“Dyadosyncrasy”, Idiosyncrasy and Demographic Factors in Turn-Taking

Julio Cesar, Gabriel · 2025

Read “Dyadosyncrasy”, Idiosyncrasy and Demographic Factors in Turn-Taking by Julio Cesar, Gabriel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-24736.txt.

Introduction

Julio Cesar Cavalcanti and Gabriel Skantze measure what shapes the tiny gaps between speakers in conversation, in a 2025 paper. Turn-taking follows universal rules yet varies a lot. The team used a large set of American English phone conversations. They measured the offset between one speaker ending and the next starting. Sex, age, and schooling were tested as factors. Individual habits were tested too.

Women and older speakers left slightly shorter gaps. Schooling showed no effect. Light topics sped the exchanges a bit. Individual differences mattered much more. The strongest effect was pair-specific. Two partners resembled each other more than either resembled themselves with others. The authors coin a term for that pair signature. Joint activity, they conclude, outweighs demographics. Dialogue systems may need to adapt per pair, not per person. The study blends linguistics with statistics.

Catalog record

CREATORJulio Cesar, Gabriel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-24736

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

“Dyadosyncrasy”, Idiosyncrasy and Demographic Factors in Turn-Taking Julio Cesar, Gabriel Abstract Turn-taking in dialogue follows universal constraints but also varies significantly. This study examines how demographic (sex, age, education) and individual factors shape turn-taking using a large dataset of US English conversations (Fisher). We analyze Transition Floor Offset (TFO) and find notable inter-speaker variation. Sex and age have small but significant effects – female speakers and older individuals exhibit slightly shorter offsets – while education shows no effect. Lighter topics correlate with shorter TFOs. However, individual differences have a greater impact, driven by a strong idiosyncratic and an even stronger “dyadosyncratic” component – speakers in a dyad resemble each other more than they resemble themselves in different dyads. This suggests that the dyadic relationship and joint activity are the strongest determinants of TFO, outweighing demographic influences. Cavalcanti Skantze KTH Royal Institute of TechnologySweden Pontifical Catholic University of São PauloBrazil Stockholm UniversitySweden keywords Turn-taking, Conversation, Sex, Age, Education, Topic, Idiosyncrasy † † email: [email protected], [email protected] 1 Introduction Humans spend a substantial portion of their lives engaged in spoken interactions, seamlessly coordinating conversations through finely tuned turn-taking mechanisms. These mechanisms, central to human communication, display strikingly universal characteristics across languages – rooted in a shared cognitive and linguistic infrastructure [ 1 , 2 ] . Notably, the rapid transition between conversational turns – often around 200 ms – highlights the efficiency of human speech processing and the predictable structure of dialogue [ 3 ] . Yet, despite these universal patterns, turn-taking is far from monolithic. Cultural, developmental, and situational factors introduce considerable variation in how turn-taking unfolds across different contexts and speakers. For instance, previous studies [ 2 ] demonstrated that while turn transition timings generally fall within a narrow range across ten languages, the mean gap durations vary markedly: Japanese speakers exhibit an average gap of only 7 ms, in contrast to Danish speakers, who average 489 ms. Developmental trajectories further underscore this variability. Research indicates that children gradually acquire turn-taking skills, initially displaying longer gaps – often between 1.5 and 2 seconds – before refining their ability to anticipate turn completions and respond more promptly as their linguistic and social competencies mature [ 4 , 5 ] . Sex differences, though long recognized, have rarely been examined with contemporary quantitative methods and extensive datasets. Evidence suggests that in same-sex interactions, men tend to adhere closely to the conversational model proposed in [ 6 ] – minimizing gaps and overlaps – while women’s conversations feature significantly more overlap [ 4 ] . In a study of multi-party human–robot interactions, adult females not only dominated speaking time but also exhibited a higher degree of overlap (27% compared to 13% for males), an effect further amplified when two females interacted [ 7 ] . Corpus studies have further confirmed that such demographic differences are part of broader “turn-taking styles”, with duration patterns correlated to sex, age, dyadic influence [ 8 ] , and other dimensions [ 9 ] . Other factors are also known to shape turn-taking dynamics. Competitive conversational contexts tend to produce shorter gaps [ 10 ] , whereas complex tasks, unfamiliar topics [ 11 ] , and increased cognitive load may be associated with longer between-speaker intervals in conversation [ 12 ] . Eye contact plays a role as well. Pauses and overlaps tend to exhibit distinct statistical properties in telephone versus face-to-face dialogues. In comparable conditions, pauses in face-to-face conversations can last up to four times longer than those in telephone dialogues [ 13 ] . In addition, speakers tend to provide answers more rapidly than non-answer responses to questions. Among these answers, confirmations are typically produced faster than disconfirmations, and this pattern appears to hold across different languages [ 2 ] . Familiarity between speakers similarly impacts conversation dynamics, as conversations between friends tend to include longer gaps [ 14 ] . Personality traits further shape conversational style [ 15 ] ; extroverts, for instance, overlap more frequently with their conversational partners compared to introverts [ 16 ] . On a higher level, there is also the impact of the Dyadic factor on turn-taking. In fact, a considerable number of studies highlight the intricate ways in which speakers mutually influence each other’s speech patterns when engaging in conversation [ 17 , 18 , 19 , 20 ] or even when reading together [ 21 , 22 ] . Accommodation processes – the dynamic adjustment whereby individuals gradually align their behaviors – can further contribute to inter-speaker similarity [ 17 ] . In conversation, speakers tend to become more like each other in different features and different ways. This phenomenon, and all other terms that may get grouped together under the word “entrainment”, is widely believed to be crucial to the success of human interactions [ 18 ] . For instance, speakers adapt the duration of gaps to those of the other participants, that is, a form of accommodation [ 19 , 20 ] . Speakers may exhibit convergence through proximity, synchrony, or other forms of speech entrainment, with turn exchanges often displaying heightened similarity [ 17 , 18 ] . Collectively, these findings demonstrate that while turn-taking is governed by universal constraints, it also exhibits notable variations at both the individual and dyadic levels – variations that have yet to be comprehensively explored and quantified. Although previous studies have significantly advanced our understanding of turn-taking, some gaps remain – particularly attempts at quantifying the relative contribution of demographic, individual, and dyadic factors. Our analysis adopts a corpus-based approach and focuses on the Transition Floor Offset (TFO) – a metric that captures the temporal gap or overlap between consecutive turns, and employs mixed-effects modeling to account for both individual and conversation-level variability. We specifically ask: How do demographic differences modulate the timing of turn transitions? How variable (idiosyncratic) can it be across different speakers? What role do dyadic interactions play in amplifying or reducing inter-speaker differences? The answers to such questions are of both theoretical and practical significance. Theoretically, it clarifies how universal turn-taking principles unfold within a diverse landscape of demographic and individual factors. Practically, these insights are important for the development of turn-taking models in conversational systems [ 23 ] . While there have been a lot of efforts recently in developing predictive models of turn-taking based on human-human conversational data [ 24 , 25 , 26 ] , the question remains how universal these models are and to what extent they should be conditioned on individual or demographic factors. 2 Materials and Method 2.1 Corpus The analyses were based on the Fisher corpus of spoken telephone conversations in US English [ 27 ] , collected by the Linguistic Data Consortium (LDC). We used both part 1 and 2 of the dataset, containing 10,950 English telephone conversations with 11,684 different speakers. Each individual speaker was involved in up to four different conversations. For our analyses, we count each speaker in each conversation as a data point, resulting in 21,900 data points. For the data collection, participants were randomly paired and engaged in spoken dialogues lasting up to 10 minutes, covering a wide range of topics that were given to the participants beforehand. One of the goals was

Read the complete text