Event descriptions require mapping event roles from an underlying conceptual representation to surface speech and gesture. Encodings in co-speech gesture tend to align with language-specific options that govern encodings in speech, but are relatively understudied for event roles that can be omitted in speech (e.g., argument-dropping languages like Turkish allow omission of core event roles, including agents and recipients). We examine the content of multimodal possession-transfer event descriptions across two typologically distinct languages (English, Turkish), differing in the grammaticality of argument-dropping. We find that language-specific encoding patterns heavily affect recipient and agent mentions in free event descriptions across modalities. Overall, Turkish speakers mentioned recipients and agents less frequently than English speakers. Although recipient and agent co-speech gestures were used more frequently in Turkish, they rarely contributed information beyond what was encoded in speech. This suggests that argument-dropping in Turkish occurs at a level of representation that is shared across modalities.