334 



SCIENCE 



[N. s. Vol. XXXIV. No. 872 



A mere catalogue of facts, however well ar- 

 ranged, has never led to any important scientific 

 generalization. For in any subject the facts are 

 so numerous and many-sided that they only lead 

 us to a conclusion when they are marshaled by 

 the light of some leading idea. 



Let US take, for instance, a catalogue of 

 variable stars such as those of Mr. Chand- 

 ler. Particulars for each star are given in 

 •separate columns, exclusive of the name 

 -and number. We might wait long for a 

 -leading idea to guide us in marshaling the 

 ■facts, and so far as I know we have waited 

 till now without anj^ such idea occurring to 

 any one. But Professor Pearson insists on 

 the plain duty of determining the correla- 

 tion between each and every pair of these 

 columns, and any others we may be able to 

 add. Anybody could have made the sug- 

 gestion, and there was plenty of elementary 

 mathematical machinery in existence for 

 carrying it out; but so far as I know no- 

 body did, any more than the critics of Co- 

 lumbus suggested how to stand up an egg. 

 But the suggestion having been made by 

 Professor Pearson, it was so clearly sound 

 that I did what lay in my power to follow 

 it up : with the result that certain correla- 

 tions were at once indicated which at least 

 pave the way for further inquiry. If we 

 can not say more than this it is simply be- 

 cause the catalogue of facts was not large 

 enough. So far from the observers having 

 wasted their energies by observing without 

 any theory to guide them, more work of the 

 same kind would have been welcome, for it 

 would have reduced the probable error of 

 the correlations indicated. As an example 

 I may quote the following. It has already 

 been mentioned that a variable-star maxi- 

 mum though it may recur after a more or 

 less definite period on the average, is sub- 

 ject to a swing to and fro like the time of 

 sunrise. Let us call the average interval 

 the day of the star and the period of swing 



the year, without implying anything more 

 by these names than appears in the anal- 

 ogy. Then I found^ that the day and the 

 year were correlated, the value of the co- 

 efficient being 



r = 0.56 ±0.08. 



Having obtained this clue, it was interest- 

 ing to use it for the elucidation of individ- 

 ual problems. The days of many stars are 

 by this time pretty well kno^vn, but their 

 years are very uncertain. In nine or ten 

 cases the assessment of the vaguely known 

 year was under revision, and in all, without 

 exception, the revised assessment tended 

 in the direction of the formula. In one 

 case {8 Serpentis) the formula suggested 

 the solution of a long-standing puzzle.* 

 Finally the inquiry is suggested whether 

 our own sun may be treated as a variable 

 star with a period or day of eleven years, 

 in which case its time of swing a year 

 should be about seventy-five years, if the 

 formula is strictly linear. There are found 

 to be indications of a swing of this order 

 of magnitude, though the time given by the 

 periodogram method is fifty-four years. ^ If 

 the relation between year and day is not 

 strictly linear these figures could easily be 

 reconciled for a case lying so far outside 

 the limits within which the formula was 

 deduced. But the ultimate successful es- 

 tablishment of the connection is of less im- 

 portance for our present purpose than to 

 notice the fruitfulness of the method of 

 suggestion, which is as mechanical as Bacon 

 himself could have wished. 



Let us admit frankly that there is an ap- 

 pearance of brutality about such methods. 

 Is our method of search to be merely the 

 old and prosaic one of leaving no stone un- 

 turned? We have been led to believe that 

 there should be more of inspiration in it; 



'Monthly Notices B. A. S., LXVIII., p. 544. 

 ' Monthly Notices B. A. S., LXVIII., p. 561. 

 '■nid., p. 659. 



