📄 On the Quantity of in Old English Words ending in -lic and -lice

Summary

Calle Martín and Miranda García address an issue in Old English phonology: whether the <i> in adjectives ending in -lic (e.g. gelic, cynelic) and adverbs ending in -lice (e.g. soþlice, lustlice) was long or short. The problem arose in practice while the authors were building MAOET, a morphological analyser of Old English texts. The analyser needed one fixed convention, but grammars and dictionaries contradict each other.

The paper first surveys the reference literature and sorts it into four camps:

• always short,
• always long,
• short in adjectives but long in adverbs,
• inconsistent or stress-dependent.

It also notes a parallel dispute over Gothic vowel length. The authors then draw on Old and Middle English texts and give three arguments for treating both suffixes as uniformly long:

• how the adverbs were formed,
• the Germanic etymon -līka, which had a long vowel,
• the later split into short Modern English -ly and diphthongised -like.

The result matters because it gives a principled, consistent lexical rule where authorities disagree. Such a rule can be applied directly in computational tagging of historical English.

Key points

• The motivation was practical: MAOET (Miranda et al. 1997) required a single standard length for -lic/-lice in order to tag reliably.
• The literature splits into four positions:
  • Campbell (1959) treats both suffixes as short, through early shortening in compound second elements.
  • Brook (1955), Bosworth & Toller, Bradley, Skeat and Lass mark both as long.
  • Clark Hall gives short adjectives but long adverbs.
  • Mitchell (1995) and Fernández (1982) are internally inconsistent, and Montes et al. (1995) make length depend on stress.
• Gothic shows a parallel debate over <i>/<ei>. Venneman (1971) argues for a length contrast. Marchand, Hamp and Jones argue the difference was tenseness or tone rather than length.
• Adverbs were formed by adding -e to -līc adjectives. Later, -līce was reanalysed as a suffix in its own right and attached to other adjectives and to nouns (e.g. heardlice, eornostlice).
• Modern -ly lost its length when its secondary stress weakened (Jespersen, Gimson). Forms in -like kept the long vowel, which the Great Vowel Shift turned into [laik].

Interesting ideas

• I1. Before choosing a convention, sort conflicting authorities into an explicit typology of positions, so the disagreement itself becomes structured data rather than noise.
• I2. Reject an analysis if it forces one derivational source to yield two variants of the same lexeme, keeping derivation consistent across base and derived forms.
• I3. Use downstream historical outcomes (the -ly versus -like split) as evidence about an unobservable earlier state, i.e. infer a hidden property from what it later became.
• I4. Let a practical tool requirement, here a tagger that needs one rule, force a scholarly decision, and document the reasoning for that rule.

For Mova

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

• I1 (learn): when witness parsers or LLM teacher tiers disagree on silver data, label the disagreement pattern (all-A, all-B, split by category, inconsistent) before deciding, so that repair rules target the systematic splits.
• I2 (tokens): ensure that -ly adverbs and their base adjectives share consistent lemma and morphology features in crate en, and flag analyses that break the adjective→adverb link.
• I3 (world): when a compiled knowledge value (e.g. a verb class or ontology attribute) is uncertain, check it against observable consequences in later or derived entries rather than trusting a single source.
• I4 (tokens): record the explicit rationale for each normalisation convention in the morphology layer, keeping every step explainable when sources such as dictionaries or treebanks conflict.