Objectives Chronic stress, characteristic of police work, affects the hypothalamic‐pituitary‐adrenal axis' control of cortisol production. Capacity to vary cortisol may be the appropriate measurement to interpret associations with chronic diseases, including obesity, best measured by variability within a person, not central tendency. Methods On each of 217 policemen, 18 saliva specimens were obtained for cortisol. Statistical models examined the associations of within‐subjects (W‐S) cortisol standard deviation (SD) and W‐S cortisol mean with waist circumference and four body composition indexes: BMI, and three derived from DEXA: fat‐mass, and trunk and extremities lean‐mass. Explained variance and the functional nature of associations are reported. Results Associations of anthropometrics with W‐S cortisol mean were not statistically significant at P < 0.05; all associations with W‐S cortisol SD were significant. The association of trunk lean mass index (LMIt) with W‐S cortisol SD dominated all models. Associations of W‐S cortisol SD with other indexes vanished when models contained LMIt ; when any other index was included in models predicting LMIt, associations with W‐S cortisol SD remained significant. The functional association between LMIt and W‐S cortisol SD is progressively “hockey stick,” monotonic increasing, and flattens at joint high values. Conclusions Results support inferences that LMIt measures visceral adiposity and W‐S cortisol variability appears to be an appropriate construct to measure in association with visceral adiposity. The “hockey stick” character of the association is consistent with other investigations suggesting obesity is associated with less W‐S cortisol variation; however, the monotonic increase and flattening of association at increasing W‐ScortisolSD values suggests a more complex association, potentially interpretable by allostasis models of causation. Am. J. Hum. Biol. 25:499–507, 2013. © 2013 Wiley Periodicals, Inc.