The emergence of word order conventions: improvisation, interaction and transmission

Marieke Schouwstra, Kenny Smith, Simon KirbyPsyArXiv 2020 · DOI

TL;DR

This paper investigates how word order conventions emerge and change in human communication. By using silent gesture in controlled lab experiments, the researchers found that while initial communication is "natural" and semantically driven (SOV for action events, SVO for mental events), interaction and transmission over generations lead to simpler, more regular word orders, often converging on the native language's dominant order (SVO).

What Problem Does This Paper Address?

The paper tackles the fundamental question of how linguistic conventions, specifically word order, arise and stabilize in human communication. It highlights a key tension: the "naturalness" of word order in spontaneous, improvised communication (driven by semantic properties of events) versus the regularity and simplicity observed in established languages. The authors aim to understand the mechanisms – improvisation, interaction, and cultural transmission (iterated learning) – that drive the shift from semantically-driven, variable word order to more fixed, conventionalized systems.

Methods

The core methodology involves a series of laboratory experiments using silent gesture. Participants were tasked with communicating simple events using only gestures, without speech. The experiments were designed to simulate cultural evolution by:

  • Improvisation: Participants initially had to create gestures to convey meaning without pre-existing linguistic rules.
  • Interaction: Participants communicated with each other, allowing for the development of shared conventions through negotiation and alignment.
  • Transmission (Iterated Learning): In Experiment 1, an "observer" learned from a communicating pair and then replaced one of them in the next "generation," simulating cultural transmission across multiple generations. Experiment 2 focused solely on interaction without transmission. Experiment 3 manipulated the frequency of different event types to see if this could influence conventionalization.

Word order was coded from video recordings, and statistical models (linear and binomial mixed-effects regressions) were used to analyze changes in efficiency (number of constituents), accuracy, naturalness of word order, SOV proportion, and entropy (a measure of regularity).

Datasets

The study used a custom-created dataset of 64 line drawings depicting intensional and extensional events. These stimuli were organized into groups to ensure all combinations of actors, patients, and actions were represented. The drawings were tested for clarity on a crowdsourcing platform.

Key Insights and Ideas

  • Naturalness vs. Regularity: The paper clearly demonstrates the trade-off between "natural" word order (semantically driven, variable) and "regular" word order (simpler, more consistent).
  • Interaction Drives Simplification: Communicative interaction alone (Experiment 2) was sufficient to drive word order simplification and regularity, similar to when transmission was also involved (Experiment 1). This suggests interaction is a primary driver.
  • Transmission's Role is Nuanced: While transmission (Experiment 1) didn't accelerate the pace of simplification compared to interaction alone (Experiment 2), it's crucial for propagating the emergent conventions across new learners.
  • Native Language Bias: The dominant word order in participants' native language (English, SVO) often emerged as the conventional order, suggesting a potential influence of prior linguistic experience.
  • Event Frequency Matters: Experiment 3 showed that manipulating the frequency of event types could lead to the conventionalization of SOV order, even when SVO was generally preferred, highlighting the role of input statistics.
  • Lineage Specificity: Word order conventions developed uniquely within different "chains" or "dyads," indicating that the emergent systems are not purely universal but are shaped by the specific interactions within groups.
  • Priming as a Mechanism: Structural priming (both self-priming and, to a lesser extent, interlocutor priming) was identified as a potential mechanism underlying word order regularization.

Key Results

  • Efficiency: The number of gestured elements per trial decreased significantly over generations in all experiments, indicating increased communication efficiency.
  • Accuracy: Communication accuracy was high from the start and generally improved slightly over generations, with dyadic experiments showing a marginally faster increase.
  • Naturalness: The proportion of "natural" word orders (SOV for extensional, SVO for intensional) decreased significantly over generations.
  • SOV Proportion: Participants showed a preference for SVO over SOV, and this preference increased over time, with the proportion of SOV decreasing.
  • Regularity (Entropy): Entropy scores decreased significantly over generations, indicating that word order became simpler and more regular.
  • Experiment 3 Impact: Skewing the event types towards extensional events (which favor SOV in improvisation) led to a higher initial proportion of SOV and allowed some chains to converge on an SOV-dominant regime.

Limitations and Future Work

  • Native Language Influence: The strong emergence of SVO, the native language order, raises the question of whether participants could converge on a different order if the input were sufficiently skewed. Experiment 3 began to address this.
  • Priming Asymmetry: The asymmetry in priming effects between Experiment 1 (comprehension priming observed) and Experiment 2 (not observed) is noted as unclear and warrants further investigation.
  • Modality: The study used silent gesture (visual-manual modality). Further research could explore how these findings translate to spoken language and other modalities.
  • Generalizability: While the lab setting provides control, future work could explore these mechanisms in more naturalistic settings or with different populations.
  • Other Linguistic Features: The paper focuses on word order. Future research could investigate how other aspects of language (e.g., lexicon, morphology) emerge and interact with word order conventions under similar evolutionary pressures.

Summary updated

Abstract

When people improvise to convey information by using only gesture and no speech (‘silent gesture’), they show language-independent word order preferences: SOV for extensional events (e.g., boy-ball-throw), but SVO for intensional events (e.g., boy-search-ball). Real languages tend not to condition word order on this kind of semantic distinction but instead use the same order irrespective of event type. Word order therefore exemplifies a contrast between naturalness in improvisation and conventionalised regularity in linguistic systems. We present an experimental paradigm in which initially-improvised silent gesture is both used for communication and culturally transmitted through artificial generations of lab participants. In experiments 1 and 2 we investigate the respective contributions of communicative interaction and cultural transmission on natural word order behaviour. We show that both interaction and iterated learning lead to a simplification of the word order regime, and the way in which this unfolds over time is surprisingly similar under the two mechanisms. The resulting dominant word order is mostly SVO, the order of the native language of our participants. In experiment 3, we manipulate the frequency of different semantic event types, and show that this can allow SOV order, rather than SVO order, to conventionalise. Taken together, our experiments demonstrate that where pressures for naturalness and regularity are in conflict, naturalness will give way to regularity as word order becomes conventionalised through repeated usage.
BibTeX
@misc{schouwstra2020emergence,
  author = {Marieke Schouwstra and Kenny Smith and Simon Kirby},
  title = {{The emergence of word order conventions: improvisation, interaction and transmission}},
  year = {2020},
  doi = {10.31234/osf.io/wdfu2},
  howpublished = {PsyArXiv},
}
Figure