This post is part of the series in response to Dunbar’s 2012 Australian seminars. See index.

Dunbar has a clear opinion on the four quadrants of operant conditioning: Ditch them!  Dunbar feels we have entered into a time of ‘quadrant worship’ when, in reality, the quadrant was only ever designed to be a memory aid. The quadrants have also led to a division in the dog community, with half the people worshipping positive rewards and negative punishment (i.e. “positive trainers”), and the other half worshipping negative rewards and positive punishment.

Here’s a little theory:  In the quadrants, positive means “you give” and negative means “negate” or take away.

Dunbar used this table to illustrate the quadrants:

Dunbar thinks this is a complicated way of viewing things.  He says that the dog doesn’t assess anything other than “did the environment get better or worse?”  He believes dogs have a binary outlook to life.  They see things as good or bad.

Funnily, the Association of Pet Dog Trainers, an association that Dunbar helped form, uses the quadrants and advocates that, first, positive reinforcement should be used to change behaviour, second preference is to negative punishment, third preference is to negative reinforcement, and the last option is using positive punishment.

Dunbar believes that rewards are the first choice (positive reinforcement), like the APDT, but then thinks that mild positive punishment (redirecting and instructing) is the second port of call.  He finds his third choice, of negative reinforcement, “powerful when done right”, but potentially damaging to the dog.  He doesn’t find negative punishment useful, and is rarely used in dog training.

He called negative reinforcement (taking away something bad) and negative punishment (taking away something good) as ‘misunderstood’ and “difficult to do both in practice”.  The lines between different quadrants, especially when considering negative punishment and negative reinforcement, become blurry and dependent on other quadrants.

This puppy is receiving positive reinforcement – gaining the reward of tugging! When the tugging stops, then negative punishment will occur. Rarely does one part of the quadrant occur in isolation.

Negative punishment is problematic as, before you do the bad (remove something good), you have to have started the good.  This means that positive reinforcement (the introduction of something is good) is always tied with negative punishment (the removal of something good). He suggested that all good things in life – putting on a leash, going on a walk, getting in the car – are training opportunities and ‘life rewards’ for the dog.

Negative reinforcement is removal of something bad… But the question is, when did the bad start? (As it acted as a punishment when it started.)  Dunbar used the example of ear pinch method of training a retrieve (basically, a dog’s ear is pinched until it bites onto the dumbbell, and then the pinch is release when it takes the dumbbell).  This method is problematic as when the ear pinch is started, then it is associated with ‘something’.  In theory, if the dog took the dumbbell, and the trainer stopped pinching, then method is effective.  But if the dog just screams and the trainer gives up, then the dog has effectively been rewarded for screaming.  Or if the dog bites, and the pinch stops, then biting was rewarded.

Negative reinforcement training methods like this are clearly problematic.  Indeed, using pain for training is problematic according to Dunbar.  Painful stimuli can ruin a dog’s temperament and should never be used.  And negative reinforcement almost always uses pain, torture, stress and pressure.

Dunbar described two scenarios where he uses negative punishment (taking away something good).  In his off lead puppy classes, he sometimes will place a bully on a tether (removing them from the good fun of playing.)  Dunbar also said he uses negative punishment “when the dog starts to blow me off”, which suggests that he removes himself if the dog chooses to disengaged from training (though I am only assuming – Dunbar didn’t get into details about that).

Another criticism of the quadrants is the diagrammatic illustration of equality – the quadrants are not equal in terms of merit, their usefulness, or anything else. It’s just a diagram!

Basically, Dunbar praised dog trainers for ‘these days’ having a solid theoretical knowledge, but, for the most part, thinks it’s rather irrelevant in dog training.  To quote his handout, “nothing holds trainers back more than endless theoretical discussion of our iconic Quadrant.”  Our next post from the Dunbar series will consider his criticisms of learning theory.

Past relevant blog post: McGreevy’s Thoughts on Operant Conditioning

External link:  Roger Abrantes’ post on “Unveiling the myth of reinforcers and punishers”