The goal of the study is to reveal the ways in which bivalent verbs are divided into the transitive and intransitive classes in Indo-European languages. Based on a questionnaire that contained 130 predicates given in contexts, I assembled a database of these verbs’ valency patterns in a sample of 23 Indo-European languages, which covers all extant genera. The data were elicited from native speakers and annotated by language experts. Transitivity prominence (the ratio of transitive verbs in the questionnaire for a given language) is shown to vary hugely across the family, with its peaks in Romance, Germanic and Balkan languages (the core of Standard Average European) and its lows in Eastern European languages and Celtic. Transitivity prominence is a diachronically unstable feature, which correlates negatively with the size of case paradigm. Large-scale areal factors shape the distribution of transitivity prominence values across languages stronger than deep genetic relationships. Deviations from transitivity can be signaled by the use of non-core devices for coding the more agent-like argument or the more patient-like argument. The former pattern is generally rare and rather homogeneous across Indo-European: it is mostly found with low-control stative verbs, such as verbs related to possession and passive emotions. The latter pattern is more diverse in terms of both verb types and areal patterns. These minor patterns include special intransitive treatment of verbs of interaction, low-affectedness verbs and verbs which presuppose the absence of the second argument; all of these patterns are typical of various small-sized pockets across the Indo-European area. The data imply that high transitivity prominence in SAE is a recent innovation. Peripheral European languages are more conservative: they preserve the typologically unusual transitivity profile which can by hypothesised for Proto-Indo-European.