Demonstration of Concept Formation in the Horse

Lisa M. Watt and Sue M. McDonnell
Equine Behavior Laboratory
University of Pennsylvania School of Veterinary Medicine
August 2001 Interim Report*

    Higher learning abilities of animals have been the focus of intensive study in recent years, spawning a fascinating branch of science known as animal cognition. One of the most important of the higher learning abilities is that of concept formation. While forming a concept, the animal presumably makes mental representations of a class, encompassing knowledge obtained regarding that class for future application. Concept formation has been studied mostly in pigeons, but also in African grey parrots, primates, dolphins, pinnipeds, dogs and horses. In pigeons, for example, a typical experiment involves positive reinforcement for demonstrated discrimination among pairs of projected images differentiated by category such as person/nonperson, tree/nontree, etc.

    In horses, four studies on concept formation and categorization have been reported. Sappington and Goldman (1994) evaluated the ability of horses to discriminate triangular objects from non-triangular objects. One out of a total of four subjects reached criterion. Flannery (1997) evaluated three horses for the concept of identity-matching-to-sample, or sameness vs different; Hanggi (1999) studied the concept of open vs filled objects (black two-dimensional objects with white centers vs all-black two-dimensional objects) using two equids; and finally, Hanggi (2001, unpublished) studied the concept of relative size using both two- and three-dimensional objects in one horse.

    The objective of the present study was to attempt to demonstrate the formation of simple concepts in equids using a novel study design which 1) is procedurally streamlined compared to those previously reported, 2) employs tactile rather than visual cues for the discrimination of objects, 3) is based on the concept of relative compressibility (e.g., harder vs softer), and 4) employs ponies as opposed to horses as subjects. The simple design was chosen in an effort to improve efficiency. We specifically aimed to reduce the inter-presentation interval and to provide a highly palatable food reward as close as possible in time and space to the response. The general design of the training and testing procedures includes an operant conditioning paradigm with initial positive reinforcement-based shaping of a nuzzling response for two-choice discrimination of hard vs soft, followed by testing and/or continued shaping of discrimination of harder vs softer among paired combinations of objects of three levels of compressibility. We concluded the study with tests of the concept using a variety of novel objects of variable and multiple increments of compressibility.

    So far, eight ponies (1 mare aged 5 y; 2 geldings aged 10 y and 11 y; and 5 stallions aged 3-27 y) have entered and/or completed the study. The training and testing apparatus consists simply of three deep plastic trays secured side-by-side to the top of a plastic cart placed in the manger of a standing stall. The outer trays served as stimulus object compartments, and the center tray forms a compartment into which food treats are dispensed. Initial two-choice discrimination of soft vs hard is accomplished using one sponge and one wooden block of similar size and shape disguised with similar covering and similarly sealed within a clear plastic bag. Continuous schedule positive reinforcement consists of palatable treats (carrots, apples, cookies, popcorn, grapes, etc., varied within and between trials to maintain appetite and interest) dispensed immediately following indication of correct choice. Subjects are initially shaped to develop sampling and then distinct selection (vigorous intense nuzzling, or grasping and lifting the correct object, or tossing the selected object from the compartment). For five of the subjects, the designated correct choice was the soft object, and for the remaining three subjects the choice was the hard object. To test for possible inadvertent visual cueing, once performance reaches levels well above chance, the procedure is repeated with the subject fully blindfolded. To then test and/or further train the concept of relative compressibility, the subject is given a two-choice discrimination task using a series of combinations of three objects of varying compressibility (solid, intermediate and soft). In this procedure, the correct response to a solid-intermediate presentation would be an incorrect response to an intermediate-soft presentation; thus, relative compressibility is inherent. The final test for the concept of relative compressibility is a series of 30 presentation of pairs of approximately 60 novel objects that differ in size, shape, color, surface color patterns, texture, and that represent multiple levels of compressibility.

    Our progress to date includes:
1. Three subjects have completed the study. On the three-level, 2-choice discrimination test, as well as the multi-level novel object test, each has demonstrated accuracy rates beyond those expected by chance alone (P < 0.05).
2. We expect that at least 3 of the 5 remaining subjects will similarly learn to demonstrate understanding of the concept of relative compressibility.
3. Progress for the shaping as well as the demonstration of relative compressibility using tactile stimuli (average of twelve 10-20 minute trials over a period of 1-3 weeks) has been far more rapid than that reported in previous studies (daily 20-40 minute trials over months).
4. Subjects have performed similarly well in the two-choice simple discrimination with blindfolds. Sampling has typically been slower and more tentative with the blindfold as might be expected, but accuracy has remained at or above criterion.
5. Subjects also appear to respond positively to the short inter-trial interval as well as to close spatial proximity of the stimulus objects and the reinforcement compartments.
6. Shorter trials (10 to 15 minutes) appear to be more effective in holding attention of the subject and vigorous response.

    Thus, we have tentatively concluded that concept formation occurs in equids, and that it can be readily demonstrated in ponies using a simple paradigm. It appears that the design of equine cognitive studies can be greatly simplified to enhance performance without compromising validity. These results represent an important step in the study of cognition in equids, as well as a contribution to comparative animal cognition. They also have practical implications for the training of horses and for assessment of this ability.

Flannery, Barbara, 1997. Relational discrimination learning in horses. Appl. Anim. Behav. Sci. 54: 267-280.
Hanggi, Evelyn B., 1999. Categorization learning in horses (Equus caballus). J. of Comp. Psych. 113 (3): 243-252.
Sappington, B.F., and Goldman, L., 1994. Discrimination learning and concept formation in the Arabian horse. J. of Anim. Sci. 72: 3080-3087.
Vauclair, Jacques, 1996. Animal cognition: an introduction to modern comparative psychology. Harvard University Press, Cambridge, MA.

Lisa Watt was born in Newfoundland and raised in British Columbia, Canada, and is a member of the Class of 2003 of Western College of Veterinary Medicine in Saskatoon. She was a 2001 Dorothy Russell Havemeyer Foundation Veterinary Student Summer Research Fellow. Lisa is a long-time horse owner, doing dressage and trail riding.

* Since completion of this interim report, Dr. McDonnell, Elkanah Grogan and Kimberly Unger have completed the study with 10 subjects total, all successfully completing the tests of concept formation.