October 10, 1913] 



SCIENCE 



517 



a discussion of this topic see the fourth edi- 

 tion, p. 295. 



At latitudes higher than 50°, where it is 

 impracticable to obtain sufficiently slowly 

 moving zenith stars with balanced A factors, 

 and where, consequently, the error in azimuth 

 will materially affect the clock correction, the 

 older methdd of observing an azimuth star is 

 still employed. The number of stars in each 

 half -set, however, following perhaps the prac- 

 tise of the Germans, is increased to six. It is 

 a fair question, in this connection, whether 

 this ratio of azimuth to clock stars is suffi- 

 ciently large. 



The time sets are so chosen, and the re- 

 versals of the instrument between half-sets so 

 planned, as to eliminate collimation and in- 

 equality of the pivots (p. 19). Inequality and 

 irregularity of the pivots, indeed, as the pivots 

 have been reground and tested (p. 46), and 

 owing to the plan of observing adopted (p. 50), 

 is thought negligible. This is in contrast to 

 the practise formerly in vogue. As for the 

 collimation, if it may be depended upon to 

 remain constant during a time set, it will be 

 eliminated entirely. With instruments re- 

 versible on each star, as already noted, in- 

 equality of pivots and collimation are rigor- 

 ously eliminated automatically by the reversal. 

 Instead of depending upon the invariability 

 of the instrumental constants for an hour, 

 this dependence is necessary for but a few 

 moments — a decided advantage. On the other 

 hand, the possibility exists that too frequent 

 reversal may disturb the azimuth; and as the 

 disturbance is likely to occur between the two 

 parts of each star observation, this is a serious 

 matter. The French have accordingly intro- 

 duced the practise of reading at suiEciently 

 frequent intervals on a meridian mark. 



It should be noted, too (p. 27), that among 

 other advantages, reversal on each star leads 

 to simplified computation. 



" It is desirable, but not necessary " (pp. 43 

 and 80, sec. 4), is the comment on the require- 

 ment of the previous edition that the same 

 stars, wherever possible, be observed at both 

 stations of a longitude determination. It is 

 now believed that errors of the star places are 



smaller than those introduced by the instru- 

 mental constants ; or by the variation of those 

 constants due to extending the observations 

 over too long an interval; or by poor bal- 

 ancing of the A factors; or by an unwise 

 choice of epoch for exchange of clock com- 

 parison signals (pp. 87 and 93, sec. 7). The 

 argument upon which this reasoning is based 

 is not conclusive; for the accidental errors of 

 the star positions alone are taken into account, 

 nothing being said of those classed as sys- 

 tematic. Yet it is probably true that great 

 inaccuracy will not result, especially if a large 

 proportion of the stars be observed in common 

 at the two stations. 



Not only is the publication marked by the 

 adaptation of a new device to the old instru- 

 ment, and the adoption of a new program of 

 observing, but also by a new method of re- 

 ducing the observations. The germs of this 

 method may be found in the old edition, p. 

 296. The use of least squares has for the 

 most part been done away with; the refine- 

 ment, evidently, is believed to be unwarranted 

 by the observed data. The result is a more 

 direct and easy method of solution. To sim- 

 plify the computations further, unsym metrical 

 threads are usually rejected (pp. 24, 79 and 

 80). Criteria for the rejection of other 

 threads are laid down on p. 80. Corrections 

 for rate (p. 24) are generally regarded as 

 unnecessary refinement. Contrary to former 

 practise, all stars observed at latitudes under 

 50° are weighted equally (pp. 79 and 80), and 

 weights generally are taboo (p. 89). 



The survey is quick to take advantage of 

 any opportunity. When the International 

 Geodetic Association commenced furnishing 

 corrections for reducing the observed to the 

 mean position of the pole, the survey began to 

 make use of these corrections. When the 

 American Ephemeris and Nautical Almanac 

 became enabled, through the omission of the 

 lunar distance tables, to extend its list of 

 stars, the survey, probably having in mind 

 also the greater ease of interpolation from the 

 Washington meridian, assigned to that ephem- 

 eris the preference formerly held by the Ber- 

 liner Jahrhuch (pp. 25 and 43) ; and from 



