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

This study attempted to demonstrate sharing/cooperation in pigeons. The first of the three sub-objectives of the current study was to produce leader and follower responding by each subject on alternate trials within sessions, in contrast to Hake, Donaldson, and Hyten (1983) in which subjects responded as either the leader or as the follower throughout an experiment. In Hake, Donaldson, and Hyten both subjects received food reinforcement in each successfully completed trial; the second sub-objective of the current study was to produce responding by the follower on any given trial that would produce food reinforcement for the leader without producing food reinforcement for the follower (i.e., to produce cooperative-like responding). The follower will "cooperate" to produce the leader's reinforcers.

The third question was whether subjects would "share' the leadership role when the subjects themselves determine the distribution of response roles on a trial-by-trial basis. That is, would both subjects respond to take the leader role (with food reinforcement) on each trial, or would a stable pattern of alternate-taking (sharing) of the leader response role across trials develop?

Method

Subjects.

The subjects were four White Carneaux pigeons maintained at between 70% and 80% of their free feeding body weights. The subjects of Pair 1 had previously been subjects in an experiment involving color discriminations and magazine cycle length discriminations. The subjects comprising Pair 2 were experimentally naive.

Apparatus.

The experimental chamber was a plywood box divided into two 205cm * 29.5cm * 32cm compartments by a clear Plexiglas partition. The clear partition was replaced by an opaque partition in same conditions. Figure 1 shows the front panel. The layout of the front panel in each compartment is a mirror image of the other.

Four translucent response keys were located in each front panel: A) two keys ("matching" keys) were arranged vertically 4.7cm from the Plexiglas partition: the lower key was 2 1.5cm from the floor of the chamber and the upper key was 7.5cm above the lower key, B) the third key (the "take" key) was 5.5cm from the outer edge of the panel and 21.5cm from the floor, and C) the fourth key (sample-key) was 7.5cm above the take-key.

A metal guard 4.9cm high protruding 5.2cm from the panel was inboard of the sample-key. The sample-key was visible only to the subject in the some compartment. All keys were 1.9cm in diameter and will require approximately 0.5N for operation.

Each panel also included a 5W houselight centered at the top of the panel, and a food magazine centered in the front panel 11.7cm from the bottom. White noise at 90db was used to mask extraneous sounds. Ventilation and additional masking noise (approximately 89db) were provided by a fan centered over the Plexiglas partition. Real-time process control for the experiment was provided by a PDP 8/A mini computer located in an adjacent room.

Procedure.

I. Preliminary training.

Throughout all phases of the study a subject responded in the same compartment of the chamber. Subject 1 of each pair was run in the left compartment of the chamber, subject 2 in the right compartment.

Keypeck training.

The naive subjects of Pair 2 were first trained individually to peck the upper of the two matching-keys by presenting groin (under conditions of food deprivation) following successively closer approximations to the keypeck.

Subjects in both Pair 1 and Pair 2 received at least three sessions of fixed ratios (FR 10, Pair 1; FR 1. Pair 2) on the upper matching-key to allow habituation to the chamber.

Training of color matching (leader responding).

Color matching-to-sample training was begun for subjects after the initial keypeck training. Color matching was trained in two steps: (a) the sample-key and the matching-key of the some color were lighted and effective, then once this two key sequence was established, (b) the take-key, the sample-key and both matching-keys were lighted and effective.

In the first step, trials started with the lighting of the houselight and the sample key. Sample key was lighted either red or green, the selection being clone at random with each color occurring with equal probability. Completion of the sample-key response requirement resulted in the lighting of only the matching-key of the same color as the sample-key, and in the sample-key being lighted white. Upon completion of the matching-key response requirement, all lights in the chamber went out, and a three second presentation of food began. The response requirements on the sample and matching keys were raised over the course of this step from FR 1 to FR 8 (Pair 1) and to FR 6 (Pair 2).

In the second step, trials began with the lighting of the houselight, and with the take-key being lighted white (throughout the study the take-key was only lighted white, when lighted). Completion of the take-key response requirement resulted in the take-key going dark and the sample-key being randomly lighted either red or green, as before. Completion of the sample-key response requirement now produced the lighting of both matching-keys. Throughout the study, the upper matching-key was lighted red, the lower was lighted green.

Completion of the response requirement on the matching-key of the same color as the sample-key (the "correct" key) resulted in termination of the trial and food reinforcement. "Incorrect" responses (responses on the matching-key not of the same color as the sample-key) restarted the chain with only the take-key and houselight illuminated. The trial was repeated with the sample-key lighted the same color as on the previous trial (on which the error occurred).

This correction procedure was intended to ensure that subjects did not develop strong position preferences and to ensure that subjects would respond on both matching-keys. The response requirement on the take-key and the lower matching-key was raised across sessions to between FR 8 and FR 10 for all subjects. The response requirement on the upper matching-key was set at FR 2. The difference in ratios between the two matching-keys was an attempt to correct a strong preference for the lower matching-key (the lower matching-key was the closer of the two matching-keys to the food magazine opening).

Once the accuracy of the color matching-to-sample appeared stable, training of social matching-to-sample (following) was begun.

Training of social matching-to-sample (following)

With the beginning of training of social-matching (follow responding), subjects were paired for the first time in the chamber. During follower training, the process control equipment determined the response role distribution (i.e., determined which subject of a pair was the leader and which was the follower). Also, food reinforcement was available for both leader and follower responding.

Leader responding was the same as in the color matching-to-sample training in all respects other than the consequences of completing the match-key response requirement. Upon completion of the match-key response requirement, the matching-keys in the compartment of the leader were lighted white.

Food reinforcement did not occur until the follower had completed its response requirement in some way. In the compartment of the subject responding as the follower, the houselight was lighted, but all response keys were darkened until the leader subject had completed its response requirements on the sample key and the take key (the take key functions are explained in the section on "free" response role distribution). Upon completion of the sample-key response requirement by the leader subject, the matching-keys of the follower subject lighted.

Upon completion of the response requirement by the follower on the matching-key which corresponded to the matching key on which the leader responded (i.e., the "correct" key), the trial ended and the food was presented to each of the two subjects, unless the leader subject had not finished the response requirement on the matching-keys in the compartment of the leader. If the leader not finished responding, the matching-keys of the follower were lighted white and became ineffective. The trial ended and food reinforcement occurred for both subjects upon the leaders completion of its matching-key response requirement.

If the follower responded on the matching key which did not correspond to that on which the leader responded (i.e., the follower responded on the "incorrect" key), all keylights and the houselight in the follower compartment were extinguished and food reinforcement became unavailable. Upon completion of the matching-key response requirement by the leader and food presentation to the leader subject, a "correction" trial began in which the same matching-key was correct, and the some subject was leader (if intra-session alternation of response roles was occurring).

If the leader responded on an incorrect matching-key, the trial immediately ended, and neither subject received food. Instead of a food presentation, a blackout period occurred which was of the some duration as the food magazine cycle.

Response role alternation.

Leader and follower response roles were alternated between subjects of a pair, as summarized in Table 1, first in blocks of several sessions, then by session, then by blocks of trials, until trial-by-trial alternation of response roles had been established. Changing of alternation modes was determined by the accuracy of the social-matching responding of the follower. Response roles and/or response role alternation modes were changed when graphed accuracy appeared stable relative to the variability in accuracy across sessions.

Matching-accuracy manipulations.

Various response requirements on the take-keys, sample-keys, and matching-keys were used in attempts to increase accuracy of the matching responding and to reduce the variability in the response accuracy across sessions. Various durations of food presentations and numbers of trials per session were used for the some purposes.

In all calculations of accuracy on the matching-keys the correction trials were not included. Accuracy on these trials may be biased by the "loose-shift" contingency of the correction procedure (cf., Hake, Donaldson, and Hyten, 1983). At this point in the preliminary training, the procedure to be used in the "sharing" procedures was in effect with two exceptions: (a) only the leaders take-key was lighted and effective, and (B.) both subjects received food reinforcement if they completed their respective response requirements.

Test of social control of inter-animal matching-to-sample.

In order to test for social control (i.e., that the responding of the follower on the matching kegs was under the discriminative control of the responding of the leader) the clear partition between compartments was replaced with an opaque partition. The opaque partition was in place for few sessions in order to avoid the possibility that the responding of subjects would come under the control of previously ineffective non-visual stimuli in the absence of visual stimuli (cf. Hake, Donaldson, and Hyten, 1983). The clear Plexiglas partition was then replaced and responding was allowed to return to previous levels.

II. "Free" distribution of the food reinforced response role

During the "free" distribution phase of the study, response roles were distributed by the subjects. Distribution of the response roles was dependent upon the responding of subjects on the take-keg. The first subject to complete the response requirement on its take-keg responded in the leader response role on that trial; the other subject responded as follower.

The procedure on each trial is detailed below. Food reinforcement of follower responding was manipulated (a) in order to determine whether or not the take-keg responding of the subjects was under the control of the differential reinforcement densities offered by the two response roles, and (b) whether or not the differential reinforcement densities associated with each response role produced any social effects on the take responding of the subjects. Food reinforcement of follower responding was alternated with no food-reinforcement of follower responding in an ABAB sequence for pair 2, and a BAB sequence for pair 1.

The original purpose of the free distribution phase of the experiment was to set up contingencies which would generate competitive responding by subjects, with an eventual shift to share responding.

Basic procedure on each trial.

Each trial was divided into three intervals which occurred in the following order: (a) a distribution interval, (b) a matching-to sample interval, and (c) a food and inter-trial interval (food/ITI).

Distribution interval.

This initial interval started with the illumination of the houselights and white illumination of the take-key in each compartment. The first subject to complete the take-key response requirement (e.g., 5 responses) responded as the leader on that trial, that is, was the stimulus bird for the matching-to-sample responding of the other subject. The other subject responded as the follower on that trial, that is, matched to the leader's sample on the matching-keys.

One subject was able to take the leader role more often than the other by completing the take-key requirement more quickly. To avoid extinction of the responding of the slower of the two subjects in a pair, the take-key response requirement was a continuously-adjusting ratio requirement. At the end of each successfully completed trial (i.e., neither subject had pecked a key designated "incorrect" during the trial) the take-key ratio of the subject taking the leader role on the trial was increased by one response, while the take-keg response requirement for the subject that responded in the follower role during the trial was lowered by one response.

Matching interval, leader responding.

The matching interval began with the sample-key in the compartment of the leader being lighted either green or red; red or green was chosen at random on each trial. A ratio of between one and five responses on the sample-key by the leader lighted the matching-kegs in the leader compartment. The upper matching-keg was lighted red and the lower matching-key was lighted green in all cases. Pecks on the matching-key lighted the same color as the color that the sample-key had been lighted (the "correct" key) resulted in the lighting of the matching-kegs in the follower compartment and resulted in the sample-key in the leader compartment being lighted white.

A number of responses were required by the leader on the correct matching-key to produce food at the end of the trial (5 to 15 responses, depending upon matching accuracy of the partner bird as follower). If the leader completed its response requirement before the follower completed its requirement, the matching-keys in the leader compartment were lighted white until the end of the trial. A minimum response requirement for the leader was used in order to increase the probability of producing enough responding to serve as a stimulus for the responding of the follower. For example, if the follower was not facing the front panel at the beginning of the trial, the leader was likely to have finished its match-key response requirement before the follower began responding if the leaders match-key response requirement was a single keypeck. The leader was more likely to still be responding if the leaders response requirement was five keypecks.

Pecks by the leader on the key not of the same color as the sample-key (i.e., a response on the "incorrect" key) under all circumstances turned off all keylights and restarted the trial with the same leader and with the same sample color as was presented in the trial in which the error was made (leader correction procedure).

Matching interval, follower responding.

The matching interval started for the follower with the matching-keys being lighted. A minimum number of responses was required by the follower on the correct matching-key for the trial to end (usually between 2 and 5 responses; response requirements on the matching-keys were manipulated in attempts to increase the accuracy of matching-to-sample responding). If the follower completed its response requirement before the end of the trial, the follower matching-keys were lighted white until the end of the trial (i.e., until the leader completed its response requirement). The "correct" key for follower responses was the matching-key which corresponded to that on which the leader was responding.

Pecks by the follower on a matching-key not of the same color (and position) as the key on which the leader responded were treated as errors. Follower errors ended the matching interval for the follower, as well as for the leader if the leader had completed its response requirement. On the next trial the same bird was leader as in the trial in which the follower error occurred, and the sample-key was lighted the same color as in the trial in which the error occurred (follower correction procedure).

When both subjects had completed their respective response requirements, all lights were darkened, the matching interval was ended, and the food/ITI interval was started.

Food/ITI.

The food and inter-trial interval was of 1.5 to 4 seconds in duration, depending upon factors discussed below. The leader food magazine and magazine light, and in some conditions during the "free" distribution phase, the follower food magazine and magazine light, were operated throughout the interval for the subject(s) that had completed the response requirement on the "correct" key. All other lights were off. At the end of the interval, presentation of the food magazine and the magazine light were terminated and the next distribution interval began.

Subjects received 100 trials per session and one session per day. The duration of the magazine cycle (and therefore the ITI, also) were adjusted in order to maintain the weight of the pigeons above 70% of their free feeding weights while receiving most of their food during a session.