The objective of the investigation was to determine the cause of wide variations (factor of 5) observed in the bulk thermal expansion of the polycrystalline graphites: ATJ‐S, AXF‐5Q, mesophase and isostatically pressed. An attempt to rationalize this spectrum on the basis of lattice thermal expansion was unsuccessful because of the relatively small differences in that lattice expansion. Addition of boron to the isostatically pressed graphite reduced the thermal expansion in the c‐direction by approximately 30%. The very low thermal expansion coefficient of ATJ‐S as compared to AXF‐5Q and the mesophase‐based graphite can be explained by the existence of very small microcracks between the layer planes in the ATJ‐S filler particles. In the AXF‐5Q and mesophase‐based graphites, very little micro‐cracking could be distinguished. The qualitative theory of Mrozowski and the more quantitative model of Buessem were examined. It was found that the results are essentially in agreement with Buessem's model. The isostatically pressed material must possess significant tessellated stresses or distortion in order to explain the very high bulk expansion.