Negative play contagion in calves

Ethical considerations

This study was carried out in accordance with the guidelines for ethical treatment of animals of the International Society of Applied Ethology. It was approved by the Institutional Animal Care and Use Committee of the Institute of Animal Science and the Czech Central Committee for Protection of Animals, Ministry of Agriculture (Permit Number 27356/2016-MZE-17214). Calves on the low-milk schedule received a milk allowance of approx. 12% of their body weight, equalling the traditional calf feeding practices²⁶.

Animals and housing

The study was conducted at the Netluky research station of the Institute of Animal Science in Prague, Czech Republic. Data were collected from August 2016 until April 2017.

Seventy-two Holstein Friesian dairy calves (31 heifers and 41 bulls) were included in the study. Calves were separated from their dams at approximately 12 h after birth and housed individually either in outdoor hutches or in individual pens in a naturally ventilated open barn equipped with curtains. In both cases, the area available for each calf was 1.4 m × 1.4 m straw-bedded lying area and 1.2 m × 1.2 m solid walking area. While in individual housing, calves were fed 3 l of milk twice per day through teat-buckets at 06.00 and 18.00 and received concentrates and water ad libitum. Calves entered the experiment at an average age of 13.3 ± 3.1 days (mean ± S.D.) and were then housed in groups of three. Calves were allocated to groups balanced by sex, age and weight. Groups entered the experiment consecutively with 1–2 groups per week. Groups were housed in a naturally ventilated open barn with curtains. Group pens were 10.1 m² consisting of a straw-bedded lying area (4.2 m × 1.4 m; approx. 2.0 m² per calf) and a concrete walking and feeding area (3.5 m × 1.2 m). The group pens were covered with visual barriers in order to avoid direct visual contact of other calves. The visual barriers in front of the respective pens were removed in order to allow video recording; however, groups that were recorded simultaneously were allocated in the barn in a way that precluded visual contact without the front visual barriers of the pens. The calves received water, hay and concentrates ad libitum, offered in buckets and were provided with fresh straw bedding three times per week. All routine farm work was done before 10.00. Calves were hot-iron disbudded at 24.4 ± 3.1 days of age (mean ± S.D.). Disbudding wounds are painful for more than three weeks²⁷, however no difference in play behaviour after disbudding was found after 27 hours²³, thus we do not expect an effect of disbudding on play in our study. On the recording days, the air temperature in the barn ranged between -4.5 °C and 29 °C with the average (± S.D.) being 7.9 (± 8.6) °C.

Experimental design and procedures

Experimental design

Groups were allocated to treatments balanced by sex composition, age, weight and point of time entering the experiment. Milk allowance, group composition and number of groups assigned to each of the treatments are displayed in Fig. 1. Calves in all treatments received three milk meals per day at approximately 06.00, 12.00 and 18.00. All calves were offered 6 l of milk per day at the beginning of week 3. For UHigh and MHigh calves, the offered milk was gradually increased to 9 l of milk per day in week four and 12 l of milk per day in week six (Fig. 1). Therefore, the total milk amount offered from the start of the experiment until the end of week eight (42 days) was 240 l for ULow and MLow calves and 420 l for UHigh and MHigh calves. Milk was offered in teat buckets. Calves were tethered for the duration of the milk meal using neck collars and were released when all calves of the group had finished their meals (i.e. calves had either emptied the buckets or stopped drinking milk; approx. 5 min). If calves did not finish the offered milk meal, the volume of the remaining milk amount was measured. The volume of unconsumed milk was then summed from the point of entering the experiment until the respective day of behaviour recording. For ULow and MLow the total volume of unconsumed milk amounted to 0.3 ± 0.9 l (mean ± S.D.; median/interquartile range: 0/0 – 0). For UHigh and MHigh the total volume of unconsumed milk amounted to 16.7 ± 18.0 l (median/interquartile range: 10/2.5–25). The average daily amount of milk refusal was 0.1 ± 0.3 l and 0.3 ± 0.6 l, when calves were four weeks and eight weeks old, respectively.

Data from two groups were excluded from statistical analysis: in one ULow group, a calf died from health issues unrelated to the experiment and one UHigh group was treated for severe diarrhoea for a prolonged time and therefore was not offered 12 l of milk in order to avoid further digestive problems.

Health and weight assessment

Calves’ health state was assessed once per week by two assessors. The following indicators of compromised health were recorded: diarrhoea, coughing/sneezing and increased respiratory rate (adapted from Gratzer, et al.²⁸; Supplementary Table 1). The overall health score was set to 0 when calves showed no or one symptom of diarrhoea or coughing/sneezing and 1 when calves showed either combined diarrhoea and coughing/sneezing or increased respiratory rate. The ratio of calves with a health score of 1 is shown in Table 1.

Calves were weighed once per week between Monday and Thursday. To allow for comparison, daily weight gain for every respective week was calculated and weights were subsequently corrected for Monday as reference weighing day. Body weights from the start until the end of the experiment (three to eight weeks of age) are presented in Supplementary Figure S1. These data show that higher milk provision in MHigh and UHigh calves resulted in faster growth.

Quantification of play behaviour

Data recording

Locomotor play behaviour of calves was quantified through leg-attached accelerometers, using a previously validated method²⁹. In this study, accelerometers were used to record running, turning and bucking/buck-kicking, as defined in Größbacher, et al.³⁰. The data used to validate accelerometer recordings for these behaviours³⁰ were a subset of the data used in this study. Accelerometers (HOBO Pendant G Acceleration Data Logger, Onset Computer Corporation, Pocasset, MA, USA; product specifications described in detail in Luu, et al.²⁹) were attached to calves’ hind legs with elastic cohesive bandages. The accelerometers were oriented with the x-axis perpendicular to the ground. Acceleration was measured on the vertical axis at 1 Hz, i.e. with one measurement per second, from 05.00 until 23.04 on two consecutive days (Tuesday and Wednesday) when calves were four and eight weeks of age and recordings were stored on the device. Accelerometers were fitted to calves from the evening before until the morning after recording days, after being programmed with an optical infrared base station with USB interface and the HOBOware Pro Software (Version 3.7.8; Onset Computer Corporation, Pocasset, MA, USA).

Behaviour classification

Data processing was performed in SAS 9.4. Always 10 acceleration measurements, representing a period of 10 s each, were evaluated. These 10 s periods were categorized into lying, standing or play behaviour using quadratic discriminant analysis. This categorization was based on six predictor variables, which were calculated for each period with the respective 10 values: mean of two highest acceleration measurements, mean of two lowest acceleration measurements, variance, maximum of absolute value of change in acceleration measurement, mean change in acceleration measurements, and total sum of absolute values of change in acceleration measurements³⁰.

In order to develop the discriminant function, a reference data set was created with randomly selected short sections of accelerometer data obtained from recordings of calves in this study. This reference data set consisted of 52 recordings with a mean (± S.D.) duration of 37.8 ± 16.8 min. Lying, standing and play behaviour were visually identified from video of these recordings applying one-zero-sampling of the respective 10 s periods, for which predictor variables were calculated. This was used as the gold standard.

The discriminant function was then applied to the entire data set in two steps to identify periods that contained locomotor play, i.e. included events of running, turning and/or bucking³⁰, based on the six predictor variables: The first discriminant function classified the acceleration data into lying and standing, based on equal prior probabilities (50:50 chance of both behaviours occurring). The second discriminant function classified all standing-periods according to their presence or absence of locomotor play, based on prior probabilities of the reference data set (3:97 chance of play occurring across all treatments). The transitions from lying to standing and vice versa were almost always falsely classified as playing, as identified from video, and reclassified into standing.

The validation of processing the raw acceleration data was accomplished through checking the agreement between the acceleration-based method and visually identified play of the reference data set³⁰. It proved that although the absolute play-levels were overestimated with the acceleration method, the method was able to truthfully quantify the inter-individual differences in locomotor play in dairy calves³⁰.

Data analysis

Data processing

The last four minutes of each recording were omitted to obtain observation durations of exactly 18 h. The number of 10 s periods of locomotor play was converted into minutes of locomotor play per recording day (18 h). Recordings were excluded for the duration of disturbance when any calf in the barn escaped their pen or a person entered the pen. If more than 1 h was missing or compromised, the entire recording day was excluded for the calves affected. If less than 1 h of the recording was missing, locomotor play was calculated on a per hour basis and extrapolated to the ‘standard’ duration of 18 h. Out of 264 recordings (4 recordings per calf overall with 2 recordings at the age of four and eight weeks, respectively), 17 recordings were excluded or missing and in 13 recordings a mean (± S.D.) of 23.9 ± 13.0 min were missing and locomotor play duration and bout frequency were extrapolated. This resulted in 245 recordings, i.e. data points, included in the model. Play bouts were assessed by counting standalone periods of play, i.e. single 10 s periods not proceeded and not followed by periods classified as play were counted as one play bout, or by counting consecutive periods of play, i.e. two or more play periods occurred in a row as one play bout. Mean bout durations were assessed by recording the duration of each bout, e.g. a play bout consisting of one play period was recorded as 10 s and a play bout consisting of 3 play periods was recorded as 30 s.

Individual play was defined as one calf performing play in a 10-s-period when no other calf in the group was performing play. Dyadic play was defined as one calf performing play in the same 10-s-period as any one other calf of the group. Individual and dyadic play were calculated as minutes per recording day (18 h). Data were only included when observations of all three calves of the group were available. If data of one of the calves in the group were partially missing, observations of the other calves for the same period of time were excluded. Then the duration of individual and dyadic play was extrapolated on a per hour basis as described above. 237 recordings, i.e. data points, were included in the analysis, whereof 15 recordings were extrapolated.

The observed and randomly expected proportion of dyadic synchronized play were calculated on the basis of dyads, i.e. the combination of always two calves of a group, according to Šilerová, et al.³¹. Both were calculated for each pair combination:

where C_sync is the number of synchronous play periods of the pair, C_A is the total number of play periods of calf A, C_B is the total number of play periods of calf B and P_dyad is the total number of recorded periods for each dyad. The randomly expected proportion of play is the proportion of synchronized play occurring by chance if the calves played independently of each other³¹.

Triadic play was defined as all three calves of a group performing play in the same 10-s-period and calculated as minutes per recording day. Only periods in which data for all calves of the group were available were included. Extrapolation of play duration (due to partially missing data) was done in 10 out of 79 recordings.

Statistical analysis

All data was analysed in SAS Version 9.4. Five separate linear mixed effects models were run with total duration of play, frequency of play bouts, mean bout duration, duration of dyadic play or duration of individual play as dependent variables. Treatment (ULow, MLow, MHigh, UHigh), age (week four, week eight) and overall health score (0,1) were included as fixed class effects, while volume of unconsumed milk and maximum daily temperature were included as fixed quantitative effects, i.e. as covariates. Age (week), nested in calf and group were included as random effects. Furthermore, the date of recording was included as a crossed random effect. The same model was used for all dependent variables. Initially, the full model contained the interaction effect of treatment and age, however this was never significant and therefore removed from the model. An auto-regressive covariance structure was selected based on the Akaike Information Criterion (AIC). All models were visually inspected for normal distribution of residuals devoid of skewness.

Source: Ecology - nature.com