Abstract
A growing body of evidence suggests that human cognition is shaped by culture, affecting processes from attention and memory to causal inference and concept representation. Such psychological differences are often described in terms of the analytic-holistic spectrum of cognitive styles. However, a unified mechanistic account of this variation remains elusive. In this paper, drawing on concepts from computational cognitive science and cultural evolution, we propose that the analytic-holistic spectrum reflects cognitive adaptations to different information ecologies. Cultural traits such as social organization, communication norms and prevalent learning modalities determine how observations are sampled, thereby affecting how much people should weigh parsimony when building mental models. We propose that information ecologies associated with holistic thought produce observations that more loosely reflect underlying structure (``weak sampling''), whereas those associated with analytic thought produce more diagnostic observations (``strong sampling''). These regimes license weaker and stronger applications of Bayesian Occam's Razor, respectively, which we show can provide a unified mechanism for cognitive differences along the analytic-holistic spectrum. By offering a computational lens on one of the most prominent empirical patterns in cultural psychology, we pave the way toward further informational-ecological accounts of how culture shapes cognition.