In face-to-face communication, the interplay between gaze behaviour and backchannels (vocal feedback) is essential for shaping conversational dynamics. This study explores the intricate relationship between these elements in dyadic face-to-face interactions across three conversational contexts: getting to know each other, a task-oriented Tangram dialogue, and a spontaneous discussion. By integrating quantitative gaze data with qualitative speech annotation, we investigate the nuanced role of backchannels in various settings. Results reveal distinct patterns in lexical choice and intonation across contexts. While the Tangram task elicits predominantly rising tokens, free conversations exhibit unexpected intonational variation. Contrary to expectations, averted/directed gaze does not explain intonational differences. Surprisingly, mutual gaze precedes only a fraction of backchannels (1% in task-oriented speech, 27% in free speech), challenging the notion that mutual gaze invites backchannels. Mutual gaze during backchanneling varies by dyad.