Research over the last forty years has highlighted the difficulties associated with teaching science in primary schools. The problems have been conceptualised in two main ways: 1. The lack of provision for science in primary schools in timetabling and resources. For example, Goodrum, Hackling & Rennie (2001) reported that in some primary schools in Australia, science was not taught at all; and 2. The attitudes of primary teachers themselves, grounded in their lack of knowledge and confidence to teach science (Gabel & Rubba, 1979; Harlen, 1997; Hodson, 2002; Wallace & Louden, 1992; Smith, 2014). For example, Hodson (2002) describes primary teachers’ attitudes towards science as “science phobia”. This paper is concerned with the assumptions made about the professional identity of primary teachers and problems associated with the scientific reform of the primary curriculum and the representation of both in science education research. To this end, an instrumental case for researching teacher agency that attends to the dialectic of person and practice (Edwards, 2016) is provided. The research, conducted at a time of product-oriented curriculum reform, is framed as a participatory ethnography and draws upon rich conversational interviews with experienced primary teachers. The teachers were involved in a school-based initiative to embed chemical science learning outcomes in their school curriculum for the first time. Open-ended interviews were conducted as professional workplace conversations between the author, employed by the school for the year as the scientist in residence, and the primary teachers. The interviews were analysed using positioning theory to preserve the social meaning of the primary teachers’ speech-action and coded using the grammar of agency (Martin, 2016). Significant disparity is shown between the ontologies of the primary teachers’ and research accounts that present mental state analyses of teachers’ lack of confidence or reluctance to teach science related to limited scientific understanding. The teachers’ agency in primary school science was found to be deeply embedded in the collective practices of primary teaching within the particular school context. The research is of interest to researchers and practitioners in science education reform and social researchers concerned with the operationalisation of human agency for social research more broadly.