Abstract A comparison has been made of impact tests by the pendulum machine (Izod) and by the falling-weight machine (Izod and Charpy), using fabric board, paper board, molded resin (Bakelite), and ebonite, each in four types of test-pieces; unnotched, saw-cut notch, V-notch with 1 mm. root radius, and V-notch with 0.5 mm. root radius. The effect of notching on impact strength varies from one material to another, and may alter their order of merit. Saw-cut and V-notches with the same root radius nearly always give about the same results, and reducing the root radius usually (but not always) reduces the impact strength. The pendulum machine always gives higher results than the falling weight-machine in Izod tests, and the ratio varies with the material, though not sufficiently to alter the order of merit. Under the conditions used in the present tests with the falling-weight machine, Charpy tests (70 mm. between supports) nearly always give higher values than Izod tests (22 mm. striking distance), but both arrange the materials in the same order of merit. Standard errors are worked out for the pendulum test results; V-notched specimens appear to give, if anything, the best accuracy. The sources of variability in test results, and hence of errors in the final value, are discussed in an attempt to compare the accuracies of the two machines. Although an exact comparison could not be made, there appears to be no great difference between the two machines as regards the accuracy of the result obtained with a given number of test-pieces. The four types of test-piece differ very little, on the whole, in this respect. With the falling-weight machine, the Charpy test appears to give rather better accuracy than the Izod test. There is no definite evidence that the amount of excess energy in the pendulum test influences the result, except perhaps with Bakelite. The variability of all four materials tested is such that a maximum excess energy limit of 25 per cent cannot be adhered to without excluding either the weakest or the strongest specimens and so obtaining a false mean. In Izod tests on the pendulum machine a 20 mm. striking distance gives higher values for fabric board than a 35 mm. striking distance; the former has the advantage that no incomplete breaks occur, as happens with the longer striking distance. Attention is drawn to certain advantages and disadvantages of the two types of machine, and to the necessity for defining a “break” in the case of laminated materials, where cracks or incomplete breaks are frequent.