The best wearing rails, which often give contradictory results with the tensile test, were comparatively pure manganese steels, low in silicon, only exceptionally up to 0.2 per cent., but generally below 0.1 per cent., and with less than 0.1 per cent. of phosphorus and sulphur. On the other hand, rails with a tendency to break or split are low in carbon, with variable proportions of manganese, but contain much silicon, 0.3 to 0.9 per cent., and often above 0.1 per cent. of phosphorus. Another series of experiments upon rails for the Finland lines made by the author in 1879-80 shows the high quality of manganese steel. These are essentially highly carburized (0.3-0.4 per cent. carbon) with 0.7 to 1.4 per cent. manganese, and have stood three and a half years' wear without a single one being broken; while those of silicon steel with 0.106-0.144 per cent. carbon, 0.592-0.828 manganese, and 0.423-0.435 silicon have failed in many cases, showing a great tendency to split. In both of the latter instances, however, the figures deduced from tensile tests of both good and bad specimens were substantially the same. The causes of the difference between the two kinds of steel the author attributes to differences in the structure of the ingot due to the agent used in "chemical consolidation," which may be either manganese or silicon, which structures are illustrated by photographs of ingot fractures. When silicon is used there is a tendency to unsoundness about the exterior of the ingot, which is surrounded by a honeycomb-like cellular casing of greater or less depth; while with manganese the vesicular cavities are more or less dispersed through the whole substance, or concentrated toward the interior of the ingot. Rails made from the former are, therefore, more likely to contain unsound portions near the outer wearing surface, and to give unsatisfactory results in wear, than those from the latter; but as the test pieces are usually cut from the center of the railhead, the tensile resistance of the interior may be equal to or surpass that of the superior material. In summing up his observations the author concludes that the method of tensile testing is mainly of value in determining the quality of the material, but that for the finished product properly arranged falling weight tests are necessary. He also considers that the test pieces should be flat bars of 2.5 to 3.5 centimeters in area, cut as near as possible to the outer surface of both head and foot of the rail. He reprobates especially the research for microscopic imperfections (mikrobensücherei) upon the fractured surfaces, as an annoyance to the producer, and perfectly useless to the consumer.--_Stahl und Eisen_, vol. iv., page 608; through _Proc. Inst. Civ. Eng_. * * * * * A NEW FORM OF SMALL BESSEMER PLANT. By Mr. A. TRAPPEN.