Natural language communication enables groups to overcome unrepresentative private evidence

Humans rely on social learning to aggregate knowledge that no single individual could acquire alone.Yet this process comes with significant epistemic risk. When each individual observes sparse and potentially unrepresentative samples of evidence, how do networks of learners overcome conflicting information to arrive at accurate collective beliefs? Formal models of this process typically assume that agents directly transmit their beliefs, but real human communication operates through natural language. Here we test the epistemic conditions under which language may help or hinder collective inference. In two experiments (N = 2,249), networks of four participants observed private samples from an underlying probability distribution and communicated over repeated rounds to estimate that distribution. We manipulated communication channel (natural language vs. numerical belief reports) and systematically varied the quality of private information: total sample size, representativeness of individual samples, and how evenly information was distributed across participants. All groups improved through social interaction, but groups restricted to numerical transmission performed systematically worse. Critically, the advantage of language emerged specifically when private samples were unrepresentative of the underlying truth, precisely when learners must appropriately weight others' evidence against their own. Analysis of message content revealed that this benefit arose from participants' ability to relay information across the network, extending the reach of evidence beyond direct dyadic exchange. These findings suggest that models of collective behavior must move beyond direct belief transmission to capture the epistemic work performed by natural language.