Fracture toughness results are often connected with varying thickness effects. Many results are contradictory, with some results indicating no size effect, some indicating increasing toughness with decreasing thickness and others indicating decreasing toughness with decreasing thickness. In this paper the causes for different thickness effects are discussed. It is shown that most of the observed size effects are due to invalid tests. The only test parameters to be regarded as valid are those that correspond to the initiation of crack extension. The theoretical thickness effects for both ductile and brittle fracture initiation are evaluated. It is shown that ductile fracture initiation is thickness independent if B ≥α(J/σy). The thickness effect in brittle fracture is explained by applying a mechanism-based statistical cleavage fracture model. A theoretical thickness correction for cleavage fracture is derived, and its validity is confirmed for a variety of materials.