| name | lang-tables-figures |
| description | Use when preparing the exhibits of a Language (LSA) manuscript — numbered examples, Leipzig-convention interlinear glosses, IPA transcription, tableaux/trees, and quantitative tables and figures. Language has specific conventions for linguistic examples that reviewers and typesetters enforce. Prepares exhibits; it does not generate data. |
Examples, Tables & Figures (lang-tables-figures)
In a linguistics paper the numbered example is the primary exhibit, and Language has firm
conventions for how examples, glosses, transcriptions, and quantitative displays are formatted. Getting
these right lets the paper read cleanly under double-anonymous review and pass into production without a
copyediting fight. Verify current specifics against the Language Style Sheet and the CUP/Overleaf
template before submission.
When to trigger
- Formatting numbered examples, interlinear glosses, trees, tableaux, or IPA
- Building quantitative tables and figures from a model
- A reader said an example was mis-glossed, unreadable, or unnumbered
- Preparing figure captions and accessibility text
Numbered examples (the core exhibit)
- Number every example consecutively —
(1), (2), (3) — with sub-parts lettered (1a),
(1b); reference them in the text by number, never "the example above."
- Give object-language data in italics (or the journal's current convention), followed by the gloss
and a free translation in single quotes.
- Cite a source for each non-elicited token; mark elicited judgments (
*, ?, #) consistently.
Interlinear glossing (Leipzig Glossing Rules)
- Use the standard three-line format: object language / morpheme-by-morpheme gloss / free
translation, with the object and gloss lines word-aligned and morpheme boundaries matched
(
- for affixes, = for clitics, . for a single-word gloss of a portmanteau).
- Use small-caps grammatical labels (
PST, 3SG, ERG, NMLZ) from the Leipzig list; define any
non-standard abbreviation in a list of glossing abbreviations.
- Keep the number of morphemes on the gloss line equal to the object line; misalignment is the most
common example error reviewers flag.
Transcription (IPA)
- Use IPA for phonetic/phonological data; state the transcription conventions and any language-
specific symbols. Use a Unicode IPA font; do not fake symbols with look-alikes.
- Distinguish phonemic
/ / from phonetic [ ] transcription consistently.
Trees, tableaux, and formal displays
- Syntactic trees, OT tableaux, autosegmental diagrams, and derivations are numbered examples too;
keep them legible in grayscale and vector where possible.
- For an OT tableau, mark the winner (
☞), violations (*), and fatal violations (*!) per convention.
Quantitative tables and figures
- Lead with the exhibit that carries the main generalization; do not bury it among model dumps.
- Report effect sizes and intervals in tables; name the model and clustering in the note. Prefer a
plot (partial effects, by-speaker variation) over a wall of coefficients where it communicates better.
- Colorblind-safe, legible in grayscale; provide alt text / a descriptive caption for every figure.
Referee/production conformance check
| Slip reviewers and typesetters catch | Language-appropriate fix |
|---|
| Gloss line misaligned with object line | match morpheme counts; align word-by-word |
| Non-standard glossing labels undefined | use Leipzig labels; list abbreviations |
| Examples referenced as "above/below" | number and cite by number |
| IPA faked with look-alike characters | use a Unicode IPA font |
| Coefficient dump where a plot would show the pattern | promote a partial-effects figure |
Calibration (exhibits as argument, hedged)
Craft heuristics fitting Language's conventions, not graded rules; confirm against the Style Sheet.
Because the numbered example is the evidence, a clean, correctly aligned gloss does more persuasive
work than any prose summary — a reader who can parse your example accepts the generalization. Illustrative:
a paper presents a key portmanteau affix with a mismatched gloss line and undefined labels; the fix
re-aligns the morphemes, glosses the portmanteau with . (e.g., 3SG.PST), adds the abbreviation list,
and promotes the by-speaker variation from a coefficient table to a figure with alt text.
Execution bridge (StatsPAI / Stata MCP)
Numbered examples, glosses and tableaux are typeset by hand and stay that way. The
quantitative exhibits should not be. Full map: execution-with-mcp.
- Model tables:
etable from the fitted result_id, so the coefficients in a table
and the effect sizes in the prose cannot drift apart between revisions.
- Predicted values:
margins_at_plot / marginsplot for interaction and
by-condition figures, with the interval and the response scale carried from the model
rather than redrawn by eye.
- Every caption names the model class and the random-effects structure — a
double-anonymous reviewer across subfields reads the caption before the method section.
Anti-patterns
- Gloss lines whose morpheme count does not match the object line
- Non-Leipzig or undefined grammatical abbreviations
- Examples that are unnumbered or cross-referenced by position, not number
- Faked IPA, or mixed phonemic/phonetic brackets used inconsistently
- A wall of coefficients where one figure would carry the generalization; missing figure alt text
Output format
【Main exhibit】the example/figure carrying the central generalization
【Numbering】examples numbered consecutively, sub-parts lettered? [Y/N]
【Glossing】Leipzig three-line, aligned, labels defined? [Y/N]
【Transcription】IPA correct, brackets consistent? [Y/N/NA]
【Quant displays】effect sizes + model note + alt text? [Y/N/NA]
【Next】lang-writing-style
Supplementary resources
