Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsA horse that stands quietly, accepts a bit or a restraint, and keeps working is not, on its own, evidence that the method was good for it. That is the central point of Cristina Wilkins’s 2026 explainer in Horses and People, which argues that compliance and a lack of visible resistance do not show that a training, equipment, restraint or handling intervention had no welfare impact. To judge a welfare claim, ask four things: what welfare state the study actually measured, how it measured it, whether each horse was assessed as an individual, and what safeguards were built in to stop harm.
What a welfare claim has to show
Welfare is a chronic state, meaning the overall quality of an animal’s life over time. A short-lived response, such as a flinch, a spike in heart rate, or a moment of apparent calm, can differ from that overall state. A study that reports a brief response has measured something real, but it has not automatically measured welfare. The 2020 review Indicators of Horse Welfare: State-of-the-Art makes this distinction the foundation of its approach.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Equine Welfare (UFAW Animal Welfare) | $79.95 | Buy on Amazon |
| 2 |
|
A Horse's Life: The Neuroscience of Equine Welfare | $24.99 | Buy on Amazon |
| 3 |
|
Equine Welfare in Clinical Practice | $129.00 | Buy on Amazon |
| 4 |
|
Equine Welfare | $74.99 | Buy on Amazon |
| 5 |
|
Equine Applied and Clinical Nutrition: Health, Welfare and Performance | $143.31 | Buy on Amazon |
Three principles follow from that distinction:
- Welfare is inferred, not seen. No single visible sign, and no continued performance, reliably captures how a horse is faring. Relevant measures have to be meaningful, clearly defined, and scientifically validated.
- Temporary states are not the same as chronic ones. Pain, fear, stress and positive emotion can each occur within a session without describing the horse’s wider condition. Check whether the study’s measures are meant to capture a lasting state or a passing one.
- Group averages can hide individual harm. If most horses cope and a few do not, a group mean may look acceptable. Individual-level reporting is what reveals this.
A reading route for any welfare study
1. Write the claim as one sentence
Before opening the paper, write the claim in plain terms, such as “this bridle reduces stress during loading” or “this training method is humane for young horses.” Then identify four things: the study question, the primary outcome, the horses and setting studied, and the intervention. A claim is only as strong as the match between those particulars and the situation it describes.
2. Test whether the population and setting match
A result in one group does not automatically generalise to other horses, management systems, methods or durations. Breed, age, training history, activity level and housing can all change how an intervention plays out. Describe the gap between the studied situation and the claim as a limit on what can be inferred, rather than assuming the finding is false. A study of horses in one controlled setting may be sound and still say little about a different yard.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
3. Check whether the measure fits the question
Ask whether each outcome measures the thing the claim is about. The 2020 review recommends checking that signals reflect chronic state, that they measure the intended construct, that they are clearly described, and that they have been validated against other welfare indicators. It also cautions that subjective perception can introduce bias, and that observers need to agree on what they are scoring.
Indicators are not equally ready for use. The 2014 AWIN-related on-farm review considered 49 candidate indicators across welfare principles and criteria. Its abstract describes some measures with potential for on-farm use and identifies gaps, notably for criteria such as absence of pain and a positive emotional state. A study that relies on a single indicator in those gap areas should be read with caution.
| Source | Date | What it contributes to a reader | Limits to keep in mind |
|---|---|---|---|
| Indicators of Horse Welfare: State-of-the-Art | 2020 | Defines welfare as a chronic state; sets validation and observer-agreement criteria for indicators | Describes criteria for judging measures; does not rate individual indicators |
| AWIN-related on-farm equine welfare review | 2014 | Lists 49 candidate indicators and shows which areas have on-farm potential and which have gaps | Read at abstract level; specific indicator performance is not stated here |
| Five Domains Model review | 2022 | Ten-stage summary: species background, candidate measures, animal identification, feasible measurement, validation, grading, and a confidence score | Organises observations; does not make a study sound on its own |
| PREPARE guideline (Smith and coauthors) | 2017 | Planning checklist for animal studies, including field work | Guidance to be adapted to each project; not a quality score |
| COMPASS Guidelines (McGreevy et al., Animals, 16(2), 206) | 2026 | 124 detailed points on welfare risks in behaviour-modification research, as summarised by Wilkins | Described as a reference rather than a verdict; check the original paper for full wording |
Organising observations with the Five Domains Model
The Five Domains Model sorts objective information into four physical and functional domains: nutrition, physical environment, health, and behavioural interactions. It then uses those observations to make cautious inferences about the horse’s mental state. Its value for a reader is diagnostic. If a study reports only behaviour and ignores health or environment, you can see which dimensions are missing.
The 2022 review’s ten-stage summary also separates validated welfare-status indices from alerting indices that flag possible future risk. Those are different tools. An alerting index can justify closer monitoring without establishing that welfare is poor, and a welfare-status index needs validation before it supports a confident verdict. A study that uses the model is not automatically sound, and not every observation supports the same level of confidence.
Methods, ethics and safeguards
Design and statistics
The PREPARE guideline, by Smith and coauthors (2017), covers planning for animal research including field studies. It asks researchers to form a clear hypothesis with primary and secondary outcomes, consider prior reviews, assess species relevance and welfare needs, and plan for reproducibility. It also covers pilot work, statistical power, the experimental unit, animal numbers, randomisation, observer bias, and inclusion and exclusion criteria. Its authors state: “The guidance in this paper should be adapted to the individual research project, animal species and location.”
In practice, look for the experimental unit. If a study with ten horses analyses hundreds of observations as though they were independent, the apparent precision is overstated. Look for whether individual horses are reported separately, and whether observers were blind to the treatment where that was feasible.
Rank #3
Safeguards and stop criteria
Wilkins describes the COMPASS Guidelines as focused on welfare risks that can be underestimated in behaviour-modification research, used alongside ARRIVE and PREPARE. The explainer recommends looking for safeguards such as monitoring, calibrated measures, disclosure of conflicts, and predefined withdrawal criteria. A useful question, as she frames it, is: “Based on the COMPASS Guidelines, what welfare safeguards are clearly described here, and what information would a horse owner reasonably want before agreeing to participate?” Treat that as a prompt for generating questions, not a substitute for expert review.
The PREPARE guidance groups its recommendations into study formulation, dialogue between scientists and animal facilities, and quality control. Its topics include ethical and harm-benefit review, humane endpoints, staff training, animal health, housing, experimental procedures, and what happens to the animals after the study.
Free tools Windows power users keep installed
One-click scans. No signup required.
What a missing detail does and does not mean
When a paper does not describe a baseline measure, a monitoring schedule or a stop criterion, the honest conclusion is that you cannot verify it from the available reporting. That is different from proving the safeguard was absent in practice. Equally, an ethical approval statement or an author’s reassurance does not show that the chosen measurements could detect a welfare impact. Each claim needs its own evidence.
Rank #4
- Used Book in Good Condition
Wilkins puts the balance plainly: “The aim is not to accuse researchers of bad faith, or to reject research outright.” The reader’s task is to state what is reported, what is missing, and how that affects confidence in the conclusion.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Comparing two studies on explicit axes
When two studies or protocols address the same practice, compare them on the same six points:
- Question and outcome: what each study actually measured, and whether that matches the claim.
- Population and context: horse characteristics, management, activity and setting.
- Time scale: baseline, intervention period and follow-up, and whether the evidence concerns an acute response or longer-term welfare.
- Measurement quality: construct relevance, validation, observer agreement, calibration and feasibility in the setting.
- Analysis: sample-size rationale, experimental unit, individual versus group reporting, bias control and confounding.
- Safeguards and transparency: stopping criteria, ethics, conflicts, stated limitations, and whether protocol or supplementary detail is available.
Systematic reviews show how much uncertainty a field can contain. Baxley, Rodriguez-Mori and Anderson (2024) included 28 papers on horses in equine-assisted services. They found that published studies rely substantially on behavioural and physiological indicators, that methods varied, and that findings were inconclusive, and they called for robust studies that address confounders. That is a review conclusion for one field, not a verdict on all equine welfare work. Separately, Hall and Kay (2024), in a Part I systematic review, retained 179 articles across two separate searches, which illustrates how much filtering a review may involve before any conclusion is drawn.
Recommended Free Tools
Best Value
- Used Book in Good Condition
Using a checklist without treating it as a verdict
A checklist helps you ask consistent questions. It does not turn a study into “good” or “bad.” Wilkins makes this explicit: “The guidelines are not a checklist for labelling research as ‘good’ or ‘bad.’” Use the questions above to decide how much weight a claim can bear, and to describe its limits accurately. A study can score well on reporting and still answer a narrow question, and a study with thin reporting can still contain sound observations.
Legal context in the UK
The GOV.UK code of practice for horses, ponies, donkeys and hybrids, last updated on 16 April 2018, summarises owners’ duties under the Animal Welfare Act 2006. These include a suitable environment, a healthy diet, the ability to behave normally, appropriate company, and protection from pain, suffering, injury and disease. The code states that breach is not itself an offence, but a court may consider compliance in proceedings for a welfare offence. This is UK law. It does not set a scientific standard for judging a study, and it should not be applied to other countries without checking their own rules.
Quick Recap
Where this approach runs out
- Visible calm, compliance and continued performance do not establish good welfare, and this approach does not treat them as proof.
- A single physiological or behavioural measure is not a complete welfare verdict unless its validation and scope have been established for that use.
- Findings do not extend beyond the population, intervention, time scale and setting that were studied.
- No population-level figure is available for how many horse people feel confident judging welfare research. Wilkins’s statement that many lack that confidence is a qualitative framing, not a measured prevalence.
- The COMPASS details here come from Wilkins’s summary of McGreevy et al. (2026). Check the original paper before relying on specific wording.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
