Snyder, R. J. and R. E. Larson.
A study was initated in September, 1952 to determine the nature of the inheritance of the tomato. In order to make accurate and unbiased comparisons between mean seed weights of inbred, hybrid, and segregating generations, it was necessary to eliminate, insofar as possible, differential environmental influences.
Seed weights are directly affected by the influence of their immediate environment within a given fruit, and by the initial sizes of the ovules from which they develop on the maternal parent. In this study, the "maternal effect" was eliminated completely by producing inbred backcross and hybrid seed populations on each of the parental lines and by producing backcross and F2 seed populations on the reciprocal hybrids. Comparisons between the various inbreds and generations were restricted to those produced on the same maternal parent.
In studies concerning the relation between seed size and heterosis in the tomato, Luckwill (New Phyt. 38:181-189, 1939) and Hatcher (Ann. Bot. n.s. 4:735-764, 1940) found that seed weight is dependent upon the number of seeds per fruit and fruit weight. Therefore, a multiple regression and covariance analysis involving those generations produced on the same maternal parent would provide a method for adjusting mean seed weights to a common seed number and fruit weight. This presents a means of eliminating differential influences from these sources of variation. In the present study, however, the covariance analysis was not directly applicable because, (1) limitations in the number of fruits available for sampling prevented unbiased random samplings, and (2) seed weight variances for those generations produced on the same maternal parents were not homogeneous.
A method used by Hatcher a (loc. cit.) whereby similar values (E values) obtained from
Fruit Weight
__________________
Number Seeds/Fruit
ratios are selected, appeared to offer a means for eliminating bias in the
selection of seed weights since highly significant positive correlations
between these E values and mean seed weights were obtained. Simple
correlation coefficients for the relationships between E values and mean seed
weights for each of the inbreds and generations are presented in the
accompanying table. Eight of the twelve correlation coefficients were highly
significant. A highly significant average correlation coefficient indicated
that approximately sixty-six per cent of the variability in mean seed weights
was associated with the variabiliy in E values.Selection of fruits having similar E values both within and between generations produced, on the same maternal parent, resulted in complete elimination of differential environmental influences of the fruits on seed weights. Neither the variance nor the adjusted mean seed weights obtained by the covariance analysis differed appreciably from those obtained by the simple analysis of variances. However, the small number of fruits available for certain generations limited the selection of similar E values to those between generations. In such instances, the covariance analysis was useful in reducing the "within generation" seed weight variances and provided a method for adjusting seed weight means to a common seed number and fruit weight.
Simple correlation coefficients for the relationships
between E values and mean seed weights per fruit for
each of the parents and generations.
_______________________________________________________________
Parents and Parents and
Generations r Generations r
_______________________________________________________________
P3 0.8971** (P2 x P3) P2 0.6431
P3 x P2 0.8304** (P2 x P3) P3 0.8278**
P3 (P3 x P2) 0.8940** (P2 x P3) self 0.8498**
P2 0.8985** (P3 x P2) P2 0.4586
P2 x P3 0.4579 (P3 x P2) P3 0.9507**
P2 (P2 x P3) 0.5224 (P3 x P2) self 0.8939**
_______________________________________________________________
* Significance at 5% level = 0.666
** Significance at 1% level = 0.798