Cât de târziu s-a separat româna?

The Reichenau Glosses can be used as an indirect snapshot of spoken Romance because glossators replace learned/older Latin words with forms they expect readers to understand.

Evidence: The paper treats the post-700 glossary substitutions as evidence about ongoing lexical replacement.

Inference: The gloss pairs create a quasi-natural experiment between older and later lexical choices.

contested · methodological

In the paper's 66-item lexical tally, Romanian aligns with the later Reichenau-side choice in 49 cases and with the older/classical side in 17.

Evidence: The published calculation is approximately 74% late-side correspondence versus 26% classical-side correspondence.

Inference: The asymmetry is treated as too large to fit a very early isolation model.

contested · empirical

From several lexical-attrition scenarios, Ungureanu estimates proto-Romanian separation from an undifferentiated proto-Romance at roughly 540–600, with uncertainty of about a few decades.

Evidence: The paper models different rates/windows of lexical attrition and converges on a mid-6th-century neighborhood.

Inference: Late shared innovations establish a lower bound for separation.

contested · historical

The published Reichenau 49/66 tally is highly inconsistent with an independent 50/50 Bernoulli null, conditional on that null and independence assumptions.

Evidence: 49/66=0.742; one-sided exact p=5.07386e-05 under p0=.5.

Inference: The descriptive skew merits explanation under the author's simplified model.

editorial_assessment · methodological

The Reichenau paper does not establish that 0.5 is the correct independent-match baseline; at a baseline near 0.639 the observed 49/66 reaches one-sided p=.05.

Evidence: Exact sensitivity threshold p0≈0.639.

Inference: Control Romance languages are needed to estimate the null empirically.

editorial_assessment · methodological

The four published Reichenau dating scenarios are not reproduced consistently by a single transparent linear interpolation using either 49/66 or the paper's rough 2/3 factor.

Evidence: Published scenarios are 400–600→~540, 400–700→600, 500–600→~530, 500–700→~540.

Inference: 540–600 should be treated as a rough author-model chronology, not a calibrated statistical interval.

editorial_assessment · methodological

The 66-item Reichenau sample needs a reproducible inclusion rule because the paper selects words described as relevant to Romanian history from a much larger, heterogeneous gloss tradition.

Evidence: The paper states it discusses only words relevant to Romanian history; historical scholarship describes >1,000 Foerster-numbered selections and around 5,000 glosses in the fuller tradition.

Inference: Selection/ascertainment bias must be tested before interpreting 49/66 as an unbiased sample proportion.

editorial_assessment · methodological

A provisional eight-language Romance control matrix does not resemble a 50/50 process: among 231 strict coded descendant cells, 205 (88.7%) are on the Reichenau/later side.

Evidence: 205/231 strict language×row cells coded late in the provisional controls.

Inference: The fair-coin null is not empirically justified for these lexical contrasts; a scholarly branch-aware control model is required.

provisional_editorial_assessment · methodological

At least three rows in the Reichenau tally are demonstrably non-binary within Romanian itself because inherited descendants of both lexical families survive: oaie/berbec, înjunghia/ucide, and agru/câmp.

Evidence: Romanian dictionary etymologies derive oaie<ovem and berbec<berbex; înjunghia<jugulare and ucide<occidere; agru<agrum and câmp<campus.

Inference: Lexical replacement cannot always be modeled as a binary winner/loser event.

editorial_assessment · methodological