Repeat after us: Syntactic alignment is not partner-specific

Ostrand, Rachel; Ferreira, Victor S. · 2019 · Journal of Memory and Language

DOI: 10.1016/j.jml.2019.104037

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates whether syntactic alignment in conversation is partner-specific (matching a specific interlocutor’s linguistic habits) or partner-independent (matching the aggregate statistical distribution of the linguistic context). While prior research established that speakers align their speech to interlocutors, it remained unclear whether this alignment is driven by communicative utility—where matching a partner prevents misunderstanding—or by a general statistical learning mechanism. The authors hypothesized that if alignment is communicatively modulated, speakers would only exhibit partner-specific alignment when it aids comprehension, such as in lexical entrainment with novel referents. Conversely, in domains like syntax, where multiple structures are universally understood and partner-specificity offers no communicative advantage, speakers should default to partner-independent alignment. To test this, Ostrand and Ferreira conducted five experiments in which participants interacted with two experimenters who held distinct, systematic syntactic preferences. For example, one experimenter exclusively used prepositional datives (e.g., “gave the book to the boy”), while the other exclusively used double object datives (e.g., “gave the boy the book”). Participants engaged in a picture-matching game, first listening to descriptions from both experimenters and then producing descriptions for each partner individually. This design allowed the researchers to disentangle whether participants tracked and reproduced the specific syntactic distribution of the current partner or simply reflected the overall syntactic statistics of the conversation. The experiments varied exposure conditions, including whether participants heard descriptions before producing them and whether partners shared or differed in their syntactic preferences. The results consistently demonstrated that participants engaged in partner-independent alignment rather than partner-specific alignment. Although participants adjusted their syntactic production to reflect the statistical prevalence of structures in their recent input, they did not tailor their syntax to match the specific preferences of the individual partner they were addressing. Instead, speakers produced syntactic structures based on the aggregated distribution of the entire linguistic context, regardless of which partner was listening. This pattern held across various experimental manipulations, indicating that speakers did not maintain separate, partner-specific syntactic models when such specificity provided no communicative benefit. These findings support the Rational Speaker Model’s claim that linguistic alignment is driven by communicative utility. The study concludes that the mechanism underlying syntactic alignment is sensitive to the social and communicative context; speakers align to aggregate, partner-independent statistics when partner-specificity does not facilitate comprehension. This distinguishes syntactic alignment from lexical entrainment, where partner-specific alignment is crucial for establishing shared referents. The results imply that the language processing system adapts hierarchically, utilizing partner-specific models only when necessary for successful communication, while relying on general contextual models for linguistic features that do not carry the same risk of communicative failure.

Key finding

Speakers align syntactically to aggregate, partner-independent statistical distributions rather than to partner-specific distributions when partner-specificity does not provide communicative utility.

Methodology

lab_experiment

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.

StageOutcomeToolModelPromptAttemptsCompleted
discover success author_sweep 2 2026-05-28
archive success canonical_url 17 2026-06-06
extract success cached 3 2026-06-10
clean success clean 1 2026-06-04
chunk success chunk 1 2026-06-04
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-04
enrich success 1 2026-05-28
promote success 1 2026-06-04
summarize success llm qwen3.6-27b-prismaquant summ-v5 2 2026-06-10
tag success vector_similarity 15 2026-06-11
verify success 2 2026-06-10

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.