📄 Indirect Responses in Biblical Hebrew

Summary

The paper looks at how Biblical Hebrew speakers reply to questions without using the expected answer form. In Biblical Hebrew, a "yes" usually repeats the question's predicate and a "no" uses a negative word. An indirect response breaks this pattern. It is a syntactically independent utterance that still connects coherently to the question. Zewi takes ten examples from Classical Biblical Hebrew prose. They cover the three question types the language attests: yes/no questions, question-word (WH) questions and alternative questions. She analyses each pair with Halliday & Hasan's three cohesion devices: anaphora (by repetition or resumptive pronouns), substitution and ellipsis. She also separates coherence (functional connectedness) from cohesion (surface linking features). Examples include:

• the Cushite's curse on the king's enemies, from which the king infers Absalom's death;
• Jehoshaphat's set phrase pledging alliance;
• Hagar giving a reason instead of a route.

Nearly every pair shows at least one cohesion device. Indirect replies are most visible after WH-questions. The study shows that implicit answers remain traceable through identifiable formal links.

Key points

• The corpus has ten question–response pairs from Classical Biblical Hebrew prose, e.g. 2 Sam 18, 1 Kgs 22:4 (cf. 2 Kgs 3:7), Gen 16:8 and 2 Kgs 9:11.
• The framework uses Halliday & Hasan's cohesion devices (anaphora, substitution, ellipsis) and Crystal's coherence/cohesion distinction.
• Only one example relies on context alone. Four use one device and five combine two.
• Anaphora (repetition or resumptive pronouns) is the most frequent device. Substitution appears in two examples (one more partially) and ellipsis in one.
• Question-tag interrogatives do not occur in Biblical Hebrew. Indirect answers are most evident with WH-questions.

Interesting ideas

• I1. Define the canonical answer pattern first, so any reply that departs from it (e.g. "yes" = repeating the predicate) is flagged as indirect and needs inference. — For Mova (frames): prag could hold a template of the expected reply per question type and route non-matching replies to implicature handling.
• I2. Even an indirect reply usually leaves a surface trace, such as a repeated noun, a pronoun mapped to the asker's person agreement, or a substitute noun; these traces can confirm that it really answers the question. — For Mova (frames): coreference could link pronouns and repeated lemmas across the turn boundary to confirm which question a reply addresses before inferring its polarity.
• I3. A wish or curse such as "may my enemies be like X" can carry the answer polarity about X, and a statement of reason can stand in for a requested place. — For Mova (world): the world model could update X's state (e.g. dead or harmed) from such comparison-wishes, and record the reason as an answer to "where/why".
• I4. Separate deep coherence from surface cohesion, and count the cases where coherence rests on context alone as the hardest category. — For Mova (learn): evaluation could keep a held-out "context-only" subset of indirect answers as a stress test, since those cases cannot be resolved by surface links.