mean=0.18, 93% HPDI [0.10, 0.27]; Experiment 2 posterior mean=0.34, 93% HPDI [0.28, 0.39]), and curiosity and interest ratings were positively correlated. Double dissociation between curiosity and interest effects on exploration replicated across experiments: - Experiment 1 (N=28): Path RE positively associated with pre-room curiosity (β=0.0059, 93% HPDI [0.0014, 0.0104]); tentative negative association with post-room interest (β=−0.0040, 93% HPDI [−0.0089, 0.0012]). Head-direction RE positively associated with post-room interest (β=0.0048, 93% HPDI [0.0016, 0.0080]); indeterminate with pre-room curiosity (β=−0.0005, 93% HPDI [−0.0030, 0.0020]). Differences between exploration types: curiosity effect difference=0.0064 [0.0017, 0.0110]; interest effect difference=−0.0088 [−0.014, −0.0030]. - Experiment 2 (N=60): Replicated patterns: Path RE with pre-room curiosity (β=0.0039, 93% HPDI [0.0003, 0.0073]); tentative negative with post-room interest (β=−0.0023, 93% HPDI [−0.0062, 0.0016]). Head-direction RE with post-room interest (β=0.0058, 93% HPDI [0.0034, 0.0082]); indeterminate with pre-room curiosity (β=−0.0009, 93% HPDI [−0.0028, 0.0011]). Effects persisted when controlling for room type. Trait curiosity moderation (Experiment 2, 5DCR): Stress Tolerance strengthened the positive link between pre-room curiosity and Path RE (interaction β=0.0041, 93% HPDI [0.0012, 0.0068]); Deprivation Sensitivity showed a potentially positive interaction (β=0.0025, 93% HPDI [−0.0003, 0.0054]); other subscales showed no reported robust interactions. Cognitive map formation (Experiment 2): Composite precision positively associated with pre-room curiosity (β=0.066, 93% HPDI [0.027, 0.10]); tentative negative association with post-room interest (β=−0.023, 93% HPDI [−0.060, 0.014]). Path RE positively associated with cognitive map precision (β=0.80, 93% HPDI [0.0076, 1.59]); no intra-individual evidence for Head-direction RE (β=−0.075, 93% HPDI [−1.22, 1.12]). At the inter-individual level, average Head-direction RE showed weak positive association (β=2.22, 93% HPDI [−0.47, 4.80]). Mediation analysis suggested a modest potential mediation of curiosity’s effect on map precision via Path RE (posterior mean=0.0028, 93% HPDI [−0.00035, 0.0077]).

Findings directly address the hypothesis that curiosity drives spatial exploration and enhances cognitive map formation. Pre-room curiosity selectively increased spatial coverage (Path RE), while post-room interest selectively increased visual scanning breadth (Head-direction RE), revealing a double dissociation between anticipatory curiosity and retrospective interest in shaping distinct exploration facets. Stress Tolerance, a trait dimension of curiosity, magnified the curiosity–exploration link, indicating that the ability to cope with uncertainty facilitates acting on curiosity. Curiosity and spatial exploration both contributed to more precise cognitive maps, suggesting that curiosity-driven exploration supports the encoding of spatial-relational details. The results align with theoretical accounts implicating dopaminergic midbrain and striatal engagement during curiosity, which interacts with hippocampus-dependent learning and memory consolidation. Analogies to animal behavior (e.g., rearing in rodents) underscore the importance of distinguishing movement-based exploration and visual scanning modes; Path RE related more strongly to within-person cognitive map precision, whereas Head-direction RE showed a potential between-person relationship. The tentative negative association of post-room interest with Path RE and cognitive map precision may reflect increased focus on specific objects or features at the expense of encoding global spatial layout, contrasting with trivia paradigms where post-information interest tends to enhance item memory. Overall, the work bridges curiosity research in semantic domains with spatial cognition, highlighting curiosity’s broader role in guiding exploration and structuring memory representations.

The study provides the first direct evidence in humans that momentary curiosity states enhance spatial exploration and, consequently, the precision of cognitive maps. Curiosity-related increases in spatial coverage and the moderating role of Stress Tolerance demonstrate how motivational and trait factors shape exploration. These insights have practical implications for designing built and virtual environments—architecture, urban planning, museums, and games—to harness curiosity and improve exploration and memory. Future work should examine neural mechanisms (dopaminergic–hippocampal interactions), incorporate complementary navigation/wayfinding assessments, analyze environmental features’ influence on interest and exploration, and ensure more balanced samples to assess generalizability across gender and other demographics.

• Cognitive map assessment relied on sketch maps; future studies should complement this with navigation or wayfinding tasks to capture additional aspects of spatial knowledge and its application. - Path roaming entropy primarily reflects spatial exploration but may also capture elements of visual exploration; Head-direction roaming entropy reflects visual exploration yet may be influenced by spatial layout. Despite overlap, most variance corresponds to their intended constructs. - Samples in both experiments were heavily skewed toward women; future research should balance gender representation to evaluate generalizability. - Detailed analyses of environmental features and their relationships to interest and exploration were not conducted; such analyses could clarify how environmental characteristics modulate visual exploration and engagement.