The morphology ↔ word-order-freedom compensation trade-off under genuine areal control: $\ge 5$ macroareas and $\ge 25$ families

Test whether, after controlling for both genealogical (family) and areal (macroarea) non-independence, a negative partial correlation holds between a morphological-richness measure and a word-order-freedom measure across a typologically and areally broad language sample. Requirements: a sample of $\ge 25$ language families spanning $\ge 5$ of Dryer's macroareas (admitting low-resource UD treebanks and/or WALS/Grambank features to reach breadth); morphological richness and word-order freedom each operationalised from a stated data source (e.g. UD-derived subject/object-order entropy or head-direction entropy for freedom; a morphological-feature count or form/lemma ratio for richness); and a partial correlation or mixed-effects model with family and macroarea as controls. Report the partial correlation coefficient (or standardised mixed-model slope) with its sign, effect size, and confidence interval.

Acceptance. FULLY RESOLVES: a reproducible analysis (script + language list with family and macroarea coding and the two feature values per language) computing the partial correlation between the richness and freedom measures under simultaneous family + macroarea control, meeting $\ge 25$ families across $\ge 5$ macroareas, reporting the coefficient's sign, effect size, and 95% CI; a CI excluding zero (either sign) at that breadth resolves the direction at scale. PARTIAL: the reproducible pipeline meeting the breadth requirement and reporting the estimate with CI even if the CI includes zero, or a documented sensitivity analysis across measure choices. Metric: partial correlation coefficient (or standardised mixed-model slope) with 95% CI, under family + area control; state the exact richness and freedom operationalisations, the control structure, and any language-inclusion/exclusion rule.