A London School of Economics study of 15,000 adults found that more intelligent people were actually less satisfied with life when they socialised with friends more often — the opposite of what was true for everyone else

A much-repeated psychology finding appears to turn ordinary advice about friendship upside down. In a large US dataset, frequent contact with friends was positively associated with life satisfaction for most young adults, but the slope reversed among people with higher verbal-intelligence scores.

This is one study, not settled consensus. The result was small, correlational and based on data collected in 2001–02 from people aged 18–28. It does not show that intelligent people dislike friends, need fewer relationships or will become happier by cancelling plans.

The most interesting part is not a licence for clever solitude. It is the reminder that an average association can conceal variation, and that frequency alone is a thin description of a social life.

The paper analysed an American cohort

The 2016 British Journal of Psychology paper was written by Norman Li of Singapore Management University and Satoshi Kanazawa of the London School of Economics. Calling it an LSE study is understandable because Kanazawa was the corresponding author there, but the participants were not recruited by LSE. The researchers reanalysed an existing American dataset.

That dataset was Wave III of Add Health, a nationally representative US cohort study that began with adolescents in 1994–95 and followed them into adulthood. Wave III interviews took place in 2001–02. Li and Kanazawa began with 15,197 adults aged 18–28, whose mean age was just under 22. The complete friendship regression included 14,513 cases after missing values.

Life satisfaction came from a single question: “How satisfied are you with your life as a whole?” Answers ran from one to five. Intelligence was estimated using an abbreviated Peabody Picture Vocabulary Test, a measure of verbal ability, with scores converted to the familiar IQ scale with a mean of 100 and a standard deviation of 15.

Socialising was equally compact. Respondents reported how many times in the previous seven days they had either hung out with friends or talked with them on the telephone for more than five minutes. The scale ran from zero to seven or more.

The reversal was statistically significant but tiny

When friendship contact was entered on its own, it had no significant association with life satisfaction. Marriage masked the relationship because married respondents were more satisfied on average and socialised with friends less often. Once current marital status was controlled, frequency of socialising had a positive association with life satisfaction. That positive association remained after controlling for age, sex, ethnicity and education.

The notable result appeared in the interaction. As verbal-intelligence scores rose, the positive link between socialising and satisfaction weakened. The interaction coefficient was negative and statistically significant, with a p-value of 0.014.

For the authors’ illustrative higher-scoring group, one standard deviation above the mean at 115.57, predicted life satisfaction was 4.106 among frequent socialisers and 4.131 among less frequent socialisers. The difference ran opposite to the general pattern, but it was only 0.0248 points on a five-point scale. The authors estimated a standardised effect size of minus 0.03.

For the lower-scoring illustration, at 81.39, frequent socialisers scored 4.159 compared with 4.116 for less frequent socialisers. That effect size was 0.05. Large samples can detect very small relationships, which is why statistical significance and practical importance should not be treated as synonyms.

“Everyone else” is shorthand, not a clean divide

The study modelled intelligence as a continuous score. It did not discover two natural types called smart loners and sociable everyone else. The examples at one standard deviation above and below the mean help visualise a changing slope, but they do not establish a threshold that tells any individual how often to see friends.

Nor did higher-scoring respondents avoid contact. Intelligence was positively correlated with frequency of socialising at 0.121. In this sample, people with higher scores socialised more often even though the model associated that frequency with slightly lower satisfaction at the upper end.

That makes the popular “geniuses prefer to be alone” reading hard to sustain. Preference was not measured. Neither were introversion, the purpose of a gathering, the number of friends present, the freedom to decline or the emotional quality of the contact. Silicon Canals has previously looked at the distinction between fewer friendships and draining relationships; this dataset cannot tell those stories apart.

The savanna explanation is more contested than the result

Li and Kanazawa framed the findings through what they called the “savanna theory of happiness”. Their argument was that modern minds may still respond to circumstances according to the consequences those circumstances would have carried in ancestral environments. Frequent contact with allies would once have supported survival, while high population density would have introduced competition. They further proposed that general intelligence helps people handle evolutionarily novel conditions, weakening those inherited responses.

The paper found the predicted interactions for population density and friendship frequency. It did not observe ancestral psychology, identify an evolved mechanism or compare modern participants with any ancestral population. The evolutionary story is an interpretation placed over the associations.

That distinction matters because the broader Savanna-IQ hypothesis has faced substantive criticism. A critical examination published in Intelligence argued that its account of general intelligence as an adaptation for evolutionary novelty rests on questionable assumptions, can generate contradictory predictions and is difficult to falsify. The interaction in the Add Health model can be reported without accepting the authors’ preferred explanation for it.

Frequency leaves out most of what friendship is

A five-minute phone call and a full day with a close friend each counted as one occasion. Hanging out with one trusted person and attending a crowded gathering also entered the same frequency scale. The item did not separate intimate conversation from obligation, chosen contact from social pressure, or support from conflict.

That missing texture is not a small detail. A systematic review of 38 studies on adult friendship and wellbeing found that friendship quality, support, maintenance, number of friends and socialising all appeared in the evidence. The review’s broad conclusion was positive, while also noting that some findings were sparse, population-specific or contradictory.

Research using recordings rather than a one-week recall item provides another angle. A pooled, multisample replication involving 486 adults found moderate associations between life satisfaction and spending less time alone, spending more time in conversation and having more substantive conversations. It found no reliable link with small talk. That Psychological Science analysis suggests that amount and content can both matter, without showing that one schedule fits everyone.

Intelligence, meanwhile, is not a simple happiness switch. A meta-analysis of general mental ability, job satisfaction and life satisfaction examined indirect routes through income and job complexity. Its model illustrates how much social and economic context can sit between a cognitive test score and someone’s answer to a happiness question.

This is not a prescription for solitude

The Add Health analysis cannot establish causal direction. More social contact might have nudged satisfaction down among some high-scoring respondents, but less satisfied respondents might also have socialised more because they were seeking support or because work, study and family circumstances required it. Unmeasured personality, health, employment and friendship quality could affect both sides of the association.

Timing adds another mismatch. Socialising referred to the previous seven days, while the satisfaction question asked about life as a whole. A busy birthday week, an exam period or a run of work events could shift one answer without representing a person’s normal rhythm.

The defensible conclusion is narrow: in this young US sample, a very small statistical interaction meant that the usual positive association between friendship contact and life satisfaction weakened and then reversed at higher verbal-intelligence scores. The finding has not earned the status of a universal law about intelligent adults.

It is useful as a warning against one-size-fits-all social advice. It is not a reason to rank people by intelligence, mistake a crowded week for friendship quality or treat friends as a cost on a clever person’s calendar.