In this study, the coupled atmosphere-surface climate feedback-responses analysis method (CFRAM) is used to quantitatively assess relative contributions of radiative and non-radiative processes to the observed temperature anomalies associated with the El Nino. In this approach, based upon the total energy balance in atmosphere-surface columns and linearization of radiative energy perturbations, local temperature anomalies are decomposed into partial temperature anomalies due to changes in oceanic dynamics/heat storage, water vapor, clouds, atmospheric dynamics, ozone, and surface albedo. Such decompositions are conducted for historical El Nino events using the ERA-Interim reanalysis and also for El Nino events in the historical climate simulations made with 16 coupled general circulation models (CGCMs) participating in the Coupled Model Intercomparison Project phases 5 (CMIP5). It is found that although most of the CGCMs analyzed are able to simulate the canonical pattern of SST anomalies and global atmospheric temperature anomalies associated with El Nino, the relative contributions from various radiative and dynamical feedbacks to such temperature anomalies vary widely among models and also tend to differ from what the observations (ERA Interim) suggest. The implications of the inter-model and model-observation differences for steps toward improving model simulations of the ENSO variability and projections of future changes in ENSO will be discussed.