ACCURACY ATTAINABLE. 13 



layer of Canada balsam. This method of blackening the vanes was 

 found easier and more satisfactory than smoking them. 



The sensitiveness was about 10 cm. deflection per square millimeter 

 of exposed surface, scale and candle at I meter. The weight was 10 

 milligrams. The vane was made heavy to avoid tremors. 



ACCURACY ATTAINABLE. 



The writer has been so frequently questioned on the accuracy of the 

 observations that a few remarks will be in order. 



First of all, nearly all the substances examined have numerous 

 absorption bands, so that we can not find an extinction coefficient by 

 means of which we may determine their transmissivity in absolute 

 value for, say, a centimeter thickness. Again, different samples of a 

 given mineral will vary in homogeneity and in purity. Different parts 

 of a given specimen will vary in thickness, homogeneity, purity, and 

 polish, so that, on remounting the same sample before the spectrometer 

 slit, the relative values of transmission may differ by several per cent. 

 The result is that such a comparison has but little meaning, and is at 

 all times of minor importance. But, after once mounting the specimen 

 before the spectrometer slit, the important question is the relative trans- 

 parency of different regions of the spectrum and the accuracy of the 

 location of the maxima of absorption. This depends upon the accuracy 

 of the observations in any region of the spectrum. With a trust- 

 worthy instrument like the radiometer, and with the specimens securely 

 mounted before the spectrometer slit, in the present research it was 

 found that for a given spectrometer setting the variation of single 

 observations would differ from the mean by only about a tenth of i 

 per cent, so that only in the extreme infra-red, or in locating sharp 

 maxima, were the observations repeated. Consequently the reality of 

 many of the small maxima can be accepted with confidence. The 

 most important question, however, is the accuracy of the location of 

 the maxima, especially those which are sharp and well defined. 



During the past few years the writer has repeatedly mounted and 

 adjusted his apparatus and has always found the location of certain 

 sharp absorption and emission bands, used as standards of reference, to 

 agree to 0.02 p., which is as accurate as our knowledge of the dispersion 

 of rock salt will permit. The adjustment of the zero of the instrument 

 was tested daily. It was found that the yellow helium line is narrower 

 and brighter and thus better adapted than sodium for making this 

 setting. 



