ABSTRACT The applicability of three previous empirical correlations proposed for estimating relative density of sandy soils based on the SPT N-value, effective overburden stress and soil gradation characteristics was investigated in the present study by using a data base of relative density obtained from high quality undisturbed samples of fine to medium sand with Fc≦20%, D50≦1.0 mm and Dmax≦4.75 mm. All undisturbed samples were recovered by in-situ freezing method. The relative density estimated by Meyerhof's method (1957) was in the range of +15%~-45% of the measured values. Meyerhof's method (1957) was modified by Tokimatsu and Yoshimi (1983) by considering the effect of fines content on the SPT N-value. The relative density estimated by Tokimatsu and Yoshimi's method (1983) is in the range of +25%~-20% of the measured values. The underestimation of relative density of Meyerhof's method (1957) was modified. On the contrary, the overestimation of relative density is more significant than Meyerhof's method. The relative density estimated by the method proposed by Kokusho et al. (1983) is in the range of +20% to -35% of the measured values even for dense sand with a relative density larger than 60%. Meyerhof's method (1957) and the method proposed by Kokusho et al. (1983) have a common disadvantage that they will extremely underestimate the relative density of fine to medium sand for SPT N-value lower than about 8. The errors in estimation of relative density by these methods are large. A simple empirical correlation (Eq. (10)) was proposed in the present study to estimate the relative density of fine to medium sand based on the normalized SPT N-value, N1. The relative density estimated by the proposed method is in the range of +15% to -30% of the measured values for N1 in the range between 0 and 50. As a whole, the proposed method is less in errors for estimating relative density compared with those estimated by Meyerhof's method (1957) and the method proposed by Kokusho et al. (1983). Based on a data base of undisturbed samples with data of fines content obtained from the SPT spoon samples, the method proposed by the authors is again compared with the three previous methods. The relative density estimated by the proposed method based on the above data base is in the range of +15% to -10% of the measured values. Among four methods, as a whole, the proposed method shows the least errors in estimation of relative density. The proposed method was also modified (Eq. (16)) by taking into account of the effect of fines content of SPT samples. The relative density estimated by the modified method based on the fines content is almost in the range of +10% to -10% of the measured values. Two empirical correlations proposed in the present study are less in errors of estimating relative density compared with three previous methods. The range of relative density estimated by the proposed method is well consistent with the range of the measured values (40% to 90%). The empirical correlations proposed in the present study should be applied to fine to medium sand with Fc≦20%, D50≦1.0 mm and Dmax≦4.75 mm.