Like many dog owners, I talk to my dog more or less constantly. “Florence,” I’ll say, as we’re walking through the park, “are you happy it’s almost fall?” (Neither of us likes the heat.) “Nope!” I’ll tell her as I drag her away from the same discarded chicken bone she’s been angling after for the past two days. “It’s never going to happen so you may as well just stop trying.” I don’t really give it much thought; mostly I just talk to her because it makes her think I’m paying attention to her. And every morning I give her what I call her daily affirmation: “Florence,” I tell her, looking her straight in the eyes, “you’re a good dog, and I love you.”
Florence is not a good dog.
Florence is an 80-pound coonhound-cattle dog mix. She’s moody, needy, distrustful of strangers, bad with other dogs, and often willfully disobedient. If Florence were a person, she’d be the friend who eats off of everyone else’s plate at the restaurant, throws in half of what she owes when the bill comes, and then looks around the table defiantly, daring someone to call her on it.
But I tell Florence she’s a good dog (demonstrably false) and that I love her (true!) because I suppose that I’m subconsciously hoping that if she’s hears it enough, it’ll sink in—that she’ll think, “Hey, maybe I am a good dog,” and start acting like one. It doesn’t make much sense, but such is the emotional self-deception of a bad-dog owner.
So when my Twitter feed recently filled up with headlines alerting me that researchers had found that dogs do indeed understand us when we talk to them, I felt pretty gratified. “Your Dog Knows Exactly What You’re Saying,” read a headline on National Geographic. “Science Confirms What Your Heart Knows: Dogs Understand What You Say to Them,” said People. I love it when my heart can say, “I told you so” to my brain.
Then I actually looked at the study.
Oops.
The study was conducted by the Family Dog Project, a research team at Hungary’s Eotvos Lorand University. Here’s what they actually did: They selected 13 dogs to put into an MRI machine and played a recording of a trainer talking to each dog. There were two variables in what the trainer said—the words (positive or neutral) and the tone of voice (positive or neutral). All the while, they scanned the dogs’ brains to see which parts lit up with activity in response to which words. As it turned out, when the trainer gave the dogs positive words, like clever and well done (apparently things people say to their dogs in Hungary), the scans showed more activity in the left hemisphere of their brains than when the dogs heard neutral words like such and yet, regardless of which tone of voice was used. Over in the right hemisphere, a known auditory region activated differently depending on whether the dogs heard praising or neutral tones (the neutral tones actually resulted in more activity—that particular brain region is more sensitive to lower pitches). But it was only when the dogs heard words of praise in a praising tone that their primary reward centers—the dopamine-pumping region of the brain that tells you when something good is happening—start going crazy.
It’s no secret that dogs can be conditioned to recognize certain words, and these dogs had been conditioned to hear the “positive” words as praise. But the researchers think their study, which was published in the Sept. 2 issue of Science, is significant for a few reasons: First, the dogs’ brains responded to the praise words even when they were spoken in a neutral tone, which they think suggests that dogs can analyze what the researchers call “lexical cues”—the words themselves—separately from emotional cues (tone) and can then integrate the two (that’s when the reward jackpot goes off). Second, the left-brain/right-brain pattern the researchers detected mirrors what goes on in the human brain—word processing is lateralized to the left side of the brain while the right brain reads emotional and social cues like tone. In other words, they conclude, the way that dogs process language is similar to the way we do. And that could in turn have implications for how language evolved in humans.
So does that mean dogs understand language?
Well, to answer that we have to keep in mind what a functional MRI, or fMRI, actually is—a way to detect what part of the brain experiences an increase in oxygenated blood flow in a given circumstance. There’s been a boom of studies in the past few years that involve performing fMRIs on dogs. That’s because a team of researchers at Emory University recently realized they could train some animals to willingly lie in an MRI machine for long enough to get a scan (now those are good dogs). Since then, there’s been an avalanche of news about what dogs do and do not understand or respond to, much of it reaffirming what we dog owners already knew to be true in our hearts. Just last month a study by the Emory team used fMRI scans to demonstrate that some dogs are more motivated by praise than by food rewards. (“Our dogs really do love us!” shouted Today.)
Exaggerated but heartwarming headlines aside, there’s still a limit to how much we can actually learn about how dogs’ minds work from an fMRI scan. It’s a pretty big leap to take a finding of “this cue resulted in increased oxygen flow to a particular part of the brain” and deduce that our dogs “love” or “understand” us, or even that they’re processing language in the same way we do. Sure, fMRI scans can detect activation patterns, but they can’t tell you what those patterns say about how the brain functions. It’s “one of the great weak spots of our field,” says David Poeppel, a psychologist and cognitive neuroscientist at NYU. “How do you connect a pattern of activation, like a map of blood flow, to function? And here’s the answer: We have no clue.” And in the case of the dog study, says Poeppel, “local patterns of activation are simply not sufficient” to know how the dogs are processing the sounds.
Even some dog fMRI enthusiasts have been skeptical of the results of the Hungarian study. Gregory Berns, who leads the Dog Project at Emory and wrote a whole book about what dog fMRIs can teach us, told Discover’s Neuroskeptic blog that it’s not clear from the study that the dogs’ left-hemisphere activation has anything to do with lexical processing. He pointed to a previous study from the Hungarian team that found that certain areas in the dogs’ left hemisphere lit up in response to decidedly nonlexical vocalizations by other dogs. Maybe it’s just that a sound that grabs the dog’s attention (such as a familiar praise word) results in more activity than one that doesn’t.
Which all points to a bigger problem: When we ask whether dogs “understand” us, we are actually asking a very complicated question about speech and language. To be fair, the Hungarian paper didn’t use the word understand—that was the media—but it did use the terminology of linguistics (like lexical representations) to describe what was going on and alluded to the potential implications for our understanding of human language. All of which is irksome for someone like Poeppel, whose lab at NYU studies language and speech in humans. (It’s worth noting that Poeppel isn’t an entirely neutral bystander here; the study’s authors claim that their findings support theories about brain-lateralization that are at odds with theories Poeppel has put forward, and he thinks the paper mischaracterizes his work.)
“What this tells me is that dogs have an auditory system, like all vertebrates do, that distinguishes frequency modulation”—intonation—“and stored sequential sounds,” or familiar words, says Poeppel. But “it’s really not about speech, it’s about auditory distinctions.” In other words, just because a dog recognizes a familiar pattern of sounds as being associated with praise doesn’t mean that the dog is processing the meaning of the word in the way we think it might.
Poeppel is more impressed by the finding that only the correct combination of praising content and praising tone lit up the dogs’ reward center. “That is cool,” he says, because it’s easy to imagine either one or the other doing the trick. But even that, he argues, has more to do with training and reward outcomes than with lexical processing—clever and well done don’t actually mean the same thing, after all. When the dogs, lying perfectly still in an MRI machine, heard well done they couldn’t conclude that they has just performed a task to satisfaction (you can imagine how that might be confusing to them).
The study itself is clever, says Poeppel—and the fact that they managed to train the dogs to stay still in a roaring MRI machine is a feat unto itself—but “the interpretation is just overcooked.”
I don’t have access to an MRI machine or a dog that would willingly lie still in one. But I do have a bad dog. So I decided to speak to her and see if there was any glimmer that she understood. Florence does know a few commands and will sometimes obey them, depending on her mood. How much of her understanding was based on what I said, versus how I said it?
“Sit,” I said to Florence, affecting what I hoped was a neutral tone. She sat. Check.
“Sit!” I said again to Florence, who was seated. Florence flopped to the ground on all fours, in the position we usually call “down,” and looked at me expectantly. Hmm.
“Okos!” I exclaimed, which Google tells me is a Hungarian word for clever, despite Florence not having done anything particularly clever. Florence looked bored.
“Good girl!”
Still bored.
“However.”
Mild panting. Not unusual.
This went on for a while. “Those Mets, huh?” I asked in a last-ditch attempt to connect.
It was at this point that Florence, seeming to have come to the realization that this exchange was unlikely to end in food, sulk-sauntered away to play with the remains of what had once been a stuffed toy chicken.
I’m not sure what I was trying to glean from this exercise. It’s true that Florence didn’t quite seem to be getting what I was putting out there, but for all I know, her blank demeanor concealed a left brain that was going nuts. (Willfully refusing to give me the reaction I’m looking for is something Florence would do.)
The truth is I don’t know what was going on in Florence’s head as I talked to her, but even if there had been an MRI machine involved, I wouldn’t have gotten much clarity.
Of course, that won’t stop me from lying to Florence that she’s a good girl, or asking her how she feels about the weather, or telling her that I realize that she’d like to go for a walk now but she needs to give me a second because I’m trying to finish up a story so just wait five minutes, please. From a scientific point of view, it’d be neat to know what she does or doesn’t understand—and what “understanding” would even mean in this context.
But as a dog owner, I guess I don’t really care. Dogs, even bad ones, provide plenty of emotional satisfaction—and opportunities for projection—as it is, without the additional expectation of actual conversation. Love your dog, pet your dog, and by all means talk to your dog.
Just don’t expect it to really listen.
