In fisheries acoustics, the standard sphere method of echosounder calibration is most commonly used (Johannesson and Mitson, 1983; Simmonds et al., 1984). To investigate the uncertainty in the method, a series of experiments were conducted in a deep tank (Demer and Hewitt, 1993). Results indicated that variations in pulse length (0.1 and 0.3 ms), water temperature (0.5 to 5.5 C), and choice of standard sphere (tungsten carbide (WC) - 33.2 mm or copper (Cu) - 30.5 mm), could cause corresponding variations in the system gains of 0.3, 0.2 and 1.5 dB, respectively. Additionally, system gain values calibrated with the 30.5 mm Cu sphere were consistently lower than those with the 33.2 mm WC sphere. Prompted by the latter result, direct measurements of target strength (TS) were made of four standard spheres (WC - 33.2 and 38.1 mm, Cu - 23.0 and 30.5 mm) in an anechoic tank. A calibrated hydrophone was used to measure the incident and reflected intensities of the pulse and the measured TS values were compared to their theoretical counterparts. The TS measurements of the 30.5 and 23.0 mm Cu spheres and of the 38.1 mm WC sphere were all about 1.5 dB larger than their theoretical values. However, the 33.2 mm WC sphere exhibited a larger difference of 2.5 dB. A system calibration by the method of self-reciprocity supported the magnitudes of these observed differences. Concluding from the results of these two sets of experiments, system calibration with an optimal standard sphere is estimated to be accurate to -1.2 dB, with a precision of 0.3 dB.