Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/mikhaeelatefrizk/affect-labeling-review

Systematic review and meta-analysis of affect labeling (Lieberman et al. 2007 paradigm). Pre-registered protocol, real random-effects meta-analysis (k=9), PRISMA 2020, RoB 2 / ROBINS-I, ~14,000-word manuscript. Open data and code.
https://github.com/mikhaeelatefrizk/affect-labeling-review

affect-labeling amygdala clinical-psychology emotion-regulation meta-analysis neuroscience open-science prefrontal-cortex preregistration prisma-2020 psychology psychophysiology replication-crisis skin-conductance systematic-review

Last synced: 28 days ago
JSON representation

Systematic review and meta-analysis of affect labeling (Lieberman et al. 2007 paradigm). Pre-registered protocol, real random-effects meta-analysis (k=9), PRISMA 2020, RoB 2 / ROBINS-I, ~14,000-word manuscript. Open data and code.

Awesome Lists containing this project

README 

          
# Putting feelings into words: a systematic review and meta-analysis of affect labeling



[![License: code MIT](https://img.shields.io/badge/code-MIT-blue.svg)](LICENSE-CODE)

[![License: manuscript CC-BY-4.0](https://img.shields.io/badge/manuscript-CC--BY--4.0-lightgrey.svg)](LICENSE-MANUSCRIPT)

[![License: data CC-BY-4.0](https://img.shields.io/badge/data-CC--BY--4.0-lightgrey.svg)](LICENSE-DATA)

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.20109595.svg)](https://doi.org/10.5281/zenodo.20109595)

[![Cite this repository](https://img.shields.io/badge/cite-CITATION.cff-green.svg)](CITATION.cff)

[![ORCID](https://img.shields.io/badge/ORCID-0009--0006--1069--9558-A6CE39?logo=orcid&logoColor=white)](https://orcid.org/0009-0006-1069-9558)



This repository is the full open-research package for a pre-registered (PROSPERO) systematic review and random-effects meta-analysis of **affect labeling** — the psychological and neuroscientific phenomenon, originating in Lieberman et al. (2007), in which putting feelings into words attenuates emotional responses.



> **Status: Preprint / working paper. Not yet peer-reviewed.** All findings are provisional and subject to revision pending external review.



## Headline result



Random-effects meta-analysis of nine peripheral-physiology effect sizes from seven independent studies:



- **Pooled Hedges' g = −0.43, 95% CI [−0.68, −0.18], p < .001**

- 95% prediction interval [−1.13, +0.27] **crosses zero**

- I² = 48.3%, τ² = 0.070



Pre-specified lab-stratified moderator analysis:



- **UCLA Lieberman/Craske axis only:** g = **−0.74**, 95% CI [−1.02, −0.47], k = 5, I² = 0%

- **Independent laboratories only:** g = **−0.13**, 95% CI [−0.41, +0.14], k = 4, I² = 27.8% — non-significant; CI crosses zero



The 0.61 g-unit gap between lab strata is larger than most moderators in the published literature and is consistent with structural author non-independence rather than incidental variation.



## For data scientists and ML researchers



The screening corpus is published as a labeled dataset suitable for training and evaluating AI-assisted SLR screening systems (see [`data/screening/`](data/screening/)). Quick orientation:



| You want… | File | Note |

|-----------|------|------|

| The 100 included papers (positive class) | `data/screening/included_papers.csv` | Generated from `references.bib` + manuscript |

| The candidate corpus (~1,800 papers) with `included` 0/1 labels | `data/screening/derived_screening_log.csv` | **The training file.** Re-derived from PubMed using the pre-registered query |

| The pre-registered query and search strategy | [`data/searches/search_strategy.md`](data/searches/search_strategy.md) | PRISMA-S compliant |

| The exclusion-reason taxonomy | [`data/exclusion_reason_codebook.md`](data/exclusion_reason_codebook.md) | The 6 full-text exclusion codes used in the original review |

| Aggregate PRISMA counts | [`prisma/prisma_counts.csv`](prisma/prisma_counts.csv) | Structured form of the published flow |



**Important honesty up front.** The derived screening log is a *re-derivation*, not the original per-paper screening record (which was not preserved in shareable form). It collapses the title/abstract vs. full-text decision distinction into a single binary label, and exclusion reasons are aggregated rather than per-paper. See [`data/screening/README.md`](data/screening/README.md) for the full caveats.



## Data Availability



All data needed to reproduce, re-analyze, or extend this review is in this repository under [`LICENSE-DATA`](LICENSE-DATA) (CC-BY-4.0). No external archive is currently used; the [`.zenodo.json`](.zenodo.json) configures a DOI mint on the next tagged release.



| Layer | Where | Source of truth? |

|-------|-------|------------------|

| Manuscript and figures | `manuscript/`, `figures/` | Yes |

| Pre-registration | `prereg/PROSPERO_preregistration.md` | Yes (canonical inclusion/exclusion criteria) |

| PRISMA flow and counts | `prisma/prisma_counts.csv` (structured), `prisma/prisma_counts.txt` (legacy) | Yes |

| Effect sizes | `meta-analysis/extracted_effect_sizes.csv` | Yes |

| Risk-of-bias assessments | `supplementary/risk_of_bias.csv` | Yes |

| Search strategy | `data/searches/` | Yes (restructured from PROSPERO) |

| Included papers list | `data/screening/included_papers.csv` | Derived from `references.bib` + manuscript |

| Candidate corpus + labels | `data/screening/derived_screening_log.csv` | Derived from PubMed re-query |



**What is *not* available.** The original per-paper screening decisions (1,571 title/abstract decisions and 282 full-text decisions) were not preserved in a shareable form. Derived labels are the closest reproducible substitute. See [`data/screening/README.md`](data/screening/README.md).



## Documentation guide



This repository ships with extended documentation under [`docs/`](docs/) for readers who want to *use* the package — not just cite it. Pick your path:



| You are… | Start here |

|---|---|

| Anyone with a quick question (year filter? where's the included list?) | [`docs/faq.md`](docs/faq.md) |

| A reader who wants the gist | [`README.md`](README.md) (this file) → [`manuscript/manuscript.md`](manuscript/manuscript.md) |

| A peer reviewer or thesis committee | [`prisma/PRISMA_2020_checklist.md`](prisma/PRISMA_2020_checklist.md) → [`supplementary/risk_of_bias_explanation.md`](supplementary/risk_of_bias_explanation.md) → [`docs/methodology-deep-dive.md`](docs/methodology-deep-dive.md) |

| A reproducer | [`docs/reproducibility-guide.md`](docs/reproducibility-guide.md) → run `make all` → [`docs/troubleshooting.md`](docs/troubleshooting.md) if anything fails |

| An ML / NLP researcher building a screening model | [`docs/for-ml-researchers.md`](docs/for-ml-researchers.md) → [`data/screening/derived_screening_log.csv`](data/screening/derived_screening_log.csv) |

| A master's student building an AI screening pipeline | [`docs/for-masters-students.md`](docs/for-masters-students.md) → [`docs/for-ml-researchers.md`](docs/for-ml-researchers.md) |

| Someone unfamiliar with SLR terminology | [`docs/glossary.md`](docs/glossary.md) (5 minutes) |

| Someone wanting to contribute back | [`docs/extending-the-corpus.md`](docs/extending-the-corpus.md) → [`CONTRIBUTING.md`](CONTRIBUTING.md) |

| A search-strategy peer reviewer | [`data/searches/PRISMA-S_checklist.md`](data/searches/PRISMA-S_checklist.md) |



## Repository structure



```

affect-labeling-review/

├── README.md                              ← this file

├── CHANGELOG.md                           ← versioned change history

├── LICENSE                                ← multi-license pointer

├── LICENSE-CODE                           ← MIT (source files)

├── LICENSE-MANUSCRIPT                     ← CC-BY-4.0 (manuscript + figures)

├── LICENSE-DATA                           ← CC-BY-4.0 (data files)

├── CITATION.cff                           ← machine-readable citation

├── .zenodo.json                           ← DOI metadata (Zenodo)

├── CONTRIBUTING.md                        ← how to report errata, propose corrections

├── CODE_OF_CONDUCT.md                     ← Contributor Covenant 2.1

├── SECURITY.md                            ← vulnerability disclosure policy

├── .editorconfig                          ← cross-editor code-style consistency

├── Makefile                               ← reproducibility entry point — run `make all`

├── requirements.txt                       ← pinned Python deps

├── environment.yml                        ← conda alternative

├── references.bib                         ← BibTeX for all references

│

├── manuscript/                            ← the manuscript and its README

│   └── manuscript.md                      ← full ~14,000-word paper

│

├── prereg/                                ← pre-registration

│   └── PROSPERO_preregistration.md        ← PROSPERO-compatible structured protocol

│

├── meta-analysis/                         ← random-effects meta-analysis

│   ├── run_meta_analysis.py               ← analysis code

│   ├── extracted_effect_sizes.csv         ← effect-size dataset (canonical)

│   ├── leave_one_out.csv                  ← LOO sensitivity output

│   └── results_summary.txt                ← plain-text summary

│

├── prisma/                                ← PRISMA 2020 flow + reporting

│   ├── build_prisma.py                    ← PRISMA flow generator

│   ├── prisma_counts.csv                  ← structured counts (canonical)

│   ├── prisma_counts.txt                  ← legacy human-readable counts

│   └── PRISMA_2020_checklist.md           ← 27-item reporting checklist

│

├── supplementary/                         ← risk-of-bias artifacts

│   ├── risk_of_bias.csv                   ← RoB 2 / ROBINS-I judgments

│   ├── build_rob_figure.py                ← traffic-light figure code

│   └── risk_of_bias_explanation.md        ← per-study rationale narrative

│

├── data/                                  ← labeled datasets (CC-BY-4.0)

│   ├── README.md                          ← dataset index + data dictionary

│   ├── exclusion_reason_codebook.md       ← 6 full-text exclusion codes

│   ├── QUALITY_REPORT.md                  ← canonical-vs-derived comparison

│   ├── searches/                          ← PRISMA-S search strategy

│   │   ├── search_strategy.md/.csv        ← canonical query, dates, hits

│   │   └── PRISMA-S_checklist.md          ← 16-item search-reporting checklist

│   └── screening/                         ← screening log (schema, template, derived)

│       ├── screening_log.schema.json      ← JSON Schema 2020-12

│       ├── screening_log.template.csv     ← empty template + 1 example row

│       ├── included_papers.csv            ← 22 confidently-identifiable of 100 includes

│       ├── all_references.csv             ← every references.bib entry + confirmed-include flag + manuscript-mention count

│       ├── derived_corpus.csv             ← 3,892 PubMed-derived records

│       └── derived_screening_log.csv      ← labelled corpus (PU framing)

│

├── scripts/                               ← derivation + validation tooling

│   ├── extract_included_list.py           ← parse references.bib → included_papers.csv

│   ├── build_derived_corpus.py            ← re-query PubMed → derived_corpus.csv

│   ├── build_derived_screening_log.py     ← join → derived_screening_log.csv

│   ├── build_quality_report.py            ← QUALITY_REPORT.md

│   └── validate_screening_log.py          ← schema + reconciliation check

│

├── docs/                                  ← extended documentation for reproducers and reusers

│   ├── README.md                          ← reading orders + table of contents

│   ├── faq.md                             ← year filter, where's the included list, how to label a re-run corpus

│   ├── reproducibility-guide.md           ← clone → byte-identical outputs

│   ├── for-ml-researchers.md              ← detailed PU-learning guide

│   ├── for-masters-students.md            ← concrete onboarding letter for ML-screening thesis work

│   ├── methodology-deep-dive.md           ← analytic-choice justifications

│   ├── glossary.md                        ← ~50 SLR / meta-analysis / ML terms

│   ├── troubleshooting.md                 ← top 10 reproducer issues + fixes

│   └── extending-the-corpus.md            ← second-coder + database-broadening guide

│

├── figures/                               ← rendered outputs (PNG + PDF)

│   ├── prisma_flow.png / .pdf             ← PRISMA 2020 flow diagram

│   ├── rob_summary.png / .pdf             ← risk-of-bias traffic light

│   ├── forest_plot.png / .pdf             ← random-effects forest plot

│   └── funnel_plot.png / .pdf             ← funnel plot

│

└── .github/

    ├── workflows/ci.yml                   ← reproducibility CI (runs on every push + PR)

    ├── ISSUE_TEMPLATE/                    ← erratum, repro, data-correction forms

    └── PULL_REQUEST_TEMPLATE.md

```



Every directory has a `README.md` orienting you to its contents and regeneration command.



## Reproducing the analyses



```bash

git clone https://github.com/mikhaeelatefrizk/affect-labeling-review.git

cd affect-labeling-review

make install       # pinned Python deps (Python 3.11)

make all           # regenerate every output: figures, counts, derived dataset

make verify        # assert outputs match the committed versions

```



Without `make` (Windows without WSL):



```powershell

pip install -r requirements.txt

python meta-analysis/run_meta_analysis.py

python prisma/build_prisma.py

python supplementary/build_rob_figure.py

python scripts/extract_included_list.py

python scripts/build_derived_corpus.py

python scripts/build_derived_screening_log.py

python scripts/build_quality_report.py

python scripts/validate_screening_log.py

```



`make all` from a fresh clone with the pinned dependency set (Python 3.11.x, see [`requirements.txt`](requirements.txt)) reproduces every output byte-for-byte. Drift on a different version stack is reported but not necessarily an error — see [`Makefile`](Makefile) target `sha` for a SHA-256 receipt.



## Re-deriving the screening corpus



`make derive` re-queries PubMed via the NCBI E-utilities API using the pre-registered Boolean query, fetches each candidate's metadata, joins against the included-papers list, and writes the labeled corpus to `data/screening/derived_screening_log.csv`. The whole derivation is deterministic given a PubMed snapshot.



Because PubMed continues to grow, today's hit count will exceed the canonical 1,842 records reported in the original PRISMA flow. This drift is reported in [`data/QUALITY_REPORT.md`](data/QUALITY_REPORT.md) (regenerated by `make quality`) and is informational, not error.



## What's verified, what's flagged



The reference list was verified against PubMed, the journal of record, and where available the original published PDFs. **Five citations were corrected** during verification, and **one fabricated reference was removed**:



- Removed: Isaacowitz & Eldesouky (2024) — could not be located in any database; closest real entity is a 2023 conference symposium.

- Corrected: "Wong et al. (2022)" → Plaisted, Waite, & Creswell (2022), *Behaviour Research and Therapy*, 148, 103997. This is a key **null finding** and the only adequately powered adolescent RCT.

- Corrected: Vlasenko et al. (2021) coauthors → Vlasenko, V. V., Rogers, E. G., & Waugh, C. E. (Wake Forest).

- Corrected: Vives et al. (2021) title → "Foreign Language Processing Undermines Affect Labeling."

- Corrected: Givon, Meiran, & Goldenberg (2025) for the *Trends in Cognitive Sciences* piece (was misattributed).

- Several DOIs and article numbers corrected.



The Burklund et al. (2024) PTSD pilot trial is flagged for a **commercial conflict of interest**: the lead author is an independent contractor at NeuroGen Technologies Inc., a private company developing affect-labeling-based PTSD interventions. This disclosure is also reflected in the risk-of-bias assessment as "high."



## Limitations explicitly acknowledged in the manuscript



1. Many original fMRI studies did not report d or g, precluding fully harmonized quantitative synthesis without re-analysis of original data.

2. Single-laboratory concentration in the affect labeling fMRI literature limits independence of the supportive evidence base; addressed by lab-stratified moderator analysis but the underlying problem requires direct multisite replication.

3. The dissociation between physiological and self-report measures means effect-size estimates depend on outcome choice in ways flagged but not fully resolved.

4. The meta-analysis includes nine effect sizes from seven studies — adequate for detecting a moderate pooled effect but underpowered for many moderator analyses.

5. Inclusion criteria excluded studies in non-English without translation.

6. **Screening was performed by a single coder.** Inter-rater agreement statistics are not available. The derived screening log (`data/screening/derived_screening_log.csv`) reproduces a binary label per paper from public sources, but the original per-paper screening decisions and exclusion reasons were not preserved.



## How to cite



The canonical citation is in [`CITATION.cff`](CITATION.cff) (machine-readable). Plain text:



```

Wahba, M. A. R. (2026). Putting feelings into words: A systematic review

and meta-analysis of affect labeling (v1.0.0).

Zenodo. https://doi.org/10.5281/zenodo.20109595

```



The repository has a permanent Zenodo DOI: [10.5281/zenodo.20109595](https://doi.org/10.5281/zenodo.20109595). GitHub also renders a "Cite this repository" button on the project page using `CITATION.cff`.



## License



Three licenses by content category:



- **Code** (`*.py`, `Makefile`, CI configs, schemas): MIT — see [`LICENSE-CODE`](LICENSE-CODE).

- **Manuscript and figures**: CC-BY-4.0 — see [`LICENSE-MANUSCRIPT`](LICENSE-MANUSCRIPT).

- **Data files**: CC-BY-4.0 — see [`LICENSE-DATA`](LICENSE-DATA).



See [`LICENSE`](LICENSE) for the at-a-glance map.



## Contributing



Errata, reproducibility issues, and data corrections are welcome. See [`CONTRIBUTING.md`](CONTRIBUTING.md) for the workflow and the issue templates under [`.github/ISSUE_TEMPLATE/`](.github/ISSUE_TEMPLATE/).



## Contact



Open an issue on this repository, or use the email in [`CITATION.cff`](CITATION.cff) for matters that aren't suitable for a public issue. ORCID: [0009-0006-1069-9558](https://orcid.org/0009-0006-1069-9558).