📄 Expanded tenses in the old English orosius a syntactic strengthening*

Summary

The study asks whether the Old English periphrasis beon/wesan + present participle in -ende (the "expanded tenses", e.g. he is laerende) behaved as a syntactic strengthening. A strengthening here means a heavier form that costs the speaker more effort but is easier for the hearer to decode. The test case is King Alfred's fairly free translation of Orosius. The corpus is the Sweet (1883) edition, and the author compares two samples:

• all 237 clauses with expanded tenses;
• a control set of 855 clauses with simple tenses, drawn on a fixed sampling grid.

Each clause is coded for several grammatical parameters, such as tense, mood and transitivity. For each parameter, the share of "marked" values is compared between the two samples using a significance index. Four parameters show significantly more marking in the expanded sample. Clauses with expanded tenses also tend to carry several marked features at once.

The diachronic claim is then set against the modern progressive. Today it is required for marked aspect, optional where it adds emotional tone, and rare or barred with stative and modal verbs. The author reads this as support for a "natural syntax" view of language change.

Key points

• Data:
  • Basic sample: 237 expanded-tense clauses from Sweet's Orosius edition.
  • Control sample: 855 simple-tense clauses, sampled by taking one clause in three from one page in three.
  • Of the 237, 53 render Latin simple tenses and 78 have no Latin counterpart, which shows the translation was syntactically independent.
• Coding: Clauses were coded for clause type and propositional modality, tense, mood, aspect, transitivity and number. Person and voice were dropped because the basic sample has only third person and active voice.
• Significance: A difference counted as significant when the index Id exceeded 2 (Pavlič 1985).
  • Tense, aspect, transitivity and number were marked significantly more often with expanded tenses.
  • Mood and modality were inconclusive, since subjunctives and negatives each fell below 10%.
• Cumulative markedness: Over half of the expanded clauses carry three marked parameters, while over half of the control clauses carry only one. The mean marking probability is also higher in the expanded sample.
• Frequent verbs: The most common verb is winnan ("make war", 34 of 237). Because the same verbs dominate both samples, the claim about lexical "heaviness" could not be tested.

Interesting ideas

• I1. The construction is admitted only if swapping it for the simple tense leaves the clause's syntax intact, which filters out adjectival participles (is interesting) and depictive ones (is in the temple teaching).
• I2. The target phenomenon is contrasted with a systematically sampled control set coded on the identical feature scheme, so any effect is measured against a baseline rather than asserted.
• I3. Instead of testing features one at a time, the paper counts how many features are marked per item and compares medians and quartiles, which captures joint complexity.
• I4. Modern progressive use is graded on four levels (obligatory, optional, exceptional, impossible), and the class of verb decides which level applies. Stative verbs in the progressive turn dynamic, and "always + -ing" adds annoyance or condescension.
• I5. Under this view the marked form is reserved for marked information: temporary duration, simultaneity, ingressive or ongoing action.

For Mova

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

• I1 (tree): The same substitution check could separate progressive aux + VBG from copula + adjectival participle when converting or validating treebank analyses.
• I2 (learn): Evaluations of a repair rule or silver-data filter could pair the affected sentences with a grid-sampled control set to show the effect is specific, not general.
• I3 (learn): A per-sentence "marked feature count" (past tense, transitive, plural, subordinate, negated) could stratify parser error analysis and reveal whether failures concentrate in heavy clauses.
• I4 (frames): The prag layer could flag progressive-with-stative as a reading shift toward an event, and "always/continually + -ing" as an attitudinal implicature cue.
• I5 (back): The generator could default to simple tenses and choose the progressive only when the meaning graph holds an ongoing or overlapping event, avoiding unmotivated -ing forms.