# Extraction report: grindrodetalndthirdpersonknowledge

## Summary
- Pipeline outputs used (preferred sources):
  - Extraction index/status: `papers/grindrodetalndthirdpersonknowledge/out/bundle.json`
  - Study design + participants + scenario wording: `papers/grindrodetalndthirdpersonknowledge/out/fulltext.md` and `papers/grindrodetalndthirdpersonknowledge/out/text.txt`
  - Numeric results used for primary effect sizes (first-block between-subject emulation):
    - Experiment 1 Table 7: `papers/grindrodetalndthirdpersonknowledge/out/tables/camelot_stream_p15_t3.csv`
    - Experiment 2 Table 8 `When presented first` columns: `papers/grindrodetalndthirdpersonknowledge/out/tables/camelot_stream_p19_t4.csv`
  - Additional supplemental within-subject results now included in the computation notebook:
    - Experiment 1 Table 6: `papers/grindrodetalndthirdpersonknowledge/out/tables/camelot_stream_p12_t2.csv`
    - Experiment 2 Table 8 Overall columns: `papers/grindrodetalndthirdpersonknowledge/out/tables/camelot_stream_p19_t4.csv`
    - Paired-samples `t` tests and omnibus within-subject analyses: `papers/grindrodetalndthirdpersonknowledge/out/fulltext.md`
- Output YAML: `papers/grindrodetalndthirdpersonknowledge/grindrodetalndthirdpersonknowledge.yaml`
- Computations:
  - The YAML now contains both designs for downstream synthesis:
    - the original first-block between-subject estimates, and
    - additional all-response within-subject estimates computed with `within_smcrp_t` (paired `t` used to recover within-person `r`).
  - `papers/grindrodetalndthirdpersonknowledge/analysis/effect_sizes.qmd` audits both sets of computations.
  - The Quarto audit also verifies that the paper's reported repeated-measures `d` values are recoverable from `t / sqrt(n)` up to rounding tolerance, which indicates that the paper is using a different repeated-measures standardization than the project metric.

## Manual interventions / non-pipeline sources (required disclosure)
- External public bibliographic records were consulted to confirm publication metadata:
  - University of Reading CentAUR record for the published article.
  - University of Manchester Research Explorer record (lists DOI `10.1111/mila.12196` and publication date July 5, 2018).
- The PDF (and rendered page images) were not consulted; all effect-size inputs still come from pipeline outputs.

## Notes / potential limitations
- This paper’s “context” manipulation bundles multiple cues (higher stakes for the *ascriber* + explicit error-possibility/skeptical-pressure cues). The extracted effects therefore represent the combined context shift, not a pure stakes-only manipulation.
- The local manuscript-style header in `out/fulltext.md` still contains the line “Published on February 10, 2010”, which is bibliographically false and was ignored in favor of the public publication records.
- The extraction now codes both the six first-block between-subject effects and six supplemental all-response within-subject effects in the YAML.
- The computation notebook audits both sets of effects, with the within-subject effects using the project's SMCRP convention.
- The paper’s own repeated-measures `d` values are comparable in magnitude to the project estimates, but they are not numerically identical because the paper appears to report `d_z = t / sqrt(n)` whereas the project uses the pooled-SD repeated-measures metric implemented in `docs/quarto_effect_template.qmd`.

## Human input needed (clearly delimited fields)
- None at present for publication metadata; DOI and published status were confirmed from public records.

## Validation / rendering
- YAML schema validation: `papers/grindrodetalndthirdpersonknowledge/grindrodetalndthirdpersonknowledge.yaml` validates against `docs/stakes_meta_schema.json` (via `jsonschema`).
- Effect-size notebook render: `papers/grindrodetalndthirdpersonknowledge/analysis/effect_sizes.html` rendered successfully via `quarto render`.
- Extraction HTML render: `papers/grindrodetalndthirdpersonknowledge/grindrodetalndthirdpersonknowledge.html` rendered successfully via `conda run -n meta-extract meta-extract render-extraction grindrodetalndthirdpersonknowledge`.