Linguistic Creativity affects Discourse Expectations related to Contiguity but not Implicit Causality

Linguistic Creativity affects Discourse Expectations related to Contiguity Relations but not Implicit Causality Oliver Bott∗,∗∗, Torgrim Solstad∗,∗∗ & Florian Kankowski∗∗ – Projects B01∗&¨ O∗∗, CRC 1646, Bielefeld University CogSci 2025, San Francisco Implicit Causality (IC) (1) a. Max admired Lea. She () / He () . . . b. Max admired Lea because ()/and so () . . . IC verbs: interpersonal verbs with cohesion-related biases ▶Coherence bias:explanation as default relation [2] ▶Coreference bias: strong remention biases [2,3,4,5,6] Biases due to discourse expectations [7,8,9] . . . meets . . . globalist strategy localist strategy change discourse maintain preferred structure discourse structure Linguistic Creativity ▶An utterance is creative if it is original and at the same time communicatively effective [10,11] ▶Only little psycholinguistic research on linguistic features of creative discourse [12,13,14] ▷IC study [14]: Biases unaffected in creative discourse ▶Research question of present study: Is this a general phenomenon or only observed in the context of IC? IC and Beyond – The Empty-Slot Account [15,16,17,18] Psychverbs [16]: (2) a. Mary fascinated Peter b. Peter admired Mary Agent-Evocator (AE) verbs [17, 18]: (3) Peter congratulated Mary (Solstad & Bott, 2022) ▶A proper name argument is not sufficient to fill the empty causal slot of an IC verb – Strong semantic requirement for specification explaining cohesion biases ▶Implicit Consequentiality [19,20,16,17] due to a general discourse principle, the Contiguity Principle [21,22] – purely pragmatic bias, open to globalist strategy? Localist vs. Globalist Strategies – Hypothesis & Examples Hypothesis: Creative discourse strategies are constrained by lexical semantics ▶IC as a semantic constraint is too strong to be overridden by creativity manipulation localist narrative strategies ▶For non-IC verbs (causative agent-patient verbs, [17]) globalist strategies Examples: Creative discourses illustrating localist vs. globalist strategies (4) a. [Max praised Mia.]prompt She had done an incredible job with her excavator and had actually brought the mountain to the prophet, shovel by shovel. (IC, explanation) b. [Mia healed Max.]prompt She healed him of his fears and worries, the suffering of the world, the agony of existence and of life in general. (Non-IC, elaboration) Study Design and Experimental Methods – Adopting the ‘Creativity on Demand’ Task from [14] Production – Baseline Trial Rating: Sample Trial Experiments on Coherence Biases Production Exp., full stop prompts ▶43 native German speakers ▷Openness to Experience score [23,24] ▶20 AE verbs, NP2 biased & 20 Non-IC verbs ▶Creativity manipulation: ▷1st block: BASELINE ▷2nd block: CREATIVE ⇓ Rating Exp. full stop prompts ▶58 native German speakers ▶naive to creativity manipulation ▶for each production ▷originality rating (slider 0-100) ▷felicity rating (slider 0-100) ▶Three ratings per production Experiments on Coreference Biases Production Exp., because prompts ▶40 native German speakers ▷Openness to Experience score [23,24] ▶same 20 IC verbs & 20 Non-IC verbs ▶Creativity manipulation: ▷1st block: BASELINE ▷2nd block: CREATIVE ⇓ Rating Exp. because prompts ▶58 native German speakers ▶naive to creativity manipulation ▶for each production ▷originality rating (slider 0-100) ▷felicity rating (slider 0-100) ▶Three ratings per production Production – Creative Trial Rating: Sample Trial Perceived Originality and Felicity of Continuations in the two Blocks Originality Ratings: ▶Creativity on demand task worked – clear differences in perceived originality (replicating [14]) Felicity Ratings: ▶Continuations provided in the creative block as felicitous as continuations from the baseline block ▶Somewhat higher felicity ratings in the because experiment than in the full-stop experiment. Easier task? Linguistic Features of Creative Productions Coherence Bias: ▶The likelihood of explanations differed between IC and Non-IC verbs ▶. . . but was unaffected by the creativity on demand task [as in 14] ▶However, creativity modulated contiguity in favor of other relations such as elaboration [cf. 12] Coreference Bias: ▶Creativity manipulation did also not affect the coreference in explicit explanations after because prompts ▶IC induced QUD remains unchanged: What was the occasion for NP2 to AE-verb NP1 on? Conclusions ▶Higher degree of elaboration in creative discourse than in default continuations – Also visible in analyses on continuation length and subordination Open Questions ▶What contributes to perceived originality? We are currently identifying the creative strategies using a highlighting task – and compare humans to LLMs References:[1] Garvey & Caramazza (1974). LI 5.[2] Kehler et al. (2008). JoS 25(1).[3] Rudolph & F¨orsterling (1997). PB 121(2).[4] Ferstl et al. (2011). BRM 43.[5] Hartshorne & Snedeker (2013). LCP 28(10).[6] Hartshorne et al. (2013). EP 60(3).[7] Koornneef & van Berkum (2006). JML 54.[8] Pyykk¨onen & J¨arvikivi (2010). EP 57.[9] Rohde & Horton (2014). Cogn. 133(3).[10] Stein (1953). JP 36(2).[11] Runco & J¨ager (2012). CRJ 24(1).[12] Skalicky et al. (2017). CRJ 29(4).[13] Johnson et al. (2023). BRM 57(7).[14] Bott et al. (2023). CogSci Proc..[15] Bott & Solstad (2014). SiTP 44.[16] Solstad & Bott (2022). LCN 37.[17] Solstad & Bott (2023). FLS 2.[18] Solstad & Bott (2024). S&P 17.[19] Stewart et al. (1998). ms.[20] Garnham et al. (2020). BRM 53.[21] Moens & Steedman (1988). CL 14.[22] Kehler (2002). CSLI.[23] McRae (1987). JPSP 52 (6).[24] Goldberg et al. (2006). JRP 40. Bott, Solstad & Kankowski, Projects B01 & ¨ O, CRC1646, Bielefeld University Materials & Analysis Scripts: https://osf.io/xz7d6/