Is your horse in the mood to learn?
In an article that foreshadows the science of horsemanship (rather than the science of training alone), researchers from the University of Sydney show why trainers should adjust their techniques to the horse’s mood and emotional state.
Behaviour scientists say that learning processes are universal and just like all beings, horses can be trained, or more precisely they learn to modify their behaviour by three distinct processes: trial and error (operant conditioning); association (classical conditioning); and getting used to things (habituation). They named these the Principles of Learning, or Learning Theory; an apt term because as many horse people will tell you, they may work perfectly in theory, but turn out to be a lot more complex to put into practice.
Researchers from the University of Sydney however, have taken a definite step towards bridging the gap between the lab and the arena with the recent publication of an article that explains why in real life, training is so complex and the outcomes so variable. They say that if your horse is in the wrong mood, and either too alert or not alert enough, it may not respond to the type of training you are using, even if you are applying it correctly. They are trying to predict why and when this is likely to happen.
The researchers are proposing that a horse’s response to training is a dynamic three dimensional landscape that is influenced by the horse’s emotional state at the time.
Taking into account how the animal ‘feels’ at the time of training may help to explain exactly why a horse works better some days than others, even at a particular time of the day, or why a horse ‘ticks’ with one rider, or under one management system and not another, even if they are under the same training regimes.
Until recently, scientists were not able to measure affective state in animals, but thanks to advances in the area of cognitive bias, they can test and determine what they call ‘emotional valence’ – whether they are in a positive mood or negative one. Cognitive bias testing has now been done in many species, from sheep and dogs to honeybees, to determine where individual animals lie in the optimism to pessimism scale, and the news is that equitation scientists in Switzerland have recently tested horses. Measuring arousal on the other hand is not new. It can be reliably determined by observing behavioural responses (position of the ears and tail, widening of the eyes, movement, etc) and confirmed by taking physiological measurements such as heart rate, cortisol, etc…
Using a very impressive set of 3-D graphs and cool (free) software that allows you to rotate and view from any angle (see http://hdl.handle.net/2123/8989), the article combines the affective and arousal state to provide a framework for predicting how horses in different emotional states will respond to training techniques in the four quadrants of operant conditioning (see the article as it appeared in the July 2013 issue by clicking on the imgage above).
Study author and PhD candidate Melissa Starling admits working in 3-D has been a brain draining experience. “When I started to come up with these graphs, I was giving myself headaches! Trying to follow the four different quadrants as well as the three dimensions was a lot to juggle in my mind, but I cannot simplify it. Training is complex and we have to take into account all the different elements.”
“The bottom line,” she says “is we should look at the animal in front of us and be able to say – if its affective state is there and its arousal there, then what should we do? How should we train each specific task? Instead of looking at an animal and just saying we want it to give ‘this’ response, so let’s make it happen; we should look at it from the perspective of what emotional state we want the animal to be in before we train it. If we do that, we might be getting somewhere”.
Although Starling’s project is mostly about detecting emotional states in dogs, she believes the same can be applied to horses, and with the collaboration of leading equine behaviour scientist Professor Paul McGreevy, the study includes two scenarios that are common in horse training; teaching the horse to target, and teaching it to go forward from a rider’s leg aid. The idea behind including graphs for horses was to develop a model that will encourage equitation scientists worldwide to experiment and test the theory proposed, to see if it is actually true.
Professor McGreevy who will be presenting this concept in a plenary at the upcoming Equitation Science Conference in Delaware, USA on the 19th July, noted that the current article offers only a conceptual approach to how different training techniques interact with affective state and arousal: “The challenge is to start populating the graphs with more data. First to develop a standardised test to determine a horse’s affective state at the time, and then compare different horses and even the same horses under different management, because as we manage them better, we should expect to see their emotional state improve. The future should see us developing measures of affective state and arousal and then moving from the conceptual landscapes to validated models, ideally for each horse and discipline, at every stage of training”.
Starling hopes that objectively measured, affective and arousal state can add information to the complex mix of interacting elements that affect training outcomes, and there are many other aspects that can be added to the mix: “Some scientists are already looking at the influence on behaviour based on laterality, or hair whorls, or height at the withers. By adding more and more objective measures, we may find that behaviour is not quite as chaotic as we thought, and not as variable. As we progress, we may be able to narrow down the number of behavioural responses we see. For example, if we say that an animal is ‘this’ level of optimism, and the base level of arousal at the time is ‘here’ – we might see that there’s only three behaviours likely to show in this particular test, whereas without that, and looking at a population there may be fifteen. I’m hoping this will help us organise how we think about behaviour, and how to predict it.”
While there is little doubt that some trainers have an instinctive ‘feel’ for an animal, taking into account emotion has, until now, been unchartered territory for equitation scientists, but Starling argues that it is not for a lack of trying: “Some people think that scientists want to boil training down to very simple robotic responses, and it often seems that’s the way we are talking, but it’s not.
“We all understand that training animals is much more complex but we have to concentrate on the things we can measure and the things that are predictive. If there’s no way to measure emotional state, then it’s really difficult to take it into account – that’s not to say we don’t care about it! I’m hoping this will change in the near future. We’ve got good statistical models that we can plug numbers into, so it is time to start collecting some data.”
So is this about making training effective, or is it also about what is fair to the horse? “Definitely both” says Prof. McGreevy. “One follows the other.”
The article titled Conceptualising the Impact of Arousal and Affective State on Training Outcomes of Operant Conditioning, by Melissa Starling, Nicholas Branson, Denis Cody and Paul McGreevy is available in the open access journal Animals and can be found on this link. We recommend you see the full paper and the interactive 3D graphs!
Does your horse see the glass as half full or half empty?
Researchers have tested optimism and pessimism bias in dogs. This type of cognitive bias testing is being widely used and is considered an objective measure because the experiment is not affected by human subjectivity like most temperament tests are.
An automated system was developed to train the dogs to touch a target with their nose when they hear a sound of a particular tone. When they do so, the system delivers them a milk reward, which can even be moderated to suit the dog’s body size.
The machine is then set to deliver a sound of a different tone, but this time when the dogs touch the target, they get a non-reward of water (something they already have free access to, to ensure they are not thirsty).
The sound cues are set on a scale so there are nine notes between the milk sound and the water sound.
The dogs must reach a trained stage where every time they hear the ‘milk’ sound they touch the target and when they hear the ‘water’ sound they choose not to touch it. That’s the first stage of training and it is all controlled by a computer so there is no human intervention to confound the results.
Now comes the interesting time – the machine is set to give them a sound that is somewhere between the ‘milk’ and ‘water’ sounds. The researchers then observe what happens and how the dogs respond to the new ambiguous cues…
Will the dog see the glass is half full and ‘have a go’ anyway, or will he see it’s half empty and stop trying?
What the researchers are finding is that different personalities have a different resilience to when things don’t go according to plan.
The most optimistic dogs are prepared to risk disappointment, and break the beam even if the tone is just one note different from the water – the non-rewarding sound – whereas the pessimistic dogs are far more cautious and don’t want to risk disappointment – to them the glass is half empty.
They imagine a similar device will soon be developed for horses, maybe one that can be bolted on the stable wall to measure your horse’s optimism as it interacts with the device.
Click below to read this article as a pdf download…