Background: Play impairments in autism are intrinsic to its definition but there is continuing debate about whether they are specific to symbolic play or found across all play behaviours. In typically developing children, play development is strongly linked to other aspects of development, to the quality of caregiving behaviour and attachment relationships; however, these associations are less well characterised in autism. Three previous empirical studies have found links between play and attachment quality in autism similar to those in typical development. Additionally, a number of recent studies have investigated patterns of child attachment in autism, finding a range of attachment styles, which argue against primary attachment difficulties in the disorder.Rationale: This study aimed to i) make a detailed study of the early development of play in autism; ii) explore for the first time the role of caregiver play; iii) extend the small literature on the relationship between child play and child attachment. Method: 49 children aged two and five years diagnosed with core autism were group-matched with 45 neurotypical children on gender and non-verbal development. Detailed structured video-coding of both child and caregiver play in a naturalistic setting was made at baseline (T1), after seven months (T2), and after 13 months (T3), using Noldus Observer. Child attachment behaviour was assessed using the Brief Attachment Screening Questionnaire at T1 and T3. Results: Children with autism showed similar patterns of play change over time to controls but at a delayed rate; with more simple exploratory play and reduced advanced functional play at all three timepoints and reduced symbolic play development identified by T3. Group differences were also found in caregiver play. Quality and complexity of child and caregiver play were largely associated. Emergence of more advanced forms of child play in both groups was predicted by level of previous child play; and also level of caregiver play, but only if it was developmentally appropriate. Children with autism showed similar responses to caregiver behaviour as typical children, despite their core social impairments. By contrast, simple play reduced over time in both groups irrespective of the nature of caregiver play. There were group differences in child attachment behaviours at T1 and T3. However, attachment quality showed no relationship to child play development. Conclusions: The play of children with autism was generally delayed compared to controls, with no evidence of a specific delay in symbolic play. In both autism and typical development, emergence of more complex forms of play was predicted by developmentally sensitive caregiver play, consistent with the Vygotskian theory of a 'zone of proximal development'. Developmental reduction in simple play however was independent of caregiver, consistent with Piagetian theory of children as 'lone scientists' in this context. Child attachment and play quality were not specifically related.