The phase structure of softening coal upon heating is discussed based on the relationship between the shear-rate-independent viscosity (eta) and the fraction of liquid (phi(1)) reported in the first part of this series paper, as well as results of previous studies on rheological property and physical structure of softening coal. Firstly, the existing models that assume softening coal to be a suspension of 'rigid' solid particles are examined focusing on their applicability to description of the observed of log(eta) - phi(1) relationship. The examination reveals that none of the models can describe the particular features the log(eta) - phi(1) relationship: linearity and insignificant sensitivity to temperature Apparent success of the existing models is attributed to problems such as no or inappropriate experimental definition of the liquid fraction as well as unsuitable estimation or measurement of the viscosity. Secondly, it is demonstrated that the linear log(eta) - phi(1) relationship is explained best by considering that the softening coal is a suspension of 'deformable' solid. Thirdly, further examination of the new suspension model draws the following characteristics of the phase structure of the softening coal. The so-called plastic domain in the softening coal consists of the liquid phase, optically isotropic solid phase and anisotropic mesophase-like spherules. The isotropic solid phase and the anisotropic phase, both of which are detected as the solid by H-1-NMR, would have very similar viscosities so that their influences on the viscosity of the plastic domain are undistinguishable. The plastic domain is suspended with coarse grains of minerals and inert organics, while they are not responsible for the change in the viscosity of the whole softening coal upon heating. The viscosity is mainly a function of the solid/liquid fractions in the plastic domain. (C) 2003 Elsevier Science Ltd. All rights reserved.