Sharing and Cooperation in Pigeons Results & Discussion, Part I Table Of Contents History Title Page List of Figures, Tables Introduction Purpose, Method Results and Discussion General Discussion References Abstract

Results and discussion

I. Preliminary training

Color-matching (leader role responding) training.

Subjects received approximately 75 sessions of color-matching to sample training. During leader role training, ratios on the take-key and on the lower (green) match-key were increased to FR 10 (except for P1S2 whose lower match-key ratio was raised to FR 20), and to FR 2 on the upper (red) match-key. The higher ratio on the lower match-key was implemented in an attempt to eliminate a preference by all subjects for the lower match-key (the closer of the two matching-kegs to the food magazine opening). Once both subjects of a pair had responded at above 60 accuracy for 5 consecutive sessions, follower role training was begun.

Social-matching (follower role responding) training.

Figure 2 shows matching accuracy during training of social matching. Subjects P1S2 and P2S2 (i.e., subject 2 of each pair) responded at above 80% correct by the ninth session of follower training. All subjects except P2S1 had responded at or above 80% correct for at least five consecutive sessions within 28 sessions of the beginning of follower training (28 sessions, P1S1; 23 sessions, PIS2; 16 sessions, P2S2). Subject P2S1 reached 80% correct after 32 sessions of follower training, and had responded at or above 80% for five consecutive sessions after 55 sessions.

Figure. 2. Percent correct matching for the follower arid leader during acquisition of follower responding (social matching)

Pair 1

Pair 2

When the first subject of a pair to receive follower training had responded at or above 80% correct far at least five consecutive sessions, follower training of the paired subject was begun. The second subject of each pair received additional training sessions after reaching criterion.

During these additional follower training sessions, the match-key accuracy of P1S1 and P2S1 decreased slightly. Match-key ratios for both subjects in both pairs were manipulated in unsuccessful attempts to increase follower accuracy.

Role-alternation training.

Figure 3 shows match-key responding during role-alternation training. For the session-by-session role alternation condition, the first 15 sessions for pair 1 (P1) and 20 sessions for pair 2 (P2), and last five sessions are shown; the first and last five sessions are shown for the half-session alternation condition through tenth-session alternation condition.

Figure 3. Percent correct for the follower and leader during response role alternation training. Subjects alternated responding as leader and as follower by full sessions (100 trials), by half sessions (50 trials), by quarter sessions (25 trials), by tenth sessions (10 trials), and by trial. Breaks in the x axis indicate sessions not shown.

Pair 1
Pair 2
Condition Description A Full session role alternations B 1/2 session role alternations C 1/4 session role alternations D 1/10 session role alternations E Trial-by-trial session role alternations

The match-key response accuracy in both the leader role and the follower role was at or above 60% correct for two subjects during all but two sessions of role-alternation training (P1S2 and P2S2, one subject from each pair). The leader-role match-responding of the other two subjects was at or above 80% for all but four sessions (P1S1) and one session (P2S1).

Follower-role match-responding was in the 60-80% range for most of role-alternation training for subjects P1S1 and P2S1. These are the same two subjects whose follower-role match-responding accuracy decreased during the previous follower training condition. During role-alternation training, various match-response and take-response ratios were used in attempts to increase the match-response accuracy for subjects P1S1 and P2S1, without success.

One subject in each pair either did not acquire follower responding as quickly as had been expected, or did not reach as high a level of accuracy as had been expected. Therefore, rather than changing immediately from follower training, in which subjects remained in a response role for tens of sessions, to a response-role alternation training condition in which roles would be switched each trial, response role alternation frequency was increased gradually across sessions. The programmed frequency of role alternation was increased across conditions from session-by-session alternation to trial-by-trial alternation.

Even with a gradual increase in response role alternation, follower match-response accuracy was expected to be disrupted by changes in role alternation frequency. However, in some cases follower accuracy actually increased with introduction of a higher alternation frequency. Figure 3 shows increases in follower accuracy for subjects P1S1 and P2S1 in changing from tenth-session alternations to trial-by-trial alternation end for P1S2 in changing from session-by-session alternation to half-session alternation. Given the lack of a systematic disruption of follower responding by the condition changes, the gradual change to trial-by-trial alternation was probably unnecessary. Time could be save in future experiments using these procedures by beginning trial-by-trial alternation immediately after the follower training.

Test of social control of follower match-responding.

The sessions plotted in the first panel of Figure 4 for each subject are the final condition of the role-alternation training (trial-by-trial alternation condition). In the second panel, the clear Plexiglas partition between the compartments of the chamber has been replaced with an opaque partition. The final panel shows match-response accuracy upon return of the clear partition.

Figure 4. Percent correct during test for social control of the follower responding.

Pair 1
Pair 2
Condition Description A Clear partition B Opaque partition

For all subjects, follower match-response accuracy decreased during the opaque-partition condition and increased to previous levels upon return of the clear partition. Three of the four subjects averaged approximately 50% correct while the opaque partition was in place (P1S1, P1S2, and P2S2). The fourth subject (P2S1) averaged approximately 56 during the opaque-partition condition.

Neither subject in pair one showed either a systematic change in accuracy during the condition with the opaque partition or above chance-level response accuracy, demonstrating control of follower match-responding by the visual social stimuli of the responding of the other subject. However, both subjects of pair two showed a change of accuracy across the condition. The follower match-response accuracy for P2S2 increased from 30% to 59% correct during the opaque-partition condition; follower match-response accuracy for P2S1 first decreased from 77% to 35% correct, then increased to 65% correct before the clear partition was re-introduced.

Subjects should have been able to respond with an accuracy of 50% by "chance" given the random selection of the key to be correct on any given trial by the process control equipment. In the absence of visual social discriminative stimuli for follower responding, follower accuracy of greater than 50% correct during the opaque partition condition indicates control by other than visual social stimuli. The above chance-level accuracy of the follower responding of P2S1 can be accounted for by a preference for the upper of the two matching keys during follower responding. Averaged across the opaque-partition condition, the follower responding of P2S1 was correct on 83% of the upper match-key trials and correct on 41% of the lower match-key trials.

The change in the follower response accuracy of P2S1 across the opaque-partition condition was accompanied by a change in the distribution of trials on which the upper key was correct as opposed to trials on which the lower key was correct. The sessions in which P2S1 responded with a higher follower accuracy were sessions in which there occurred an approximately even split in the number of trials with the upper key correct and with the lower key correct; the sessions with lower P2S1 follower response accuracy are those sessions with more trials on which the lower key was correct (e.g., the third opaque-condition session, in which P2S1 responded with a follower accuracy of 35%, had twice as many lower-key trials as upper-key trials). The preference for the upper match-keg by P2S1 did not exist before the opaque-partition condition, but did continue through later conditions. Subject P2S2 showed an even stronger upper key preference, initially responding on the upper key on all lower-key trials. This upper key preference gradually decreased across the sessions of the opaque-partition condition and was accompanied by an overall increase in follower accuracy for the subject.

The decreased follower response accuracy in the absence of visual social discriminative stimuli demonstrates social control of the follower responding.