What Would Test the Third Place Hypothesis

A 2025 study in Health & Place tracked twelve categories of third place — coffee shops, libraries, museums, recreation centres, restaurants and others — across United States census tracts from 2010 to 2021, and found closures in every category between 2019 and 2021. Losses were concentrated in tracts with higher social vulnerability and in rural areas. Summarising the work for a general audience, the University of Colorado Boulder’s Institute of Behavioral Science framed the disappearance of senior centres and churches as a driver of loneliness rather than a symptom of it.

That framing is a causal claim, and it is now doing policy work. The Surgeon General’s 2023 advisory made strengthening social infrastructure one of six pillars of a proposed national strategy. The WHO Commission on Social Connection’s June 2025 report calls on all Member States to treat social connection as a public health priority, citing an estimated 871,000 deaths a year associated with loneliness. Neither rests on a study that measured what happens to individual loneliness when a specific place closes.

The claim is plausible. It is also largely untested at the level at which it is asserted. Splitting it into its constituent links shows where the evidence is strong, where it is thin, and what a study would have to do.

Link one: the places closed

This is the best-evidenced step, with one large caveat about timing. The Health & Place exposure window that shows the sharpest decline — 2019 to 2021 — is the pandemic. Closures during that period are confounded with lockdowns, mortality among older patrons, the shift to remote work, and a collapse in demand that would have depressed foot traffic regardless of whether a venue’s doors stayed open. The study establishes that availability fell and that it fell unevenly. It cannot separate venue loss from everything else that happened simultaneously.

Link two: fewer places, less contact

Here the evidence is weakest, and the causal arrow is genuinely contested. Ray Oldenburg’s 1989 account of third places was an argument about their function, not a measurement of their effect. Eric Klinenberg’s 2018 work on social infrastructure comes closer, drawing on the 1995 Chicago heat wave to argue that shared physical spaces shaped survival — but that comparison is between neighbourhoods that differed in many ways at once, not between the same neighbourhood before and after a closure.

Against the venues-first ordering stands Robert Putnam’s account of civic decline, which documented falling club membership and informal socialising across decades that mostly preceded the closures in the Health & Place series. Time-use evidence reported by WBUR in early 2024 shows face-to-face socialising declining before the pandemic and continuing to decline after it. If Americans stopped going out first, the closures are downstream, and building new venues will not reverse the trend.

Distinguishing these two orderings is the single most important empirical task in this literature, and no supplied study does it.

Link three: less contact, more loneliness

AARP’s 2018 survey of 3,020 adults aged 45 and older, using the 20-item UCLA Loneliness Scale, found that 33% of those who had spoken with their neighbours were lonely, against 61% of those who never had. The gap is large. It is also cross-sectional, from a single wave, and entirely compatible with lonely people withdrawing from neighbourly contact rather than the reverse. AARP’s 2025 follow-up reports rising loneliness in the same age band, and is unusually valuable because the instrument is consistent across waves — a rarity in this field.

This link also runs into the isolation–loneliness distinction. A closed library removes an opportunity for contact, which is a structural change to a person’s network. Loneliness is the subjective appraisal of that network. Work published in Scientific Reports in late 2024 found the relationship between the two varies by age, which means a place-based intervention could plausibly reduce isolation without moving loneliness at all. Holt-Lunstad’s 2015 meta-analysis found both independently predictive of early mortality — isolation OR 1.29, loneliness OR 1.26 — so an intervention that shifted only one would still be worth having, but it would not be the intervention that was promised.

Link four: loneliness, then disease

The final link is the strongest. The American Heart Association’s 2022 scientific statement put social isolation and loneliness at roughly a 30% increased risk of heart attack, stroke, or death from either. That statement also named the absence of intervention evidence as the central research gap — a judgement that has not been overtaken.

What a test would look like

Four features would move this from hypothesis to evidence.

Individual-level panel data with geocoded exposure. Repeated loneliness measurement on the same people, using one instrument, linked to the opening and closing of venues within a defined radius. Prevalence estimates in this field range from 37% moderate-to-severe in the 2022 HINTS-6 data published this November to 57% in Cigna’s 2025 report from mid-2024 fieldwork, almost entirely because the questions differ. A place-based study that switches instruments mid-stream will produce noise indistinguishable from effect.

Exogenous closures. Library branch closures driven by municipal budget shortfalls, or venue losses following transit changes, provide variation not generated by declining local demand. That is what separates link two from Putnam’s ordering.

Isolation and loneliness measured separately, with a time-use intermediate. If venue loss reduces contact frequency but not loneliness, that is a finding, and a policy-relevant one.

Attention to what the place is for. The 2022 qualitative meta-synthesis of social prescribing found benefit came from structured, purposeful participation rather than contact alone. A café supplies contact; a choir supplies purpose. Any test that pools twelve venue categories will average across that difference.

The intervention trials that do exist — HEAL-HOA’s volunteering trial in Hong Kong, and the befriending trial in Australian residential aged care that cut UCLA scores by 2.71 points at sixteen weeks — put people into activities. They do not test whether building the room changes anything. Meanwhile the systematic review protocol registered in July 2025 notes that only one peer-reviewed randomised trial of social prescribing in older adults exists, seven years after the UK strategy began funding it.