Measurements of the diffusive transport of multiply scattered ultrasonic waves show that the energy velocity is very similar in magnitude and frequency dependence to the group velocity. Our data are accurately described using a theoretical model that accounts for the renormalization of scattering by the coupling between neighboring scatterers, quantitatively predicting the scattering delay that causes the strong frequency dependence of these velocities seen in our experiments. This gives a unified physical picture of the velocities of energy transport by both diffusive and ballistic waves.[ S0031-9007(97) PACS numbers: 43.35. + d, 43.20. + g, 62.30. + d In recent years, there has been a tremendous revival of interest in the propagation of classical waves through inhomogeneous media containing random scatterers [1,2]. When the scattering is strong, the propagation is typically very well described using the diffusion approximation, and this has successfully facilitated the interpretation of a wide range of fascinating wave phenomena. Even though the transport is diffusive, it is, nevertheless, still essential to define several propagation velocities. These include the group ͑y g ͒ and phase ͑y p ͒ velocities which characterize the ballistic, or unscattered, component of the incident waves. For diffusive waves, the relevant velocity is the energy velocity y e which is defined as the ratio of the energy flux to the energy density [3], and which corresponds to the average local velocity of energy transport in the diffusion process, since y e is related to the wave diffusion coefficient D by D y e l ء ͞3. Here l ء is the transport mean free path, or the distance the waves must propagate until their direction is randomized. Within this scenario, the energy velocity represents a distinctly different type of velocity, as wave coherence is no longer relevant. The first calculations for diffusive waves by the Amsterdam group [4], carried out for low volume fractions of scatterers f, correctly accounted for the extremely low value of y e found in light scattering experiments; they also predicted a strong frequency dependence for y e in the vicinity of Mie resonances as a result of the temporary storage of wave energy inside the scatterer [4][5][6][7]. However, these calculations fail as f is increased. Following an idea proposed by Sheng [2], recent Coherent Potential Approximation (CPA) calculations [8,9] have attempted to rectify this deficiency for higher volume fractions, and have suggested that the large variations of y e with frequency are washed out with increasing f; these results are purported to be in excellent agreement [8,9] with the limited experimental data that currently exist for y e [10,11]. However, in the intermediate frequency regime where resonant scattering occurs, such CPA theories become suspect [12], raising questions about their reliability. In addition to these questions about the behavior of y e , there have been conflicting ideas about the possible connection between the energy velocity ...