August i6, 1894] 



NATURE 



385 



works out yearly to an amount much more than equal to the 

 whole cost of the original article. 



I believe that experimental work in an engineering labora- 

 tory can educate this critical sense of proportion very admirably 

 in a number of ways. In the first place, it directs quantitative 

 work into very varied channels, and not along one particular 

 line. Secondly, it compels the observer to combine a number 

 of measurements in such a way that the relative importance of 

 accuracy in e^ch can be seen. In the case of an engine trial, 

 for instance, the combined results are affected by the accuracy 

 of measurements of the dimensions of the machine, by the 

 apparatus and methods used for measuring the water, by the 

 indicator, and by its springs, by the speed counter, by the ther- 

 mometers, and so on. An error of I per cent, in counting the 

 revolutions is just as important as an error of i per cent, in 

 measuring the water, or in measuring the mean pressure. I am 

 afraid that one could point to a good many cases in which this 

 ha< been more or less forgotten. Then, by making a series of 

 measurements all in absolute quantities, the relative importance 

 of each quantity to the desired total resul; can be seen. Thus 

 it will be found that changes in ceriain quantities affect the 

 total result to a very small extent, while changes in others affect 

 it very largely, so that not only is the accuracy with which 

 different quantities can be determined very different, but also 

 the same degree of accuracy is of very different importance 

 according to the particular quantity to which it refers. Once it 

 is found that a final result is exceedingly little affected by a par- 

 ticular set of changes, it ceases to be of importance to measure 

 or observe those changes in any extremely minute way, and of 

 course the reverse holds equally good. Finally, and this per- 

 haps is the most important matter of all, measurements in such 

 a laboratory are made to a great extent under the complicated 

 conditions under which the actual final result has to be ob- 

 tained in practical work. They are not made under the con- 

 ditions which insure the greatest individual accuracy of each 

 result. 



It will be seen that throughout, but particularly in the two 

 \\-.\ points which I have mentioned, the work of an engineer- 

 mi; laboratory is in intention and in essence different from that 

 of a physical laboratory. The aim of the latter is to make its 

 problems as simple as possible, to eliminate all disturbing ele- 

 ments or influences, and to obtain finally a result which possesses 

 the highest degree of absolute accuracy. In most physical in- 

 vestigations the result aimed at is one in which practically 

 absolute accuracy is attainable, although attainable only if 

 infinite pains be taken to get it. It is the business of 

 the physicist to control and modify his conditions, and 

 to use only those which permit of the desired degree of 

 accuracy being reached. In such investigations it sometimes 

 becomes almost immoral to think of one condition as less im- 

 portant than another. Every disturbing condition must be either 

 eliminated or completely allowed for. That method of making the 

 experiment is the best which ensures the greatest possible accu- 

 racy in every part of the result. The business of the engineer, 

 on the other hand, is to deal with physical problems under 

 conditions which he can only very partially control, and the 

 conditions are a part of his problem. He does not, for instance, 

 experiment with a steam engine so made that it can work with 

 a Carnot cycle. It is in the nature of the case that he must 

 experiment wiih a much less perfect machine. In burning fuel 

 he does not use apparatus especially made to absorb the 

 whole heat of combustion, but in the nature of the case has 

 to investigate the behaviour of apparatus in which a very large 

 part of that heat is unavoidably wasted. So one might go on 

 through an immense number of instances. Perhaps the whole 

 matter may best be summed up by saying that in a physical 

 laboratory the conditions of each expeiiment arc under the 

 1 control of the experimenter, and are subservient to the experi- 

 ment itself. In an engineering laboratory the conditions form 

 part of the experiment. However much more difficult or com- 

 plicated they tender it, they still unavoidably form part of it — 

 an experiment under any other conditions, or with those condi- 

 tions removed, would ipso facto be irrelevant. 



A critical training m matters mechanical is, however, only 

 too similar to the celebrated training of the Mississippi pilot 

 which so nearly broke the heart of Mr. Mark Twain. When- 

 ever the whole matter seems to be completely mastered from 

 one point of view, it is only to find, with a little more ex- 

 perience, that from another point of view everything looks 

 different, and the whole critique has to be started afresh. 



NO. 1294, VOL. 50] 



Machines cannot be finally criticised — that is to say, they can- 

 not be pronounced good or bad — simply from results measurable 

 in a laboratory. One wishes to use steam plant, for instance, 

 by which as little coal shall be burnt as possible. But clearly 

 it would be worth while to waste a certain amount of coal if a 

 less economical machine would allow a larger saving in the cost 

 of repairs. Or it might be worth while to use a machine in 

 which a certain amount of extra power was obviously em- 

 ployed, if only by mean-; of such a machine the cost of attend- 

 ance could he measurably reduced. In fact, what may be 

 summed up in the phrase the " worth-whileness " of economies, 

 is in itself a matter upon which a whole paper might be 

 written. Unfortunately, the latter points which I have men- 

 tioned are just such as cannot easily be piieasured in laboratory 

 work, or, indeed, in any other way whatever, except by actually 

 using the apparatus in question. All that can lie said is thii a 

 careful training in the critical measurement of comparatively 

 simple points fits a man more than anything ehe to gauge 

 accurately the importance of such other matters as I have men- 

 tioned. No doubt tfiere are many men in whom the critical 

 faculty is insufficiently developed to allow them ever to be of 

 use in these matters, but to those who are intellectually capable 

 of the "higher criticism" it must be, I think, of inestimable 

 benefit to have had a systematic training in the lower. 



Is there, then, any general standpoint from which mechanical 

 criticism can be directed? Certain points are obvioui, but 

 probably the whole matter cannot easily be generalised. A 

 city has to be supplied with water ; there are three requisites : 

 that the water should be of proper quality, of sufficient quan- 

 tity, and that it should be brought in at a reasonable co5t. But 

 in such a case the first two are so enormously more important 

 than the third, that the ideal is comparatively simple (of course, 

 this is quite a different thin^ from being simply reached). A 

 city has to be supplied with electric light ; the essential con- 

 ditions are similar. But in this case there are so many qualities 

 which are equally proper, and there are so many diflferent ways 

 of bringing it in in sufficient quantity, that the third point — 

 namely, the cost — becomes especially important. A faciory 

 has to be driven by steam power : the amount of p^wer that is 

 wanted can be produced by so many different types of engine 

 and boiler — all capable of approximately equal economy, and 

 all claiming equal freedom from breakdowns — that the choice 

 is a peculiarly difficult one from the critical point of view. 

 j It seems almost inipossilde that a criticism on any one basis 

 could meet all the three cases which I have supposed unless 

 ) that basis were that the thing supplied should be the absolutely 

 fittest, having regard to all the conditions of each case and the 

 relative importance of each condition. Possibly in all cises we 

 could get at some generalisation which would show us which 

 was the absolutely fittest, if only the necessary data were in any 

 way complete, which they very seldom are. Perhaps in one 

 sentence we may say that that scheme, or system, or machine, 

 will be the absolutely best in any particular case which will the 

 longest survive and maintain its place in its particular environ- 

 ment. I cannot doubt that this development of Darwinian ideas 

 in the world of the inorganic is a legitimate one. Of course the 

 problem would be comparatively easy in each particular case if 

 only the environment would stand still. It would even be com- 

 paratively easyif we knew how the environment was going to 

 alter, but this we are unable to do. We only know that it cer- 

 tainly -Mill change and will go on changing, and that therefore the 

 things which we make now have not got to survive in the con- 

 ditions in which we make them, but have got to survive through 

 some new sets of conditions of which we know nothing. I do 

 not think the difficulty is in any way met by the popular method 

 of guessing at what will be wanted fifty years hence, which 

 gent rally means simply guessing at something very big. It is of 

 no use making our ships or our engines of a type which we 

 choose to imagine will be that of fifty years hence. If we do 

 they will be of no use to day, and for that very reason they will 

 not even be in existence, useful or other, at the end 

 of the fifty years. Sufficiently sad illustrations of this 

 will occur to everyone in very different directions. I 

 hope I shall not be considered churlish in saying that 1 do 

 not think that the men who have worked on this principle have 

 really been far seeing, or have really brought us much forward. 

 They have been men often of genius, often of great personal 

 fascination, always of immense imagination. liui they have 

 proceeded by methods essentially opposed to anything like the 

 gradual evolution which must occur in technical as it does in 



