Two experiments were conducted to study changes induced by stage of lactation and milk ejection in the cisternal compartment of the udder in dairy cows. In experiment 1, 18 cows grouped according to stage of lactation were used 12 h after milking for measuring alveolar and cisternal milk volumes (by cannula) and cisternal area (by ultrasonography) in the front quarters. Cisternal milk and cisternal area were correlated (r = 0.74 to 0.82) for all stages of lactation. As lactation advanced, volumes of alveolar and cisternal milk and cisternal area decreased. Proportion of cisternal milk varied between stages (early, 33.2%; mid, 23.1%; and late, 42.6%). In experiment 2, 7 cows were used to show return of milk from cisternal to alveolar compartments when milk ejection was induced without milking. Cisternal area was measured before (0 min) and after (3, 15, 30, and 60 min) an i.v. oxytocin (OT) injection administered immediately before normal a.m. and p.m. milking times. Cisternal area increased dramatically from 0 to 3 min (98%) and decreased slowly thereafter. The 0- and 3-min data provide clear evidence of milk ejection, and their difference indicated cistern elasticity. Maximum cisternal area in each cow was similar for the 8- and 16-h milking intervals, indicating that in both cases the cistern was completely full of milk. Decrease in cisternal area after 3 min was significant at 15, 30, and 60 min. Decreased cisternal area was interpreted as the reflux of cisternal milk to the alveolar compartment. We termed this 'cisternal recoil.' In conclusion, ultrasonography was a useful method to evaluate dynamic changes in cisternal milk throughout lactation and after udder stimulation in dairy cows. Evidence exists that udder cisterns decrease when lactation advances and milk returns to the alveolar compartment when cows remain unmilked after milk ejection.