13/08/2026
I continue to see inaccurate opinions from horse people, where they claim horses are able to plot, plan and strategise against us.
I saw a paper published at the ISES Conference currently running about language horse people use that describes horses as being “naughty” or “stubborn” and how that can influence the responses people choose, including increasing the likelihood of suggesting sedation or punishment.
People like to share an incorrect interpretation of a study about how horses learn and behave.
I thought I’d re-share my old post, where I addressed these articles.
**I’ve put a red line through the heading, as it’s inaccurate click bait. I took a photo of the article so people would know what I was talking about without having to share the link. Please don’t just read the heading and believe what you read 😉
- - - - -
There’s a couple of articles currently circulating about equine cognition, whether they can "plan ahead and think strategically", based on a recent study about model based learning in horses. I’ve taken a photo as I don’t want to share the link and perpetuate misinformation.
The click bait articles, (not the study), are rather alarming in their interpretations of the research and people are then extrapolating this misinformation further.
Here’s a couple of examples of magical mind reading and I believe also misrepresenting the study and suggesting that they knew what the horses were thinking and scheming:
“That was enough for the horses to go: ‘OK, let’s just play by the rules.’”
“It suggests that, rather than failing to grasp the tenets of the game, the horses had understood the rules the whole time but, astutely, had not seen any need to pay much attention to them in the second stage.”
- - - - - - -
Yikes! That’s a bit of mind reading there! Horses don’t think or plan like this, they didn’t know the “rules” and they weren’t deliberately behaving that way, they did not have enough information to know what to do, that’s all.
- - - - - -
Let me just pop this nice little quote here about Skinner and how good (errorless) training happens:
“In [Skinner’s] system, errors are not necessary for learning to occur. Errors are not a function of learning or vice-versa nor are they blamed on the learner. Errors are a function of poor analysis of behavior, a poorly designed shaping program, moving too fast from step to step in the program, and the lack of the prerequisite behavior necessary for success in the program” (Rosales-Ruiz, 2007).
- - - - - -
What actually happened in the study was that a group of horses firstly were trained to touch their nose to a target, a laminated card and were positively reinforced for that behaviour. They underwent two 1 hour training sessions and were positively reinforced for any kind of touch to the laminated card. That’s a substantial amount of target training and weirdly this was done before conditioning the bridging stimulus. They could have both conditioned the bridging stimulus AND trained the nose target at the same time and it may have made the bridging stimulus more salient by doing it this way.
Then they separately conditioned the meaning of a secondary reinforcer, the bridging stimulus, by pairing the sound of a whistle with food. They undertook two 15 minute sessions involving 3 minute continuous sessions interspersed with 2 minute breaks. Not a lot to condition a bridging stimulus!
The researchers then introduced a stop signal where the horse’s behaviour was marked with the whistle and positively reinforced with food when their nose touched the target only when the stop light was NOT on. Any nose targets during the stop contingency (when the stop light was on) were deemed errors.
What they found was that the horses didn’t discriminate between when they could nose target and when they weren’t supposed to.
This is not surprising as presentation of the laminated target is generally considered the Discriminative Stimulus ie. the cue, and there was already a lot of R+ history built on this behaviour and potentially a lot less on the conditioning of the bridging stimulus.
Generally when we are teaching things like nose targets, in the teaching phase, it’s always a good idea to present the target as the cue to touch it and remove it after the behaviour has been performed. This starts to build some discrimination into the behaviour and if we want a target present at all times, which they did in the study, then we have other strategies to teach stimulus control (put the behaviour on cue). Not only using the bridge as they did in the study, but for example combining the new behaviour with another known behaviour. The aim is that the horse cannot touch the target uncued, if they are performing another behaviour that is incompatible with nose targeting. We can interject the cue for the known behaviour to prevent the other behaviour being offered off cue. We can also use our bridging stimulus in the same way. This is an example of “errorless learning” where we do not need to resort to punishment to train the horse and prevent errors.
The researchers went away and designed an additional step where they introduced negative punishment in the way of a 10 second timeout where the person with the target and the food removed themselves.
I feel this study illustrates a number of processes of operant conditioning that we already know exist. Yes, reinforcement increases behaviour, punishment reduces behaviour and consequences influence future behaviour.
The study focused on model based learning and yes horses are problem solvers, they can learn this way, they can learn and behave in extremely nuanced and subtle ways, they're not just throwing out behaviour randomly in all directions like a water sprinkler.
I’d also question the role of extinction in the process of introducing the stop light and how that influenced behaviour and especially the number of errors. I’d also question the brief training time for “charging” the whistle and why the behaviour was taught before the bridge was conditioned and done separately.
Whilst I understand that this research needs to be done for people to even consider how equines learn, from my perspective as an equine R+ trainer, I would consider this one of the most basic training sessions. BUT they did not set the learner up for success, which is always my aim. R+ training has become very nuanced in how we can avoid errors and I love the errorless mindset.
Finally, to quote B.F. Skinner, the father of Operant Conditioning, “the implication that learning occurs only when errors are made is false”.
Link to the original study in the comments.