PNL Volume 13
Swiecicki, W. K., Plant Experiment Station, Wiatrowo, Poland
P. Blazczak, J. Hauke, and S. Mejza Agricultural Academy, Poznan, Poland
After several years of research work at Wiatrowo to breed cultivars with
a higher protein content, our progress has been less than expected. The basic
difficulty was the negative correlation between % protein content and seed
yield, a difficulty discussed also by Jermyn and Slinkard (Canad. J. Pl. Sci.,
1976, 56:427). To learn more about the problem we decided to subject
the results of breeding tests to statistical analysis. First a preliminary
analysis of correlation and regression was made. The analysis was based on
the observations, from 1973-76, of seed yield and % protein content in four
crosses: L 2813 x L 6003 (mean protein content 28.9%), L 3509 x L 2813 (30.6%),
L 2815 x L 2813 (29.4%), and L 2813 x L 4318 (30.9%). In the first stage
the significance of the coefficient of correlation between seed yield and
% protein content for particular hybrids and for particular years was studied.
There was a high, negative correlation always above (according to the absolute
value) -0.95 between these factors.
Next the coefficients of correlation and regression were analyzed over
all years. There was general lack of homoegeneity among the coefficients
of correlation, reflecting differences in some or all of the variables studied.
Crosses were also compared with respect to coefficients of correlation and
regression. Three of these crosses showed substantially different correlation
in all possible clusters of years whereas L 2813 x L 4318 showed a certain
regularity (homogeneity of coefficients of correlation) over years.
From a breeding point of view, it was of interest to know if the negative
correlation between seed yield and % protein content was equally strong at
different levels of protein content. In order to study this problem, the
cross L 3503 x L 2813 was chosen because the most data were available. Using
a computer, the data were searched for those instances in which higher seed
yield was not associated with lower protein content. This analysis showed
that when protein content was not more than 27% a larger seed yield did not
cause significant reduction of protein content. Since this conclusion was
based on only one cross, we checked for verification in following years.
Coefficients of correlation were analyzed and coefficients of equations of
regression were compared for other crosses in which protein content was not
more than 27% (1977 - 17 crosses, 1978 - 7 crosses, 1979 - 15 crosses). The
results from 1979 (Table 1) are given as examples verifying the conclusions
as stated.
The conclusions presented above show the possibilities of increasing
the protein yield in cultivars, either by improvement of yield when protein
content is not more that 27%, or by increase of protein content to the
threshold level of 27% in high-yielding cultivars. For example, some varieties
bred in Poland give a seed yield of about 3-4 tons/hectare, yet their protein
content is only 22%. If, according to the above hypothesis, the protein content
is increased to the theoretical 27%, protein yield would be greater by 150-200
kilos per hectare.
These observations demonstrate how difficulties faced in routine breeding
inspire more detailed investigations.
PNL Volume 13 1981