I have been thinking a lot recently about why dogs don't seem to generalise behaviours we teach them in a number of environments, and why it is so important to proof different behaviours. I encounter, as I'm sure many other people do, owners complaining that dogs are not performing a behaviour that the dog 'knows'. As a result, people often get angry and upset, claim the dog as 'stubborn' or 'independent' and make the behaviour into something it isn't, i.e. that the dog is 'intentionally' performing this behaviour because they don't respect you etc.

Whilst reading some stuff on cognition recently, I came across a few experiments into babies' abilities to identify different objects, or perform certain behaviours in different scenarios, which people may be interested in here. I'll give the links and descriptions of a few below, but in general, many studies have found that babies (around a year old) struggle to transfer behaviours into different contexts. In fact, their brains and bodies interact with the environment in such a close-knit, dynamical way, that behaviours are very context-dependent and need to be proofed in different environments to become reliable.

For example, a baby who encounters a steep slope when crawling will often plummet straight down it, without any fear. They soon learn to tackle slopes in more safer ways, however, and avoid really steep slopes. Nevertheless, once the babies' bodies are strong enough to walk, and they encounter these slopes again, they go through the same sequence of having no fear, and toppling down the slope awkwardly yet again, before learning to select and be more careful walking down different slopes. This seems to indicate that behaviours are very context-dependent, and babies learn about different environments in relation to their body positions and perceptions. This study can be found here (although I realise that not everyone has access! But if you do...):
JSTOR: An Error Occurred Setting Your User Cookie

There are many other studies supporting the view that learning is 'time-locked' or multimodal, where babies rely on the sensorimotor interactions between brain, body and environment to perform a behaviour. See:
http://informatics.indiana.edu/rocha...pers/smith.pdf

So, with dogs, when we teach behaviours in specific circumstances, those behaviours are time-locked in those moments- with the dog relying on their posture and their sensorimotor capacities for memory of that behaviour. Their memories are locked in the workings of the brain, body and environment as a whole, and when put in different situations, those memories are not easily transferred.

As soon as we change the scenario, e.g. ask for a 'Sit' when we are in a different position (standing slightly further away), or when the dog is in a different position, or we are in a different environment, the dog has trouble relating the behaviour to a number of different contexts. They really don't 'know' or have any memory of the behaviour we ask because learning it the first time was most probably in a different setting.

This really brings home the importance of proofing, for me, in the early stages and getting a dog used to many different experiences during socialisation. As in babies, the more they learn about different environments, the more a 'map' of behaviours is created with their sensory abilities informing each other. Thus, the more easily dogs would be able to transfer and generalise behaviours.

Also, it gives a valid reason for why dogs should not be called 'independent' or 'stubborn' because they are supposed to know a command. In reality, they don't because it hasn't been proofed or taught in different environments yet.

Great reminder and interesting explanations. I've always been fascinated by how poor dogs are at generalizing. Try teaching your dog a new, fairly easy trick, then change YOUR position (like sit down instead of standing up) and ask for the same thing - most dogs will look at you like you've lost your mind.

Interestingly, I've noticed that the more cues my dogs learn, the better they are at generalizing the verbal or signal cue to any situation. I've also noticed that operant dogs seem to generalize quicker than those taught though escape/avoidance techniques.

By the same token, though dogs are terrible at generalizing, the opposite is also true in that they pay SO much attention to context that it makes it very easy for them to adapt their behavior based on context. So when someone says to me I'm an idiot for letting my 80 pound therapy dog jump on me and mouth me when we're playing, I say, no, not really, because he understands context. When we're in a hospital setting, away from home, wearing certain gear, around a certain type of people with specific smells, movements etc., he knows the expectations are different. Same with dogs who perform in more than one venue, they get it...

I think sometimes we don't give dogs enough credit for the abilities they do have and blame them wrongly for the abilities they don't. Reminds me of one of my favorite quotes:
"Everyone is a genius. But if you judge a fish on its ability to climb a tree, it will live its whole life believing that it is stupid." ~Einstein

In relation to this I've noticed the presence of the clicker helps them. It's like my boy's training anchor if he gets confused. I bring it out and suddenly he knows what to do lol.

Definitely needing to fling me and Dino back into training and socialisation. I seem to have stopped my random training sessions in the train station and other random places.

Interesting thread!!

This is exactly what has been shown in babies. A lot of the behaviours they learn and the so called 'random' movements we see babies do, e.g. flinging their hands, putting things in their mouths, touching objects, has been shown to be very important in setting up their sensory systems to work together dynamically and efficiently, informing each other of information. The more they do this, the more they are able to generalise behaviours, so I'm sure it's very similar in dogs. Their 'beliefs' become more adaptable and applicable in different settings.

Yep, I think this is the most important issue the articles I have read emphasise. It's the ability for behaviour to be controlled by a range of sensory stimuli, so dogs learn to discriminate very well between contexts. This is the new way of looking at cognition, rather than some only-found-in-the-brain approach, i.e. cognition encompasses everything to do with a behaviour, including the physical environment it is performed in. The brain doesn't control everything we do, many behaviours are controlled mostly by our bodies or the physical environment.

I hate breed discrimination- not only BSL obviously, but the 'border collies are more clever than spaniels' discrimination too. Each dog is built in different physical formats, have different genetic make-ups and develop in distinct ways. If we are going to talk about intelligence, it cannot be to discriminate between dogs, it needs to be understood that intelligence is about the range of abilities one has, and their adaptability, not some marginalizing entity such as clever versus stupid.

YES!! Hannah Brannigan talks about how with every cue you teach, you are teaching not only a behavioral response, but also an emotional response. Its so true if you watch a dog perform too.
Really gives you pause when you think about creating super-reliable, strong behaviors. Those that the dog has formed a positive association with are absolutely going to be more reliable long term and in the face of distractions.