Safety performance data collected over a five year period, at a large Australian infrastructure project was analysed. The analysis examined the temporal relationships between the safety performance indicators measured at the project, including traditional lagging indicators, as well as expected leading indicators. The purpose of the research was to uncover time dependent relationships and explore causal relationships between indicators. The analysis revealed complex interactions between safety indicators over time. Notably, the expected leading indicators behaved as both leading and lagging indicators in relation to the project total recordable injury frequency rate. This finding suggested a cyclical relationship between management actions relating to safety and the rate of safety incidents. This cyclical relationship is unlikely to produce long term sustained improvement in safety performance. The expected leading indicators of safety were also inter-related with one another in complex ways. The results indicate that assumptions underpinning the use of leading indicators should be reconsidered. In particular, the findings challenge the assumption that leading indicators measured at one point in time can predict safety outcomes at a subsequent point in time. The collection and use of different types of safety indicator data should be reconsidered.