Reward-Based Training Versus Punishment: Which is Better?

Last updated on:

In this article, we’re exploring a scientific paper that was published in 2004. We’ll try to break down the article into plain English to help you understand what the researchers had to say.

Dog training methods: their use, effectiveness, and interaction with behavior and welfare
EF Hiby, NJ Rooney, and JWS Bradshaw
Animal Welfare 2004, 13: 63-69


The authors set out to lay a foundation for their paper in the abstract. This is where they start out with some basic facts and assumptions:

  • Dogs used to be trained mostly with punishment-based training, but positive reinforcement is becoming more popular.
  • Punishment includes scolding, leash corrections, alpha rolls, or anything else that scares or hurts a dog.
    Punishment’s goal is to reduce a behavior.
  • Positive reinforcement includes giving your dog treats, toys, praise, or something else she likes.
    Reinforcement aims to increase a behavior.

The researchers asked 364 people about their training methods.

They wanted to see if training methods related to pet behavior or training effectiveness. They found:

  • 66% of people used vocal punishment
  • 12% used physical punishment
  • 60% used praise
  • 51% used food rewards
  • 11% used toy or play rewards

The researchers asked the owners to rate their dog’s obedience on eight different tasks.

They found that owners who used positive reinforcement rated their dog’s obedience higher. Trainers who used punishment to train rated their dog’s compliance lower.

The researchers asked the owners about 16 problem behaviors. Owners who used punishment reported more common behavior problems in their dogs. Dogs trained using punishment might behave worse because they’re more anxious.

The bottom line: punishment correlated with worse obedience and more problem behaviors. Positive training methods might be more useful to pet owners.


For the remainder of the paper, I will just pull out the highlights, as I see them. I won’t go through the entire paper sentence-by-sentence. That would be insane!

Training methods are a welfare concern.

Traditional training is based on aversion/punishment. Past research indicated that punishment-based training has many bad side effects for dogs, including:

  • Causing suffering (Beerda et al., 1997)
  • Health risks through increased stress
  • Aggression towards other dogs (Roll & Unshelm, 1997)

This paper aims to investigate the training methods of UK dog owners. It also examines the effectiveness of each training method.

People will only adopt positive training methods if they’re at least as effective as punishment based training techniques.

The bottom line: the goal of this study was to see what training methods were used by the owners, and how effective these training methods were.


The researchers stopped talked to a bunch of dog owners and gave them surveys. They asked that owners fill out surveys for their youngest dog that was over 1 year old.

The surveys focused on four main areas:

  • Demographics. This included the age, sex, breed of the dog and information on the owners, among other things.
  • Training Methods. How was the dog toilet trained? How did the owner react when the dog chewed on something? How did the owner react if the dog stole their food or objects? How did they teach their dog to come, sit, drop it/leave it, and heel?
  • Obedience. Owners rated their dog’s obedience on a scale of 1 to 5 for each of the seven tasks above. 5 was most obedient. They also scored their dog’s “overall obedience.”
  • Problem behaviors. Owners examined a list 16 common behaviors their dog displayed. They then indicated if their dog displayed these problems in the past, currently, or never.

The researchers threw out a bunch of surveys that were useless due to missing data or other human errors. Then they ran a lot of statistics and did some math. They categorized people as punishment-only, reward-only, combination, or miscellaneous. They did some more math in order to explore the relationship between training method and number of behavioral problems.

The bottom line: researchers distributed a survey to a bunch of people with dogs. The survey asked about problem behaviors and training methods in dogs. The researchers then did a bunch of math to see how training methods related to problem behaviors.


This is where the researchers show us what they found. It’s a lot of numbers.

In many cases, different training methods were used for different things by the same people. For example, many owners used punishment when teaching dogs not to chew but rewards when teaching a dog to come.

  • 20.2% of owners used only reward-based training
  • 9.8% used only punishment
  • 60.4% used a combination
  • 9.6% used “miscellaneous” methods or did not mention their methods.

The maximum score of “obedience” was 40. The average dog in this study scored 33.

  • People who rewarded their dogs more often also scored their dogs higher. The researchers didn’t find a relationship between punishment frequency and obedience scores.

The researchers then found the average obedience score of each group of dogs in relation to the training methods used. The training methods fell into this order:
1) Reward-based only
2) Combination reward/punishment
3) Punishment only
4) Miscellaneous training methods

In other words, reward-based trainers rated their dogs higher than combination trainers, who rated their dogs higher than punishment-based trainers. But it turns out punishment-based training might give you better compliance than no methods at all!

The researchers then looked deeper into specific tasks that dogs were trained to do. They found:

  • Dogs trained using play specifically as a reward were much more likely to give up or leave items.
  • Dogs trained using rewards were most obedient while heeling.
  • Dogs given a different item to chew on were least likely to chew on inappropriate items. (For example, giving your dog a chew toy to encourage him not to chew on a table).

97.2% of dogs showed at least one problematic behavior. The most common problem behaviors were over-excitement, barking, and fearful behaviors.

Dogs that were punished more displayed more problematic behaviors. Owners who used reward-based training reported the fewest over-excitement problems.

Owners who used punishment, either alone or in combination, reported the most separation-based problems.

The bottom line: the researchers found that dogs trained using reward-only were also rated highest for obedience. They found that the more rewarded, the better-behaved. They also found that dogs that were punished displayed the most problem behaviors.


This is where the researchers talk about all the numbers found above.

There’s still a general belief that punishment is the best technique. Reward-based training outperformed punishment in several tasks and punishment is not the most effective training method.

Owners who used punishment did not report higher compliance in the 7 tasks studied.

In fact, owners who rewarded their dogs the most had the highest overall reported obedience. In other words, the more owners rewarded their dogs, the better their dogs behaved.

Dogs trained using only reward-based training were more obedient than dogs trained using punishment (alone or in combination).

Dogs trained using more rewards were more obedient than dogs who were rewarded less often.

Other studies have found that dogs that are more obedient are more likely to satisfy their owners. Satisfied owners, in turn, are less likely to give their dogs up to a shelter. Therefore, training methods and their success have an impact on whether or not dogs end up in animal shelters.

The more a dog was punished, the more it displayed problem behaviors.

The authors admit that perhaps well-behaved dogs “rewarded” their owners for using reward-based training. The emphasized that regardless, punishment clearly does not improve obedience in dogs compared to reward-based training.

Past research linked punishment-based training and dog-dog aggression. Punishment-based training might increase anxiety, which causes more separation-related stress and over-excitement in dogs.

The authors say: “Punishment-based training is not effective at reducing the incidence of problem behaviors, and its use seems to be linked with the increased occurrence of potential problems.”

The bottom line: The authors admit that punishment can be used effectively in dog training, as shown by years and years of training. Most owners are inexperienced trainers who likely are administering punishment incorrectly. This causes anxiety and stress for the dogs. “For the general dog-owning population, reward-based training methods may produce a more balanced and obedient animal.”

Conclusions and Welfare Benefits

  1. Dogs trained using punishment are no more obedient than dogs trained using other methods
  2. Dogs trained using punishment are more likely to have problematic behaviors
  3. Dogs with problem behaviors often are more anxious
  4. Dogs with problem behaviors are more likely to be given up to a shelter

Because reward-based training resulted in better obedience and fewer problem behaviors, they’re a better option for the average dog owner!

That’s that for our first-ever dive into science-based dog training. I hope that my “translation” helped you better understand some of the science surrounding my reward-based training techniques. Still lost? Let me know! 

4 thoughts on “Reward-Based Training Versus Punishment: Which is Better?”

  1. While I agree with the conclusions, I don’t think the study proves them. The basic issue being that correlation is not causation. Sure, good behavior correlates with rewards and bad behavior correlates with punishment. What’s not demonstrated is which is the cause and which is the effect. One could equally well conclude, “In fact, owners whose dogs had the highest overall reported obedience rewarded their dogs the most. In other words, the better their dogs behaved the more their owners rewarded them.”

    Causation could be studied by changing the training pattern and watching for a change in behavior. If rewards stop, does good behavior stop with it? If punishment stops, does good behavior start? Etc. Demonstrating causation would be more convincing than correlation.

    • Oh yes, this is certainly just correlation. It’s not the only paper to find this correlation, though – and such a body of correlations suggests a very strong link indeed. Pairing that with welfare concerns, especially when undereducated owners attempt to apply a given technique, is definitely compelling to me.


Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Join up with our pack

Get tons of great dog training tips and tricks you won't find anywhere else, along with exclusive deals and discounts only for pack members.