In 1990, the South Dakota Department of Transportation (SDDOT) noted significant discrepancies between its ride-quality measurements and those taken by a contractor paving a portland cement concrete project. The contractor's measurements were consistently smoother than SDDOT's and would have generated incentive payments approximately twice as large. About half of the observed difference could be attributed to increased pavement roughness after paving, but the rest appeared to result from differences between the department's manual profilograph and the contractor's computerized unit. Analysis revealed that a numerical filtering algorithm used by the computerized profilograph strongly attenuated profile features with wavelengths shorter than 10 ft. Such attenuation was observed directly on the computerized unit's profile traces. Because of the attenuation, SDDOT considered the computerized measurements unsuitable calculating incentive payments. However, SDDOT could not use its own measurements as a basis for payment because they were not taken within the specified 48-hr period after paving. To estimate a fair incentive payment, SDDOT developed a correlation between the computerized and manually interpreted profile indexes for the project. Using the correlation, SDDOT awarded an incentive payment approximately midway between its original estimate and the contractor's. SDDOT has suspended use of computerized profilographs pending improvement of the filtering algorithm. Preliminary experiments indicate that although the computerized profilograph's first-order filter attenuates profiles too strongly and produces artificially low profile indexes, a third-order filter might generate higher profile indexes than does a manual interpreter. This suggests that a second-order filter might best approximate a human's visual interpretation of the profile. Further research is needed to confirm this hypothesis and to establish a foundation for standard filtering procedures.