Phenotypic and approximate genetic correlations between milk yield per lactation, days open and four-month calf weight were determined, using records from 5,514 calvings in 14 Venezuelan dual purpose herds. Cows were classified according to breed group as zebu-type (Z), and crossbreds of medium (M) or high (H) levels of European dairy breed inheritance. The approximate genetic correlations were calculated, first, using the deviations (D) of each record from contemporary averages and, second, using the cows' estimated breeding values (EBV's) based on fictitious values for heritability and repeatability. In both cases, the correlations were calculated from records on pairs of traits on the same animals, rather than on relatives. The analyses were carried out separately for each breed group. For milk yield:days open, they were also separated according to milking system (with or without calf suckling) and level of milk yield (above or below the average breed group yield, within milking system). The correlations between milk yield and days open were 0.21** (phenotypic), 0.36** (D) and 0.16** (EBV). The phenotypic and D-based values increased consistently with level of milk yield within breed group and milking system, reaching 0.34** (above average, suckled cows) and 0.39** (above average, non-suckled cows). All correlations tended to be higher in the H cows than in the rest.

The correlations between milk yield and calf weight were low but positive in all breed groups (0.11**, phenotypic; 0.09**, D; and 0.16**, EBV). For calf weight and days open, all correlations were close to zero. It was concluded that although the phenotypic correlation between milk yield and days open was very weak in suckled cows with saleable yields up to about 1800 kg, there may be a positive genetic correlation at this level, as well as at the higher levels of milk yield (above 2700 kg, without suckling) where the phenotypic correlation was more clearly manifest. Fertility should therefore be taken into account in the selection of all dual purpose cows, until more precise information is obtained. Selection for saleable milk yield in suckled cows should slightly favour four-month calf weight.

The lack of information from which to develop appropriate selection procedures is recognised as an important obstacle to the genetic improvement of tropical dual purpose cattle in Latin America. Milk yield, fertility and growth rates are evidently traits of vital economic importance in dual purpose systems, but there are very few estimates of heritability or genetic correlations available for the mixed crossbred populations which predominate. The probability of obtaining precise information on the genetic parameters in these populations in the near future is low because the restricted use of artificial insemination and lack of sire identification reduces the volume of records on related animals.

The objective of the present study was to determine the phenotypic correlations between the three traits mentioned above and, using approximate procedures, estimate the probable magnitude and direction of the genetic correlations. These should provide an acceptable temporary base for selection until more precise information becomes available.

The data were obtained from 14 commercial herds which participate in the Dual Purpose Cattle Research Project of the Central University of Venezuela. The farms are situated at less than 350 m above sea level with an average annual temperature of about 26 ºC and dry seasons extending from three to seven months per year. All herds are kept on grazing and raise male calves to at least 12 months of age, but the production systems vary greatly in intensity. Cows on eleven farms are milked in the presence of the calves which are raised by restricted suckling. The amount of milk allowed to the calves varies between and within herds, depending on the farmers' criteria, which include the age and condition of individual animals. Thus, the proportion of milk which is extracted at milking ("saleable" yield) is not constant or definable from existing information. Extensive use is made of natural pastures and cereal stubbles, with little or no supplementary feeding. The other three farms are more intensive. Milking takes place without the calves which are reared artificially and thus the cows' total yield is extracted. Pastures are based on cultivated grass species, especially Cynodon, and feed supplements are provided. In general, artificial insemination is used occasionally and cows are usually kept with the bulls. The cows on all farms are mainly mixed crossbreds and are classified on the basis of external appearance as zebu-type (Z), or crossbreds of medium (M) or high (H) grades of European dairy breed inheritance. Although they are racially very heterogeneous, the major influences on the herds' genetic composition come from the Venezuelan Brahman, American Brown Swiss and American Holstein breeds.

The traits included in the study were milk yield per lactation, days open and calf weight at four months of age. Milk yield was estimated from weekly or monthly samples and all lactations, including those of zero yield which accounted for 2.9% of the total, are included in the data. Days open were calculated from calving intervals where these were complete, or otherwise from expected calving dates based on pregnancy diagnoses made by experienced technicians. Calf weights were available from the 11 herds which used restricted suckling. Records were corrected for sex and age at four months, and zero assigned to those dams whose calves died (13.7% of the total). The least squares adjusted average values for these traits obtained from a recent overall analysis of the Project data were 1,509 kg milk per lactation, 145 days open and 71 kg calf weight (Vaccaro et al 1996).

The records used refer to a total of 5,514 calvings and three different data sets were used to estimate the correlations between the traits studied. First, the phenotypic correlations were calculated using the raw data (R), unadjusted except for the age and sex correction described above for the calf weights. The second and third data sets were used to generate approximate genetic correlations. The second set consisted of the records from all 5,514 calvings expressed as deviations (D) from corresponding contemporary means, taking into account herd (14), calving year (1990-1994), season (wet, dry) and number (1st, later). These factors have consistently been found to be the most important non-genetic sources of variation in analyses of the Project data and are routinely incorporated into estimates of the breeding values of cows in the cooperator herds. The third data set consisted of the expected breeding values (EBV's) for the three traits of 2,449 cows in the 11 herds which had calf weights available. The EBV's were calculated according to the standard procedure described by Legates and Warwick (1990). Deviations derived from contemporary means consisting of less than 10 records were excluded. The proportions of EBV's which incorporated records from one, two, three and four of more lactations per cow were 41%, 25%, 18% and 16%, respectively. In the absence of precise estimates, the following hypothetical values of heritability and repeatability were used: 0.20 and 0.30 (milk yield), 0.15 and 0.25 (days open) and 0.30 and 0.40 (calf weight).

The correlations were calculated overall and separately for each of the three breed groups (Z, M and H). Since preliminary results suggested that the relationship between milk yield and days open might be affected by the level of milk production, the analyses of the R and D data sets were repeated for cows in each breed group whose records were above or below the group's average yield in the suckled and non-suckled milking systems, respectively.

Table 1 summarises the phenotypic correlations derived from the R data set. The highest total correlation was that between milk yield and days open (0.21), and a higher value was found for the H cows (0.27) than for the M (0.13) or Z (0.11) animals. Table 2 shows the results separated by milking system and level of milk yield. In the suckled cows, the correlations were all low. None of them exceeded 0.19 and they only reached the 1% level of statistical significance in the H breed group. Also, there was no clear pattern in the comparisons between milk yield levels within the three breed groups. In the non-suckled cows, however, a much higher total correlation was obtained for the high-yielding cows (0.39) compared to the lower producers (0.18), and a similar pattern was observed consistently within breed groups, except for the very small Z sample.

Table 1: Phenotypic correlations between traits, according to breed group (zebu type or medium or high grade of crossbred European dairy)
	Zebu	Grade European
Traits	-type	Medium	High	Total
Milk yield and:
Days open	0.11**	0.13**	0.27**	0.21**
(n)	(1053)	(2571)	(1890)	(5514)
Calf weight	0.23**	0.12**	0.14**	0.11**
(n)	(1121)	(2259)	(917)	(4297)
Calf weight and:
Days open	0.05	-0.13**	0.02	-0.06 **
(pon)	(923)	(1939)	(733)	(3595)

Number of observations in parenthesis
* P < 0.05
** P < 0.01t

Table 2: Phenotypic correlations between milk yield and days open, according to breed group, milking system and level of milk yield#
	Zebu	Grade European
	-type	Med	High	Total
With suckling
Mean yield,kg	944	1248	1799	1285
Correlations:
Low yield group	-0.03	0.01	0.11*	0.02
(n)	(607)	(1322)	(485)	(2560)
High yield group	0.10*	-0.01	0.19**	-0.01
(n)	(416)	(788)	(342)	(1400)
Without suckling
Mean yield,kg	2739	2920	3268	3155
Correlations
Low yield group	0.37	0.06	0.17**	0 .18 **
(n)	(18)	(231)	(556)	(803)
High yield group	0.34	0.38**	0.35**	0.39**
(n)	(12)	(230)	(507)	(751)

Number of observations in parenthesis
# Classified as above (high) or below (low) the average milk yield for each breed group, within milking system. In the 'total' columns, records were reclassified as above or below the overall milking system average.
* P < 0.05
** P < 0.01

The next phenotypic correlation in order of magnitude was that between milk yield and calf weight (Table 1). Although the overall value (0.11) was very low, the correlations were consistently positive in all breed groups and slightly higher in the Z cows (0.23) than in the crossbreds (0.12 to 0.14). The correlations between calf weight and days open were generally close to zero (Table 1).

Table 3 shows the results based on the D values. The positive correlations between milk yield and days open were stronger in all breed groups than those shown in Table 1, ranging from 0.26 in the Z group to 0.42 in the H cows. Broken down by milk yield level and milking system, Table 4 shows that most correlations reached statistical significance, even for the low-yielding cows in the suckled system. Significant values for high-yielding cows in both systems ranged from 0.26 to 0.45, and mostly exceeded those of the lower producers. In general, the correlations were also lower in the zebu-type cows and tended to increase with rising levels of European breed inheritance.

Table 3: Correlations between traits based on deviations from contemporary means, according to breed group.
	Zebu	Grade European
	-type	Med	High	Total
Milk yield and:
Days open	0.26**	0.32**	0.42**	0.36**
(n)	(1053)	(2571)	(1890)	(5514)
Calf weight	0.09**	0.11**	0.10**	0.09**
(n)	(1121)	(2259)	(917)	(4297)
Calf weight and:
Days open	0.04	-0.01	0.02	0.01
(n)	(923)	(1939)	(733)	(3595)

Number of observations in parenthesis
** P < 0.01

Table 4: Correlations between milk yield  and  days open based on deviations from contemporary means, according  to  breed group, milking system and level of milk yield#
	Zebu	Grade European
	-type	Med	High	Total
With suckling
Mean yield,kg	944	1248	1799	1285
Low yield group
Correlation	0.06	0.18**	0.22**	0.16**
(n)	(607)	(1322)	(485)	(2560)
High yield group
Correlation	0.26**	0.33**	0.45**	0.34**
(n)	(416)	(788)	(342)	(1400)
Without suckling
Mean yield,kg	2739	2920	3268	3155
Low yield group
Correlation	0.53*	0.13*	0.19**	0.19**
(n)	(18)	(231)	(556)	(803)
High yield group
Correlation	0.51	0.36**	0.36**	0.39**
(n)	(12)	(230)	(507)	(751)

Number of observations in parenthesis
# Classified as above (high) or below (low) the average milk yield for each breed group, within milking system. In the 'total' columns, records were reclassified as above or below the overall milking system average.
* P<0.05
** P<0.01

The correlations based on D values involving calf weight remained positive but low for milk yield, and did not reach significance for days open (Table 3).

Table 5 shows the results using the EBV's. The correlations of milk yield with days open are somewhat lower in each breed group than those based on the D records, and failed to reach significance in the Z breed group. The results are consistent with those of Table 4 in terms of the increase in the magnitude of the correlations according to degree of European-breed inheritance. In the case of milk yield and calf weight, the low, positive correlations also follow the trends observed before. The correlations between calf weight and days open remained non-significant.

Table 5: Correlations between traits based on expected breeding values
	Zebu	Grade European
	-type	Med	High	Total
Milk yield and:
Days open	0.04	0.13**	0.24**	0.16**
(n)	(526)	(1143)	(609)	(2449)
Calf weight	0.15**	0.16**	0.22**	0.16**
Calf weight and
days open	-0.05	-0.05	0.05	-0.04 *

** P < .01
Number of observations in parenthesis. In the 'total' column, records of an additional 171 cows with unidentified breed groups are excluded.

The results derived from the R data set estimate the phenotypic correlations between the three traits studied, while those based on the D and EBV values provide two different approximations of the genetic correlations. Both the D and the EBV-based values are likely to overestimate the true genetic correlations because, in contrast to standard procedure, the covariance between breeding values was measured on the same animals instead of on relatives. This may have inflated the covariance for non-genetic reasons. Although the correlations based on the D values are subject to error where the numbers of contemporary records were low, they are free from the bias which may have affected the EBV data set due to the possibly erroneous values assumed for heritability and repeatability. The values of the latter used for days open may seem especially questionable, but measurements such as days open probably underestimate genetic variation in cow fertility (ALPA 1988a), and the evidence of moderate levels of heritability in American zebu (ALPA 1988b), which contribute importantly to the dual purpose populations, was also taken into account in fixing the values used. Despite the methodological shortcomings, the consistency of the results obtained lends some validity to the conclusions which are drawn.

First, the phenotypic correlation between milk yield and days open appears to become manifest at higher levels of milk yield in the dual purpose system. The results from Table 2 suggest that the correlation is non-existent or very weakly positive in suckled cows where yields excluding milk consumed by calves average up to about 1800 kg, according to breed group. This would cover a large proportion of the dual purpose enterprises throughout tropical Latin America. However, the results of Table 4 indicate that, even in the suckled system, cows in all breed groups which exceed their contemporaries in milk yield also tend to exceed them in days open. This suggests that there may be a positive genetic correlation between the two traits, even at quite low production levels where the phenotypic correlation is weak or absent.

Allowing 600 kg milk for consumption by the calf in the suckled system, the total yield would have averaged about 1800 kg, or up to about 2400 kg in the case of the H cows. According to Table 2, the phenotypic correlation between yield and days open only reached moderate importance (above 0.30) at levels of total yield above this, as found in the non-suckled system where means ranged from 2739 to 3268 kg, according to breed group. This coincides with evidence from other samples of dual purpose cattle in the region that long calving intervals and specific reproductive problems are more common in high yielding cows (González 1990; Navarrete et al 1995). Tables 4 and 5 suggest that this observed correlation has a genetic base which may be of low to moderate importance.

Earlier references to possible genetic correlations in these populations were derived from sub-sets of the present data (Vaccaro et al 1994), and are therefore superseded by this study. Although the true magnitude of the genetic correlation is not yet clear, the consistency of the evidence based on the D and EBV data sets suggests that it may be real, especially in higher-yielding cows, irrespective of breed group. Therefore, until more precise information is available, fertility should be included in any selection program for dual purpose populations, whatever the level of intensity. Even though the antagonism between milk yield and fertility may be less evident in zebu and lower-yielding crossbreds, selection for milk yield alone would be expected to compromise their fertility over the longer term.

The second conclusion of practical importance refers to the relation between saleable milk yield and calf weight. Although saleable yield here represents an undefinable fraction of the total because of the variable milking-suckling procedures, such variation is characteristic of commercial farms throughout the region, and the present results may therefore be considered to be representative of the system as commonly employed. Earlier evidence concerning the correlation between saleable yield and calf weight was contradictory (Aragon and Deaton 1981), but more recent studies based on larger data sets and calf weights at or before weaning found the phenotypic relation to be either zero or slightly positive (Vaccaro et al 1994). The present study, containing the highest number of observations, supports the evidence of a low, positive phenotypic correlation in all breed groups and suggests that it may have a genetic basis. Selection for milk yield should therefore slightly favour pre-weaning calf weights, or at least not prejudice them. Also, since there is no evidence of any phenotypic or genetic relationship between calf weight and days open, inclusion of dam fertility in the selection program would not be expected to make any long term difference to the weights of the calves.

This research was carried out thanks to the financial support of the International Development Research Centre, Canada. The authors also express their deep appreciation to the cooperating farmers.

(Received 13 June 1996)