208 MR. C. F. M. SWYNNEKTON ON THE 



made to yield other conclusions if the experimenter is found to have been misled 

 through bias or error or ignorance, is evidence ; and that it is only right to 

 place this evidence on record and at the disposal of all who mnv at any time 

 care to study it and draw those other conclusions if they can. Even the 

 giving of sample experiments is no guarantee against erroneous deduction in 

 those that are not given. 



To give only one example of the very great danger attaching to incomplete 

 publication in a matter of this kind, 1 may say that, with increased knowledge, 

 I myself now draw quite different conclusions from my own first five hundred 

 experiments from those I drew at the time. Yet five hundred is a large 

 number^ capable of forming a paper nearly ;islong as the whole of the present 

 one, and had I at that time decided to publish them in full my decision would 

 almost certainly have met with the friendly criticism it has met with to-day. 

 The result, had I contented myself with the publication of my conclusions, 

 would have been utterly to mislead. So now, I can be by no means sure that 

 I have extracted all conclusions of importance that can be extracted from my 

 experiments— I know that I have not — or that the discovery of some new 

 complicating factor still unknown to me may not jet invalidate some part of 

 my present conclusions. This, as I shall relate, actually happened to me not 

 very long ago. I feel, therefore, that the only safe course is to place on 

 record the exact details of ivJiat the animals actually did. 



B. Are the Experimental Results probably to be relied on?* 



The reader may object (very reasonably) that records of wild birds actually 

 attacking butterflies are even now all too scarce ; that tens of thousands of 

 their stomachs have been examined, in almost every case with negative results ; 

 that M^here wild birds liavf. been seen to eat butterflies these have very 

 frequently indeed belonged to so-called nauseous species and that destruction 

 by birds cannot therefore fairly be regarded as discriminative. Finally, even 

 admitting that discrimination may take place, what proof is there that a 

 captive bird^s preferences will fairly represent those of a bird in the wild 

 state ? If a pet and in the habit of receiving dainties from its owner, may it 

 not tend to eat on trust even unpleasant insects ? With appetite and digestion 

 impaired by fear or fretting or if stinted of its natural range of food, is it 

 likely to be a suitable subject for this kind of experiment at all ? Captive 

 birds have eaten what they are not known to eat in the wild state and refused 

 prey that in the wild state they have been found to eat. " Unless, therefore/'' 

 the reader might conclude, " you can first satisfactorily meet these very real 



* This, and the introduction genevall}?, was originally written in relation to the experi- 

 ments witli insects. 3£zitatis mutandis, a great deal of it is applicable to the experiments 

 in which vertebrates and plants were used, A good case can be stated, I think, for the 

 reliability of these too. 



