Concerns about nuclear power plant safety have stimulated research in thermal hydraulics of reactor cooling systems. Complementary efforts in computer code development and in experiments using scaled models are being made. The applicability of the experimental results to a full-size power plant system depends on the scaling criteria on which the test facility is designed. Several sets of scaling criteria can be found in the literature, not all of them compatible with one another. A critical review and clarification of a number of these scaling criteria are presented. Specifically, the commonly recognized linear scaling, volume scaling, and several sets of single- and two-phase scaling criteria recently derived by M. Ishii are examined in terms of their limitations and interrelationships. It is shown that (a) as far as thermal-hydraulic modeling is concerned, Ishii's time-distorted scaling is the most general one to date, (b) Ishii's scaling offers the model designer a flexibility in the height of the model, and (c) both the linear and volume scaling are special cases of the Ishii scaling, and each has its own practical limitations.