Measurements of ellipticity of Rayleigh waves recorded in the U.S. Midwest have been examined for azimuth dependence, effects of interference, and repeatability, as well as the hypothesis that a single station may be used to determine local structure. Time- and frequency-domain analyses were performed for each event, with more consistent results from the time-domain method. Results indicate that for the period range of 10 to 50 sec, ellipticity depends primarily upon local structure and does not exhibit significant azimuthal dependence. Most ellipticity values for a given period are repeatable within 5 per cent of other measured values from all source regions, with the greatest deviation being about 10 per cent. The cause of the deviations is attributed to interfering waves and/or poor signal-to-noise ratios. Interference effects result in scatter in ellipticity values. An ellipticity peak in the period range of 18 to 22 sec has variable magnitude for different events, depending upon the amount of interference present and the signal-to-noise ratio. Interference effects also manifest themselves as sharp decreases in group-velocity observations even after filtering. Model studies show that ellipticity peaks can exist, which are due to the layered structure and not necessarily to interference effects. Ellipticity measurements (10- to 50-sec-period range) from a single station are useful for determination of a crustal model for the vicinity of the recording station, but should be used in conjunction with other available geophysical and geological data. Ellipticity measurements are shown to be of special value for model determination in areas with sedimentary layering, a result in agreement with the Boore-Toksöz (1969) study.