Association of Temperament and Cognitive Ability in Horses

Jennifer R Swallow, Elkanah Grogan, and Sue M McDonnell

Equine Behavior Lab, New Bolton Center
University of Pennsylvania School of Veterinary Medicine

Understanding the association of temperament and learning ability in horses has been a long-term goal of horse owners, breeders, trainers, and behaviorists.  While there are strong beliefs and considerable anecdotal evidence of a link between temperament and learning ability, the systematic study of temperament and cognition in horses are both in their infancy.  Currently a few laboratories have begun investigating methods for objectively measuring temperament in horses.  In general, the approach has been to apply to the horse various objective behavior tests that have been developed for use in other species, such as open field and novel object tests.  The earlier methods of using subjective survey and rating scales to evaluate temperament are still an important component of temperament evaluation and are often judged useful in validating results from the objective tests.  In the field of equine cognition, considerable progress has been made in recent years on the study of operant associative and discriminative learning, as well as concept formation.  A few studies have investigated temperament characteristics in learning situations, and most results are promising but still preliminary.

The goal of the study reported here was to adapt and develop methods of assessing temperament in horses for use in our laboratory for study of the association of temperament with cognitive ability. Each of 10 ponies were administered a battery of four objective temperament assessment tests (Familiar Open Field Isolation, Neutral Novel Object, Handling Challenge, and Visual Auditory Startle) adapted from published work done in other laboratories.  For each of eight temperament characteristics (calmness, cautiousness, compliancy, curiousness, fearfulness, nervousness, reactivity, and social dependency), subjects were ranked from 1 (least) to 10 (most) based on specific frequencies, durations and latencies of behavior across the four objective temperament tests. In the 8 weeks following completion of temperament testing, in a separate study (Grogan and McDonnell, in progress) each subject was evaluated in four series of associative operant conditioning trials representing simple associative learning tasks involving visual, tactile, auditory, and olfactory modalities.  As of this interim report, subjects have completed the visual and tactile series.  Based on those interim results, subjects were ranked 1 (least efficient) to 10 (most efficient) for learning efficiency in the visual and tactile associative operant tasks. Spearman rank order analyses were used to evaluate both a priori and post priori hypothesized associations of temperament and cognitive ability.
 

Interim Findings

Objectively Assessed Temperament Characteristics and Learning Efficiency
            A total of 99 specific hypothesized associations of objectively assessed temperament measures and learning efficiency were evaluated.  As expected, with only 10 subjects evaluated thus far, significant associations were not demonstrable.  Of interest, the frequency of non-calm behaviors in the Visual Auditory Startle test tended to be significantly associated with learning efficiency rank in the tactile task (Spearman rho = 0.71, 8 df, p < 0.05).  Also, the total frequency of snort or blow vocalizations and eliminations (urination or defecations) in the Familiar Open Field Isolation test tended to be significantly associated with learning efficiency in the visual task (Spearman rho= 0.70, 8 df, p < 0.05).  Curiously, these associations were in the direction opposite to our a priori hypothesis. This could represent an unexpected association, or could be an artifact of the small sample size or of the particular haphazard sample available for this preliminary study.

Subjectively Assessed Temperament Characteristics and Learning Efficiency
            For each of the eight temperament characteristics we obtained subjective rankings of the 10 subjects from each of two evaluators familiar with the subjects.  A total of 32 hypothesized associations of these subjective rankings with the objective temperament test results or directly with learning efficiency ranks were evaluated.  Again, as expected with only 10 subjects evaluated thus far, significant associations were not demonstrable.  However, even with only 10 subjects, some tendencies for significant associations were found. Social dependency assessed subjectively tended to be significantly associated with learning efficiency rank in the tactile learning task (Spearman rho=0.83, 8 df, p<0.01). Reactivity assessed subjectively also tended to be significantly associated with learning efficiency rank in the tactile task (Spearman rho=0.70, 8 df, p<0.01). These two associations were also in the direction opposite to that hypothesized.  Reactivity assessed subjectively tended to be associated with the frequency of non-calm behaviors in the Visual Auditory Startle test (Spearman rho=0.77, 8df, p < 0.01). Curiousness assessed subjectively tended to be associated with the frequency of gazing at the object in the Neutral Novel Object test (Spearman rho=0.71, 8 df, p < 0.05).

Objectively Assessed and Subjectively Assessed Temperament Characteristics
            Though not significant with only 10 subjects evaluated thus far, associations of objective temperament assessments with subjective temperament assessments appeared promising.  A composite ranking based on sum of the ranks of the objective and subjective temperament assessments tended to be significantly associated with tactile learning efficiency rank for the temperament characteristics of fearfulness, reactivity, and social dependency (Spearman rho=0.75, 8 df, p<0.05; rho=0.71, 8 df, p <0.05; rho=0.76, 8 df, p<0.05 respectively).  The composite ranking for social dependency also tended to be significant with the combined tactile and visual learning efficiency rank (rho=0.80, 8 df, p<0.01).  All of these associations are in the direction opposite to that hypothesized.

This study represents an important preliminary step in our laboratory’s investigation of temperament in horses and its association with cognitive ability.  Our design would likely yield significant results with a greater number of subjects more representative of the variation in temperament and learning ability in the general horse population.  Our initial estimates are that 30 to 50 subjects would greatly improve the power of the design. These results suggest that eventual accurate prediction of cognitive ability or performance based on assessed temperament will require multivariate techniques requiring large sample sizes.

This is Dorothy Russell Havemeyer Foundation Project.

Jennifer Swallow is a graduate of Penn State and member of Penn’s Veterinary Class of 2006.
She is a 2003 Dorothy Russell Havemeyer Foundation Veterinary Student Summer Research Fellow.