Materials and methods Directory UMM :Data Elmu:jurnal:A:Applied Animal Behaviour Science:Vol66.Issue3.2000:

Ž . They found no relations between the backtest modified after Hessing and other behavioural tests and the intra-test consistency was low. Using factor analysis, they did find correlations between several behavioural traits, excluding the backtest. Three personality factors are suggested by these experiments: aggression, sociability and Ž . exploration Forkman et al., 1995; Jensen, 1995; Jensen et al., 1995a,b . The objectives of our studies are threefold. Firstly, we investigated if individual characteristics of pigs, measured with the backtest, are consistent. Secondly, we determined if these individual characteristics were related to sow, boar or gender. Thirdly, we studied the relationship between these individual characteristics and the performance of the pigs. If backtest results in pigs are related to performance, this would mean that a selection of the best performers can be made at an early age. Also, if backtest results are heritable and related to performance, a selection in sows can be made according to their own backtest results.

2. Materials and methods

2.1. Housing and animals The studies were performed in farrowing and fattening units at the farm ‘The Tolakker’ of the University of Utrecht, the Netherlands. In two studies, 823 and 566 fattening pigs, born from Dutch LandraceGreat Yorkshire sows, were used. The piglets were born in conventional farrowing pens. A heating lamp and floor heating were used if necessary. An all-in-all-out system was used consequently. Male pigs were castrated at approximately 5 days of age and tails of all pigs were docked, according to standard Dutch procedures. Iron was given to the piglets by injection at 3 days of age. At weaning, around 28 days of age, the sow was removed from the pen and the piglets remained in the same pen until 9 weeks of age. At that time, the fattening period started. All fattening pens had a partly slatted, concrete floor with floor heating. Feed and water were supplied ad libitum. Pigs were slaughtered at Ž . approximately 110 kg liveweight around 6 months of age . 2.2. Backtest Ž . A backtest was performed as described by Hessing et al. 1993 . In this backtest, each piglet was put on its back and restrained in this supine position for 1 min. One hand was placed loosely over the head of the pig, the other was placed loosely on the hind legs. Each series of wriggles that the piglet made without a pause was counted as one escape attempt. The total number of escape attempts is called ‘backtest score’. In the first of our studies, all piglets were tested twice during the suckling period, at 10 and 17 days of age. In the second study, piglets were tested at 3, 10 and 17 days of age. 2.3. Data In the first study, 823 pigs were followed from birth until slaughter. The following performance parameters were recorded for each pig: weight at birth, at weaning, at approximately 9 weeks and at slaughter, lean meat percentage at slaughter and carcass grading. Lean meat was calculated using muscle- and backfat thickness, measured on Ž . Ž . carcasses with the Henessy Grading Probe HGP Engel and Walstra, 1993 . Live wŽ weight just before slaughter was calculated as slaughtered weight 83 y slaughtered . x weight 0.0025 q 1.3. Gender, sow and boar identity were recorded. Carcass grading Ž . meatiness in the slaughterhouse was recorded using the SEUROP system. With this system, carcasses are graded according to the muscularity of the slaughtered pig: S s Superior, E s Excellent, U s Very good, R s Good, O s Moderate and P s Poor. In the second study, data from 566 piglets were collected until weaning. Weight at birth and at weaning were recorded, as well as gender and health of the piglet. Sow identity, sow reaction to piglet removal and medical treatments of the sow were recorded. Sows were backtested before the start of the study, at the age of 10 and 17 days. Sows were divided into three backtest score groups, based on the outcome of two Ž . backtests, according to Hessing et al. 1994b : Ø An animal was classified as LR when the total number of escape attempts was less Ž . than four, with a maximum of two attempts per test e.g., 0 q 0 or 1 q 2 ; Ø An animal was classified as HR when the total number of escape attempts was more Ž . than four, with a minimum of two attempts per test e.g., 2 q 3 or 4 q 4 ; Ž . Ž . Ø The other animals were classified Intermediate I e.g., 0 q 3 or 2 q 2 . The reaction of the sow to piglet removal was scored as ‘quiet’ or ‘aggressive’. A Ž . sow was ‘aggressive’ when she reacted hostile e.g., making noise, jumping up when one of her piglets was taken out of the pen for testing. A sow that did not react was classified as ‘quiet’. 2.4. Statistical methods 2.4.1. Consistency of backtest scores The Pearson correlation coefficient is used to determine the consistency of backtest scores in the successive tests. 2.4.2. Backtest scores and related factors Relations between factors such as birth weight, gender, boar and sow, and backtest scores were calculated for both studies. The effect of test day was also calculated. 2.4.2.1. First study. Preliminary analysis showed no effect of the different pens, so this variable is not used in the models. Ž . To determine the effect of test day 10 or 17 days of age on backtest score per test Ž . y , the following model was used: y s m q piglet number q testday q ´ i j i j i j. Since there was no effect of test day in the first study, the rest of the models were not corrected for test day and the mean number of escape attempts in the two tests was calculated for each animal. To determine the effect of AI or natural service and the effect of boar number on Ž . backtest scores, the mean backtest score per litter was calculated y . This means that Ž . first the mean score per animal was calculated from the results of the two tests and these figures were used to calculate the mean score per litter. The following models were used: y s m q AI q ´ , i i i y s m q boar number q ´ . i i i Ž . To determine the effect of gender, mean backtest score per animal from two tests Ž . y was calculated and the following model was used: y s m q litter q gender q ´ . i j i j i j Ž 2.4.2.2. Second study. To determine the effect of litter, gender and test day 3, 10 or 17 . Ž . days of age on backtest score per test y , the following model was used: y s m q litter q gender q ´ q test day q ´ , i jk l i j 1 :i jk l 2 :i jk l Ž . in which ‘litter’ is the combined effect of sow and boar each sow is used only once , ´ 1 is the errorterm of litter and gender nested within piglet number and ´ is the errorterm 2 of the total model. The relation between sow reaction, sow health and sow backtest scores was calcu- lated using logistic regression. Since sow reaction and sow health were related, we used only sow reaction and sow backtest scores in the final models. To determine the effect of sow reaction and sow backtest score on backtest score of the piglets, mean backtest score per litter was calculated. This means that first, the mean Ž . backtest score per animal was calculated from the results of the three tests , and these figures were used to calculate the mean score per litter. The following model was used: y s m q sow reaction q sow backtest score q ´ . i j i j i j 2.4.3. Relations between backtest scores and production parameters Ž For the first study, relations between production parameters lean meat percentage, . carcass grading and daily weight gain and backtest scores were calculated. Since no Ž . day-effect was found, the mean number of escape attempts mean of two tests per Ž . animal is used. To determine the effect of mean backtest score on daily weight gain y , the following model was used: y s m q litter q weight q gender q backtest score q ´ , i jk l i j k l i jk l in which y is daily weight gain in the suckling, growing or fattening period and ‘weight’ is birth weight, weight at weaning or weight at the start of the fattening period. To determine the effect of mean backtest score on meat percentage and carcass Ž . grading y , the following model was used: y s m q litter q weight q gender q backtest score q ´ , i jk l i j k l i jk l in which y is lean meat percentage or carcass grading and ‘weight’ is birth weight, weight at weaning or weight at the start of the fattening period. 2.4.4. General Ž . For all models, the Shapiro–Wilk statistics W was calculated as a statistical test for normality and the skewness and kurtosis of the rest value of the models were deter- Ž mined. All calculations were performed using the statistical package, SAS SAS . Institute, 1989 . P-values - 0.05 were considered to be significant.

3. Results