Paratactic Causal Processing in Mandarin: How Iconicity and Connectives Shape Offline Judgments and Online Reading

The Iconicity Hypothesis and the Connective Facilitation Hypothesis predict that canonical cause–effect ordering and explicit causal connectives facilitate discourse comprehension, but both accounts derive primarily from hypotactic Indo-European languages, leaving their applicability to paratactic languages such as Mandarin unclear. This study examined how clause order and the causal connective yīnwèi (because) shape Mandarin causal discourse processing through offline coherence judgments and online eye-tracking in a 2 × 2 design manipulating clause order (iconic cause–effect vs. non-iconic effect–cause) and connective presence. Offline judgments showed that iconic ordering and yīnwèi both enhanced perceived intelligibility and local coherence, with the connective providing a particularly strong compensatory benefit in non-iconic structures. Eye-tracking, however, revealed no reliable global effects of clause order or connective presence; effects were instead confined to the cause-clause region, where yīnwèi reduced first-run dwell time across both orders, indicating early facilitation, but increased regression-path reading time in non-iconic sequences, reflecting later integration costs. These findings reveal a dissociation between reflective coherence evaluation and incremental online processing. We argue that this pattern is best captured by a staged probabilistic cue-integration framework that refines, rather than rejects, the Iconicity and Connective Facilitation Hypotheses, treating clause order and connective marking as informative but non-obligatory cues that readers dynamically weight alongside semantic plausibility and pragmatic context. Explicit connectives thus reduce early relational uncertainty but may incur later costs when predictive cues conflict with non-canonical structure, suggesting that iconicity and connective marking function as probabilistic rather than universally facilitative cues.