Anne is a founding member of APDT Ireland and is CAP2 certified (with distinctions).
She works with a large canine welfare organisation, A Dog's Life, as a training/behaviour consultant and board member through which she sees hundreds of dogs and their people every year.
She works with a large canine welfare organisation, A Dog's Life, as a training/behaviour consultant and board member through which she sees hundreds of dogs and their people every year.
1. Aversives and How Dogs Learn
To get us started we first need to understand what aversives are and how they affect dogs and their ability to learn.
Aversives are things that an animal will work to avoid. They are things the individual finds unpleasant.
To truly understand what aversives are and how they work, we need to take a closer look at how dogs (and all animals) learn.
There are lots of ways that animals learn and it depends on the brain and complexity of it that determines how learning happens. The main types of learning that we use when teaching dogs, and probably the main ways that dogs learn are through what is known as conditioning, operant and classical conditioning. Don’t worry about the technical terms!
Classical conditioning involves the dog learning involuntarily (they are not thinking, they are responding) and this governs how the dog feels about a situation and all the things present. The dog learns to predict that something is going to happen because of the presence of something in the environment. The dog learns that this signals something good is about to happen (feel happy, anticipatory) or the dog might learn that this signals that something bad is about to happen (feel afraid, anxious).
This clip is a funny covering of the predictive power of classical conditioning and I use it in all my lectures on this topic:https://www.youtube.com/watch?v=Eo7jcI8fAuI
Operant conditioning teaches the dog what results they can get out of behaving in a certain way. The dog learns that a behaviour they carry out may result in one or more of the following:
- something I like might become available (in technical speak this is known as positive reinforcement)
- something I like might be taken away or withheld (in technical speak this is known as negative punishment)
- something I don’t like might start (in technical speak this is known as positive punishment)
- something I don’t like might be taken away (in technical speak this is known as negative reinforcement)
The dog learns to carry out behaviours that result in things they like becoming available and behaviours that result in things they don’t like being taken away.
The dog learns to avoid behaviours that result in things they don’t like and behaviour that result in things the dog likes being taken away.
All but positive reinforcement (the first one) involve the application of aversives.
BUT it is not so straight forward. Positive reinforcement cannot work without negative punishment (an aversive).
You only feeding a treat for the dog sitting is positive reinforcement but you are also withholding that treat for standing or any non-sitting behaviour (negative punishment).
The point here is that nobody can say that they don’t use aversives or that they only train with positive reinforcement.
Aversives are everywhere in life – it rains, the alarm goes off on Monday morning, you stand on a piece of lego, you touch a hot plate and so on.
It is incorrect to say that aversives ‘don’t work’. They absolutely do modify behaviour (cause learning).
We have evolved to avoid the fire if we get burnt; we have evolved to learn to avoid things that predict the fire and the burning.
But it depends on your definition of what ‘works’.
To get us started we first need to understand what aversives are and how they affect dogs and their ability to learn.
Aversives are things that an animal will work to avoid. They are things the individual finds unpleasant.
To truly understand what aversives are and how they work, we need to take a closer look at how dogs (and all animals) learn.
There are lots of ways that animals learn and it depends on the brain and complexity of it that determines how learning happens. The main types of learning that we use when teaching dogs, and probably the main ways that dogs learn are through what is known as conditioning, operant and classical conditioning. Don’t worry about the technical terms!
Classical conditioning involves the dog learning involuntarily (they are not thinking, they are responding) and this governs how the dog feels about a situation and all the things present. The dog learns to predict that something is going to happen because of the presence of something in the environment. The dog learns that this signals something good is about to happen (feel happy, anticipatory) or the dog might learn that this signals that something bad is about to happen (feel afraid, anxious).
This clip is a funny covering of the predictive power of classical conditioning and I use it in all my lectures on this topic:https://www.youtube.com/watch?v=Eo7jcI8fAuI
Operant conditioning teaches the dog what results they can get out of behaving in a certain way. The dog learns that a behaviour they carry out may result in one or more of the following:
- something I like might become available (in technical speak this is known as positive reinforcement)
- something I like might be taken away or withheld (in technical speak this is known as negative punishment)
- something I don’t like might start (in technical speak this is known as positive punishment)
- something I don’t like might be taken away (in technical speak this is known as negative reinforcement)
The dog learns to carry out behaviours that result in things they like becoming available and behaviours that result in things they don’t like being taken away.
The dog learns to avoid behaviours that result in things they don’t like and behaviour that result in things the dog likes being taken away.
All but positive reinforcement (the first one) involve the application of aversives.
BUT it is not so straight forward. Positive reinforcement cannot work without negative punishment (an aversive).
You only feeding a treat for the dog sitting is positive reinforcement but you are also withholding that treat for standing or any non-sitting behaviour (negative punishment).
The point here is that nobody can say that they don’t use aversives or that they only train with positive reinforcement.
Aversives are everywhere in life – it rains, the alarm goes off on Monday morning, you stand on a piece of lego, you touch a hot plate and so on.
It is incorrect to say that aversives ‘don’t work’. They absolutely do modify behaviour (cause learning).
We have evolved to avoid the fire if we get burnt; we have evolved to learn to avoid things that predict the fire and the burning.
But it depends on your definition of what ‘works’.
2. Aversives, Stress and Learning
Stress is generally thought of as all bad. It isn’t. A level of stress that the individual can handle is referred to as mental stimulation and it drives performance and motivation. When stress puts more pressure on the individual beyond what they can handle at that time, it becomes distress. This is the not-so-nice type of stress that can be damaging and the one we want to reduce and avoid.
Stress is simply anything that causes the body to change and adapt.
Learning is stressful and a lot of it is beneficial stress but because it awakens the stress response systems in the body it can of course tip over and begin to cause distress.
When an animal is taking in information from the world around it there are two main parts of the brain that process that information and determine the animal’s responses.
i. There is the reactionary part, the amygdala, which will be engaged if the brain determines that the world is not safe. This is an emotional part of the brain and causes the dog to experience the fight or flight response (a complex physical and emotional response involving lots of processes and the release of different chemicals) – we will be familiar with this experience – it’s butterflies in your tummy, rapid heart and breathing, trepidation and possibly a feeling of anxiety.
ii. There is the thinking part, the cortex, which will take a little more time to process information and work out the best course of action. This is the higher more complex part of the brain that is less concerned about immediate issues of survival (that’s the amygdala’s job!) and allows the animal to weigh up the situation and make the best decision.
What’s interesting is that between these two areas of the brain, when one is engaged the other doesn’t work.
If the amygdala is dealing with keeping the animal safe and alive, the thinking cortical part isn’t working. This makes sense – if you are in mortal danger or likely to be, you don’t want to hang about thinking about it and taking your time – your brain will cause you to react quickly and efficiently to get you to safety.
If the thinking cortical part of the brain is working the amygdala and its panicking is not active. This also makes sense – if you have time to think, process and work it out panicking and reacting isn’t going to help you make a sensible decision.
So, if we keep stress low and at a manageable level the cortical part of the brain remains engaged and working, while also keeping the panicking amygdala turned off so preventing further stress responses.
And if stress is higher we keep the amygdala engaged we will get panic and very limited learning because the thinking part of the brain can’t engage.
As we have covered, all learning is stressful to some degree. But if we apply aversives we increase the stress response and tip into distress – exposure to aversives causes a withdrawal and avoidance response (remember we have evolved to avoid the fire or things signalling that fire is coming). This limits learning because the thinking part of the brain can’t engage under this level of stress.
Stress is generally thought of as all bad. It isn’t. A level of stress that the individual can handle is referred to as mental stimulation and it drives performance and motivation. When stress puts more pressure on the individual beyond what they can handle at that time, it becomes distress. This is the not-so-nice type of stress that can be damaging and the one we want to reduce and avoid.
Stress is simply anything that causes the body to change and adapt.
Learning is stressful and a lot of it is beneficial stress but because it awakens the stress response systems in the body it can of course tip over and begin to cause distress.
When an animal is taking in information from the world around it there are two main parts of the brain that process that information and determine the animal’s responses.
i. There is the reactionary part, the amygdala, which will be engaged if the brain determines that the world is not safe. This is an emotional part of the brain and causes the dog to experience the fight or flight response (a complex physical and emotional response involving lots of processes and the release of different chemicals) – we will be familiar with this experience – it’s butterflies in your tummy, rapid heart and breathing, trepidation and possibly a feeling of anxiety.
ii. There is the thinking part, the cortex, which will take a little more time to process information and work out the best course of action. This is the higher more complex part of the brain that is less concerned about immediate issues of survival (that’s the amygdala’s job!) and allows the animal to weigh up the situation and make the best decision.
What’s interesting is that between these two areas of the brain, when one is engaged the other doesn’t work.
If the amygdala is dealing with keeping the animal safe and alive, the thinking cortical part isn’t working. This makes sense – if you are in mortal danger or likely to be, you don’t want to hang about thinking about it and taking your time – your brain will cause you to react quickly and efficiently to get you to safety.
If the thinking cortical part of the brain is working the amygdala and its panicking is not active. This also makes sense – if you have time to think, process and work it out panicking and reacting isn’t going to help you make a sensible decision.
So, if we keep stress low and at a manageable level the cortical part of the brain remains engaged and working, while also keeping the panicking amygdala turned off so preventing further stress responses.
And if stress is higher we keep the amygdala engaged we will get panic and very limited learning because the thinking part of the brain can’t engage.
As we have covered, all learning is stressful to some degree. But if we apply aversives we increase the stress response and tip into distress – exposure to aversives causes a withdrawal and avoidance response (remember we have evolved to avoid the fire or things signalling that fire is coming). This limits learning because the thinking part of the brain can’t engage under this level of stress.
3. Using Aversives
As already discussed we can’t actually avoid using aversives or being exposed to them (just in everyday life) but we can be careful in their application by understanding how they affect the learner (the dog) and how they work.
Using rewards in learning requires just as much skill as using aversives. But if we make mistakes when using rewards (such as poor timing or associations) we might get a dog that is frustrated and there will be an inhibition of learning. (I am not down playing this.)
But when we make mistakes in the application of aversives there are more serious side effects that are likely to be seen.
This is primarily why, from a professional ethics and liability point of view, I will teach pet owners how to implement reward based training because the mechanical skill required is tough to get right so I don’t want to give tools to pet owners where the inevitable mistakes create health, welfare and safety issues for the dog or people.
Everything else aside, aversives are most effective when:
- they are intense, introduced at maximum intensity and consistently applied
- they must be introduced during or within less than 3 seconds of the behaviour to be suppressed
- any prior rewards that the undesired behaviour gained the dog must be eliminated
First off, we are not very good at being consistent with our dogs and we are especially bad at consistency with the application of aversives because it we might not be there to catch it every time and in time.
The aversive applied MUST outweigh the reward the dog gets out of doing the behaviour. This is especially relevant where the dog is exhibiting a behaviour due to stress. If a dog is stressed in a social situation and uses distance increasing signalling (e.g. growling) to get someone away from them and then an aversive is applied (which itself increases stress) that dog will be strongly rewarded by getting the scary thing to move away (because of the growling) and is likely to do it again and take the following aversive.
Therefore when we apply an aversive and believe that we are “really teaching them a lesson” we are not always getting what we think we are:
- we might affect that behaviour in the future
- we might not affect that behaviour in the future
- we might affect other behaviour/s in the future
These effects of the application of aversives are referred to as ‘fallout’ or side effects. They include some of the following:
a) Negative Associations: there is more to learning than just the behaviour we can see. That is there because the dog gets one of those four possible options (operant conditioning). But underlying that is an emotional response (governed by the other type of learning, classical conditioning).
If an aversive is applied the dog will not necessarily be able to pin point which thing in the environment predicted its arrival and will form associations between the unpleasant feeling and all things present at that time, including you, children, other dogs and animals, the room you are in, the location, other people and so on.
Because of this lots of remote products have been developed so that aversives can be delivered without it directly appearing to come from the trainer. Negative associations will still occur to other things in the environment.
Therefore before an aversive is even presented the dog may already be engaging the panicking part ofthe brain just by you or something related being present.
b) Behaviour Vacuum: we now get that aversives cause the amygdala to start panicking and engage fight or flight. But, we have selectively bred domestic dogs to not fight with us and not flee from us – we have essentially selected for dogs that will take our s**t.
The third survival option therefore is ‘freeze’ and that’s when dogs for the most part do when faced with aversives, especially when delivered by someone to whom they are bonded.
So when the world is dangerous everything stops – the dog stops offering behaviour.
This is something that makes people presume that their application of aversive is ‘working’. The dog stops doing (everything) so it must be working!!
Vacuums are always filled with some behaviour and that might not be one you want either.
c) Suppression: as above all behaviour is suppressed. This is most dangerous when aversives are applied to warning behaviour such as growling, snarling etc.
Growling and other warning behaviour is communication that’s letting you know in no uncertain terms that the dog is uncomfortable and has been for some time.
If we apply aversives in response to this and the dog’s behaviour is suppressed we have not made the dog feel any better about the situation (he probably feels a lot worse about it and your presence now) but now we have taken away his ability to warn….
This leaves the dog little choice but to escalate his communication attempts to snarling, snapping, biting.
Indeed that the dog is growling is probably because our previous ignoring of their more subtle signalling behaviour that is caused it to escalate in the first place.
d) Reinforcing for the punisher: because of this suppression the person believes what they are doing, on the surface, “works” so they are likely to do it again.
Dogs are very appeasing deferent animals (again something we have selected for in them) and when they are frighted, startled etc. they often respond with strong appeasing behaviours that people interpret as them being sweet and apologetic. Dogs don't have a moral compass like ours so wrong and right isn't relevant for them - they do not understand being apologetic to us. But, they do learn that this behaviour often stops the aversive. For people though this generally encourages them to use an aversive again.
e) Timing: if not timed really really efficiently the aversive can be associated with other goings on and this can be dangerous especially where children or other dogs are present.
f) Potential for Abuse: because it’s quite difficult to use them correctly aversives and the threat of aversives will often need to be used over and over, with little change or effect on behaviour. Just making the dog feel like crap and not giving him the tools to learn how to avoid yucky things (through learning) is abusing our power that we have over these animals.
g) Punishing Nothing: often times people will say that they are punishing the dog ‘for not walking nicely’, ‘for not greeting politely’ and so on. If the dog is not doing it, continuing to apply aversive isn’t going to give him the information he needs to do the desired behaviour.
As already discussed we can’t actually avoid using aversives or being exposed to them (just in everyday life) but we can be careful in their application by understanding how they affect the learner (the dog) and how they work.
Using rewards in learning requires just as much skill as using aversives. But if we make mistakes when using rewards (such as poor timing or associations) we might get a dog that is frustrated and there will be an inhibition of learning. (I am not down playing this.)
But when we make mistakes in the application of aversives there are more serious side effects that are likely to be seen.
This is primarily why, from a professional ethics and liability point of view, I will teach pet owners how to implement reward based training because the mechanical skill required is tough to get right so I don’t want to give tools to pet owners where the inevitable mistakes create health, welfare and safety issues for the dog or people.
Everything else aside, aversives are most effective when:
- they are intense, introduced at maximum intensity and consistently applied
- they must be introduced during or within less than 3 seconds of the behaviour to be suppressed
- any prior rewards that the undesired behaviour gained the dog must be eliminated
First off, we are not very good at being consistent with our dogs and we are especially bad at consistency with the application of aversives because it we might not be there to catch it every time and in time.
The aversive applied MUST outweigh the reward the dog gets out of doing the behaviour. This is especially relevant where the dog is exhibiting a behaviour due to stress. If a dog is stressed in a social situation and uses distance increasing signalling (e.g. growling) to get someone away from them and then an aversive is applied (which itself increases stress) that dog will be strongly rewarded by getting the scary thing to move away (because of the growling) and is likely to do it again and take the following aversive.
Therefore when we apply an aversive and believe that we are “really teaching them a lesson” we are not always getting what we think we are:
- we might affect that behaviour in the future
- we might not affect that behaviour in the future
- we might affect other behaviour/s in the future
These effects of the application of aversives are referred to as ‘fallout’ or side effects. They include some of the following:
a) Negative Associations: there is more to learning than just the behaviour we can see. That is there because the dog gets one of those four possible options (operant conditioning). But underlying that is an emotional response (governed by the other type of learning, classical conditioning).
If an aversive is applied the dog will not necessarily be able to pin point which thing in the environment predicted its arrival and will form associations between the unpleasant feeling and all things present at that time, including you, children, other dogs and animals, the room you are in, the location, other people and so on.
Because of this lots of remote products have been developed so that aversives can be delivered without it directly appearing to come from the trainer. Negative associations will still occur to other things in the environment.
Therefore before an aversive is even presented the dog may already be engaging the panicking part ofthe brain just by you or something related being present.
b) Behaviour Vacuum: we now get that aversives cause the amygdala to start panicking and engage fight or flight. But, we have selectively bred domestic dogs to not fight with us and not flee from us – we have essentially selected for dogs that will take our s**t.
The third survival option therefore is ‘freeze’ and that’s when dogs for the most part do when faced with aversives, especially when delivered by someone to whom they are bonded.
So when the world is dangerous everything stops – the dog stops offering behaviour.
This is something that makes people presume that their application of aversive is ‘working’. The dog stops doing (everything) so it must be working!!
Vacuums are always filled with some behaviour and that might not be one you want either.
c) Suppression: as above all behaviour is suppressed. This is most dangerous when aversives are applied to warning behaviour such as growling, snarling etc.
Growling and other warning behaviour is communication that’s letting you know in no uncertain terms that the dog is uncomfortable and has been for some time.
If we apply aversives in response to this and the dog’s behaviour is suppressed we have not made the dog feel any better about the situation (he probably feels a lot worse about it and your presence now) but now we have taken away his ability to warn….
This leaves the dog little choice but to escalate his communication attempts to snarling, snapping, biting.
Indeed that the dog is growling is probably because our previous ignoring of their more subtle signalling behaviour that is caused it to escalate in the first place.
d) Reinforcing for the punisher: because of this suppression the person believes what they are doing, on the surface, “works” so they are likely to do it again.
Dogs are very appeasing deferent animals (again something we have selected for in them) and when they are frighted, startled etc. they often respond with strong appeasing behaviours that people interpret as them being sweet and apologetic. Dogs don't have a moral compass like ours so wrong and right isn't relevant for them - they do not understand being apologetic to us. But, they do learn that this behaviour often stops the aversive. For people though this generally encourages them to use an aversive again.
e) Timing: if not timed really really efficiently the aversive can be associated with other goings on and this can be dangerous especially where children or other dogs are present.
f) Potential for Abuse: because it’s quite difficult to use them correctly aversives and the threat of aversives will often need to be used over and over, with little change or effect on behaviour. Just making the dog feel like crap and not giving him the tools to learn how to avoid yucky things (through learning) is abusing our power that we have over these animals.
g) Punishing Nothing: often times people will say that they are punishing the dog ‘for not walking nicely’, ‘for not greeting politely’ and so on. If the dog is not doing it, continuing to apply aversive isn’t going to give him the information he needs to do the desired behaviour.
4. What can we do instead?
As we have discussed, we can’t teach or learn without aversives. But when applying aversives we don’t need to cause the dog too much stress that ultimately inhibits learning.
With an understanding of canine signalling we can better determine the emotional experience of the dog and therefore have a better picture of how what we do affects the dog and its ability to learn.
It is not good enough to just ignore behaviour that we don’t like. And it is not fair of professionals like me to just say that you need to reward behaviour that you want. It’s much more complex than that.
The first thing we need to remember is that because dogs have no choices we need to take responsibility because we are primates with big brains ;-)
We know that dogs are going to do things that are in conflict with what people want – all the behaviours that want to apply aversives to such as barking, pulling jumping, peeing, chewing biting, growling, humping, destroying, running away and so on. All normal, natural and necessary dog behaviour that we happen not to like!
Dogs are wonderfully unashamedly hedonistic – they do what works. If they find a behaviour rewarding they do it... a lot. Dogs do what works.
All that normal dog behaviour (that we don’t like) dogs must do – it is part of the package. So, we need to prevent them getting their kicks out of the behaviour we don’t like and provide them with more acceptable outlets for normal, natural, necessary dog behaviour.
This is where management comes in. This means that we take responsibility for limiting the dog’s ability to make mistakes. Remember, they didn’t ask to live in the human world with our rules.
Management means preventing the dog ever being rewarded for behaviours that we don’t like – for example, crate training prevents dogs from chewing the wrong thing, peeing in the wrong place, jumping up on the wrong person, barking at the wrong times and so on.
If the dog does make a mistake, it’s on us – we need to take responsibility that our management has failed and we need to step it up.
But management is just the beginning…. Remember that dogs must carry out those behaviours that we don’t like so if we are managing their behaviour really well they don’t get to do those behaviours, and this in itself can be damaging. We need to provide acceptable outlets so that dogs can get their jollies but in a safe and acceptable way.
More on enrichment with acceptable outlets here:http://pawsitivedawgs.wordpress.com/2012/04/13/enrichments/
While managing and allowing for acceptable outlets for normal behaviour, we also need to actively teach the acceptable behaviours. Remember that dogs do what works so if we limit their ability to make mistakes and provide them with rewards for behaviours we like (the dog will like them too as they are associated with good things!) then we will see the behaviours we like, next to no behaviours we don’t like and have minimal need for aversives.
More here:http://pawsitivedawgs.wordpress.com/2010/12/27/preparefortheworst/
More on solving behaviour ‘problems’ here:http://pawsitivedawgs.wordpress.com/2011/12/28/problemsolving/
But sometimes management and our best intentions fail – that’s ok, we are human! So, what do we do if faced with behaviour that we don’t like?
There are two options – do you want to stop the behaviour RIGHT NOW or do you want to teach the dog and affect future behaviour?
Generally if we are surprised by undesired behaviour we just want to stop it and I think that’s a better option in most scenarios. This should be done as non-confrontationally as possible. As we are just stopping and not teaching we don’t want to lash out and get all the fallout from aversives and none of the learning.
Divert the dog’s attention, get them away from the situation by tossing food, leading them away, opening the door to allow them out and so on. Remove them from the situation to prevent any further problems and to allow you to calm down too.
Sometimes we might want to use aversives in teaching, and this is much more difficult and we need to be careful when applying it to the individual dog.
We might withdraw the reward the dog wants or we might withdraw/remove the dog.
Withdrawals are generally very brief to allow the dog to learn to choose the right behaviour – they can then be rewarded for that with access to that reward again.
Removing the dog is more akin to a time out. Time outs must be used with care and the trainer needs to earn them. If you want to use a time out make sure to have rewarded the desired behaviour at least ten times for every one time out. If you can’t achieve this ratio there is something not quite right with your management and supervision – again the responsibility of the primate at the other end of the leash.
Using times out should allow the dog to calm a little and in effect manages their behaviour too. They shouldn’t last any longer than 20 seconds to see learning. The dog should be timed out in a boring place where they are confined, safe, and have nothing with which to entertain themselves.
More here: http://pawsitivedawgs.wordpress.com/2010/12/27/quickfixes/