In a gray, linoleum-floored laboratory in Vienna, two dogs sit in side-by-side enclosures. One dog pulls a handle, and a tray laden with sausage moves down to where the other dog can reach it and excitedly gulp it down. How many times the first dog does this—giving a gift to another, with no benefit to itself—is a measure of what cognition researchers call “prosociality,” or, in essence, generosity. Specifically, how many times it gives food to a friend, or a strange dog, or just offers it to an empty enclosure can help researchers understand whether prosociality exists in dogs, part of a larger quest to understand which creatures are capable of generosity and how and when the trait evolved.

In the past, the Messerli Research Institute scientists who run this particular facility at University of Vienna have found that pet dogs give more to their friends—dogs who live in the same house with them—than to strangers. That fits with the idea that prosociality reinforces social bonds and may have evolved because it contributes to cooperation.

In a new study, however, they report that changing the procedure from simple handle-pulling to a more complex task gives different results. With this task—dogs touching buttons with their noses, cuing a researcher to slide a plate of food under the barrier to the other chamber—dogs still give more to friends than strangers. But, they also give food to an empty chamber when a friend or a stranger is leashed nearby. And, weirdly, they give even less food to a stranger than an empty room. These results don’t fit with the earlier, clearer signs of prosociality in dogs, says Mylene Chaumette, one of the first authors of the paper. It suggests that the task may have been too complex for the dogs to understand exactly what they were doing—they may have been performing the new task simply out of excitement from having learned a new trick, or reacting to the proximity of their friends, rather than comprehending the goal.

That observation matters, because prosociality studies have generated some contradictory results. One kind of test, for instance, has shown prosocial behavior in macaques and capuchins but not chimps or tamarins. Other studies do find prosociality, with different tests. And not all experiments include a control situation where researchers keep another animal that is unable to get at the food near the test subject. To be sure that a behavior is really prosocial—that the animal intends to be giving to an individual—these kinds of controls are needed, says Chaumette. The heterogeneity of the tests has made findings difficult to compare.

So it’s useful that these new results help hammer home how even slight changes in tests can alter outcomes. If the researchers had performed either one of the experiments in isolation, Chaumette says, rather than both, they would have come to different conclusions. The results also confirm what researchers in primates have already learned: “The more difficult the task, the less likely you find prosociality,” says Chaumette. Even human children are less likely to show prosocial behavior as tasks grow more complex.

That said, the researchers believe the handle-pulling test is simple enough for dogs to understand. And they are going to continue their testing using both tasks in another canine: “[There are plans] to replicate it with the wolves to investigate whether prosociality is due to domestication or not,” says Chaumette. Dogs have been profoundly shaped by their relationship with the human species. But wolf packs raise pups together, and defend territory together—will they give presents of food to each other as well? They might give more than dogs, which would lead to the unsettling conclusion that we've made them more selfish, rather than less.