[en] For fracture mechanics assessment of the safety of metal structures some material characteristics are necessary, which should be independent of the dimensions of the tested specimens. For the required minimum dimensions of the specimens some conditions are known, which, however, are inexact or contradictory, especially in the case of elastic-plastic materials behaviour. In this work the influence of specimen dimensions on elastic-plastic fracture mechanics values is investigated for structural steel in three states of heat treatment (normalized, quenched and tempered, coarse grain annealed). Standard specimens with greatly differing dimensions are tested according to J-integral- and COD-standards and are evaluated with regard to fracture initiation. Conditions for specimen dimension for structural steel are derived from the test results, depending on micro-structure and test temperature. If these conditions are not satisfied, the material characteristics measured in the upper shelf region are somewhat too low. Testing at different temperatures shows here, that there is a contrary influence on the material characteristics by specimen size and test temperature. The comparison of J-integral- and COD-test results shows that both methods produce practically equal fracture toughness values and that there is plane strain condition in the specimens at fracture initiation. From plates of the above-mentioned quenched and tempered steel several pressure vessels were welded. Artificial flaws were put into them and they were subjected to burst tests. The burst behaviour of the pressure vessels can be assessed with elastic-plastic material characteristics and fracture mechanics calculations. (author)