THE INDIVIDUALITY OF SOWS IN REGARD TO SIZE OF LITTERS BY CARL HALLQVZST ANIMAL BREEDIXG INSTITUTE, WIAD, ELDTOMTA, SWEDEN I N order to judge the selectional value of such characters as litter size and litter weight of sows it is important to know if different sows at repeated farrowings show any marked individuality in this respect. It is also of importance to learn if the individuality may be more or less strongly marked at different ages of the litters. It seems that the stage of greatest individuality should be the best moment to obtain succesful selection. LUSH and MOLLN (1937) have studied some of these characters by means of analysis of variance, and they use the intraclasscorrelation as a measure of the differentiation between individual sows. In the present study the numbers of pigs per litter for a number of sows have been analysed and the coefficient of the intraclasscorrelation has been determined with the usual formula total variancevariance within sows total variance THE INDIVIDUALITY IN REGARD TO NUMBER OF PIGS AT FARROWING. The material used was obtained from the results of farrowings in the Wiad herd. All farrowings from sows giving at least 3 litters are included. The number of sows is 117 with 596 litters, on an average 5,i litters per sow. The total number of pigs, dead or alive at farrowing, is 6333 or lo,6 pigs per litter. The analysis of variance of the numbers of pigs per litter at farrowing is recorded in Table 1. In the material of LUSH and MOLLN there were differences between herds, a cause of variation which is absent in the present material. Our data have not been corrected for age of sow, but this cause of \ ariation is at least partially eliminated by grouping the sows according to the number of their litters. A simple comparison between intersow variance and total variance (Table 1 a) gives a coefficient of 0,zg. If the sows are grouped, as done by LUSH and MOLLN, according to their birthyear in contemporary
128 CARL HALLQVIST sows (Table 1 b) the same coefficient is obtained. In the present material the year of birth of the sow is of no importance and has, consequently, been disregarded in the following analyses. TABLE 1. Analysis of variance of number of pigs farrowed per litter. Degree of free dom squares Mean square Variance ratio I ntraclasscoefficient 9 Total... Between sows... 595 116 8,93 19,~s 6,35 8,93 G.35 8,ss = 0,29 1) Total... Between groups of contemporary sows... Within groups of contemporary sows... Between sows Within sows... 595 6 589 110 479 8433 10,18 8,92 20,os 6,35 Total... Between breeds Within breeds... Between groups of sows with equal number ol litters... Within groups of sows with equal number of litters Between sows Withinsows... 585s 1 584 12 572 102 470 4360,sti 1501,ts 2859,47 8,45 45,3i 8,3s 4488 The significance is denoted thus: P < 0,001; * 0,001 < P < 0,111; 0,ol <P < 0,m; p>o,o5, where P is the probability that only random variation occurs. In the group with 10 litters there was one sow only and therefore omitted in this case. In Table 1 c the material has been grouped according to breeds (Large White and Improved Swedish Landrace) and number of litters
THE INDIVIDUALITY OF SOWS 129 per sow. This number has varied from 3 to 10 and the number of sows with different numbers of litters is as follows: Total Total Number of litters... 3 4 5 6 7 8 9 10 sows litters Number of sows with different numbers of litters... 28 26 18 17 15 7 5 1 117 596 The grouping according to number of litters must be important. For sows with a low number of litters only the first litters are recorded and it is known that these have a lower number of pigs than the later ones. Furthermore, sows with markedly low numbers of pigs in their first litters are often slaughtered, and in the above material the poorest sows are, therefore, to be found in the classes with only few litters. It is therefore to be expected that the lowest numbers of pigs per litter should be found among the sows with small numbers of litters. Table 1 c shows that in the present material the difference between breeds is statistically insignificant and has but very slight influence on the coefficient of correlation. The number of litters per sow, as expected, is of consid,erable importance. If breed and number of litters are simultaneously considered, the coefficient is only O,zo, which value, however, is considerably higher than that obtained by LUSH and MOLLN. TABLE 2. Number of pigs at different ages of the litters. Farrowing, including deadhorn... )), alive... Alive after 1 week... )) )) 2 weeks... 0 )) 3 ))... )) )) 4 ))... )) )) 5 ))... )) )) G ))... Total 5060 4670 4001 3893 385 1 3723 3658 3585 % of farrowed 100,o 92,3 79,i 76,~ 76,t 73,a 72,3 70,8 Average number per litter THE INDIVIDUALITY IN REGARD TO SIZE OF LITTER MEASURED AT DIFFERENT AGES OF THE LITTERS. It is of interest to know at which age of the pigs the size of the litters should b,e determined so as to show most clearly the differentiation Heredltaa XXVIII. 9
130 CARL HALLQVIST between the sows. In a study of this problem we have utilized all sows having at least 3 litters with all pigs unweaned at the age of six weeks. The material analysed has thus been limited to 97 sows with 461 litters. For each of these litters we know the total number of pigs at farrowing, the number of pigs alive at farrowing and the number of pigs alive at the end of 1, 2, 3, 4, 5, and 6 weeks after farrowing. TABLE 3. Analysis of variance of numbers of pigs per litter at different ages of the litters. Between Within [ntra SOWS Variance class Age of litters ratio coeffi I 460 I 96 364 cient Farrowing(including deadborn) I I sows squares 4468,043 1725,891 Mean sauare 9.71 1i.w 27x2~ 52 7.53 17,s~ 7.53 2939 I 0,22 Farrowing (alive) squares 3468,181 Mean square 7,54 1348,881 14,o~ 2119,310 5,s~ 1 week 2 weeks 3 weeks I Mean square I 5,85 1 lo,.so weeks squares 2561,343 791,086 Mean square 5,57 8.24 I squares Sum Of I 2524,048 I 739,467 Mean square 1 5,49 I 7,70 I I 6 weeks I * as in Table 1. I squares 1 2642,831 1 936,863 1 sum of 1 I I Mean sauare I 5.75 I 9.76 squares 2631,307 914,3613 I Mean sauare I 5172 I 9152 1 Mean square I 5,63 1 7,;s 1701,207 4,87 1705,968 4.09 1770,257 4,ao 1784,581 4,911 1845,~ 5,07 0,18 o,l3 0,11 0,lO
THE INDIVIDUALITY OF SOWS 131 The numbers of pigs at each time are given in Table 2. For each of these eight sets of data the coefficient of intraclasscorrelation has been calculated (Table 3). The sole object of this study was to determine the relative sizes of coefficients, determined at different ages of the litters, therefore variations in breed and in number of litters per sow have been disregarded. It is evident from Table 3 that the coefficient tends to decrease with increasing age of the litter, a phenomenon which was to be expected. Though the sows differ in milk production and care of their offspring, the environment increases in importance with the ageing of the litters and concomitantly the differences between the sows decrease. The coefficient of correlation, as determined on the number of pigs at farrowing, is smaller in this material than in that presented in Table 1, although the difference is hardly significant. TESTING THE INDIVIDUALITY BY SELECTION. THE PRACTICAL RESULTS OF SELECTION. The analysis of variance of the present material gave a coefficient of intraclasscorrelation of k 0,so. In order to test this figure and to demonstrate its significance in practical breeding work the following calculations were made: The sows have been grouped according to the number of pigs in their first litter. The total average of this number is 9,3. In group A are placed all sows with 9 pigs or less in their first litter, in group B those with 10 or more. In group A there are 56 sows with an average number of 6,s pigs, in group B 61 sows show an average of 11,~. The difference b,etween the two averages is 4,8 pigs. The coefficient of intraclasscorrelation being O,zo, it is to be expected that 20 % of this difference will remain between the two groups, if the average number of pigs in the subsequent litters is determined. This 20 4,8 expected difference is ~ = O,%. The empirical data are as 100 follows: The sows in group A have given 220 later litters with an average of 10,4 pigs per litter. The sows of group B have 114 as an average of 259 later litters. The difference 1,o fits the expected difference of 0, remarkably ~ well. If no selection is made the average number of pigs in the later litters is 10,g. The rejection of all sows with a number of pigs below the average in the first litter increases the average number of pigs in the subsequent litters by 0,5 or somewhat less than 5 %.
~ The 132 CARL HALLQVIST The difference between individual sows is most pronounced when determined on the number of pigs at farrowing and is considerably less if determined on the number alive six weeks later. From the pigbreeder s point of view, however, the selection of mother sows should be as efficient as possible with respect to the economical value of the litter at the time of weaning or at some fixed age near this time. The number of pigs alive six weeks after the first farrowing may be more closely correlated with the number and weight of the subsequent litters at weaning than the number of pigs at birth in the first litter. The former number may, therefore, be a better basis of selection, in spite of the fact that it offers a poorer measure of the differentiation between sows. It is also quite possible that the selection will be most effective if it is based on both numbers: pigs at the first farrowing and pigs alive six weeks later. In order to test the effects of the different modes of selection the following calculations were made: The 97 sows have been divided into four groups according to the numbers at birth of the first litter and the numbers alive six weeks later. The four groups are: 9 : 7. 9 pigs or less at birth, 7 or less alive after six weeks. 9 : 8. Same numbers at birth, 8 or 9 alive after six weeks. 10 : 7. 10 pigs or more at birth, 7 or less alive after six weeks. 10 : 8. Same numbers at birth, at least 8 alive after six weeks. With this grouping it was possible to study the effects of different modes of selection of mother sows after the results of the first farrowing. Table 4 shows the results of three different modes of selection, calculated with respect to number of pigs at weaning as well as with respect to total weight of litter at this time. three,selection experiments, carried out on the material are as follows: (1) The 97 sows are grouped according to the number of pigs at birth in the first farrowing. A: 9 pigs or less, 9 : 7 i 9 : 8. B: 10 pigs or more, 10 : 7 l 10 : 8. The average number of pigs per litter at weaning and the mean weight of litter at the same time are determined for all subsequent litters of sows in group B. (2) The sows are grouped according to the number of pigs in the first litter alive after six weeks.
THE INDIVIDUALITY OF SOWS 133 A:. 7 pigs or less, 9 : 7 f 10 : 7. B: 8 pigs or more, 9 : 8 4 10 : 8. The average result of the subsequent litters of group B is determined as in (1). (3) The sows are grouped according to numbers at birth and six weeks later. A: 9 pigs or less at farrowing and 7 or less six weeks later, 9: 7 k 9 :8f 10: 7. B: 10 pigs or more at farrowing and 8 or more at six weeks, 10 : 8. The average results of all subsequent litters of group B are determined as in (1) and (2). The results of the 3 different modes of selection are compared with the average result of all subsequent litters from all 97 sows, i. e. the whole material without selection. The corresponding data are given in Table 4 as selection number 0. It is evident from Table 4 that the number of pigs alive 6 weeks after the first farrowing is of more value in practical breeding than the number born in the first litter. The rejection of all sows below the average in the latter number only increases the result in the subsequent litters at weaning by 2,s % in number and 1,4 % in weight, whereas the rejection according to number alive after six weeks has an improving effect of 4,4 76 and 3,4 %, respectively. The best result is obtained if the selection is made as in (3). In this case all sows are eliminated which have a fecundity below the average in their first litter and also all sows with good fecundity but poor capacity for maintaining their offspring until weaning age. By this selection an increase of 5.8 % is gained in number and of 3.6 % in weight at weaning of the subsequent litters. It is of interest to note that the selection according to number has a smaller effect on the weight of litter than on the number. Environmental factors must have a greater influence on weight than on number, and that was to be expected. According to the data of the present paper the individuality of sows in producing litter of a certain size is more pronounced than in the material of LUSH and MOLLN. There seems to exist some possibilities of making a selection based on the number of pigs in the first litters of the sows. This conclusion is also supported by the results of certain German investigations (HARING und HAGEN, 1939). Studies of
100,o 82,s 101,r 84,1 103,r 843 1 103,6 Group B Group A Number of pigs at 6 weeks in 4 CI Selected sows Rejected sows later litters 1 u 0 v1 Group Number Group Number Per litter yo/ ' of sows of sows 0 10:8+10:7+9:8+9:7 Y7 0 7,77 100.0 9:8+9: 7 44 7,99 102,s 1 : 1 10: 8 + 10: 7 10:8+ 9:s 10:s 42 z; I 10 : 749: 7 46 8,ii 104~ 10: 7 f 9 :8 +9:7 55 1 8,22 I 105,s Litter weight ai 6 weeks in late] litters ' pcr2iteri 81,r
THE INDIVIDUALITY OF SOWS 135 the same kind but bearing on the individuality of the weights of pigs and litters are now in progress. SUMMARY. The results of 596 farrowings of 117 sows of the Wiad herd were studied, all sows giving at least 3 litters. The birthyear of the sow had no influence on the number of pigs per litter. There was a small and insignificant difference in this respect between the two breeds studied, viz. Large White and Improved Swedish Landrace. If the material is grouped according to breed and number of litters per sow there is an intraclasscorrelation of f 0,zo. 461 litters from 97 of the sows were kept together at six weeks. On this material the coefficients of intraclasscorrelation was determined on the following 8 sets of data: number of pigs at birth; number of pigs alive at birth; number of pigs alive at the end of 1, 2, 3, 4, 5 and 6 weeks alter birth. The value of the coefficient decreases with increasing age of the litters. If all sows are divided into two groups, above and below the average with respect to number of pigs born in the first litter, the difference in this number between the groups is 4,s. As the coefficient of intraclasscorrelation is O,ao, it was to be expected that there should be a difference of 0,96 between the averages of the two groups in regard to all subsequent litters. The actual difference observed is 1,0, and by rejection of all sows below the average in the first litter the result in the subsequent litters is improved by 5 %. The sows were grouped according to number born in the first litter and number alive six weeks later, and the effect of selection, according to these numbers, on the average number and weight of all subsequent litters at weaning was studied. The selection according to number alive after six weeks proved to be more effective than the selection according to number born. The best result of the selection was obtained when due regard was paid to both these numbers. LITERATURE CITED. 1. HARING, F. and HAGEN, H. 1939. Einfluss der Erstleistung auf die Hohe der Lebensleistung in der Schweinezucht. KiihnArchiv, Bd. 52. 2. LUSH, J. L. and MOLLN, A. E. 1937. The degrees to which litter size is a constant characteristic of sows. The American Society of Animal Production. Record of Proceedings.