How Many Friends Do You Really Need? New Research Tests Loneliness Targets

Loneliness has quietly become one of the most pressing public health challenges of our time, with decades of research linking social disconnection to depression, anxiety, cognitive decline, cardiovascular disease and premature death. As governments around the world begin to treat social connection as a matter of population health rather than private circumstance, an obvious question has emerged: if we can recommend 150 minutes of exercise per week or specific dietary targets, can we also prescribe how many friends people should have, or how often they should see them? A new study published in SSM – Population Health puts that idea to one of its most rigorous empirical tests yet, and the answer is a nuanced one: thresholds appear to exist, but they shift depending on the outcome measured and the mental health of the person being measured.

The research, led by Jorge Andrés Delgado-Ron of Simon Fraser University and Kiffer George Card of the University of Victoria, alongside colleagues across Canada, set out to evaluate whether candidate numeric targets for social connection are stable enough to anchor population-level guidelines. Two initiatives have shaped this debate. The first, led by Julianne Holt-Lunstad, proposed quantitative benchmarks for network size, interaction frequency and relational diversity based on the available evidence. The second, the Canadian National Recommended Public Health Guidelines for Social Connection, deliberately avoided numeric cutoffs altogether, instead emphasizing universal design principles such as accessibility, inclusion and environments that enable connection for everyone, captured in statements like “Make social connection a priority throughout your life.” The new study was designed to determine whether the data themselves favor one approach over the other.

To do this, the team analyzed repeated cross-sectional data from the Canadian Social Connection Survey, an annual online survey of Canadians aged 16 and older conducted between 2021 and 2024. After excluding the 2021 baseline because of instability in the loneliness measure during the pandemic period, the final analytic sample comprised 1,325 participants with a median age of 62. Respondents reported how many close friends they had and how many days in the past week they spent at least five minutes socializing with friends. Loneliness was measured with the validated six-item De Jong Gierveld Loneliness Scale, which distinguishes emotional loneliness, the felt absence of close attachment figures, from social loneliness, the lack of belonging to a broader network. Secondary outcomes included depressive symptoms and generalized anxiety symptoms measured with the PHQ-2 and GAD-2 screeners, and self-esteem measured with the Rosenberg Self-Esteem Scale.

The statistical approach was deliberately flexible. The researchers fitted generalized additive models, or GAMs, with tensor product smooths to visualize the joint, non-linear effects of friend count and socializing frequency, then used segmented regression to estimate breakpoints, the specific values at which the slope of an association changes. Effective degrees of freedom near one indicated a straight-line relationship, while values of two or higher signaled curvature or threshold-like behavior. This two-stage design allowed the team not merely to assume diminishing returns but to formally test where they begin, a step that previous studies of social connection and health had rarely taken.

The headline findings reveal two distinct thresholds. For social loneliness, the benefits of additional close friends were steep up to an estimated breakpoint of roughly 6.7 friends, with each additional friend associated with a meaningful reduction in loneliness scores, and the association flattened to near zero beyond that point. For emotional loneliness, the breakpoint arrived much earlier, at approximately 3.5 friends, after which additional friendships contributed little. This divergence makes conceptual sense: emotional loneliness is relieved by a small number of confiding, attachment-based relationships, whereas social loneliness reflects integration into a wider network and responds to breadth of ties. The thresholds for the secondary mental health outcomes, ranging from 3.0 friends for depressive symptoms to 3.7 for generalized anxiety, aligned closely with the emotional loneliness breakpoint, suggesting that the largest gains across most outcomes occur as friend counts rise from zero to about four.

Time spent socializing told a different story. The association between days spent with friends and both forms of loneliness was essentially linear, meaning each additional day conferred a small but steady benefit. For depression and poor self-esteem, however, the relationship was non-linear, with the strongest improvements occurring up to about 1.6 and 1.5 days per week respectively. For generalized anxiety, benefits accumulated linearly with each additional day. Taken together, the authors suggest that spending at least two days per week socializing with friends is a reasonable floor, while additional time continues to help counter loneliness for those willing and able to socialize more often.

The most striking results emerged when the sample was stratified by mental health screening status. Among people who screened negative for both depression and anxiety, the social loneliness breakpoint was only about 2.5 friends. For those screening positive for depression it was around 4.0 friends, but for those with anxiety alone it rose to approximately 8.0 friends, and for those with both conditions to about 7.0. The pattern for socializing days was even more heterogeneous: among people with depression, social loneliness paradoxically increased with more socializing up to about 3.6 days before declining, while among those with anxiety, loneliness fell up to roughly 4.0 days and then trended upward. Three-dimensional response surfaces also revealed that people with anxiety who reported many close friends but spent little time with them showed high social loneliness, whereas people with both conditions showed elevated loneliness even at high levels of both exposures.

These subgroup patterns carry a caution for anyone hoping to distill social health into a single number. Fixed targets are attractive because they are simple to communicate, easy to monitor and useful for surveillance and evaluation. But the study shows that a cutoff derived from the whole population can misrepresent the needs of subgroups: a target calibrated to the average person would underestimate how much connection people with anxiety may need, while potentially overlooking paradoxical responses among people with depression. The authors note that loneliness and depression can form a self-reinforcing cycle in which social withdrawal and difficulty communicating about mental health exacerbate both conditions, and that interventions may need to move beyond simply increasing contact, supporting safe disclosure and helping networks respond supportively instead. Prior research on existential isolation similarly suggests that contact alone does not guarantee feeling understood.

The study has limitations that the authors acknowledge candidly. The measures of social exposure capture quantity rather than the quality, reciprocity or emotional depth of relationships, and thresholds may differ across lines of racialization, gender identity and other axes of marginalization. The screening tools used are brief instruments rather than diagnostic assessments, the sample skewed older, female and toward higher prevalence of mental health symptoms, and the cross-sectional design precludes causal inference. Some subgroup analyses rested on sparse data, particularly at high levels of both exposures, and the authors could not fully rule out mischievous responding among the small number of participants reporting extreme values. Still, the sample spanned all Canadian provinces and three years of post-pandemic data collection, and the flexible modeling approach represents a methodological advance in testing, rather than assuming, threshold effects.

The broader conclusion is that no single numeric target can adequately capture the complexity of social connection as a determinant of health. The optimal level of connection appears to be context-dependent, shaped by the outcome of interest, life stage, social norms and individual mental health. The findings support a hybrid message: for most people, meaningful gains come early, going from isolation to a handful of close friends and from no social contact to a couple of days per week of socializing, so recommendations to increase socialization remain broadly beneficial. But the specific breakpoints observed here should not be enshrined as universal targets. Instead, the authors argue, guidelines should emphasize flexible, inclusive and context-sensitive principles, as the Canadian guidelines do, with person-centered approaches such as social prescribing translating those principles into tailored, community-based support. In a field racing to write its first population guidelines, this study offers a sobering and useful reminder that the arithmetic of friendship is not one-size-fits-all.

Subject of Research: Testing whether quantitative thresholds for social network size and interaction frequency are stable across loneliness and mental health outcomes in a Canadian sample.

Article Title: Are quantitative targets for social network size and interaction frequency appropriate for public health guidelines on social connection? Evidence from loneliness and related mental health outcomes

Article References: Delgado-Ron, J. A., Orpana, H., Roddick, C., Mulligan, K., Pinel, E. C., Helm, P. J., Oliffe, J. L., Coplan, R. J., Benoit, C., Joordens, S., Skakoon-Sparling, S., & Card, K. G. (2026). Are quantitative targets for social network size and interaction frequency appropriate for public health guidelines on social connection? Evidence from loneliness and related mental health outcomes. SSM – Population Health, 36, Article 101970. https://doi.org/10.1016/j.ssmph.2026.101970

Image Credits: AI Generated

DOI: Not provided

Keywords: loneliness, social connection, public health guidelines, mental health, depression, anxiety, self-esteem, social network size, Canadian Social Connection Survey, threshold effects, emotional loneliness, social loneliness