All ultrasound exposure indices that have been formulated up to the present time assume that the propagation of ultrasound is linear. In fact, under most exposure conditions for which biological effects may be a concern, sound propagation is highly nonlinear. A nonlinear propagation model has been used to evaluate the nature of the effects that occur under realistic exposure conditions encountered in diagnostic procedures. Because of the way that the thermal index is defined, it turns out that ignoring nonlinear propagation leads to underestimates of tissue temperature increments that typically are less than 40%. As currently implemented, the mechanical index may be underestimated by more than a factor of 2 because it ignores the saturation of the sound fields that result from nonlinear propagation. For large propagation distances in soft tissues (e.g., 10 cm at 3 MHz in liver), however, it is physically difficult to exceed tissue pressures corresponding to MI>2 because of these same saturation phenomena.