The analytical, within-subject and between-subject components of variation were estimated for amylase activity and pancreatic isoamylase activity in serum measured using newer analytical methods. Desirable analytical imprecisions based on within-subject variation were CV less than or equal to 4.4% and CV less than or equal to 7.0%, respectively. Conventional population-based reference intervals were not useful because of marked individuality; clinical decision-making points should be derived from the desired sensitivity and specificity. Serial results must change more than about 30% and 40% respectively before significance (P less than or equal to 0.05) can be claimed. Similar data on total amylase and pancreatic isoamylase activities in random and first morning urines showed that the use of conventional reference intervals was appropriate. Very large changes (greater than 100%) were required before a difference in serial results was significant. Calculation of the urine amylase/creatinine ratio appeared to confer no advantage. Derivation of the ratio of pancreatic isoamylase to total amylase activity in serum or urine was unlikely to provide additional information of value in either diagnosis or monitoring.