A key assumption in the reconstruction of extensive air showers using the air fluorescence technique is the hypothesis that fluorescence is proportional to energy deposition at all depths in the shower. This ansatz, along with the supposition that particle distribution and energy loss can be well modeled by modern shower simulation software, must be thoroughly verified in order to validate the air fluorescence technique. We report here the results of the first direct measurement of air fluorescence yield as a function of shower depth, as performed in the thick-target phase of the FLASH (FLuorescence in Air from SHowers) experimental program at the SLAC Final-Focus Test Beam facility. We compare observed fluorescence light yields as a function of shower depth to concurrently measured charged particle yields, to the predictions of the EGS and GEANT software packages, and to empirical energy-loss models. We also examine the extent to which the relative yield versus shower depth is independent of wavelength within the fluorescence spectrum.