Conditioned stimulus

[Joy]

If a dog barks when the doorbell rings is this a conditioned stimulus? And if so, must a dog have had an unpleasant experience (even just such as not liking seeing a stranger at the door) to condition it?
Am I right in thinking that an unconditioned stimulus is an involuntary response, like salivating when they see food?

And how would you define stimulus control?

This doesn't relate to any practical concerns, just geekishness!

 

[snowbunny]

An unconditioned response is something that happens without thinking when a stimulus (unconditioned stimulus) presents. So, drooling at food, jumping at a loud noise etc.
A conditioned response is a behaviour that is learned in response to a stimulus. So, yes, a dog barking at the doorbell is a conditioned response to a stimulus. It doesn’t mean that it is or was unpleasant; we’d need to dissect the dog’s body language to determine that. The dog has likely paired it with people arriving, and that might be a worrying event or an exciting one.

Stimulus control is the act of putting a behaviour on cue, and not having that behaviour present (within a given context) unless it is cued. In general terms, it’s an animal behaving in one way when there is a certain stimulus and another way in the absence of that stimulus. For example, during a training session, you want your dog to sit every time you say “sit”, but if you don’t give the cue, the dog shouldn’t sit. Stimulus control is a bit confusing when we apply it to our dogs because of course we can’t have our dog only performing behaviours on cue in their daily lives; they need to be able to sit without being told to around the house! So it is very contextual - that’s ok though because all of the environment becomes part of that cue or stimulus.
An example is Squidge; she adores the behaviour of “front feet on” and will perform it on walks all the time, in the hope of earning a treat. Sometimes she gets one because she’s adorable
But in a training context, I want her to put her feet up only on my cue “hup” and at no other time. If she does it uncued, or in response to another cue, it means it’s currently not under stimulus control.

Does that make sense?

I think Rachael (@Oberon) explains it better than I am

 

[snowbunny]

I should say that sometimes there is a fine line between conditioned and unconditioned responses when the resulting behaviour is something like barking.
Take a human context: if something scares the bejeebers out of me, I might jump in a startle response (unconditioned response) but I also might swear loudly. That has to be a conditioned response as swearing itself is learned, but I’m not stopping and thinking, “this seems like an appropriate time to use my potty mouth”, it just comes out. Technically, it’s a conditioned response, but in practical terms, I’d be more successful treating it like an unconditioned response as it is so closely linked to that unconditioned startle response.
Similarly with dogs who alert bark; it’s hard to say whether the bark itself is conditioned or unconditioned but in practical terms, it’s far more effective to consider it reflexive and work therapeutically in that context.

 

[Joy]

Thanks. I need to digest this for a bit.

 

[snowbunny]

It was an interesting question! I can’t say I’m 100% on the full extent of stimulus control. There’s an interesting video about it from ClickerExpo a few years back called Control is an Illusion. I’ve watched it twice and need to go back and watch it a few more times still, I think

 

[snowbunny]

Here:
https://videos.clickertraining.com/CFVEX17SO01

“Stimulus control is not about how much control you have over your animal, but about how precisely you are able to communicate when reinforcement is available for a behavior and when it’s not. Stimulus control is also a reflection of overall dog and handler fluency at the moment the cues are learned. We’ve all seen or worked with dogs that fidget, pant, bark, and throw every behavior they’ve ever been clicked for when they are not quite sure what you want. These responses are not caused by over-arousal or the use of food in training, but by unclear criteria, unclear context cues, unhelpful defaults accidentally reinforced in past training sessions, and the resurgence of incomplete behaviors that were never fully put on cue to start. “

 

[Joy]

Stimulus control is not about how much control you have over your animal, but about how precisely you are able to communicate when reinforcement is available for a behavior and when it’s not.

What about when the handler gives the cue (stimulus) and receives the desired behaviour but doesn't offer reinforcement - for example in a competition environment?

 

[Oberon]

Do you mean will it be damaging in some way to not give reinforcement? The simple answer is that is the behaviour is very, very well learned (I.e. it’s a ‘habit’) then you’ll get the behaviour and the lack of reinforcement on the odd occasion won’t hurt (or diminish the behaviour next time).

The longer answer is that you want to prepare for competing by building up long behaviour chains. A behaviour chain is a way of building up duration. First you teach a bunch of separate behaviours, making sure they are well learned and strongly reinforced. Then you start adding them together. So, don’t just ask for a sit. Ask for a sit, then a drop, then deliver reinforcement. Next time add a stand as well before delivering the reinforcement.

Basically you are building up a string of behaviour with one reinforcer delivered at the end. When each behaviour is well learned and has a strong history of being reinforced, even just being asked to do that behaviour is reinforcing. The behaviour has become associated with good things (food/reinforcers) and, so, via classical conditioning, the cue itself has acquired positive characteristics. Also the cue offers a chance to perform a behaviour which means the chance to get a reinforcer. So a cue is a like a ticket to a lottery you know you are going to win. As a result, the chance to perform each behaviour reinforces the behaviour that came just before. In this way, strong behaviour chains can be built up.

So, when you perform a bunch of behaviours in the competition ring, you are reinforcing them all.....if you have taught them in the form of a behaviour chain that is reinforced at the end. Just a tip...in the ring, act like the whole thing is one long behaviour...engage with your dog the whole time. Don’t ‘stop’ or act like it’s ended at any point...keep looking at your dog...be excited...act like it’s one long party (which you dog should think it is as you have trained it that way).

 

[Oberon]

I think Rachael (@Oberon) explains it better than I am

I don’t think so! But thanks for the positive reinforcement ?

 

[Joy]

Good tips, thanks. I was more trying to see how the term 'stimulus control' fitted into a situation where someone cues a behaviour but doesn't reinforce - not necessarily in competition but in well established simple behaviours, like a sit maybe.

 

[Oberon]

Haha, sorry for the long post you didn’t ask for ?

I’d say that the behaviour is under stimulus control when the dog performs it on cue and doesn’t perform it at any other time.

It’s only worth a dog’s while to perform learned behaviours if there’s a possibility of reinforcement (which in training should be true even if you don’t actually reinforce a behaviour every time...the dog should think it’s possible). To put behaviours under stimulus control, humans need to make sure that reinforcement is only given for already taught behaviours when a behaviour has been performed in response to a cue. Behaviour performed off cue gets nothing from the handler. As Snowbunny said, dogs will throw familiar behaviours at you when you have not asked for them if they are not clear about when it’s worth their while and when it’s not. Dogs that toss out uncued behaviours have not learned that it’s only worth their while when the cue for that behaviour is given. It means that the behaviour is not under stimulus control, as it’s given by the dog in the absence of a cue. That means the handler has been sloppy about their training.

 

[snowbunny]

What about when the handler gives the cue (stimulus) and receives the desired behaviour but doesn't offer reinforcement - for example in a competition environment?

I was more trying to see how the term 'stimulus control' fitted into a situation where someone cues a behaviour but doesn't reinforce - not necessarily in competition but in well established simple behaviours, like a sit maybe.

We're still talking about the dog only offering the behaviour (within a given context) when the stimulus (cue) is given. That is the essence of stimulus control: the animal is very clear that the cue releases them to perform that particular behaviour, which can earn reinforcement.
Just because you're not offering reinforcement every single time doesn't mean anything in this context; you're working on a variable-ratio reinforcement schedule (where you reinforce the behaviour sometimes) not a continuous reinforcement schedule (where you reinforce the behaviour every time). If you never reinforce a behaviour, that behaviour will diminish - assuming you're not using punishment to achieve the behaviour - but reinforcement doesn't always have to come in the form of food or toys. For many dogs, reinforcement comes from the "work" itself, once the rules are established. We see this with retrieving all the time. We can develop long and complex behaviour chains with the reinforcement at the end, and each step between the first behaviour and the last is rewarded as a part of the chain.
Let's say that the behaviour chain you want is:
sit-down-heel-turn-sit
At first, you'd work on sit-reward. Then, depending on if you're chaining or back-chaining, you'll add in another behaviour. Back-chaining is often the most beneficial way of working with chains, so I'd add the turn to make turn-sit-reward. The dog knows that the sit is followed by the reward, so the sit in itself becomes rewarding. So the reward for the turn is being cued to sit. The reward for sit is their food or toy or whatever. By working backwards, we end up where every piece of the chain is being rewarded by the next piece. So even though to our minds we might think that the only reward is that food or toy at the end, in actual fact, each piece of the chain is being reinforced by the next. That's what happens in the obedience ring.

If you're truly never reinforcing a certain behaviour (let's say sit), then it will die out. If you think you're never reinforcing the sit on cue yet it isn't dying out then you're inadvertently using reinforcement, which will either be positive reinforcement that you'd not really considered (environmental, interaction etc) or negative reinforcement (the dog will sit in order to avoid you becoming grumpy and nagging them to sit).

So, to recap back to the original question:
The term "stimulus control" is used in a sense of the dog getting the clarity of understanding of what the cue is and what the behaviour is that accompanies it. The behaviour won't be offered for any other given cue, because the dog understands that that's not the cue for that behaviour. A dog trained using positive reinforcement will perform a behaviour on cue in the absence of a reinforcer because of the reinforcement history - that is a slightly separate issue to stimulus control: during the teaching phase, you will be using a high rate of reinforcement to associate the cue with the behaviour, and thereby developing that stimulus control (if you're being clear and consistent). Once the dog has developed that stimulus control, then you will lower and/or vary your reinforcement schedule.

 

[Joy]

Thanks both of you.

 

[Stacia]

My brain has gone into meltdown! I go by the philosophy to keep dog training as simple as possible

 

[snowbunny]

It is simple, these are just definitions which are really clean once you understand them. Having standardised definitions adds clarity and, therefore, simplicity once everyone speaks the same language

In fact, keeping it simple is one of the foundations of stimulus control - if you don't keep your cues clean and simple, then your dog won't understand and will become frustrated.

Personally, I don't like the term "stimulus control", but once you know that it just means "your dog understands what a specific cue means and so pairs the behaviour accurately with the cue", then you can use it to ensure you are on the same page as other people who are speaking your language. And, remember, Joy is learning to become a trainer so it's important to understand all this terminology; even though it's almost never going to be appropriate to use it with clients, it helps trainers ensure they're all on the same page. If someone says to her, "You need to ensure your clients have their cues under stimulus control", she will need to understand what is meant by that.

I don't think the average person and their dog need to understand the behavioural science behind dog training; in fact, I think it can be counter-productive and they simply need to be given the tools to do the things they want to do without all the geekiness, but that geekiness is absolutely fascinating to those cognitive learners amongst us who want to learn the "whys and wherefores" of it all. And, after all, "he who knows better, does better" - these terms are from the psychology that has led us away from believing in the dominance model and towards using kinder methods. Those who get geeky with it - the likes of Ken Ramirez, Kathy Sdao, Bob Bailey et al - are the ones who aren't content to sit back with how we've always done it and are looking at ways to get the same results with more progressive methods. I'm sure they don't talk to their base-level clients in terms of stimuli etc, but they can talk to one another in those terms and therefore know with absolute clarity what they are talking about. The language isn't there to confuse people - although it can be confusing when you're first exposed to it - but actually to remove confusion. It's just a new subset of vocabulary we need to get used to. If we're that way inclined

I likened it the other day to being a ski teacher. Someone teaching beginners to ski doesn't need to understand the biomechanics of skiing, and they wouldn't be doing their clients any favours by approaching the lesson by talking about it. A good ski teacher simply needs a big toolbox of tactics that will help people to find their balance, what feels right, and allows them to build confidence in what they're doing.
But to be at the top of your game, you absolutely need to understand about biomechanics. It's the driving force behind everything we do on the snow, and leads the changes in the technology we use to ski.
You can be a skier without understanding any of it. You can be a good ski teacher without understanding much of it, but the great ski teachers will have a very keen understanding of the biomechanics, the physics, the technology, the psychology... every facet of the environment they are in. I won't ever be convinced that there is such a thing as "overthinking", or that we should suppress a desire to understand more about the subjects in which we are interested. For those people that aren't interested, there are tools and techniques that are tried and tested and wonderful for getting results. For those of us who are driven by a thirst for understanding, there is always so much more out there to learn, and that is truly magical!

 

[Stacia]

I must admit I speed read your answer to Joy as rushing OH to doctor and head not in gear as worrying about him. I have heard and read about the four quadrants, but could never get my head around positive and negative punishment. I can see that it is necessary to understand that but can one be an instinctive dog trainer?

 

[snowbunny]

I think if you’re only instinctive you will instinctively use punishment to a lesser or greater extent because our whole society is punitive and so that’s what we are used to.
I don’t think you need to understand all these terms to be a positive reinforcement trainer (if your own dogs). And I don’t think you need to understand all these terms to be someone who teaches basic skills to others. But I do believe that anyone who wants to act in a professional capacity with dogs should understand all of this, and it never hurts the rest of us who enjoy learning and want to be the best trainers we can be.

 

[MaccieD]

And I don’t think you need to understand all these terms to be someone who teaches basic skills to others. But I do believe that anyone who wants to act in a professional capacity with dogs should understand all of this,

If you are training others you ARE working in a professional capacity and SHOULD have a clear understanding of learning theories. Part of the problem in dog training is unqualified/untrained people advising/teaching others and their dogs. Although my first love us puppy training I spend much of my time on behaviour cases correcting and dispelling misinformation - sad but true.

 

[Stacia]

I don't think instinct necessarily involves punishment, but I suppose it depends on whose instinct!

 

[snowbunny]

I think most people instinctively want to correct unwanted or “bad” behaviour. It’s an ape thing, and also very much a cultural thing. We punish bad behaviour (in our human society) far more than we reward the good.

If you can manage to never think in terms of using corrections then you are well practiced in the techniques of R+, but I would be hugely surprised if that were wholly instinctive rather than learned. Even the biggest names in the R+ world talk about how they have to sometimes battle their inner monkey.