Most of the existing plotting position formulas have been proposed for use in the analysis of systematic flood records, but little has been reported on the plotting formulas for historical or non-systematic flood samples. In particular, no previous investigations have specifically examined the probability plots for the Pearson type I I I (P3) distribution in the analysis of historical flood information. The present paper suggests a new plotting position formula for the P3 distribution for use with both systematic and historical flood records. The proposed formula has a simple structure as do most existing formulas, but it is more flexible because it can take explicitly into account the skewness coefficient of the underlying distribution. Further, results of graphical and numerical comparisons have demonstrated that the suggested formula provided the least bias in flood quantile estimation as compared with many available plotting formulas, including the well-known Weibull formula. Finally, results of a numerical example using actual flood data have indicated the practical convenience of the proposed plotting formula. It can be concluded that the formula developed in this study is the most appropriate for the P3 distribution in the analysis of flood records considering historical information. 138 VAN THANH-VAN NGUYEN AND NOPHADOL IN-NA Nguyen et al. (1989) have proposed an unbiased plotting formula specifically for this distribution and for the analysis of systematically recorded flood series. The present study, a continuation of a previous one, emphasizes the development of a more general plotting position formula for the P3 distribution and for use with both systematic and historical flood records. Results of graphical and numerical comparisons have demonstrated that the plotting position formula proposed in this study can provide the least bias in flood quantile estimation as compared with several existing formulas. Note that only the bias in flood quantile estimates is examined in this study. The question of whether plotting positions should be unbiased in terms of exceedance probabilities or unbiased in term of flood estimated is not addressed here. This question should be best considered in the context of the use which is to be made of the results. Such consideration is beyond the scope of this paper. Nevertheless, the advantages related to the use of unbiased plotting positions in the estimation of flood quantiles have been shown in many previous works (see, e.g., Kimball, 1960;Cunnane, 1978). Further, the flood record length in the present study is assumed to be known. The effect of misspecification of the true record length on flood quantile estimates have been previously studied by Hirsch andStedinger (1987), andHirsch (1987).