Abstract
Phonetic convergence—the phenomenon where speakers' speech patterns align during interaction—often presents a methodological challenge: balancing acoustic control with social authenticity. This study introduces a novel "music game" paradigm that requires pairs to coordinate specific vocalizations to perform a melody. Results from a large-scale field study (N=77 pairs) demonstrate significant phonetic convergence within 15 minutes, specifically within the task context, providing a robust framework for studying real-time linguistic alignment.
Introduction
In linguistics, phonetic convergence occurs when speakers begin to sound more like one another during a conversation. Studying this phenomenon typically involves a "methodological trade-off." Shadowing tasks offer high acoustic control but lack social reciprocity, while spontaneous conversations provide authenticity at the cost of precise acoustic measurement. To address this, we developed a "middle ground" task: a phonetic variant of the procedural coordination task that is highly controlled yet deeply interactive.
Methodology
The Music Game Paradigm
The core of the study is a two-player music game where pairs must "play" a melody by pronouncing specific consonant-vowel (CV) sequences: /ki/, /ka/, and /ku/.
- Roles: One player acts as the Director, who sees the "sheet music" (the required sequence and timing of sounds), while the other is the Follower, who sees an empty interface.
- Coordination: Players must coordinate their vocalizations to progress. As the game advances, melodies require "chords"—simultaneous vocalizations where one player’s sound must overlap with or be nested within the other’s.
- Technical Implementation: A real-time vowel recognizer (implemented in Matlab) triggered the game's mechanics based on audio input from headworn microphones.
Experimental Setting
The study was conducted as a "field lab" at Lowlands Science 2019, a large Dutch music festival. This environment allowed for the recruitment of 77 pairs of speakers, most of whom had pre-existing relationships (friends, partners, or family).
Results
Phonetic Alignment
Statistical analysis revealed a significant increase in phonetic similarity for the /a/ vowel within task-specific sequences (beta = -12.6, p = .01). This alignment occurred rapidly, within the 15-minute gameplay window. Interestingly, this effect was context-dependent; no significant convergence was observed in general sentence production (beta = -1.2, p = .81).
Emergent Communication and Robustness
Beyond acoustic alignment, pairs developed emergent communication systems to solve coordination challenges. For instance, some Directors used vocal repetitions (e.g., "/kaka/") as cues for the Follower. The observed convergence remained statistically significant even after controlling for external factors such as substance use among festival participants.
Discussion and Conclusion
The findings suggest that the necessity of coordination—acting as a single unit to achieve a goal—is a powerful driver of phonetic alignment. The lack of "spillover" into general sentences may be attributed to the participants' pre-existing relationships; pairs may have already reached a baseline level of alignment over years of interaction, leaving little room for further general convergence.
This paradigm successfully demonstrates that experimental control can be maintained without sacrificing the interactive nature of communication. Future work will involve analyzing acoustic data during gameplay to determine the temporal dynamics of convergence and how it evolves throughout the interaction.
Summary updated