384 Intelligence and Miscellaneous Articles. 



It would be impossible to determine by inspection of these 

 measurements what is the real relation of the objects measured. 

 The Mean of the first series being 2 1 0, while that of the second 

 series is 162, those who look only to averages without testing 

 their significance will conclude offhand that the first object 

 is greater than the second. But in fact the real value under- 

 lying the first series is only half the corresponding value for 

 the second series ; the former being the twice-mean-square- 

 of-error for aggregates of ten digits taken at random, the 

 latter for aggregates of twenty digits. The real relation 

 would come out if we went on long enough. I have gone on 

 long enough, in the case of the first series, to get within a 

 thirty-fifth part of its theoretical value, namely 165. I obtain 

 160 as the mean of two hundred and eighty terms of the first 

 series — a number of observations which corresponds to a pro- 

 bable errror of about 5, the error which I have incurred. If 

 I went on long enough with the second series I should, 

 doubtless, get equally near its real value, which is 330. But 

 to the unitiated such statistics are hopelessly misleading. 

 Blind palpation is sure to err. 



XLVII. Intelligence and Miscellaneous Articles. 



COMPKESSIBILITY OF LIQUIDS. BY M. F. AMAGAT. 



TN measuring very high pressures I have adopted the principle of 

 -*- the differential manometer; the conditions to be realized for 

 obtaining exact measurements is that the pistons be completely 

 mobile and at the same time perfectly tight. 



M. Marcel Deprez had the idea of dispensing with the leather of 

 the small piston, and making the escape extremely small by a con- 

 venient adjustment. This device is insufficient for very high 

 pressures, especially in the conditions of my experiments ; this is 

 also the case with the use of goldbeater's skin, adopted by M. 

 Cailletet. 



On the other hand, numerous experiments have shown me that 

 the membrane, on which the large piston rests, introduces several 

 sources of error. I have altogether dispensed with the leather and 

 the membrane, and have solved the difficulty by using a viscous body 

 suitably chosen. The large piston, which only receives the reduced 

 pressure, rests on a cushion of castor-oil, which transmits the pres- 

 sure to the mercury; the small piston, which receives all the 

 pressure at the top, becomes quite tight if, after being soaked in oil, 

 and put in its place, it is wetted on its base with a sufficiently viscous 

 liquid, such as molasses, which answers perfectly. In these con- 

 ditions, the pistons being even somewhat free, there is no real leak, 

 but only an extremely slow oozing, which does not affect the mea- 

 surements, and this up to pressures higher than 3000 atmospheres. 



