The acoustic output of medical ultrasonic equipment is usually measured in water but to determine the safety (or performance) of a machine it is necessary to know the in situ acoustic pressure levels inside a patient. At present, when estimates of these levels are made, a linear propagation model is used although in practice the propagation is nonlinear. This paper shows that such a model can lead to large errors (80% or more in pressure) and in particular that in situ predictions of the peak-negative acoustic pressure are too low. To describe the field in water and in tissue it is necessary to take account of diffraction, attenuation, and dispersion as well as nonlinear propagation. It is difficult to use acoustic output levels in water to predict in situ values because the amplitude in water approaches a limit, an effect known as saturation. Nevertheless, a method of making such predictions is presented in this paper and is validated by comparison with experiment. The method is relatively time-consuming to implement and has not yet been applied to pulsed fields, so there is a need for more complete and simpler methods.