and checked with those of subsequent researches, it is essential that the units and standards employed should have the same meaning then as now, and, therefore, that such standards should be capable of being reproduced with the highest attainable order of accuracy. We may, perhaps, say that the limit of such attainable accuracy is the accuracy with which two of the standards can be compared, and this is, roughly speaking, about one-half of a micron—some say as small as three-tenths of a micron. For such work neither of the three methods described above of producing a standard is sufficiently accurate. As before stated, the results obtained by them vary among themselves by quantities of the order of one part in 50,000 to one part in 20,000. Since the whole meter is 1,000,000 microns, an order of accuracy of one-half of a micron, which can be obtained with a microscope, would mean one part in 2,000,000, which is far beyond the possibilities of any of the three methods proposed.
We now turn to the interference method. Some preliminary experiments showed that there were possibilities in this method. The fact to which we have just drawn attention—namely, that the wave lengths are the same to at least one part in 500,000—looks particularly promising and leads us to believe that, if we had the proper means of using the waves and of multiplying them up to moderately long distances, without multiplying the errors, they could be used as a standard of length which would meet the requirement. This requirement is that a sufficient number of waves shall produce a length which may be reproduced with such a degree of accuracy that the difference between the new standard and the one now serving as the standard cannot be detected by the microscope.
The process is, in principle, an ideally simple one, and
