(In answer to a couple of comments: The title of the post is correct. I am addressing my dogs and asking if they want to come in. Sorry if it comes off as clunky,)
What do my dogs understand when I ask them a question?
A while back I read a suggestion that we should stop giving our dogs one-word verbal cues and start asking them questions instead. In full sentences.
Talking to our dogs is no biggie–most of us talk to our dogs all day, right? But doing so instead of carefully trained and clarified cues when we need a certain behavior? Several claims followed the suggestion. First, that if we ask our dogs verbal questions as prompts for behavior we are not actually giving them cues. Second, that dogs understand that when we ask them questions we are giving them choices. Finally, that asking dogs questions with the intention of giving them choices takes us outside of the realm of operant conditioning.
Whoa. I’m going to try to unravel this group of statements.
First, about the cues. Any change in the environment that can be sensed by an animal can function as a discriminative stimulus or cue. For example, a whistle from a human, a hand signal, a sound in the environment, or an odor can all be discriminative stimuli. A verbal question from a human is a series of sounds. It can be a discriminative stimulus, or can contain one. There’s no reason in the world to exempt questions from that definition.
This is Learning Theory 101. So far so good?
Second, about the choices. While dogs may come to understand that a questioning tone acts as a predictor of certain things, we do not know that they understand questions semantically as we do. I’ll expand on that below.
Finally, about learning theory. Both operant learning and respondent conditioning are going on all the time, whether we want them to be or not. Antecedents set the stage for behaviors. Consequences affect whether the behaviors increase or decrease. You can’t magically step away from antecedents and consequences by using a sentence with a particular inflection.
And it turns out that I might be just the right person to demonstrate this. I can demonstrate a few things about questions and choices because of how I communicate with my dogs. And I can show you a video of the results.
“Do You Want To Go Outside?”
Several years ago I realized my dogs didn’t always need to go out or come in from the back yard every time I did. I had adult dogs and knew their habits well. I went outside more often than any individual dog needed to for elimination, so I started giving them a choice. I had long had the habit of asking, “Do you want to go outside?” as my cue for that, although I wasn’t originally giving a choice. But over the years, I started to let a dog stay inside if she hung back and I knew going out wasn’t essential. Likewise, if I asked the dogs whether they were ready to come in and someone wanted to stay in the yard longer, I accommodated that.
Plenty of people do this. There are people out there casually giving their dogs choices about activities all the time and letting their dogs vote with their feet. Plenty of us look for ways to let our dogs make decisions about what they’d like to do and when. That’s been going on long before choice became a trendy word in dog training.
So How Does That Question Work?
But what about this “asking” thing? How did my question become a cue for “choice”? Did my dogs have some innate understanding of my words or inflection? Not necessarily. Science has started to show that dogs can detect inflections and familiar words–no surprise–but there are not yet data showing that they parse that information as humans do. (See a list of articles on dogs and human language here.)
But there was an obvious way for them to learn about those “question” cues. And that was through the consequences.
Cues in positive reinforcement training are opportunities for reinforcement. The thing we tend to forget is that using any cue is offering a choice, even when we’d rather not be doing so. There are decades’ worth of studies about reinforcement characteristics and the likelihood of a response from an animal.
When we want a dog to respond reliably to a cue, we use high value reinforcers on a dense schedule. We also limit access to reinforcers for the behaviors we don’t want. So what would we do if we want the animal to have a more of a choice? The opposite! We would use a weak-ish reinforcer ourselves so as not to stack the deck in our direction. We would make sure that there are reinforcers available for other choices. And we would not penalize the dog for making those other choices.
And that’s what I did. For example, when inviting the dogs to come in the house from the yard, I used low value reinforcers. Each dog who came in got a piece of kibble. Enough to sweeten the deal, but not so much that it overpowered the value of the other choice. Any dog could stay in the yard instead. There were naturally reinforcing activities in the yard, and all they’d miss would be one piece of kibble. (Which they could get later, although I don’t know if they thought about it that way.)
I deliberately paired a cue with weak reinforcement, and I didn’t intervene if they chose the “stay in the yard” option. I could have skipped the food reinforcer entirely. Coming in the house is often reinforcing by itself. But I want to reinforce my dogs–in some way–any time they come to me after I have cued it.
“Real” Recall Cues
If that weakly reinforced cue were the only way I had to call my dogs, I’d be in trouble. A strong recall is a safety issue. I also have a strongly reinforced recall cue for each of my dogs. These cues are designed and trained to a level that the dogs’ responses are reliable even in the face of large temptations. When I use those recall cues, I pay well. I bring out the meat and fish.
So again, for the choice recall I pay low. If they choose not to respond, I let them go about their business. For the “real” recall cue I pay high value, and keep it practiced so that their responses are reliable.
The differences in the dogs’ responses are a result of the quality of the reinforcement. They are not necessarily a result of questioning tone. It’s not some intrinsic quality of the antecedent. It’s the consequence that is tied to it.
Movie: Two Different Recall Cues
The movie shows what Zani does when she hears her heavily reinforced recall vs. what she does when she hears her “come if you want” cue.
Wait–What’s the Real Cue?
This is important. When I invite my dogs inside, is my verbal question the only cue? Nope. There’s something more salient than what I’m saying. It’s the fact that I am up on the porch, headed to the back door.
There are tons of cues like this in our dogs’ lives. If I am about to get my dogs’ supper, the time of day plus the fact that I’m walking to the kitchen show that. Other actions strengthen the message, including my gathering up the dogs’ food toys and getting out the food. It’s common for something other than our exact verbal cue to be the most noticeable cue for our dogs. My friend Debbie Jacobs has a great little movie where she is out with her dogs in the woods. Her dogs are all out of sight. She calls out, “Overhead slide projectors!” Her dogs come running. Her dogs are not responding to the specific words. It’s enough that they are out in the woods and she calls out some words, any words.
People who say that our dogs have advanced understanding of human language are making extraordinary claims. At the same time they are often ignoring dogs’ masterful powers of observation. Extraordinary claims require strong evidence. Stronger than saying, “When I ask my dog a question, she understands she has a choice, because this study showed that dogs have a language center in their brains just like humans.” It’s difficult to control variables enough to show evidence for this outside a laboratory. I’ll be discussing the Law of Parsimony and the idea of falsifiability of claims in a future post.
Often when we scratch the surface of a recommendation that sounds attractive–”Ask your dogs questions instead of ordering them around with robotic monosyllables!”–we find that the claimed results may not be as advertised. Or they may not be happening for the reasons cited. In response to the inevitable question: I talk to my dogs all day every day. But when I am actually trying to impart information to them, I try to be very clear. I believe that it is humane and loving to give clean, clear cues to dogs and not to overestimate their language capabilities.
The blanket recommendation of using complex conversational cues strikes me as odd for another reason. I currently have three dogs, and have trained another handful. Not a huge sample, but that’s part of the point. I have perceived big differences in how easily they learn verbal cues even within this small group. My little hound mix Zani has a really hard time learning verbal cues (though she’s great at inflections). I’m sure part of it is my poor training mechanics, but even so, there is a clear difference in how many repetitions Zani needs to learn a verbal cue compared to my dog Summer, for instance. And Zani is no slouch in the brains department. So I am concerned about wholesale recommendations to switch over to sentences, since for some dogs verbals are the thing they pay the least attention to.
Finally, I think focusing on dogs’ supposed understanding of human language is very human-centric. I’d rather put that energy into reading them better and learning their language. Dogs are already saying yes and no to things all the time, if only we would listen.
I’ve provided this handy list of the recent journal articles on this topic: Dogs and Language. If you want to comment about the findings, please quote the actual articles and not the blog posts or major media articles about them. Many have been sensationalistic.
Copyright 2016 Eileen Anderson