The energy-flux model of seismic coda, developed by Frankel and Wennerberg (1987), is used to derive path-averaged estimates of scattering () and intrinsic attenuation () for northeastern North America. The model predicts the amplitude of the coda wave versus time as a function of frequency, , and . A nonlinear inversion scheme is developed that allows for the estimation of and as a function of frequency by fitting the model to a narrow band-pass-filtered envelope of the seismic coda for each seismogram at discrete frequency points, namely, 1, 5, 10, and 20 Hz. The inversion is performed on seismograms from earthquakes recorded by the Massachusetts Institute of Technology New England Seismic Network over a 15-year period between 1981 and 1995. Preliminary results indicate that scattering is the dominant mechanism of energy dissipation and that the effects of intrinsic attenuation are secondary. The scattering is strongest at 1 Hz and decreases with increasing frequency. The scattering is also strongest at shorter propagation distances and decreases substantially as the propagation distance increases. For example, at a frequency of 1 Hz, the value of is approximately 0.22 at an epicentral distance of 10 km and decreases rapidly to a value of approximately 0.05 at an epicentral distance of 50 km. Conversely, intrinsic attenuation is negligible at shorter propagation distances and increases as the propagation distance increases, and it does not show a strong frequency dependence. One interpretation of these results is the presence of a strong scattering region at shallow depth, with the scattering decreasing with increasing depth, and with a subsequent increasing of intrinsic attenuation at greater depth. Possible mechanisms for the scattering include the presence of a weathering layer near the surface, the presence of fractures in the shallow crust, and topography.