Minimal attention has been paid to the factors influencing compositional shifts in residual gas during the desorption process from coal. Two anthracite samples were collected from the Zhina Coalfield in China to study compositional shifts in residual gas. The entire desorption process was divided into five stages, comprising thirteen substages. Low-pressure N-2 adsorption tests showed that mesopore and macropore volumes and surface areas increased continuously as the particle size of pulverized coal decreased. The increase in porosity with decreasing particle size is attributable to the existence of many closed pores in the coal. The residual gas contents of the two samples were 8.34 and 7.22 cm(3)/g ("as-received" basis). The desorption rate of residual gas gradually declined with time in a 95 degrees C thermostatic water bath. Measured concentrations of the chemical gas components CH4, C2+, N-2, and CO2 show distinct variations during the desorption process. The factors influencing compositional shifts in residual gas are ambient temperature, heating time, and pulverized coal particle size. N-2 concentrations were highest during the pulverization process and relatively low during the heating process, which is in contrast with the variation trend of C2+ and CO2 concentrations. CH4 and N-2 concentrations decreased steadily with heating time in a 95 degrees C thermostatic water bath, whereas C2+ and CO2 concentrations showed the opposite trend. This is because C2+ and CO2 have stronger adsorption affinities than CH4 and N-2 do, and a large proportion of N-2 and CH4 could desorb preferentially from coal. C2+ and CO2 concentrations increased with decreasing coal particle size, whereas CH4 and N-2 concentrations showed the opposite trend. Relatively high concentrations of CO2 and C2+ in closed pores may be related to the sealing effect of closed pores, which prevents the escape of early generated products. In addition, increased pressure in closed coal pores should retard C2+ cracking at high temperatures. Thus, it is necessary to focus on the compositional shifts in residual gas and its influencing factors. Failure to account for these aspects may lead to typical analytical errors, which will propagate in the assessment of CH4 abundance in coalbeds.