Few things are more adorable—or destructive—than a new puppy. When they pee on rugs, chew furniture, and get aggressive with other pups, their stressed-out owners usually turn to dog training. Now, a novel study suggests programs that use even relatively mild punishments like yelling and leash-jerking can stress dogs out, making them more “pessimistic” than dogs that experience reward-based training.

“[Punishment] training may seem to work in the short run … but these methods can have future negative consequences,” says Marc Bekoff, an evolutionary biologist at the University of Colorado in Boulder who was not involved in the new study. “[These dogs are] living in perpetual stress.”

Previous studies have suggested that although both reward-based and punishment-based training methods are effective, punishment-based training can have negative effects. But those studies tend to focus on police and laboratory dogs instead of family pets, and most used shock collars, which have been banned in several countries, as punishment.

To find out how companion dogs react to more routine punishments, scientists led by Ana Catarina Vieira de Castro at the University of Porto in Portugal recruited 42 dogs from reward-based training schools, which use food or play to encourage good behaviors. The team also enlisted 50 dogs from aversive-based programs, which use negative reinforcement like yelling and leash jerking to train dogs, or even pressuring their rumps to get them to sit.

The researchers videotaped the dogs during training and tested their saliva before and after for the stress hormone cortisol. Dogs in the negative reinforcement programs showed more stress-related behaviors during training, such as lip licking and yawning, and they had higher levels of cortisol in their saliva than when at home, the team reports on the preprint server bioRxiv. Dogs in the reward-based training group showed no changes in cortisol levels during training or at home.

To find out whether these effects lingered, the researchers measured how 79 of the dogs responded to a potential food reward. First, they trained the dogs to associate one side of a room with a delicious sausage. If a dog found a bowl in that part of the room, it would contain sausage. But bowls on the other side of the room would be empty.

Then, the researchers placed an empty bowl at various positions between the two extremes and measured how quickly the dogs approached it. An “optimistic” dog would run excitedly to a bowl in the middle, whereas a “pessimistic” dog would move more slowly. (In humans, an equivalent might be a glass half empty versus glass half full mindset.) Such “pessimistic” mindsets have been associated with separation anxiety and other problem behaviors in dogs. In the test, the more punishment a dog had received, the more “pessimistic” it was, and the more pronounced the results.

“This was a careful study,” Bekoff says. And although the paper does not address which method is more effective at training dogs, Bekoff says this and other findings provide more than enough evidence that dog owners should avoid aversive-based training.

That’s often easier said than done, because many dog training schools don’t advertise their methods, and such training is not regulated—at least in the United States, says Zazie Todd, a dog trainer and animal psychology blogger. She adds that dog owners should look explicitly for keywords like “reward-based,” and avoid schools that use language like “balance training” or “dominance methods.”

Bekoff agrees and says owners should take the time to talk to the trainer and to other owners who have worked with them. “[Reward-based training] may take time, but so what? At least the dog isn’t living in fear or constant stress.”