The coarsening of theta -Al2O3 crystals to a 'critical size' is often interpreted as the first step in the shear nucleation of alpha -Al2O3. The existence of this so-called critical size has also been used to explain the observation that alpha -Al2O3 "nuclei" are generally twice as large as the crystals in the theta -Al2O3 matrix. This paper discusses the important issues in the nucleation of alpha -Al2O3 from theta -Al2O3. A few key experiments are also presented to clarify the nucleation process. It is concluded that a critical theta -Al2O3 crystal size is not a prerequisite for alpha -Al2O3 nucleation, but is primarily a result of the incubation time required to produce alpha -Al2O3 nuclei by diffusional nucleation. It is proposed that the large observed alpha -Al2O3 crystal size also does not result from a shear nucleation event in a 'critical size' theta -Al2O3 crystal, but is due to the intrinsically low alpha -Al2O3 nucleation density, together with rapid growth of alpha -Al2O3 after nucleation.