ATP distribution at rest was studied in 11 Thoroughbred horses, 8 of which had been admitted for clinical post mortem examination. A biopsy of the middle gluteal muscle was obtained from 6 horses immediately after death, following which a larger piece was taken by dissection from the same area. From the other 2 horses, only cut sections were obtained. The risk of damage to fibres dissected from the inner portion of the cut samples was considered less than for fibres from biopsies. From a separate 3 standing horses, a single biopsy of the middle gluteal muscle was taken at rest. Following freeze-drying, single fibres were dissected, weighed and analysed for ATP by luminometry. Fibre type was determined by staining for myosin ATPase. Variance in ATP content due to weighing error (se), analytical precision (sp) and the variability in content within a fibre (SDf) were determined in 16 fibres. Total variance (SDt) in ATP content was estimated both for cut sections and biopsies. Data were analysed by ANOVA using an hierarchical model to estimate the specific variance within (s̀wf) and between (s̀bf) fibres. ATP distribution was the same in cut sections and biopsy samples. The mean content of all fibres at rest was 20.14 mmol/kg dm. Values for se, sp, SDf and SDt were 0.49, 0.61, 1.61 and, 3.38 (cut sections) 4.28 (biopsies) mmol/kg dm, respectively; s̀wf was 1.53 mmol/kg and s̀bf 2.91 and 3.92 mmol/kg dm in cut sections and biopsy samples, respectively. ATP contents did not differ between fibre types. We conclude that most of the variance observed in measurements of single fibre ATP contents is due to variance between and within fibres rather than to the imprecision of the procedures used.