📄 The Development of Old English Diphthongs

Summary

Steponavičius follows the history of four Old English diphthongal sounds, /ie(:) io(:) eo(:) ea(:)/, from their Germanic sources to their eventual disappearance. He treats spelling as the main evidence. Diphthongs were written as digraphs, so the letters scribes picked for each element reveal sound changes, not only mergers. Any inference from spelling, however, must agree with what is known of earlier and later stages. The data are early records: Bede's Historia Ecclesiastica, the Liber Vitae, the Corpus, Epinal and Erfurt glossaries, Kentish charters, and, for the late stage, Middle Kentish texts such as the Ayenbite of Inwyt. The study reconstructs a phonetic path, roughly [iu eu æu] → [io eo æo] → [io eo æa]. It dates the merger of /io/ and /eo/ by when scribes stopped keeping the two digraphs apart. It then shows how the diphthongs became single vowels or split into /j/ + vowel, which partly restored front rounded and front low vowels. The author's explanation is that the diphthongs were poorly integrated into the overall vowel system.

Key points

• Around the 7th century, non-West Saxon dialects had three falling diphthongs, high, mid and low, each long or short. Early West Saxon also had /ie(:)/, which the author reads as a relic of an i-umlaut series.
• Kentish charter 6 writes every /ea(:)/ as aea. The author regards this as conservative or confused spelling, not reliable phonetic evidence.
• In the 8th–9th centuries, only Northumbrian texts such as the Liber Vitae still keep io and eo consistently apart. The early Mercian glossaries mix them only a few times, which is too little to prove a merger.
• In Southern Northumbrian (Rushworth), eo spellings for long /ea:/ outnumber ea by about three to one.
• In late Kentish, /ea:/ and /io:/ split into /j/ + vowel mainly at the start of a word or after a single dental stop. Elsewhere they became single vowels, and word-final /io:/ ended up as /i:/. In 11th-century West Saxon and other dialects, /eo(:)/ became /ø(:)/.

Interesting ideas

• I1. A spelling convention can carry information beyond what it was designed to record, and the choice of letters for each part of a digraph serves as a measurable signal.
• I2. A merger is inferred only when two forms are used interchangeably, not from a few stray confusions, so a frequency threshold separates noise from real change.
• I3. Each reading of the evidence is accepted only if it agrees with the stage before and the stage after.
• I4. Whether a digraph alternates with a simpler spelling in the same word shows whether it stands for one sound or two.
• I5. The loss of a whole subsystem is explained by how poorly it fits the larger system, not by isolated sound changes.

For Mova

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

• I1 (tokens): Nonstandard or variant spellings in input could be logged as signals of dialect or register instead of being normalized away silently.
• I2 (learn): Self-induced repair rules could fire only when the conflation of two labels exceeds a threshold, with rare mismatches kept as negative controls.
• I3 (world): Proposed entity states or event orders could be checked against neighbouring states in the text, and readings that break continuity rejected.
• I4 (tree): Treebank dialect conversion could use alternation patterns within a word or construction to decide whether a variant is notational or structural.
• I5 (world): When ontology categories are rarely linked to the rest of the knowledge graph, that could flag them as candidates for merging or pruning.