a b s t r a c tThe search for weakly-interacting massive particle (WIMP) dark matter is multi-pronged. Ultimately, the WIMP-dark-matter picture will only be confirmed if different classes of experiments see consistent signals and infer the same WIMP properties. In this work, we review the ideas, methods, and status of directdetection searches. We focus in particular on extracting WIMP physics (WIMP interactions and phasespace distribution) from direct-detection data in the early discovery days when multiple experiments see of order dozens to hundreds of events. To demonstrate the essential complementarity of different directdetection experiments in this context, we create mock data intended to represent the data from the nearfuture Generation 2 experiments. We consider both conventional supersymmetry-inspired benchmark points (with spin-independent and -dependent elastic cross sections just below current limits), as well as benchmark points for other classes of models (inelastic and effective-operator paradigms). We also investigate the effect on parameter estimation of loosening or dropping the assumptions about the local WIMP phase-space distribution. We arrive at two main conclusions. Firstly, teasing out WIMP physics with experiments depends critically on having a wide set of detector target materials, spanning a large range of target nuclear masses and spin-dependent sensitivity. It is also highly desirable to obtain data from low-threshold experiments. Secondly, a general reconstruction of the local WIMP velocity distribution, which will only be achieved if there are multiple experiments using different target materials, is critical to obtaining a robust and unbiased estimate of the WIMP mass.