The aim of this study was to validate the recordings of a commercial triaxial accelerometer (HOBO Pendant G data loggers) with respect to visual observations of grazing and ruminating time of dairy cows. Seven lactating Holstein cows with a mean body weight of 602 ± 45 kg were used for the study. Grazing and ruminating times were recorded using HOBO loggers (31.5 hours of total observations for grazing and ruminating) that were attached to the lateral-medial side of the jaw using a strap attached to the head of each cow in a position such that the X-axis was parallel to the ground, the Y-axis was perpendicular to the ground pointing upward, and the Z-axis was parallel to the ground pointing away from the sagittal plane; these relative positions were defined when the cow was in a natural head-up position. Median acceleration (m/s2) readings in the X-axis >0.175 and <0.95 indicated grazing activity, whereas readings in the Z-axis >−0.275 and <0.0875 indicated ruminating activity. The degree of vertical tilt (° Y-axis) was used to determine grazing position (readings >0° and <61°) and ruminating (readings >0° and <25.8°). Results showed a significant (P < 0.001) relationship of estimated grazing time against visual observations when acceleration (X-axis) was used only; nonetheless acceleration (Z-axis) was not significant (P > 0.05) for estimating ruminating time. When acceleration and tilt were used for estimating grazing (X- and Y-axis) and ruminating (Z- and Y-axis) time, the slope in both showed a significant (P ≤ 0.001) relationship. The prediction R2 in both activities indicated that acceleration (X- and Z-axis) and tilt (Y-axis) of HOBO loggers explained 0.961 and 0.945 of the variance in visual observations per cow/day. Therefore, the validation of the HOBO loggers was successful on a per cow/day and per day basis.