Sources & method

Sources

1. Logan, Lively & Pisoni 1991 (JASA 89(2) 874–886, doi:10.1121/1.1894649, PMID 2016438, PMCID PMC3518834): the founding identification program (multiple natural exemplars, pretest–posttest, variability importance). 4 independent citing records; AIP page 403s automated clients (record verified, page unread). DOI
2. Zhang, Cheng & Zhang 2021 (JSLHR 64(12) 4802–4825, doi:10.1044/2021_JSLHR-21-00181, PMID 34763529): multi-talker immediate-gains advantage g=.28 (k=13) + generalization to untrained talkers g=.36 (k=9; large full-estimate generalization/retention), with heterogeneity + publication-bias sensitivity. OSF + SemanticScholar + figshare; PubMed page cookie-walled (record verified, abstract via secondaries). DOI
3. Uchihara, Karas & Thomson 2025 (SSLA 47(3) 794–827, doi:10.1017/S0272263125100879, online 04.06.2025, CUP): canonical = talker + context variability + feedback; pre-post g=0.92 (k=96), treated-control g=0.67 (k=32); retention confirmed; generalization “to some extent” (complete vs partial defined); moderators incl. talkers/tasks/phones. Full text read (1.5MB). DOI · CUP full text
4. Recasens & Sánchez-Miret 2025 (Languages 10(1):12, 10 pp. PDF via UAB repository): 6 talkers (3M+3F, mixed Bucharest-lifelong/Moldavian-born/Spain-resident), read speech, 7 stressed vowels × controlled consonantal contexts (4 place conditions) × real + nonsense words, F1/F2 + duration + stats. Scoped finding: [ɨ]/[ə] distinct along F1 in careful stressed speech — no exit from the scope. Re-fetched 2026-09-17, 3.1MB, sha256 4119339d…. PDF
5. Pointers & context (not design evidence): Sarlin-2014 vowel chart (via the Wikipedia Romanian-phonology pointer, tertiary; /ɨ//ə/ in place); De Gruyter pronunciation note (learner comparison, sheet pointer); Renwick 2014 (the reason /ə/ is a broad convention); HansPub RO→CN paper (reverse-direction pointer); 2024 perception→production meta (the reason for perception-only scope). Wikipedia Romanian phonology
6. Overlap baseline (verified live 2026-09-17): doua-note (vowel formant table i/e/â-î/ă/a/o/u + reference words fir/tren/cânt/casă/pas/nord/nor + method note — linked to vowels only); dexonline (spelling-and-meaning channel); auzi (thresholds instrument, in preparation, different mechanism). doua-note · dexonline
7. Never cited: Chinese-equivalence assertions, efficacy/transfer promises, progress reporting, waveform acoustics, video/marketing content, “merge” diagnostics, pinyin analogies, expired methodologies as current.

Build journal (2026-09-17)

• All four design records re-verified: DOIs resolve (two publishers 403 robots, marked), Recasens re-fetched identical, CUP full text 200.
• Siblings verified live: doua-note words + formants + note all present; dexonline works; auzi unpublished (404) → mechanism mentioned, no dead link.
• All five recording gates (bank/transcriptions/phonetician/validation/ceiling) unpassed → test fully frozen; shipping sheets + paradigm + method (the no-script baseline promoted to v1).
• CN read: terms, buttons and metadata on one convention; no pinyin, no Chinese-phonology assertions, no efficacy language.

Maintenance & corrections

Pairs reviewed annually; recordings re-done only on protocol change; HVPT literature re-read on new metas; aggregate data monitored as outcomes, curriculum-revision proposals logged; any audio/pair change = new manifest version + note. Any contrast with <4 valid talkers or a failed ceiling → withdrawn. Corrections via the site contact page, audio evidence with originals attached.