the transits of circumpolar stars would be affected in the same way as those of stars south of the zenith, and the error of the computed azimuth would increase the effect upon the computed clock corrections. If the azimuth of the instrument had been determined from south stars, the effect would not appear in the clock corrections. With this line of reasoning, the difference between clock corrections observed at sunset and sunrise would appear to be due to an atmospheric deviation at those two epochs of the day.