PRELIMINARY INVESTIGATION OF EQUINE LEARNING AND MEMORY
LAUREL J. DUNN
DEPARTMENT OF PSYCHOLOGY
MISSOURI WESTERN STATE COLLEGE
------------------------------------------------------------------------
Bold face text passages in the article are from the web site developer.
ABSTRACT
Six horses demonstrated observation and discrimination learning and memory ability. The purpose of this study was to gain further knowledge in the area of equine learning. Performance on each task was compared in a single subject design. Subjects learned to discriminate between a black and a white bucket. The criterion for learning was set at 80% correct black bucket choice. All subjects successfully performed the discrimination task by the eighth session. Observation learning was unsuccessful; no subject reached the 80% correct criterion. Five horses were tested for memory retention of the discrimination task three weeks after the initial learning. All subjects performed the discrimination by the second session of two. These data support existing results from similar learning and memory tasks.

------------------------------------------------------------------------
INTRODUCTION
A horse's performance relies, not only on its predisposition and genetics, but also on the trainers' and handlers' competence, and the horse's health and environment. Worth may be determined by its aptitude for a specific type of work or its performance versatility. The ease in which a horse may be trained will add or detract from worth.

A reliable method of assessing a horse's ability to learn (its trainability), would be beneficial to most horsemen. Currently there is no efficient or accurate way to determine a horse's trainability. Equine research has only begun to investigate learning ability, emotionality, physiology, genetics and how these things relate to performance and to each other (Wolff, Hausberger, and Le Scolan, 1997). If horses' intelligence can be correlated to trainability, then intelligence may be of predictive value.

Assessing a horse's intelligence poses difficulties. Perception, evolutionary function and communication differences in the horse, as with most research animals, constrain methodology and need to be identified. Research in equine perception has shown horses are able to perceive color. Results have been contradictory however, as to the exact colors they are able to see. Interpreting information from a study done in 1994, Budiansky (1997) suggests horses are capable of seeing only reds and blues, but not greens. The ability to perceive color was also investigated by Smith & Goldman (1999). Horses were able to correctly distinguish red, yellow, green and blue from a series of varying intensities of gray. Horses have been found to have at least dichromatic vision, seeing red and blue, although color-blindness may occur in some individuals.

Evolution may play an active role in the types of information attended to and ability to learn certain concepts (Budiansky, 1997). Horses may also attend to differing cues than humans would normally attend. It has been suggested, artificially selected breed differences play a role in breed traits such as emotionality, behavior and learning ability (Mader & Price, 1980).

Questions also remain concerning the ability of horses to learn through observation. Learning curves have been obtained, showing the observation group's initial acquisition rate higher than the control group. Subsequent learning, however, suggested no statistically significant difference between the two groups (Baer, Potter, Friend, & Beaver, 1983). These results indicate observation of another horse performing the task has more of a priming effect than an overall increase in learning acquisition. Studies done in the past have measured the ability of a horse to observationally learn from another horse, rather than observation of a human performing a task (Linberg, Kelland, & Nicol, 1999; Clarke, Nicol, Jones, & McGreevy, 1996).

Horses are capable of higher order learning. They are able to make at least 20 concurrent associations, compared to a mouse, which makes seven and a rat, which makes eight (Budiansky, 1997). Evelyn Hanggi, Ph.D. has done research illustrating the horses' capability in choosing one object over another, depending on the situation. This conditional concept is the seventh level of Roger Thomas' Hierarchy of Learning. Horses are also able to demonstrate ability in accomplishing the eighth, and last, level. This concept involves placing value on an object, where red is better than blue, but blue is better than green and so on. Contrary to popular belief, horses are capable of complex cognition (Pottenger, 1998).

To some extent scientists have demonstrated horses' ability in memorization, spatial, and instrumental learning (Wolff & Hausberger, 1996). Recent studies have been done assessing instrumental learning and associative learning (Pottenger, 1998). In general horses are capable of learning many different tasks, but how this relates to performance and how performance on learning tasks relate to trainability, has yet to be ascertained. The concept of a relationship has been investigated, but no statistically significant differences were found between discrimination reversal learning and performance (Sappington, McCall, Coleman, Kuhlers, & Lishak, 1997). Further equine research, involving all types of learning needs to be done.

Research into human cognition and intelligence indicators and predictors, is still being done and many questions remain. In the area of equine research, information about these concepts, are even further behind and moving at a slower rate. The purpose of this study was to provide further data, extending previous research in the area of equine learning and memory. A discrimination learning, observational learning and memory task was measured and compared in a single subject design using six horses.

------------------------------------------------------------------------
METHODS

PARTICIPANTS
Six horse subjects were drawn from and tested at, Rose Red Farm in Northeast Kansas. There were four mares at age five, consisting of three grade and one Thoroughbred. There was a two-year-old Quarter horse stallion, and a gelding Shetland pony over the age of 20. All horse subjects have differing backgrounds, breeding and training. All subjects are handled by and familiar with, the experimenter.

Under non-experimental conditions, five of the horses are used for breeding purposes and the last of the six, as a companion for new horse arrivals. Subjects are pastured together and fed Brome hay during winter. All subjects, except the Shetland pony, receive a 12% protein sweet feed every day during fall and winter. Shelter and water are openly accessible at all times, to all subjects and each is familiar with the surroundings in which they have been tested.

MATERIALS
Three testing procedures were done and measured. The first test was a discrimination learning task and the last a measurement of retention of the discrimination task. The equipment was a black and a white bucket identical, except in color, separated by a four by four feet wooden partition. Paper and pencil were used to record subject's discrimination, and results for the memory task. The regular 12% protein feed was given as a positive reinforcer for all testing procedures. The second test was to assess the horse's ability to learn through observation. Two purple buckets and a three feet high by five feet long wooden table for bucket placement were used.

PROCEDURE
To insure sufficient motivation all subjects were given the morning feed at regular feeding time; testing for all tasks started no earlier than three hours after this and no later than one hour before the afternoon feeding. To minimize separation anxiety, all horses were separated, but within view of the test subject.

Test 1. Each horse demonstrated discrimination learning ability. Testing was done in a small rectangular paddock. The positive reinforcer was always in the black bucket and the horse learned to choose this over the white bucket to the criteria of at least 80% of the time. The horses were given eight sessions composed of ten trials, to learn the discrimination. Two sessions were given on Saturday and Sunday, for two consecutive weekends. Patterns for the right and left placement of buckets were chosen prior to each session. No bucket was in the same position more than three consecutive trials.

Each horse was held by the experimenter six feet away and centered on the wooden partition separating the black and white buckets. Once the assistant had prepared bucket placement for the trail, the horse was released and allowed to choose. A choice was made by selecting one side of the partition over another and so, one bucket color over another. Subject, session, trial, bucket placement, color and selection by the subject, were recorded on a pre-designed sheet.

Test 2. This task involved observational learning assessment. Testing was done in one area of a round pen sixty-six feet in diameter. A purple bucket was placed at each end of the wooden table. Alternation of reinforcement placement was chosen prior to session start, as in the previous testing procedures. The subject, held by the experimenter four feet away from the buckets, was allowed to view the assistant placing the grain reinforcer in one of the buckets. The assistant faced away from the buckets, to minimize unconsciously given cues for correct bucket choice. The subject was then released to make a selection. The subject, trial, reinforcer placement and selection were recorded.

The subject had successfully learned through observation when the correct bucket was selected 80% of the time during two consecutive sessions consisting of ten trials each.

Test 3. Three weeks following initial discrimination learning, five subjects were tested for memory retention. The subject again discriminated between the black and white bucket, choosing the black bucket at least 80% of the time during a session of ten trials.

------------------------------------------------------------------------
RESULTS
A simple linear regression was calculated predicting subjects' discrimination learning based on session number. A significant regression equation was found (F (1, 46) = 52.181, p < .001), with an R2 of .531. Subjects' predicted discrimination is equal to 33.81 + 7.024 (session number). Subjects' average correct discrimination choice increased 7.024 for each session and is shown in comparison to mean results of all subjects in the observation task in Figure 1.

A simple linear regression was also calculated predicting subjects' observation learning scores based on the session number. The regression equation was not significant (F (1, 46) = 1.269, p > .05) with an R2 of .027. The number of completed sessions could not be used to predict observational learning scores. Observation and discrimination percentages for each subject can be seen in Table 1.

Mean correct response for five subjects on the discrimination learning memory task was calculated for each of the two sessions. The mean for the first session was eight correct with a Standard deviation of 1.225. The mean for the second session was 9.4 correct responses with a Standard deviation of .548. Subjects correctly responded about 80% of the time on the first session with a mean increase to 94 % correct on the second session (Table 2).

------------------------------------------------------------------------
DISCUSSION
The majority of horse training is done through negative reinforcement, while most laboratory experiments use positive reinforcement. Because of this, questions arise as to the validity in generalizing these experimental findings to a horse's trainability. However, a study done in which horses were given tasks to learn with both positive and negative reinforcement, showed a high positive correlation between the two (Budiansky, 1997).

This study assessed discrimination learning, observation learning and memory in six horses. All horses were able to learn the discrimination task and each session produced a predictable mean increase in learning. There were individual differences between each horse, with approximately 50% of the variance accounted for. From a subjective observation it was noted that the horses which learned quickest had also been the most difficult in training to ride.

Observation learning and imitation is one theory explaining the horses' acquisition of stereotypic behavior, such as cribbing (Cooper, 1999; Siegal, 1996). Another theory is that horses stabled in the same environment tend to behaviorally react in the same manner. Experiments have failed to confirm an equine ability to learn from observing other horses or from humans (Budiansky, 1997; Baer, Potter, Friend, & Beaver, 1983).

Results from the observational learning task lend further credence to the theory that horses may not necessarily acquire bad habits from other horses, but from conditions under which they are subjected. Not one horse from the six was able to demonstrate the ability to learn the appropriate bucket choice, even after just observing placement of the grain reinforcer.

Horses have demonstrated, under experimental conditions, to have excellent memories. Mariner and Alexander (1994) demonstrated horses' ability to run errorless trials through a maze, two months after initial conditioning. From personal experience most owners and trainers know horses will remember associations once made, will be retained for extreme periods of time. It may also be difficult to reverse a behavior, once the horse has learned or made the association (Sappington, McCall, Coleman, Kuhlers, & Lishak, 1996).

In this study subjects were assessed three weeks after initial acquisition of the discrimination task. In affirmation of memory ability, four horses achieved at least 85% accuracy by the second session. One horse obtained a score of 60% on the first session, but by the second session 90%. The below average performance for the one horse on the first session may have been due to nervousness or excitability.

Generalization of these findings to the equine population can be made when viewed in conjunction with prior research. Although all subjects were representative of average horses at pasture, the sample size was small, selection biased. All subjects were selected for reproductive potential, not learning capability (trainability). Sample size, specific breed, emotionality and level of training should be taken into consideration when generalizing to the population.

Results from further investigation, controlling these variables may be beneficial in providing accurate predictors of a horse's performance potential. Operant conditioning tasks correlated with evaluations of a horse's trainability or past successful performance may prove useful for assessing future equine purchases.

REFERENCES
Baer, K. L., Potter, G. D., Friend, T. H., & Beaver, B. V. (1983). Observation effects on learning in horses. Applied Animal Ethology, 11, 123-129.

Budiansky, S. (1997). The nature of horses: Exploring equine evolution, intelligence, and behavior. New York: The Free Press.

Clark, J. V., Nicol, C. J., Jones, R., & McGreevy, P. D. (1996). Effects of observational learning on food selection in horses. Applied Animal Behavior Science, 50, 177-184.

Cooper, J. (1999). Learning in horses. Available: http: equinevetnet.com/pages/ animalscience/behavior/papers/learning.html (1999, February 5).

Lindberg, A. C., Kelland, A., & Nicol, C. J. (1999). Effects of observational learning on acquisition of an operant response in horses. Applied Animal Behavior Science, 61, 187-199.

Mader, D. R., & Price, E. O. (1980). Discrimination learning in horses: Effects of breed, age and social dominance. Journal of Animal Science, 50, 962-965.

Pottenger, D. (1998). Sense and sensibility. Horse Illustrated, November, 48-55.

Sappington, B. F., McCall, C. A., Coleman, D. A., Kuhlers, D.L., & Lishak, R. S. (1997). A preliminary study of the relationship between discrimination reversal learning and performance tasks in yearling and 3-year-old horses. Applied Animal Behavior Science, 53, 157-166.

Siegal, M. (1996). U. C. Davis book of horses: A complete medical reference guide for horses and foals. New York: Harper Collins.

Smith, S., & Goldman, L. (1999). Color discrimination in horses. Applied Animal Behavior Science, 62, 13-25.

Wolff, A., Hausberger, M., & Le Scolan, N. (1997). Experimental tests to assess emotionality in horses. Behavioral Processes, 40, 209-221.

Wolff, A., & Hausberger, M. (1996). Learning and memorization of two different tasks in horses: The effects of age, sex and sire. Applied Animal Behavior Science, 46, 137-143.

Controlled Study of Early Handling and Training of Icelandic Foals
Hrefna Sigurjónsdóttir1 and Víkingur Gunnarsson2
1Iceland University of Education, Stakkahlid, 105 Reykjavik, Iceland
Voice 354- 5633858       Fax 3545633833       hrefnas@khi.is
2Hólar Agricultural College, Skagafjördur, Iceland vikingur@holar.is

Introduction

Many scientists agree that information on the learning abilities of horses should be used in developing training methods (Fiske and Potter, 1979; Mader and Price, 1980; McCall, 1990) but nevertheless research where such knowledge is put to the test is limited (Flannery, 1997). Foals that are handled continuously between 4 and 22 months of age were quicker learners and were easier to train than those that got less handling (Fiske and Potter, 1979). The handling involved being lead, brushed, and acclimated to restraint. In some studies more relaxed foals learned more and were easier to train later on than stressed foals (Fiske and Potter, 1979; Heird et al, 1986) while no such correlation was found in another study (Mader and Price 1980). It is commonly believed, and has been established with tests, that younger horses are quicker-learners than mature (Mader and Price, 1980; Houpt, 1982), but the question remains to be answered if a certain age is better than another and if so, if the optimum time differs between breeds.

The idea of imprint training (Miller, 2000) is controversial but seems to be popular in the USA. Miller found that foals that had been treated by veterinarians at birth were easier to handle later on. He associated this with what Konrad Lorenz called imprint learning (1937), which involved irreversible learning taking place during a sensitive period early in life. Miller claims that the idea has been tested scientifically but no such papers are covered by Wed of Science. To our knowledge one systematic research has been carried out whose results will be published this year (Jennifer Williams, personal communication). Miller (2000) claims that it is very important to stroke and handle the foals quickly after birth  (hence the reference to the imprinting concept) and desensitize them to touch and other interventions. A study by Mal et al (1994) did not support the claim that very young foals are more sensitive to treatment than older foals. It is possible that what we are witnessing when the foals seem relaxed after stroking is not “imprinting” but “learned helplessness”.

In Iceland where mares give birth out in the field within the herd (usually herds of mares, sub-adults and sometimes geldings) it is customary not to disturb the mare and foal for the first hours of the foal’s life. That fact and the custom in many places to let the foals stay with their dams within large groups in the wilderness from 1-2 months of age till autumn, would mean that continuous handling would not have a practical value for most Icelandic horse owners. We wanted to test the idea that early handling and training of foals should calm the foals considerably and teach them valuable lessons, which would make training later on easier.

Methods

Subjects. Twenty-two foals, 10 colts and 12 fillies, from Holaskóli (Holar Agricultural College) and 3 farms in Skagafjördur, Iceland, were treated 4 times. In the control group there were equal number of foals from the same farms, 9 colts and 13 fillies. All the mares and the foals were kept in fields close to the farms with many other horses. During the treatments and the tests the dam was caught in an enclosure and taken under halter to a barn with the foal following. The mare was kept standing close to the foal (1-3 m) and attended to (given silage /stroked) if it was stressed.   

Treatments. The first treatment which involved rubbing the foal all over its body while being held lying down as recommended by Miller (2000) and inserting the finger into the mouth and the ears (but not as often nor as fast as Miller does), was carried out within the first 24 hours of the foal’s life. The second treatment which involved repeating the first one but this time the foal was standing, was done 1, 2, 3 or 4 (1 foal) days later. The foal was held and given a firm push at the both ends until it stepped backwards and forward. Then it was haltered and the head pulled gently to both sides a few times. The third treatment was done 1, 2, 3 or 4 (2 foals) days later. The second treatment was repeated and in addition the foal was lead. A body rope was put around the foal in such a way that by lifting the head the rope got tight around the hindquarter. The foal was encouraged to follow its dam the first round but then it was lead away from the dam. The fourth treatment was carried out 4, 5, 6 or 7 days later and involved repeating the third one. The reactions of the foals were noted and the time it took to carry out the different parts of the treatments. The behaviour of the dams was also noted. The foals were handled between 15th of May and the 20th of June 2000.

Test. Between the 10th and the 13th of October 2000 all handled and control foals were tested by the same person (experimenter) who had not been involved in the treatments in the spring and remained blind to group assignments. An assistant helped. All the foals had been in the mountains during the summer. During the test procedures, the dam was always within reach. 16-20 m2   indoor area was used as a test area in all the farms. The time it took to catch the foal and hold it still was taken, then it was haltered, legs lifted and hoofs hammered, ears, gums, belly and area underneath the tail stroked, the response to 1-3 pushes to the front and the hindquarters scored and also the response to the lead to left and right and how well it followed. The experimenter graded foal on each test. The time it took to carry out all the tests was measured and finally the foal was graded (on scale 1-5) on overall compliance.

Video recordings. The behaviour of the foals during the test was classified and measured (frequencies and time). The program Observer (produced at Noldus, Wageningen, Netherlands) was used in the recording and analyses. The behaviour classes used were: walk, run, flee (run fast), stand still, shake head, pull against the lead by standing, fall on the ground, and strongly resist.

Temperament of the dams. The owners of the mares answered a questionnaire rating on the scale 1- 3 or 4 each of the following: ease of training, ease of handling, ease of approach, and general level of nervousness.

 Statistical analyses. Chi square tests were used to compare groups (early handling/control) in the tests where the distributions pointed towards significant differences. Mann –Whitney U tests and t- tests were carried out to compare time measurements and frequencies. Correlation tests were carried out on performance (grades) in spring and autumn tests and on foal resistance characteristics and temperament features of the dam. Finally, stepwise regression was carried out on the resistance score as a dependent variable and treatment and the four temperament characteristics of the dams as independent variables.

Results

The early handled foals were in general not calmer than the control foals when being rubbed in the autumn and were similarly difficult to catch in the beginning of the test. There was some indication that foals in the early handled foals scored better than the control foals in three tests, i.e. in the response to the lead to the left and right and to pushing on the breast and also in the final grade. However, in no case was the difference significant. Also, in no case was there a significant correlation between the grade the foal got in the 4th treatment in the spring and the grade in the autumn test. 

            The foals differed in difficulty being caught, haltered and lead. The analyses from the video recording revealed that the resistance of control foals was greater than that of early handled foals, measured as total time (p<0.01) and frequencies of strong resistance (p<0.02). There was no difference between the sexes.

            Also, in the experimental group, the strength of foal resistance to handling was correlated with the temperament scores of the dam in 3 of the 4 attributes tested (ease of training: r = - 0.510, p<0.02; ease of handling: r = -0.556, p<0.01;  general level of nervousness: r = 0.483, p<0.03). In the control group, no significant correlations were found in comparable tests.

            In the stepwise regression analyses two variables explained a significant portion of the variation in the dependent variable (strong resistance): the treatment ( F = 10.19, p< 0.01) and ease of handling of the dam (F = 4.14, p<0.05).

Discussion

These results suggest that the 22 foals in the experiment did not learn significantly by being handled, haltered, and lead during the first 2 weeks of their life. Some foals had learned some of the tasks and appeared to retain compliance later, but due to high individual variation the overall differences between early handled and control were not significant. However, early handled foals were calmer and on average resisted less when tested 3 months later compared to control foals. When analysed in detail, it became clear that the difference is due to the effect the whole treatment had on foals whose dam’s were in general easy to handle. This is not surprising and supports research and the common knowledge that it takes more time to work with more stressed horses (Fiske and Potter, 1979). 

            Although we handled the foals in a gentler manner than recommended in Robert M. Miller’s video they were nevertheless forced to lie down and restrained throughout the treatment. On the basis of our results we conclude that such a forced early handling should not to be recommended for Icelandic foals. That is because foals whose dams are difficult to handle do get stressed and they did not learn to lead. Also, the foals that had learned to be more relaxed would probably have learned that through gentler handling. Many trainers do in fact recommend early handling procedures that are more pleasing for the foals and which aim to form a bond between the foal and the trainer built on trust (Tellington-Jones, 1998; Líndal, 2001).

Acknowledgements

We thank Stefán Már Stefánsson who analysed some of the data and wrote the first report of this study with us and kept all records. He, Gudrún Magnúsdóttir, Andrés Ívarsson and Snorri Sigurdsson helped with the experiment in the spring, and the autumn test was done by Valberg Sigurdsson with Stefán, and us assisting. The farmers at Hofstadasel, Keldudalur and Vatnsleysa provided facilities and foals and helped in many ways, as did the staff at Holar. Holar Agricultural College and Iceland University of Education supported the study.

References
Fiske JC, Potter GD (1979)  Discrimination reversal learning in yearling horses.  J Anim Sci 49: 583-588.
Flannery B (1997) Relational discrimination learning in horses.  Appl Anim Behav Sci 54 (4): 267-280.
Heird JC, Whittaker DD, Bell RW, Ramsey CB, Lokey CE  (1986) The effects of handling at different ages on the subsequent learning ability of 2-year horses.  Appl Anim Behav Sci 15: 15-25.
Houpt KA, Parson MS, Hintz HF  (1982) Learning abilities of orphan foals, of normal foals and of their dams.  J Anim Sci 55: 1027-1032.
Líndal B (2001) Early training- a video. See http://www.eidfaxi.is/shop/
Lorenz  K (1937)  The companion in the bird’s world.  Auk 54: 245-273.
Mader DR, Price EO (1980) Discrimination learning in horses: effects of breed, age and social dominance.  J Anim Sci 50: 962-965.
Mal ME, McCall CA  (1996) The influence of handling during different ages on a halter training test in foals. Appl Anim Behav Sci 50: 115-120.
Mal ME, McCall CA, Cummins KA, Newland MC (1994)  Influence of preweaning handling methods on post-weaning learning ability and manageability of foals. Appl Anim Behav Sci 40: 187-195.
McCall CA  (1990) A review of learning behavior in horses and its application in horse training.  J Anim Sci 68: 75-81.
Miller RM  Early learning- video.  VideoVelocity, PO Box K, Virginia City, NV 89440, (800) 284-3362.
Miller RM  (2000) Molding the equine mind.  Equus 271: 60-67.
Tellington-Jones L (1998) Handle and halter foals without trauma. TTeam Up With Your Horse 1(6): 8-12.