Lameness remains one of the greatest welfare challenges for the dairy industry, but there is little consensus on how best lameness can be surveyed on farm. This longitudinal study evaluated how measures of lameness incidence change with sampling frequency. Weekly visits to 6 farms in British Columbia were performed between May and December 2017, and 93 animals without previous history of severe claw horn lesions were repeatedly locomotion scored from 8 weeks prepartum until calving. Lameness was scored using a 1 to 5 numerical rating; cows were considered lame when they scored 3 (mildly lame) on 2 consecutive visits, or ≥4 (severely lame) on one occasion. The weekly data were used to create 3 data sets representing 1) weekly assessment, 2) assessment every other week, and 3) assessment every third week. Incidence of lameness was 31% when using the complete data set (i.e., weekly scores), versus 19% and 11% for assessments every 2 and 3 weeks respectively. Farm incidence ranged between 18 and 50% (median 30%), 6–50% (23%) and 0–50% (6%) when assessment occurred every 1, 2 and 3 wk. Differences in farm incidence depending on assessment frequency was tested using paired sample t-tests. Compared with weekly visits, farm incidence was significantly lower when assessment occurred every 2 (t(5) = 3.5) or 3 weeks (t(5) = 4.2). Of the animals that became lame during the dry period, average time from the first visit until diagnosed as lame was 20.5, 24.1, and 25.1 d when scoring was performed weekly, every other week and every 3 weeks respectively. These results suggest that (1) a high proportion of the animals become lame during the dry period, (2) estimates of lameness incidence are dependent upon sampling frequency, with low sampling rates likely leading to underestimation of the number of animals becoming lame, and (3) frequent assessment shortens the time before lameness is diagnosed.