VARIOUS steels, or even different heats of the same steel, with identical mechanical properties when determined by ordinary tests, vary considerably in behaviour in service; and such differences of performance are neither due to any variation in composition nor to any other known factors. That these differences exist is now known but their causes are still very much matters for metallurgical conjecture. It is known that steels tempered from the austenitic range and tempered to give a martensitic structure become increasingly ductile with rise of temperature. The ordinary tests, however, fail to give any indication of the variations in behaviour, on tempering, between even different low‐alloy steels, to say nothing of different heats of any one particular steel. The notched‐bar impact‐bend test is capable of revealing differences in impact properties among steels which have been tempered to levels of mechanical strength identical with the results obtained by ordinary tensile tests. The notched‐bar test is, however, highly complex, involving a certain amount of tension, compression, bending and shear, and its full significance is not yet clearly understood. Of its three characteristic features: high speed, notch, and bend action, the notch is the most potent embrittling factor. Recent investigations have shown that static notched‐bar tensile tests give more fully understandable results and evaluate metals in very much the same manner as impact tests.