📄 Ongitan: A Case Study of Evidentiality in Old English Perception Verbs

Summary

The paper examines how Old English expressed the source of a speaker's knowledge without grammatical evidential markers. It focuses on ONGITAN, the most general Old English verb of perceiving. The authors analysed more than 500 contexts drawn from two sources, the Helsinki Corpus of English Texts and the Toronto Microfiche Concordance. Because most of these texts are narrative, the knowledge is usually attributed to a character, so the authors use Langacker's term "conceptualizer" for whoever holds the knowledge, rather than "speaker". They adapt Anderson's criteria to separate two uses: plain perception, where ONGITAN has a concrete object, and evidential use, where it qualifies a proposition. They then classify the evidential uses by how the information was obtained. Finally, they study the verbs ONGITAN co-occurs with and the conjunctions and adverbs that surround it. Their main finding is that inference dominates the evidential uses, ranging from a sensory pole to an intellective pole. The study shows that a single lexical verb can carry a detailed source-of-knowledge system in a language that has no grammatical evidentials.

Key points

• The authors are Ana Laura Rodríguez and Eugenio Contreras (Universidad Complutense de Madrid). The printed header gives Selim 11 (2001–2002), pp. 97–115, not 2019.
• The data consist of more than 500 ONGITAN contexts, sourced from the Helsinki Corpus and the Toronto Microfiche Concordance. Bosworth-Toller lists eight senses, from sensory ones (see, hear, feel) to cognitive ones (judge, understand, recognize).
• Evidential use is identified by three conditions:
  • ONGITAN is not the main predication.
  • It introduces an event or state of affairs.
  • It is a concept-taking complement (as in perceive your fear vs perceive you are afraid) that signals the non-evidential reading.
• The typology, building on Willett, has two branches:
  • Direct evidence: visual observation.
  • Indirect evidence: hearsay, inference from perceived clues, and "mental construction" (reasoning, deduction, belief). Belief is split off from reasoning as its own type.
• Function words fall into two groups:
  • Words that introduce the evidence: forþon þe, siþþan, mid þy.
  • Words that introduce the inferential act itself: forþon, forþyþe, þa…þa, þa, þæt.

Interesting ideas

• I1. The syntactic type of a perception verb's complement (a nominal object vs a clause) separates literal sensing from a source-of-knowledge reading, and this test is cheap and transparent.
• I2. Knowledge is attributed to the character whose perspective the narrative adopts, not to the narrator.
• I3. Sources of knowledge are graded along a scale from direct sight, through hearsay and inference from clues, to pure reasoning and belief.
• I4. Connectives split into two groups: those that mark the evidence (because, since, by that) and those that mark the conclusion drawn (therefore, then).
• I5. When ONGITAN is paired with a sensory verb, the order of the two verbs mirrors the order of the actions, while with intellective verbs the order carries no meaning.

For Mova

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

• I1 (frames): when see/notice/perceive governs a ccomp rather than an obj, tag the embedded proposition with an evidential-source role instead of creating a sensing event.
• I2 (world): store facts as beliefs indexed by the character who holds them, so that answers about stories can tell the narrator's facts apart from a character's conclusions.
• I3 (answer): keep this grade on each belief so readers can answer "how did the fox know?" questions with an explainable source type.
• I4 (frames): use mark/advmod lemmas around cognition verbs to build explicit evidence → inference links between clauses.
• I5 (world): treat order inside coordinated sensory-plus-cognition verb pairs as an event-sequence cue, but not for pairs of purely mental verbs.