Theoretical models based on several hypothesized attenuation mechanisms are discussed in relation to published data on the effects of pressure and fluid saturation on attenuation. These mechanisms include friction, fluid flow, viscous relaxation, and scattering. The application of these models to the ultrasonic data of Toksoz et al (1979, this issue) indicates that friction on thin cracks and grain boundaries is the dominant attenuation mechanism for consolidated rocks under most conditions in the earth's upper crust. Increasing pressure decreases the number of cracks contributing to attenuation by friction, thus decreasing the attenuation. Water wetting of cracks and pores reduces the friction coefficient, facilitating sliding and thus increasing the attenuation. In saturated rocks, fluid flow plays a secondary role relative to friction. At ultrasonic frequencies in porous and permeable rocks, however, Biot-type flow may be important at moderately high pressures. 'Squirting' type flow of pore fluids from cracks and thin pores to larger pores may be a viable mechanism for some rocks at lower frequencies. The extrapolation of ultrasonic data to seismic or sonic frequencies by theoretical models involves some assumptions, verification of which requires data at lower frequencies.