Many studies have been devoted to investigate the effects of AlN filling on the thermal conductivity of AlN/polymer composites. However, inconsistency was often found among their experimental results and many of them did not report some important experimental parameters (e.g., oxygen content of AlN), thus limiting their applications. In this work, we investigate the effects of essentially all the important experimental parameters on the thermal conductivity of AlN/epoxy composites including particle size, oxygen content, filler content and surface treatment by using the AlN particles synthesized by a SHS method developed in our previous study. The results show that a higher filler content, a larger filler particle size, a lower oxygen content and a surface treatment of the filler result in a higher thermal conductivity of the composites. The most influencing factor affecting the thermal conductivity of the composites was found to be the volume content of the filler. Particle size shows a pronounced effect only when the filler content is increased to higher than~40vol%. Oxygen content and surface treatment do not play important roles until the filler content reaches~ 50vol%. Many models have been developed to predict the thermal conductivity of composites with inorganic fillers. However, most of them consider only the effects of filler content and sphericity of the filler particles. Filler particle size, which has a significant effect on the thermal conductivity of AlN/polymer composites at high filler contents was ignored in these models. In this study, we developed a particle size factor and incorporated it into Bruggeman theoretical model. This modification enables the model to predict the effect of filler particle size on the thermal conductivity of the composites.