A sufficiently large data base of repeated integral thermal-hydraulic loop tests has been accumulated recently from which generalized conclusions can be drawn. Evidence obtained from experiments performed in the University of Maryland College Park (UMCP) loop show that qualitative as well as quantitative differences exist between integral and separate effect tests. For separate effect tests, flow conditions are controlled continuously and usually steady (or quasi-steady) states are of interest. Integral facilities are “closed” systems and reactor safety oriented investigations center on transient behavior for which only initial conditions can be specified. It is shown that integral flow systems have a generic capability of amplifying (or damping) small perturbations and usually can operate in one of several possible alternate flow states. These characteristics can lead to two distributions of interexperiment variations; the differences can follow a Gaussian distribution or a bifurcation. In the UMCP test program, several examples of repeat experiments whose trajectories fall outside a Gaussian distribution were observed. Such experimental results have implications for the planning of experimental test programs and for the verification process of computational models.