THE ORIGIN OF VARIATION 41 



similar results claimed to have been established by Pavlov 

 have now been withdrawn by the author (see Macdougall, 



1927)- 



xix. Harrison (1927). Oviposition of the sawfly Pontania 



salicis. 



Harrison found that the galls of this sawfly in any limited 

 area tended to occur on only one species of willow, though in 

 the whole range of the sawfly many species of willow were 

 attacked. He therefore took sawfly galls from one willow and 

 exposed them in a locality where only another species was 

 available. In the most convincing experiment, in the first 

 year there were few galls and many of these aborted, but later 

 the sawfly became entirely attached to the new host, and, when 

 tested five years later, the original host was no longer attractive. 

 Harrison regards these experiments as a proof that an induced 

 habit-change is inherited. It is possible to regard them as an 

 evidence for ' larval memory ' (see remarks on biological races, 

 pp. 50-52), the oviposition of the females being influenced 

 by the nature of the site in which their larval life was spent. 

 It has been suggested that crosses between certain moths show 

 that oviposition-response segregates as a typical unit character 

 and that therefore such responses must always be germinally 

 fixed. We do not doubt that in many cases the oviposition- 

 response is germinally fixed and it is possible that the temporary 

 fixture by means of ' larval memory ' is a sort of ' half-way 

 house.' But this is by no means proved, and, indeed, any 

 attempt at direct proof would be likely to meet with invincible 

 technical difficulties. 



xx. Tornier and Milewski (literature in MacBride, 1924). 

 Experiments with ' fancy ' types of Carassius (Goldfish). 



Certain domesticated ' fancy ' breeds of Goldfish have been 

 cultivated for a long time in China and Japan. They are 

 characterised principally by abnormal development of the 

 fins and the snout and head and by certain colour-changes 

 (Ryukin and Ronchu types, e.g.). In the course of a long 

 period of culture these aberrant types have been detected, iso- 

 lated and bred from for ' fancy ' purposes. It is said that they 

 breed true, but how far this is a fact is uncertain. In experi- 

 ment {e.g. Milewski's) they seem to be relatively unstable. 

 Tornier discovered by experiment, both on Carassius and other 

 forms, that the abnormal structural features were due to 



