In Fourier transform infrared (FT-IR) photoacoustic spectroscopy, advanced scanning and data-handling techniques have placed increasing emphasis on the phase of the photoacoustic signal. Unfortunately, there is no agreement on the best material to use as a phase reference. We have examined the frequency dependence of the signal from several candidate phase references and found that cell effects dominate the absolute phases and magnitudes observed. The absolute phase is exceptionally fast at low frequencies and exceptionally slow at high frequencies because of the cell effects. Accordingly, details such as sample position must be scrupulously controlled to achieve accurate, reproducible results. Because of the cell effects, no candidate material behaves like an ideal phase reference. If relative phases are used, however, glassy carbon comes closest to the ideal, differing from theory by no more than 8° at any frequency examined.