HALL. — DEVIATIONS OP FALLING BODIES. 345 



low could not be kept uniform during all the changes of position of the 

 pans indicated by Fig. 6. Accordingly, I regard the mean value found 

 for the easterly deviation as entitled to considerably less confidence than 

 the mean value found for the southerly deviation. 



At times the tallow mixture adhered well to the bottom and side of 

 the pan. At other times it would loose hold on the pan, during cooling, 

 to such an extent as to be easily removable in a single cake. To pre- 

 vent error from this condition it was the practice, whenever there was a 

 gap between the tallow and the side of the pan, to force slips of paper 

 on thin cardboard into the gap in such a way as to pack the tallow in 

 place. Moreover the studs, which (see Fig. 5), on and after April 7, 

 projected from the outside of the edge of the pan, were really the heads 

 of screws which extended a considerable distance into the tallow, and 

 these screws helped much to hold the tallow from rotation or sliding. 



On July 10 soft wood (white wood) about 2 cm. thick was used in 

 two of the pans instead of tallow. The mean of the range of balls in 

 these two pans on this night was decidedly less, in both latitude and 

 longitude, than the usual range in pans filled with tallow. Unfortunately, 

 however, the impression made by a ball dropping upon wood is ellipti- 

 cal, because of the grain, so that when the ball is replaced in the hollow 

 it has made it does not lie securely, but is likely to roll or rock a little, 

 making determination of its proper position somewhat difficult. The 

 results for the night of July 10, with wood, were omitted from the final 

 calculations. 



Two possible sources of error remain to be considered, — the action of 

 air and the faults of levelling. Great care was taken from the first to 

 prevent or reduce draughts. The description, previously given, of the 

 apparatus shows that chinks at the releasing end were stuffed with cotton, 

 if not otherwise closed. The cotton may have been slightly dis{)laced at 

 times, but certainly there was very little freedom of circulation through 

 the top inclosure just before or just after a ball was dropped. The cloth 

 tube (c in Fig. 1) was continuous, being woven in one piece, and at its 

 lower end the chink about it was stuffed with cotton. Below tliis tube 

 the balls fell about two meters through air inclosed only by the walls of 

 the constant temperature room, which is about 6 m. long, 4 m. wide, and 

 3 ra. high. The one door leading into this room was closed while the 

 balls were falling. Two incandescent electric lamps of eight or sixteen ■ 

 candle power were available for lighting the room ; but usually only one 

 of them was in use at a time. These were suspended in such a way as 

 to light up the plumb-line and the brass tip over the balls to advantage, 



