Predicting the frequency of roadside encroachments is a necessary first step to analyzing the costs and benefits of a proposed roadside improvement. The results of the classic Hutchinson and Kennedy study, which analyzed encroachments onto the median of divided highways, have been used for this purpose. That study asserted that encroachment frequencies have a nonlinear dependence on average daily traffic (ADT) such that there is a decrease in encroachment rate for roadways carrying between 4,000 and 5,000 vehicles per day. A review of the data from that study has revealed that the ADT counts were confounded with time, over which the observed encroachment rate varied. With the use of statistical techniques not available to that study's authors, it was shown here that the assertion of ADT dependence is not well supported and that the observed changes in encroachment frequencies can also be explained by a simple model consisting of a time trend and an effect due to winter weather. The resulting recommendation is that only the portion of Hutchinson and Kennedy's study yielding a relatively constant encroachment rate be used in future analyses of roadside safety improvements.