Do dogs generalise badly?

So we supposedly know that dogs are bad at generalising.

I recently asked a trainer "how do I generalise my clicker retrieve?" and got the answer "turn over the chairs in your kitchen". So this set me off on a journey...

Why are dogs bad at generalising? There aren't many people that have an answer (or, to be honest, are interested ;D ;D ;D ). But I've always been struck by how very little is written by experienced trainers on the process of proofing. The instructions are often limited to "generalise this properly" "go through a proofing process" and so on.

The clicker retrieve is absolutely classic. Wherever I look, the instructions end at a perfect retrieve in the kitchen. Who has actually generalised a clicker retrieve? Am I the only one who needed to do this? Surely not! And why did so very few people (actually none) think the clicker retrieve was the answer to my problems (bring back the dummy) if it didn't work "out in the field" once it worked in my kitchen?

I came across an interesting (although possibly not correct) answer. The reason dogs are bad at generalising is because they select irrelevant cues when we train. Could this be the case? In which case, varying potentially irrelevant things may improve matters. I wondered if this is actually what experienced trainers do and this accounts for the gap between how they say they train and the results?

So I have been carrying out a series of tests on Charlie to try to identify the irrelevant cues. It's interesting. He is obviously highly tuned in to items around my visual cues. So a hand signal in front of a bush, looks different to him if the branches of that bush are tied back for example. And he definitely marks on my movement towards my treat bag after the click. So if I'm not wearing a treat bag, his performance worsens.

This might be obvious to everyone else : ;D. Sorry. But it is useful to me. It changes the way I train in the garden - I'm introducing a huge range of small variations before I think I'm done in the garden. I'll be interested to see whether I have an improvement when I change locations....

 

Re: Do dogs generalise badly?

John Rogerson banned treat bags on our recall course for that very reason

I definitely think experienced trainers do a bunch of stuff they don't register and as a result that info doesn't get passed on.

For example you often hear about the value of different locations, this may address the backdrop question as well as many other factors but only if you think about the new locations in the right way. If you have a dog that does fly-bys you may end up in front of barriers a lot. When you step out in the open all bets could be off

 

Re: Do dogs generalise badly?

[quote author=bbrown link=topic=7888.msg110721#msg110721 date=1411072600]
If you have a dog that does fly-bys you may end up in front of barriers a lot. When you step out in the open all bets could be off
[/quote]

Exactly.

 

Re: Do dogs generalise badly?

Interesting stuff 8) (and not at all obvious - to me at least)

Will the testing be confined to your garden or will you try to implement it elsewhere too? Is it the size of variations that has the biggest impact or the number (would one huge old oak tree behind you be more of a challenge than 10 young saplings?)?

Perhaps you could write a mini article about it when the testing has finished? It's such a key part of training that I'm sure others would find it useful.

 

Re: Do dogs generalise badly?

I have some questions about this. Please excuse my ignorance though. :-\. I think it's very interesting but I don't think I know enough about it or understand it properly.

Taking recall as an example.

If I understand correctly a dog that has excellent recall in the garden might not have any recall if you took it out of the garden and into a field for example.

Is this because the dog doesn't understand the recall cue anymore, whistle or verbal, in a different environment or does the dogs recall need proofing against the field as if it's a distraction ?

At what point do you no longer have to do this though ? I'm assuming those on the forum who work their dogs and take them to working tests in fields they've never been to before don't have to quickly re train the recall. :. Does that mean at some point the dog learns to generalise that cue ?

Of course the real question which I doubt anybody can answer is why can't they generalise ? :-\

Sorry if I've misunderstood and asking daft questions. :

 

Re: Do dogs generalise badly?

[quote author=UncleBob link=topic=7888.msg110748#msg110748 date=1411077963]
Perhaps you could write a mini article about it when the testing has finished? It's such a key part of training that I'm sure others would find it useful.
[/quote]

I suspect that explaining this is quite beyond me - but I will keep my notes, of course. I am experimenting in a tiny, modest, way in order to understand.

The concept of irrelevant cues is put forward in "clicker revolution". Kay Lawrence claims not to need to proof...

But it is helping me, I think. I switched everything round in the garden (the placeboards, the cone, my chair) and found the cues I thought Charlie knew, completely fell apart. I hadn't changed distance, distractions, duration or difficulty. I just changed how the visual cues looked...and found out Charlie didn't understand them.

 

Re: Do dogs generalise badly?

[quote author=Jen link=topic=7888.msg111179#msg111179 date=1411157697]
Is this because the dog doesn't understand the recall cue anymore, whistle or verbal, in a different environment or does the dogs recall need proofing against the field as if it's a distraction ?
[/quote]

I don't know. I think that Charlie's cue to go to his left is my arm extended to my right. What he sees as the cue is a purple arm in front of a yellow bush in "a picture". So if I'm wearing a blue top, and switch to the other side of the garden, so there is a brown fence behind me, the "picture" Charlie sees is completely different - and it's a bit unreasonable to expect him to know it means the same as the other picture.

I fear that I'm probably rambling in a mess of amateur misunderstanding now and experienced people are rolling their eyes...

 

Re: Do dogs generalise badly?

Interesting. I think you're right that the problem is not actually 'dogs don't generalise' but rather, that dogs notice everything, visual, aural and, er, smelly, and don't prioritise the cues in the same way we humans do (we're obsessed with verbal communication!). So, in effect, the 'generalising' is about improving the 'signal/noise ratio' for the dog, so the poor love has some hope of working out what cue we floundering humans are trying to associate with what action!! Basically, it's a process of elimination for the dog : 'OK, that changed, but she still seems to want the same thing. Mmm, what is she doing that's staying the same?' Except, obviously without the internal monologue :

Most of the time, when I think my dog is doing stuff wrong or not 'getting it', she's actually way ahead of me. Couple of months ago I couldn't work out why she'd stay planted on dummy throws when I was going to go and pick it up myself (for steadiness practice) but would run in ahead of the command if I was actually going to send her. Mind reader? Probably not, I think I must have been raising my eyebrow or turning my head or hand in a particular way before I gave the command. I cured 'her problem' by making myself think 'mock throw' thoughts every time right up until I spoke ;D ;D I feel so humble sometimes - dogs are amazing

I do think proofing against distractions is a different issue though- sometimes other stuff is just more interesting than me :

 

Re: Do dogs generalise badly?

[quote author=Merla link=topic=7888.msg111223#msg111223 date=1411160884]
so the poor love has some hope of working out what cue we floundering humans are trying to associate with what action!! Basically, it's a process of elimination for the dog : 'OK, that changed, but she still seems to want the same thing. Mmm, what is she doing that's staying the same?' Except, obviously without the internal monologue :
[/quote]

Exactly. It helps to appreciate it, I think. Plus work out that the poor dog thinks it's the angle of your arm against the top of the fridge that's the thing... ;D ;D ;D

Need to think about distractions though....

 

Re: Do dogs generalise badly?

This is interesting. I am sure that at least some dogs generalise very well. We used to have a labrador who would come round to stay with us every weekend. At his own house, he was not allowed in the dining room. At our house, he was allowed in our dining room. Once, his owners came round to our house, and he was obviously and immediately confused about whether or not he was allowed into the dining room. I'm sure it wasn't a cue from his owner, she was at the other end of the house at the time... he must have had some general concept about "the room where they all go to eat food". I thought that was very impressive!
So it seems to me the issue must be more about picking up the "wrong" cues from us, like you are suggesting.

 

Re: Do dogs generalise badly?

Yes, I think so.

I suspect that more experienced people adjust for this. So, for example, I see that people make very exaggerated body movements at first on directional cues. This makes the visual cue much, much more obvious to the dog. But it's not like the dog couldn't see a less obvious cue, so I wonder if it's that by exaggerating the relevant cue against the irrelevant ones, this speeds things up (it seems, this morning, anyway).

 

Re: Do dogs generalise badly?

I think there's something in that obvious cue thing. I see experienced people with young dogs being extremely animated, vocal, encouraging etc whereas with their older dogs they're completely minimalist. If I asked them I think they'd talk about not taking anything for granted in terms of the dog understanding and performing the correct behaviour, making sure the dog knows what you want and not letting the dog go wrong.

 

Re: Do dogs generalise badly?

I think so. Right, back to the beginning for us this morning, with more obvious cues and changing the position of the target...

 

Re: Do dogs generalise badly?

In that case then are working dogs that can be taken to a shoot or field trial anywhere in the country more able to generalise or have they just been trained in so many different situations they've learnt to ignore the irrelevant cues ? ???

 

Re: Do dogs generalise badly?

I was suggesting that experienced trainers train in a way that includes fewer irrelevant things in the cues. So they get better results and don't experience what I do - surprisingly fragile cues. So they won't have to proof as much as I do (which sometimes seems to be to a degree that it's impractical).

If you are on facebook, there is a more advanced discussion of this on Clicker Revolution, and just starting on Clicker Gundogs.

I think the idea is that by changing what you think are irrelevant things, you make your cue training much more robust. You can do this by training in lots of different places, or you can stay in the same place and change the surroundings (obviously there are limitations to this). This is why I got the answer "turn over your kitchen chairs" when I asked how to generalise my clicker retrieve.

 

Re: Do dogs generalise badly?

So a friend who competes with her labs makes a point of taking her young dogs as many different places as possible and thinks about the kind of terrain the tests are likely to be on. I think when you've been competing for a while you get used to the locations that the clubs use for tests and trials. She also uses very exaggerated cues with her young dog additionally maintaining that cue so if the dog "refers back" she's still got her left arm out for example.

A really simple example is I haven't trained anywhere yet in Hampshire that has dry stone walls but if I look at some of the spaniel training photos online you see dogs much younger than Riley tackling walls with nonchalance.

You could say the retrieve is exactly the same. A straight line over an obstacle but Riley wouldn't look at it like that. The same is true of whether I'm set against the gentle green backdrop of rural Hampshire, the more open and scarred Salisbury Plain or an as yet untried grouse moor......

 

Re: Do dogs generalise badly?

It most definitely makes you think. ???

 

Re: Do dogs generalise badly?

The practical difference for me is that I have spent much, much longer working on changes - cones instead of placeboards, swapping things round, making things look different. What I know is that the behaviour is much weaker than I would have assumed. Before, I'd have tried to move on faster. Without changing so much, I'd think that the cues were stronger than they are. Another 2 days in the garden for me at least...

 

Re: Do dogs generalise badly?

Yes.

When I was recall training Tatze and trying to phase out treats on every recall, she knew which recalls I was going to treat on and which not, then became reluctant on the no-treat ones!

I had to minutely scrutinise my own body language to discover the cues she was picking up on - clever girl!

 

Re: Do dogs generalise badly?

It's like having ADHD

Us ADDers don't have problems focussing - we have problems focussing on too many things. Our senses work overtime - so that pencil on the floor has to be picked up before we can even begin to listen properly!