Individualized Target Training Facilitated Transfer of Group Housed Capuchin Monkeys (Sapajus apella) to Test Cubicles and Discrimination of Targets on Computer Touch Screens

Abstract

Animals in captivity often experience fear, anxiety and aggression during non-voluntary procedures, leading to adverse behaviors and ineffective outcomes for both animals and caretakers. Negative reinforcement and punishment, often due to ignorance regarding animal learning, can hurt animal welfare. However, voluntary participation through positive reinforcement training (PRT) can decrease stress related to these procedures and increase desired behaviors. Our goal was to demonstrate the positive effects of "target training" on animal welfare by training 10 captive capuchin monkeys (Sapajus apella) in two experiments designed to facilitate movement from a group home enclosure to a test cubicle. In Experiment 1, each monkey was assigned an individualized target (a unique shape/color combination). In daily training sessions, the animal was rewarded with a click-sounding stimulus and a food reinforcer for (a) touching the target, (b) following the respective target into a test cubicle, and (c) touching progressively smaller targets until progressing to digitized images on a computer touch screen. All 10 animals learned to approach and touch their individual physical target in one or two sessions and were able to successfully transition this behavior to an image of their target on a touch screen, although they made more errors with the touch screen. In Experiment 2, the animals were presented with other animals' targets and novel targets. The seven animals in this experiment all touched their target at higher-than-chance rates in Trial 1 without explicit discrimination training, but only five reached the learning criteria for the task (>83% correct for three consecutive testing days. These results demonstrate that target training can make voluntary movement from group housing to test cubicles easier and benefit future animal care and procedures.

Keywords: animal husbandry; animal welfare; capuchin monkey; learning theory; positive reinforcement training; target training.

Figure 1

A scaled diagram of the enclosure and test cubicles.

Figure 2

Photographs of the facility from the testing room (top) and the entry area (bottom). On the white wall on the right side of the bottom photograph, four of the 16 sliding doors to the testing cubicles are visible. These doors can be secured open or closed, so caretakers and trainers have the option to close either the mesh screen door, the solid metal panel door, or both.

Figure 3

Photograph of one of the eight upper-level cubicles used in this experiment taken from the testing room. At the rear of the cubicle, the solid guillotine door leading to the capuchin enclosure is closed. At the front of the cubicle, the cubicle door consists of a stainless-steel mesh upper portion and a Plexiglas^TM window that can be secured at several different heights.

Figure 4

Images of the physical targets assigned to each monkey in the colony, together with the wavelengths of the light they reflected in the visible spectrum. The targets for Jamie, Lucky, and Marcel do not reflect light from a single well-defined band. Targets were originally made out of laminated construction paper and were replicated as precisely as possible when moved to the computer touch screen use. Note the overlap of shape and color for several of the targets.

Figure 5

Capuchins completing the target behavior through mesh (top left), using the pole target (top right), across the Plexiglas^TM (bottom left), and on the touch screen (bottom right).

Figure 6

Mean Number of 20-trial sessions taken to reach the criterion of two consecutive errorless 10-trial blocks for each of the four training stages in Phase 1. Errorless responding would result in reaching the criterion in the initial session. Different letters represent significant differences in mean sessions to learn the task; error bars show standard errors.

Figure 7

Response error rate by session in Phase 2: Stage 6, averaged across capuchins. Error rate was calculated by determining the number of errors divided by 27 trials per session. No significant differences were found across sessions.

Figure 8

Error rate on the first trial for different trial types. No significant difference was found between conditions (F(2, 12) = 0.68, p = 0.53, partial Eta² = 0.10). Error rate was calculated by determining the number of errors divided by the number of trials of that type presented in the 53-trial session. Reference line shows chance performance; error bars show +/− 1 SE. Performance was significantly better than chance in the “different color, same shape” and the “different color, different shape” conditions, but not in the “same color, different shape” condition.

Figure 9

Mean error rate decreased significantly across trial blocks. Error rate was calculated by determining the number of errors divided by the number of trials of that type presented in the 53-trial session. There was no significant effect of trial type (F(2,12) = 1.22, p = 0.33, partial Eta² = 0.16).