The behavior of human circadian rhythms could be interpreted within the two-oscillator regime: one for the circadian pacemaker driving temperature/plasma melatonin rhythm and the other for the sleep-wake rhythm, tentatively called the SCN (suprachiasmatic nucleus) oscillator and non-SCN oscillator, respectively. Recently, the existence of the non-SCN oscillator was demonstrated through showing the possibility of non-photic entrainments by the shifted sleep schedule, and its interactions with the SCN oscillator were disclosed through analyzing the re-entrainment processes. Based on these experimental results at the behavioral level, we developed a phase oscillator model consisting of mutually coupled SCN and non-SCN oscillators, and an extra-oscillator representing an overt sleep-wake rhythm. Our model successfully reproduced the experimental results of the non-photic entrainments. In addition, our phase oscillator model and the widely accepted two-process model were shown to share a possible mechanism underlying interactions found between the SCN and non-SCN oscillators during the re-entrainment processes. Therefore, it was suggested that the interactions indicated essential dynamics of the SCN and non-SCN oscillators. Then, a possible mechanism underlying the interactions at the behavioral level was reduced into the conditions of coupling strength and intrinsic angular velocity in the coupled realistic oscillators. These conditions led us to the hypothesis at the organ level that a part of the SCN oscillator is non-photically entrained through being coupled with the non-SCN oscillator. Of course, its reality should be assessed further. This top-down approach is one of practical strategies for modeling the hierarchical biological system.