During every summer Olympic Games, the sport of gymnastics rises briefly to the world’s attention as the public admire the incredible skills and feats performed by fit muscular bodies on a range of apparatus. The gymnastics they watch consists of performances in which bodies assemble with apparatus. This thesis utilises an Actor Network Theory (ANT) perspective to follow this assembling of gymnastics in the five codes of competitive gymnastics competed in New Zealand: women’s artistic gymnastics, men’s artistic gymnastics, rhythmic gymnastics, trampolining and competitive aerobics. This thesis is a descriptive ethnography of the world of high performance gymnastics. It begins by examining some of the controversies that have operated to both criticise and rework the sport. Next, the gymnasts are followed through the selection processes that lead them to become members of national squads and teams. It then moves to the training gymnasium and examines the variety of non-human actants that work in the gymnasium to assemble gymnastics. The next two chapters examine how gymnasts are found to enrol and assemble with video technologies and sports science professionals in their efforts to improve performance. Following this, gymnasts are observed to produce a routine at a competition which is translated into a score and ranking through the highly complicated and laborious process of judging. Finally, the thesis concludes with the story of Angela McMillan, New Zealand’s most successful athlete within the gymnastic codes. Throughout are a range of accounts from participants, together with observations, describing attempts to secure the stabilisation of gymnastics as an actor-network that produces internationally successful athletes. All the networks followed involve a continual process of enrolling, un-enrolling, translating and mediating, with power constantly shifting and being shared between various heterogeneous actants including coaches, parents, the national federation and the international federation. At times these networks stabilise with particular actants, such as sports scientists or technologies, being enrolled, while at other times the paths of the networks come to an end as particular assemblages or actants, such as physical ability tests, are no longer enrolled. In contrast to a perception that successful high performance sports include key actors and resources, this thesis shows how the networks that produce high performance gymnasts are highly unpredictable and messy, with humans and non-humans both equally influential in affecting every branch of the networks. Processes such as talent identification, training and judging are found to be complicated and unstable.