Temporal dynamics of cognitive map formation in early- and late-onset blindness

Cognitive maps encode spatial relationships between locations and support flexible navigation. However, how these mental representations form in the absence of visual experience remains unclear. Here, we introduce a multisensory virtual navigation paradigm that allows to track the temporal dynamics of non-visual cognitive map formation. Sixteen early blind (EB), 17 late blind (LB), and 29 sighted controls (SC) learned the layout of a tactile maze. Participants repeatedly performed virtual pointing (estimating directions between locations) and navigation (reaching locations) tasks, which measured cognitive maps across multiple stages of learning. This method also enabled algorithmic inference of cognitive maps, providing insights into how mental distortions are progressively corrected. Although there were no group differences in average navigation performance, EB showed slower knowledge accumulation compared to LB and SC. In addition, both EB and LB had difficulties translating cognitive maps into first-person perspectives, resulting in reduced pointing and cognitive map accuracy. Yet, cognitive map accuracy improved progressively in all groups and a subset of EB and LB achieved expert-level performance with high navigation and pointing precision. In sum, this study provides a scalable framework for tracking alterations in cognitive map formation in blindness and other neurological conditions. Importantly, it demonstrates that cognitive map formation in the absence of vision is experience-dependent and trainable. Spatial disadvantages often observed in EB and LB thus do not reflect cognitive deficits but result from adaptive behavioral strategies constraining the use of allocentric cognitive maps.