Accurate estimation of daily grazing time and nycthemeral grazing pattern in ruminant nutrition research may result in a better understanding of sward–animal interactions and of the behavioural constraints affecting daily intake at pasture. Portable devices which automatically record grazing activities are needed. The Kenz Lifecorder Plus device (LCP, Suzuken Co. Ltd., Nagoya, Japan), originally developed for measuring human daily activity level, has been used recently to record daily grazing time and pattern in dairy cows. The aim of this validation study was to compare grazing activity duration as recorded simultaneously by the Lifecorder Plus fitted on the cow neck and by trained observers. A total of 250 h of visual observations on six grazing cows over 12 days were performed. The relative prediction error averaged 0.12, 0.07, 0.05, and 0.03 of actual grazing time when accuracy was investigated per hour, per sequence of observation (approx. 5 h), per day (approx. 21 h), and per cow (approx. 41 h), respectively. Error was mainly random, the proportion of mean square prediction error related to mean bias or line bias being always very small. This high accuracy shows no overall confusion between grazing and other activities, such as rumination, when cows were at pasture. No activity was detected when cows were ruminating. Walking time to and from the milking parlour was detected as activity due to much greater speed than while grazing. This can be easily filtered out when the raw data are processed so as not to over-estimate grazing time, particularly if some cows are fitted additionnally with Lifecorder on the leg to identify these periods. In conclusion, the high accuracy of the Lifecorder Plus to detect grazing activities when cows are at pasture, particularly at the day scale or when comparing cows over several days, as well as its robustness and simplicity in use, make it a very suitable device for recording the grazing time and pattern of dairy cows at pasture.