The Hypothesis that low-frequency variation of ENSO predictability may be caused by stochastic processes is tested using an Intermediate Coupled Model (ICM). Three perfect model scenario prediction experiments were carried out, each of which consists of large ensemble predictions. In the first case (reference case), the model is solely forced by stochastic forcing without any explicit decadal-varying background. In the other two experiments, decadal-varying backgrounds with different amplitudes were introduced into the ICM in addition to the stochastic forcing. By comparing the low-frequency variations of ENSO predictability in these experiments, it is found that only when the amplitude of the decadal-varying background is unrealistically strong, the effect of decadal-varying background on ENSO predictability is comparable to or larger than the stochastically induced low-frequency ENSO variation. Therefore, to the extent that the simple coupled system can represent reality, it seems that one can not simply reject the null hypothesis that the decadal variation of ENSO predictability may be stochastically induced.