Social Buffering
Other people can lower a physiological stress response, but presence alone does not do it. What buffers cortisol reactivity is shared exposure: someone undergoing the same stressor alongside you, not someone merely there to support you. Filetti, Tsakonas, Donzella, Thomas and Megan Gunnar's 2025 study of 235 adolescents found cortisol reactivity fell only when a friend was tested alongside the participant, and rose when the friend was present purely as a supporter. An audience is not a buffer. A co-sufferer is.
What it is
The common assumption about stress is that company helps, that having someone nearby, on the phone, in the room, dampens whatever the body is doing in response to a threat. Social buffering research complicates that assumption. The physiological benefit of other people is not simply a function of their presence. It depends on what those people are doing while they are present.
The clearest recent demonstration comes from adolescent stress research. Filetti, Tsakonas, Donzella, Thomas and Gunnar ran 235 young people, aged 11.2 to 15.1, through an online version of the Trier Social Stress Test, a standard laboratory procedure that reliably induces measurable cortisol reactivity by putting someone through an evaluated, socially threatening task. Participants went through it alone, alongside an unfamiliar peer who was also being tested, alongside a friend who only supported them from the sidelines, or alongside a friend who was tested at the same time. The pre-registered prediction was specific: buffering would depend on whether the friend shared the stressor, not on whether the friend was simply liked, present, or encouraging.
The result matched the prediction exactly, and in one direction it ran against what most people would guess. Cortisol reactivity dropped when the supporting friend was also being tested. It rose when that same friend was there only to cheer the participant on but was not undergoing the stressor themselves. An encouraging audience made the stress response worse, not better. Recovery afterward did not differ across conditions, so the effect was specific to the size of the spike rather than to how quickly someone calmed down once the task had ended.
The mechanism this points to is shared exposure, not company. A friend who is watching someone sweat through a task, however warmly, is in a different physiological category from a friend who is sweating through it alongside them. That distinction has a direct bearing on how tightly bounded groups form: a group built around a shared stressor has a reason to exclude people who did not go through it, because their presence does not do the same physiological work.