📄 On the excrecent Middle English

Summary

Anna Hebda (SELIM, 2019) examines the unetymological p that sometimes appears between /m/ and a following dental in Middle English, as in dempd or empti. Standard handbooks mention it only in passing and attribute it to mistimed articulatory gestures. They do not say how common it was across dialects or periods. The study uses the online MED and OED for attestations. Its counts, however, come only from Early and Late ME texts with secure dialect localisation, covering the Northern, East Midland, West Midland, Southwest and Southeast areas.

Every word in which /m/ came to stand before d, t, s, θ or n was examined. The words were split into two groups: /m/ before an oral dental, and /m/ before another nasal. Forms without p clearly dominate everywhere except the West Midlands. Sporadic insertion shows no consistent pattern in time or place and looks like scribal habit, for example a way to make minim strokes easier to read. Only empty acquired p permanently. The paper therefore questions the purely articulatory account, especially for /m/+/n/ clusters.

Key points

• The quantitative corpus spans five ME dialect areas and two periods, Early ME and Late ME. Texts include the Ormulum, Trinity Homilies, Ayenbite of Inwyt, Gawain, Pearl and Rolle. The MED and OED serve for attestation, not for the statistics.
• The target words include past forms of dēmen, drēmen and nemnen, along with emptī, dēmester, somnour, hērsumnesse, him-self and sŏm-tīme. They fall into two groups: /m/ before an oral dental and /m/ before a nasal.
• Sporadic insertion touched nearly every item except emptī and its derivatives. The earliest cases come from East Midland, West Midland and Southwestern manuscripts.
• Empty is the only word with permanent epenthesis:
  • The earliest ME case is in St. Katherine (SW).
  • The form later spreads to the Midlands.
  • No form of the word occurs in the Northern texts.
  • The last p-less form found is in Brevint's Saul at Endor, cited from the OED.
• The localized-text counts for empty are very small, and the MED shows more forms with p than without. The author therefore treats the change as well advanced in Late ME, despite what the corpus table alone suggests.

Interesting ideas

• I1. Frequencies are computed only from texts of known provenance, while dictionaries are used just to attest that forms exist, which keeps uncertain sources from biasing the statistics.
• I2. Splitting the items by environment (/m/ before an oral consonant versus /m/ before a nasal) shows that the textbook mechanism cannot apply to one group, which forces a search for another explanation.
• I3. Spelling variation can come from writing practice, such as helping readers separate minim strokes, rather than from pronunciation.
• I4. Separating sporadic variants from permanent lexical change avoids treating occasional scribal forms as the norm.
• I5. The paper weighs a tiny corpus table against a larger dictionary sample and refuses to draw conclusions from single-digit counts.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (learn): Silver-data statistics could count only sentences from sources with verified origin, using broader web or LLM material merely to confirm that a construction occurs.
• I2 (learn): Evaluations can include a subset where the assumed cause of a repair rule logically cannot operate, serving as a negative control on whether the rule really captures its stated reason.
• I3 (tokens): Archaic or noisy spellings like dempd, emty or dampne should be mapped to the same lemma, treating them as orthographic variants rather than distinct morphology.
• I4 (world): Compiled lexical knowledge could flag variant forms as occasional or established, so rare spellings do not create spurious entries in the ontology.
• I5 (learn): Evaluation reports could flag results drawn from very small slices and cross-check them against a larger independent source before claiming a trend.