Taking into account the effects of a problem with bridge calibration and a missing data point improves the agreement between theory and experiment. Specifically, in Fig. 5(a) the three a-axis permittivity points immediately above, but not including, 1 MHz frequency are too low because the bridge (Hewlett-Packard 4191A RF Impedance Analyzer) was not automatically calibrated at these frequencies. Deleting these three points leaves a nearly fiat permittivity vs frequency curve, as expected because there is no piezoelectric effect for electric field along the a axis of this crystal.Using this result to interpret the c-axis permittivity data in Figs. 4(a) and 6, we assume that the points for the corresponding frequencies are low also. The 1 MHz c-axis point (which is at a calibrated frequency) was mistakenly omitted from these figures. Its permittivity value is near the permittivity values at higher frequencies for which the bridge was calibrated. This indicates that the piezoelectric resonance only becomes evident below 1 MHz.The 4191A's lowest frequency is 1 MHz. The other bridge used, the Wayne-Kerr Model 642S Precision Component Analyzer, covers the 20 Hz to 300 kHz range. At the upper end of this range, a sharp upturn in c-axis permittivity is seen in Figs. 4(a) and 6 as expected for a piezoelectric resonance. The rest of the resonance must occur between 300 kHz and 1 MHz, the frequency range inaccessible to us.One might attribute the permittivity upturn near 300 kHz to an artifact caused by distributed parameters, a problem which often occurs at the upper end of the kHz range for audio-range apparatus. However, the a-axis permittivity shows no such upturn, so we believe in the reality of the c-axis upturn. Also, the large change in c-axis permittivity level from the audio to the MHz range, and the lack of such change in a-axis permittivity level, gives us added confidence that the c-axis permittivity change with frequency can be attributed to piezoelectric resonance. The fit in Fig. 6 is improved by including the 1 MHz point and ignoring the three uncalibrated points immediately higher in frequency. Erratum: Inelastic-phonon-scattering effect on the behavior of the thermoelectric power of metals [Phys. Rev. B 49, 13 215 (1994)] K. Durczewski and M. AusloosSeveral typographical errors appear in this paper. The correct equations are given below. These changes do not affect the conclusions of the paper. u, z(E, E' -E, q ) =u2, (E, e'e, q) = r(eg)+(e'e)](q a) u2z(E, E'e, q)=[(eg) +2(E' -E)(eg)](q a) +(Ee') [(q a) +(k a) ] (6b) (6c) P+(t)=P+(t, T, E"iE~i)= (10) 2 ksT~2 (ksT) P,2(0) 2 ieF i 3 [E~fe, P"(0)