In the first part of this dissertation we use an Intermediate Coupled Model to develop a quantitative test to validate the null hypothesis that low-frequency varia- tion of ENSO predictability may be caused by stochastic processes. Three "perfect model scenario" prediction experiments are carried out, where the model is forced ei- ther solely by stochastic forcing or additionally by decadal-varying backgrounds with different amplitudes. These experiments indicate that one can not simply reject the null hypothesis unless the decadal-varying backgrounds are unrealistically strong. The second part of this dissertation investigates the extent to which internal atmospheric variability (IAV) can influence ENSO variation, and examines the un- derlying physical mechanisms linking IAV to ENSO variability with the aid of a newly developed coupled model consisting of an atmospheric general circulation model and a Zebiak-Cane type of reduced gravity ocean model. A novel noise filter algorithm is developed to suppress IAV in the coupled model. A long control coupled simulation, where the filter is not employed, demonstrates that the coupled model captures many statistical properties of the observed ENSO behavior. It further shows that the development of El Ni~no is linked to a boreal spring phenomenon referred to as the Pacific Meridional Model (MM). The MM, character- ized by an anomalous north-south SST gradient and anomalous surface circulation in the northeasterly trade regime with maximum variance in boreal spring, is inherent to thermodynamic ocean-atmosphere coupling in the Intertropical Convergence Zone latitude. The Northern Pacific Oscillation provides one source of external forcing to excite it. This result supports the hypothesis that the MM works as a conduit for extratropical atmospheric influence on ENSO. A set of coupled simulations, where the filter is used to suppress IAV, indicate that reducing IAV in both wind stress and heat flux substantially weakens ENSO variance. Furthermore, the resultant ENSO cycle becomes more regular and no longer shows strong seasonal phase locking. The seasonal phase locking of ENSO is strongly tied to the IAV in surface heat flux. The ENSO cycle is strongly tied to IAV in surface wind stress.