Extended Data Fig. 7: Association between age, home environment, country and wayfinding performance.

a, Wayfinding performance as a function of age for participants who grew up in city, suburb, mixed and rural environments. Data points correspond to the wayfinding performance averaged within 5-year windows. b, Difference in the effect of growing up outside cities on wayfinding performance across countries. We fit a linear mixed model for wayfinding performance, with fixed effects for age, gender and education, and random environment slopes clustered by country, as in Fig. 2a. Suburbs, mixed and rural environment slopes are represented, with City environment as baseline. Positive values correspond to an advantage compared to growing up in cities. Countries are ranked according to their suburb slope. The slopes of the different non-city environments are highly correlated: Pearson’s r(suburb, mixed) = 0.97, p < 0.001, r(suburb, rural) = 0.72, p < 0.001, r(mixed, rural) = 0.53, p < 0.001. The country ranking is very similar to the one with only 2 classes (city / non-city): Spearman’s r(non-city, suburb) = 0.85, p < 0.001, r(non-city, mixed) = 0.73, p < 0.001, r(non-city, rural) = 0.94, p < 0.001. P values are from a t-test testing the hypothesis of no correlation against the alternative hypothesis of a nonzero correlation. c, Pairwise differences between random environment slopes shown in panel b, averaged over countries. We show that the average difference in effect size between the city environment and the other 3 environments (city-rural, city-mixed, city-suburb) are around 10 times larger than the difference between the ‘non-city’ environments (rural-mixed, mixed-suburb, rural-suburb). This supports the approach to cluster together rural, mixed and suburb environments. All error bars correspond to standard errors, n = 397,162 participants.