Total internal reflection microscopy is used to measure the total potential energy of interaction between a 6 mu m polystyrene (PS) latex bead and either a bare glass microscope slide or a glass slide spin-coated with a 1 mu m thick PS film, when the two interacting bodies are separated by 10-300 nm of aqueous solution having an ionic strength between 0.5 and 3 mM. In particular, these are the first measurements of van der Waals interaction between microscopic bodies of PS across water, for which the dielectric spectra are well-known. Under these conditions the bead is levitated above the slide by double-layer repulsion. After the gravitational contribution is subtracted, the potential energy profile displays a minimum of 0.5-2.3kT formed by long-range van der Waals attraction and shorter-range double-layer repulsion. The attraction was detectable at distances up to 200 nm. At separation distances greater than 100 nm (energy < 0.5kT), the measurements agree well with predictions using Lifshitz theory to predict the interaction of two PS half spaces, coupled with Derjaguin＇s approximation to account for the curvature of the sphere. At all separations, both retardation and screening are very important to the van der Waals interaction. As the separation becomes smaller than 100 nm, the measured interaction becomes weaker than predicted. Using atomic force microscopy (AFM), we observed asperities with heights up to 10 nm on the spin-coated PS film and up to 30 nm on the latex bead. If our experimental "zero" separation corresponds to contact of the largest asperities, the separation distance used in the theory should be larger than that measured. Shifting the theoretical curve by the sum of the asperity heights causes the theory to shift from overpredicting the van der Waals attraction to underpredicting it. We suggest a new theory in which roughness is treated as a diffuse film whose composition varies from pure PS at the inner surface to pure water at the outer surface. The composition profile can be determined independently from the histogram of elevations measured with AFM.