Qualitative researchers can discard data that are unsaturated or unrelated to research questions, but what do we do when these data affect us, or 'haunt' us, 'long after collecting "it"' (Taylor 2013, 691)? In this paper, we draw upon Sara Ahmed to guide our engagement with 'discarded data': young children's gendered accounts of violence that unexpectedly arose during interviews about rest-time and relaxation in childcare. We show what it feels like to be affected by children's accounts throughout the non-linear and zig-zagged 'data analysis' processes. Applying Ahmed's conceptualisation of power as 'directionality', we critique the power of qualitative research conventions to define our focus as researchers and pay attention to what children raise: how they are directed towards gendered futures. We find that, in children's accounts of violence, boys have more agency than girls as they participate in -- and respond to -- violence.