Wednesday, 19 September 2012

Now where’s my treat?

Trainers often advise owners to use treats to train their dogs, but some owners want to phase them out as fast as they can. Shouldn’t a dog be prepared to work for just verbal praise and affection? That’s the question asked in a recent study by Erica Feuerbacher and Clive Wynne – and they didn’t just test dogs, but wolves too!

A brindle boxer is happy to receive a treat
Photo: LarsTuchel / Shutterstock

The question is interesting for practical reasons, since it’s useful to know how to motivate a dog if you want to train one. But it’s a very interesting question for another reason too. Some scientists have suggested that dogs are uniquely tuned in to human contact; in other words, that in the process of evolving from wolves, dogs have developed special abilities to read human emotions and communication. If this is the case, then social contact with humans should be a valuable reward in training sessions with dogs, but not wolves.

The study involved five separate experiments, four with dogs and one with hand-reared wolves. Two different types of dog were included: those who lived in homes with their owners, and those in a shelter. It might be expected that shelter dogs, starved of the usual amount of human contact, would respond especially well to social interaction as a reward. On the other hand, if a relationship with someone is needed for that interaction to be valuable, dogs with owners would respond more.

The experiments all used the same task: a simple nose-touch to the hand. In the food condition, the dogs and wolves were rewarded with a small piece of food. For the dogs, it was a piece of Natural Balance (except for one dog with allergies, who was given a small piece of potato instead). For the wolves, it was a small piece of sausage, because this is what the wolf trainers recommended.

In the social condition, dogs were rewarded with 4 seconds of social contact – petting either side of the head combined with verbal praise. (One of the wolves did not like physical touch, so he just received praise). This is only a short time of social interaction, but the length of time it took coincided with the length of time taken to give a treat.

If the dog or wolf touched the experimenter’s hand, the hand was removed to shoulder height and then the reward was given (food from the other hand or the social interaction). For the owned dogs, the owner carried out the experiment, and for the wolves at Wolf Park, a trainer did the experiment in each wolf’s pen, with another trainer present for safety reasons.

The results showed that across all three types of animal – shelter dogs, owned dogs, and wolves – food was a better reinforcer than social interaction. Although there were individual differences between animals, the use of social interaction as a reward did not lead to many nose touches. On the other hand, when food was used as a reward, many more nose touches were recorded, and the time interval between them was shorter. The wolves did more nose touches than the dogs, and in fact the best performing wolf produced 33% more nose touches than the best performing dog. It’s not known if this is because of a wolf’s greater intelligence, or because they had some prior training that was useful (but some of the owned dogs had also had relevant prior training from their owners in their normal lives).

The results are fascinating. Instead of suggesting that dogs have developed special abilities to understand humans, it seems that our special relationship might come about because of ongoing reinforcement. The authors say, “…domestication has not necessarily resulted in dogs being more sensitive to human behavior than wolves. The comparison with wolves confirmed that the relative effectiveness of social interaction for hand-reared wolves was the same as for dogs.”

Of course, only a brief period of social interaction was used in this study, and it might be that longer periods would work better. If this were the case, though, I would have expected the dogs to nudge the experimenter to ask for more fuss. And while the food reward for the wolves was sausage, which might be a better motivator than the treat given to the dogs, it’s still the case that social interaction did not really motivate the wolves or dogs. So next time someone says they want to phase out treats in dog training, this study provides evidence that it would be a good idea to stick to the treats after all.

From your dog’s point of view, what’s the best reward in a training session? For my dogs, sausage is definitely a favourite.

Reference
Feuerbacher EN, & Wynne CD (2012). Relative efficacy of human social interaction and food as reinforcers for domestic dogs and hand-reared wolves. Journal of the experimental analysis of behavior, 98 (1), 105-29 PMID: 22851794

If you enjoyed this, you might also like: How do hand-reared wolves and dogs interact with humans? (April 2013)

42 comments:

  1. This was interesting, but I've already seen that my dogs are highly food motivated. What I'm trying to figure out, is how when transitioning over to praise instead of treats, you can keep dogs motivated. Because in the competition obedience ring, food is not allowed...

    ReplyDelete
    Replies
    1. It's not the goal to phase out treats totally. If you don't want to give treats all the time, you can give treats on an intermittent base. Sometime you do, some you do not give a treat. It's more surprisingly/ challenging for an experienced dog and in that way you can keep your dog higly motivated. This study provides evidence that a treat is a strong motivator but there are also a lots of individual differences. An interesting and very actual topic on the field!

      Delete
    2. As a psychology major, I definitely second this comment. The established term in this field is Variable Ratio (VR) Reinforcement, and is the explanation as to why games of chance can be so addictive. A great write-up on B.F. Skinner's work can be found on good ol' Wikipedia (just don't reference it for scholastic work.) :)

      http://en.wikipedia.org/wiki/Reinforcement

      Delete
    3. Yup. Intermittent rewards seem to be more effective, although my dog knows when a treat is due and won't let me off the hook. Fussy little Yorkie. p.s. the sitter thinks a bark means he wants to be picked up, not a treat for good behaviour, so she picks him up all the time. I think he's probably a percentage hitter, like young men in my youth, who know they'll get a treat eventually. lol

      Delete
  2. Thank you both for your kind and interesting comments. Deborah is right that a variable reinforcement schedule, in which sometimes you don't give a treat, works better than rewarding every single time. Another thing is that people often want to phase out treats too quickly, and end up with a grumpy dog!

    This study didn't look at different reinforcement schedules. In order to compare the food treat with social interaction, everything else had to be the same.

    Collie222, good luck in the obedience ring! I hope you and your dog have lots of fun.

    ReplyDelete
  3. My 9mth old wolfdog begs to differ...

    ReplyDelete
  4. The wolf would be more motivated for food because domestic animals just have it given to them, the wolf must work. Therefore, when food is so easy they'll repeat the steps necessary to get it, especially when compared to something that isn't necessary for life--which attention/affection is not required to live.

    ReplyDelete
    Replies
    1. Thanks for your comment. I didn't include it in my blog post, but the full article gives information about when the dogs and wolves had been fed relative to the time of the experiment. The wolves in this study were captive wolves.

      Delete
  5. I never use treats to train my cats, but they respond just fine. This is weird to learn that they at least seem to me more sociable than dogs. Maybe cats just love a chin or back scratching more than dogs.... ? They sure do love chin scratching.

    ReplyDelete
    Replies
    1. Maybe a future study can test cats too! That would be interesting. It's nice to know that you train your cats, since some people (wrongly) say that it's not possible. Thanks for your comment.

      Delete
  6. Fascinating article. I would love to be a part of something like this. From what I've concluded through training my dogs, food is the way to go. My dachshund would definitely prefer sausage.

    ReplyDelete
    Replies
    1. If you're in Florida, my understanding is that the Canine Cognition Lab is recruiting participants for other studies. You can read more here: http://www.psych.ufl.edu/~caninecognition/Participate.html.

      If you're in another part of the world, it's always possible that a university near you has a canine behaviour lab that needs more dogs to participate.

      Delete
  7. This is silly, how can you base a study on 5 dogs, not to mention the dog being fed potato, this can be very dangerous even fatal in high amounts. A dog requires structure, leader and you should be in control of all it's resources. It seems harsh but they respect you for it. I'm not saying starve your dog or restrict them.No dog responds just to food you just affection. It's being a fair reliable leader.

    ReplyDelete
    Replies
    1. (sorry, put the wrong url and couldn't edit it)

      Just to correct a misunderstanding, it was five experiments, not five dogs. And I am sure they did not harm any of the animals! But thank you for your concern.

      Dogs can eat cooked potato and it is an ingredient in some commercially available dog foods. According to the ASPCA (http://www.aspca.org/Pet-care/ask-the-expert/ask-the-expert-poison-control/potatoes some raw potato is fine, so long as it isn't green. Just as humans shouldn't eat green potato, nor should dogs.

      Delete
    2. The concept of 'respect' is something an animal is unlikely to be able to comprehend, this is a high level cognitive reasoning that if a dog were able to understand would suggest they have the same level of conciousness and intelligence as humans. Dogs respond well to consistency as they are able to learn rules through consequences. Yes you can be your dog's 'leader' if you like but you can also train them to make their own choices. I prefer to think of dog ownership as a partnership where the dog's needs are also recognised and met. It is important to train dogs using positive reinforcement to show them what you do want rather than punishing them when they do things we don't want but never showing them what you actually want them to do. Using food can be an excellent way of doing this and there is no reason not to do so, it will only imporve your relationship with and control of your dog. If you use some of your dog's usual meal for training and make them work for it you also avoid adding additional food to the diet which could cause weight gain. There really is no downside to using food in training if done correctly.

      Tamsin Peachey
      MSc Clinical Animal Behaviour
      Director
      BestBehaviour Services Ltd

      Delete
  8. I don't see the conclusions as being valid. Just because a dog enjoys treats more than being pet does not indicate that the dog has less, more, or the same ability to "read" feelings. they're mutually exclusive, but possibly related, variables.

    it's also unlikely that a dog's ability to read emotions is at all genetic. dogs aren't, and rarely have ever been, bred for their pet-like behavior. they're bred for color, size, nose sensitivity, aggressiveness, intelligence, etc. not for their ability to be best friends. and those that are bred that way, like service dogs (primarily seeing eye dogs), might prefer the petting.

    since it is VERY unlikely that there is a genetic ability to read human emotions, it only leads that it's a learned event. All learned events are due to reinforcement. it's easily known through anecdotes. if a dog spends every day for a while with anyone, the dog will learn all the habits. how the person smells, laughs, smiles, etc. it knows when something is different. it's like if you had a picture in your room for 10 years and all of the sudden it was gone. you recognized something small was different, and not normal, and you want to rectify it.

    also the test variables seem sketchy. dogs have different tastes. maybe they would prefer one treat to a petting over another. some dogs like to be pet in different areas and are indifferent to others.

    these conclusions are akin to "blind guys don't jump as high as people who can see".

    maybe there's more in the journal article that would quell my issues. i don't know. but as a wildlife biologist, i guess i am left unhappy.

    ReplyDelete
    Replies
    1. Thanks for your thorough comments, which touch on wider issues as well as the research mentioned in this post.

      You are right that it would be interesting to compare different treats, and that different dogs might prefer different treats. It could be a question for future research. For this study, it says they took advice from trainers (dog trainers and the wolf trainers) as to what would be an appropriate treat, typical of what is used in training.

      The treats and the petting had to be, as far as possible, the same across all the animals and all the trials.

      Of course we can't make assumptions about genetics from this study, but many people would have expected the dogs to respond more to social interaction, because somehow through evolution they have come to like being around people, instead of fearful like wolves.

      It must be fascinating to work as a wildlife biologist!

      Delete
  9. Freeze dried liver, jerky, or doggy sausage.

    ReplyDelete
  10. Very interesting article, but there's one thing I can't really piece together from my experience with my dogs. Often times, they will act out behavior that they know has never resulted in treats, but always results in love and affection...which has clearly become a trained habit of theirs.

    Our little girl will lick your arm, softly bark (almost more like talking than barking), roll on her back and wave her paws at you, and many other things that generate physical affection from us. At no point in her life though has she ever been given food from those actions though.

    So clearly, she has actually been "trained" over time and had these behaviors reinforced...with the only motivator being that she gets pats and hugs usually.

    Our older male dog is similar. Put his head in your lap, stands on his hind legs and stretches on you, various things that generate affection, but never food.

    I feel that maybe better variables need to be tested in the research:

    1. Amount of praise and type of praise.
    2. Relationship with the tester...I'm sure a dog wouldn't desire as much physical attention from someone relatively new to them as they would with their humans of many years. I know the owner was used for the dogs, but maybe you need a similar relationship for the people testing the wolves.
    3. More standardized food treats. My dogs would definitely go more nuts over the pieces of sausages being given to the wolves, than the treats or the potatoes.

    ReplyDelete
    Replies
    1. Thanks for your comments. You make some interesting points. Of course, one study can't answer everything at once, or it would get too complicated. That's why scientists are always saying 'more research...'. I think some of your comments would make good ideas for future research.

      It was only the shelter dogs that were tested by someone they didn't have a long-standing relationship with. The other dogs were tested by their owner, and the wolves were tested by a trainer who regularly worked with them (with another trainer present for safety reasons, who did not take part in the study).

      One of my dogs is like your girl, he has a little routine he goes through when he wants affection :)

      Delete
    2. It seems to me that it's not strange at all that your dogs are reinforced for some things by affection and for some things by food; it's that way with all of us. I am reinforced by money for some things, but by respect and affection for other things. Thing is, with dogs, we're teaching them obedience exercises and expecting praise to be an adequate motivator, and that is a mismatch. Reinforcing social behaviors with social reinforcers is sensible. Reinforcing precision heeling, or recalling during play, or something even more difficult, with something social, is less sensible, and less effective.

      Knowing what your dog wants to work for is the first step to good dog training, and it is only the arrogant trainer that decides ahead of time what you will be giving out. You should always control and give out what your dog wants most, if you are trying to change behavior in any lasting way. In a given dog, perhaps it is really affection that works best, but I'm never going to assume it and make that dog work for only affection when the task is demanding and there are competing activities he'd also like doing. The right reinforcer for the job is important, as is matching the reinforcer to the situation. I don't reinforce social behaviors with treats, and I don't reinforce non-social behaviors with pets.

      Delete
  11. Yes and No. The study in question is based on a rather poor experimental design and analysis. (1) It doesn't use any statistical techniques to control for other explanatory variables (e.g. varying intelligence and sociability measures of each canine) (2) Sample size is way too small for any statistically significant results (3) They don't look for interaction effects (e.g. does praise PLUS food have a synergistic effect in dogs that might not appear in wolves?)

    On a personal experience level, older puppy mill dogs I've trained over the year would often forgo food and water to be with their human. How would this study explain that?

    ReplyDelete
    Replies
    1. Somehow, my dog knows when I am getting up (I'm in bed a lot), and when I am at a break in using the computer. He'll jump right up then to get my attention. Otherwise, he entertains himself.

      Delete
    2. Thanks for your comments. The full paper gives a lot of information about the dogs, including things like how long the shelter dogs had been in the shelter, what breed they were, and the reason they were there. One of the things I like about this paper is the amount of detail given about all of the studies; someone else could replicate the study, if they wanted, but also it takes account of the fact that the animals were all tested in a real-life setting, not in the lab.

      Maybe a future study could look at treats plus praise. They didn't investigate it here (as I said above, it's not possible to include everything in one set of experiments, as things get unwieldy).

      I'm really interested in your comment about former puppy mill dogs. Kudos to you for the training you have done with them. You are right that they would make another interesting group of dogs to test, since they will likely have had very little normal human contact during their lives.

      Delete
  12. Where do you work? I am a college student majoring in Psychology and my heart is with animal behavior.

    ReplyDelete
    Replies
    1. Psychology is a fascinating subject, isn't it? Congratulations on taking it for your major. The study reported here was done at the University of Florida, so if you are thinking about grad studies, you could look at them?

      Delete
  13. Rewarding a dog by patting it on the head is not a reward--even friendly dogs don't usually like that. Tolerate, yes. Enjoy? No. Saying this proves that they don't like human interaction is like assuming a dog is deaf because it's not responding to your verbal command.

    Many working dog trainers use toy rewards, where the reward is interaction with the handler. That actually produces more results for most dogs than just treats.

    ReplyDelete
    Replies
    1. Luckily they didn't pat the dogs on the top of the head. You are right that many dogs wouldn't like that. They report that it was around the neck.

      Yes, many working dogs are trained using toys; also for agility, a game of tug at the end is commonly used. This study didn't look at toys, but it could be an interesting idea for future work. Thanks for your comments!

      Delete
  14. The experiment is interesting, as are some of the replies. I am part of a neuropsychology research team at Emory University that is using fMRI to answer many of the same questions pondered by the UF behavioral psychologists (Feuerbacher and Wynne)cited in the article. As with many studies, the initial research may raise more questions than it provides answers. It appears that the questions posed in the Feuerbacher/Wynne study were: 1) "As a primary reinforcer, which do participating canids prefer between human social interaction in the form of tactile contact in combination with praise versus a food reward (Natural Balance/potato, sausage)?" and 2) "Is there a difference amongst domestic pet dogs, domestic shelter dogs, and captive wolves regarding which primary reinforcer they prefer?" In essence, the experiment asked, "Which drive is more pronounced, the drive for human social contact or the drive for food?" However, since human social contact was inherent to each repetition, regardless of whether the dog received petting/praise or food as an ultimate reward, the experiment did not accurately examine whether the dogs preferred human social contact or food. A human experimenter was present for each repetition. In addition, the behavior of initiating tactile contact with the experimenter was required for the dog to receive either petting/praise or the food reward.

    Therefore, to more accurately determine whether the dog/wolf prefers human social contact or food, the experimenter should have been placed out of site and the participating dogs/wolves should have been taught to touch nose or paw to both of two pads, one that would bring a human inside the pen to pet/praise and possibly play and the other that would cause a pellet of food to drop into the pen. Then, the experimenters could record which pad the dogs/wolves depressed most frequently. Alternatively, the experimenters could have one pad and two sub-groups within each primary group of pet dogs, shelter dogs, and wolves, one sub-group that received only human contact after performing the behavior and one group that received only food and then the experimenters would record the behavioral frequency of each sub-group.

    Consequently, regardless of intentions, the UF experiment actually asked, "When a dog or wolf already experiences human social contact, which does it then prefer between an elevated magnitude of contact or a food reward?" Nevertheless, the experiment is highly relevant to pet training situations, as the experiment effectively supports the use of food-based positive reinforcement when teaching new behaviors and dispels the beliefs of trainers that remain resistant to incorporating food. In addition, the experiment is interesting in questioning the extent that selective breeding has advanced the co-evolutionary process between domestic dogs and humans.

    For those interested in learning more about our work at Emory, feel free to view a press release and video (http://esciencecommons.blogspot.com/2012/05/what-is-your-dog-thinking-brain-scans.html) and our journal article published in PLoS One (http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0038027).

    ReplyDelete
    Replies
    1. Thank you for your detailed and thoughtful comments. I'm especially intrigued by your ideas for a study with the experimenter not present. One of the things I find so fascinating about the behavioral sciences is the tension between what makes a perfect experiment and the fact that people and animals can behave differently in the real world than the lab. I found your ideas really interesting.

      As you say, one of the strengths of this particular study is that it is closely related to pet training situations. That's useful because I hear so many people (and some trainers) saying that they don't like to use treats. Thanks for taking the time to share your thoughts about the advantages and disadvantages of this kind of design.

      Many thanks, too, for the link to the research at Emory. I will check it out!

      Delete
  15. This might be a bit off topic, but my dog exhibits some strange behaviors that I can only assume are indicative of social preference over food.

    For example, she'll only usually eat once we're done eating, even if the food is there for her. Also, if I feed her right before leaving for work, and my girlfriend isn't home, she won't eat until we come back. I'm assuming it must be some kind of evolutionary leftovers from being in a pack, where she will not eat until the pack leader is present and has already eaten.

    I also have an indoor-only cat, who loves bringing socks into the living room, dropping them on the floor and meowing at me, as if to say "Look! I brought you something!"

    Ahh pets.

    ReplyDelete
    Replies
    1. It sounds like your dog does not like to eat when alone. Some dogs (and cats) are like that. Without more information, it's hard to say why, but I don't think it's to do with the 'pack leader'. One of the great things about all the research going on into dogs at the moment is that we understand them better than ever (even if there's still a lot to learn). Many of the older ideas about packs and being pack leader don't stand up to the current evidence.

      Having said that, I think it does show that your dog misses you when you're not there. I wonder why she prefers to wait for you to finish eating? Maybe she likes to be with you while you eat? It sounds like she is very sociable with you. It is interesting to think about how behaviours have come about. There is still a lot to learn.

      I love your story about your indoor-only cat. It's fantastic! Thank you for sharing. While an outdoors cat might bring you a present of a dead bird or mouse, your indoor cat brings you socks instead! That's lovely.

      Delete
  16. My dog is definitely food-oriented, and actually loves small pieces of cooked carrot - this makes me feel a little better, as she is still a pup and we do lots of training. I hope I'm not contributing too much to the puppy fat this way! :)

    ReplyDelete
  17. Hi Jenny,
    That's interesting. My dogs like carrot too, but I hadn't thought of using it in training! Pumpkin is another favourite.

    ReplyDelete
  18. While you have a small puppy, use part of the dog's daily food allowance (kibble of course is easier than raw!) as training treats. Therefore your dog won't get overweight. Every mealtime is a training opportunity! Some dogs are just not food motivated and the use of a "special" toy that the dog only gets to play with as a training reward can work just as well if he is toy motivated.

    ReplyDelete
  19. It seems to me a huge design flaw to use extremely low value food (low protein kibble and plain potato? really? ) for dogs and extremely high value food for the wolves. My dogs would work a lot longer for sausage than kibble and 2/3 of my dogs wouldn't eat a piece of plain potato, certainly not repeatedly )

    ReplyDelete
    Replies
    1. The Natural Balance food reward used for the dogs is actually a kind of baloney-style product commonly used for training purposes, not kibble. I would have to assume that the dog being fed the potato as reward must actually like it well enough to want to work for it, if it was being considered a reward for them.

      Delete
  20. What is most interesting about this article is the initial premise: "Some scientists have suggested that dogs are uniquely tuned in to human contact; in other words, that in the process of evolving from wolves, dogs have developed special abilities to read human emotions and communication. If this is the case, then social contact with humans should be a valuable reward in training sessions with dogs, but not wolves." I actually think that in the process of evolving from wolves (who are hunters) dogs (who became opportunistic scavengers) have developed special abilities to grab food that was either dropped or tossed by the humans they were evolving alongside. This is my little theory as to why food is such a great motivator for dogs. It's been so from the very beginning of our relationship with them. Pretty difficult to undo aeons of evolutionary hard-wiring.

    ReplyDelete
  21. This article has many valid merits With each of these articles we read, we tend to look over at our own dogs to determine if the article holds any weight. The mechanics of this one does, although the specifics don't. Dogs place their own 'value' on each reward, and will literally jump through hoops until they get it. I feel I'm blessed to have 5 dogs that will ignore food if a disc is present. Merely holding that disc enables new skills to be taught. Companion animal psychology isn't my area of expertise, but with five being constantly observed, one can't help but notice certain trends and tendencies, and yes, they most certainly are in tune and have those mentioned communication abilities.

    ReplyDelete
  22. Would be interested to know how motivated the dogs or wolves would be to participate in the training if the social contact reward used was interaction through play instead of just pat/verbal praise? Would it increase the percentage of repetitions of touch if a play style that was engaging for the animal, maybe involving an object like a ball or stick or tug toy, was used?

    ReplyDelete
  23. "The concept of 'respect' is something an animal is unlikely to be able to comprehend, this is a high level cognitive reasoning that if a dog were able to understand would suggest they have the same level of conciousness and intelligence as humans - See more at: http://www.companionanimalpsychology.com/2012/09/now-wheres-my-treat.html#sthash.7yYKo8x0.dpuf"

    I would be careful saying what is likely a dog would understand or do. Franz deWaal has already shown that dogs and other animals show what he calls a sense of "fairness" which is also a high level cognitive ability. Hare, Kaminski, Miklosi and others are all experimenting to find out more of just what dogs can and do "have" in terms of emotions. They are of course comparing these to human models, so there may or may not be direct correlations. We just don't know.

    So saying that this is not so in terms of "respect" is just as unfounded as saying dogs need respect, need to show respect or need to receive respect, know what respect is or don't know what respect is. We simply don't know.

    What dogs DO know is, what is in their best interests. Getting choked, pronged, shocked is not i their best interest. Getting food, positive attention, and the primary reinforcers is. Given the choice of what I'd rather do to and/or with my dog, I'd rather pay her for her work with positive reinforcers than with a military hand salute.

    ReplyDelete