1890.] MICROSCOPICAL JOUEN,\L. 231 



We next have to find the greatest errors of the intervals from the mean ; 

 G is the greatest, and S the least. .Calcnlation shows that G is 1-20,000 

 in. too large, and S 1-40,000 in. too small. But, on returning to Pow- 

 ell's scale, we find a much closer agreement than this. Taking the 

 .001 in. first, we find the mean to be 628.0. Three out of the ten in- 

 tervals agree to that mean to ± i , this being " the constant of the in- 

 strument and observer," they are without sensible error. Four inter- 

 vals agree to ±: 2, which is less than 1-300,000 in. ; two lines B and H 

 agree to ±: 3, wdiich is less than 1-300,000, and one interval G is -j-4, 

 viz., 1-157,000 in., too large. Now, as we found that ± i was the limit of 

 observation, we may say that the scale, with the exception of B, G, and 

 H, has no sensible error. Practically speaking, G is the only interval 

 that is out, and its error is small in comparison with the other scales. 



The next scale is the i-ioomm. The .01 mm. is too small a quan- 

 tity to treat in the above way ; it must be left until we have objectives 

 as perfect as those we have at present, but of double their power. All 

 that can be done is to take several of the divisions. Eight sets of three 

 each were measured on Powell's new scale ; the variation from the 

 mean was less than 1-200,000 in. Rogers' is a very well ruled scale ; 

 it is. however, difticult to observe, the lines being without pigment, 

 and it is mounted dry. The lines under these circumstances present 

 the usual black and white diffraction images. It is, on that account, 

 very difficult to maintain an equable focus during measurement. 



In Rogers' scale the greatest error is in interval G, where it amounts 

 to four divisions, or somewhat less than 1-100,000 in. Thirteen out of 

 twenty intervals have practically an insensible error. One cannot 

 speak with the same certainty with regard to this plate as to the others, 

 because of the focal difficulty. Different readings give discordant re- 

 sults ; therefore, in this case, more must be allowed for the " constant 

 of the observer and instrument." With regard to the i-ioths of a mm. 

 on Powell's scale, they were examined by a power of 600 diameters 

 with a dry lens. The mean was 987 ; six intervals had no sensible 

 error, but C and G had an error of three divisions, which is equivalent 

 to 1-100,000 in. Rogers' gave a very similar result. 



The error of the internal D, in the Zeiss scale, was 1-30,000 in. 



I next compared the length of the mm. on the three scales, that is, 

 Powell's, Rogers', and Zeiss' with each other. I detected a slight but 

 insensible difference of ±r. All that now remains to be done is to 

 compare the inch and the mm. scale on Powell's plate, liy measure- 

 ment we foand that 30// gave a screw value of 741-25 ; therefore, the 

 value for .1 mm. would be 2470.8, and the value for 1-1,000 in. .001 X 

 2470.8 ^ .003937 = 627.59. 



The value actually measured was, as we saw above, 628.0. Here 

 again there is no sensible discrepancy. In conclusion, 1 feel sure that 

 such an accurately ruled micrometer, and one so clear to read, will 

 prove extremely useful to microscopists at large. 



Before closing, I would like to bring to your notice a screw mi- 

 crometer made for me by Mr. Powell, which contains some slight 

 modifications from the usual forms and suggested to me by practical 

 experience. 



First, with regard to the lens portion, I have substituted a compen- 

 sating positive for the old form of Huyghenian or Ramsden. This 

 yields far better images when making measurements with apochromatic 



