THE GREAT STAR MAP 255 



of so many figures in the proof sheets is no trivial matter, since 

 it is important to avoid mistakes — astronomers know the trouble 

 which may be caused by a wrong figure, once it gets into 

 print. 



In dealing with so many figures it is important to devise 

 tests of their accuracy. In mathematical calculations, a whole 

 series of operations can sometimes be tested from a single result ; 

 if that is correct, there cannot be a mistake in any of the opera- 

 tions, at any rate not one mistake alone ; there may be two or 

 more which exactly compensate one another but this risk is 

 more remote. Just so in money accounts, if the totals check 

 one another, it is usually fair to assume that the individual items 

 are correct. In the case of the Star Map no such economical 

 tests are possible ; for no connection between the positions of 

 the stars is known to us. We must be content to check each 

 star by itself and for this purpose two measures are made of it 

 under different conditions. After all the stars on a plate have 

 been measured, the plate is taken out of the measuring apparatus, 

 turned round through 180° and put in again for remeasurement. 

 The second set of measures gives an independent check on the 

 first ; in this way nearly all the unintentional slips are detected : 

 so that when the results are printed and compared with both 

 sets of measures, they are substantially correct. This is proved 

 in the following ways : according to the decisions of the Confer- 

 ence the plates of each series are to cover the sky tv/ice, with a 

 certain overlap of adjacent plates, so that every star appears on 

 at least two plates. After the measures have been printed off 

 the two independent measures for each star have been compared 

 in many thousands of instances ; the number of errors found is 

 remarkably small. Again, some of the plates accepted in the 

 first instance did not come up to later standards and were re- 

 peated : here again comparisons between old and new measures 

 have detected a few mistakes, but not many. These checks 

 have given confidence in the general accuracy of the work. 



( To be continued) 



