The following study was sent to me recently:
http://seab.envmed.rochester.edu/jeab/articles/2012/jeab-98-01-0105.pdf
The question being evaluated was the relative value of food or social interaction as a reinforcer for dogs being asked to perform a simple task (nose touch).
To look at this question, fourteen shelter dogs were evaluated – they were offered opportunities to work for either food or personal interaction (4 seconds of petting/praise) and their responses were recorded. In addition, four “pet” dogs had the same test run with their owners, to help clarify if the results were a factor of whether or not the dog’s personal relationship with the trainer was important.
It comes as no surprise to me that the dogs performed much better when food was the reinforcer rather than four seconds of personal interaction.
So far so good.
Here’s where things start to go awry: One of the conclusions was that: “Social interaction functioned as a low-value reinforcer.” Further, for the one dog that worked just as hard for praise/petting as for food, one of their possible explanations was that the dog may have been “conditioned” (by the pairing of food with praise prior to the shelter), to value praise.
The good news is that they did point out that individual dogs might have different responses to the value of praise.
Why does this research disturb me?
First, because four seconds of petting/praising is NOT social interaction. It is four seconds of petting/praising. That is not genuine. We do not praise dogs (nor people) in that manner. When my dogs do something well, I do not praise/pet for a predetermined amount of time and then cut it off, asking for more work. How can a dog possibly perceive that as genuine interaction? Here’s another way to ask the question: Was the researcher genuinely pleased with the dog, or simply going through the motions as dictated by the research design? My gut tells me it’s the latter; remember, scientists, are expected to be detached.
Praise is GENUINE; it only works if you mean it! If you offer it to train your dog as part of an experiment then yes, it is likely to fail. Some dogs will take what they can get – any praise is good praise. But most dogs accept praise from those they care about – when they believe it is real. Really, it’s not so different from people.
Second, the value of praise is built over time with a specific person and style of interaction. If I want my dogs to work for praise, I will have to practice it – not by pairing it with food, but by finding the type of praise that works for that particular dog. Like humans, dogs are individuals – unlike eating, social interaction is not a one size fits all proposition. Each of my dogs works best for a different type of personal interaction, and none of them would work for praise from a stranger. It took me time to find those interactions, and with the youngest dog I’m still actively searching. This doesn’t mean it’s not real, but it does mean that it’s not nearly as obvious as food to use in training.
Is it possible that the four pet dogs that also struggled to work well for praise were distressed by the loss of food as a motivator rather than the choice to use praise as an option? The control of the four pet dogs only works if they didn’t have a history of food based training because now we have a punishment situation…where’s my cookie? You can’t just switch motivators and expect it to be fine.
If you did this study with a toy (ball or tug) instead of food, you would almost certainly get the same results. Most dogs do not value toys as much as food, unless you make an effort to build the value of toys. As a rule, my young dogs prefer food to toys, but after I work on developing their love of toys, they will, without exception, choose the toys over the food. Indeed, if I bring out food they may act like they are being punished. Note that I never paired the toys with the food; in my mind that kills the dog’s natural interest in toys. I simply use the toys in a way that brings out a dog’s natural interest in playing with objects – the underlying primary motivator of pursue, grab and fight.
Is this preference for toys over food the same for all dogs? Of course not. Dogs will have innate interests and tendencies, just like people! I’m sure there are dogs that care a lot about food and relatively little about personal interaction. And there are other dogs that are highly motivated by personal interaction or toys, but not as much by food. It is noteworthy that the breeds of dogs most often selected for AKC performance events are also the breeds of dogs that seem to take most easily to toys and personal interaction as effective reinforcers.
I have no argument with the researcher’s conclusion that food is the reinforcer of choice for training most dogs – it is certainly the easiest for the average person to use, it is effective with a high percentage of dogs, and you can get in a whole lot of repetitions in a short time. But that conclusion does NOT support the second conclusion that the researcher came to; that genuine praise is a relatively low-value reinforcer, any more than it would be correct to say that toys are a relatively low-value reinforcer simply because most dogs will initially work better for food than toys.
I worry that these kinds of conclusions devalue how we view dogs. When dogs are subjects, controlled by their desire to eat and fundamentally uninterested in humans, then we give them “object” status. Objects have no innate value beyond whatever monetary or sentimental value we may attribute to them. The odds that people will do ugly things to dogs go up when they are objects. When you view dogs as selfish takers rather than as partners, you undermine the innate value of dogs and people as a team – the middle that grows between them. It’s not so different than the years when Skinner’s behaviorism ruled child rearing – the children’s need for cuddling and love was scoffed at since only food, warmth and a clean diaper made the radar. Some of those babies died. Apparently, no one could quantify the need for love, so they concluded that it did not exist.
When I sit by a whelping box and watch four-week old puppies wag their tails and smile up at me, I’m hard-pressed to say they don’t feel good in my presence. At that age no one has given them food, so they cannot associate food with humans; the puppies simply enjoy my company. For sure there are other explanations about why they choose to interact with me, but the one that “feels” right is that they have an innate interest in humans. No, I’ve never studied it. I’ve simply lived with dogs and puppies, watched them carefully, and played with them for thirty years.
I’ve never considered myself anti-science, so why a rant?
Because science forms the backbone of much that I believe about dogs – that training is best when dogs are working towards positives rather than away from aversives. I believe that because it is my experience, and it happens to be backed up by research. I want to believe that science can come to genuinely accurate conclusions that I can use when making my own decisions about how to proceed in a given circumstance with a specific dog.
I find it hugely damaging to human/canine relations to suggest that we are a secondary reinforcer; that dogs only interact with us because they are hoping for food. My guess is that the purpose of this research was to support the use of food in training – that is good. Unfortunately, I know that many people will interpret it to mean that dogs and people cannot have mutually enjoyable social interactions – that is bad.
If you are a scientist, then for one minute, stop with the science and sit with a few dogs. Pet those dogs. Watch them interact with each other and with you – no food involved. After you do this, tell me that the only reason your dog stays with you is that you have been conditioned as a secondary reinforcer.
This reminds of of someone, commenting, once… on how we should all start wearing the same hair-styles and clothes, because that’s how alien races are on Star Trek… aren’t all dogs the same too? 😉 Of course not!
A lady at work, the other day, said someone said to her that he thought are not capable of thinking… hmm… I said, go to any dog sport trial sometime, and you’ll see those dogs thinking. It’s sad that people do not understand.
There is good science and there is bad science. I’m a scientist, I agree with you. Bullshit.
I’ve mentioned before my previous dog was a Samoyed that loved to work. She taught me that when training she preferred ball playing as a reward to eating her breakfast. I never paired ball playing with food. In fact the day I finally figured out what she preferred was when she walked away from her breakfast (raw) in my hand to dig a tennis ball out of my training bag. She wanted the relationship not just food.
Interesting observations but not science…and not research. Dogs from a shelter with no “relationship” with the testers and probably starved are probably going to act better for food. Some don’t even LIKE “praise” from strangers. How many people would prefer ANYTHING to being rubbed on the head or hugged by a complete stranger in a contrived situation. I think it is testament to the huge value of social involvement of ANY kind to dogs that they even tolerated the experiment in the first place. JMO
I am a scientist… AND I actually took my dog (who is not terribly food motivated) to participate in some experiments they were doing at Harvard…. many of which were to evaluate the studies they were planning on doing, and collect some preliminary data and tweak the studies so that they could get meaningful data. The most notable one was a variation on the “dog follows pointing finger to find treat” experiment. So they warmed the dog up by having the grad student point at the bowl… dog gets treat… dog expects praise and gets it from me, but not the grad student (who was trying to not interact with the dog, and kept a very still face the whole time). The second time my dog got the treat, sniffed the grad student, came back to me, but was definitely wary about the game, the third time, she walked up to the grad student, looked right at her face, turned her back on her and lay down. It was hilarious. The graduate students and I had a long long discussion about the true meaning of their experiments based on their experimental design, some of the tweaks they might want to make in their choice of dogs with and without prior training, and the signs of stress in a dog (Mine didn’t do any more experiments, as she was pretty stressed to be in a room where people were actively ignoring her with stern expressions, which is interesting on its own as she is not particularly an attention-seeking dog). They had interacted with lots of dogs, but I don’t think they understood just how closely dogs respond to subtle body language, much less glaring body language. I hope I gave them a lot of good information that they can use to inform their future studies
My interpretation is that like most animal studies, it is a broad guideline. I’ve worked in a big chain dog training Co. for almost 5 years and I see hundreds of different dogs a week with their owners.
I think that accomplished performance dog people are not the norm and most people do not interact with dogs the way they do.
I have people all the time come to me and some of the first things out of their mouths are “I don’t want to use food, my last dog was trained by love.” For me this is something I can direct those people towards. Learning to play properly and interact meaningfully with your dog so that it becomes a primary reinforcer is something that is learned for the human, if you are going to a big chain Co. (which MANY dog owners do) then you aren’t trying to learn to play with your dog in any other way than for it to blow off some steam. I can talk until I am blue in the face, I can demo my own dogs, it doesn’t matter. They want the most direct result from A to B with the minimum amount of work.
Private clients, that is a little different, but think of the vast majority of dog owners, this is a great article for them.
I get where you are coming from, but why not tell people what is probably the truth? Their last dog was likely trained with a combination of compulsion and praise; you recommend a combination of food and praise. Most people would rather train with kindness if it is presented that way. Telling people their dog doesn’t care about their praise is likely to have negative ramifications for how they view their dogs in the future.
I think “doesn’t care about their praise” is a little harsh. I didn’t leave with that taste in my mouth, I left with, isn’t as effective to get clear, quick and lasting results. I didn’t think they were making a point about an overall relationship with a dog. Obviously there are those of us that spend everyday all day conscious of that and other’s whose needs aren’t the same as ours. I think it was more sterile, the training, more specifically 4 secs of “good dog, pat, pat” and its efficacy in a controlled environment, as pointed out not very realistic, but thus are the challenges of dogs and the studies of dogs :).
I had all the same gut reactions you did, Denise! Thanks for posting. This “study” wasn’t close to “science.”
This is an interesting post. I am still learning a lot about playing with my dog, but I know she likes playing with me and would rather do a lot of play activities over food for a reward.
Well with a quick skim of the article. The first problem I noticed in the study was they cited physiological dog studies which conclude that increases in good hormones, that create good feelings, happened after 5 to 23 minutes of social interaction (see page 2). But, in their own study they only had social interaction for 4 seconds. Which according to all those studies they cited, wasn’t nearly long enough social interaction to have a pleasing affect, physiologically, for the dogs.
The writing of this is akward, since they make a giant intellectual leap from nose-touching exercises to “why dogs interact with people at all”. They start out by posing this very broad question, but then proceed to find some answers through nose touching exercises. In the world of science this might be fine, but for those involved personally with dogs it’s of little value.
Personally, I would never have average dog owners read something as tedious as this. I don’t think making someone read a study like this will endear them to positive training methods. I would think it might make them throw up their hands in exhasperation instead.
I remember reading a book (in German) by a German scientist a decade or two ago that purported to debunk a lot of the “lassie come home” type myths about dogs. It was a well-written book that tried to explain why dogs were acting on instinct/drive rather than out of an emotional basis.
What I remember most about the book however was the tongue-in-check epilogue — in which the researcher (PhD biologist or zoologist, don’t remember exactly) said that none of his foregoing observations about dogs applied to dachshunds. He himself had a dachshund and they were smart and loyal, etc., and part of the family, no matter what. I found it really interesting that this author, after laying out a straight “scientific” case, was able to say that his emotional reality in living with his dog made him question his scientific findings — or at least accept the ambiguity between his emotional and his intellectual reality.
THis is what makes you such a great trainer, balancing science with psychological or emotional needs (both human and canine). I think a fine example of this in my own dogs was when I adopted my 2nd dog (the Siberian, Juno) for my GSD mix Loki. For a long time at the beginning, I thought Juno was anti-human social, not interested in play and even had low food motivation in the beginning. We used high tasting treats at first to garner a relationship. But let me tell you, human anti-social was the last thing from this dog’s mind, no matter how much she tried to ignore or run away from us. Over time, as I worked both dogs independently and together as a team, in at home games (and games YOU taught me and Loki), she would do HER BEST “I’m am so cute, leave that GSD alone and get over here and give me some attention NOW” behaviors (even if i had no food). So if i were casually working Loki (not a formal session or anything like that) or even just giving him a pet, she would roll over, show her belly, let her tongue hang out, she’d do her “sit pretty,” play bow, whatever it took for me to leave Loki and go to her. Yes, indeed, she became a “hussy for human attention.” (at least from me) I think that’s probably the ultimate you can ask for any dog. One who really wants to be with you and is maybe even a little jealous but acts in a socially acceptable way on that jealousy (versus being pushy or naughty).
RE: the comment above about the “Big Chain Co.” I look at it as a variety of people own dogs for a variety of reasons and have a limit on how much time/energy they really want to invest. For you and me (and most on this list) watching and interacting with dogs and dog behavior is fascinating and rewarding psychologically for us. People’s lives are so busy with work and family’s and their own children, they have little focus for observing –and i mean REALLY observing–dog behavior and communication, let alone the energy to put in to getting the dog bonded in such a way quickly. I spend so much of my time with both my dogs, they really are a team with me and I haven’t fixed every behavioral issue with Loki, I know exactly why and what makes him behave as he does. If he were with a person with half the time I give him and half the interest, I’m not so sure he’d do as well as his is now. For the first few years, even with the time I gave him, I was still perplexed by his behavior. After 4 years, lots of time investment and actual INTEREST, I feel I get him and our relationship is such a strong partnership now.
Some people just want a dog for very superficial reasons and even if the time was availble, they have little interest in knowing more than teaching in a logical (scientific) way. This is why I never really feel so strongly about one method of teaching over another. Because I really look at it as “working with the dog in front of you” (what works best for that dog) with a balance of what work best for the handler too (how much time he/she can invest and what type of partnership s/he looks for with that dog). Hopefully there’s something that fits both dog and handler in a healthy way.
I should have mentioned that I have a degree in zoology and behavioral sciences and worked as a zoo keeper for an AZA accredited zoo for four years and still have many close friends in the biz, but started working with dogs so that one day I might actually pay off my student loans. That is my animal training foundation and they hammer in that absolutely no anthropomorphism. I always find it interesting in the dog world the idea of dogs having emotion. I was always taught that emotions in animals is a dangerous concept, and in exotics more specifically. I was taught that it is a spectrum of stress, not all stress being bad stress (in the wild obviously they experience stressors frequently). I approach clients like a zoo keeper and solve problems like a zoo keeper to this day.
I agree though, my dogs deff seek me out for affections, but again I spend all day everyday aware of my relationship with them. However, I see every day dogs blow off their owner’s affections out in public….and I think it is a disservice to not expose your average dog owner to science…I have no agenda, just to accomplish whatever task I am being paid to do for the individual dog/human team.
Im not trying to start anything, had read the article and was looking for a forum discussing it. It was an interesting study, I thought, as I think most are. Most animal experiments have flaws, it is hard to control all factors of animal and human behavior, again the exciting challenges of working with animals and people!
I would love to discuss this further but I think this is the wrong forum. If you add me as a friend on facebook, we could start a thread there – I know that my friends are heavy in both trainers and scientists. I don’t know if people will behave, but heck….it might be worth a try.
This is a very interesting discussion. If you haven’t already you might take a look at Patricia McConnell’s blog as well; she frequently holds discussions on this topic of dogs and emotions as well as discussing the latest research.
I susprised you to write that the concept of emotions is dangerous. In fact as far back as 25 years ago the 1986 Home Office code of practice and the Animal Welfare Act (1987) mandated not only physical well being but also their psychological welfare.
Yes and that is why providing animals with enrichments is probably a zoo keeper’s most important roll. However it is framed as a continuum of stress not specific emotions. A zoo keeper’s job is to familiarize themselves with their animals and notice subtle changes in their behaviors. The conversation with the curator would go “I believe the great horned owl is stressed because I have observed A, B and C.” Not, “I think the great horned owl is sad because…” Not attributing human emotions does not negate empathy and wellbeing, I think quite the opposite actually it forces you to try your darndest to think like said animal and figure out what in its environment is not working towards a low stress life without our own human emotions clouding ALL possible combinations of contributing factors, psychological and physiological.
I share your on again off again relationship to science when it is brought up in dog training conversations. Great dog training stands outside of science at many times.
If it can’t be readily and reliably reproduced by another researcher, it is not science. The skill and acumen of the researcher cannot be taken into account or it is not science. I look at dog training as an artform based on scientific principles of learning.
Dogs work for many reinforcers, just as humans do. The dogs are the ones that need to value the reinforcement, if we hope to influence their responses to our requests, and they are the only arbiters of what’s interesting to them. If I paid you in peanut hulls versus dollars, how long would you come to work? While you can build interest in toys for many dogs, there are some for whom they will never hold as much interest as a piece of steak. Other dogs will work for a tug or a Frisbee all day long and refuse food for quite a long time so long as the preferred “thing” is available. What really bothers me as a trainer and, hopefully, a student of dog behavior, is the pervasive desire of some trainers to require that dogs work for reinforcements that are not valuable to the dogs, then punish them when the “positive reinforcement doesn’t work.” I’ve used “chasing the squirrel” as a reinforcement for dogs that like to chase (“Come” and “sit” then you get to chase!). Premack is very powerful. I find that some studies are more designed to foster a point of view than to actually create an unbiased study. That’s why I like to look at data, not the extra verbiage in the conclusions, i.e. the “might be because…” statements.
I’m a post-grad student working on dog behaviour, cognition and emotion… Honestly, I am yet to meet a dog scientist who isn’t a wealth of information on dog behaviour and thoroughly passionate about all things dog. It is a big commitment to put 7+ years into tertiary education. I for one took a 60% pay cut to start my last 3-ish years. Many dog scientists cop it from dog owners and there is frequently controversy, often misplaced. You don’t walk into that without a deep wonder and heartfelt appreciation of everything a dog is and isn’t, and an all-abiding passion for their welfare and the human-dog bond. Often I hear comments from dog enthusiasts that suggest dog researchers don’t actually know anything about dogs, or pointing out the myriad flaws in what they have termed ‘not science’. It is tiresome and oftentimes offensive. Researching dog behaviour is quite hard. Most dogs come from their own unique environment, and there are countless breeds and breed mixes, and the list of factors that we know can affect behaviour is long and I don’t think anyone believes we have an exhaustive list yet, or even that we understand any one of those factors very well. We are all aware of these problems, but our problem is how to capture that in data. We can’t publish it if we don’t have the data. And behavioural data is notoriously complicated and variable, which means you need a lot of it to be able to find any significant patterns. And then you have to figure out how to analyse it. The statistics I’m using are intensely complicated and not really powerful enough. I’m attempting to teach myself an even more complicated method of analysis because I badly want to lay down the foundations to start capturing some of the richness in dog behaviour that we all know exists but can’t do anything with. And it will all be for nothing if I can’t get dog owners everywhere to contribute to the data collection. Data is currency, here. I can’t access enough dogs on my own to make a difference.
So in defence of dog scientists, yes, we do love dogs. Yes, we realise there are a lot of factors we can’t capture in data at this point. No, that does not mean we should not try to make sense of what we can capture. Science is not friend or foe. It is what you make of it. You can produce the most thoroughly, carefully defined piece of work and people out there will take it the wrong way and misquote you. It’s something we just have to live with.
I couldn’t agree with you more. It’s insulting and it usually comes from people who have never done any science.
This really hit home with me, Denise. I agree with everything you ‘ranted’ about. I sometimes feel like my dog only loves food or toys, but not me. I get so frustrated with her and myself, and think where the heck did I go wrong? I just need to find something that makes her energy explode and really get her going when training. She’s 8.5 years old, and I’m still searching for something that ticks her when we train. She does like to chase toys, but I need to act crazy an get her really going. Any suggestions on how to get her to find value in our training sessions?
Okay I am a scientist and a dog owner and a learning trainer. There are so many problems with the study as you have presented it. First, the researchers presumably knew nothing about the history of the 14 shelter dogs. Those dogs could have had 1) good experiences with humans, 2) bad experiences with humans, 3) no experiences with humans except once sheltered. When we talk about a social species, like the dog, context is extremely important. So if you had a dog that had a bad experience with humans, then yes, certainly food would be a higher motivator than human interaction (that’s a “duh” moment). Same would be true for a dog that was basically feral. Only those dogs that had had good experiences with humans *might* value human interaction and then that would largely be dependent on the type of interaction.
Seond, let’s look at the 4 pet dogs. Again, think of the range of human-dog pet relationships. These can be “I have a dog to protect my house and yard and just be a dog” — which is basically no real relationship. Could be “I have a dog and I pay attention to it when I have time and energy, but mostly it just hangs out.” Both of those types of pets are likely to find food a helluvah lot more rewarding than interactions with their owners. But for those of us who play with our dogs, train our dogs, play some more with our dogs, cuddle our dogs, and treat our dogs like a real entity with feelings and needs, well those dogs might actually prefer human interaction to food. And even then, if the dog was hungry, that might change the context too. My dogs love interacting with me because we play shaping games, tug games, training games and even wrestle. But by god if it is dinnertime, then they want food, not me (well, only me inasmuch as I deliver the food). Did the study bother to standardize hunger levels or to remove the issue of hunger and timing of food delivery? Without reading the study, one cannot say how any of these factors were standardized or even if they were taken into consideration — and if they were not, then this study is just junk science, since social behavior is always contextual in nature.