Side notch type CTOD tests with various specimen thicknesses including full thickness were carried out by using a narrow gap welded joint with plate thickness of 200 mm to ensure the effect of specimen thickness on critical CTOD for as-weld joint. All the specimens were intended to have a fatigue pre-cracked notch at HAZ. After completion of CTOD tests, notch location was checked for each specimen. At the same time, the peak temperature of welding thermal cycle at each point was analyzed from the macro structure of the welded joint and the part in the vicinity of the weld metal was classified approximately according to metallography due to estimated peak temperature, such as simple coarse grain region, coarse grain region with island martensite (C+M^*), simple island martensite region, simple fine grain region etc., It was investigated what kinds of region were cut across by the front line of the fatigue pre-crack and their crossing length for each specimen. As a result, it becomes clear that brittle fracture initiated from the most locally embrittle zone which was suggested by thermal cycle calculation. Moreover critical CTOD's were converted to those for full thickness CTOD specimens by using equations which represents the mathematical effect of specimen thickness on critical CTOD and was confirmed by experiments using base metals with various strength levels. From the result of the relation between a maximum crossing length of a fatigue pre-crack front for the most local brittel zone and the converted critical CTOD, it becomes clear that the critical CTOD for full thickness CTOD specimen can be estimated by using small size CTOD specimen sampled adequately from the joint.