The present work focuses on how signers and speakers demonstrate forms of engagement through visible cues (gaze, body orientation, postures, suspension of actions) with or without providing visual or vocal feedback in multi-party conversations. The dinners involve complex participation frameworks (Goffman, 1981, Goodwin, 1981) and coordination of activities (Haddington et al., 2014). Analyses are carried out on the DinLang Corpus (Morgenstern et al., 2021) which includes video recordings of French middle-class speaking and signing families during dinner at home. The specificity of family dinners is that participants are constantly alternating between languaging (Linell, 2009) and eating, which requires a finely tuned orchestration of their bodies. How do participants thus manage to signal their engagement while managing these different activities simultaneously? We present multimodal analyses of both signers and speakers from 4 signing families and 4 speaking families (two dinners per family) with children between 3 and 12 years old. Three major trends emerge: participants may display signs of active participation without providing backchanneling while suspending their current activity, or they may do both simultaneously; in other cases, however, participants may withdraw from the current participation arrangement by focusing on a single activity. By observing the subtle coordination of bodies in those complex multiparty interactions, this study illustrates the subtle impact that addresses' and overhearers’ engagement and backchanneling have on complex participation frameworks and on interactive dynamics, as participants constantly navigate between different activities and modalities.