COMMON SIGNS OF ERROR. 139 



other. It is therefore a good rule to assume their pre- 

 sence until all their possible sources have been exhausted, 

 and to take such steps as will quickly disclose them. The 

 most speedy way of rendering them evident is by means of 

 variety of experiments. Constant errors are often occult 

 ones, and are frequently due to a similar cause ; two phe- 

 nomena may vary together, and yet be only coincident, 

 one being the occult error. They will, however, probably 

 vary according to different rates, or one will disap- 

 pear without the other, if the conditions of the experi- 

 ment are sufficiently varied. Minute traces of iron, salt, 

 sulphur, ammonia, organic matter, air, water, and other 

 bodies have, on very many occasions, been the unsuspected 

 causes of errors in experiments in magnetism, chemistry, 

 spectrum analysis, &c. 



In every research, in order to reduce the sources of 

 error to the minimum, we simplify the experiments as 

 fast and as completely as we can; but even after we 

 have done this, many sources of error remain which we 

 are quite unable to exclude, without preventing the phe- 

 nomenon which we wish to examine. Those which remain, 

 are not, however, always termed errors, but interferences, 

 coincidences, or concomitant circumstances. In every 

 case, as long as any large, errors remain, we must try to 

 discover them. 



Small errors are more frequent and probable than 

 large ones, and usually more difficult to detect. Fre- 

 quently, where the phenomenon we wish to observe is a 

 very minute one, the coincident phenomena or inseparable 

 sources of error are very much larger than the true effect. 

 If there is a source of error, we always avoid it if we can, 

 or we neutralise or balance it ; or if we can do neither, we 

 make it as small as we can, and the true effect as large as 

 possible. We also endeavour to keep it of uniform mag- 



