Seismic verification of a structure by application of true earthquake motions is not practicable. Expected ground motions due to seismic events vary with geographic location, mounting conditions differ, etc. Consequently, it is often essential to give the structure a seismic classification which is as general as possible. Deterministic signals will be preferred, e.g. sine beat and sine sweep functions. The problem is then to connect the parameters of the test signal with the characteristic properties of an assumed earthquake-induced motion. This is generally effected by comparing responses of linear single-degree-of-freedom (SDOF) systems to the two types of signals. In many cases this works out well for mechanical systems as steel structures. Test programs are, however, more and more frequently produced for other types of equipment, with the implicit assumption that the same relationship exists between excitation and response as is the case in the SDOF system. In structures with several degrees of freedom, where the modes of motion are coupled, or in systems with significantly non-linear properties, the accuracy of the comparison is disturbed. The aim of the present work is to show that this kind of disturbances has to be taken care of already when planning the test program. The work is performed as a case study for a structure which has been tested according to customer specifications, and which, although conceptually simple, contains some properties typical for a "difficult" system. Test measurements are extended with independent measurements, and experimental results are compared with results from analyses of models built on various degree of idealization.