A tracer (51Cr-EDTA) study was undertaken with juvenile (20 g) freshwater rainbow trout, Oncorhynchus mykiss, using an experimental design that minimized stress effects during feeding and drinking trials. A calculation procedure was developed, where feed intake (pellet number), tracer intake (mL), water in stomach contents (g) and drinking rate (mL·kg–1·h–1) are essential to discriminate between the major stomach water sources prandially and postprandially: water bound in food; initial water absorption of pellets; prandial water intake; postprandial water intake and endogenous stomach secretion. We put forward the hypothesis that intake of dry food with a minor water content (10 %) may impose a demand for water to moisture the feed up to the level in natural feed (75 %) as preparation for gastric emptying, whereafter food is ready to pass from the stomach through the pyloric sphincter. Moisture content of pellets increased from 9.4 to 24.9 % in the pre-meal period. The pellets were ingested with 4.0 to 19.3 μL water per pellet, reflecting high inter-individual variation. Prandially moisture content rose to 52 % and further increased to 56 % in the delay period. Moisture content was ca. 65 %, when pellets began to disintegrate and move through the sphincter in accordance with the hypothesis. Stomach secretion contributed 34–44 % of the stomach water and ingested water 25–35 %. The sampling and calculation procedure gave convincing evidence for the detailed stomach water budget and this individual approach can be very useful in comparisons of artificial and natural diets.