The growth history of a macroscopic crack at high temperatures is studied using a crack growth criterion based on void growth and coalescence. Two models of crack growth are considered: a discrete model in which the crack advances by discrete jumps, each with a length equal to the void spacing, and a continuous model in which the crack advances continuously because of the accumulation of damage ahead of the crack tip. The resulting crack growth histories based on the two models are compared. Our results indicate that, irrespective of the void growth mechanisms, the continuous and discrete models give practically the same results. The effect of void nucleation is included in the analysis by the requirement that no void growth occur below a critical void nucleation stress. For the continuous crack growth model, an exact solution for the entire crack growth history is obtained as a function of the path-independent integral C* for a wide variety of void growth mechanisms. Both discrete and continuous models show that crack growth exhibits stop, and go, behavior before approaching steady state. The amplitude of these stop and go episodes is shown to decrease exponentially to zero with a characteristic distance R, where R is the nucleation distance. It was found that the steady-state crack growth rate is approximately proportional to C* and independent of the void growth mechanisms. The authors also show that the presence of a breakpoint commonly observed in the experimental crack extension rate versus C* curves can be explained by examining the entire crack growth history predicted by the models.