Curiosity and spatial learning are closely related in human cognition, especially when navigating novel environments where flexible representations must be encoded and updated. Theta-band oscillations (4–8 Hz) have been closely linked to spatial memory, but their role in curiosity-guided exploratory navigation remains unclear.

In this study, 37 participants completed a virtual navigation task across novel and familiar environments while scalp EEG was recorded. We introduced a novel approach that compared dynamic theta power with time-aligned curiosity estimates generated by a self-supervised model of intrinsic motivation (Random Network Distillation). We also examined how environmental context influenced theta power and dynamic effective connectivity and their relation to exploratory behaviour.

No significant differences in (dynamic) theta power or effective connectivity emerged between familiar and novel conditions. However, exploratory behaviour (indexed by Roaming Entropy) was positively associated with frontal theta during novel navigation (ρ = .34, p = .047). Curiosity scores from the RND model did not predict theta activity, suggesting that predictive novelty alone may not capture the neural correlates of spatial learning.

These findings suggest that theta power persists across all environments and that exploratory behaviour may be linked to theta power during novel navigation. Methodologically, we demonstrate the value of using dynamic theta power to capture time-sensitive neural fluctuations during naturalistic navigation. We also apply a simple, interpretable model of intrinsic curiosity to human data, enabling direct, time-resolved comparisons with neural activity. While model–brain convergence was limited, this framework offers a starting point for aligning neural dynamics with computational models for spatial learning and curiosity.

Key words: Spatial learning, curiosity, theta oscillations, EEG, computational modelling, Random Network Distillation, intrinsic motivation, Roaming Entropy, dynamic power analysis, effective connectivity, virtual navigation