268 PROCEEDINGS OF THE SOCIETY. 



particular about this. The error, if any, from this source must be very 

 small, if not insensible ; but, if existing, it was a constant affecting all 

 alike, and hence unimportant in relative measurements such as these 

 were. Again, the differences of the micrometer readings are given in 

 nearly every instance, the first reading being usually zero or so near it 

 as to involve no error of the screw. The screws themselves are excellent 

 for their purpose. Their errors were not magnified by the whole optical 

 combination, as were the stage micrometers ; and, so used, their errors 

 were insensible. 



Now, as to micrometers used. None except " Cm. A " were selected 

 rulings, but mainly such as I could buy in open market or borrow from 

 friends. My own were not selected from the outset, but purchased in 

 open market and carefully standardised, or were ruled by myself on the 

 Rogers engine which I owned for over ten years. These were carefully 

 ruled on the same part of a carefully calibrated screw ; and were no 

 better and no worse than others ruled on the same machine. 



Now as to glass micrometers. My experience is that if the lines are 

 filled with fine graphite, and then covered with balsam, they will not 

 deteriorate if the balsam is kept carefully sealed in. The graphite, un- 

 fortunately, will not always fill in the fines with perfect uniformity ; but 

 some portions of the scale can easily be found which will furnish good 

 lines. Personally I prefer a metal surface or a line ruled through a thin 

 film of silver on glass, sealed with balsam, and covered. These last are 

 elegant. I have never seen any except those made by myself. Zeiss 

 uses this sort of surface in his Abbe test plate, but I make use of a 

 much thinner film, so thin, in fact, as to be translucent. The trouble 

 with realgar is that there is no guarantee of permanency ; and it is 

 disquieting to find a micrometer upon which one has bestowed a large 

 amount of research, become so impaired as to invalidate all one's work. 

 I have micrometers ruled on glass, the line filled with graphite, and 

 others ruled through silver on glass covered as above stated, that I have 

 used for fifteen years, and they are still as good as when first ruled. 



There is one other item I wish to refer to, viz., the probable error. 

 The tables given furnish the data for computing this error, and I think 

 I sent you in a later communication the probable error of one or more 

 of the series. The labour of computing the probable error of all was 

 more than I felt able to undertake at the time the work was completed. 

 The probable error does not take count of constant errors, and, under 

 the appearance of great accuracy inferred from a small probable error, 

 may lie concealed constant errors of unknown magnitude. These errors 

 I have taken great pains to eliminate, whether or not with entire 

 success is not for me to determine. 



In conclusion, I wish again, through you, to thank the Society for 

 the candid and entirely fair manner in which my paper has been 

 criticised. I expected and desired it. Candid criticism is much more 

 valuable than praise to me. It has disclosed weak points which I shall 

 endeavour to correct in future work. 



Very respectfully yours, 



M. D. Ewell. 



