To clarify the feeding strategy of pelagic larvae of stone flounder in Mutsu Bay, the dietary composition and prey size was investigated from February to April during 1989–1999. Diets were compared with the numerical and volumetric compositions and frequency of occurrence of each prey species. Mensuration formulae were applied to estimate individual prey volume in the diet, while the length of planktonic species was measured from net and water samples. Prey shapes were assumed as sphere, cylinder, ellipsoid, pyramid, two elliptical cones, or a combination of ellipsoid and cylinder. Prey-size range increased as the larvae grew. Preflexion larvae fed mainly on copepod nauplii. Flexion and postflexion larvae ingested primarily appendicularians, with a suggestion that these larvae might depend on some parts of the microbial food web. Low frequencies of flexion and postflexion larvae with empty guts (1.7 and 1.4%, respectively) might be derived from feeding on slow-swimming appendicularians. From a size comparison between 'house'-like organ length and trunk length of the appendicularian Oikopleura sp., almost all house-like organs with trunks in the larval diet were nonexpanded 'house rudiments', not expanded 'houses'. Thus, stone flounder larvae may not chew the houses, but swallow the house rudiments with trunks.