The aim of this study was to estimate heritabilities and determine genetic and phenotypic correlations for BW, breast meat yield, and meat quality traits in turkeys. We also aimed to estimate the growth curve and the heritabilities of its parameters. The phenotypes used in this study were measured on an F2 cross between 2 turkey lines with a different genetic background and selected for different traits. The variances obtained are relevant to the F2 cross and cannot be directly applied to existing breeding stock. The estimates do provide a useful benchmark for breeders interested in the potential for correlated responses in meat quality from selection on growth and yield and for breeders who contemplate the estimation of heritabilities in their breeding lines and/or adding these traits to their breeding objectives.
In the present study, body weight was considered to be a single trait across both sexes, with sex used as a fixed effect in the analyses. This was in contrast with other studies, where parameters were estimated separately for males and females [13, 26, 27]. Parameters were not estimated separately in our analyses because those estimates would have been based on a subset of our relatively small population. Joint analysis of males and females seems warranted because genetic correlations between BW of males and BW of females at the same age were found to be high. In addition to sex, hatch date was included as a fixed effect in the analyses because it was found to play a significant role in BW and other traits in the study [28, 29].
In the present study, univariate models were used for the estimation of heritability and bivariate models for the estimation of genetic and phenotypic correlations . Multivariate analyses were performed for small groups of related traits and results were not different from those obtained from univariate and bivariate models. Combining traits did not always result in convergence of the REML estimation. A common environmental variance (c
) was found significant for some traits (BW01 & BW17) and not for others which further complicated the estimation from multivariate models.
We found heritability estimates for BW traits in the expected range, except for BW01 and BW17, which is attributed to the strong common environment effect at those early ages. Results are in range with previously reported heritability estimates. BW traits at various ages were reported to have an average heritability of 0.41 in a review of eighteen reports by Arthur and Abplanalp . Similar results were also reported by Buss , who observed heritability in the range of 0.23 to 0.71 for BW traits at different ages.
The common environment effect had a large impact on the estimates of heritability for BW traits, especially at early ages. Neglecting the common environment effect would have resulted in an overestimation of heritabilities at early ages. For comparison, we estimated heritabilities without including the common environmental effect (results not shown), and found that the estimated heritability of body weight was increased at all ages, but especially for BW01 and BW17. Similar conclusions were reached by others regarding the effect of common environment on the estimation of heritability [12, 32–34]. In our study, c
was found to decrease with increases in age and it was close to zero at later ages. The direct genetic component was found to increase with age which could be attributed to the initiation of expression of the animal's own genetics.
In the present study, the BW of day old turkey chicks had a heritability close to zero. Tullett and Burton  found in a study on broilers that 97% of the variation in chick weight at hatching was due to two factors: fresh egg weight and weight loss during incubation. Moreover, North  found that egg weight represented 70% of the chick weight. Taken together, these results suggest that day old BW was not heritable, but egg weight or egg size was heritable.
Our heritability estimates of the other production traits, including PBM, BrL, and BrW, were also consistent with reports from other groups. Our heritability estimate for PBM was 0.30, similar to values found by Le Bihan-Duval et al.  in chickens. The comparison is made to chicken because it is the closest related species to the turkey for which values are available. Our heritability estimates for breast length and breast width were low and quite close to each other. These results were in agreement with the work of Adeyinka et al.  on chickens. Our heritability estimate for PDL at 0.12 was the first reported for turkey meat, and somewhat inconsistent with the heritability of 0.26 found in chickens by Le Bihan-Duval et al. . Besides the estimate being made in different species there were also differences in the measurement of traits with Le Bihan-Duval et al.  measuring PDL from the whole breast muscle while a smaller breast meat sample was used in our study.
The heritabilities in the present study for pHu, a* and b* at 0.09, 0.30 and 0.15 were found roughly in agreement with the results of Le Bihan-Duval et al.  in turkeys, while our estimate of heritability for L*, 0.27, was somewhat higher that the value of 0.12 obtained Le Bihan-Duval et al. . A possible explanation can be sought in the different fixed effects included in the models by these two studies which in turkey may have explained a bigger part of the residual variance for this particular trait L*.
Sengul and Kuraz  concluded that Gompertz, Logistic, and Richards models all performed well for describing growth in turkeys. The logistic and Gompertz models have fixed growth forms with points of inflection at about 50 and 37% of the asymptote, respectively . These parameter models are special cases of the more flexible Richards model, which has a variable point of inflection specified by the shape parameter . The growth models (Logistic, Gompertz and Richards) also differ slightly from each other in the interpretation of other parameters . Here, we choose to use the logistic growth model for the analyses of growth. The Aswt (upper asymptote) had high heritability, consistent with that found by Mignon-Grasteau et al.  who used the Gompertz model in chickens. We found low heritability estimates for tmid and scale which was not in agreement with the results reported by Grossman and Bohren  in chicken but the heritability estimate for tmid from our study was in agreement with the results from Le Rouzic et al.  in chicken who used a Gompertz growth model. Inconsistency in the results of the present study and the study by Grossman and Bohren  for tmid and scale could be due to the difference in species, differences between methods for the estimation of genetic parameters (based on correlation among full-sibs in Grossman and Bohren ) or because of the high margin of error reported in the study by Grossman and Bohren . The differences we observed between the estimates of growth curve parameters for males and females were similar to differences observed by Sengul and Kuraz  in white turkeys and by Barbato and Younken  in chickens.
In the present study, the genetic correlations among all the BW traits ranged from 0.86 to 0.99. Genetic correlations were higher for measurements taken close together in age and declined somewhat as the measurement were taken farther apart in age. Similar results on genetic correlations among multiple BW traits were reported by Kranis et al. and Chapuis et al. [12, 26], who applied various mixed models and performed multivariate analyses. We found high genetic and phenotypic correlations among all the BW traits and the Aswt; the correlations generally increased as the age of the birds increased. Genetic and phenotypic correlations between the BW120 and Aswt were both found close to 1, reflecting the similarity of the upper asymptote and BW at the later ages. The parameters tmid and scale showed a strong positive genetic and phenotypic correlation while both have negative genetic and positive phenotypic correlations with Aswt. The negative genetic correlation between Aswt and tmid is considered favorable since individuals with high Aswt will take less time to reach tmid making that individuals with high asymptotic weight can be identified earlier. Similarly, positive genetic correlation between tmid and scale is also considered favorable and logical because for birds that take less time to reach 50% of the asymptotic weight we will automatically see shrinkage in the scale. A smaller value for scale also means asymptotic weight will be approached earlier. In other studies a negative genetic correlation was also observed between Aswt and exponential rate of decay of the specific growth rate (k) by Mignon-Grasteau et al.  and between Aswt and scaling parameter by Narinic et al.  who applied the Gompertz model in their work on chickens and quails respectively.
In our study, pHu showed highly negative genetic correlations with PDL, a* and b* whereas correlation with L* was moderately negative. These negative genetic correlations of pHu were in agreement with the previous work of Le Bihan Duval et al. on turkey and chicken [2, 5, 37]. The increase in positive genetic correlation of PDL with BW traits at later ages could be due to the increase in glycogen contents of breast muscles with age, which also had a strong negative genetic correlation with pHu . The negative genetic correlation of pHu with L* and b* would explain off color meat (PSE) with low pHu and high drip loss and vice versa which was in agreement with the results from previous studies [6, 7].
In our study, both the PDL and PBM were recorded in percentages, and the genetic and phenotypic correlations between these traits were close to zero. We found that PBM had positive genetic and phenotypic correlations with BrL and BrW. The high genetic and phenotypic correlation between BrL and BW traits was also observed by Adeyinka et al.  in chickens. The positive genetic and phenotypic correlation of PBM with BrL and BrW will be useful in selection for increased PBM which is an important trait but can only be recorded after the animal is killed.