📄 Old English verbs of prohibition. Grammatical behaviour and class membership

Summary

The paper tests whether six Old English verbs that A Thesaurus of Old English (TOE) groups under "prohibition" form a genuine verb class: bewerian, forbēodan, foresacan, forwiernan, stīeran and tōcweþān. The author holds that shared meaning is not enough and that class members must also behave alike grammatically. She combines Levin's verb classes and alternations with Role and Reference Grammar (RRG), which supplies Aktionsart, macroroles (Actor/Undergoer), the PSA and juncture/nexus types. Each verb's attestations are analysed for three things:

• the arguments it takes and their case marking,
• the syntactic configurations it appears in (simplex clause, linked core with inflected infinitive, subjunctive dependent clause),
• whether it takes part in the Nominalisation and Undergoer alternations.

Only bewerian, forbēodan and forwyrnan use all three configurations and both alternations. Their syntax centres on the patient of prohibition, which acts as the pivot of the construction. The other three verbs appear in only one configuration and have divergent meanings. The thesaurus grouping therefore does not hold as a grammatical class. The study shows that meaning-based lexical groupings need checking against syntactic behaviour.

Key points

• Classes compared: a TOE label of six verbs, set against Levin's criterion that class members share alternations and against RRG semantic-role analysis.
• Shared profile of all six verbs:
  • macrorole transitivity 2 and a preference for active voice;
  • up to three syntactic arguments, usually a nominative plus a genitive or dative;
  • the dative patient is usually placed ahead of the verb, or after the auxiliary but before the main verb.
• The consistent core (bewerian, forbēodan, forwyrnan):
  • they appear in all three configurations shown in Figures 1–3;
  • they take part in the Nominalisation and Undergoer alternations, the latter treated as raising from the linked clause.
• Differences in linking to PSA within the core:
  • forbēodan lets both the object and the patient of prohibition become PSA, including case-preserving passives;
  • bewerian cannot, because it does not keep the dative under passivisation;
  • forwyrnan's genitive object cannot passivise.
• The excluded verbs (foresacan, stīeran, tōcweþān): they appear only in simplex clauses and have accusative or genitive objects. The author proposes removing them from the class.

Interesting ideas

• I1. Validate a meaning-based verb grouping by checking whether its members actually share alternations and argument frames, and split it when they do not.
• I2. Identify the participant around which a construction's complexity turns (co-reference, case, passive linking, raising) and treat it as the pivot.
• I3. Describe one verb's complement options as separate configurations (simple object, infinitival core sharing an argument, finite subordinate clause) that all map to the same roles.
• I4. Do not supply an omitted dummy subject or matrix clause when no principled way to restore it exists; analyse the attested structure directly, for example as a case-preserving passive.
• I5. Record when a verb blocks a transformation, such as a passive on a genitive or dative object, as a lexical property.

For Mova

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

• I1 (world): Audit compiled verb classes from VerbNet or WordNet against the frames observed in parsed text, and flag members whose syntactic profile diverges from the rest of their class.
• I2 (frames): For verbs like forbid or prevent, mark the "prohibited party" role as the pivot, so it can be resolved consistently in both the matrix and the embedded clause.
• I3 (frames): Map xcomp, ccomp and plain object variants of a verb onto one role frame, so "forbade him to go", "forbade that he go" and "forbade his going" yield the same who-did-what-to-whom.
• I4 (tree): Prefer analysing elliptical or subjectless passives as they stand over inserting silent tokens, which keeps the parse explainable and avoids invented material.
• I5 (answer): Use such per-verb constraints to check whether a paraphrased question (for example a passive rewording) is a legitimate restatement of the text before answering it.