A theory of the origin of intraseasonal oscillations of the tropical atmosphere is presented and tested by simple model experiments. This study forces on the validation of the basic theory against key features of the observed 40–50 day oscillation. It is shown that the observed eastward propagation of intraseasonal oscillation in the tropical atmosphere arises as an intrinsic mode of oscillation resulting from an interaction of convection and dynamics via the so-called “mobile” wave-CISK mechanism. Through this mechanism, the heat source feeds on the east-west asymmetry of forced equatorial waves. As a result, Kelvin waves are selectively amplified, which in turn causes the heat source to propagate eastward. This mechanism also prevents small-scale waves from immediate destabilization, contrary to the results of traditional wave-CISK theory. The “mobile” wave-CISK establishes a new dynamics equilibrium state between convection and the wind field to form a wave packet or collective motion with relatively fixed horizontal and vertical structure. Relative to the steady state solutions with stationary heat source the new equilibrium state has suppressed Rossby-wave resonance to the west and enhanced Kelvin-wave response to the propagating heat source. Results also suggest that the periodicity of the oscillation is determined by the time taken for the Kelvin wave to complete one circuit around the globe in the equatorial region. The propagation speed (∼19 m s−1) of the model disturbance, which is about twice as fast as the observed, is found to coincide with the real part of the complex phase speed of the model's unstable normal mode modified by internal heating. The speed and the growth rate are dependent on the vertical structure of the heating profile and the static stability of the basic gate. In addition to the eastward propagation, many observed features, such as pressure and wind distribution, amplitude modulation by SST, and dominance of low wavenumber response, are well simulated in the idealized experiments. The theory also predicts that the low-frequency disturbance should have a westward tilt with height. This is partially confirmed in real observation and in GCM simulations. While the basic theory appears to explain some fundamental feature of the 40–50 day oscillation, large discrepancies still exist The possibility of examining further detailed features of the oscillation in the present theoretical framework is also discussed.