Annual-peak discharge records at 50 sites with 5 or more years of record were used to determine individual site log-Pearson Type III frequency curves. The log-Pearson Type III frequency curve values for 2-, 10-, 25-, 50-, and 100-year recurrence intervals at 37 of these sites with 10 or more years of record were regressed against basin characteristics. Drainage area and mean annual rainfall proved to be the only independent variables significant at the 95 percent confidence level. The regression equations are: Q2 = 0.033 A* 776 (Ann P) 2 ' 11 SER = +51,-34 percent Q 1Q = 3.72 A-822 (Ann P) 1-29 SE = +45,-31 percent Q25 = 25.7 A' 826 (Ann p)* 953 SER = +50,-33 percent Q50 =89.9 A-830 (Ann p)-734 SE^ = +55> _36 percent Q 10Q = 286 A* 832 (Ann P)' 531 SER = +61,-38 percent where Q is the T-year recurrence interval peak discharge, A is the drainage area in square miles, (Ann P) is the weighted mean annual precipitation in inches, and SE is the standard error of estimate of the regression. All flood data in Puerto Rico through December 1975 were analyzed. Shortterm records were extended to include historic flood records. Frequency curves were computed by methods that tend to reduce time-sampling errors. A stepwise regression technique was-used to relate flood frequency at the gaged sites to areal parameters.