Try Wonders for freeHow the literature is organised
The oldest works listed here are about Facebook, and the field still carries that imprint. Kross et al. (2013) text-messaged people five times a day for two weeks and found that the more they had used Facebook at one point, the worse they felt at the next — with no such effect from interacting with people directly. Verduyn et al. (2015) split the behaviour: in a laboratory experiment and an experience-sampling study, passive use undermined affective wellbeing by increasing envy, independently of active use.
A second cluster, from 2016 onwards, moves to adolescents and clinical outcomes. Lin et al. (2016) surveyed 1,787 US adults aged 19 to 32 and found the highest quartile of time on social media had 1.66 times the odds of depression, with a dose-response trend. Woods and Scott (2016) found, among 467 Scottish adolescents, that night-time use predicted poorer sleep quality even after controlling for anxiety, depression and self-esteem. Keles, McCrae and Grealish (2020) systematically reviewed 13 studies, 12 of them cross-sectional, and found correlations across all four domains they examined — time spent, activity, investment and addiction — with heavy caveats about design.
A separate, smaller strand treats social media as a resource. Naslund et al. (2016) argue that online peer-to-peer connection can help people with serious mental illness challenge stigma and find support, while naming the risks: misleading information, hostility, and greater uncertainty about one's own condition.
One caution about searching. A title search on this topic also returns studies that detect depression from social media posts using machine learning, which are a different literature; those were excluded from this page. Pandemic-era studies of media exposure were not excluded, because several are among the most-cited works on the topic — Gao et al. (2020) on 4,872 Chinese citizens in the first days of the outbreak, Geirdal et al. (2021) across four countries, Masciantonio et al. (2021) across four platforms. Read them as evidence about a particular period, not about social media in general.
Main debates
The main debate is about the size and direction of the effect, not its existence. Verduyn, Ybarra and Résibois (2017) answered "it depends on use": passive use relates negatively to subjective wellbeing through social comparison and envy, active use positively through social capital and connectedness, with the first relationship more robust than the second.
Then the reviews multiplied, and disagreed. Valkenburg, Meier and Beyens (2022) found 25 reviews published between 2019 and mid-2021 — seven meta-analyses, nine systematic and nine narrative — and report that most interpreted the association as weak or inconsistent while a few called it substantial and deleterious. Valkenburg (2022) extends the count to 27 and asks why meta-analyses drawing on the same evidence base disagree, particularly about time spent, active use and passive use.
That last point undercuts the active/passive distinction itself. Valkenburg, Van Driel and Beyens (2022) found 40 survey-based studies using 36 different operationalisations of active and passive use, producing 172 associations, most of which did not support the hypothesis. Their conclusion is that time spent is too coarse a measure, and that future studies need to take the content, the sender and the receiver into account — which is the strongest argument available for not designing yet another screen-time survey.
Where recent work is heading
Recent highly cited work is more careful about causality. Braghieri, Levy and Makarin (2022), in the American Economic Review, used the staggered introduction of Facebook across US colleges as a natural experiment, found a negative effect on student mental health and on self-reported academic impairment, and attribute it to unfavourable social comparisons.
Others look at specific mechanisms — body image among adolescent girls (Choukas-Bradley et al., 2022), augmented-reality face filters (Javornik et al., 2022), problematic use and where support comes from (Meshi and Ellithorpe, 2021) — or refuse to treat "social media" as a single exposure. Masciantonio et al. (2021), surveying 1,004 people, found passive Facebook use negatively related to wellbeing through upward social comparison, active Instagram and Twitter use positively related through social support, and TikTok use not related at all. Ostic et al. (2021), with 940 users in Mexico, found an overall positive indirect effect running through bonding and bridging social capital. Both are arguments for naming the platform and the behaviour in your research question rather than the medium.