We used the immunogold method on ultrathin cryosections to measure intracellular amylase (Am) concentrations in subcellular compartments of rat exocrine pancreatic cells. Previously, the quantitation procedure was characterized in a model system consisting of Am dispersed at known concentration in a matrix of gelatin. Variations in labeling efficiency, due to differences in matrix density, were equalized by embedding in 30% polyacrylamide (PAA). Here we applied these model conditions to rat pancreas and established intracellular Am concentrations [Am]. Specimen blocks were composed of tissue and a reference layer of gelatin mixed with a known Am concentration ([Am]r), both fixed in glutaraldehyde. Cryosections of the PAA embedded blocks were immunogold labeled for Am. The labeling density was measured in the reference layer (LDr) and in structures in exocrine cells that were involved in Am synthesis and transport (LDs). In each of these structures the Am concentration ([Am]s) was calculated from: [Am]s = [Am]r. LDs/LDr In this way we measured average concentrations ranging from 63 mg/ml in rough endoplasmic reticulum to 261 mg/ml in secretory granules. Concentration of Am appeared to occur mainly in the most cis- and the most trans-Golgi cisternae. To check whether sterical hindrance was an inherent bias to the [Am] measurements in compartments that contained high concentrations of the enzyme, the labeling efficiency for Am in intact isolated secretory granules in gelatin and embedded in PAA, was compared with the efficiency when the granules were lysed and approximately 50 times diluted in gelatin before PAA embedment. It appeared that Am was detected with similar efficiency under both conditions. This demonstrated that sterical hindrance did not cause errors in the measurements of cellular Am concentrations.