In sensor array processing, high-resolution eigenbased direction-finding algorithms are known to be sensitive to sensor gain, phase, and mutual coupling errors. In order to approach the algorithms' theoretical direction-finding performance in actual arrays, methods are required to measure these errors and compensate for them. The authors develop calibration procedures based on least squares fitting of known measured direction vectors. After calibration, any appropriate direction-finding algorithm may be applied. Experimental results from an ultrasonic array testbed demonstrate the performance of the calibration algorithms. A previously proposed autocalibration algorithm is discussed. The algorithm is shown to give nonunique solutions for linear arrays.