Bulk raw milk samples from 2 farms, selected for their consistently low or high acid degree values (ADV) in milk, were used to determine whether any notable flavour difference could be detected between samples with widely different ADV. Milk samples were homogenized and pasteurized to give (i) low ADV, (ii) naturally high ADV and (iii) induced high ADV (as (i) but with delayed pasteurization). 46 of 72 consumers could detect a difference (P<0.05) between (i) and (ii), and between (i) and (iii) milks, and 33 could detect a difference between (ii) and (iii) milks. Concept scores, given by a trained panel, for rancid flavour intensity of (i), (ii) and (iii) milk resp. with ADV of 1.57, 5.37 and 7.67 were 0.08, 0.22 and 0.18 (slightly or moderately rancid) whereas a laboratory-prepared rancid sample (LPRS) with ADV of only 3.04 received a score of 0.30 (very rancid). It is concluded that the ADV provides an indication of changes in free fatty acid concn. but does not reliably predict rancid flavour intensity in farm milk or LPRS.