Effects of main sources of bone biopsy sample variability have been examined. Variability was assessed in double iliac crest biopsies of 12 subjects with normal or pathological bone. Components of variance were estimated as follows: two biopsies per patient; three specimens at different distances from compact bone; three sections per specimen; three microscopic fields per section. The following indicators were measured: bone volume (BV/TV); osteoid volume (OV/BV); osteoid surface (OS/BS); osteoblast surface (Ob.S/BS); eroded surface (ES/BS); osteoclast surface (Oc.S/BS); osteoid thickness (O.Th). Sources of variability were assessed by ANOVA for random effects. On the basis of the results, between fields variation gave the main contribution to the error of single measures (BV/TV, 50%; O.Th, 70%; OV/BV, Ob.S/BS, OS/BS, Oc.S/BS, ES/BS more than 80%). Distance from compact bone affected mostly the BV/TV (40%) and the O.Th (10%) error. When bone specimens at intermediate distance from cortical bone are examined, variations due to different biopsies, different section and different microscopic field are largely reduced by measuring 12 microscopic fields for BV/TV (14%) and 48 microscopic fields for the other indicators (O.Th 16%; OV/BV, Ob.S/BS, OS/BS, Oc.S/BS, ES/BS more than 30%). The precision can be only slightly improved by further increasing the number of the microscopic fields.