The measurement of tissue lipid and glycogen contents and the establishment of normal levels of variability are important when assessing changes caused by pathology or treatment. We measured hepatic and skeletal muscle lipid and glycogen levels using 1H and 13C MRS at 3 T in groups of subjects with and without type 2 diabetes. Within-visit reproducibility, due to repositioning and instrument errors was determined from repeat measurements made over 1 h. Natural variability was assessed from separate measurements made on three occasions over 1 month. Hepatic lipid content was greater in subjects with diabetes relative to healthy subjects (p = 0.03), whereas levels of hepatic and skeletal muscle glycogen, and of intra- and extra-myocellular lipid, were similar. The single-session reproducibility values (coefficient of variation, CV) for hepatic lipid content were 12% and 7% in groups of subjects with and without diabetes, respectively. The variability of hepatic lipid content over 1 month was greater than the reproducibility, with CV = 22% (p = 0.08) and CV = 44% (p = 0.004) in subjects with and without diabetes, respectively. Similarly, levels of variation in basal hepatic glycogen concentrations (subjects with diabetes, CV = 38%; healthy volunteers, CV = 35%) were significantly larger than single-session reproducibility values (CV = 17%, p = 0.02 and CV = 13%, p = 0.05, respectively), indicating substantial biological changes in basal concentrations over 1 month. There was a decreasing correlation in measurements of both hepatic lipid and glycogen content with increasing time between scans. Levels of variability in intra- and extra-myocellular lipid in the soleus muscle, and glycogen concentrations in the gastrocnemius muscle, tended to be larger than expected from single-session reproducibility, although these did not reach significance. Copyright © 2013 John Wiley & Sons, Ltd.