This work addresses the estimation of modal parameters from measured vib ration data. A linear time-invariant model is fitted to the data and eig enfrequencies, damping ratios, mode shapes and, whenever possible, modal scaling factors, are extracted from the model. The current operational modal testing practice, where the model is estimated from output-only da ta, suffers from three important problems: the modal scaling factors can not be determined, the frequency content of the excitation can not be c ontrolled, and most often the accuracy of the modal parameters that are estimated in a single test can not be assessed. The goal of this work is to provide solutions for these problems. In order to solve the first two problems, it is proposed to apply measur ed artificial forces to the structure in addition to the unmeasured ambi ent excitation, and to identify a model that accounts for both excitatio n sources. Several simulation and measurement examples illustrate that i n this way the frequency content of the operational excitation can be en larged, the modal scaling factors can be determined, and the quality of the modal parameter estimates can be improved. Furthermore, the vibratio n response levels of the forced and ambient excitation are allowed to be of the same order of magnitude, which makes it possible to use small an d practical actuators on civil structures. In order to solve the third problem, expressions for the asymptotic prob ability density function of the estimated modal parameters are derived f or some system identification methods. From simulation and measurement e xamples, it follows that with the derived expressions, good estimates fo r the confidence intervals of the modal parameters can be obtained.