Lying behavior in dairy cattle can provide insight into how cows interact with their environment and can serve as a useful indicator of animal welfare. We tested the accuracy of the Onset Pendent G data logger for measuring lying behavior in dairy cattle (n = 12 cows in each of 2 experiments). Loggers sampled position at 4 intervals (6-s, 30-s, 60-s, and 300-s), and data were smoothed using 3 filtering methods, using an Excel macro to convert readings (i.e., lying bordered by standing or vice-versa) occurring singly, or in runs of 2 or less, or 6 or less. Cows wore the logger for at least 1.5 d on the lateral side (Experiment 1) or the medial side (Experiment 2) of the hind leg above the metatarsophalangeal joint and were simultaneously video recorded. The logger accurately measured lying behavior, including which side the cow was lying on: for example, predictability, sensitivity, and specificity were > 99% using 30-s sampling and the single event filter compared to continuously scored video. The 6-, 30- and 60-s sampling intervals were comparable for all aspects of lying behavior when using the 6, 1 and 1 event filters, respectively. Estimates of lying time generated with the 300-s, 1-event filter were highly correlated (r>0.99) with estimates from video, but the number of lying bouts was underestimated by 6.2 and 23.0% in Experiments 1 and 2, likely because short lying bouts were missed (on average, 1.5 and 23.8% bouts less than 300 s in Experiment 1 and 2, respectively). In summary, the data logger accurately measured lying behavior when the sampling interval was ≤60 s and single readings indicating lying or standing were filtered from the data set.