EXPLANATION OF FOEM AND COLOUEING. 241 



Half an hour later two Gharaxes candiope only had been eaten. 



Half an hour later again two Gharaxes hriUus only had been eaten. 



I next put in a handful of grasshoppers, etc., in great variety and two more 

 Gharaxes hrutus. The bird descended, ate a number of the grasshoppers, and 

 returned to her perch, having taken no notice of the butterflies. 



Three hours later (evening) there remained one Gharaxes hrutus, one Pyrameis 

 carchoi, one Junonia, one Papilio lyceus. a number of the bird's favourite species of 

 grasshoppers and some termites. The butterflies eaten since I last looked were 

 two Gharaxes hrutus (two of the five had been eaten previously), one Gharaxes 

 candiope (two of the three had been eaten at the outset), two out of three Pyrameis 

 cardui, two out of three Precis cehrene, the only Eurytela hiarhas, and all three 

 Neptis saclava. 



[The first butterflies eaten were four out of six large Gharaxes, the smaller 

 butterflies amongst which they were scattered having been apparently ignored. 

 This might seem to merely confirm my other experiments which have always seemed 

 to place these two species of Gharaxes in Grade 1. Actually, however, an exceed- 

 ingly important factor in determining the bird's choice is likely to have been size. 



Against this we find that two out of nine big butterflies were left at the end, 

 and also only two out of ten small ones. But it is possible that through stillness 

 or position the two large butterflies left escaped the roller's observation, while some 

 of the smaller ones were more prominent or accessible or happened to move. 



Then as to relative pleasantness. Three butterflies out of fourteen that the 

 other mode of experimentation places in or near Grade 1 were left uneaten and 

 only one (and that the pleasantest) of the five that it places below Grade 1 : 

 21 7o against 20 7o- These figures again would appear to indicate indiscriminate- 

 ness — and " indiscriminateness " there may have been if the roller, on the occasion 

 of any of her unwitnessed descents, had been hungry enough to eat Neptis saclava, 

 the lowest-gTade butterfly put down : and she obviously had been hungry enough. 

 Under these circumstances Neptis saclava would be in a sense as acceptable to the 

 roller as a Precis or Pyrameis, and would certainly be picked up and eaten if it 

 caught her attention first or were more accessible. 



The one definite preference apparently shown (apart from that for Gharaxes at 

 the commencement) was for the pleasanter grasshoppers as against any of the 

 butterflies. Yet even this conclusion is not secure. Was the handful of insects 

 put down together in a prominent mass, or was it scattered evenly through the 

 butterflies already there ? Would the roller at that moment not perhaps have 

 descended with equal readiness and eaten as heartily had the handful been one of 

 high-grade butterflies ? Was she temporarily tired of butterflies and wanting a 

 change through having eaten so many ? Or, did she actually and usually (as was 

 very possible) prefer the grasshoppers to the pleasantest butterflies ? 



The experiment, a poor one, unfortunately answers neither these questions nor 

 any other. It was mostly unwatched, and such portions as ivere witnessed were 

 not sufliciently fully recorded.] 



Exp. 2Q,-~May 24. Had evidently not gone down to eat recently as she was 

 unmistakably hungry in manner. She crushed and readily ate a Neptis agatha, but 



