The SCIENCE
REAL SCIENCE.
Not the junk "positive" pseudo-science
Relevant scientific articles
Positive reinforcement training better than normal training for recall and livestock chasing?!
The below study by Lucy China, Daniel Mills & Jonathan Cooper claims that positive reinforcement methods are more effective than normal balanced training or e-collar methods. It is simply another piece of ludicrous junk science.
The study involved dogs presented for livestock chasing and worrying in the UK. They were split into 3 groups (1/3 positive reinforcement, 1/3 normal training without e-collar, 1/3 normal training with e-collar), and then given 5 days of training. The training involved having sheep and chooks inside a tiny enclosure where they couldn’t run around, and the dogs outside the enclosure on lead, close to the trainers.
The dogs were never tested off lead or at a distance to their owners!
But Lucy China, Daniel Mills and Jonathan Cooper concluded that positive reinforcement got better results!
The first link below is to the study itself.
The second link is to Rebecca Sargisson and Ian McLean’s commentary on that study. The third link is to Prof. Doug Elliffe’s commentary on the study.
They conclude that, “China et al.’s results shed no light on the possible behavior of the dog off-lead or when the owner is absent“. They also state: “We have concerns about the statistical analyses… These statistical anomalies are not minor issues, as they potentially compromise or even reverse the key results in the paper.” And: “The problem for which the dogs were referred to trainers was not directly addressed, and the real-life context in which problem behaviors occur was not replicated in the study.”
In other words, just more biased, junk pseudo-science from Lucy China, Daniel Mills and Jonathan Cooper.
Efficacy of Dog Training With and Without Remote Electronic Collars vs. a Focus on Positive Reinforcement
– China, Mills & Cooper 2020
Commentary: Efficacy of Dog Training With and Without Remote Electronic Collars vs. a Focus on Positive Reinforcement
– Sargisson & McLean 2021
Commentary on China, L, Mills, D.S., & Cooper, J.J. (2020). Efficacy of dog training with and without electronic collars vs. a focus on positive reinforcement.
– Eliffe, 2020
Evaluation of an aversion based program designed to reduce predation of native birds by dogs: An analysis of training records for 1156 dogs
– Arnja Dale 2017
The acquisition and maintenance of dogs aversion responses to Kiwi training stimuli across time and locations
– Arnja Dale 2013
Behavioural differences and similarities between dog breeds: proposing an ecologically valid approach for canine behavioural research
– Pongracz & Dobos 2025
Comparison of the Efficacy and Welfare of Different Training Methods in Stopping Chasing Behavior in Dogs
– Anamarie Johnson & Clive Wynne 2024
Learned Helplessness
Reward-only trainers often claim that punishment will cause “Learned Helplessness”, and therefore you should never use any punishment (anything the dog doesn’t like). In fact, learned helplessness never occurs except with non-contingent and inescapable punishment – i.e. abuse. Note the title of the original study below: “inescapable shock”. These experiments were very cruel – no sane training ever involves inescapable, continuous, non-contingent abuse, but that’s what these experiments involved.
Note also, that learned helplessness can also occur with positive reinforcement. Reward-only advocate Susan Garett recognizes this when she notes: “I have seen many “shut down” dogs with handers that would never so much as raise their voices to their dogs.” Karen Pryor gives various examples of herself causing learned helplessness (and aggression) in the course of her positive reinforcement training. Also, the study below by O’Rourke agrees, and concludes: “The present results demonstrate that helplessness can be induced using positive reinforcement.”
So as an argument against the use of punishment, the argument that it causes “learned helplessness” is completely fallacious.
Two of the original studies are below. The third is O’Rourkes showing that positive reinforcement can cause learned helplessness.
There are also Martin Seligman’s books such as “Learned Optimism” (which is quite a good read – see Books to read).
The fourth study (Learned Helplessness – Now you see it, now you don’t) casts some doubt on the whole notion of learned helplessness.
The fifth study (Learned Helplessness 50 years on) reveals Seligman and Maier’s new insights – they now believe Learned Helplessness is not a real phenomenon – it should actually be called Learned Control – animals don’t learn to be helpless, they learn to CONTROL their environment through aversives.
Learned Helplessness – Effects of inescapable shock upon subsequent escape and avoidance responding
– Overmier & Seligman 1967
“Fallout”
One of the common arguments from the reward-only ideologues is that punishment always causes “fallout”. For example, reward-only trainer Susan Garett (like so many others) claims:
“there is always fallout with any form of punishment – even something as benign as a time out.”
She is wrong. Completely and utterly.
There is NO negative long term “fallout” with a sound use of punishment, whether “time outs” or any other form. You only have to consider that fact that dogs use punishment with each other all the time – a growl means “No”, and if that isn’t heeded a bite will follow. For example a bitch weaning her pups (see video below), or an old dog chewing a bone when a pup gets too close. Or when a pup gets accidently stood on. There is no terrible “fallout” – the pup just learns.
Many of these myths stem from some of Skinner’s ideological leanings, backed up by some of his poor experiments (which have since been refuted). They have also come largely from Murray Sidman’s influential (but mostly erroneous) book, “Coercion and its Fallout”.
The two articles below deal with some of the problems with Sidman’s work:
Misc
More to come…
Other Junk Pseudo-Science
So much of the so-called “science” the reward-only ideologues use in support of their position is, quite simply, junk.
Many people would be surprised to learn that often these “studies” are nothing more than small sample size surveys – questionnaires of the general dog-owning public. Anyone who has ever noticed political polls getting the results completely wrong (and these generally have far larger sample sizes) can understand how useless these are.
And a careful analysis shows that the conclusions often don’t match the results, and simply show the author’s bias. As Professor John Staddon points out (see Books to read ), they are often just efforts in confirmation bias, and activism. Then other studies cite these junk pseudo-science surveys as if they are fact, and a whole house of cards is built on nothing. It is time to bring the reward-only house of cards crashing down.
American Veterinary Society of Behavior Position Statement
A great example of relying on junk pseudo-science.
The least inhibitive, functionally effective (LIFE) model: a new framework for ethical animal training practices.
– Eduardo Fernandez, 2023
A failed attempt by a reward-only ideologue at the University of Adelaide’s School of Animal and Veterinary Sciences to improve on Lindsay’s LIMA model (which has itself been hijacked by the reward-only ideologues).
More to come…
Trust THE SCIENCE?
There is just so much junk, even fraudulent, pseudo-science out there, that we must be very discerning about what we do, and don’t, believe. Dr. Peter Ridd gives a good overview.