Rice-based (Oryza sativa L.) rainfed lowlands are the major cropping system in northeast Thailand. Average yields are low, which is generally explained by frequent drought events, low soil fertility, and poor fertilizer response. However, neither the relative importance of these factors nor their interaction is well understood. Therefore, we analyzed an existing database on fertilizer trials conducted between 1995 and 1997 at eight different sites in northeast Thailand with the objective to determine indigenous nutrient supplies, internal efficiencies, and recovery efficiencies of applied nutrients in rainfed lowland rice. Of particular interest was the effect of variety type (traditional) and water supply on these components. Comparison of N, P, and K concentrations in grain and straw (average N-P-K grain concentration of 11.0-2.7-3.4 g kg(-1); average N-P-K straw concentration of 5.2-0.9-16.4 g kg(-1)) in the traditional-type varieties used at all trial sites with literature values showed no differences for these parameters between traditional and modern-type varieties or between irrigated and rainfed environments. In contrast, internal efficiencies of N, P, and K (average IEN: 46 kg grain per kg N uptake; IEP: 218 kg grain per kg P uptake; IEK: 25 kg grain per kg K uptake) were much lower than reported for irrigated systems, and the difference was greatest for K, which is mainly accumulated in the straw. Indigenous nutrient supply (average INS: 38 kg ha(-1); IPS: 10 kg ha(-1); IKS: 89 kg ha(-1)) and recovery efficiency (average REN: 0.28 kg kg(-1); REP: 0.13 kg kg(-1); REK: 0.49 kg kg(-1)) were low but comparable to the lower values reported from irrigated systems. Average seasonal field water resources seemed to reduce the indigenous nutrient supply but had no or little effect on internal efficiency and recovery efficiency. We concluded that the main reason for the low system productivity without and with fertilizer in northeast Thailand is the dominant use of traditional-type varieties with low harvest indices, which was the dominant cause for the observed low internal nutrient efficiency. Therefore, intensification of rainfed systems through substantially increased nutrient inputs can be recommended only where varieties with an average harvest index of close to 0.4 or higher are available.