Reverberation is an important cue for determining the distance and location of sounds, and can be used to infer the size of a space. Although simple models of reverber- ation assume a single exponentially decaying late rever- beration tail that is independent of direction, in practice, many rooms feature multiple exponential decays with di- rectional anisotropy [1, 2, 3]. In this paper, a framework for the directional analysis of spatial room impulse response reverberation decays is presented. The framework uses a recent neural net- work approach for multi-exponential decay analysis in conjunction with a spherical filterbank analysis. Addi- tionally, the paper introduces the common-slope model of directional reverberation, in which directional or spa- tial decay variations are described in terms of exponential amplitudes, while fixing the corresponding decay times.