An indigenous purple nonsulfur bacterium Rhodopseudomonas palustris WP3-5 was used to produce hydrogen phototrophically from acetate (HAc) and butyrate (HBu), which are the major soluble products from acidogenic dark fermentation. Statistical experimental design methodology was applied to identify optimal composition of the two carbon substrates in the medium, leading to better H2 production performance of R. palustris WP3-5. Three performance indexes were used to assess the effectiveness of the phototrophic H2 production; they were H2 yield (Y H2), maximum H2 production rate (Rmax) and maximum cumulative H2 evolution (Hmax). An overlay contour plot was used to determine the optimal concentration range of HAc and HBu, taking into account all three performance indexes (i.e., Rmax, Hmax, and Y H2) simultaneously. With the response surface analysis, R. palustris WP3-5 could produce H2 efficiently with the best Rmax, Hmax, and Y H2 of 39.5 ml/h, 2738 ml, and 51.6%, respectively. This performance is superior to most reported values in the literature, indicating that the statistical experimental design is an effective tool to improve phototrophic H2 production with R. palustris WP3-5.