📄 "Old English Noun Declensions."

Summary

Juan de la Cruz (University of Málaga; the printed header reads Selim 3, 1993, pp. 26–42) addresses a teaching problem. The traditional historical names of Old English noun classes, the "A declension" for masculines and neuters and the "Ō declension" for feminines, seem to contradict the attested forms. Wulf has no -a, and giefu ends in -u. Students who know Latin are further confused when handbooks match the OE A-class to the Latin second declension and the Ō-class to the Latin first. The article first surveys the major OE declensions: A, Ō, weak/N, I and U. It then places them in an Indo-European frame and gives reconstructed, Gothic, Old English and Latin paradigms side by side, for example wulf/lupus, wylf/lupa and geoc. Its central point is that Indo-European short o became Germanic a, and long ā became Germanic ō. Final short vowels were then lost under initial stress, and analogy reshaped the endings. A few relics, chiefly the -u of short-stemmed nouns, still show the old stems. This explains both the labels and the Latin correspondences without heavy apparatus.

Key points

• The "great" declensions are three:
  • A: masculines and neuters, including -ja and -wa stems.
  • Ō: feminines, including -jō and -wō stems.
  • Weak N: e.g. nama/naman, which parallels Latin nomen/nominis. Only three neuters belong to it.
• The I and U declensions survive with few members. Only nine nouns are known in the U class, because most moved into the A and Ō classes.
• The key sound correspondences are IE short o = Germanic a (Latin 2nd declension) and IE long ā = Germanic ō (Latin 1st declension). Vowel quality shifts while length is kept.
• The -a in wulfas/wulfa is not the original stem vowel. It comes from the stem vowel contracting with the ending.
• The -u in N.Ac.pl. neuter geocu/scipu and in N.s. feminine giefu reflects one old ending, a collective -ā. This parallels Latin templa beside rosa.
• Cognates can fall into different classes across languages. Examples are genus/cynn, pecus/feoh, cor/heorte and dens/tōth.

Interesting ideas

• I1. An inflection class can keep a historical label that its surface forms no longer show, so the class ID and its realized endings are best stored as separate layers.
• I2. Analogy can override regular sound change, as with G.s. giefe and Ac.pl. giefa. A model should therefore expect a regular rule plus a recorded analogical exception, not one rule for everything.
• I3. When one form serves two functions, as when the neuter plural matches the feminine singular, the shared origin explains the ambiguity better than calling it a coincidence.
• I4. The argument rebuilds a hidden form by lining up several independent witnesses (Gothic, Latin, OE relics), each of which preserves different traces.
• I5. The paper explains confusing terminology by tracing it step by step to its source, aiming at the actual confusion learners report.

For Mova

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

• I1 (tokens): the morphology tables in crate en can keep an abstract paradigm ID apart from surface suffix rules, so irregular or leveled forms do not force a mislabeled lemma class.
• I2 (tokens): lemmatizer rules could carry an explicit "regular vs. analogical override" flag, which keeps each lemma decision explainable.
• I3 (tree): number/case syncretism in morphology can be passed to the parser as an unresolved feature set and settled from syntactic context, instead of being guessed early.
• I4 (learn): this mirrors silver data agreed by several witness parsers, where each witness may be partial but their alignment recovers the most trustworthy analysis.
• I5 (answer): student-facing explanations can name the learner's likely misconception first and then give the chain of reasoning, instead of stating only the final answer.