Individualized monitoring of rumination time (RT) could be used to guide nutritional recommendations for improved production, health and well-being of dairy cows. One commercial application is the QWES-HR Tag system (HR Tags), which uses a microphone and microprocessor in neck collars for sound-based detection of RT; filtered sounds in 12 × 2 h intervals are automatically integrated for estimation of 24 h moving RT, but inaccurate detection or processing of sounds could significantly affect estimations. A study was conducted to compare the automatic detection of RT by HR Tags against observed values obtained by supervised inspection of 24-h soundtracks. The RT of 5 lactating Holstein cows on ryegrass/white clover or orchardgrass/white clover pastures was monitored for 6 non-consecutive days, using 2 sound-based methods (1) HR Tags and, (2) Acoustic Halters, which included a digital recorder and one directional microphone pressed to the animal’s forehead (Control). Five halters were used with halters rotated across cows according to a completely randomized design. Cross-validation of HR Tags (24-h and 2-h periods) was conducted by inspection of the root of the mean squared prediction error (RMSPE) and concordance correlation coefficient (CCC) between methods. The HR Tags consistently underestimated RT by 18%, regardless of the period of measurement (Table 1). Similarly, high RMSPE and low CCC indicated limited accuracy and low precision by HR Tags, which were slightly improved for 24 h evaluations (Table 1). Further research is needed to improve the automatic sound-based detection of rumination by commercial HR Tags. Table 1. Estimation of 24-h and 2-h rumination time by 2 acoustic methods

Rumination time/measurement period		Control	HR Tags
24 hour (n = 29)	Mean (min)	534	437
	SE (min)	19.2	25.4
	RMSPE (min)	90.7
	CCC (%)	36.6
2 hour (n = 349)	Mean (min)	44.3	36.3
	SE (min)	1.7	1.2
	RMSPE (min)	30.6
	CCC (%)	23.8