The Science of
CONNECTION
(RESEARCH)
I’ve spent 15+ years reading everything I can get my hands on about why relationships matter — and then translating it for stages, boardrooms, and books. This is the research I keep coming back to: the studies that changed how I think, the ones audiences ask me about afterward, and the numbers that make the case that friendship isn’t a soft subject.
Use it, cite it, share it with your team. If you’re an event planner or journalist looking for data to make the case internally, this page is for you.
Last updated: August 2026
Highlights
THE
PHYSICAL HEALTH AND LONGEVITY
Lacking social connection increases your risk of premature death about as much as smoking 15 cigarettes a day.
— U.S. SURGEON GENERAL’S ADVISORY, 2023
15
CIGARETTES A DAY
50%
INCREASED LIKELYHOOD OF SURVIVAL
People with strong social relationships have a 50% increased likelihood of survival.
HOLT-LUNSTAD META-ANALYSIS OF 148 STUDIES
2.3X / 2.8X
MORTALITY RATE
In the landmark 9-year Alameda County study, the most socially isolated men were 2.3x more likely to die than the most connected — and for women, 2.8x. The effect held independent of smoking, drinking, obesity, exercise, and baseline health.
BERKMAN & SYME, AMERICAN JOURNAL OF EPIDEMIOLOGY, 1979
60-70%
INCREASED MORTALITY
Social isolation raises mortality risk roughly on par with obesity, physical inactivity, and heavy drinking — it belongs on the same list of major risk factors, not a softer one.
HOLT-LUNSTAD ET AL.; DANISH POPULATION COHORT
“I am not aware of any other factor in medicine that has a greater impact on our survival than the healing power of love and intimacy. Not diet, not smoking, not exercise, not stress, not genetics, not drugs, not surgery.
DEAN ORNISH, M.D.
3 of 9
LONGEVITY HABITS
Of the nine lifestyle habits shared across the world’s longest-lived communities, three are about relationships.
BLUE ZONES “POWER 9”
MENTAL HEALTH
Fragmented
Sleep
Loneliness doesn’t shorten how long you sleep — it fragments it. Lonely people wake more often through the night, measured by actigraphy, even controlling for depression, anxiety, and stress.
KURINA, HAWKLEY, CACIOPPO ET AL., SLEEP, 2011
50%
DEMENTIA RISK
Roughly half of U.S. adults report measurable loneliness — and social disconnection raises dementia risk 50% in older adults.
U.S. SURGEON GENERAL’S ADVISORY, 2023
7X
ANXIETY OR DEPRESSION
Workers experiencing frequent loneliness are more than 7x more likely to suffer anxiety or depression than those with strong social connection.
SYSTEMATIC REVIEW, LONELINESS IN THE WORKPLACE
STRONGEST PREDICTOR
OF LIFE SATISFACTION
Our relationships are the strongest predictor of life satisfaction — stronger than income, IQ, social class, or genes.
HARVARD STUDY OF ADULT DEVELOPMENT (85+ YEARS RUNNING)
THE WORKPLACE
Employees with a best friend at work are 7x more likely to be engaged and less likely to get injured on the job.
GALLUP
7x
MORE LIKELY TO BE ENGAGED
70%
FEWER SAFETY INCIDENTS
Business units in the top quartile of engagement have up to 70% fewer safety incidents than bottom-quartile units.
GALLUP META-ANALYSIS (82,000+ BUSINESS UNITS, 1.8M EMPLOYEES, 73 COUNTRIES)
45%
ISOLATED AND ALONE
45% of professionals now report feeling isolated and alone at work, nearly double the previous year. KPMG calls it “the friendship mirage”: connection on the surface, loneliness underneath.
KPMG FRIENDS AT WORK, 2025
20%
FRIENDSHIP PREMIUM
57% of employees would take 10% below market salary for a job with close work friendships and turn down 10% above market without them. A 20% premium on friendship.
KPMG FRIENDS AT WORK, 2025
$154B
ANNUAL COST
Loneliness costs U.S. employers an estimated $154 billion annually in stress-related absenteeism alone. Lonely employees miss five additional workdays a year and are nearly 2x as likely to intend to quit within 12 months.
CIGNA LONELINESS INDEX, JOURNAL OF ORGANIZATIONAL EFFECTIVENESS
2.4X
MORE LIKELY TO STAY
Professionals with close friends at work are 2.4x more likely to stay with their company.
GALLUP
TRUST AND PSYCHOLOGICAL SAFETY
Google studied 180 teams across 250 attributes to find what makes a team effective. The answer wasn’t talent, seniority, tenure, or team size. It was psychological safety, by a wide margin.
Psychological Safety
GOOGLE PROJECT ARISTOTLE, 2012-2016
MORE ERRORS
Edmondson’s original finding was counterintuitive: the highest-performing hospital teams reported MORE errors, because they were safe enough to admit them.
AMY EDMONDSON, ADMINISTRATIVE SCIENCE QUARTERLY, 1999
CAN I TRUST YOU?
We ask two questions about a new person, in a fixed order. First: “Can I trust you?” Only then: “Can I respect you?” Competence is evaluated only after trust is established.
FISKE, CUDDY & GLICK
61%
HINDERS PERFORMANCE
Half of CEOs report experiencing loneliness, and 61% of those say it hinders their performance. 70% of first-time CEOs name isolation as a significant challenge.
RHR INTERNATIONAL / HARVARD BUSINESS REVIEW
FRIENDSHIP ITSELF
It takes roughly 50 hours together to become casual friends, 90 to become friends, and 200+ to become close friends.
200+
HOURS TO BECOME CLOSE FRIENDS
JEFFREY HALL, UNIVERSITY OF KANSAS
5:1
POSITIVE TO NEGATIVE
Gottman could predict divorce with over 90% accuracy from 15 minutes of conflict, and the differentiator was a ratio: 5 positive interactions for every 1 negative. Outside of conflict, successful couples ran closer to 20:1.
GOTTMAN & LEVENSON
36 QUESTIONS
45 MINUTES
Two strangers, 36 escalating questions, 45 minutes. Significantly closer than a control group making small talk for the same 45 minutes. Compatibility and good intentions produced no effect. Mutual disclosure did.
ARON ET AL., PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN, 1997
HALF
OUR CLOSE NETWORK
We lose about half our close network every seven years, even though the size of that network stays roughly the same.
MOLLENHORST, UTRECHT UNIVERSITY
4X
NO CLOSE FRIENDS
The share of Americans with no close friends has quadrupled since 1990, from 3% to 12-17%, depending on survey year.
SURVEY CENTER ON AMERICAN LIFE
Reseach Library
THE FULL
PHYSICAL HEALTH
Social Networks, Host Resistance, and Mortality (The Alameda County Study)
Lisa F. Berkman & S. Leonard Syme — American Journal of Epidemiology, 1979
The study that started the field. Berkman and Syme followed 6,928 adults in Alameda County, California, over nine years, using a social network index that counted marriage, close friends and relatives, church membership, and other group affiliations.
The findings:
-
Men with the fewest social ties had a mortality rate 2.3x that of men with the most
-
Women with the fewest ties: 2.8x
-
The effect held independent of self-reported baseline health, socioeconomic status, smoking, alcohol consumption, obesity, physical activity, and use of preventive health services
The 17-year follow-up (Seeman, Berkman et al., 1987) extended the finding to adults 70 and older, controlling additionally for depression and perceived health — with social ties still significantly predicting survival.
The Berkman-Syme Social Network Index remains in use as a measurement instrument nearly 50 years later — reconstructed in the Framingham Heart Study, a Danish national population cohort (which found social isolation associated with 60-70% increased mortality), and dozens of other studies.
On the “relationships beat health habits” framing: this study is the origin of that claim, and it’s worth stating precisely. What Berkman and Syme demonstrated is that the social-ties effect held independent of health practices — including a cumulative index of smoking, drinking, obesity, physical activity, and preventive care use. In other words, being socially connected predicted survival regardless of how well or badly you were treating your body.
The stronger popular version — that people with poor health habits and strong relationships outlived people with good habits and weak relationships — is a reasonable reading of the independence finding, but it’s a popularization rather than a reported result in the paper. The defensible claim is the one above: social connection predicted survival on its own, not explained away by lifestyle. That’s remarkable enough, and it holds up to scrutiny.
WHY I CHOSE THIS ONE
Because it’s from 1979, and we’ve all had almost fifty years to act on it. Berkman and Syme followed nearly 7,000 people for nine years and found that the most isolated among them died at more than twice the rate of the most connected — and it held up even after controlling for smoking, drinking, and exercise. Which means we’ve known relationships were a survival issue since before I could read. We just filed it under “nice” instead of “urgent.”
A naming note, since this study travels under a few different names: it’s the Alameda County Study. The same cohort also produced the well-known “Alameda 7” lifestyle findings, which is likely why the two sometimes get blended together. And a precision note on a claim you may have heard: what the study established is that social ties predicted survival independent of health practices — not that connected people with bad habits outlived isolated people with good ones. The first version is documented. The second is a popularization.
Social Relationships and Mortality Risk
Julianne Holt-Lunstad, Timothy B. Smith, J. Bradley Layton — PLoS Medicine, 2010
The landmark meta-analysis. Reviewing 148 studies following participants for an average of 7+ years, Holt-Lunstad and colleagues found that people with stronger social relationships had a 50% increased likelihood of survival. This made social disconnection a more significant mortality risk than obesity, physical inactivity, or excessive drinking.
Holt-Lunstad is a distinguished professor of psychology and neuroscience at Brigham Young University, director of the Social Connection and Health Lab, and served as lead scientific editor for the U.S. Surgeon General’s Advisory. Her later synthesis (World Psychiatry, 2024) covers mental and physical health outcomes and documents rising rates of disconnection globally. Her data draws on more than 300,000 people worldwide.
Our Epidemic of Loneliness and Isolation
U.S. Surgeon General’s Advisory (Dr. Vivek Murthy) — 2023
An 81-page advisory with 325 references, reviewed by 49 peers, that formally declared loneliness a public health epidemic.
-
Mortality risk comparable to smoking up to 15 cigarettes a day
-
Loneliness increases premature death risk 26%; social isolation, 29%
-
29% increased risk of heart disease; 32% increased risk of stroke
-
50% increased risk of dementia in older adults
-
About half of U.S. adults reported measurable loneliness even before the pandemic
-
Americans aged 15-24 had roughly 70% less in-person social interaction with friends than two decades prior
-
Social isolation among older adults accounts for an estimated $6.7 billion in excess annual Medicare spending
-
Isolation diminishes task performance, limits creativity, and impairs executive functions including reasoning and decision-making
Accompanied by a National Strategy to Advance Social Connection — the first U.S. framework of its kind.
Love and Survival
Dean Ornish, M.D. — 1998
Ornish is best known for demonstrating that comprehensive lifestyle change can begin to reverse coronary heart disease without drugs or surgery. What’s less known is his own conclusion about which intervention mattered most:
“Love and intimacy are at the root of what makes us sick and what makes us well, what causes sadness and what brings happiness, what makes us suffer and what leads to healing. I am not aware of any other factor in medicine — not diet, not smoking, not exercise, not stress, not genetics, not drugs, not surgery — that has a greater impact on our quality of life, incidence of illness, and premature death from all causes.”
He calls loneliness and isolation “spiritual heart disease” — and argues it’s the more consequential epidemic.
WHY I CHOSE THIS ONE
Because it’s a cardiologist saying it. Not a friendship person. Not me. The guy who spent his career on diet and heart disease says relationships matter more.
Blood Pressure and the Accumulating Cost of Loneliness
Hawkley, Masi, Berry & Cacioppo — Chicago Health, Aging, and Social Relations Study
A population-based sample of 229 adults aged 50-68 tested annually over five consecutive years. Cross-lagged panel analysis found that loneliness at study onset predicted increases in systolic blood pressure two, three, and four years later — demonstrating that loneliness doesn’t just correlate with poor cardiovascular health, it accumulates.
LONGEVITY
The Harvard Study of Adult Development
Ongoing since 1938 — Harvard Medical School
The longest-running longitudinal study of adult life, now tracking participants and their descendants for over 85 years. Its central finding, repeated across decades of directors: the quality of our relationships is the strongest predictor of long-term health and happiness — outperforming cholesterol levels, income, social class, IQ, and genetics.
Freshman-Year Friendships Predict Wellbeing Fifty Years Later
William Chopik et al. — Innovation in Aging, December 2024
Researchers applied new machine-learning analysis to an existing longitudinal Harvard dataset following the Class of 1964. The finding: the quality of friendships formed in the first year of college predicted wellbeing, relationship quality, and mental health fifty years later.
The Blue Zones “Power 9”
Dan Buettner and research team — National Geographic / Blue Zones
Buettner’s team of medical researchers, anthropologists, demographers, and epidemiologists identified nine lifestyle characteristics shared across the world’s five longest-lived populations (Okinawa, Japan; Sardinia, Italy; Nicoya, Costa Rica; Ikaria, Greece; Loma Linda, California).
Three of the nine are explicitly relational:
-
Loved Ones First — prioritizing family proximity and commitment
-
Belong — all but five of the 263 centenarians originally interviewed belonged to a faith-based community
-
Right Tribe — being born into or intentionally choosing social circles that support healthy behavior
The moai: In Okinawa, home to the longest-lived female population in the world, people form moai — committed groups of typically five friends who support each other for life. Originally created to pool village financial resources, they evolved into lifelong social support structures beginning in childhood. Okinawan women live on average eight years longer than American women.
The team cites Framingham Study research showing that smoking, obesity, happiness, and loneliness are all socially contagious — the composition of your network shapes your health behaviors, not just your mood.
Buettner’s team estimates the average person’s life expectancy could increase 10-12 years by adopting a Blue Zones lifestyle.
A note on rigor: some of Buettner’s centenarian counts have been contested by demographers (Austad, Pes, and Willcox among them, with a peer-reviewed exchange running through 2025). The Power 9 behavioral findings have held up considerably better than the raw longevity statistics. Worth citing the behaviors with more confidence than the ages.
MENTAL HEALTH
Loneliness and Sleep Fragmentation
Lianne Kurina, Kristen Knutson, Louise Hawkley, John Cacioppo, Diane Lauderdale, Carole Ober — SLEEP, 2011
One of the most elegant studies in the field, because it isolates a specific mechanism. The research team used actigraphy — wearable measurement, not self-report — on 95 members of a traditional communal agrarian society in South Dakota.
The finding: loneliness didn’t change how long people slept. It changed how well. Higher loneliness scores predicted significantly higher sleep fragmentation — more awakenings through the night — controlling for age, sex, BMI, sleep apnea risk, and negative affect (a composite of depression, anxiety, and perceived stress).
The interpretation the authors offer: our social nature may be partly manifest through a dependence on feeling secure in our social environment in order to sleep well. If your nervous system doesn’t believe someone has your back, it doesn’t fully stand down at night.
Cacioppo and Hawkley (2003) had proposed sleep disturbance as a key mechanism through which loneliness damages health — arguing that fragmented sleep represents the loss of a fundamentally restorative behavior, with downstream metabolic, neural, and hormonal consequences.
Related findings across the literature: loneliness prospectively predicts poorer sleep efficiency (Cacioppo et al., 2002), greater wake time after sleep onset, and increased daytime dysfunction regardless of sleep duration (Hawkley et al., 2010).
WHY I CHOSE THIS ONE
Loneliness doesn’t shorten our sleep — it fragments it. Researchers measured this with wearables, not self-report, and it held up after controlling for depression, anxiety, and stress. Which means some part of us refuses to fully stand down until it trusts that somebody has our back. We are, quite literally, sleeping with one eye open.
Loneliness and Social Isolation as Risk Factors
Julianne Holt-Lunstad — American Journal of Lifestyle Medicine, 2021
Documents the mechanisms — effects on biomarkers and health-relevant behaviors — through which social connection influences physical and mental health, and argues clinicians have a role in assessing and mitigating isolation. Builds on a National Academy of Sciences consensus committee report.
Social Connection as a Critical Factor for Mental and Physical Health
Julianne Holt-Lunstad — World Psychiatry, 2024
A comprehensive synthesis. Finds robust evidence that social connection factors independently predict mental and physical health outcomes, with the strongest evidence on mortality. Documents rising rates of disconnection across multiple societal indicators.
The Stress Reactivity Pathway
Multiple studies, synthesized in National Academies of Sciences, Engineering, and Medicine, 2020
Two mechanisms appear to operate together: relationships help individuals cope with stressors, and loneliness independently produces a stronger physiological stress response (Hackett et al., 2012; Brown, Gallagher & Creaven, 2018). Loneliness isn’t only the absence of buffering — it’s an active stressor of its own.
Loneliness by Generation
Cigna, Gallup, and Meta-Gallup surveys
Consistently, younger adults report higher loneliness than older adults — reversing the common assumption. Cigna’s research has found roughly 79% of Gen Z, 71% of Millennials, and 50% of Boomers reporting isolation. Other datasets find younger adults roughly twice as likely to report loneliness as those over 65. Gallup’s 2025 work identified younger men in the U.S. as among the loneliest demographic in the West.
A caveat worth naming: loneliness prevalence figures vary substantially across datasets depending on how the question is asked. The directional finding — that younger adults report more loneliness than older ones — holds consistently. Any specific percentage does not. Important to ciite the source alongside the number.
THE WORKPLACE:
THE BUSINESS CASE
Friends at Work
KPMG — 2024 and 2025 surveys
The most current and most striking workplace friendship data available. Across surveys of roughly 1,000-1,100 full-time professional employees:
On the value employees place on work friendship:
-
87% say close work friendships are “very important” — up from 81% in 2024
-
Generational split: 90% of Gen Z vs. 77% of Boomers
-
57% would accept 10% below market salary for a role with close work friendships
-
57% would decline a role paying 10% above market if it precluded work friendships
-
Combined: roughly a 20% salary premium on friendship at work
On
On outcomes:
-
83% say work friendships make them more engaged
-
81% more satisfied
-
80% more connected to their workplace
-
78% say work friendships support their mental health
-
68% report positive impact on job performance
-
41% credit work friends with introducing them to internal advancement opportunities; 36% to external ones
On the lonliness side:
-
45% report feeling isolated and alone at work at least sometimes (2025) — nearly double the 25% in 2024
-
Remote employees and those in telecom, media, and tech report the highest loneliness
-
69% of those without work friends feel isolated
-
Women (30%) report loneliness at work more than men (19%)
On employer competitiveness: companies rating strongly across five domains — work friendships, salary, work-life balance, learning opportunities, and culture — could increase candidate preference by 28 points above market average. Strong work-life balance plus frequent learning opportunities alone lifts employer preference to 71%.
The framing: KPMG’s U.S. Vice Chair of Talent and Culture, Sandy Torchia, describes a “friendship mirage” — universal connection on the surface, increasing loneliness underneath, with surface-level connection mistaken for authentic engagement.
One finding worth flagging: 99% of KPMG respondents said they were open to the idea of AI workplace friendships. Whatever you make of that, it’s a live conversation in every organization right now.
WHY I CHOSE THIS ONE
Twenty percent. That’s what we’re collectively willing to pay — in actual salary — for coworkers we’d call friends. Ten percent less to have them, ten percent more to decline a job without them. Whenever someone tells me friendship at work is a soft topic, this is the number they should believe.
The Cost of Loneliness to Employers
Cigna Loneliness Index — published in Journal of Organizational Effectiveness: People and Performance
Based on an Ipsos survey of nearly 6,000 U.S. employees. The business case, in dollars:
-
Loneliness costs U.S. employers an estimated $154 billion annually in stress-related absenteeism alone
-
Lonely employees miss more than five additional workdays per year
-
Lonely employees are 2x as likely to call out sick, and 5x more likely to miss work due to stress
-
Lonely employees are nearly 2x as likely to intend to quit within the next 12 monthsof non-lonely peers)
-
Lonely employees are less likely to go the extra mile (63% vs. 74%
The intervention finding — this is the important one: employees who had the job resources of social companionship, good work-life balance, and satisfaction with communications were 53% less likely to be lonely than other employees. Loneliness at work isn’t fixed; it’s addressable through specific, nameable conditions.
Some broader analyses place total economic cost as high as $406 billion when accounting for productivity loss and turnover. At an average replacement cost of $15,000-$20,000 per departure, even modest loneliness-driven turnover creates costs most organizations never attribute to the real cause
Engagement, Safety, and the Gallup Meta-Analysis
Gallup — Q12 meta-analysis, multiple editions
Gallup’s meta-analysis is the largest workplace study ever conducted. The 2016 edition examined 82,000+ business units and 1.8 million employees across 230 organizations, 49 industries, and 73 countries. The most recent editions cover 183,806 business units across 53 industries and 90 countries, with data on more than 3.3 million employees.
Comparing top-quartile to bottom-quartile engagement business units:
OUTCOME
DIFFERENCE
Safety incidents
Up to 70% fewer (2016 meta-analysis); 63% fewer in more recent editions
Absenteeism
78% lower
Turnover (high-turnover orgs)
21% lower
Turnover (low-turnover orgs)
51% lower
Profitability
23% higher
Productivity (sales)
18% higher
Quality defects
32% fewer
Theft/shrinkage
32% fewer
Customer loyalty
10% higher
Why the safety number matters for connection work: Gallup’s own framing is that having a best friend at work makes employees “less likely to get injured on the job.” The mechanism is intuitive once named — people who trust their teammates speak up about hazards, admit near-misses, and stop unsafe work. People who don’t, don’t. (This connects directly to the psychological safety research below: in safety-critical industries, psychological safety predicts near-miss reporting rates, incident disclosure, and willingness to stop unsafe work.)
WHY I USE THIS ONE
If we act like friendship is just about morale and belonging and how people feel on a Tuesday, then we are leaving out the part where it keeps people from getting hurt. Up to seventy percent fewer safety incidents in the most connected teams. Seventy. The reason isn’t complicated: we tell our friends things we don’t tell our coworkers. That we’re tired. That we skipped a step. That something looks wrong. A team that can’t admit a near-miss can’t prevent the accident — and admitting things is what friendship is for.
The Best Friend at Work Finding
Gallup — Q12, item 10
Gallup’s long-running finding: employees who have a best friend at work are 7x more likely to be engaged. Related findings:
-
Professionals with close friends at work are 2.4x more likely to stay with their company
-
Women who strongly agree they have a best friend at work are more than 2x as likely to be engaged (63% vs. 29%
-
Those with a work best friend are better at engaging customers, produce higher-quality work, have higher wellbeing, and are less likely to be injured on the job
-
Women with a best friend at work are less likely to be actively watching for other job opportunities, and more likely to trust their coworkers’ integrity and ethics
-
1 in 5 employees worldwide feels lonely at work (Gallup State of the Global Workplace, 2024)
Gallup’s own explanation for including the item: “we instinctively look to develop friendships at work in which we feel respected and trusted.”
Workplace Belonging and Retention
BetterUp
BetterUp’s research on workplace belonging found a 167% increase in employee willingness to recommend their company to others when they feel a sense of belonging, alongside measurable performance gains and turnover risk reduction. Related work has found employees with a strong sense of belonging take substantially fewer sick days.
Social Connections in the Workplace
Birmingham, Holt-Lunstad, Herr & Barth — American Journal of Health Promotion, 2024
Peer-reviewed research extending Holt-Lunstad’s public health framework specifically into workplace settings — useful when you need the workplace claim backed by a clinical-literature citation rather than a consulting survey.
Friendship, Innovation, and Resourcefulness
Multiple studies
A study in Current Psychology found a positive link between workplace friendship prevalence and innovative behavior. Research from the University of Bath’s School of Management found supportive friendships between coworkers to be key to unlocking resourcefulness and original thought.
Workplace Loneliness and Engagement
Moderated mediational model — 274 employees across domestic and multinational organizations
Workplace loneliness reduced work engagement, which in turn increased job dissatisfaction. Notably, the harmful effect on engagement was weaker for individuals with a higher need to belong — a counterintuitive finding suggesting that those most motivated to connect may be more resilient to isolation’s effects.
LEADERSHIP &
WHO IS LONELIEST AT WORK
The Loneliness of Leadership
RHR International / Harvard Business Review
-
50% of CEOs report experiencing loneliness in their role
-
Of those, 61% believe it hinders their performance
-
70% of first-time CEOs report that feelings of isolation are a significant challenge
Corroborating research: HEC Montréal found that while 25% of CEOs report frequent loneliness, 55% acknowledge significant bouts of it — and one in five consistently downplay their loneliness entirely, reflecting the pressure leaders feel to appear in control. A 2018 study in the Journal of Leadership & Organizational Studies found senior managers lonelier both inside and outside work because of role demands — particularly acute for CEOs, who have no true peers within the organization.
Seniority Increases Isolation
Gallup and Perceptyx
Gallup’s data found senior leaders score 10 points higher on loneliness than the employees who report to them — along with higher stress, anger, and sadness. Gallup’s 2026 reporting found managers report higher engagement than individual contributors and significantly worse mental health: stress +7 percentage points, anger +12, sadness +11, loneliness +10.
Perceptyx found roughly 4 in 10 workers lonely, with senior leaders 2x more isolated than their teams.
The pattern is consistent: the more responsibility someone holds, the lonelier the role becomes — and it isn’t a shortage of people. A leader can be surrounded all day and have no one to be honest with.
Who Else Is Loneliest at Work
KPMG, Gallup & Perceptyx — multiple workplace surveys
Drawing across the datasets:
-
Remote employees, and those in telecommunications, media, and technology report the highest workplace loneliness (KPMG, 2025)
-
Gen Z places the highest value on work friendship (90% “very important”) and reports the highest isolation
-
Women report loneliness at work more than men (30% vs. 19%, KPMG 2024)
-
Senior leaders and managers — see above
-
Those without work friends — 69% feel isolated, versus far lower rates among those with them
WHY I USE THIS ONE
Because we often mistakenly think our leaders only need to focus on connecting their teams, when in fact they are quietly the loneliest one in their building. Gallup found senior leaders score ten points higher on loneliness than the people who report to them. Half of CEOs say they’re lonely; sixty-one percent of those say it’s affecting how they perform. And it isn’t a shortage of people — you can be surrounded from eight to six and still have no one you can say the true thing to. The last people we need lonely are our leaders as it reduces engagement, empathy, optimism, and vision.
TRUST &
PSYCHOLOGICAL SAFETY
Project Aristotle
Google — 2012-2016
Google set out to answer one question: what makes a team effective? Over two years, the research team analyzed 180 teams (115 engineering, 65 sales), conducted 200+ interviews, and examined 250+ team attributes. Teams ranged from three to fifty people, median nine.
Google’s working hypothesis was a composition hypothesis — that the best teams are built from the best combination of people. The data said otherwise. No combination of individual traits reliably predicted team performance. Not talent, not seniority, not tenure, not team size, not introversion/extroversion, not whether members socialized outside work.
What predicted performance was a set of shared behaviors and norms — how members treated one another. Five dynamics emerged, in order of importance:
-
Psychological safety — can we take risks here without feeling insecure or embarrassed?
-
Dependability — can we count on each other for high-quality work on time?
-
Structure and clarity — are goals, roles, and plans clear?
-
Meaning — is the work personally important to us?
-
Impact — do we believe the work matters?
Psychological safety wasn’t just first — it was the runaway leader, and the foundation the other four depend on. Without it, dependability erodes (people hide problems) and clarity suffers (nobody asks).
The gap Project Aristotle left: it proved psychological safety matters without explaining how to build it. That gap is the whole reason relational skill-building work exists.
WHY I CHOSE THIS ONE
Because Google went looking for the right people and found the right relationships instead. They studied a hundred and eighty teams across two hundred and fifty attributes, fully expecting to discover the perfect combination of talent and tenure and personality. None of it predicted anything. What predicted everything was how team members treated each other. I love this study because a company with every possible resource set out to solve teams like an engineering problem, and the answer came back human.
Psychological Safety and Learning Behavior in Work Teams
Amy C. Edmondson — Administrative Science Quarterly, 1999
The paper that made psychological safety measurable. Edmondson’s inquiry began with an anomaly in hospital data: teams that reported better relationships also reported more medication errors. The obvious reading — better teams make more mistakes — was wrong. The real finding was that psychologically safe teams surfaced errors that unsafe teams buried.
The published study tested the model across 51 work teams in a manufacturing company. Key findings:
-
Team psychological safety — defined as “a shared belief held by members of a team that the team is safe for interpersonal risk taking” — was associated with learning behavior
-
Team efficacy was not, when controlling for psychological safety
-
Learning behavior mediates between psychological safety and team performance
That mediating variable matters enormously for practice: the path from safety to
performance runs through specific learning behaviors — seeking feedback, experimenting, discussing errors, asking for help. You can’t shortcut from “we want better performance” to “let’s declare ourselves psychologically safe.”
Edmondson locates psychological safety at the team level, and explicitly distinguishes it from interpersonal trust (which operates between two individuals). She also distinguishes it from lowered standards: the combination of high standards and psychological safety is the specific condition that produces learning, and it requires leaders to frame work for learning rather than for evaluation — and to model intellectual humility themselves.
Her 2018 book The Fearless Organization is the accessible synthesis.
Timeline of the concept: Edmondson 1999 (the empirical turn) → Google’s Project Aristotle 2012-2016 → New York Times coverage 2016 (mainstream breakthrough) → The Fearless Organization 2018 → ISO 45003, the first international standard for psychological health and safety at work, 2021 → Australia’s WHS regulations making psychosocial hazard management a legal requirement, 2022.
Psychological Safety in Safety-Critical Industries
An important extension for construction, manufacturing, healthcare, and energy audiences: psychological safety doesn’t only predict productivity and innovation. It predicts near-miss reporting rates, incident disclosure rates, and workers’ willingness to stop unsafe work.
Edmondson’s original hospital finding is the template — high-performing teams report more errors because they’re safe enough to. In an industrial context, a team that can’t admit a near-miss can’t prevent the incident.
This is where the Gallup safety data (70% fewer safety incidents in top-quartile engagement units) and the psychological safety literature converge on the same mechanism: people who trust each other tell each other the truth, and telling the truth prevents harm.
The Warmth-Before-Competence Finding
Susan Fiske, Amy Cuddy & Peter Glick — Research in Organizational Behavior, 2011; and related work
Fifteen years of research establishing that we evaluate other people along two fundamental dimensions, and in a specific order.
The first, unconscious question is: “Can I trust this person?” — an assessment of warmth, encompassing trustworthiness, kindness, friendliness, and empathy. It’s fundamentally a judgment of moral character: does this person intend to help or harm me?
Only after that question resolves do we ask the second: “Can I respect this person?” — an assessment of competence, encompassing intelligence, skill, efficacy, and power.
Cuddy’s framing of why: from an evolutionary standpoint, knowing whether someone deserves our trust mattered more to survival than knowing whether they were capable. Competence, she notes, is evaluated only after trust is established.
The dimensions trace back to Asch’s classic 1946 study, in which manipulating warmth information while holding competence constant produced a radical change in overall impression.
The organizational implication, which Cuddy and Matthew Kohut developed in “Connect, Then Lead” (Harvard Business Review, 2013): many leaders lead with competence signals and underinvest in warmth. But warmth is the conduit — trust is the medium through which ideas travel. Leaders who lead with competence alone may be respected without being trusted, feared without being liked. If people don’t trust you, they aren’t open to your ideas.
WHY I CHOSE THIS ONE
People decide whether to trust us before they decide whether we’re impressive — in that order, every time, without meaning to. Which means warmth isn’t the soft preamble to the credible part. It’s the medium the credible part travels through. If someone doesn’t trust you, they aren’t actually listening to how smart you are.
Bowling Alone: America’s Declining Social Capital
Robert D. Putnam — Journal of Democracy, 1995; expanded as Bowling Alone, Simon & Schuster, 2000
Putnam’s foundational work documenting the decline of American social capital — which he defines as “connections among individuals: social networks and the norms of reciprocity and trustworthiness that arise from them.”
He tracked decline across seven separate measures: political participation, civic participation, religious participation, workplace networks, informal networks, mutual trust, and altruism. The decline began in the 1960s-70s and accelerated sharply through the 1980s and 90s.
On trust specifically: Putnam found successive generations trust one another less, and pointed to indirect indicators alongside survey data — including the boom in employment for police, lawyers, and security personnel after a century of stability. Americans socializing with neighbors more than once a year fell from 72% (1974) to 61% (1993).
On causes, his own attribution: generational change (~50%), electronic entertainment (~25%), suburban sprawl (~10%), time and money pressure (~10%).
n’t open to your ideas.
A note on rigor: later work has both supported and complicated Putnam’s thesis. Costa and Kahn found parallel declines in volunteering and membership; Fischer and others have argued some measures moved less than the framing suggests, and that a few indicators rebounded in the 1990s. The broad decline is well supported. The size of it is still argued about.
FRIENDSHIP:
HOW IT ACTUALLY WORKS
How Many Hours Does It Take to Make a Friend?
Jeffrey A. Hall — Journal of Social and Personal Relationships, 2019
Two studies (N=355 recently relocated adults; N=112 first-year university students tracked over nine weeks) establishing time thresholds for friendship development:
-
~50 hours together to move from acquaintance to casual friend
-
~90 hours to move from casual friend to friend
-
200+ hours to become good or close friends
The regression detail: the probability of transitioning from friends to good/best friends exceeds 50% after roughly 119 hours over three weeks, or 219 hours over three months.
Critically, Hall found that what you do with the time matters: leisure activities predicted closeness, while hours spent working alongside someone counted for considerably less. People who never progressed past acquaintance typically had spent under 30 hours together. And some adults reported hundreds of hours with colleagues while still describing them as acquaintances. Text messages and social media interaction contributed minimally to forming new friendships, though they can maintain existing ones.
The student cohort formed close friendships fast precisely because they had the hours — some reported spending a third of all waking hours in a month with one eventual close friend. Working adults rarely have anything close to that.
Hall drew explicitly on Dunbar’s social brain hypothesis in designing the studies.
WHY I CHOSE THIS ONE
It gives people permission to stop feeling broken. You’re not bad at friendship — you just haven’t put in 200 hours with anyone since college.
Dunbar’s Number and the Friendship Layers
Robin Dunbar — University of Oxford
Dunbar observed a correlation between primate brain size and social group size, and extrapolated a natural human limit of roughly 150 stable relationships. He later refined this into nested layers organized by emotional closeness:
LAYER
SIZE
WHAT IT IS
Intimate
~1.5
Most intimate; typically romantic
Support clique
5
The shoulders-to-cry-on friends who drop everything
Sympathy group
15
Core social partners; trusted with your children
Affinity group
50
The big-weekend-barbecue people
Active network
150
The weddings-and-funerals group
Acquaintances
500
Recognized, but no consistent relationship
Dunbar’s explanation for the layers is time: social time isn’t infinite, and relationship strength correlates directly with time and effort invested. He has noted that beyond 150, relationships become effectively one-way — the reciprocity that defines the number breaks down. The innermost five consume roughly 40% of total available social time.
Independent validation: analysis of mobile phone records from 35 million people (deliberately chosen to predate widespread smartphone and Facebook use) found layer structures closely matching Dunbar’s predictions for the innermost and outermost layers, with more variability in between — approximately four closest contacts, 11 close friends, 30 acquaintances, 129 others.
Two findings with real practical weight: life satisfaction correlates with the inner layers (5-15), not the outer ones. And digital platforms make maintaining weak ties cheaper without raising the ceiling on close relationships — 5,000 followers and 5 close friends use entirely different cognitive systems.
An organizational note Dunbar has made: companies commonly lose their informal culture crossing roughly 150 employees, because relationships can no longer hold the group together and process has to take over.
The Magic Ratio
John Gottman & Robert Levenson — longitudinal studies beginning in the 1970s
Gottman and Levenson asked couples to work through a real conflict in 15 minutes, recorded it, and followed up nine years later. From the tapes, they were able to predict which couples would divorce with over 90% accuracy.
The differentiator was simpler than anyone expected: the balance of positive to negative interactions. Stable, happy couples maintained roughly 5 positive interactions for every 1 negative during conflict — the “magic ratio.” Outside of conflict, the ratio among successful couples ran closer to 20:1. Unhappy couples hovered near 1:1 or worse.
Two details that matter for application:
-
Negativity is asymmetrically powerful. One negative interaction does more damage than one positive interaction repairs — which is why the ratio has to be lopsided rather than merely favorable. As Gottman puts it: the bad is more powerful than the good.
-
The positives are small. Not grand gestures. Displaying interest, asking open-ended questions, making eye contact, well-timed acknowledgment, humor, affection, validation. Gottman’s phrase is “small things often.”
Gottman’s team observed thousands of couples across 50+ years of research at the University of Washington “Love Lab.”
WHY I CHOSE THIS ONE
Because it’s marriage research, and it lands the same in a room about coworkers. Positivity isn’t optional politeness — it’s the ledger every relationship is running whether you’re tracking it or not.
A note on rigor: the 90% predictive accuracy figure comes from retrospective modeling on the original cohorts and has been debated in the methodological literature — predicting forward is a harder standard than explaining backward. The 5:1 ratio itself is robust and widely replicated. Worth knowing which of the two claims is which.
The Experimental Generation of Interpersonal Closeness
Arthur Aron, Edward Melinat, Elaine Aron, Robert Vallone & Renee Bator — Personality and Social Psychology Bulletin, 1997
Better known as “the 36 questions.” Aron’s team wasn’t trying to invent a dating protocol — they had a mundane methodological problem: how do you reliably produce closeness between strangers in a lab, in under an hour, so you can study intimacy at all?
Their solution: pair strangers and have them work through 36 questions over 45 minutes, escalating gradually from small talk to genuine disclosure. A control group spent the same 45 minutes making small talk.
Study 1 found significantly greater post-interaction closeness in the disclosure condition than in the small-talk condition. Some participants reported closeness comparable to their closest existing relationships. One pair from the original experiment married and invited the lab to the wedding.
The findings that matter most are the ones people skip:
-
Studies 2 and 3 found no significant effect for whether pairs were matched on attitudes, whether they were led to expect mutual liking, or whether getting close was made an explicit goal. Compatibility and intention weren’t the active ingredient.
-
Escalating mutual self-disclosure was.
-
Pairs who bond in 45 minutes and never meet again typically don’t maintain it. Aron’s own framing: the questions are a doorway, not a guarantee — walking through it together doesn’t mean you end up in the same room afterward.The closeness is real but doesn’t automatically persist.
The questions or variants have since been used in hundreds of studies, including work on cross-race friendship and prejudice reduction, and in police-community relationship programs.
WHY I CHOSE THIS ONE
It’s the cleanest proof that vulnerability is a mechanism, not a mood. You can manufacture closeness in 45 minutes with structured disclosure — and you can’t manufacture it with compatibility, good intentions, or small talk. (That’s why my speed-friending sessions work.)
Half Your Close Friends in Seven Years
Gerald Mollenhorst — Utrecht University, 2009
Mollenhorst surveyed 1,007 people aged 18-65, then re-interviewed 604 of them seven years later. The findings:
-
Only 48% of the people in someone’s close network were still in it seven years later
-
Only 30% still held the same position (discussion partner, practical helper)
-
Yet network size stayed strikingly stable — people replaced who they lost
Most of these weren’t dramatic ruptures. Jobs changed. Neighborhoods changed. Kids started at different schools. The weekly coffee became monthly, then quarterly.
Mollenhorst’s larger finding was about context: our friendship choices are constrained far more by opportunity to meet than by preference. People repeatedly choose friends from contexts where they’ve already chosen a friend, and whether our friends know each other depends heavily on where we met them. Part of the project “Where Friends Are Made: Contexts, Contacts, Consequences,” led by Beate Völker.
WHY I CHOSE THIS ONE
It’s the most freeing statistic I know. Friendship turnover isn’t failure — it’s normal. And it means building new friendships isn’t a young person’s task, but rather a lifelong one.
The Friendship Recession
Survey Center on American Life — American Perspectives Survey and American Social Capital Survey
Tracking close friendship counts over three decades:
-
Americans reporting no close friends: 3% (1990) → 12% (2021) → 17% (2024 survey)
-
Americans reporting ten or more close friends: 33% (1990) → 13% (2021)
-
Americans with three or fewer close friends: 27% (1990) → 49%
-
Average close friend count: 2.94 (1990) → 2.23 (2021)
The gender split: men with six or more close friends fell from 55% to 27%. Men with zero close friends rose from 3% to 15% — a fivefold increase. One in five unmarried men not in a romantic relationship reports having no close friends. Women’s numbers declined too (roughly 2% to 10% with no close friends), indicating a structural trend rather than one confined to one group.
A finding that matters more than it looks: Americans with one close friend are not measurably less lonely or isolated than those with none. Those with a couple are only modestly better off. Size matters — the inner layer needs more than one person in it.
Related: American Time Use Survey data shows time spent with friends dropped roughly 37% between 2014 and 2023. By middle age, Americans devote about 30 minutes a day to maintaining friendships.
Note: the 12% vs. 17% figures reflect different survey years and instruments, not a contradiction. Cite the year alongside the number.
WHY THREE THINGS:
THE CONVERGENCE ARGUMENT
This section explains where the Three Requirements framework comes from — and why it isn’t a separate theory competing with the research above, but a synthesis of it.
Every body of relationship research on this page studies a different relationship, in a different setting, using different language. Marriage researchers study couples. Organizational psychologists study teams. Epidemiologists study social networks and mortality. Communication scholars study friendship formation. They rarely cite each other.
And yet the same three conditions keep surfacing under different names.
Positivity — someone we enjoy. The emotional payoff of being together.
Consistency — someone we can rely on. A track record built over time.
Vulnerability — someone who knows us. Mutual disclosure of what’s real.
Here’s the same three showing up across the literature, in the researchers’ own terms:
RESEARCH TRADITION
POSITIVITY
CONSISTENCY
VULNERABILITY
Gottman (marriage)
The 5:1 ratio of positive to negative interactions during conflict — 20:1 outside of it
“Small things often”; decades of predictive data built on accumulated interaction
Turning toward bids for connection; letting yourself be known
Edmondson (psychological safety)
Without fear of embarrassment, rejection, or punishment
A shared belief — necessarily formed over repeated experience
Able to bring up problems, admit mistakes, ask for help
Interpersonal risk-taking
Project Aristotle
(teams)
Norms of how members treat one another
Dependability
Interpersonal risk-taking
Fiske / Cuddy / Glick (trust)
Warmth: friendliness, kindness, empathy
Reliability signals; trust accrues
Trust as prerequisite for openness to influence
Hall (friendship formation)
Leisure activities predicted closeness; work hours didn’t
50 / 90 / 200+ hours — time is the mechanism
Everyday talk and disclosure as a distinct predictor
Aron (interpersonal closeness)
Small talk control group didn’t produce closeness
Closeness generated in a lab doesn’t persist without continued contact
Escalating self-disclosure was the active ingredient
Dunbar (network structure)
Life satisfaction tracks the inner layers, not the outer
Layers exist because social time is finite
The support clique: those who know you well enough to drop everything
Blue Zones / moai (longevity)
Companionship as the modern function
Groups committed for life, formed in childhood
“Someone will always be there for you”
The claim, stated plainly
Every other quality we name in relationships is one of three things: a synonym for one of these (warmth, safety, reliability, honesty), an illustration of one (showing up to the funeral, admitting the mistake, laughing together), or an outcome of them (trust, belonging, engagement, loyalty, retention).
Take psychological safety, the most rigorously studied construct in the organizational literature. Edmondson’s own definition contains all three: a shared belief (consistency — a track record, since a shared belief can’t form in a single interaction) that the team is safe for interpersonal risk-taking (vulnerability — bringing up problems, admitting error, asking for help) without fear of embarrassment, rejection, or punishment (positivity — the expectation of a warm rather than hostile response).
Psychological safety isn’t a fourth thing. It’s what the three look like when they’re present on a team.
Why this matters practically
Project Aristotle established that psychological safety is the strongest predictor of team effectiveness — and then stopped. It proved what matters without explaining how to build it. Edmondson identified the mediating pathway (learning behaviors) and located the construct at the team level, but her work describes the climate rather than the mechanics of human bonding.
That gap is the whole reason a relational framework is useful. If the three conditions are the common substrate underneath trust, safety, friendship, and marriage, then they’re also the diagnostic: when a team relationship is struggling, one of the three is missing, and you can identify which. That’s what the Relational Results assessment measures — positivity, consistency, and vulnerability, scored individually and rolled up to a team view.
The framework doesn’t compete with the research on this page. It’s downstream of it.
A note on what kind of claim this is: it’s a synthesis, not an empirical finding. No study has ever tested whether there are exactly three requirements. What’s demonstrable is that three dimensions recur with unusual consistency across research traditions that rarely cite one another, and that treating them as the operative variables produces something you can actually diagnose and act on. That’s how frameworks get built — but it’s a different kind of claim than the studies above it, and it deserves to be labeled as one.
SHASTA’S
OWN RESEARCH
The Relational Results Workplace Assessment
A proprietary diagnostic measuring team and organizational relational health across three dimensions — positivity, consistency, and vulnerability — the same framework underlying Frientimacy and The Business of Friendship. Individual participants privately evaluate their own experience; results roll up into a team score showing leaders where they’re thriving and where there’s room to grow.
The Frientimacy Quiz
More than 10,000 people have taken Shasta’s self-assessment, scoring their own relationships across the Three Requirements.
Qualitative Research
Fifteen years of friendship coaching, group facilitation, and workplace relationship mediation — thousands of conversations about how people actually form, maintain, deepen, and lose friendships.
The Six Points of Connection Study
U.S. Chamber of Connection — 2026
A national study, conducted through the U.S. Chamber of Connection where Shasta serves as Chief Friendship Officer, identifying the six points of connection that distinguish the lives of people who are socially thriving.
SHASTA’S
FRAMEWORKS
The Three Requirements of Relationship™
The Three Requirements of Relationship™ (also called the Frientimacy Triangle) — Positivity, Consistency, Vulnerability. Every healthy relationship has all three; every struggling relationship is missing one.
The Five Levels of Relationship
The Five Levels of Relationship — Curiosity → Exploratory → Familiarity → Commitment → Community.
The Belonging Wheel
The Belonging Wheel — from U.S. Chamber of Connection research, 2026.