Ensemble numerical simulations are performed using an atmospheric general circulation model to study the responses of the atmosphere to volcanic aerosol forcing and sea surface temperature (SST) forcing during the two years following the Pinatubo eruption. The simulated surface air temperature (SAT) anomalies, forced respectively and jointly by the Pinatubo aerosol and the observed SST anomalies, are compared to those derived from observations using a singular value decomposition (SVD) model. The simulated land SAT anomalies, forced only by the observed SST anomalies, match well in magnitude and time the leading mode of SAT anomalies from the SVD analysis, which represents the ENSO-forced SAT variations. When forced by both the Pinatubo aerosol and the observed SST anomalies, the model captures the observed surface cooling over land in June--July--August and September--October--November 1992 but misses the surface warming over northern Eurasia in December--January--February (DJF) 1991--1992 and DJF 1992--1993. In the atmosphere the simulated temperature anomalies in the troposphere (stratosphere) forced by the Pinatubo aerosol are sensitive (insensitive) to the initial conditions and the type of prescribed SST. The signal of SST forcing is stronger in the troposphere and near the surface than in the stratosphere, while the signal of the Pinatubo aerosol forcing is strongest in the lower stratosphere. The simulated global-mean stratospheric temperature anomalies induced by the Pinatubo aerosol are generally 1¿C to 1.5¿C larger than observed. Empirical data analyses and numerical model simulations showed that this discrepancy is explained in part by the influences of the quasi-biennial oscillation and the observed ozone depletion, which the ensemble numerical simulations did not resolve.