Maize is one of the most important crops contributing to grain production in Argentina. National production has strong fluctuations between years because of climatic variability affecting crop growth and yields. It is broadly recognised that ENSO contributes highly to interannual variability of climate and its signal has been identified in our cereal production region. Several authors found that precipitation above normal values occurs in the period November-February during El Niño years, but precipitation under normal values are identified during La Niña years for the period October-December. Consequences for crop growth and yield should be different in each phase. In this work historical maize yields from Pergamino, a site representative of the main maize production region, are used to assess the relationship between grain yields and ENSO phases during the period 1925-1994. Two approaches in defining ENSO phases were selected, one based on SOI variations and the other in SST changes. Results obtained shows that upper and lower yields are related to El Niño and La Niña years respectively but a stronger association between extreme yield values and ENSO phases was obtained with the JMA SSTA ENSO definition. Precipitation anomalies for November and December presented a strong association with yield variations in Niño and Niña years. During Niña years the determinant month seems to be November but during Niño events should be more important precipitation occurred during December. To assess the impact of weather on crop production the functional model CERES-Maize, that has been previously calibrated and validated for local conditions, was used. Outputs obtained using historical climatic data and those obtained by a conditioned ENSO-phase weather generator were compared. Observed climatic data gave good estimation for medium and high yields, but simulated yields showed greater variability, in particular for the lowest levels. We found that this lower range of yields was coincident with higher not harvested areas, in particular in Niña years, and this led to overestimate observed grain production. Using generated climatic data maize yields were underestimated all over the range of values. As the weather generator was developed using climatic data from the last century, long term variability (in particular dry periods) was included, leading to generate lower rainfall patterns and consequently lower yield levels