The performance of a pair spectrometer of the Walker and McDaniel type is discussed from both a theoretical and an experimental point of view. It is shown that the energy of the γ-ray may be measured to a first approximation by the product of the distance between the inner edges of the slits which define the coincidence counters and the highest value of the magnetic field at which coincidences are found. A more accurate value of the energy of the γ-ray may be obtained by adding a small correction, called the "toe" correction, to the result so obtained. The magnitude of this correction is dependent on the γ-ray energy and on the width of the slits and is obtained from calculations of the shape of the coincidence peak based on the Bethe–Heitler formula. The correction depends very little on the length of the slits or on the thickness of the radiator. The relation between the calculated correction and the slit width has been examined experimentally. It is shown that the ultimate accuracy in energy measurement depends on the homogeneity of the magnetic field and on the error involved in the calculated toe correction. The latter error can be minimized by using very narrow slits and may be eliminated by plotting the value of magnetic field at the upper limit of the coincidence spectrum against the slit width and extrapolating the curve so obtained to zero slit width. The uncertainty in the magnetic field of the present instrument introduces a possible systematic error of about 0.05%. Measured values of the energies of a number of neutron capture γ-rays are compared with the values derived from the energy balance in (d, p) reactions. The results are in good agreement (within 0.1%).The relative counting efficiency as a function of energy is calculated theoretically and found to be in good agreement with experiment from 2.75 to 7.4 Mev. Above 7.4 Mev. no good method of checking the theoretical efficiency has been found. The absolute value of the counting efficiency at 2.75 and 7.38 Mev. has been measured by an ionization chamber method and is found to be in rough agreement with theoretical estimates.