Vitamin H: the science of happiness is weaker than you think
We’d like to know what really makes people happy, but it’s a hard problem
Imagine that a pharmaceutical company were marketing a new mood-enhancing vitamin, one that makes you happy – meaning that it improves your self-reported subjective well-being (SWB), as psychologists like to put it. I’m sure we’d demand rigorous evidence before recommending it for daily use, no? Yet behavioural recommendations for “happiness” circulate freely, especially among so-called influencers, backed by far less scrutiny.
A paper published in Nature Human Behaviour a few years ago by University of British Columbia’s Dunigan Folk and Elizabeth Dunn, entitled “A systematic review of the strength of evidence for the most commonly recommended happiness strategies in mainstream media,” addressed precisely this problem on the basis of the most rigorous data available so far.
Aristotle says in the Nicomachean Ethics that happiness is what all human action aims at, and Google agrees: “how to be happy” is searched for more than “how to get rich,” which is saying something!
Folk and Dunn wanted to investigate just how much scientific evidence is there on behalf of various types of advice floating around, so they took an in-depth critical look at the available literature in what is sometimes called “positive” psychology, that is, the psychology of everyday life, as distinct from pathologies.
Before we delve into what they found, it’s good to keep in mind the relevant background: the so-called replication crisis that hit the field in 2011. It was around that time that it became clear just how many findings in the psychological literature were likely false positives. This was the result of a number of key culprits: p-hacking, small underpowered samples, no pre-registration, and selective reporting.
P-hacking (or data dredging) refers to the practice of manipulating data analysis until statistically significant results (probability of a chance result, p < 0.05) are found. Researchers might try different combinations of variables, exclude outliers selectively, or stop collecting data only once significance is reached, rather than adhering to a fixed sample size. This effectively tests multiple hypotheses on the same dataset without adjusting for the increased likelihood of chance findings, leading to many “significant” results that do not reflect true effects.
Small, underpowered samples are due to resource constraints, which do not allow researchers to use large sample sizes. When the power of a study is low, however, the few positive results one might find tend to emerge because of statistical noise rather than an actual signal, making the results nearly impossible to replicate in larger, better-powered follow-up studies.
Lack of pre-registration means that researchers have complete flexibility to change their hypotheses or analysis plans after seeing the data. Without a publicly timestamped record of the original plan, there is no way to distinguish between confirmatory research (testing a specific prediction) and exploratory research (looking for patterns). This flexibility may allow researchers to present post-hoc findings as if they were predicted in advance, obscuring the true rate of hypothesis testing.
Finally, selective reporting, closely tied to the other three issues, involves publishing only the studies or outcomes that yield positive, significant results while filing away negative or null outcomes (the “file drawer problem”). This creates a distorted literature where the published consensus suggests strong, consistent effects that simply do not exist in the broader set of conducted research, rendering attempts to replicate those specific findings largely futile.
Together, these practices created a feedback loop where non-replicable findings were published as facts, reinforcing a culture that rewarded novelty over rigor until the field began shifting toward open science practices like registered reports and data sharing. To be clear: the replication crisis is not a reason to shrug off psychology as a scientific endeavor. On the contrary, it is a good, if painful, example of how science at its best works: it was psychologists themselves, including Nobel Prize winner Daniel Kahneman, who pointed out the problem and suggested and implemented the necessary solutions.
Back to Folk and Dunn’s paper: because of the replication crisis, they set high methodological standards, focusing only on (a) experimental (not correlational) designs; (b) pre-registered studies; and (c) studies with sufficient statistical power (≥80% power to detect d = 0.43, which is the average published social psychology effect size).
Folk and Dunn also stayed away from so-called meta-analyses, the practice – diffuse in psychology and biology – of aggregating a large number of published studies and producing overall statistics summarizing any underlying trend. This was a wise move on their part, because meta-analyses are known to yield estimates of effect sizes that are roughly three times those of large-scale pre-registered replications.
The study in Nature Human Behavior focused on five strategies that are often recommended for increasing SWB. Here is how the authors themselves describe them:
Gratitude: expressing gratitude for people and events in one’s life by writing in a gratitude journal, thinking about how lucky one is, and writing gratitude letters;
Social interactions: socializing and reaching out to close friends and family, interacting with strangers and increasing sociability;
Mindfulness / meditation: engaging in mindfulness meditation, breathing exercises and loving kindness meditation;
Exercise and physical activity: engaging in physical activity such as lifting weights, running, walking, playing sports or yoga;
Nature exposure: increasing the time one spends around nature by getting fresh air, going for nature walks or adding plants to one’s home.
And here, briefly, are the results:
1. Gratitude – Verdict: modest support, with caveats
This turns out to be the best-supported strategy of the five. Two well-powered, pre-registered studies show short-term boosts in positive affect from gratitude expression. However, the popular recommendation to write gratitude letters is surprisingly weak, while gratitude lists seem to fare better. Moreover, benefits fade quickly when the practice stops. Overall, then, the approach is reasonably solid, but not the slam dunk the wellness industry often implies.
2. Social interaction – Verdict: fairly solid, but narrow
Folk and Dunn found good evidence that talking to strangers boosts mood (even on the London Underground!). Acting extraverted also shows benefits, but there’s a striking gap: there is almost no rigorous research to be found on the most common recommendation, that is, spending more time with close friends and family. So the most intuitive piece of advice is actually the least tested.
