Language statistics shape spatial processing

Recent theories suggest that space serves as a scaffold for high-level cognition, including reasoning and language. Here, we tested a reverse scenario: namely, whether humans organize novel spatial information solely by relying on linguistic regularities. This possibility was firstly investigated by quantifying fine-grained systematic information embedded within city names. Computational evidence based on more than 40,000 cities across 5 European countries revealed systematic language-to-space regularities as indexed by both surface-level (i.e., letters and bigrams composing the city name) and distributional (i.e., as extracted from a word embedding model) linguistic information. Next, to test whether these linguistic regularities influence spatial judgements, we asked human participants to indicate the geographical location of never experienced but linguistically plausible strings of letters, presented as names of fictional cities (pseudocities). Results from multiple behavioral experiments revealed that participants' performance aligns with prior language-to-space bindings. This pattern was fully replicated both computationally and behaviorally in experiments focusing on USA and Mexican cities. These findings indicate that prior linguistic, sublexical regularities influence the organization of new spatial maps, suggesting a deeper interdependence between language and spatial cognition.