The J -integral has been proposed as an engineering fracture criterion for elastic-plastic behavior in metals. The J -integral, as is defined, has several advantages over some other currently used fracture parameters. But the experimental evidence so far obtained to show its practical usefulness have been limited. Using an ordinary ship steel, the authors investigated some practical aspects of the J -integral as a fracture criterion. The value of J was evaluated from several different methods including numerical (finite element analysis) as well as experimental (load vs. load point displacement record) procedure. The specimen used are center-notched tension, compact tension and bend specimens with varied dimensions. The critical values of J were evaluated from brittle fracture tests using these specimens. The method to evaluate J -value from single load-displacement record, originally proposed by Rice et al., has been modified by the authors to yield better results. Relation between J and COD and their equivalence are discussed. It is concluded that J concept is almost equivalent to the COD concept and thus J_C would be one of the useful measures of fracture toughness, especially in case of fracture under large scale yielding condition.