Current Findings

Window: 2017-2020 CAPES article records for 63 CAPES-7 programs in the 12-field comparison set.

Journal metric source: SJR 2024 quartile file from the UKZN mirror. Treat this as a proxy until historical 2017-2020 SJR exports are added.

Main Contrast

Linguistics/Letras differs sharply from the comparison fields on three related outcomes:

SJR indexed: 12.2% in Linguistics/Letras vs. 73.4% in the comparison fields.
English: 8.3% in Linguistics/Letras vs. 82.2% in the comparison fields.
Q1/Q2 with all articles in the denominator: 6.0% in Linguistics/Letras vs. 62.3% in the comparison fields.

The indexing gap is not a nuisance variable. It is substantively meaningful: in journal-based fields, not being present in SJR is itself evidence that a journal is outside the main international indexing system used for quartile comparisons. This does not mean every indexed journal is high quality, but it does mean that treating unmatched articles as simply missing would make the comparison too generous to fields that publish heavily outside indexed venues.

Conditional Quartile Interpretation

Use two denominators:

All-article denominator: asks what share of total journal output lands in indexed Q1/Q2 venues.
Indexed-only denominator: asks, conditional on being in SJR, how strong the matched journals are.

For Linguistics/Letras:

473 of 3,878 article rows match SJR.
233 of 3,878 are Q1/Q2 using the all-article denominator.
233 of 473 matched rows are Q1/Q2 using the indexed-only denominator.

So the first result is the stronger field-level placement result; the second result is useful as a within-indexed-journals diagnostic.

Language Mechanism

English publication is strongly associated with indexing and Q1/Q2 placement.

Linguistics/Letras English articles: 36.0% SJR indexed, 27.3% Q1/Q2.
Linguistics/Letras non-English articles: 10.0% SJR indexed, 4.1% Q1/Q2.
Comparison-field English articles: 81.9% SJR indexed, 71.4% Q1/Q2.
Comparison-field non-English articles: 34.4% SJR indexed, 20.4% Q1/Q2.

Adding English to descriptive logit models reduces the Linguistics/Letras field penalty:

SJR indexing focal-field odds ratio: 0.050 without English, 0.188 with English.
Q1/Q2 focal-field odds ratio: 0.039 without English, 0.160 with English.

This supports a mediation interpretation: field norms shape language and journal-audience choices, which in turn shape SJR indexing and quartile placement.

Brazilian Journal Proxy

Among SJR-indexed articles, SJR journal country provides a partial proxy for Brazilian versus non-Brazilian journals.

Linguistics/Letras indexed Brazilian journals: 13.4% English, 39.1% Q1/Q2.
Linguistics/Letras indexed non-Brazilian journals: 45.2% English, 68.1% Q1/Q2.
Comparison-field indexed Brazilian journals: 49.4% English, 29.2% Q1/Q2.
Comparison-field indexed non-Brazilian journals: 95.5% English, 89.9% Q1/Q2.

This proxy cannot classify non-indexed articles by journal country, so a stronger test would require an external ISSN-country registry or a curated Brazilian-journal list.

Caveats

The current quartile match uses 2024 SJR as a proxy because the official Scimago historical CSV download is blocked from this environment.
Historical SJR files for 2017-2020 should replace the proxy before publication.
CAPES 2013-2016 article-detail datastore is incomplete, so processed outputs currently use 2017-2020.
Professor-level denominators are based on observed faculty authors because filtered CAPES docente datastore calls returned empty files.