Evaluation of Short-Term Stall Size Housing on the Behavior and Welfare of Domestic Horses (Equus caballus)

Abstract

Single-stall housing is a common system used for horses in various settings, including racing stables, riding schools, teaching and research facilities. Adequate stall space is necessary for horses to lie down and make postural adjustments, both of which are essential for equine rest and welfare. The Guide for the Care and Use of Laboratory Animals recommends a stall size of 144 ft². However, not all stalls meet this size requirement, and there is limited literature supporting its use in terms of animal welfare. This study used a crossover design to compare behavioral and physiologic variables indicative of welfare of horses housed in the standard recommended 144-ft² stall compared with a smaller stall (136 ft²) to determine whether the smaller stalls affected horses’ welfare and behavior. Eight horses (1,380-1,640 lb) were paired with preferred associates resulting in 4 pairs. One horse from each pair was randomly assigned to be housed in either the standard stall (144 ft²) or the smaller stall (136 ft²) for 4 days (from Monday morning through Friday morning), while the paired horse was housed in the opposite-sized stall. After this initial stall period, all horses were turned out to pasture for 9 days before being housed in the alternate-sized stall for an additional 4 days. Hay consumption, fecal output, and serum cortisol were quantified daily. Movement and duration of recumbency were measured with an accelerometer, and continuous video recordings of all horses were obtained while housed in the stalls (total of ∼96 hours per horse per stall condition). Behavioral variables (resting, locomotion, standing, consummatory, elimination, and stereotypic behavior) were scored every 5 minutes from the recordings and compared between stall conditions. Although several measures changed across days in both stall conditions, stall size did not affect the horses’ hay consumption, fecal output, or cortisol concentrations, nor the duration of sternal or lateral recumbency as determined by accelerometer data. However, the duration of walking recorded by the accelerometer different between the 2 stall sizes with horses spending more time walking in the smaller stall (136 ft²). Analysis of video recordings also found no difference in percentage of time spent in sternal or lateral recumbency, rolling, standing, stand resting, walking, eating, or drinking. In conclusion, time budgets, food intake, fecal output, and cortisol concentrations were similar in horses housed in the standard-sized stall (144 ft²) compared with the smaller stall (136 ft²), suggesting that the reduced stall size did not result in observable changes in equine behavior or physiologic welfare indicators compared with the standard-sized stall.

Introduction

Horses are singly housed in stalls for a variety of reasons, including shows, competitions, breeding, hospitalization, and leisure.¹ Stall size is a critical factor in equine housing and management, with potential implications for behavior and welfare. The Guide for the Care and Use of Laboratory Animals² recommends that all animals should be housed under conditions that provide sufficient space as well as supplementary structures and resources required to meet physical, physiologic, and behavioral needs. Environments that do not adequately address those needs can negatively affect brain development, disrupt normal physiologic function, and lead to maladaptive behaviors.² A 12- × 12-ft (144-ft²) stall is widely regarded as the industry standard in the United States, as recommended by the Guide for the Care and Use of Laboratory Animals² and the Guide for the Care and Use of Agricultural Animals in Research and Teaching.³ However, these standards lack strong scientific background, raising the question of whether alternative stall sizes could provide comparable or even improved welfare conditions for horses.

Historically, horses have been housed in a variety of stall sizes, ranging from 10 × 10 ft (100 ft²) to larger enclosures such as 14 × 14 ft (196 ft²) or 16 × 16 ft (256 ft²).^4–7 Although the 12- × 12-ft stall has become the widely accepted standard in research and industry settings, our review of the literature found no peer-reviewed evidence supporting the origin of this recommendation within the Guide for the Care and Use of Laboratory Animals or other regulatory guidelines. Most published stall-size recommendations appear to derive from university extension publications and long-standing industry practice rather than empirical research.

Animal welfare is a complex, multifactorial construct that is challenging to define and assess. Typically, analysis of both physiologic and behavioral variables is used to strengthen the validity of the assessment.^1,8 One means of assessing welfare is to quantify the animal’s time budget, that is, the amount of time it engages in behavioral activities.^8–11 The current study used video recordings to compare time budgets of horses when housed in standard 144-ft² stalls to time budgets when housed in 136-ft² stalls. In addition, accelerometer data were used to quantify duration of active and resting behavior, particularly of recumbency.

Sleep is a universal maintenance behavior that is essential for health, welfare, and productivity in animals.^12,13 Although horses can drowse and sleep while standing due to the stay apparatus of the forelimbs and hindlimbs,¹² they must lie down to achieve rapid eye movement sleep and complete a full sleep cycle.^14,15 Rapid eye movement sleep deprivation, often caused by pain, environmental insecurity, or inadequate space, can result in increased daytime drowsiness, poor performance, collapse, and injury.^16–18 There are studies investigating the relationship between management factors or stall design and recumbency in horses.^19–25 Several studies have shown that horses housed in larger stalls or with greater lying‐area allowance display increased recumbency and rolling behaviors, which are considered important indicators of rest and welfare.^26–28 Conversely, insufficient stall space has been flagged as a risk factor in welfare assessments, associated with restricted movement, reduced rest postures, sleep deprivation, and elevated behavioral indicators of frustration or stress.^4–29 Nevertheless, no studies have determined the optimal size for single-housing horses in stalls.

As stress can negatively affect the health and well-being of animals,^30–32 its assessment and regulation are indispensable.³³ In response to stress, the HPA axis is activated, initiating the release of cortisol,³⁴ making blood cortisol concentration a useful biomarker of the acute stress response.³³ Stress in horses also activates the sympathetic–adrenal–medullary axis, leading to increased sympathetic tone and, therefore, decreased gastrointestinal (GI) motility.^35–37 In addition, transitioning from pasture to stall housing can decrease water intake and reduce fecal output, potentially impacting GI health and increasing the risk of colic impactions.³⁸ Hay intake and fecal output were therefore measured in this study as indicators of GI function.

Our study goal was to address a key question: Could stalls smaller than the current recommended size (144 ft²) still meet performance and welfare standards for horses? Specifically, we investigated whether a small reduction in stall size affected physiologic and behavioral indicators of welfare, namely feeding behavior, fecal output, blood cortisol concentrations, movement, and resting behavior. We hypothesized that stalls smaller than the current guidelines (136 ft² compared with 144 ft²) would not significantly impact equine behavioral time budgets or physiologic indicators of welfare.

Materials and Methods

Humane care and use of animals.

This study was performed in a facility accredited by AAALAC, International and was approved by the IACUC at Cornell University.

Study site and animals.

Eight horses from a teaching and research herd at Cornell University were included in the study. The group included 6 mares and 2 geldings ranging in age from 8 to 22 years (Table 1). The average weight of the horses was 1,456 lb (661 kg), ranging from 1,380 to 1,640 lb, and their heights ranged from 15.5 to 17.2 hands (62 to 68.8 in.). The breeds represented in the study included Oldenburg (n = 3), Hanoverian (n = 1), Westphalian (n = 1), Dutch Warmblood (n = 1), Irish Sport Horse (n = 1), and a Clydesdale/Quarter Horse cross (n = 1). Although some variation in age and sex was present, the small sample size and paired crossover design limited the ability to statistically model these factors. All horses were healthy and maintained under identical management, housing, and feeding practices throughout the study period.

Table 1.

Demographic Characteristics and Group Assignment of the 8 Horses Included in the Study

Horse	Pair	Group	Sex	Age (y)	Breed	Weight (lb)	Height (hands)	Height (in.)
1	4	A	Mare	22	Oldenburg	1,485	15.5	62
2	4	A	Mare	11	Dutch Warmblood	1,446	16.8	67.2
3	3	A	Mare	14	Hanoverian	1,380	16.0	64
4	3	A	Mare	16	Oldenburg	1,470	16.2	64.8
5	2	B	Mare	14	Hanoverian	1,380	16.0	64
6	2	B	Mare	16	Oldenburg	1,450	15.5	62
7	1	B	Gelding	8	Clydesdale/QH X	1,640	17.2	68.8
8	1	B	Gelding	14	Irish Sport horse	1,480	17.0	68

Experimental design.

A crossover study design was used to evaluate the impact of stall size on horse behavior and physiologic welfare parameters (Figure 1). Horses were assigned to 1 of 2 stall sizes. The smaller stalls measured 135.9 ft², with internal dimensions of 141.3 in. by 138.5 in. The larger stalls measured 144.6 ft², with internal dimensions of 150.0 in. by 138.8 in. For the purposes of this paper, these are referred to as 136- and 144- ft² stalls, respectively. The 144-ft² stalls were modified from existing larger foaling stalls by adding temporary panels to achieve the target stall size (Figure 2). Construction materials and fixtures matched those used in the rest of the barn to minimize any environmental variation. To eliminate potential bias from novel scents associated with the newly modified stall walls, horses not enrolled in the study were housed in the reconstructed stalls for 1 week prior to the start of the study.

Figure 1.

Crossover Study Design for Assessing Physiologic and Behavioral Responses to Stall Size. Each horse spent 4 d in a 144-ft² stall (blue box), followed by a 9-d washout period in pasture, and then 4 d in a 136-ft² stall (red box). ‘N’ and ‘S’ denote the north and south orientations of the barn. Figure created with BioRender.com.

Figure 2.

Modified Stall With Temporary Panel Extensions Used to Achieve the Target Stall Size of 144 ft². Temporary wooden panels were installed to standardize stall dimensions during the study.

The study was conducted between November and December. Temperature and humidity were recorded daily throughout the study period; however, they were not included in the final analysis, as no substantial fluctuations were observed. All horses in this study were housed inside an enclosed barn. The barn was equipped with multiple ventilation features, including large ceiling fans, a continuous ventilation strip along the upper perimeter of the walls, and a ridge vent extending along the roofline. Each stall had a window, and the barn featured large garage-style doors that could be opened to enhance airflow, but no mechanical HVAC (heating, ventilation, and air conditioning) system was used during the study.

Horses were moved from pasture to their assigned stalls on Monday morning and remained continuously housed until Friday morning, at which point they were returned to pasture. During this period, horses did not receive any turnout or in-hand exercise. Weekday-only design was chosen to accommodate husbandry staff schedules and ensure consistent measurement of hay intake and fecal output. In addition, this design avoided housing horses through the weekend without turnout, which would have exceeded the standard of care at our institution.

To minimize stress and maintain social stability, horses were housed in adjacent stalls and positioned to always maintain auditory contact with their paired partner and visual contact with horses across the aisle with other horses. This was done to minimize potential stress related to social separation. A 9-day washout period was implemented between stall conditions, during which all horses were returned to pasture before reassignment. Horses were paired with preferred associates (as determined by the manager of the equine facility), resulting in 4 pairs (Table 1). A random number generator was used to order the pairs (1-4). Pairs 1 and 2 were assigned to group A (first group to be housed in stalls during weeks 1 and 2 of data collection), and pairs 3 and 4 were assigned to group B (initially on pasture; housed in stalls during weeks 3 and 4 of data collection). Within each pair, a stall order (large then small compared with small then large) was randomly assigned to one horse and the opposite order was assigned to the other horse. The north compared with south side of the barn was also randomly assigned for the first pair in group A, with the opposite side assigned to the second pair. The same randomization process was applied to group B (Figure 1). All random assignments were done using the RAND function in Excel.

Physiologic data.

To assess the stress response of horses in each stall condition, GI motility was evaluated by measuring daily hay consumption and fecal output.

Hay consumption.

Horses were fed orchard-timothy hay on the stall floor twice daily. Mares received 3 flakes (average 8.8 lb) at both the morning and afternoon feedings, and geldings received 4 flakes (average 13.2 lb) at both the morning and afternoon feedings. In addition, all horses were provided with 1 cup (190-200 g) of Inspire Min-A-Vite Lite mineral and vitamin grain supplement (Kent Nutrition Group, Muscatine, IA) twice daily. However, as all horses consumed their grain immediately, only hay consumption was used as a dietary intake measure. Hay was weighed before feeding and again 24 hours later to calculate intake.

Fecal output.

Feces were quantified once daily as an indicator of GI function. Manure was collected during stall cleaning each morning and weighed using a calibrated large livestock scale. Broad forks were used to sift out bedding material when collecting feces to improve the accuracy of the fecal weight measurements. Fecal consistency was qualitatively observed during daily collection but was not formally scored, as the study prioritized objective, quantitative measures of GI function, specifically fecal output by weight, to ensure consistency and reduce subjectivity in data collection.

Cortisol.

To assess physiologic stress, daily serum cortisol concentrations were measured in all horses. On the first day of data collection for each group, blood was collected from the horses on pasture immediately prior to being moved into the stalls and then daily during stall housing. Collections were performed each morning between 7:30 and 8:30 am to control for circadian variation in cortisol concentrations. Horses were sampled using a rotating schedule to ensure balance and minimize time-of-collection bias in cortisol measurements. Specifically, the first horse sampled on one day was moved to the end of the sampling order the following day. This approach reduced potential confounding effects related to the time of day each horse was sampled and minimized the likelihood that horses would associate the person’s presence with blood collection, which could induce anticipatory stress. Although not a formal Latin square design, this rotation helped balance order effects across days. Each sample was obtained via jugular venipuncture using a sterile vacuum phlebotomy tube needle (Exel multi-sample, 20 gauge × 1.5 in.; Patterson Veterinary, Loveland, CO). Approximately 3 mL of blood was collected into silicone-coated red-top tubes (Monoject; Patterson Veterinary, Loveland, CO). To minimize potential vein trauma, venipuncture sites alternated between the right and left jugular veins. The time taken to obtain each collection was recorded to account for potential cortisol elevation associated with venipuncture stress. After collection, blood samples were kept at room temperature for 30 minutes and then centrifuged at 2,000g for 10 minutes. From each sample, 1 mL of serum was transferred into microcentrifuge tubes (JorVet; Patterson Veterinary, Loveland, CO) and labeled with the horse’s name, date, and time of collection. Samples were then stored at −20 °C until analysis. At the conclusion of the study, all serum samples were submitted in a single batch to the Endocrinology Laboratory at the Animal Health Diagnostic Center, Cornell University (Ithaca, NY) for cortisol analysis.

Accelerometer data.

Data loggers (HOBO Pendant G data loggers [UA-004-64]; Onset Computer, Bourne, MA) were used to record equine posture and three-dimensional movement in this study. These triaxial accelerometers record gravitational force on the x-, y-, and z-axes, and can be configured to sample data at user-defined intervals. Each accelerometer was placed on the third metacarpal bone using synthetic cast padding (3M synthetic cast padding; 3M, St. Paul, MN) to reduce the risk of pressure sores and was secured with a cohesive bandage (Pivetal Not One More Vet cohesive wrap; Patterson Veterinary, Loveland, CO) (Figure 3). After 2 days, loggers were removed for data download (HOBOware software; Onset Computer, Bourne, MA) and then replaced on the contralateral third metacarpal bone. The downloaded data were then transferred to an Excel datasheet. The loggers were configured to record gravitational force values at 10-second intervals. Placement orientation was standardized so that (1) the x-axis was positioned horizontal to the ground, capturing forward and backward motion; (2) the y-axis was vertical to the ground, capturing side-to-side motion; and (3) the z-axis recorded upward and downward motion. This orientation was based on protocols described by Williams and Hatto³⁹ and DuBois et al.⁴⁰ To estimate steps automatically, the absolute values of acceleration from the x-, y-, and z-axes were summed based on protocols described by Allen⁴¹ and Ringgenberg et al.⁴² Peaks in the summed acceleration were interpreted as steps. Because movement patterns varied between horses and across gaits, a threshold was set individually for each horse and gait based on percentiles. Specifically, thresholds were determined using the 98.7th percentile for walking.⁴¹ This means only the highest 1.3% to 1.6% of summed acceleration values—those most likely to represent true steps—were counted.⁴¹ The corresponding threshold value is the summed acceleration at that percentile, and values below this were excluded. Steps identified from the accelerometer data were then validated against manually counted steps from the video recordings.

Figure 3.

Placement of the Accelerometer Midway Along the Lateral Side of the Metacarpus.

Sternal recumbency was identified by a sustained reduction in x-axis acceleration, a downward shift in x-axis tilt, and a stable midrange z-axis tilt, indicating that the horse was lying on its sternum with minimal forward–backward movement. Lateral recumbency was identified by a marked shift in z-axis tilt toward a sustained maximum or minimum value, accompanied by minimal variation in x- and y-axis acceleration. This pattern indicated that the horse was lying fully on its side, with the accelerometer registering a consistent change in orientation relative to gravity.

To validate accelerometer classifications, video footage for each horse was reviewed and verified with the corresponding accelerometer timeline. Behaviors observed on video (walking, sternal recumbency, and lateral recumbency) were compared with the accelerometer-derived activity states for the same timestamps. A board-certified veterinary behaviorist independently reviewed the same video segments and scored behaviors. Accelerometer patterns appeared to correspond closely with observations obtained from video-based assessments confirming the accuracy of the accelerometer data for posture and locomotor detection.

Behavioral observations.

Continuous behavioral recordings were made for an average of 88.6 hours (91.9 hours in the 144-ft² stalls and 85.3 hours in the 136-ft² stalls) from approximately 8:15 am Monday through approximately 8:15 am Friday for each horse for both stall conditions using a remote camera system installed in each stall (Figure 4) (Axis M1075-L box cameras; Axis Communications, Lund, Sweden). An ethogram adapted from Houpt et al. (1988) and McDonnell (2003) was used to record the behavior of each horse. To minimize observer variation with behavioral data, all videos were scored by one person (P.J.P.).

Figure 4.

Video Monitoring Setup Used to Record Continuous Behavior Within Each Stall.

Behavioral states were recorded as a yes/no occurrence using instantaneous sampling every 5 minutes. Behavioral ‘state’ variables were mutually exclusive and included lateral recumbency, sternal recumbency, stand resting (defined as a relaxed posture characterized by a lowered head and neck, partially closed eyes, and flexion of one hindlimb), standing (defined as an upright, alert posture with even weight-bearing on all 4 limbs), walking, and eating. All occurrences of behavioral ‘events’ (drinking, urination, defecation, roll, and stereotypic behaviors) were recorded to the nearest 5-minute mark. The percentage of observation time spent performing each behavior was then calculated for each horse.

Although stereotypic behaviors were recorded as part of the ethogram during the study, horses were not formally evaluated for stereotypies prior to study initiation. However, no stereotypic behaviors were reported by husbandry staff or noted during routine observation in the weeks leading up to the study.

Due to technical difficulties, video recordings during the final week of data collection were missing from 4:00 pm until 8:00 am for 2 days for 2 horses, from 1:00 am until 8:00 am for 2 days for 1 horse, and from 10:00 pm until 8:00 am for 2 days for 1 horse. In addition, one horse manipulated the camera in and out of viewing range during the last 2 days of data collection. In total, 735.3 hours of video footage were analyzed for horses in 144-ft² stalls and 682.8 hours of video recording were analyzed for horses in 136-ft² stalls. All usable observations were included in the results.

Statistical analysis.

All physiologic and behavioral data were analyzed using linear mixed effects models with fixed effects stall size, day (24-hour period from 8:00 am to 8:00 am the following day), the interaction between stall size and day, and fixed effects of sequence and period. A random effect of horse nested in sequence was included in the models, and the repeated measures taken on the horses over days within a period were assumed to have an autoregressive AR(1) (first-order autoregressive) covariance structure. The fixed effects were tested using F tests, and when the interaction between stall type and day was statistically significant, post hoc comparisons between stall types were performed for each day with a Bonferroni correction to control the type I error rate. Model assumptions of normality and homogeneous variance were assessed visually using residual plots. The variables sternal recumbency and lateral recumbency were log transformed. The behavior data were summarized as the proportion of each day spent engaged in the behavior, and a logit transformation was applied to the proportions. Model assumptions of normality and homogeneous variance were assessed visually using residual plots. The fixed effects were tested using F tests, and when the interaction between stall type and day was statistically significant, post hoc comparisons between stall types were performed for each day with a Bonferroni correction to control the type I error rate. Estimated marginal means and SEs were obtained from the models for use in figures. Estimates for models fit to log- and logit-transformed variables were back-transformed to the original scale, and SEs were calculated. Only one horse displayed a stereotypic behavior (cribbing) once while housed in the small stall, and therefore this behavior category was eliminated from the analyses. Data are presented as least squares means (LSMs) ± SE derived from the mixed model with a significance threshold set at P < 0.05. Statistical analysis was performed using the SAS statistical software package (SAS/STAT 15.3 User’s Guide [2023]; SAS Institute, Cary, NC).

Results

Hay consumption.

Across all days, horses consumed a similar amount of hay in the 144-ft² stalls (27.6 ± 1.5 lb) and the 136-ft² stalls (28.5 ± 1.5 lb). No significant stall size × day interaction was detected (P = 0.69). Although hay consumption varied significantly by day (P = 0.015), it did not significantly differ between stall sizes (P = 0.27), indicating that the pattern of increasing hay consumption over time was similar for both stall sizes (Figure 5A).

Figure 5.

Physiologic Data of Horses (Least Squares Mean ± SE) Across the 4-d Stall Period for the 2 Stall Sizes: 144 ft² (Blue) and 136 ft² (Red). Panels show hay consumption (A), fecal output (B), blood cortisol (C), accelerometer-derived walking activity (D), sternal recumbency (E), and lateral recumbency (F). A significant effect of day was detected for hay consumption, fecal output, and sternal recumbency. Walking activity showed significant effects of day, stall size, and the stall size × day interaction. *, P < 0.05; **, P < 0.001.

Fecal output.

Across all days, horses produced similar amounts of feces in 144-ft² stalls (48.3 ± 3.8 lb) and 136-ft² stalls (52.4 ± 3.9 lb). No significant stall size × day interaction was detected (P = 0.44). Although fecal output varied significantly by day (P < 0.001), it did not significantly differ between stall sizes (P = 0.20), indicating that the decreasing pattern over time was similar for both stall sizes (Figure 5B).

Blood cortisol.

Serum cortisol concentrations remained within normal physiologic ranges (2-6 µg/dL) throughout the study for all horses in both stall conditions (3.7 ± 0.8 µg/dL in the 144-ft² stalls, and 3.9 ± 0.9 µg/dL in the 136-ft² stalls). No significant stall size × day interaction was observed (P = 0.82). Blood cortisol concentrations did not differ by stall size (P = 0.30), and no effect of day was detected (P = 0.69) (Figure 5C).

Accelerometer data.

Walking.

Horses spent an average of 114.6 ± 65.3 minutes walking over a 4-day period in the 144-ft² stalls and 108.1 ± 43.5 minutes in the 136-ft² stalls. A significant stall size × day interaction was observed (P = 0.027). A significant effect of stall size was detected (P < 0.001), and a significant effect of day was also detected (P < 0.001). Overall, horses spent significantly more time walking daily in the 136-ft² stalls (LSM ± SE: 34.2 ± 1.5 minutes) compared with the 144-ft² stalls (21.7 ± 1.5 minutes). Walking time also varied across days, with the highest walking activity observed on day 2 (42.7 ± 2.6 minutes), followed by a decline on days 3 and 4 (27.0 ± 2.6 and 23.3 ± 2.6 minutes, respectively) (Figure 5D).

Sternal recumbency.

Sternal recumbency values (log-transformed LSM ± SE) showed no significant stall size × day interaction (P = 0.52). Although sternal recumbency varied significantly by day (P = 0.003), it did not significantly differ between stall sizes (P = 0.51), indicating that sternal recumbency was similar for both stall sizes (Figure 5E).

Lateral recumbency.

Lateral recumbency values (log-transformed LSM ± SE) showed no significant effect of stall size (P = 0.48). No stall size × day interaction was observed (P = 0.74). No effect of day was detected (P = 0.12). Lateral recumbency values were therefore similar between the 144-ft² (1.0 ± 0.7) and 136-ft² stalls (1.2 ± 0.7) (Figure 5F).

Behavioral data.

Based on behavioral observations, there were no significant stall size × day interactions for sternal and lateral recumbency (P = 0.63 and P = 0.80, respectively). Likewise, the proportion of time spent in sternal or lateral recumbency was not affected by stall size (P = 0.80 and P = 0.09, respectively) nor was there a significant effect of day for either behavior (P = 0.43 and P = 0.87, respectively). No significant stall size × day interaction was detected for standing or stand resting (P = 0.95 and P = 0.81, respectively). In addition, standing and stand resting were not affected by stall size (P = 0.90 and P = 0.12, respectively) nor was there any significant effect of day for either behavior (P = 0.75 and P = 0.70, respectively). There was no significant stall size × day interaction found for rolling or walking (P = 0.67 and P = 0.57, respectively). Similarly, rolling or walking was not affected by stall size (P = 0.49 and P = 0.44, respectively) and there were no significant effects of day detected for either behavior (P = 0.08 and P = 0.44, respectively). No significant stall size × day interaction was detected for either drinking or eating (P = 0.54 and P = 0.81, respectively). Moreover, drinking and eating behavior were not affected by stall size (P = 0.74 and P = 0.17, respectively). A significant effect of day was observed for drinking behavior (P = 0.001). However, no effect of day was detected for the eating behavior (P = 0.71). There was no significant stall size × day interaction detected for either defecation or urination (P = 0.96 and P = 0.27, respectively). Defecation frequency was statically significant between stall sizes, with a higher frequency observed in the smaller stall (P = 0.02), but no effect of day was found (P = 0.14). Urination frequency did not significantly differ by stall size (P = 0.88) but there was a significant effect of day (P = 0.01) (Table 2).

Table 2.

Effects of Stall Size and Day on Behavioral Parameters (LSM ± SE of % of Observations) From Video-Based Observations of Horses Housed in 144-ft² and 136-ft² Stalls

	144-ft2 stalls		136-ft2 stalls		P value
Behavior data	LSM (% of observations)	SE	LSM (% of observations)	SE	Stall size × day interaction	Stall size effect	Day effect
	1.86	1.19	2.03	1.41	0.63	0.80	0.43
Lateral recumbency	0.32	0.18	0.29	0.12	0.80	0.09	0.87
Standing	18.60	2.00	18.8	2.10	0.95	0.90	0.75
Stand resting	33.8	4.60	32.70	4.40	0.81	0.12	0.70
Rolling	0.15	0.08	0.25	0.12	0.67	0.49	0.08
Walking	1.02	0.32	1.31	0.41	0.57	0.44	0.44
Drinking	8.27	1.62	8.79	1.71	0.54	0.74	0.001
Eating	38.80	2.90	39.20	2.90	0.81	0.17	0.71
Defecation	4.27	0.46	5.11	0.54	0.96	0.02	0.14
Urination	1.60	0.39	1.57	0.38	0.27	0.88	0.01

Discussion

Our study investigated whether horses used for teaching and research differed in behavioral and physiologic welfare indicators when housed in 136-ft² stalls compared with being housed in the currently recommended 144-ft² stalls. We selected the largest horses in the herd to ensure that any potential effects of reduced stall size on behavior or physiologic measures would be detectable and not confounded by insufficient space for larger-bodied horses. The results demonstrated no statistically significant differences between the 2 stall sizes in hay consumption, fecal output, serum cortisol concentrations, locomotor activity, or time spent in sternal or lateral recumbency, rolling, standing, stand resting, walking, eating, or drinking. These findings suggest that a modest reduction in stall size (as compared with the recommended 144 ft²) may not adversely affect short-term measures of welfare in horses.

Horses are valuable teaching animals at Cornell University and at many other universities in both agricultural and veterinary settings. In addition to their educational role, horses serve as important models for studying joint diseases and osteoarthritis, contributing significantly to translational orthopedic research.^43–46 Furthermore, equine recurrent uveitis is a well-recognized model for human autoimmune uveitis, offering critical insights into disease mechanisms and treatment approaches.^47–49 Beyond teaching and research, the U.S. horse industry plays a significant economic role, with an annual direct contribution of $177 billion, influencing sectors such as agriculture, tourism, veterinary services, and technology.⁵⁰ Given the important role of horses in teaching, research, and the broader economy, evaluating the impact of housing conditions on equine welfare is essential.

Stress in horses can lead to increased sympathetic tone and, therefore, decreased GI function and increased risk of colic.³⁶ Our study measured hay intake and fecal output as indicators of GI function. Stall size did not influence hay consumption, as horses consumed similar amounts of forage in the 144-ft² and 136-ft² stalls. This finding suggests that short-term confinement in slightly smaller stalls does not substantially alter feeding motivation or access to hay. Instead, day-to-day variation had a stronger effect on intake, with horses gradually increasing consumption over the 4-day period regardless of stall size. This progressive rise in hay intake may reflect acclimation to the stall environment, increased comfort with the study routine, or normal biologic variation in appetite across days. Because the stall size × day interaction was not significant, the pattern of increased intake appears consistent across both housing conditions, indicating that horses adjusted similarly to each stall size over time. Collectively, these results suggest that hay consumption was stable across the 2 stall dimensions evaluated, and short-term changes in stall size are unlikely to meaningfully impact feeding behavior.

Fecal output did not differ between the 2 stall sizes, indicating that short-term housing in a slightly smaller stall did not measurably alter GI function. Instead, day to day variation had a stronger influence, with fecal mass declining steadily from day 1 through day 4 regardless of stall assignment. Similar reductions during the first days of confinement have been reported previously; Williams et al.³⁸ observed a significant decrease in fecal output during the transition from pasture to stabled management, particularly within the first 5 days. Evidence of reduced large-intestinal motility in stabled horses further supports this interpretation. Williams et al.⁵¹ demonstrated that large colon motility, assessed via ultrasound, was significantly lower in stabled horses compared with pastured horses. This reduction in motility, combined with behavioral and hydration adjustments that accompany stall confinement, likely contributes to the progressive decline in fecal mass observed in our study. Taken together, these findings indicate that the decrease in fecal output is driven by horses’ acclimation to stall housing rather than by the modest difference in stall dimensions.

Cortisol plays an important role in assessing the welfare and health of horses, as it provides a snapshot of the horse’s physiologic state.^33,34 An elevation in blood cortisol typically reflects activation of the HPA axis.⁵² Numerous studies have demonstrated a positive correlation between cortisol concentrations and other physiologic stress indicators in horses, including increased heart rate, respiratory rate, and rectal temperature.^53–58 In our study, serum cortisol concentrations remained within the normal reference range (2-6 µg/dL) for all horses across all stall conditions. Furthermore, there were no significant differences between cortisol concentrations measured on pasture, in the 144-ft² stalls, or in the 136-ft² stalls. These findings support the interpretation that horses did not experience a measurable increase in physiologic stress when housed in stalls slightly smaller than the standard size stalls.

Evaluation of behavioral time budgets, that is, the amount of time an animal engages in behavioral activities, is often used as a measure of animal welfare because a change in behavior can be an indication of compromised welfare.^9,10,59–61 Moreover, the use of accelerometers is a validated method of recording sternal and lateral recumbency in horses.^39,40,62 Our study compared both accelerometer data and time budgets of horses confined in 2 stall sizes, 136 ft² and 144 ft². According to the behavioral time budgets obtained with video recordings, none of the behaviors analyzed was significantly different between the 2 stall sizes.

Accelerometer-derived behaviors provided insight into how horses used the 2 stall sizes during short-term housing. Although movement was spatially restricted, horses were still able to walk and change position normally within the stall. These small-scale locomotor behaviors can serve as sensitive indicators of restlessness, frustration, or environmental responsiveness during confinement. While walking time showed a significant stall size × day interaction, the biologic relevance of this effect appears limited. Horses walked only a small proportion of the total observation period in both stalls, and the largest day-to-day variation occurred independently of stall size. The increase in walking on day 2 followed by a decline on days 3 and 4 reflects a typical acclimation pattern to stall confinement rather than an effect of stall size. The fact that this pattern occurred in both stall sizes indicates that it is driven by confinement itself rather than the modest difference in available space. Visser et al.⁶³ reported that horses show temporary increases in walking/movement under new conditions, followed by stabilization, supporting the interpretation that these changes represent an adjustment process rather than a sustained response to spatial constraints. In contrast to walking behavior, both sternal and lateral recumbency were remarkably consistent across stall sizes. Sternal recumbency fluctuated modestly across days, but without a stall × day interaction, indicating that these changes were unrelated to stall dimensions. Lateral recumbency, a sensitive indicator of comfort, deep rest, and sleep quality in horses, did not differ across days or stall types. The absence of differences in either form of recumbency suggests that horses were able to achieve typical resting postures in both stall sizes, and that neither space was restrictive enough to reduce the occurrence of lying down. Interestingly, the accelerometer data indicated that horses spent more time classified as ‘walking’ in the 136-ft² stalls than in the 144-ft² stalls, whereas video-based behavioral scoring did not detect a significant difference in walking between stall sizes. This discrepancy likely reflects differences in how each method captures movement. Accelerometers record continuous changes in acceleration and may categorize small shifts in posture, brief steps, or circling within a confined area as walking, even when such movements are subtle enough to be scored as standing by an observer. Although based on continuous video monitoring, behavioral observations were made by observations taken every 5 minutes, which can miss behaviors of short duration. In addition, walking represented only a small proportion of activity within the 24-hour period in both stall sizes, so relatively modest changes in micromovements can reach statistical significance in the accelerometer model without manifesting as clear differences in overt locomotion on video. The behavioral data were also influenced by partial loss of video footage during the final week of data collection, which reduced the amount of observable walking behavior for several horses and may have limited the sensitivity of the video analysis compared with accelerometry. In contrast to walking behavior, both sternal and lateral recumbency were remarkably consistent across stall sizes. Sternal recumbency fluctuated modestly across days but not between stall sizes. Lateral recumbency, a sensitive indicator of comfort, deep rest, and sleep quality in horses, did not differ across days or between stall types. The absence of differences in either form of recumbency suggests that horses were able to achieve typical resting postures in both stall sizes, and that neither space was restrictive enough to reduce the occurrence of recumbent behavior. The results of walking behavior, together with the absence of differences in sternal or lateral recumbency and the lack of a cortisol response, suggest that the increased accelerometer-derived walking time in the smaller stall reflects a minor change rather than a serious indication of heightened stress or compromised welfare.

Based on behavioral observations, stall size did not affect the percentage of time spent in sternal or lateral recumbency, walking, rolling, standing, or stand resting. No significant effect of day or stall size × day interaction was detected for these behaviors. Although defecation frequency differed significantly between stall sizes, with a higher frequency observed in the smaller stall, fecal output was the same in both stalls. Drinking and urination exhibited a significant effect of day, varying across days without a consistent directional trend, but were not influenced by stall size. However, these day-to-day differences should be interpreted cautiously, as the overall observation period was limited. Consequently, observed day effects may reflect transient acclimation, novelty, or short-term behavioral adjustment rather than stable changes in welfare or housing preference. Importantly, no consistent stall size × day interactions were detected for any behavior, suggesting that temporal variation was not differentially influenced by stall size. While these findings provide insight into short-term behavioral dynamics during stall housing, longer observation periods are needed to determine whether such day effects persist and have biologic or welfare significance.

Although recumbent behavior is considered a sensitive indicator of comfort and sleep quality in horses,^20,64,65 substantial individual variation such as age, social rank, musculoskeletal pain, and environmental conditions can all influence recumbent behavior in horses,^{9,17,20–22,26,27,66} stall size can be a major limiting factor.²⁶ In our study, 3 horses did not display lateral recumbency in either stall size, and 1 horse (horse no. 4) did not lie down in either sternal or lateral recumbency at any point during the study. Similar patterns have been described in stall-housed horses, characterized by reduced frequency of lying down.^64,65 These individual differences, combined with the short duration of each stall phase, likely explain the absence of recumbency in these horses. Importantly, the same horse failed to lie down in both stall sizes, indicating that this behavior reflects individual resting patterns rather than an effect of stall dimensions.

The findings from this study suggest that a 136-ft² stall size does not compromise short-term welfare as compared with 144-ft² stalls in adult horses under the conditions tested. Horses had comparable feed intake, resting behavior, time budgets, and physiologic parameters in both stall sizes. This may have important implications for equine facility design and management, particularly in settings where space is limited or where preexisting stalls measure <144 ft². While larger stalls are often considered ideal, these results indicate that smaller stalls (as compared with the recommended 144-ft² stalls) can still support essential behaviors and meet welfare standards for healthy adult horses housed individually.

Several limitations of the study should be acknowledged. First, the stall size difference evaluated was relatively modest (144 ft² compared with 136 ft²), which may have limited our ability to detect meaningful behavioral or physiologic differences. These sizes were not selected based on experimental criteria but rather reflect the existing stall configurations at our institution, which may restrict generalizability to facilities with a broader range of stall sizes. Second, the study included 8 of the largest horses, which would be most greatly influenced by a reduction in stall size; however, this sample may not capture the full spectrum of responses present in a larger or more diverse population. Although some variation in age and sex was present, the small sample size and paired crossover design limited the ability to statistically model these factors. All horses were healthy, in regular work, and maintained under identical management, housing, and feeding practices throughout the study period. Third, the observation period was limited to 4 days per stall condition, raising questions about whether longer-term housing would produce cumulative or delayed effects. In addition, physiologic stress indicators such as heart rate, heart rate variability, respiratory rate, and rectal temperature were not measured in this study, which may have provided further insight into subtle stress responses not captured by cortisol or behavioral data alone. Another limitation of this study was the loss of video footage during the final week of data collection. Due to technical issues, several overnight recordings were incomplete for a subset of horses, and one horse intermittently displaced the camera, resulting in periods without visual coverage. Although sufficient video was available to complete behavioral scoring for all horses, these gaps reduced the total amount of footage collected and should be noted as a constraint on the completeness of the behavioral dataset. Although environmental parameters such as temperature and humidity were monitored and showed no marked variation during the study, changes in environmental conditions could interact with stall size and influence horse behavior or welfare outcomes. Future studies should explore a wider range of stall dimensions, longer observation periods, a larger sample of horses of various sizes, and seasonal or climatic variability to fully assess the long-term welfare implications of reduced stall size.

The role of attentive caretakers and husbandry staff is crucial in maintaining the welfare of horses housed in stalls. While our study did not find stall size to be a significant factor affecting short-term welfare indicators, it is important to recognize that not all horses adapt equally well to stall housing. This underscores the importance of daily observation and individualized care. Caretakers who are attuned to each horse’s normal behavior and needs can identify subtle signs of discomfort or stress that may not be captured by standard measurements. Such attention is particularly important in research and teaching herds, where housing and handling conditions vary. Ensuring that husbandry personnel remain vigilant and responsive to individual variations in behavior and well-being is essential for supporting overall welfare, regardless of stall size.

Given their multifaceted roles, evaluating and optimizing housing conditions for horses is essential for ensuring their welfare and for supporting the validity and reproducibility of research outcomes. This study contributes to the evaluation of stall size recommendations by assessing the impact of housing horses in 2 different-sized stalls on key welfare indicators. Future studies should evaluate whether existing stall-size recommendations align with horses’ actual behavioral and physiologic needs and should also investigate whether environmental enrichment can improve comfort and welfare in smaller stalls without requiring major structural changes.

Conflict of Interest

The authors have no conflicts of interest to declare.

Funding

This work was internally funded by residency funds from the Cornell University Center for Animal Resources and Education (CARE) and the College of Veterinary Medicine.

Data availability

Data are accessible by contacting the corresponding author.