ON LIGHT. 227 



particulars of their motions being known (as well as of 

 that of the planet itself, and therefore of the size and 

 situation of its shadow), there would be no difficulty in 

 making such prediction (starting from the time of some 

 one observed eclipse of each as an epoch) ; provided 

 always each eclipse were seen at the identical moment when 

 it actually happened. Moreover, on that supposition, the 

 times recorded of all the subsequent eclipses ought to 

 agree with the times so predicted. This, however, proved 

 not to be the case. The observed times were some- 

 times earlier, sometimes later than the predicted ; not, 

 however, capriciously, but according to a regular law of 

 increase and decrease in the amount of discordance, the 

 difference either way increasing to a maximum, then 

 diminishing, vanishing, and passing over to a maximum 

 the other way, and the total amount of fluctuation to 

 and fro being about i6 m 27 s . Soon after this discrep- 

 ancy between the predicted and observed times of 

 eclipse was noticed, it was suggested that such a dis- 

 agreement would necessarily arise if the transmission of 

 light were not instantaneous. This suggestion was con- 

 verted into a certainty by Roemer, a Danish astronomer, 

 who ascertained that they always happened earlier than 

 their calculated time when the earth in the course of its 

 annual revolution approached nearest to Jupiter, and 

 later when receding farthest : so that in effect the ex- 

 treme difference of the errors or total extent of fluctua- 

 tion the i6 m 273 in question is no other than the 

 time taken by light to travel over the diameter of the 

 earth's orbit, that being the extreme difference of the 



