Laughter is the most frequently transcribed paralinguistic feature in social research interview transcripts, occurring even where the transcriber gives no other indication of how words were said. It is thus a useful starting point for reconstructing aspects of interaction from the traces in standard social science research transcripts. First, we examine the practices of a transcriber in recording laughter by comparing transcripts from one project to the audio recordings. We then analyse the placement of these tokens in transcripts from other projects, considering their relation to the immediately preceding and following talk, drawing on Wallace Chafe’s (2007) interpretation of laughter as the expression of a feeling of ‘non-seriousness’. The laughter marks a transition away from and back to a serious frame. We argue that attention to the recording of laughter as a variable transcription practice can draw the attention of researchers using standard orthographic transcripts to interviewees’ orientations to topics and to the interview process itself.