1869.] 



Corrections to Pendulum Observations. 



493 



cess of reducing our observations, came to the conclusion that it would 

 be best to exclude all experiments made at temperatures above 100°, 

 and also those where great differences in the readings for temperature 

 occurred, from our final results. The principle which guided us was not 

 to vitiate good observations by doubtful ones ; and the following small Table, 

 showing the temperature-readings during four experiments, taken quite at 

 random, will show best how we proceeded: — 



t 



Thermometer 

 in front. 



Thermometer 

 behind. 





Upper 

 therm. 



Lower 

 therm. 



Upper 

 therm. 



Lower 

 therm. 





7I - 20 



71-15 

 71-30 

 71-85 

 72*00 

 7* '9° 



70-20 

 70-20 

 70-00 

 69-60 

 69-55 

 69-50 



70-0 

 70-4 



70- 6 



71- 4 

 71-4 

 7*"4 



68- 5. 



69- 4 

 69-7 1 



7i*4 ( 

 7 r6 

 707 J 



S 



£ V 

 |l 



Mean 70-9 



70-6 







70-8 







III. 



99'5 

 97'4 

 95*5 

 82-5 

 816 

 80-9 



98-7 

 97-0 

 94' 3 

 79*9 

 79'3 

 79-2 



92-4 

 916 

 9°*3 

 78-5 

 77*4 

 75'5 



9 I "5^ 

 90-6 

 88-i ( 

 76'2 { 

 76-0 

 7.4" y 



a . 



£ £ 



a 



X ! 



m 



Mean 88-8 



837 







862 







II. 



Thermometer 

 in front. 



Upper Lower 

 therm, therm. 



85- 6 



86- i 

 867 

 78-4 



79'4 

 812 



837 

 83-8 

 83-6 

 78-8 

 80-4 

 82-0 



!2- 5 



Thermometer 

 behind. 



L T pper 

 therm. 



^4-0 

 83-3 

 82-2 

 8o-8 

 79'4 



Lower 

 therm. 



83*5 

 83-0 

 82-6 

 82-1 



81- 2 



82- 8 



82-45 



82-5 



IV. 



108-4 



107-3 



100-6 



99*5 



107-6 



106-5 



99*7 



99-0 



105-7 



103-8 



98-5 



97*3 



97'4 



96-5 



931 



9 r 3 



96-3 



95-1 



92*2 



90-9 



959 



947 



91-8 



9°'3 



101-3 



953 



Although in the rejected experiments the means of all readings of 

 both sets of thermometers approach each other, still there occurs a 

 fall of nearly 20° at the end of an experiment, as compared with that 

 temperature which is recorded at the beginning, besides differences 

 of nearly 8° between the thermometers in front and those at the 

 back of the pendulum. That the temperature of the latter during an 

 experiment is represented by the arithmetical mean of such discordant 

 readings, we think most unlikely; and hence these experiments and 

 similar ones were not used, although, of course, the number of available 

 experiments was thereby reduced. 



The following experiments, which represent the final results of this 

 temperature-investigation, deserve at least some confidence, although we 



