Discriminant analysis was used to identify combinations of variables, chara
cterizing patterns of jaw movement (JM) during eating and rumination, that
minimize errors when classifying series of successive JM of knows behaviour
into rumination and eating. JM data from a non-lactating jersey cow were r
ecorded continuously for 72 h. The cow was fed 1800 g of barley straw and 1
375 g of dried sugarbeet pulp pellets twice daily. Average times spent eati
ng and ruminating daily were 294 and 462 min. Successive rumination and eat
ing JM were grouped into series when pauses between JM were equal to or lon
ger than a fixed time interval corresponding to the minimum time interval b
etween successive rumination cycles. Time intervals between successive JM,
amplitude (the magnitude of the jaw's movement) and number of chews per ser
ies of successive JM were calculated for all rumination and eating series a
nd included in the discriminant analyses. The combination of the four varia
bles, basic chewing rate (BCR), mean of amplitude/standard deviation of amp
litude, ln(number of chews) and standard deviation of time intervals betwee
n successive jaw movements, gave the least minutes of misclassified chewing
behaviour per day (3%, of total daily chewing time) and yielded a quadrati
c discriminant function. The order of the variables mentioned rank the vari
ables according to their ability to classify the series. Combinations of th
e variables BCR and ln(number of chews), and other variables based on mean
and standard deviation of time intervals between successive JM and amplitud
e of JM, can be used with advantage for classifying series of successive JM
into rumination and eating behaviour.