Goffman’s critique of the speaker-hearer dyadic model (1974) and cross-cultural studies suggest that children experience and learn language not just in dyadic but also in situated multiparty interactions (Ochs and Schieffelin, 1984). Family dinners involve several simultaneous activities, among which eating and languaging are central. In this multi-activity setting (Haddington et al., 2014), children and adults play different participants’ roles (speaker, addressee, bystander…) (Goffman, 1981). Children can be integrated in collaborative talk and broaden their experience of various interactional practices. They are socialized to conduct multiparty interactions and are dynamically involved in various participation frameworks. This study focuses on the role of gaze and how it may indicate (dis)engagement in different participation frameworks throughout the conversation in both signing and speaking family members. We analyzed dinners in four families composed of two adults and two children. Two families used French and two families used French Sign Language (LSF). The data was systematically coded for 1) audible or visible languaging or acting produced by all participants, 2) which participants the acting or languaging was directed to, and 3) all participants’ gaze orientation towards people or objects. Quantitative results from our coding are complemented with qualitative analyses of chosen extracts to provide a fuller picture of the identified features of dinner talk. Our analyses underline differences in gaze and turn distribution patterns according to the varying status of the participants, while they conduct eating actions that require moving objects and bodies. We highlight differences according to whether the participants use a spoken or a signed language and we show how parents adapt to and scaffold their children’s management of gaze as they need to orient their visual attention during family dinners.