We carry out a detailed comparison between winds derived from F2 peak heights and winds obtained from incoherent scatter radar (ISR) line-of-sight velocity measurements. A total of 34 incoherent scatter radar experiments at Millstone Hill spanning all seasons and levels of solar activity are included in this study. For two experiments we compare results from five different wind-derivation techniques. According to work by Titheridge [1993, 1995a, b>, neutral winds derived from the servo model are inaccurate during the sunrise and morning period because of a shift in the zero-wind F peak downward from the balance height. To investigate this effect, we determine a correction factor cfac to be applied to the servo model c parameter as a function of time of day for summer, equinox, and winter at both solar maximum and solar minimum. Our results confirm that a sunrise decrease in cfac is necessary to bring about best agreement between the servo model winds and the winds derived from the ISR ion velocity data at Millstone Hill. However, the effect is not large, so that a constant cfac for each season/solar activity level usually introduces less error than other factors which may result in differences between the servo model and ISR winds. These factors include measurement errors in hmF2 and the ISR line-of-sight ion velocities, spatial variations in the wind field above the station, and the assumption that hmF2 is the peak in the O+ altitude profile.¿ 1997 American Geophysical Union