This paper presents an analysis of combustion data from a spark ignition engine running at a variety of mixture strengths at low speed and load conditions. In each case, the heat release per-cycle is studied using statistical tools and nonlinear time-series analysis techniques to illustrate the systematic variation of the combustion. Measures are presented here which aim to detect these variations at less severe operating conditions than have been previously reported, and which, in turn, enables their use in regions more representative of those experienced in production calibrations. It is shown that as mixture strength reduces from an A/F of 14.5 to 20.5 the degradation in combustion increases, as expected. More importantly, as the variability increases it begins to exhibit a systematic structure, which contains repeated sequences that suggest the presence of deterministic behaviour.