l'J2 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1937 



moon on a scale which was wide horizontally and narrow vertically. 

 In order to get the niaterinl on a smaller piece of paper, I altered the 

 horizontal scale so that the horizontal scale for a century was about 

 the same as the vertical. A peculiar feature appeared. The curve 

 was not like an ordinary sine curve, which has flat tops and bottoms 

 and which is characteristic of motions produced by continuous forces. 

 It had sharp pointed tops and bottoms and was much more like a 

 series of intersecting straight lines than a continuously changing curve. 

 In other words the changes that took place were comparatively sudden 

 and then stopped, as if an occasional blow had been given. 



I had three of these sudden changes which were large: One about 

 1790, which, however, might have been gradual, as the observations 

 are not very accurate; one about 1897 ; and a third about 1917. Since 

 the last one there has been no further large change. The last change 

 disposes of Newcomb's sine curve — it will not lit at all. When the 

 sun curve was compared, it showed the same features as the moon 

 curve and in particular the three sudden changes of direction at the 

 same times. There could then hardly be any doubt about the source 

 of the deviations. 



Other workers have added informative evidence from the planets 

 Venus and Mars. DeSitter gathered all the evidence from various 

 sources including some from the other satellites and in general showed 

 that the result was correct. But there was a fly in the ointment, as 

 usually' happens in pioneer scientific work. Although the changes in 

 rate shown by all the bodies occurred at the same dates, the amounts of 

 the changes shown by the sun and planets were less than those shown 

 by the moon in the ratio 1.28 : 1. DeSitter estimated the probable 

 error of this factor to be ±.08 so that it is probably real and not a 

 mere effect of inaccurate observations. This factor, if real, indicates 

 that the moon has something to do with the changes but has only a 

 small efTect. 



The amounts of these changes in the rate of the earth's rotation are 

 considerable from the astronomical point of view. The greatest of 

 which we have any certain knowledge is that wliich occurred about 

 1897 and which changed the apparent length of the year by a second. 

 This is getting near the best that the new Shortt clocks can show, but 

 the latter have not yet quite attained the accuracy which is needed 

 for the purpose. Perhaps if we had a dozen such clocks, kept going 

 under the very best conditions, we might, by averaging the times they 

 would show, detect such a change. There is some interest in this 

 question at the present time because the deviation of the earth from 

 showing correct time is now greater than it has ever been since observa- 

 tions were made with sufficient accuracy, and consequently it is 

 reasonable to expect that a new change may soon occur. 



It is natural to ask at this stage how the observations of the moon 

 are made and the nature of the errors to which they are subject. 



