| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* School of Land-based Studies, Nottingham Trent University, Brackenhurst College Campus, Southwell, Nottinghamshire, England NG25 0QF and
and
School of Psychology, University of Nottingham, University Park, Nottingham, England NG7 2RD
2 Correspondence:
E-mail:
carol.hall{at}ntu.ac.uk.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Key Words: Discrimination • Height • Horses • Learning • Stimuli • Vision
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
To assess the effect of stimulus position on performance, horses in this study were trained to perform a simple two-choice, black/white discrimination with stimuli either at ground level or nose height. The aim was to test the prediction that stimuli presented at ground level would be easier for the horse to discriminate and result in an improved learning rate, relative to stimuli presented in an identical way but at a higher level.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Eleven experimentally naive horses from the equestrian center at the Brackenhurst College campus of the Nottingham Trent University began the pretest training. Three of these horses did not learn to open the stimulus box in two training sessions and so were excluded from the study. Eight subjects learned to push open the stimulus box to obtain a reward within the first training session. The three mares and five geldings were of varying types, ridden for 2 h, 6 d/wk. Their heights ranged from 152 to 165 cm, with a mean height of 157.87 cm. Ages ranged from 6 to 16 yr, with a mean age of 10 yr. All horses were stabled during the study and turned out on their day off. They were all accustomed to eating forage and concentrate rations at various levels, from the ground to above nose height, and all had been fed carrots at some time.
Test Area and Apparatus
The test area was located in an enclosed barn with a concrete floor. Skylights in the roof provided daylight. The half of the building used for testing was fenced off along the long side using galvanized wire mesh barriers (120 cm in height) and screened from view by sheeting to a height of 300 cm. A gap of 10 cm in this screening allowed the experimenter to view the subject performing the trials while remaining outside the test area. The test area was 5 m wide and 10 m long; a "starting line" of masking tape was placed on the floor 6.5 m from the end wall where the stimuli were displayed. Two identical wooden boxes were placed against the wall, each 125 cm from the side of the test area with a gap between the two boxes of 150 cm. The stimulus box was either located on the floor for ground-level presentations or on a table for high-level presentations, 70 cm above ground level. Each table had a top measuring 120 x 60 cm. A rubber mat was placed under the stimulus box when it was placed on the table to prevent it from moving when the horse tried to open it. See Figure 1
for a plan of the test location.
|
for a plan of the stimulus box.
|
Training
Each horse was introduced to the stimulus box and shaped to push the flap of the box open with its nose (by tapping the box). Once they could open the box to obtain the reward, either black or white was designated as their correct stimulus (four horses in each of the two conditions: black correct or white correct) and pretest training commenced. Training sessions were held twice a week and comprised 10 separate trials. The training sessions took 20 to 40 min according to individual performance. For each separate trial, the horse was released at the starting line, and during the first session, the handler walked by its side toward the stimulus boxes. During the first training session, if a wrong choice was made, the horse was allowed to change its selection and obtain a reward for the correct choice. The horse was then led behind the screens, the stimulus boxes were reloaded with carrots, and the position of the cards was altered. After the first three trials of the second session, the horse was taken back to the starting line following a wrong choice and had to return to make another selection. This procedure was repeated without altering the presentation of the stimuli until the horse made the correct choice, these repeat corrections being counted as one trial. The latter protocol was adopted during the experimental trials. During the training sessions (only), after three repeated errors within one trial, the horse would be guided to the correct box. In the testing phase, there were no such forced corrections.
Throughout the training and experimental testing periods, the left/right position of the positive stimulus was varied randomly, up to a maximum of three consecutive choices on one side to avoid spatial cues from becoming more important than visual cues. The handler led and released the horses from either side in order to control for directional influences. During the first session, the initial height of the stimulus for four of the subjects was on the ground and on the table for the other four to control for any effects of order of presentation (two high, two low in the black designated correct group; two high, two low in the white designated correct group). The height of presentation in the subsequent sessions was alternated, high for one whole session and low for the next session.
Pretest training was complete once the horse could freely approach the stimulus boxes from a distance of 6 m at both the high and low presentation heights, then select one of the boxes to obtain reinforcement. This was accomplished during the first two sessions for all of the horses.
Experimental Testing
Test sessions were carried out twice a week and each session comprised 10 trials. The height at which the stimuli were presented was alternated from one session to the next, and it was the same for all 10 trials within a single session. Two consecutive training sessions within one week constituted a single training set, with the first sessions at high and low height presentations forming the first training set. The position of the stimuli (left/right) was varied as in the training sessions, and equal numbers of left and right presentations of the correct stimulus were included in each experimental session.
At the start of each session, the horse was led into the barn, the doors were closed, and the horse was positioned behind the starting line, directly facing the stimulus boxes. The horse was then released and allowed to approach the boxes to make its selection. A correct choice was rewarded by access to the carrot via the unlocked flap before the subject was caught and led behind the screens. The number of trials during which the horse made a correct selection on the first attempt was calculated as a percentage of the total number of trials and resulted in an accuracy score. An incorrect choice resulted in the horse being caught by the handler before it could try the correct box and being led back to the starting line to try again. If an incorrect choice was made, the same presentation was repeated until the horse made the correct choice. The initial choice would be scored incorrect and repeated errors with the same stimulus presentation were counted up within any one trial. The number of error runs (on first or subsequent attempts) was calculated as a percentage of the total number of runs (whether correct or incorrect) in that session. Thus, the error scores were not simply the obverse of accuracy scores and reflected perseverance in making an incorrect choice.
At the end of each trial, the horse was led behind the screens while the experimenter repositioned the stimuli according to the prearranged, semi-random order. When no change of stimulus position was required, the cards were removed and replaced in the same box to control for possible auditory cues. After a period of 30 s, the horse was led back to the starting line to commence the next trial. Both accuracy and error rates were calculated as percentages for the session.
The overall learning criterion for the discrimination task was reached once 70% accuracy was attained on four consecutive sessions. Because sessions alternated, this criterion included two sessions at high presentation and two at low presentation (i.e., two training sets). The total number of trials required for each horse to reach the individual criteria at each of the high and low positions (two scores of 70% or over, attained consecutively at a single height and independently of the scores at the alternate height) provided an additional measure of the effect of stimulus height on learning.
Data Analysis
To assess the effect of stimulus height on performance of the visual discrimination, the number of trials taken to reach the criterion of 70% correct on two consecutive sessions was calculated for each presentation height separately (accuracy scores of 70% or above for two consecutive sessions at a single presentation height, regardless of interim scores at the other height). Mean accuracy and error rate scores for the two heights of presentation were calculated. Two-way mixed ANOVA were conducted with the within subjects factor of height (of stimulus) and the between subjects factor of stimulus (black or white designated correct). The dependent variables were trials to criterion, accuracy and error rates. The interaction between height and stimulus designated correct was also investigated.
Learning rate was examined using training set scores (the combined means of all eight subjects for accuracy and error rate at the different heights of stimulus presentation for each individual session). Because the horses took different numbers of trials to reach the training criterion (see above), learning rate over the training sessions could only be assessed for the first 10 sessions of training (for which there was complete data). These data were analyzed as five training sets (at both high and low presentations) in a repeated measures design with the factors of height and training set; again, the between-subjects factor was stimulus. Planned comparisons were made by paired samples t-tests (one-tailed) in order to assess the predicted effect of height of stimulus on performance as training progressed.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Trials to Criterion
The number of trials taken to reach the overall learning criterion ranged from 76 to 282, with a mean of 182.88 ± 27.51 trials. Separate scores for each height of stimulus presentation ranged from 25 to 204 trials for the low presentations (mean = 97.63 ± 21.48); 66 to 269 for the high presentations (mean = 162.88 ± 24.53). The number of trials taken to reach the criterion of 70% accuracy for two consecutive sessions at a single presentation height was less when the stimuli were presented in the low position. There was a main effect of stimulus height (P = 0.014). The trials to criterion were not affected by which stimulus was designated correct and there was no significant interaction between height and black/white color of stimulus.
Performance Accuracy
Overall accuracy scores ranged from 61.54 to 77.29% (mean = 67.93 ± 1.96). Mean accuracy scores for the different presentation heights were 74.00 ± 2.85% at the low presentation, 61.86 ± 2.41% at the high presentation. Accuracy was found to be significantly better when the stimulus was presented at ground level. There was a highly significant main effect of height (P = 0.004). No significant difference in accuracy was found in relation to the stimulus designated correct. A marginal interaction (P = 0.051) between the stimulus designated correct and its height was found in the accuracy scores, showing a tendency for the positional effect to be greater when the positive stimulus was black.
To compare learning rates at the two heights of presentation, mean accuracy scores for all of the first 10 sessions (five high, five low) are shown in Figure 3. Statistically, there was again a main effect of height (P = 0.044), but the interaction between training set and height was only marginal (P = 0.066). Thus, although accuracy of performance at the outset was similar regardless of the stimulus height, it was consistently better with the low-level presentations on subsequent training sets, significantly so for training sets three (P = 0.022), four (P = 0.043), and five (P = 0.005).
|
Overall error rates ranged from 21.84 - 34.25% (mean = 28.55 ± 1.57). Mean error rates for the different presentation heights were 23.25 ± 2.76% at the low presentation and 33.85 ± 1.31% at the high presentation. The error rate was significantly lower when the stimulus was presented at ground level. There was a highly significant main effect of height (P = 0.008). No significant difference in error rate was found in relation to the stimulus designated correct and there was no interaction between height and stimulus designated correct.
Although the overall effect of height was very clear, there was no evidence for an effect on learning rate with respect to repeated errors over the first 10 sessions at the two different heights of presentation. Mean error scores at the different presentation heights (five high and five low training sets) are shown in Figure 4. For the error scores in the early stages of training, the effect of height was marginal (P = 0.078), and there was no evidence for any interaction between training set and height. However, the planned comparisons confirmed that performance was again significantly worse (reflected in more repeated errors) at the higher position in training sets three (P = 0.019), four (P = 0.04), and five (P = 0.015).
|
Thus, on all three measures of performance, there was a clear advantage in presenting the stimuli at ground level. There was no significant difference in any measure of performance in relation to the stimulus designated correct.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
The results of the present study are consistent with what is known about the horse’s visual abilities and the factors that should improve visual discrimination learning. The visual field of the horse is constrained by the anatomy and physiology of the visual system as well as by the position of the head and the level at which the eye is carried. The lateral position of the horses’ eyes, the size and curvature of the cornea, size and horizontal shape of the pupil, and angular extent of the retina provide the horse with extensive monocular vision. The binocular portion of the visual field is limited to between 65 (Crispin et al., 1990) and 80°(Harman et al., 1999) in front of the horse. The latter investigation concluded that this binocular overlap was located down the horse’s nose and not directly ahead as was previously thought. Harman et al. (1999) also found that a blind area existed in front of the forehead. In order to get the clearest possible picture of the visual stimuli, the image must be projected onto the area of the retina with the highest ganglion cell density. This area has been found to coincide with the area responsible for binocular vision, the temporal end of the visual streak (Hebel, 1976; Harman et al., 1999; Guo and Sugita, 2000).
Thus, the position of the head and consequently the level at which the eye is carried is important in projecting the visual image onto the most sensitive areas of the retina, particularly while the horse is in motion (Saslow, 1999), as when approaching the stimulus boxes in the present study. If the visual field of the horse is as Harman et al. (1999) concluded, then the position of the horse’s head on approaching the stimulus boxes at different heights will need to vary accordingly. When the horse lowers its head, the binocular field is directed toward the ground and this should allow the ground-level stimuli to remain visible as the horse approaches them. By contrast, if the horse failed to raise its head sufficiently when approaching the high level presentation used here, the independent evidence on the nature of the visual field would suggest that the stimuli should disappear from view, in which case the horse would then make a "blind" choice. The preference shown for the ground level stimuli in this study is consistent with the observation that horses prefer to eat from the floor or from low-level receptacles (Houpt, 1991). In this position, the horse has its binocular field directed toward the ground and has the benefit of being able to scan the lateral horizon for potential threats with its monocular fields (Harman et al., 1999).
Although retinal ganglion cell density has been found to be greatest at the temporal end of the visual streak (Hebel, 1976; Harman et al., 1999; Guo and Sugita, 2000), the total numbers and exact density of these cells has been debated. A recent study into the structure of the equine retina has found large gaps between ganglion cells in most parts of the equine retina (Ehrenhofer et al., 2002). The majority of these ganglion cells were found to be very large and to have input from many amacrine cells, indicating the sensitivity of the visual system to subtle changes in illumination levels and stimulus motion. It is only in the area of the visual streak and a small area close to the optic disc, where there is a well-balanced ratio of photoreceptor, bipolar, and ganglion cells, that the horse possesses any real visual acuity (Ehrenhofer et al., 2002). Even in this area, it is thought that the horse has a limited ability to see detail (Saslow, 2002; Timney and Keil, 1992). Given the limitations of the equine visual system, it is important to present visual stimuli in a position that optimizes their perception by the horse. The present study provides direct evidence that equine visual learning can be enhanced by ground-level presentations and the associated lowering of the head, even in the simple task of discriminating between black and white stimuli. This advantage is likely to be even more important in more complex tasks of visual discrimination.
This variation in visual ability in relation to head position may account for some differences in the results of studies into equine perception. For example, the earliest published study into equine color vision (Grzimek, 1952), involved stimuli presented at ground level. The results of this study do not correspond with those of more recent studies, where the stimuli were presented to the horses at nose height (Pick et al., 1994; Macuda and Timney, 1999). Grzimek (1952) found that horses were able to select a green stimulus from various shades of gray; the two more recent studies concluded that they could not. Similarly, with a presentation height of 1.22 m from the ground, Smith and Goldman (1999) found individual differences in the color discrimination ability of horses. Three horses successfully discriminated green and yellow from gray, one horse performed at chance levels for these colors. A study into color vision in fallow deer (Birgersson et al., 2001) concluded that this ungulate could discriminate greens from grays, with brightness cues controlled for, when the stimuli were presented at ground level. Food selection by both fallow deer and horses involves mainly green stimuli and is carried out at ground level. Given the positional differences in the performance of visual discriminations demonstrated in the current study, the effect of the height of the stimulus on the ability to discriminate specific colors should be investigated further.
During ridden work, the horse must be allowed to alter the position of its head in order to obtain a complete visual picture, particularly while in motion (Saslow, 1999). This study provides controlled experimental evidence to suggest that by lowering the head, the horse can better assess ground conditions to improve footing. The results of this study also highlight the importance of the visual appearance of ground level stimuli to the horse. In designing floor surfaces for use in various locations (e.g., stable flooring, ramps, and flooring for trailers), this factor should be considered. Further work is required to assess the visual features of ground surfaces that will optimize horse performance.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
Received for publication December 12, 2002. Accepted for publication March 14, 2003.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. M. Gearhart, C. C. Gearhart, and S. M. Petersen-Jones A Novel Method for Objective Vision Testing in Canine Models of Inherited Retinal Disease Invest. Ophthalmol. Vis. Sci., August 1, 2008; 49(8): 3568 - 3576. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
