A Better Rule for Looking Up
Upward social comparison is neither a universal source of harm nor a universal engine of improvement. The evidence supports a more conditional conclusion: a person can compare themselves with a superior standard, experience a socially meaningful threat, and later show a behavioural adjustment, but those outcomes depend on what is being compared, whether the gap seems actionable, what other feedback surrounds it, and how the result is measured. The useful judgment is therefore not “Is comparison good or bad?” It is “What kind of comparison is this, and what can the evidence justify saying about it?”
That question brings together the two research syntheses without pretending that they studied the same exposure. The online upward-comparison meta-analysis found an average association with poorer psychological functioning across multiple outcomes, strongest in the cross-domain model for social-evaluative negative emotions. The randomized-trial meta-analysis found small average effects when comparison was deliberately used in behaviour-change interventions. One result describes a naturally occurring self-evaluative relationship; the other describes a designed package tested against a control. The apparent paradox dissolves when the comparison process, the setting, the outcome, and the evidence design are kept in view.
One process, two settings
The most compact model has four questions. What is the social standard? On which dimension is it superior? How does the person appraise the gap and their ability to respond? What outcome is observed afterward? This model keeps the primary subject—upward social comparison—visible while allowing its consequences to branch.
A comparison can become a status threat when the standard concerns acceptance, appearance, competence, or another part of the social self, and when the gap is meaningful but not clearly controllable. The immediate result may be envy, body dissatisfaction, loneliness, or fear of negative evaluation. These emotions are specific signals about social position or belonging. If similar comparisons recur, they may be associated with wider indicators such as anxiety, depression, lower wellbeing, or lower self-esteem. The Frontiers review’s proximal–distal framing helps organize those outcomes, but it does not prove that one causes the next. The evidence shows differing average associations, not a confirmed path from a moment of envy to a later disorder.
The same basic process can instead become an actionable reference when the dimension is malleable, the standard is understandable, and the person can see a credible next step. A structured intervention may show a participant’s energy use beside a social standard, or provide another measurable comparison on health, performance, or service behaviour. The intervention can prompt action even if the comparison is not emotionally pleasant. When randomized trials find a shift in the intended direction, they provide evidence about the comparison-containing intervention relative to its control, not proof that all upward comparison improves wellbeing.
These branches can overlap. A participant may feel threatened and still change behaviour. A person may feel temporarily inspired and later become preoccupied with monitoring a superior target. A comparison intervention could produce a small group-average behavioural effect while also imposing emotional costs on some participants, although the RCT synthesis did not directly quantify that trade-off. An ordinary online comparison can also supply useful information in a particular moment. The model does not classify people as resilient or fragile; it identifies conditions that can change what the same relative difference means.
The branches are not competing theories of two different people.