Despite a growing body of knowledge about what content, processes and arrangements for learning may result in more effective initial teacher education, there remains a problem with the variability of outcomes from teacher education programmes. This paper reports on a multi-perspective exploration of what influences learning to teach in valued ways during initial teacher education. Framed by complexity theory, which emphasises the non-linear nature of social phenomena, the paper presents an analysis of 76 maps of influences on learning to teach (made by teacher candidates, teacher educators, mentor teachers and policy makers), looking for differences and patterns that might point the way to explanations about teacher candidates' varying ability to enact practice that improves outcomes for all learners.