characterizing rumination patterns of dairy cows using ...

Characterizing rumination patterns of dairy cows using spectral analysis. K A Beauchemin, R G Kachanoski, G B Schaalje and J G Buchanan-Smith J ANIM SCI 1990, 68:3163-3170.

The online version of this article, along with updated information and services, is located on the World Wide Web at: http://jas.fass.org/content/68/10/3163

www.asas.org

Downloaded from jas.fass.org by guest on April 16, 2012

CHARACTERIZING RUMINATION PAlTERNS OF DAIRY COWS USING SPECTRAL ANALYSIS K. A. Beauchemin*, R. G. Kachanoski3, G. B. SchaaljG and J. G. Buchanan-Smith4 Research Station, Agriculture Canada, Lethbridge, Alberta T1J 4B1 and University of Guelph, Guelph, Ontario NlG 2W6 ABSTRACT

Spectral analysis techniques were used to characterize the cyclical variation in rumination behavior of cows. Four Holstein cows were fed twice daily a diet of 60% highmoisture shelled corn-based concentrate, 15% firstcut alfalfa-grass hay and 25% secondcut alfalfa silage. The number of minutes that each cow spent ruminating was determined for 15-min intervals during six consecutive days. Rumination data then were characterized using Fourier harmonic analysis to decompose the total s u m of squares into 288 orthogonal components due to different rumination wavelengths. Rumination patterns for all cows consisted mainly of wavelengths that were harmonics of a 24-h cycle, indicating a circadian pattern of rumination. Differences in rumination patterns between cows occurred mainly at wavelengths of less than 2 h. Rumination patterns of two of the four cows were more complex, and consisted of high-frequency, non-24-h harmonic wavelengths in addition to the circadian pattern. Spectral analysis can be used to identify the component cycles of rumination patterns of individual animals, which can then be used to determine the effects of dietary or other manipulations on rumination behavior. (Key Words: Rumination, Time Series, Ingestion, Behavior.) J. Anim. Sci. 1990. 68:3163-3170

lntroductlon

The cyclical pattern of rumination activity by cattle and sheep is well documented. The greatest portion of rumination occurs at night, although animals also display a distinct period of rumination during the day (Castle et al., 1950; Gordon, 1958; Pearce, 1965; Gordon and McAllister, 1970; Gwffioy, 1974; Woodford and Murphy, 1988). Examination of the circadian pattern of rumination of animals in relation to dietary manipulations would be advantageous. One

‘LRS Contributionno. 3878925. 2 A ~ c n l h u eCanada, Res. Sta., P. 0. Box 3000, Main, Lethbridge. Alberta T1J 4B 1. %pt. of Land Resource Sci., Univ. of Guelph. ‘ b p t . of Anim. and Pod. Sci., Univ. of Guelph. Received July 24,1989. Accepted March 15,1990.

problem with doing this, however, is the question of how to summarize and compare long series (often hundreds) of rumination measurements. Murphy et al. (1983) characterized rumination patterns of sheep and goats by a least squares fitting of rumination time (min/ h) to a cosine equation. However, this ascribes all rumination activity to a single wavelength and a simple sinusoidal shape. Although the daily rumination profile appears to be cyclical, it may consist of other wave components of differing frequencies. Spectral analysis techniques have been used widely to examine cyclic variation in many types of data (Webster, 1977; McBratney and Webster, 1981; Grant et aL, 1985; Stroup et al., 1987); they may prove useful to characterize and summarize the cyclical pattern of rumination behavior. Spectral analysis refers to decomposition of the total s u m of squares of a time series into components due to periodicities at each of several discrete frequencies, as

3163

Downloaded from jas.fass.org by guest on April 16, 2012

3164

BEAUCHEMIN ET AL.

well as the interpretation of these components. 0730 and 1500, the refused concentrate was This analysis is done, using least squares, by removed and the alfalfa silage was fed. At this fitting a s u m of sine and cosine functions of varying amplitudes and wavelengths to the time series data. The objective of this work was to characterize rumination behavior of meal-fed dauy cattle using spectral analysis. Materials and Methods

Rumination data from a study reported by Beauchemin and BuchananSmith (1990) were used. Eating and ruminating behavior of four lactating Holstein cows were m r d e d using transducers consisting of two strain gauges5 bonded to a strip of spring steel6 formed into a concave shape (Beauchemin et al., 1989). These gauges, covered with a protective coating and attached to a leather halter, were used to monitor jaw movements. The transducers were linked to a computerized data acquisition system consisting of a 12-bit analog to digital converter7 installed in an lBM personal computer. An algorithm was used to process the digitized signal into discrete chews. Chewing during eating was distinguished from chewing during rumination using a scoring system based on the duration of chewing activity, the number and standard deviation of chews per second, and the duration of the preceding and succeeding pauses. The number of minutes that individual cows spent eating and ruminating during each 15-min interval was recorded for six consecutive days (a total of 576 m r d e d intervals). This sampling period was chosen to provide sufficient replication to establish diurnal rumination patterns. Cows received a diet composed of 60% high-moisture (67.6% DM) shelled corn-based concentrate (C), 15% first-cut alfalfa-grass hay (H), and 25% secondat alfalfa silage (S). Ingredients were fed separately in two equal portions M y . Hay was either offered before grain (diet H-C-S) or blended together and offered with silage (diet C-S+H). All cows were offered concentrate at 0600 and 1330. At

time, cows fed C-S+H also received hay; hay was coarsely (approximate length of 5 cm) chopped with a bedding c h o p $ , which facilitated hand-mixing with silage. Silage was available until the subsequent feeding of hay (€3-CS) or of concentrate (C-S+H). Feed refused from the afternoon feeding was offered during the am. feeding in addition to regular portions. Refusals of concentrate, silage and hay at the start of the subsequent p.m. feeding were weighed and used to calculate daily feed intake. Composition of diets and ingredients is given in Table 1. Cows were fed at 90% of their individual ad libitum feed intake, which was determined prior to the experiment using a completely mixed diet. Cows were housed in tie stalls in a barn artificially lit between 0500 and 1700. Spectral Analysis Techniques. Cattle dnate in bouts of activity that last from 30 s to 2 h (Ruckebusch, 1988). When monitored for several days, rumination appears to exhibit a degree of regularity in pattern. These patterns can be complex, consisting of many overlying patterns that mask regular patterns. For example, animals can ruminate heavily at one time of day, demonstrating a circadian pattern corresponding to a 24-h wavelength; yet, rumination also may occur during the day at regular intervals. Spectral analysis can be used to decompose the complex rumination patterns into individual cyclic components. As variation in time spent ruminating per unit of time is an indication of rumination activity, decomposing this variation into components due to various periodicities should be very useful in studying rumination behavior. In our experiment, rumination data for each cow were characterized using Fourier harmonic analysis (Bloomfield, 1976). This enables a variable measured over a finite period of time, at discrete and equally spaced intervals, to be exactly represented by an orthogonal finite Fourier series (Kachanoski et al., 1985). For rumination time per 15-min interval, the series is given by the following equation:

5EA-06-lZAD-lZ0, Intertechnology Ltd., Toronto, Ontario. %2 cm in length and 5 cm in width.

7FC Mate Lab Master, Techmar Inc., OH. 'Chop'n Bed, Model E,Leo Equipment, St. & G Quebec.

Downloaded from jas.fass.org by guest on April 16, 2012

3165

SPECTRAL. ANALYSIS OF RUMINATION PA-S TABLE 1. COMPOSITION OF DIETS AND INGREDIENTS Ingredients

Alfalfa silage

High-moisture

Soybean

Item Diet composition, %abc Nutrient content Dry matter, % Estimated NE],M u g d

COm

meal

55.3

4.7

25.0

Alfalfapass hay 15.0

67.6 2.13

89.4 1.94

37.2 1.37

95.7 1.38

Crude protein Neutral detergent fiber Acid detergent fibex

11.3 12.9 2.8

58.4 10.1 5.2

% DM basis

20.8 44.4 37.8

15.8 56.5 37.7

8Composition(% dry matter basis) for diets H-C-S, hay (H)allocated prior to concentrate (C) followed by silage (S), and C-S+H, hay blended with silage and offered after concentrates. b i e t s were supplemented daily with 200 g of minerals providing 115 g salt, 14.4g P, 2.9 g Mg, .27 g Mn, .52 g Zn,520 mg Cu, 72,400 IU vitamin A, 18,600 IU vitamin D and 130 IU vitamin E. 'Diets were formulated to provide 17% acid detergent fiber and 16%crude protein (drymatter basis). %I = net energy for lactation.

where Z, = number of rumination minutes in the j* 15-min interval during the 6 d monitoring period; j = 1, 2 ..., N, Ak, Bk = Fourier series coefficients; = mean value of Z N = total number of 15-min intervals (576), N/2 if N is even; and m = (N-1)/2 if N is odd. Fourier series coefficients were calculated directly, as follows:

separate proportion of the total sum of squares for rumination activity. To determine whether there were consistent cyclic patterns in the rumination times, a white noise test based on Fisher's Kappa (Fuller, 1976) was carried out for all of the 288 frequencies for each animal. This test is helpful in determining whether differences in periodogram values are due to random error or are the result of actual periodicities. The white noise indicated that there were significant periodicities, and it was hypothesized that the significance was due to a circadian rhythm only. To test t h i s hypothesis, the white noise test was applied to the periodogram values for each animal excluding the 48 values corresponding to a 24-h cycle and all its harmonics. The first (k = 1) sine and cosine harmonic If the latter test indicated that there still were pair has a cycling frequency equal to 1/N &e., significant periodicities among the non-circaperiod equal to the length of the sample). The dim frequencies, the maximal periodogram last (k = N/2) sine and cosine harmonic pairs values were removed one at a time from the have a cycling frequency equal to 112 (Le., 1/2 set and the white noise test was recomputed cycle per 15-min interval or each period equal until the test no longer was significant at the to twice the sampling interval). The sum of 5% level. Frequencies corresponding to the squares associated with the kth harmonic pair periodogram values that were removed were is called the periodogram value or intensity recorded, and the sum of the remaining periodogram values was divided by the approand is given by this equation: priate degrees of freedom to give the error mean square for the series. Ik = N/2(AZ+BZ). Finally, the 48 periodogram values corresponding to a 24-h cycle were analyzed. Each periodogram value has two degrees of Because the periodogram value is equivalent to freedom, except the last, which has one. an ANOVA sum of squares, F-tests were Because the Fourier harmonics are orthogonal, applied to the periodogram values. The perioeach of them (288 in this case) accounts for a dogram values were divided by 2 to get the

Downloaded from jas.fass.org by guest on April 16, 2012

3166

BEAUCHEMIN ET AL.

TABLE 2.MEAN "TINTAKE! AND CHEWING ACTIWTES OF COWS DURING A 6-DAY INTERVAL cow ID Item

1096

1210

1158

1159

Diet receiveda Dry matfa intake, kgld

H-C-S

H-C-S

C-S+H

C-S+H

16.80

(Wb

19.14 (23)

16.73 (.20)

14.73 (e181

4.33 (.17) 224.7 (31.1) 3 16.6 (20.3)

3.82 (.14) 207.9 (28.7) 362.1 (27.6)

Neutral detergent fiber intake, kg/d Eating time, mi4d

R

'

.' gtime,min/d

4.34 (J7) 134.5 (17.2) 284.2 (25.1)

4.97 315.7 (42.9) 355.8 (43.0)

%-C-S =hay given prior to concentrate followed by silage; C-S+H = hay blended with silage and offered after concentrates. bStandard deviations given in paentheses (number of observations= 6).

numerators of the F- statistics and the denominator was the error mean square mentioned in the preceding paragraph. The P-value of an individual test had to be less than . W 2 in order to be declared significant so that, by Bonferroni's formda, the overall rate of Type I errors would be less than .20 for each animal (Hochberg and Tamhane, 1987). Thus, in summary, a set of frequencies associated with significant periodmities was obtained for each animal. The non-circadian frequencies were identified by the sequential use of white noise tests and the circadian frequencies were identified by use of ANOVA F-tests. Fourier series coefficients associated with the significant frequencies were used to compute the amplitude and the phase angle for each of the component periodicities and to give predicted rumination times for each time interval.

this period ranged between 14.73 and 19.14 kdd for the four cows vable 2). On the average, 84.2% of the total eating activity occurred between 0500 and 1700, which may be related to our feeding schedule. Time spent eating was variable and ranged from 134.5 to 315.7 min/d. Part of the discrepancy between eating times for individual cows is related to the inaccuracy of using jaw movements as an indication of eating time (Beauchemin et al., 1989). Only partial distinction between mastication of feed and grooming, licking and drinking activities is possible for recorded jaw movements. Cows on both diets ruminated during the day between feeding times as well as during evening and early morning hours, confirming patterns observed by Woodford and Murphy (1988) for cows fed twice daily. However, only 36.9% of total rumination activity ocm e d between 0500 and 1700. Periodogram values were plotted against Results and Discussion wavelength (period) for each cow in Figure 2. Figure 1 shows eating and ruminating A number of significant wavelengths were patterns of each cow during the 6 d monitoring identified for each cow (Table 3). For cows interval. Both eating and ruminating activity 1096 and 1210, sigmficant wavelengths were displayed a strong degree of regularity in all harmonics of a circadian cycle. In contrast, pattern during this period. cows 1158 and 1159 had more complex Eating patterns were similar for all cows, rumination patterns consisting of hmonics of although meals started sooner for cows fed diet a circadian cycle, as well as non-circadian H-C-S (cows 1096 and 1210) because of the wavelengths in the high frequency range. distribution of long hay before concentrate. Significant spectral peaks at 24-h Peak time of eating coincided with the twice wavelengths indicated a pattern of rumination daily feeding pattern, although some eating that was repetitive from day to day for all also occurred randomly throughout the cows. Others also have reported a circadian 6 d period. Average dry matter intake during rhythm of rumination (Murphy et al., 1983).

Downloaded from jas.fass.org by guest on April 16, 2012

SPECTRAL ANALYSIS OF RUMINATION PAlTERNS

3167

E1

60

i

0 12 0 12 0 12 0 12 0 12 0 12 I Day 1 Day 2 I Day 3 j Day 4 Day 5 Day 6 1

12

,

~

I

~

I

/

Time of day Figure 1. Time spent eating (E)and ruminating (R) each hour (12 = nooo; 0 = midnight) by individual cows during a 6 4 interval. Cows 1096 and 1210 received bay (H) before concentrate (C)followed by silage (S) (diet H-C-S), and cows 1159 and 1158 received hay blended with silage and offered after concentrate (diet C-S+H).

Although the circadian basis of rumination milking schedules and patterns of lighting each may be partially derived from a biological day (Metz, 1975). clock, undoubtedly it is also influenced by All cows, except 1096, exhibited a si@environmental factors such as feeding and cant 12-h rumination behavior pattern. Rumi-

Downloaded from jas.fass.org by guest on April 16, 2012

3168

BEAUC3IEMIN ET AL.

TABLE 3. DESCRIPTIONOF SIGNIFICMeanruminati04 15-min interval,

cow

min

Diet H-C-Sc 1096

2.91

1210

Diet C-S+H? 1158

3.69

3.75

Wavelengthb. h

Cycles per Mintnval

24 8 1.85 1.41 1.20 24 12 8 1.6

6 18 78 102 120 6 12 18 90

24 12

6 12 18 78 95 104 112 115 120 138 147 6 12 18 107 128 133 134 144

8

1159

3.25

1.85 1.52* 1.38* 1.29* 1.25* 1.20 1.04 .98* 24 12 8 1.34* 1.13* 1.08* 1.07* 1.o

WAVELENGTHS

Periodogram, min/15min2 314.0

448.O 304.0 281.8 282.8 325.6 540.4 8902 360.4 1059.0 262.4 873.4 249.2 358.4 535.8 358.8 441.0 616.8 269.2 369.0 570.6 1202.8 336.6 377.6 535.8 460.0 633.6 263.4

Amplitude, h

Phase h

1.044 1.249 1.031 1.008 .990 1.M3 1.370 1.758 1.119

3.606 1.358 243 .224 .030 1.554 1.454 1.406 .212

1.919 .953 1.740 .934 1.110 1.368 1.107 1.240 1.452 .774 1.124 1.401 2.051 1.087 1200 1.379 1256 1.474

1.388 .498 .557 .328 .057 .235 249 .016 .162 .039

.960

.028

1.299 .185 .704 .286 .024

.2a6 .085 .234

vests for non-circadian wavelengths were based on Fisher’sKappa. Tests for circadian wavelengths used Bonfemni’s procedure and F-statistics. ~ A Uwavelengths are lmmonics of 24 h, except boose followed ty an asterisk. ‘H-C-S =hay given prior to concentrate followed by silage; C-S+H= hay blended with silage and offered after concentrates.

nation patterns shown in Figure 1 indicate that these three cows displayed a prominent diurnal-nocturnal pattern of rumination over the 6 d period. Although rumination peaked between meals and again at night approximately 12 h later, the specific peak times varied somewhat between cows, indicating some individuality in behavior. In contrast, the pattern of rumination exhibited by cow 1096 differed somewhat from that of the other cows; the period of maximum rumination activity each day for this cow varied from day to day. All cows also displayed significant 8-h rumination wavelengths. This probably was a reflection of the feeding schedule of two main meals, approximately 8 h apart. As expected, no rumination occurred consistently during these times each day.

In agreement with observations by Metz (1975), all cows exhibited some significant short-term rumination behavior, as indicated by significant short-term spectral peaks. However, the number of significant short-term wavelengths was greater for cows 1158 and 1159 than for cows 1096 and 1210, mainly due to noncircadian wavelengths in this high frequency range. The significance of regular short rumination periods is unclear, but may, in fact, be beneficial to the dairy cow. Occurrence of nunination periods immediately after meals or between eating periods may enhance the animal’s ability to buffer products of fermentation produced subsequent to eating, because salivary output increases during rumination (Seth et al., 1974). This could minimize

Downloaded from jas.fass.org by guest on April 16, 2012

SPECTRAL ANALYSIS OF RUMINATION PAlTP.RNS

3269

1250 - Cow 1096 1000

-

750

-

500

-

1250 - c o w 1210 1000 -

750

-

250 0

:

.. . . . .. ..:... . . ..*-... . .='. - ." '. *

i

I

144 7 2 4 8

.

7 -

24

I

12

-!-*#-

a

4

2

1

05

Wavelength (h) figure 2. Periodograms (min/15min)2 for cows during a 6-d interval (meanperiodogram represented by dotted line).

fluctuations in ruminal fluid pH that are detrimental to cellulolytic bacteria (Kaufmann et al., 1980). Because rumination aids in particle size reduction and clearance of undigested feed from the reticulorumen (Ulyatt et al., 1986), regular frequent periods of rumination may help to maintain a continuous flow of digesta from the reticulorumen, alleviate distention of the foregut and promote higher feed consumption. The amplitude and phase of the significant wavelengths listed in Table 3 were used to compute predicted d a t i o n patterns for individual cows. These predictions are compared in Figure 3 with rumination patterns actually recorded during the 6-h interval. The R2 values for the predictions were all low (.11

45

I

12 0 12 0 12 0 12 0 12 0 12 0 12 :Day 1iDay 2iDay 3iDay 4iDay 5/Day 6

Time of day Figure 3. Actual (A) and predicted (E)' rumination times (per hour) for individual cows during a 6-d interval (12 = noon; 0 = midnight).

to .30). Accuracy of predictions could be

enhanced by considering a greater number of wavelengths; however, the intent of this work was not prediction, but rather decomposition of patterns into significant components in order to characterize the dominant d a t i o n pattern and facilitate comparison of patterns between cows. The non-circadian wavelengths resulted in predicted rumination patterns for cows 1158 and 1159, which differed slightly from day to day. In contrast, predicted rumination patterns for cows 1096 and 1210 were similar from day to day because all wavelengths were harmonics of a 24-h cycle. The amplitude of predicted rumination activity occurring during the day between feeding times was less than that

Downloaded from jas.fass.org by guest on April 16, 2012

3170

BEAUCHEMIN ET AL.

chevre et le mouton. Ann. Zootech. (Paris) 23:63. Gordon, J. G. 1958. The effect of time of feeding upon rumination. J. Agric. Sci. (Camb.) 51:81. Gordon, J. G. and I. K. McAllister. 1970. The circadian rhythm of rumination. J. Agric. Sci. (Camb.) 74291. Grant, C. D., B. D. Kay, P. H.Groenevelt,G. E. Kidd and G. W. Thurtell. 1985. Spectral analysis of micropenetrometer data to characterize soil structure. Can. J. Soil Sci. 65:789. Hochberg,Y. and A. C. Tamhane, 1987. Multiple Comparison Procedures. John Wiley & Sons, New York. Kachanoski, R. G., D. E. Rolston and E. ddong. 1985. Spatial and spectral relationshipsof soil properties and microtopography: I. Density and thickness of a horizon. Soil Sci. Soc. Am. J. 49:804. Kaufmans W., H. Haganeister and G. Duben. 1980. Adaptation to changes in dietary composition, level and frequency of feeding. In: Y. Ruckebusch and P. implications Thivend @ Digestive I.) Physiology and Metabolism in Ruminants. pp 587-602. AVI Publishing Co. Spectral analysis can be used for describing Westport, CT. cyclical variation in rumination behavior of McBratney, A. B. and R. Webster. 1981. Detection of ridge and furrow pattern by spectral analysis of crop yield. cows. Complex rumination patterns can be Int. Stat. Rev. 4945. decomposed into component cycles; the signifMetz, JH.M. 1975. Time patterns of feedmg and rumhation icant component cycles then can be used to in domestic cattle. Meded. Landbouwhogesch. Wagmake comparisons between various patterns. 75-12. 274 pp. Cyclic variation exists in rumination behavior, Murphy, M. R, R. L. Bald- M J. Ulyan and L. I. Koong. 1983. A quautitativeanalysisof Nmination patterns. J. wavelengths reflect primarily a circadian h i m . Sci. 561236. rhythm partially derived from environmental Peace, G. R. 1%5. Rumination in sheep. III. Experimentcues. Non-circadian, short-term patterns also ally induced variations in the circadian pattern of were observed, although their significance is rumination. Aust. J. Agric. Res. 16549. unclear. Spectral analysis could be used to Ruckebusch, Y. M. 1988. Motility of the gastro-intestinal tract. In: D. C. Church (Ed.) The Ruminant Animaldetermine the extent to which rumination Digestive Physiology and Nutrition. pp 64-107. patterns are unique for individual cows, and Rentice-Hall, Englewood Cliffs. NJ. whether these patterns can be altered by Seth, 0. N., G. S. Rai, P. C. Yadav, M. D. Pandry and J. S. Rawat. 1974. Effect of diet and rumination on the rate changing feeding stimuli such as diet composiof secretion and chemical composition of parotid tion, time of feeding or other factors. saliva of Bubalus bubalis and Bos indicus. Indian J. Anim. Sci. 44:717. Stroup, W. W., M. K. Nielsen and J. A. Gosey. 1987. Cyclic variation in cattle feed intake data: characterization Literature Cited and implicationsfor experimentaJdesign. J. Anim. Sci. Beauchemin,K. A. and J. G. Buchanan-Smith. 1990. Effects W1638. of fiber source and method of feeding on chewing Ulyatt, M J., D. W. Dellow, A. John, C.S.W. Reid and G. C. activities, digestive function, and productivity of diary Waghom. 1986. Contributionof chewiugduring eating COWS. J. Dairy Sci. 73:749. and rumination to the clearance of digesta from the BaUChemin,K. A., S. zelin, D. Germcr and J. G. B~chananmminoretidum.ln: L. P. Milligan, W. L. GroMmand A. Dobson (Ed.) Controlof Digestion and Metabolism Smith. 1989. An automatic system for quantifying eating and rUmiaatingactivities of dairy cattle housed in Ruminants. pp 498-515. Prentice-Hall, Englewood in stalls. J. Dairy Sci. 722746. cliffs, NJ. Bloomfield, P. 1976. Fourier Analysis of Time Series: An Webster, R 1977. Spectral analysis of gilgai soil. Aust. J. Introduction. John Wdey & Sons, New York. Soil Res. 19191. Castle, M. E.,A. S. Foot and R J. Halley. 1950. Some Woodford, S. T.and M. R. Murphy. 1988. Effect of forage physical form on chewmg activity, dry matter intake, observations on the behviour of dairy cattle with pdcul81 reference to grazing. J. Dairy Res. 17215. and rumen function of dairy COWS in early lactation J. Fullex, W. A. 1976. Introduction to Statistical Time Series. Dairy Sci 71:674. John Wiley & Sons, New York. Zerbe, G. 0.and R H. Jones. 1980. On applicationof growth Geoffrey, F. 1974. Etude compark du comportement curve techniques to time series data. J. Am. Stat. alimentaire et merycique de deux petits ruminants: la Assoc. 75507.

occurring at night for three cows, but not for cow 1210 (Figure 3). The extent to which rumination patterns are unique for individual cows, and whether these patterns are affected by dietary treatment, needs to be determined using a greater number of cows fed a wider range of diets than used in the present study. This could be done by transforming the data to discrete Fourier coefficients, selecting a set of potentially important frequencies, and then comparing the Fourier coefficients for these frequencies using a multivariate approach as described by Zerbe and Jones (1980).

Downloaded from jas.fass.org by guest on April 16, 2012

Citations

This article has been cited by 1 HighWire-hosted articles: http://jas.fass.org/content/68/10/3163#otherarticles

Downloaded from jas.fass.org by guest on April 16, 2012