482 KEroRT — 1882. 



taken was published, not only the metalloids but many metals have been fomid 

 to give complicated spectra at one temperature and much simpler spectra at 

 higher temperatures. Such are the channelled spectra of sodium and potas- 

 sium, first descrilsed by Roscoe and Sclmster, the channelled spectra of silver, 

 bismuth, and other metals described by Lockyer and Roberts, and the ultra-violet 

 channelled spectrum of tin recently photographed by Dewar and myself, 

 Mitscherlich's hypothesis gives us a rational explanation of such multiple spectra 

 produced hj the same substance, and it has Ijeen accepted, in one form or another, 

 by all spectroscopists since he wrote. Nevertheless, the existence of multiple 

 spectra cannot be taken as a proof of allotropic modification imless the possibility 

 of a chemical combination be excluded. The channelled spectrmn which mag- 

 nesium gives in hydrogen was mistaken by more than one observer for that of 

 some modification of the simple metal, until it was shown that magnesium in 

 nitrogen or other gases does not give it if no hydrogen is present, and that its per- 

 sistence in hydrogen at high temperatures depends, as does the permanence of 

 chemical combinations, on the pressure of the gas. If, however, homogeneous mole- 

 cules are dissociated by heat, so also are heterogeneous molecules, formed, as we say, 

 by chemical combination, split up by elevation of temperature, to unite again on 

 cooling, or when the pressure is increased within certain limits. Nor is there any 

 essential difference in character between a chemical compound and an element be- 

 yond that of facility of decomposition. If we could not so easily resolve them into 

 their constituents, and were to disregard the difierence of their spectra, no one- 

 would suppose ammonia to be differently constituted from potassiimi, or cyanogen 

 from chlorine. Indeed, chemists have long been in the habit of considering the 

 imion of two atoms in a molecule of ordinary hydrogen or chlorine as a species of 

 chemical combination ; but when we find that the combinations of particles of the 

 same kind are as definite as those of particles of different kinds, and that they 

 are both subject to precisely the same mechanical laws, we are hardly justified 

 in regarding the forces by which they are produced as essentially difl'erent. To get 

 rid of a gratuitous hypothesis in chemistry must be a great gain. 



But it may be asked, why stop here ? Why may not the chemical elements- 

 be further broken up by still higher temperatures ? A priori and from analogy 

 such a supposition is extremely probable. The notion that there is but one ele- 

 mentary kind of matter is at least as old as Thales, and underlies Front's hypothesis 

 that the atomic weights of om* elements are all multiples of that of hydrogen. 

 This famous hypothesis has gone up and down in the scale of credibility many 

 times during the present century. About seventeen years ago the publication of 

 Stas' new determinations of combining weights, carried out on a scale never before 

 attempted, and with all the refinements which the growth of our knowledge could 

 suggest, was thought to have given it its death-blow. But a reaction has set in 

 since that time. The periodic recurrence of the properties of elements with regular 

 additions to the atomic weights, like octaves in a musical scale, put forcibly before 

 us by Mendelejeff, makes it diflicidt not to think that there is a simple relation 

 between the atomic weights, though there may be causes producing slight perturba- 

 tions of such a relation. Quite recently a fresh revision of the combining weights 

 has been made on the other side of the Atlantic by Professor F. W. Clarke. Pie has 

 collected all the determinations made by different observers, and after rejecting such 

 as from defective methods were imtrustworthy, has applied to the remainder such 

 corrections as newer experiences have suggested, and then deduced from the cor- 

 rected nimibers the most probable values by the methods of the theory of errors. 

 Professor Clarke has done a piece of work of the highest utility, for which chemists 

 must be grate fid ; nevertheless, he has not carried the revision so far as it might be 

 carried. He has, to begin with, rightly separated the several sets of observations, 

 and deduced the most probable nimiber from each set by itself, but in combining 

 the various sets for the final determination of the numbers adopted, he has treated 

 the residts obtainfid by different methods as if they were a set of observations all 

 presumably of equal value, so that the most probable numbers could be deduced 

 by the method of least squares. He has not attempted any discussion of the diffe- 

 rent methods with a view to an estimate of the relative values of the results 



