Originally proposed as a method for measuring the power gain of the cochlear amplifier, Allen-Fahey experiments compare intracochlear distortion products and ear-canal otoacoustic emissions (OAEs) under tightly controlled conditions. In this paper Allen-Fahey experiments are shown to place significant constraints on the dominant mode of reverse energy propagation within the cochlea. Existing Allen-Fahey experiments are reviewed and shown to contradict the predictions of compression-wave OAE models recently proposed in the literature. In compression-wave models, distortion products propagate from their site of generation to the stapes via longitudinal compression waves in the cochlear fluids (fast waves); in transverse traveling-wave models, by contrast, distortion products propagate primarily via pressure-difference waves whose velocity and other characteristics depend on the mechanical properties of the cochlear partition (slow waves). Compression-wave models predict that the distortion-product OAEs (DPOAEs) measured in the Allen-Fahey paradigm increase at close primary-frequency ratios (or remain constant in the hypothetical absence of tuned suppression). The behavior observed experimentally is just the opposite-a pronounced decrease in DPOAE amplitude at close ratios. Since neither compression-wave nor simple conceptual "hybrid-wave" models can account for the experimental results--whereas slow-wave models can, via systematic changes in distortion-source directionality arising from wave-interference effects--Allen-Fahey and related experiments provide compelling evidence against the predominance of compression-wave OAEs in mammalian cochlear mechanics.