The adoption of cultural variants by learners is affected by multiple factors including the prestige of the model and the value and frequency of different variants. However, little is known about what affects onward cultural transmission, or the choice of variants that models produce to pass on to new learners. This study investigated the effects on this choice of congruence between two contexts: the one in which variants are learned and the one in which they are later transmitted on. We hypothesized that when we are placed in a particular context, we will be more likely to produce (and therefore transmit) variants that we learned in that same (congruent) context. In particular, we tested the effect of a social contextual aspect–the relationship between model and learner. Our participants learned two methods to solve a puzzle, a variant from an “expert” (in an expert-to-novice context) and another one from a “peer” (in a peer-to-peer context). They were then asked to transmit one method onward, either to a “novice” (in a new expert-to-novice context) or to another “peer” (in a new peer-to-peer context). Participants were, overall, more likely to transmit the variant learned from an expert, evidencing an effect of by prestige bias. Crucially, in support of our hypothesis, they were also more likely to transmit the variant they had learned in the congruent context. Parameter estimation computer simulations of the experiment revealed that congruence bias was stronger than prestige bias.