If you have an Internet connection, you’ve probably read at least one article in the past few days claiming that science has shown that dogs hate being hugged. We at are not here to argue that you should go squeeze the life out of a puppy. But we are here to squeeze the life out of this misleading science coverage.
According to numerous outlets, a scientific study has found that the majority of dogs dislike being hugged, based on cues of distress found in a random assortment of photos pulled from the Internet.
The problem here is that what’s being reported as a “study” is, in fact, an op-ed written in the magazine Psychology Today by a single researcher.
“This is a set of casual observations,” Stanley Coren, the retired University of British Columbia professor who penned the column, told The Post. He reiterated that his data collection wasn’t part of a peer-reviewed study.
Peer review is a process that all scientific studies should go through. It’s a fancy way of saying that other scientists – ones who don’t have a vested interest in the study’s success – have vetted it as being sound. Not all peer-reviewed studies are right or even good, but it’s a basic benchmark for scientific research. It means someone other than the study lead has looked at the data and deemed it legitimate, and that other scientists have had the chance to poke holes in the study’s methodology and conclusions.
(Once a peer-reviewed study exists out in the world, we shouldn’t really say anything can be concluded from it one way or another until other scientists have performed more experiments and come up with the same results, because one never really knows what unknown factors or biases might make a result pop up once or twice in a blue moon – but that’s another issue entirely.)
If the data isn’t published, no one can really comment on it. Several animal behavior researchers I contacted said as much – they couldn’t really tell me whether or not they agreed with Coren’s conclusions, because they had no real sense of what kind of data collection he’d done.
“This is interesting preliminary data which might serve as a good starting point for a formal study,” Evan MacLean, co-director of Duke’s Canine Cognition Center, wrote in an email. “But it’s important to note that (to my knowledge) this is not a peer-reviewed empirical paper so I would caution against any firm conclusions before the work can go through this important part of the scientific process.”
That doesn’t mean this isn’t an interesting dataset: Coren searched Google and Flickr for photos of dog hugs, picked 250 at random, then removed ones where there were obvious reasons for the dog to be stressed and rated the remainder based on established facial cues. But a peer reviewer might argue that Coren had no real information about the environment in which the photos were taken – and the emotional state of the dog beforehand. They’d want to know there was some kind of control, which in this case would mean a way for scientists to see what each dog was like when the variable of a hug was removed.
They might ask for more information on how “random” the photo selection actually was, because any kind of bias might skew the results. They’d probably want to know whether the facial cues had been assessed by a single researcher who went in expecting a specific result, or by participants who knew what facial cues to look for but not what they meant in the context of the study – something that would make them more likely to give unbiased assessments. They might suggest that dog owners are more likely to post hug photos online when their pets make “funny” faces that actually indicate distress. They might ask whether the awkward “hugs” in posed photos were actually comparable to natural cuddles. A peer reviewer might ask what ratio of the photos featured kids versus adults, and whether that might affect a dog’s stress levels – and so on.
It probably seems like I’m being really, really picky. But that’s what peer review is for. Welcome to science, it’s a pain in the butt.
MacLean pointed out that the indicators of stress used in the study are generally accepted, but not totally unequivocal. The natural floppiness of a dog’s ears might make a researcher rate them as “ears down” as Coren did, and the visibility of the white part of a dog’s eye – another indication of stress used in the experiment – can vary based on the direction the dog is looking in. And while a dog sticking her tongue out might mean she’s stressed, it can also be totally normal:
Coren says he’s glad that his data is getting so much attention, because he set out to make folks more aware of the risks of hugging dogs who might become agitated. But he isn’t surprised it’s being reported as if he did a full-scale study.
“The trick might be that the word ‘data’ appears,” he said. “If you use science words, people are impressed. It immediately elevates it, in some people’s minds, to a stereotypical lab full of people in white coats. It’s really just an observation or a measurement which can be used to answer a question.”
Coren hopes some other scientist will take his idea and run with it, using the untapped online photo dataset to perform a more thorough study fit for publication. But, he added: “I’ve been retired since 2007. I wasn’t going to go open up my old lab just to do this.”
But Rachel, you say, I just want to know whether or not to hug my dog. MacLean – and some other animal experts we reached out to – did agree that squeezing your dog probably isn’t a great idea.
“I would advise against hugging dogs, at least in the conventional human form of hugging,” MacLean wrote. “This is essentially a primate behavior (for example, we see similar embraces in nonhuman apes), but not something that dogs do with one another naturally. However, there are lots of ways to have close body contact with dogs that don’t require wrapping your arms around them in a confining manner.”
In other words, don’t put your dog in a chokehold. Hopefully you knew that already.
But has science shown that the well-meaning cuddles your dog seems to love are actually loathed? Not even close.