Although rice accounts for approximately forty-four percent of land under cultivation and forty-six percent of caloric intake in Madagascar, most farmers cannot produce enough rice to feed their families. Total rice production increased little in the country during the 1990s, and yields were stagnant and well below world average yields. Because of the importance of rice for both family income and nutrition and because of the significant role upland rice cultivation plays in deforestation in Madagascar, intensification of lowland rice production has been a major focus of many development interventions. The System of Rice Intensification (SRI) is a method that has been promoted and closely followed in Madagascar for more than ten years. SRI is remarkable for its dramatic increases in yields achieved without external inputs. While the method is more labor intensive, the doubling or even tripling of yields would appear to make the method extremely profitable. Despite its apparent potential and intensive extension efforts in some areas, adoption rates have generally been low, and the average rate of disadoption (the percentage of households who have tried the method and who no longer practice it) for study area was 40 percent. Using survey data from five communities in Madagascar, a sample selection model and a probit model were used to study three related choices farmers must make concerning the SRI. The first is whether to try the method, and then, conditional on having tried the method a farmer must decide how much of his lowland rice area to put in the SRI. Finally, for all subsequent years, a farmer must decide whether to continue using the method. While many adoption studies have looked at the role of liquidity and labor, few have done so in a dynamic setting. We take advantage of reliable recall data and NGO records on technology use to construct a panel data set, which was used to analyze the effects of learning, labor, liquidity, and social benefits over a five-year period. Learning has been the focus of several recent adoption studies, but this study tries to separate the learning effects from the effects of compliance with authority figures and with social and cultural norms. The results show that planting-season labor and liquidity constraints are extremely important factors in the adoption decisions of Malagasy farmers. Although the method is low-external input, the poorest farmers are not able to take advantage of the technology. This result is very important for development policy-makers in Madagascar who have often seen rice intensification as the most important means by which to alleviate rural poverty in Madagascar.