Imaging of blood flow perfusion is an area of significant medical interest. Recently, the advantages of using the total integrated Doppler power spectrum as the parameter that is encoded in color has been shown to result in an approximately threefold increase in flow sensitivity, a relative insensitivity to acquisition angle and lack of aliasing. We have taken this mode a step further and demonstrated the potential for quantifying blood flow using correlation-based algorithms applied to the power signal. We show that phi(tau) = phi(0)e-VT, where phi(tau) is the two-time correlation of the fluctuation in the power signal, and v is the specific flow (reciprocal of mean transit time). Scans of a dog's blood, pumped at a constant rate through gum rubber tubing, were obtained using a Diasonics Spectra 10-MHz linear array transducer at standard range-gated spectral mode (PRF = 1400 Hz, wall filter = 50 Hz, sample gate = 1.5 mm). A fixed Doppler angle of 68 degrees was used. Five different flow rates were tested, and the velocities determined by power decorrelation were compared to the mean velocities calculated from the Doppler shifts by linear regression (R2 = 0.987). We believe the results are very encouraging for using power decorrelation in perfusion evaluation.