犬は協力を選ぶ？ある研究でわかったこと

One 2026 study tested whether pet dogs would keep trying to earn a larger reward together when each dog could instead choose a smaller reward alone. Across 23 household pairs, the dogs coordinated on the harder joint option in about 5% of included choice trials in both reward conditions. They usually chose the individually solvable option, and the cooperative choice became less common as sessions continued.

The question behind the study

The researchers used a stag-hunt game, a structured choice in which two participants can work together for a higher-value reward or choose an individually solvable option for a lower-value reward. They asked whether dogs would adjust their choice to a partner and invest enough effort to make the joint option work.

The paper examined one demanding apparatus in one volunteer sample. It did not test whether a dog is friendly, loyal, independent, or capable of cooperation in ordinary home life. The result applies to the task, its reward rules, and the dogs who took part.

Who took part

The study tested 23 pairs of pet dogs at the Clever Dog Lab in Vienna, Austria. The dogs were between 1 and 12 years old. Each pair had lived with the same owners and with each other for at least one year. Owners reported no food aggression between the dogs. Eleven pairs entered the typical-payoff condition and 12 entered the same-reward condition.

The sample came from voluntary recruitment through the lab's participant database. Four pairs were not tested because one dog did not learn the trained action after three training sessions. During the first test session, 16 dogs received extra encouragement after repeated non-response. All of those dogs then made a choice, and no animal was excluded at that stage.

How the game worked

Each pair completed six sessions of 30 trials on different days. The dogs chose between two sides of an apparatus:

• The individual option, called hare, could be solved by one dog and produced a low-value reward.
• The joint option, called stag, required both dogs to choose the same side and pull their assigned rope or drawer at similar times and intensities.

In the typical-payoff condition, the larger reward was available only when both dogs coordinated on the joint option. In the same-reward condition, the larger reward could be earned from either the individual or joint side. A food-preference test helped the researchers choose the higher- and lower-value foods for each pair.

The platform could be rotated to 18 marked angles, so the sides were not always in the same place. One dog learned to pull a rope with its mouth and the other learned to pull a handle with its paw. The dogs practiced on a different apparatus to limit familiarity with the test equipment. The study recorded choices and successful food access from video, and 20% of the videos received a second coding pass. Interobserver reliability was above the paper's stated threshold for every coded variable.

What the researchers counted

These denominators are the included choice trials reported in the paper. The analysis included trials in which the dogs made a choice, so they should not be read as a simple count of every scheduled opportunity.

The individual side was also more reliable. When dog 2 chose hare, food access was reported at 90% to 95% across the partner-choice conditions. When the dogs coordinated on stag, dog 2 reached the food in 8% of the same-reward trials and 21% of the typical-payoff trials.

What changed across sessions

Dogs in the typical-payoff condition showed more early coordination than dogs in the same-reward condition, but that coordination declined more sharply in later sessions. Across both conditions, dog 2 became less likely to choose the joint option as sessions continued. The partner's choice did not produce the predicted interaction in the final model.

Distance mattered as well. Both dogs were more likely to coordinate on the joint side when it began closer to them. That finding is a reminder that a choice can reflect the immediate layout of a task as well as the reward relationship between the options.

Evidence-led insights

An easier reward can win when the joint option is unreliable

The dogs often returned to the individual side, where one correct action was usually enough to reach food. The joint side required the right choice from both dogs, close timing, similar force, and persistence. A cautious inference is that the full cost and reliability of an option shaped the dogs' later choices. The study does not establish a general preference for working alone.

Cooperation here was a coordination problem, not one simple social decision

A pair could fail after both dogs selected the joint side if their actions were not well timed or forceful enough. The authors also note that minimal training left the dogs to explore the consequences of the apparatus. This means a low success rate can reflect physical coordination, reward motivation, or learning the task, as well as any attention to the partner.

One unusually successful pair changes the interpretation

Aquila and Nox, one pair in the study, chose the joint side in 44% of trials and succeeded in 16% of those trials across sessions. Their performance shows that the apparatus was solvable for at least one dyad. It does not show that their breed, age, training, or relationship caused the difference, because the study was not designed to test those explanations.

Experience and context deserve equal weight

The experiment's early exploration, later shift toward the individual option, and strong pair-level difference all point to a context-sensitive result. The authors compare this minimal-training design with studies in which dogs received more direct instruction or worked with predictable human partners. Those studies are not interchangeable with this one, but they help explain why a task can measure experience and apparatus demands alongside social coordination.

What the study cannot tell us

This paper cannot tell us whether a dog prefers solitude, dislikes another dog, lacks a bond, or would cooperate in a familiar daily routine. It also cannot rank breeds or multi-dog households. The dogs lived together and were screened by owner report for food aggression, which narrows the sample rather than making it representative.

The authors state that the design cannot fully separate an inability to understand the partner's role from difficulty with the physical coordination or means-end demands. The higher-value food may also not have been worth the extra effort for most pairs. One joint-choice percentage therefore describes this experiment, not a stable social trait.

Observable choices in this task are not a diagnosis of fear, anxiety, aggression, stress, or another condition. If repeated interactions at home involve conflict or distress, an individualized assessment from a qualified behavior professional is more appropriate than using this experiment as a test.

A careful takeaway for dog families

When two dogs choose separate activities, read the moment before reading the relationship. The reward, the layout, the effort required, the dogs' prior experience, and the chance of success can all change what they do next. A single failed attempt at a shared activity does not establish that the dogs cannot cooperate or do not enjoy one another.

If you offer a joint enrichment activity, keep it familiar and easy enough to observe without forcing the dogs to compete. Watch the pair's repeated pattern and comfort, then adjust the activity if one dog is repeatedly crowded out. This practical observation goes beyond what the study tested, so it should remain a gentle household choice rather than a claim about the dogs' underlying abilities.

Sources and limits

Wallner Werneck Mendes J, Cimarelli G, Vindevogel M, van Peer I, Ladurner G, Range F. Pet dogs prefer to work alone than to engage in a challenging cooperative task with conspecifics. PeerJ. Published January 27, 2026; 14:e20609. DOI: 10.7717/peerj.20609.

The open-access full-text paper provides the study design, methods, results, coding details, and limitations. The publisher record provides the formal article record. The source was checked again on October 6, 2026. Reported findings and reader-facing inferences are kept separate in this brief.