"String-Pulling" in Birds. Iii. the Successful Response in Greenfinches and Canaries

Abstract

Abstract1. Eighteen greenfinches, of which six were wild trapped adults, five were aviary-reared juveniles and seven were hand-reared juveniles; and also sixteen domestic canaries of which six were adults and ten were juveniles were tested in a "string-pulling" situation. 2. The experiment consisted of (1) a training period when birds were fed repeatedly at the point on a perch where the string and bait were to be suspended; (2) a test period when birds were given up to twelve half-hour trials with the bait suspended on a 4" string; and (3) for those birds which failed in the test period, a period of training in string-pulling. Half the juveniles tested were successful in string-pulling, but none of the adults. 3. Results showed that the finches used more varied methods of pulling up the string than do other species (e.g., tits) which habitually use the feet in feeding. It was also shown that the methods used by juveniles tended to be more varied than those of adults. 4. To this greater variety of behaviour is in part imputed the greater success of the juveniles in the string-pulling situation; this factor is, however, considered to be additional to those previously reported: that juveniles spent a greater proportion of each test trial in responding to the string and/or bait, and a diminished tendency to become extinguished in comparison with adults. 5. The way in which behaviour developed throughout the test trials was adaptive in the sense that activity which was reinforced (e.g., by the movement of the bait following a pull on the string) tended in general to increase at the expense of unreinforced activity. 6. Consideration of individual performances during the test and training trials suggested that the birds showed no evidence of insight into the nature of the experimental situation; success tended to occur intermittently and the improvement in performance was on the whole gradual. This suggested that the performance is closer to "trial-and-error" than to insight learning.