The seeing noise for perpendicular planes of vibration in a stellar source has been found to be coherent. Hence, an instrument has been designed and constructed that eliminates seeing noise from photoelectric polarization photometry by observing simultaneously the intensities of perpendicular planes of vibration and measuring the ratio, or the difference for ratios near one, of the two planes. The instrument consists of a Wollaston prism to separate the two perpendicular planes of vibration, a field lens of long focal length, 255 mm, to focus the objective of the telescope on the large unobstructed cathodes of two EMI 5311 multiplier phototubes, and a depolarizer to serve as a "zero" of intensity. The difference in the signals from the two phototubes, one receiving radiation in one plane of vibration and the second, radiation in the perpendicular plane, is measured with an amplifier the output of which is recorded on a Brown recorder. The internal accuracy of such an instrument is limited, not by seeing noise, but primarily by the shot noise in the photocurrent. From a series of 24 observations of ~ Herculis with the 82-inch reflector of the McDonald Observatory the probable error of a single observation has been found to be ~0.o00I6 mag. The largest deviation from the mean was o.ooo6 mag. This method of observing polarization should approach similar accuracy to the non-compensated DC photometer near the 10th magnitude with the 8~-inch reflector of the McDonald Observatory. The time required to complete an observation is no greater than other methods employed heretofore. Yerkes and McDonald Observatories, Williams Bay, Wis. and Fort Davis, Tex.