## Executive summary

* **Question:** Across the available empirical literature, how is short-form video (SFV) engagement (e.g., TikTok/general SFVs; usage vs “addiction”) *associated* with cognitive functioning and mental health outcomes, and what moderates those associations? ([ResearchGate][1])
* **Headline finding (with numbers):** In meta-analysis, higher SFV engagement was associated with **poorer cognition** (**r = −.34**, 14 studies) and **poorer mental health** (**r = −.21**, 61 studies). Stronger domain-level associations included **inhibitory control** (r = −.41), **attention** (r = −.38), **stress** (r = −.34), and **anxiety** (r = −.33). ([ResearchGate][1])
* **Evidence strength:** **Moderate** for *correlational association patterns* (large total N, sensitivity analyses, limited publication-bias impact), but **weak for causal inference/directionality** (predominantly cross-sectional correlational evidence; very high heterogeneity). ([ResearchGate][1])
* **Implications:** Findings most strongly implicate *problematic/addictive* engagement measures (stronger negative associations than simple duration/usage in several analyses) and point to potential value in “mindful use” supports (e.g., reminders/time limits), while avoiding causal claims. ([ResearchGate][1])
* **Main limitation:** The review cannot establish whether SFV engagement *leads to* poorer cognition/mental health, whether poorer functioning *drives* SFV engagement, or whether both are shaped by third factors; the evidence base is also highly heterogeneous. ([ResearchGate][1])

---

## Key findings

* **Overall cognition association:** Greater SFV engagement correlated with poorer cognition (**r = −.34**, 95% CI [−0.42, −0.26], k=14). → **Paper location:** Results (Cognitive correlates) + Table 1. ([ResearchGate][1])
* **Cognitive domains differ:** Negative associations were **moderate** for **attention** (r = −.38) and **inhibitory control** (r = −.41), **weak** for language (r = −.16), memory (r = −.14), working memory (r = −.21), and **non-significant** for reasoning (r = −.13, p = .281). → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Mental health overall:** Greater SFV engagement correlated with poorer mental health (**r = −.21**, 95% CI [−0.25, −0.17], k=61). → **Location:** Results (Mental health correlates) + Table 1. ([ResearchGate][1])
* **Mental health domains differ:** Stronger negative associations were observed for **stress** (r = −.34) and **anxiety** (r = −.33); **depression** and **loneliness** were weaker (both r = −.23); **sleep** r = −.22; **well-being** r = −.14; **affect** r = −.13. → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Null domains (mental health):** SFV engagement was **not** significantly associated with **body image** (r = −.10, p = .090) or **self-esteem** (r = −.08, p = .193). → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Moderator—how SFV is measured matters:** For cognition, **intensity** showed a very large negative estimate (r = −.55) but was based on **k=1**; **addiction** was moderate (r = −.37) and **duration** was weaker (r = −.20). For mental health, **addiction** was stronger (r = −.32) than **duration** (r = −.10) and **frequency** was non-significant (r = −.05, p = .229). → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Moderator—platform specificity:** For mental health, **general SFV** studies showed a stronger negative association (r = −.27) than **TikTok-specific** studies (r = −.15), with a significant moderator test. → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Moderator—age group:** Youth vs adult samples did **not** significantly differ in effect size for cognition or mental health. → **Location:** Table 1; Results narrative. ([ResearchGate][1])
* **Publication bias impact (as assessed here):** For cognition, trim-and-fill imputed 1 study and adjusted the mean slightly (to r_adj = −.35) without changing interpretation; for mental health, funnel/Egger/trim-and-fill suggested minimal bias impact. → **Location:** Results narrative. ([ResearchGate][1])
* **Heterogeneity is very high:** Cognition **I² ≈ 95.73%** and mental health **I² ≈ 96.43%**, indicating substantial between-study variability not resolved by the tested moderators. → **Location:** Results narrative. ([ResearchGate][1])

---

## What it shows / suggests / does not show

### What it shows

* A consistent **negative correlation** between SFV engagement and (a) cognition and (b) mental health, with domain-specific variation (some outcomes stronger, some null). ([ResearchGate][1])
* Associations tend to be **stronger when SFV engagement is operationalised as “addiction”** versus simpler exposure measures in several analyses. ([ResearchGate][1])

### What it suggests

* Some domains (e.g., **reasoning**, **body image**, **self-esteem**) may be less consistently related to SFV engagement *as currently studied*, and results may depend on content/context and measurement choices. ([ResearchGate][1])
* Cross-platform generalisation is **uncertain** because included platform-specific evidence was effectively limited to TikTok, despite search terms including other platforms. ([ResearchGate][1])

### What it does not show

* **Causality or directionality** (SFV → outcomes vs outcomes → SFV), because the evidence base is largely correlational and cross-sectional. ([ResearchGate][1])
* That reducing SFV use (or changing platform features) will necessarily improve cognition or mental health (intervention efficacy is framed as an open question). ([ResearchGate][1])

---

## Methodology & limitations

**Design**

* Systematic review + random-effects meta-analysis; protocol preregistered on **PROSPERO (CRD42024587550)**; data available on **OSF**. ([ResearchGate][1])

**Search & selection**

* Database search completed **28 Oct 2024** across PsycInfo, PubMed, Scopus, Web of Science; grey literature via ProQuest sources; forward/backward searching and manual journal searches followed in early Nov 2024; no date/language restrictions noted (translation tool used for 2 Mandarin articles). ([ResearchGate][1])
* Dual screening with high agreement (Cohen’s κ **.87** title/abstract; **.94** full text); 2,495 records screened; 71 studies included in review (70 in meta-analysis). ([ResearchGate][1])

**Measures & analysis**

* Effect size metric: correlations (r); group-comparison statistics converted to r when needed; random-effects models; moderators included age group, platform/type, SFV measure, and outcome domains; sensitivity analyses across model type, study quality (MMAT), study design, and declared conflicts. ([ResearchGate][1])

**Main validity threats (and impact)**

* **Directionality/causality:** predominantly correlational evidence means associations may reflect reverse causality and/or confounding. **Impact:** limits decision-making about interventions or policy effects. ([ResearchGate][1])
* **Extreme heterogeneity:** I² >95% for both main syntheses indicates effects vary widely by study/context/measurement. **Impact:** pooled averages may mask important subgroups/conditions. ([ResearchGate][1])
* **Sparse evidence for some domains/moderators:** several domain estimates rely on very few studies (sometimes k=1–2). **Impact:** low power and unstable estimates. ([ResearchGate][1])
* **Platform coverage gap:** included studies largely focused on “general SFV” or TikTok; no evidence for other major SFV products is synthesised here. **Impact:** limits platform-specific recommendations. ([ResearchGate][1])

---

## Evaluation & gaps

**Overall evidence strength:** **Moderate** (for association patterns). The dataset is large, preregistered, and includes sensitivity/bias checks, but causal inference is constrained and heterogeneity is very high. ([ResearchGate][1])

**Biggest alternative explanation:** **Reverse causality + confounding by user state and content exposure** (e.g., stress/anxiety prompting SFV use for relief; content characteristics shaping affect), which the paper explicitly flags as plausible given cross-sectional correlational predominance. ([ResearchGate][1])

**Follow-up studies (grounded in the paper’s identified gaps)**

* **Longitudinal, temporally dense panel studies** with objective usage logs and repeated cognitive/mental health assessments to establish temporal ordering. ([ResearchGate][1])
* **Randomised experiments** manipulating SFV exposure and/or feed characteristics (e.g., pacing, reward density, break prompts) to test causal effects on attention/inhibitory control and stress/anxiety. ([ResearchGate][1])
* **Platform-comparative studies** spanning TikTok, Reels, Shorts (and others) to test whether correlates differ by platform design/ecosystem. ([ResearchGate][1])
* **Content- and motivation-stratified designs** (including qualitative/experience sampling) to test whether associations depend on content type, intent, and emotional response rather than time alone. ([ResearchGate][1])
* **Better-isolation designs for “SFV-specific” effects**, controlling for broader social media use consistently to reduce confounding. ([ResearchGate][1])

---

## Questions answered / raised

### Answered (with answers)

* **Is SFV engagement associated with cognitive performance?** Yes—negative association overall (r = −.34; k=14). ([ResearchGate][1])
* **Which cognitive domains show the strongest associations?** Attention and inhibitory control are strongest in this synthesis (r ≈ −.38 to −.41). ([ResearchGate][1])
* **Is SFV engagement associated with mental health?** Yes—negative association overall (r = −.21; k=61). ([ResearchGate][1])
* **Which mental health domains are most strongly associated?** Stress and anxiety (r ≈ −.34 and −.33). ([ResearchGate][1])
* **Are body image and self-esteem associated with SFV engagement in this meta-analysis?** No statistically significant associations were found. ([ResearchGate][1])
* **Do effects differ between youth and adult samples?** Not in the tested moderator analyses. ([ResearchGate][1])

### Raised (new questions the paper surfaces)

* **Directionality:** Does SFV engagement worsen stress/anxiety, or do stressed/anxious individuals preferentially engage with SFVs (or both)? ([ResearchGate][1])
* **Content dependence:** Which content types (and algorithmic delivery patterns) drive negative vs neutral/beneficial associations? ([ResearchGate][1])
* **Platform specificity:** Do Reels/Shorts differ from TikTok/general SFV in cognitive and mental health correlates? ([ResearchGate][1])
* **Measurement validity:** How much of the stronger “addiction” association reflects construct overlap with distress/impulsivity rather than SFV behaviour per se? ([ResearchGate][1])
* **Boundary conditions:** For whom (e.g., by baseline activity, traits, motivations) are associations larger/smaller, and under what usage patterns? ([ResearchGate][1])
* **Underpowered domains:** Are reasoning/body image/self-esteem truly near-zero, or are nulls driven by sparse or heterogeneous evidence? ([ResearchGate][1])

[1]: https://www.researchgate.net/publication/397584857_Feeds_feelings_and_focus_A_systematic_review_and_meta-analysis_examining_the_cognitive_and_mental_health_correlates_of_short-form_video_use/fulltext/69162bed9d514a24b432284e/Feeds-feelings-and-focus-A-systematic-review-and-meta-analysis-examining-the-cognitive-and-mental-health-correlates-of-short-form-video-use.pdf "Feeds, Feelings, and Focus: A Systematic Review and Meta-Analysis Examining the Cognitive and Mental Health Correlates of Short-Form Video Use"