Bilingual children are most likely to speak both of the languages presented to them if they hear them from both parents and if no more than one parent speaks the language of the society at home (De Houwer, 2007). If one language is considerably underrepresented in the input, especially when it comes from only one parent, children combine longer stretches of speech from their dominant language with only individual words from their weaker language (e.g. Gaskins, Backus & Quick, 2019). As imperatives give rise to some of the earliest constructions in child speech (Keren-Portnoy 2006), we investigate their occurrence in the child’s input. More precisely, we ask how their usage compares in the input if delivered by one as opposed to two speakers in naturalistic interactions with a child aged 1;5–2;5 exposed to Polish and English from birth. Ten hours of data are derived from recordings made with the child’s Polish-speaking mother only and ten with her mother and a second Polish speaker. Our data show that type frequencies of imperatives are very similar in the two contexts. However, it is only when two speakers interact with the child that the meaning of imperatives is sometimes illustrated by one adult acting on the demands of the other. These findings are discussed with reference to the concepts of language frequency and transparency and the importance of multi-speaker contexts for the acquisition of a heritage language.