Zusammenfassung: BASIC programs to analyse the polymodal frequency distribution into normal distributions were studied and a Maximum-Likelihood program was compared with a Least-Squares program and its variations. The Maximum-Likehood method is the most suitable procedure for the problem. The X super(2) minimum method is more suitable than the Least-Squares method for normal data, but the latter is more suitable than the former for abnormal data which have a few separate parts at the end of a distribution. These methods are easy to apply for a good estimation. Parameters are stable where an obvious minimal value is recognized between neighboring distributions, but the confidence intervals of the parameters are larger than for the parts where it is not recognized.

Zusammenfassung: With respect to the sexual dimorphism in growth of fish, computer simulation of individual growth variation was made in terms of coefficient of variation (CV) and skewness (sk) of frequency distribution. Normal distributions of initial body length, specific growth rate in exponential growth coefficient in asymptotic growth and asymptotic value (each consisting of 500 monosexual distributions) were supposed. In exponential growth both CV and sk of the bisexual distributions increased more quickly than the two monosexual distributions. In asymptotic growth (Bertalanffy and Logistic) CV of the bisexual distribution was always greater than the monosexual distributions, and sk of the former was greater in positive value and smaller in negative value than those of the latter all the time.