A statistical analyses based on the maximum likelihood method are performed on cleavage fracture toughness data for A508 Class 3 forging to investigate the effect of test temperature and specimen thickness on precleavage ductile crack growth and the effect of all three on cleavage fracture toughness. For the purpose of the analyses, Weibull distributions have been assumed for precleavage ductile crack growth and cleavage fracture toughness with a threshold of zero. The analysis have shown that distributions with higher threshold levels are required for cleavage fracture toughness when precleavage ductile crack growth exceeds 0.2 mm. The equations derived for precleavage ductile crack growth and cleavage fracture toughness can be used to estimate the probability of cleavage failure in relation to prescribed values of: (i) ductile crack growth, (ii) fracture toughness for both a given amount of ductile crack growth and for precleavage ductile crack growth greater than the given value of ductile crack growth, (iii) ductile crack growth and fracture toughness. A procedure is proposed for the definition of onset of the upper shelf in relation to the three criteria given above.