Content uploaded by Nicol A. Arellano Véliz
Author content
All content in this area was uploaded by Nicol A. Arellano Véliz on Apr 03, 2024
Content may be subject to copyright.
Available via license: CC BY 4.0
Content may be subject to copyright.
Beyond Words: Speech Synchronization and
Conversation Dynamics Linked to Personality and
Appraisals
Nicol Alejandra Arellano-Véliz
University of Groningen
Ramón Daniel Castillo
University of Talca
Bertus F. Jeronimus
University of Groningen
Elske Saskia Kunnen
University of Groningen
Ralf F.A. Cox
University of Groningen
Research Article
Keywords: interpersonal coordination, synchrony, dyadic interactions, personality, speech coordination,
speech synchrony
Posted Date: March 28th, 2024
DOI: https://doi.org/10.21203/rs.3.rs-4144982/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: No competing interests reported.
Beyond Words: Speech Synchronization and Conversation Dynamics Linked to Personality and Appraisals
Abstract
We studied how personality differences and conversation topics predict interpersonal speech synchronization,
leading/following dynamics, and nonverbal interactional dominance in dyadic conversations. 100 undergraduate
students (50 same-gender dyads) had a 15-minute conversation following three topics (introduction/self-
disclosure/argumentation) in our laboratory. Their speech synchronization and turn-taking (speech/silence)
dynamics were assessed through nonlinear time-series analyses: Cross-Recurrence Quantification Analysis (CRQA),
Diagonal Cross-Recurrence Profiles (DCRP), and Anisotropic-CRQA. From the time series, we extracted five
variables to operationalize speech synchronization (global and at lag-zero), leading-following dynamics, and
asymmetries in the interacting partners’ nonverbal interactional dominance. Interaction appraisals were also
assessed. Associations between personality traits Extraversion/Agreeableness, speech synchronization, and
nonverbal interactional dominance were tested using mixed-effects models. Speech synchronization and nonverbal
interactional dominance differed across conversational topics and peaked during argumentative conversations.
Extraversion was associated with increased speech synchronization, and nonverbal interactional dominance,
especially during an argumentative conversation. Extraversion homogeneity was associated with more symmetry in
turn-taking dynamics during a self-disclosure conversation. Speech synchronization was generally associated with
positive post-conversational appraisals such as wanting to meet in the future or liking the conversation partner,
especially in extroverted individuals, whereas introverts seemed to value less swift dynamics. High Agreeableness
predicted less speech synchronization during argumentative conversations, and increased speech synchronization (at
lag-zero) predicted reduced perceived naturality in agreeable individuals. This may suggest a trade-off between
maintaining swift speech dynamics and the natural flow of conversation for individuals high in Agreeableness.
Keywords: interpersonal coordination, synchrony, dyadic interactions, personality, speech coordination, speech
synchrony
1
1. Introduction
The course and dynamics of a conversation between two partners (a dyad) and how they experience the
interaction are affected by contextual factors (i.e., situations) and individual differences (e.g., personality traits, see
Harley, 2013). Interpersonal speech dynamics such as the temporal attunement of speech and silence turns show
how dyads synchronize during conversations (through turn-taking dynamics), the leading-following dynamics they
exhibit, and how potential nonverbal interactional dominance asymmetries emerge from the mutual influence of
interacting partners. We present a study with four aims. First, we use complex dynamical systems theory and speech
recordings during conversations to quantify differences in overall speech synchronization through turn-taking
behaviors, leader-follower dynamics (temporal domain), and differences or asymmetries in nonverbal interactional
dominance between the interaction partners (e.g., one person tends to “dominate” the conversation to a greater
extent through speech or silence episodes). Second, we examine whether these speech dynamics differ across three
types of conversations, introduction, self-disclosure, and an argumentative conversation. Third, we examine how
dyadic speech dynamics differ as a consequence of their personality traits, and study dyadic combinations (one or
both low/high scores) of Extraversion (sociability) and Agreeableness (nurturance). Fourth, we examine how
interpersonal speech synchronization and personality traits influence how both interacting partners appraise their
conversation in terms of the perceived quality of the interaction and rapport. We conclude by discussing our study
results and their fit in the broader literature of interaction and personality theory.
1.1. Interpersonal Speech Dynamics
Human communication extends beyond spoken words and comprises a complex flow of interpersonal
dynamics within conversations. Language, viewed as a complex adaptive system, operates through a set of
interacting elements distributed across the body and social environments (Ellis & Larsen-Freeman, 2009; Di Paolo
et al., 2018; Lund et al., 2022). These language elements modulate perceptions, emotions, and thoughts, and convert
these experiences into meaningful language expressions (Scheidt et al., 2021). In this way, the behavior of each
interacting partner is influenced by characteristic adaptations –i.e., perceptual information, situations, social
motivations, and other individual differences (Asendorpf, 2017; Beckner et al. 2009; Mischel & Shoda, 2005), such
as gaze, gestures, movement, and speech synchronization (Fusaroli et al., 2014). Recognizing language as a complex
adaptive system has implications for interpersonal settings when interacting partners are connected via verbal and
nonverbal communication (Thibault, 2004; Falandays et al., 2018; Scheidt et al., 2021). These joint dynamics
2
emerge in dyadic interactions where both partners shape the unfolding conversation (Reuzel et al., 2013), and dyads
constitute the majority of human social interactions (Peperkoorn et al., 2020).
People mutually adjust in a conversation in terms of turn-taking, pauses, speech duration, speech rate,
response latency, vocal intensity, and movement, among others, which synchronize across timescales (Fowler et al.,
2008; Reuzel et al., 2014; Bloomfield et al., 2021). Such interpersonal speech dynamics are key to this study in
which we focus on speech synchronization through turn-taking behaviors, the presence of leader-follower dynamics,
the differences or asymmetries in nonverbal interactional dominance between the interaction partners; as well as
individual differences in the personality traits Extraversion and Agreeableness in these conversation dynamics.
1.1.1. Functional Synchronization in Social Interactions
Speech synchronization can be described in terms of self-organized sets of coupled components that behave
as a single functional unit –such as a conversation (Bernieri et al., 1988; Shockley et al., 2009). Such sets of
components are self-organized and context-sensitive because each individual actively structures exchanges with the
environment to generate and maintain systemic stability (Varela et al., 1991, 2017). In this way, interacting partners
develop stable interaction patterns that are softly assembled (context-sensitive), self-organized, and adapted to reap
their opportunities (or affordances) and goals in specific situations, whether those motives are affiliative,
competitive, or problem-solving, among others (Fusaroli et al., 2014; Kelso et al., 1984; Shockley et al., 2003). In
essence, the actions of one member of the dyad impact the actions of the other as they start behaving as a coupled
system (Shockley et al., 2009).
In the context of our study, speech synchronization refers to interdependent dynamics among
conversational elements (speech and silence turns) that exhibit coherence and are coupled over time. Speech
synchronization does not necessarily mean that interacting partners show the same action simultaneously as co-
occurring states, but involves compensatory dynamics that allow for the emergence of functions to achieve specific
goals (Nowak et al., 2017). Therefore, we employ the definition of speech synchronization in terms of reciprocity
and the nonverbal coordination of turn-taking behavior, which creates a conversational rhythm or confluence of
performances (Reuzel et al., 2013). We explore speech synchronization, the temporal dimension of leading and
follower dynamics, and differences or asymmetries in nonverbal interactional dominance through the turn-taking
behaviors exhibited in conversations (Reuzel et al., 2014).
3
1.1.2. Nonverbal Interactional Dominance in Speech and Personality
Speech synchronization occurs as interacting partners find an optimal, predictable, and cyclic rhythm in the
conversation, which allows them to take turns, and exhibit fewer interruptions or prolonged silences (Warner, 1992,
Reuzel 2013). These optimal and synchronized rhythms of communication are linked to affiliation (Hove & Risen,
2009), cooperative efficiency (Delaherche et al., 2012), and positive affect (Warner, 1992). Speech is not distributed
equally in interpersonal interactions (Bales, 1973), and sometimes one partner influences the interaction and talks
more (i.e., interactional dominance), which is associated with prestige and greater influence in decisions in
interactions, as well as higher status in other contexts (Meeker, 2020). According to some authors, the speed of
speech onset, or given opportunities to speak (response latency), indicates conversation dominance (Berger et al.,
1972; Fişek et al., 2005; Meeker, 2020). Other nonverbal indicators of communication imbalances include temporal
leader-follower dynamics, regarding having the initiative in the turn-taking structure (Reuzel et al., 2013, 2014).
People who more often initiate speech, and lead the dyadic conversation dynamics over time resulting in
asymmetries or imbalances in conversation dynamics. On the other hand, possible differences in nonverbal
interactional dominance between the interaction partners refer to the extent and average duration of asymmetric
episodes in the influence of one conversation partner's nonverbal behavior on the other (Reuzel et al., 2014).
Conversation dynamics, such as speech synchronization, leader-follower dynamics, and nonverbal
interactional dominance, allow individuals to fulfill their communicative needs. However, conversation dynamics
are also influenced by situational factors and other differences between the interacting partners (Mischel & Shoda,
2005). Thus, we study how variation in Extraversion and Agreeableness are associated with such dynamics.
1.1.3. Personality Conceptualizations and the Emergence of Dyadic Systems
Dynamic personality models suggest that although human affect, behavior, thoughts, desires, and
“predispositions” for action change continuously, due to the interplay of intrinsic mechanisms and external forces,
they converge into stable personality patterns over time (Bleidorn et al., 2022 for lifespan trajectories, also Nowak et
al., 2005; Sosnowska et al., 2019; Revelle & Wilt, 2020). Personality differences have also been operationalized as
tendencies or patterns to optimally engage with the world, thus stylistic differences in person-environment fit
(Hovhannisyan & Vervaeke, 2022). Personality is then a label for how individuals structure and navigate the flow of
(social) affordances across contexts (Gibson, 1979; Chemero, 2003; Satchell et al., 2021). This more contextualized
4
and dynamic view of personality is broader than trait concepts known as characteristic adaptations (Nguyen et al.,
2021).
Personality describes how system components differ between conversation partners, and similarity and
dissimilarity in personality may promote or hamper the emergence of conversational elements at the higher-order
level of the dyadic system that were not present for each individual alone (Mischel & Shoda, 2005; Nowak et al.,
2020). We test how the personality characteristics of each interaction partner affect dyadic speech synchronization
during different conversation types, and how interacting partners appraise these interactions.
We focus on the personality dimensions of Extraversion and Agreeableness, as they are key to social
behavior (Goldberg et al., 1998; McCrae & Costa, 2008; Koole et al., 2001). The three other broad personality
dimensions of the Five Factor Model are more relevant to other outcome domains, such as work
(Conscientiousness), affective experience and health (Neuroticism), and intellectual/creative life (Openness; see
Peabody & Goldberg, 1989; Cuperman & Ickes, 2009; Larsen et al., 2020). Extraversion captures one’s sociability
and high scores indicate liveliness, gregariousness, cheerfulness, outgoing, and adventure-seeking behaviors,
compared to introverted individuals (McCrae & Costa, 2008). Agreeableness reflects the tendency toward altruism,
cooperation, and prosocial interactions (Larsen et al., 2020; McCrae & Costa, 2008), while “disagreeable”
individuals tend to lack concern for others (De Young, 2015; Hovhannisyan & Vervaeke, 2022).
Previous studies of individual differences in speech patterns demonstrated that the amount of speech of an
individual will be affected by factors such as the talkativeness of the interacting partner (Borgatta & Bales, 1953;
Leaper & Ayres, 2007; Oben & Brône, 2016), which partly reflects their personality traits (Cuperman & Ickes,
2009). For instance, two extroverts are likely to cover a wide variety of themes and more relatable/common ground
topics in their conversation, while two introverted speakers tend to be more concise and engaged in focused problem
themes (Thorne, 1987). Dyads composed of two introverted or two extroverted individuals are likely to take more
and longer speaking turns, and to disclose more personal information, versus mixed dyads (introverted/extroverted,
see Arellano-Véliz et al., 2024; Cuperman & Ickes, 2009). Homogeneous introverted or extraverted dyads evaluated
their interactions as more positive, natural, and engaging, and felt more accepted and encouraged to lead the
conversation and to interact more in the future. In the case of Agreeableness, mixed dyads (agreeable/’disagreeable’)
tended to self-disclose more personal information, and agreeable individuals tended to evaluate their interaction
more positively (vs. disagreeable individuals; also see Arellano-Véliz et al., 2024). Overall, the question remains
5
whether (higher-order) conversation constraints such as a conversational topic and personality traits can play a
significant role in explaining speech synchronization.
1.2. The present study
We propose that speech synchronization dynamics are an emergent factor with a complex nature, and are
influenced by personality and contextual differences, among others. To study interpersonal speech dynamics, we
analyze conversational behavior (non-verbal) in time series through the dynamics of overall speech synchronization,
leader-follower dynamics, and differences or asymmetries in nonverbal interactional dominance. We aimed to
elucidate the influence of (high-level) situational constraints (e.g., Paxton & Dale, 2017), which we operationalized
as specific conversational topics (i.e., introduction, self-disclosure, argumentative), and individual differences in the
personality traits Extraversion and Agreeableness. Furthermore, we scrutinized how these factors shape interaction
partners' perceptions and evaluations of the conversation.
1.2.1. Cross-Recurrence Quantification Analysis to Quantify Synchronization of Speech, Leader-Follower
Dynamics, and Nonverbal Interactional Dominance
To explore speech synchronization, leader-follower dynamics, and differences or asymmetries in nonverbal
interactional dominance we employed three related nonlinear time-series techniques based on Cross-Recurrence
Quantification Analysis (CRQA, e.g., Zbilut & Webber, 1992; Marwan et al., 2007; Cox et al., 2016), which are
outlined in Table 1 and detail in the Methods section. CRQA offers a powerful means to study the temporal patterns
of speech dynamics in interpersonal settings and is especially suitable for studying temporal interdependencies
between two time series (e.g., Zbilut et al., 1998; Marwan et al., 2007; Cox et al., 2016). We focus on instances
where turn-taking behavior was observed as one person was speaking and the other was silent (similar to Reuzel et
al., 2013, 2014).
The nonlinear time-series techniques allowed us to study interpersonal speech dynamics in three different
domains. First, we measured speech synchronization (through turn-taking dynamics) globally (at all lags), and
simultaneously (at lag-zero), by identifying occurrences where one interacting partner is silent while the other is
speaking. Second, we analyzed balance during conversations by quantifying the overall leading-following dynamics
(imbalance in having the initiative in turn-taking structures). Third, we explored asymmetries in nonverbal
interactional dominance by assessing the influence of one interacting partner over the other (i.e., the influence of
6
one interacting partner on the other). Detailed information about the techniques and variables extracted can be found
in the Methods section and Table 1.
Complex dynamic systems techniques can help to study dyadic interaction dynamics and the role of
personality traits that surface in reciprocal influences between individuals (e.g., Mischel & Shoda, 2005). Individual
differences influence the dyadic system which has emergent properties that would not be present otherwise, for
example, a talkative person can generate more opportunities for action in a dyadic interaction. This integration of
CRQA with dyadic systems and personality traits provides a comprehensive framework for understanding the
multifactorial nature of interpersonal dynamics.
1.2.2. Expectations
This study aimed to answer four research questions with the following hypotheses (See Table 1 for the
operationalization of the variables).1
1) How are speech synchronization, leader-follower dynamics and differences/asymmetries in nonverbal
interactional dominance explained/influenced by (H1a) conversational constraints i.e. different conversational
topics? (H1b).
First, we expected that the type of conversation would significantly explain part of the variance of speech
synchronization, and that self-disclosing (e.g., Thorson et al., 2021) and argumentative (Tschacher et al., 2018)
conversations result in higher synchronization (H1a; see Arellano-Véliz et al., 2024). In the first case, it could be
linked to affiliative drives, whereas in the latter, it could be related to competition (Tschacher et al., 2018) or
achievement goals (Allsop et al., 2016). Generally, differences in speech synchronization, leader-follower dynamics,
and nonverbal interactional dominance (H1b) were expected, especially in the argumentative task compared to the
introduction.
2) To what extent do personality traits and dyad composition explain variation in speech synchronization?
We focused our primary hypotheses on the social traits of Extraversion and Agreeableness based on
previous studies (Funder & Sneed, 1993; Cuperman & Ickes, 2009; Arellano-Véliz et al., 2024).
High Extraversion is characterized in the literature by social enjoyment and talkativeness; while low
Extraversion has been linked to (perceived) socially reserved and ‘awkward’ behavior (Funder & Sneed, 1993;
1 Since this study is part of a larger data collection and a subsample was considered, we delineate our hypotheses
based on Arellano-Véliz et al. (2024) which focused on synchronization and coupling of body motion, as well as
pioneer literature by Funder & Sneed (1993), Cuperman & Ickes (2009) and Tschacher et al. (2018).
7
Cuperman & Ickes, 2009). Therefore, higher scores on Extraversion were expected to increase dyadic speech
synchronization while introverted dyads were expected to synchronize less (H2a). When at least one of the dyadic
partners is extroverted we expect more speech synchronization (H2b; Arellano-Véliz et al., 2024).
Additionally, we expected that high scores on Agreeableness would be associated with stronger dyadic
coupling thus increased speech synchronization (H2c, vs. disagreeable dyads).2 The link between Agreeableness and
dyadic synchronization roots in the reported warmth, ‘cheerful’, and friendly behavior linked to this trait (Funder &
Sneed, 1993; Cuperman & Ickes, 2009), which would contribute to better dyadic coupling (e.g., Arellano-Véliz et
al., 2024). Low Agreeableness was expected to hamper dyadic synchronization in dyads with at least one low-
agreeable individual (H2d). An exception to this would be the argumentative conversation, where competing goals
can increase the synchronization of speech in low-agreeable individuals (H2e).
3) To what extent do personality traits and dyad composition explain variation in distinct patterns of leader-follower
dynamics and differences/asymmetries in nonverbal interactional dominance?
Higher scores of Extraversion would be associated with a more imbalanced distribution of speech when
interacting with introverts, where they could take a leading behavior and exhibit asymmetries in nonverbal
intersectional dominance (H3a). Higher scores on Agreeableness are expected to explain balanced and symmetric
dynamics of speech in the conversations, conversely to individuals with low scores on Agreeableness (H3b).
4) How do speech synchronization, leader-follower dynamics, and nonverbal interactional dominance impact the
perceived quality of interactions among participants?
We expected that higher levels of synchronization of speech (H4a), balanced (leader-follower), and
symmetric (lower differences in nonverbal interactional dominance) interactions (H4b) would be positively linked
to perceived interaction quality, indicating that individuals who exhibit more aligned speech patterns will perceive
the interaction more positively. In addition, based on previous studies, it is expected that higher scores on
Extraversion (H4c) and Agreeableness (H4d) will explain positive perceptions of the interaction, especially linked to
synchronized interactions.
2 We based this hypothesis in the results found in (Arellano-Véliz et al., 2024) evidencing increased entropy
(interpreted as coupling) between agreeable individuals.
8
2. Method
2.1. Participants
Data presented in this paper were collected between 2021 and 2022 in the context of a larger multimodal
experimental project (see Arellano-Véliz et al., 2024). From 300 screened participants, 112 undergraduate students
participated in a 15-minute conversation in same-sex dyads in a laboratory room. A subsample of 100 participants
(50 same-sex dyads) aged 18-33 (mean = 20.54, SD = 2.74; 72 females, 28 males) constituted the final sample, as
these data were suitable for study analyses (i.e., audio files integrity was adequate to be analyzed, which was not the
case in six dyads). They received ECTS credits for participating in the study and provided informed consent
according to the ethical requirements to conduct research with human participants (ethical approval code PSY-1920-
S-0525).
In principle, we designed the laboratory study by focusing on the (dis)similarity in socially relevant traits
scores, particularly, Extraversion and Agreeableness, either 0.5 SD above (“high”) or 0.5 SD below (“low”) the
sample mean (e.g., Li et al., 2020). In our models, we used the full sample of 100 participants as we modeled the
personality traits of interest continuously while preserving the dyadic structure in a parsimonious way. In addition,
we provide plots and descriptives following this thresholding dyadic matching (similar to the threshold described by
Li et al., 2020 for low/high scores and the approach of Cuperman & Ickes, 2009).
2.2. Instruments
2.2.1. Big Five Personality Traits: International Personality Inventory Pool—120 (IPIP-NEO-120)
Personality traits were assessed utilizing the IPIP-NEO-120 (Johnson, 2014) through the Qualtrics online
platform before the laboratory study was conducted, approximately 10 days before the laboratory study took place,
and participants provided informed consent for this screening part of the study. The IPIP-NEO-120 is a self-report
questionnaire comprising 120 items that measure the Big Five personality traits Extraversion, Neuroticism,
Openness to Experience, Agreeableness, and Conscientiousness, and their 30 facets (5*6). The completion of the
questionnaire takes between 10 and 20 minutes on average according to the author (Johnson, 2014). The
psychometric properties (Johnson, 2014) are consistent with the psychometric properties of the NEO-PI-R scales
(McCrae & Costa, 2008), which indicates that the IPIP-NEO-120 is reliable and valid. Furthermore, in a sample of
501 individuals, the scales of the IPIP-NEO-120 and NEO-PI-R exhibited high correlations (Extraversion 0.85,
Neuroticism 0.87, Openness to Experience 0.84; Agreeableness 0.76 and Conscientiousness 0.80 (all p <.01),
9
Johnson, 2014). The IPIP-NEO-120 exhibited good internal consistency (Cronbach's alpha of 0.84, 0.88, 0.85, 0.81,
and 0.84, respectively). The IPIP-NEO-120 is publicly available and has cross-cultural robustness, thus, suitable for
an international sample.
2.2.2. Self-disclosure paradigm
We employed the self-disclosure paradigm to create an affiliative conversation (Aron et al., 1997). This
paradigm consists of an experimental protocol in which both interacting partners ask and answer a set of questions,
which become increasingly more personal. The protocol aims to create interpersonal closeness in an experimental
context. The original version has three sections with 12 questions each, taking approximately 45 minutes to
complete. We shortened the protocol to three sets of questions each (9 questions in total) since this part of the
conversation lasted 5 minutes in our study. They were asked to choose at least one question from each set (e.g.,
“What would constitute a ‘perfect’ day for you?”; “Is there something that you’ve dreamed of doing for a long
time? Why haven’t you done it?”; “How do you feel about the relationship with your family?”).
2.2.3. Perception of the Interaction (appraisals)
After the dyadic conversation, participants were asked to complete a modified version of the Perception of
the Interaction questionnaire (Cuperman & Ickes, 2009) to assess the self-reported interaction experience. The
scores go from 1 (“not at all”) to 5 (“very much”). The original version of this questionnaire assesses the first-person
perspective (e.g., “To what extent did you feel accepted and respected by the other person”?) and the third-person
perspective (e.g., “To what extent do you think your conversation partner felt accepted and respected by you?). In
the present study, we report the first-person questions, as our research focused on individual experience, instead of
actor-partner interdependence effects. To preserve nuanced interpretations the Perception of Interaction
questionnaire items were used as variables, each provided in Table 2, instead of clusters of items, following the
precedent by Funder and Sneed (1993) and Cuperman & Ickes (2009).
2.3. Procedure
Participants were invited to participate in the experimental study, and provided with a heart rate transmitter
belt when they arrived in the laboratory, although these data are not part of this paper. After their arrival,
participants were asked to read and sign the informed consent, and then fill in an affect score. Subsequently, a
microphone was attached to their clothing, and instructions about the conversation task were given. Participants
engaged in their interpersonal tasks standing face-to-face on a balance board (designed for measuring postural
10
control, also not reported in this paper), at a fixed distance of 1.5 meters. A camera positioned approximately 4.5
meters away recorded the interaction from a sagittal perspective.
The conversation followed an approximately 15-minute semi-structured interaction schedule with three
parts of approximately 5 minutes each, although participants had the freedom to finish each part before moving to
the next one. In each conversation, participants (1) introduced themselves, (2) self-disclosed, and (3) engaged in an
argument or debate. In the introduction phase, participants were instructed to briefly introduce themselves to their
interacting partner within 5 minutes. Some general themes were provided as examples in case guidance was needed.
The subsequent "self-disclosure" phase involved participants asking each other questions using a modified version
of Aron et al.'s (1997) self-disclosure paradigm. Participants selected questions to discuss for 5 minutes, and both
individuals were required to answer each question. They were given the freedom to decide what they wanted to
share with their partner. In the argumentative phase, participants selected a conversational topic from a pool of
around 20 and took opposing sides (pro/against) on topics such as "Are strict lockdowns a valid measure during the
pandemic to keep people safe?”; “Are dating apps a good platform for meeting a romantic partner?"; “Should pre-
adolescents and adolescents use social media freely?” For these example items, participants had to choose a
position, such that the dyads had conflicting arguments, and they discussed as many topics within 5 minutes, with
the time controlled by an alarm. However, participants were instructed to conclude their conversation before moving
on to the next phase, therefore, each part could last for longer than 5 minutes if necessary. Following the interaction,
each participant completed questionnaires on affective state, interpersonal closeness, and interaction appraisals.
2.4. Data Processing and time series generation
Data streams were recorded using the Lab Streaming Layer software (Kothe et al., 2019) and each audio
stream was cleaned using Adobe Audition to remove the background noise by employing the default ‘noise print’
and ‘DeNoise’ functions on each audio file until the background noise and the voice of the interacting partner in the
background were mostly canceled. With the resulting cleaned files we used the voice activity annotation function on
the Praat software (Boersma et al., 2023) to code silence and speech segments, being categorized as ‘0 = silence’
and ‘1 = speech’ similar to the procedure described by Reuzel et al (2014). This procedure resulted in a dichotomous
time series consisting of 1’s and 0’s for each interacting partner (see Figure 1). Since the software is sensitive to the
phonetic level (phoneme and syllable boundaries), we resampled the time series closer to the utterance level,
11
defining 1 second per silence and speech segments (e.g., see Situation Model by Pickering & Garrod, 2004; Abney
et al., 2014).
Figure 1. Representation of the time series generation
Note: The figure represents the coding of the time series, which represents the speech (1) and silence (0) segments
by seconds in the time series of each interacting partner (Figure adapted from Reuzel et al., 2013).
2.5. Time series technique and statistical analyses
2.5.1. Categorical Cross-Recurrence Quantification Analysis (CRQA)
Cross-Recurrence Quantification Analysis (CRQA) was used to quantify the degree of speech
synchronization or dyadic coupling. CRQA is a nonlinear bivariate correlation technique based on recurrence
analysis to quantify the temporal similarity or coupling properties between time series (Zbilut et al., 1998; Marwan
et al., 2007; Wallot & Leonardi, 2018). CRQA is the extended bivariate form of the original Recurrence
Quantification Analysis (RQA), and it is used to study the coupling of two time series, in the case of our study, both
conversational partners (Marwan et al., 2007). The main tool, the cross-recurrence plot is built by identifying and
noting down in a 2D plot all instances where the behavior of the two-time series match (e.g. Cox et al., 2016; Xu et
al., 2020; Wallot & Leonardi, 2018; see Figure 2), and these recurrence measures describe the temporal dynamics of
the interacting systems across all possible lags or time scales (Zbilut et al., 1998; Marwan et al., 2007).
We employed a categorical form of CRQA given the categorical (dichotomous) nature of our data,
computed with one time series for each interaction partner with codes “1” and “0” for speech and silence
occurrences respectively (see Figure 1). In this case, a match (i.e. recurrence) is noted down as a dot in the cross-
recurrence plot, when one interaction partner is silent and the other one speaks (which are “1" and “0” or “0” and
“1” combinations in the respective time series). Consequently, a line in the cross-recurrence plot (i.e. sequence of
matching points), corresponds to the prolonged co-occurrence of speech and silence. CRQA enables us to quantify
the synchronization or behavioral attunement between interacting partners during the conversation, from the
12
temporal patterns present in the time series (Reuzel et al., 2013). CRQA identifies recurrences in the behavior
between dyadic partners throughout the conversation, which can be simultaneous or in temporal proximity (with
some delay) (Cox et al., 2016).
We conducted two follow-up analyses: Diagonal Cross-Recurrence Profile (DCRP) and anisotropic CRQA
(aCRQA). With DCRP we analyzed the leader-follower dynamics in the conversation. That is, it quantified the
imbalance between the interaction partners with respect to having the initiative in the turn-taking structure. aCRQA
enabled us to investigate possible differences in nonverbal interactional dominance between the interaction partners.
That is, it quantified the extent and average duration of asymmetric episodes in the influence of one conversation
partner's nonverbal behavior on the other. Each of these analyses will be explained in more detail below. All
variables in this study are defined in Table 1, such as RRglobal, which quantifies the recurrence rate across the entire
cross-recurrence plot.
2.5.2. Diagonal Cross-Recurrence Profile (DCRP)
We examined balance/imbalance in leading-follower dynamics during the dyadic conversations with
Diagonal Cross-Recurrence Profile analyses (DCRP, Richardson & Dale, 2005). DCRP quantifies the number of
recurrences at different lags across the main diagonal or line of synchrony of the cross-recurrence plot, see Figure
2B (Wallot & Leonardi, 2018; Tomashin et al., 2022). The main diagonal of the cross-recurrence plot is called the
Line of Synchrony (LOS) and captures simultaneous matching behaviors at lag-zero. That is, occurrences when one
participant speaks (code “1”) while the other is silent (code “0”), at the same time. The diagonals parallel to the LOS
display instances where these recurrences (matching behaviors) occur with some delay, which increases the further
one moves away from the LOS. The distribution of recurrences at one side of the LOS indicates how much (and how
quickly) the behaviors of one participant in the time series are followed by the matching behaviors in the other
participant, for different lags. Importantly, it is likely that the recurrences on both sides of the LOS are
asymmetrically distributed (Wallot & Leonardi, 2018), creating a diagonal cross-recurrence profile (DCRP). The
DCRP quantifies leader-follower imbalances in the conversation (Dale et al., 2011; López Pérez et al., 2017). We
computed the absolute Quotient of the DCRP (QDCRP) to indicate the overall conversational imbalance in leading and
following between the interaction partners (see Table 1; Richardson & Dale, 2005; Dale et al., 2011).
13
2.5.3. Anisotropic CRQA (aCRQA)
To further examine asymmetries in nonverbal interactional dominance in conversations we used
Anisotropic CRQA (aCRQA; Cox et al., 2016; Xu et al., 2020). Categorical time series tend to form horizontally or
vertically oriented rectangular structures in cross-recurrence plots which indicate coupling between two time series
(see Figure 2A). aCRQA quantifies the relative influence of each interacting partner on the other by capturing
matched behaviors over time and separately quantifying the distribution of vertical and horizontal structures in the
cross-recurrence plot (Cox et al., 2016; Xu et al., 2020). These structures represent behaviors performed by each
partner, matched for extended periods by the other. Thus, one partner effectively captures the behavior of the other
during certain episodes. The difference between horizontal and vertical structures indicates whether one interaction
partner captures the other partner to a larger degree. aCRQA analyzes horizontal lines and vertical lines separately
and quantifies their (relative) differences, in amount, and length, among others. Differences in these measures
indicate an asymmetry in dominance, that is, unequal coupling strength between the interacting partners (Cox et al.,
2016; López Pérez et al., 2017).
Differences in Laminarity (LAMARD) and Trapping Time (TTARD) quantify overall asymmetries in the
conversation (see Table 1). It is important to note that asymmetries in nonverbal interactional dominance indicate
that one of the interaction partners tends to exhibit more control, capturing the other partner into matching behaviors
for extended periods (van Dijk et al., 2018). This is not necessarily a negative aspect of the interaction, as the term
‘dominance’ might suggest. Dominance merely describes potential asymmetries in the coupling strength of the
dyadic system during the conversations. We focus on turn-taking dynamics (speech and silence episodes) during the
conversation, expressed by horizontal and vertical rectangular structures in the plot, which capture the behaviors
displayed by each interaction partner respectively. These behaviors can also represent the creation of opportunities
(or affordances) for episodes of speech and silence in the conversation. Therefore, nonverbal interactional
dominance can have a positive social communicative effect in such cases.
14
Figure 2.
Cross-Recurrence Plot and Diagonal Cross-Recurrence Profile
Note: Panel 2A. The figure represents a categorical cross-recurrence plot. In this case, the matching between the two
interacting partners (time series) was defined as one person speaking (categorized as 1) and the other person being
silent (categorized as 0). In the plot, these occurrences are represented by the blue lines or blocks, whereas the non-
occurrences –both speaking (1-1) or both in silence (0-0)– are represented by white spaces. Panel 2B. Diagonal
Cross-Recurrence Plot (DCRP) representation. Adapted from Wallot & Leonardi (2018). The profile can help
determine the coupling direction of time series in terms of leading and following dynamics at different lags along
the line of synchrony (LOS). Each line parallel to the line of synchrony represents a particular delay or lag in the
alignment of speech dynamics between both interacting partners. A lag of 0 indicates synchrony or simultaneous
recurrence. In the context of turn-taking, this represents moments when both individuals are engaged in speaking or
listening at the same time, suggesting coupling, reciprocity, and attunement.
15
Table 1.
Cross-Recurrence Quantification Analysis: Measures, Description, and Interpretations
Measure
Formula and Description
Interpretation
Overall Speech Synchronization (coupling) - CRQA
Recurrence
Rate
RRglobal
𝑆𝑢𝑚 𝑜𝑓 𝑟𝑒𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑝𝑜𝑖𝑛𝑡𝑠 𝑖𝑛 𝐶𝑅 𝑝𝑙𝑜𝑡
÷ 𝑠𝑖𝑧𝑒 𝑜𝑓 𝐶𝑅 𝑝𝑙𝑜𝑡
Measures the overall rate (proportion) of recurrent
points (matches) between the two time series across
the entire cross-recurrence plot. Matching was
specified as one person speaking while the other
was silent (1-0, 0-1 occurrences).1
RRglobal estimates synchronization across all
possible lags (delays). Higher RRglobal indicates
more speech synchronization across all possible
lags (greater engagement in turn-taking episodes),
and more structured and responsive interactions. A
low(er) RRglobal indicates less synchronization,
more interruptions, silences, and a more irregular
pattern of coupling.1
Recurrence
Rate through
the Line of
Synchrony
RRLOS
𝑆𝑢𝑚 𝑜𝑓 𝑟𝑒𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑝𝑜𝑖𝑛𝑡𝑠 𝑜𝑛 𝑡ℎ𝑒 𝑙𝑖𝑛𝑒
𝑜𝑓 𝑠𝑦𝑛𝑐ℎ𝑟𝑜𝑛𝑦
÷ 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑙𝑖𝑛𝑒 𝑜𝑓 𝑠𝑦𝑛𝑐ℎ𝑟𝑜𝑛𝑦
(𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑡𝑖𝑚𝑒 𝑠𝑒𝑟𝑖𝑒𝑠)
The rate (proportion) of recurrence (matching) on
the line of synchrony represents the instances
where speech and silence co-occur
(simultaneously). It represents the percentage of
synchrony.1,2,3
RRLOS is a measure of synchrony or coupling in
speech at lag-zero (simultaneous). Higher RRLOS
suggests more co-occurrence of speech and silence
thus synchronization at the same time and,
therefore, responsive interactions. Conversely, a
lower RRLOS indicates less synchronization, more
interruptions, and silences at lag-zero
(simultaneously).1
Leading-Following Dynamics (DCRP)
Quotient of
Diagonal
Cross
Recurrence
Profile,
QDCRP
(absolute)
|(𝑅𝑅𝑟𝑖𝑔ℎ𝑡 − 𝑅𝑅𝑙𝑒𝑓𝑡 ÷ 𝑅𝑅𝑟𝑖𝑔ℎ𝑡 + 𝑅𝑅𝑙𝑒𝑓𝑡)|
RRright and RRleft refer to the recurrence rates on the
left and right sides of the LOS, respectively, within
the DCRP (see Figures 2B and 6).1, 2, 3
QDCRP indicates the absolute overall degree of
balance/imbalance between the left and right sides
along the LOS.4 Leading-follower dynamics can be
inferred from this measure, where 0 represents a
totally balanced interaction (i.e. equally leading
and following), and 1 represents the situation
where one of the interacting partners is leading
during the entire interaction.1
16
Asymmetries in Nonverbal Interactional Dominance (aCRQA)
Relative
Difference in
Laminarity
of
Anisotropic
CRQA,
LAMARD
|(𝐿𝐴𝑀𝑣𝑒𝑟 − 𝐿𝐴𝑀ℎ𝑜𝑟)| ÷ (𝐿𝐴𝑀𝑣𝑒𝑟
+ 𝐿𝐴𝑀ℎ𝑜𝑟)
LAMver, and LAMhor refer to the Laminarity of
vertical (interacting partner “A”) and horizontal
lines (interacting partner ”B”), respectively.
Laminarity is a recurrence measure that quantifies
the presence of vertical and horizontal lines in the
DCRP, indicating periods of repeated fixed
behavioral patterns in the dynamics of the
conversation. The relative difference in laminarity
compares the laminarity along specific directions
(i.e. horizontal and vertical), reflecting variations in
the leading-following (balance) dynamics between
conversational partners.4
Asymmetric laminar patterns indicate conversation
dominance. A higher relative difference in
laminarity suggests the overall asymmetry's
magnitude in the interactions' nonverbal
interactional dominance. If one dyadic partner
more often initiates the talking and silences this
steers episodes of silence or speech in the other.
Hence, one person “affords” or gives the other
opportunities to display behavior to the partner,
and thus is more “dominant” or influential in the
conversation.4 Low LAMARD implies a more
“balanced” interaction when nobody consistently
takes the lead over the other.
Relative
Difference in
Trapping
Time of
Anisotropic
CRQA,
TTARD
|(𝑇𝑇𝑣𝑒𝑟 − 𝑇𝑇ℎ𝑜𝑟)| / (𝑇𝑇𝑣𝑒𝑟 +𝑇𝑇ℎ𝑜𝑟)
TTver and TThor refer to the Trapping Time of
vertical (interacting partner “A”) and horizontal
lines (interacting partner ”B”) respectively.
Trapping Time quantifies the average duration for
which one system "traps" or captures the behavior
of the other system. It quantifies the average
temporal persistence of one system's influence on
the other.4
One partner's speech and silence episodes lead
("trap") the other partner in relatively longer or
shorter episodes of silence and speech,
respectively. A longer trapping time indicates a
sustained influence of one speaker on the other or
nonverbal interactional dominance. A shorter
trapping time su
ggests swift reciprocal interactions, where both
speakers influence one another. It can be indicative
of a higher interactional sensitivity or responsivity.4
Note. CR plot= Cross-Recurrence plot, RR= Recurrence Rate, CRQA= Cross-recurrence quantification analysis.
aCRQA= Anisotropic Cross-recurrence quantification analysis. DCRP= Diagonal Cross Recurrence Profile. LAM=
Laminarity. LOS= Line of Synchrony. TT= Trapping Time. References: 1 Reuzel et al. (2014); 2 Richardson & Dale
(2005), 3 Wallot & Leonardi (2018), 4 Cox et al. (2016). CRQA analyses were performed using Marwan’s CRP
toolbox (2024, available at https://tocsy.pik-potsdam.de/CRPtoolbox) on MATLAB (2022).
17
2.5.4. Models
First, we performed maximum likelihood mixed-effects models with two levels to test our research
questions regarding interpersonal synchronization of speech, task effects, and personality traits. Level 1 was the task
or type of conversation (3 observations), nested within the dyadic structure (Level 2). We employed the “lme4” R
package, and the degrees of freedom and significance were calculated using the Satterthwaite method, which is
suitable for small sample sizes and complex model structures (Bates et al., 2015). First, we performed models
predicting the effect of the task on each speech synchronization and nonverbal interactional dominance variable. The
predictor variable was the task (conversational topic), a categorical variable with three levels (introduction/self-
disclosure/argumentative), where the introduction was the baseline and the response variables were RRglobal, RRLOS,
QDCRP, LAMARD, and TTARD, respectively. Next, we modeled the effects of the personality traits Extraversion and
Agreeableness and tasks on speech synchronization, leader-follower dynamics, and nonverbal interactional
dominance. The response variables were RRglobal, RRLOS, QDCRP , LAMARD, and TTARD, the predictors were the
personality traits of each interacting partner and task.3 We report both estimates and standardized beta weights (𝛽)
which can be interpreted as effect sizes (e.g., Paxton & Dale, 2013). For linear mixed effects, all continuous
predictors were standardized before being incorporated into the models to obtain beta weights. Particularly,
personality variables were centered by subtracting the mean and scaled by the standard deviation (R core team,
2022).
Subsequently, to explore the effects of speech synchronization and nonverbal interactional dominance on
the perceptions of the interactions (appraisals assessment), generalized linear models were conducted. We selected
as predictors one variable of speech synchronization (RRLOS) as this is the basic measure that provides information
on speech synchronization simultaneously (lag-zero); a variable informing about leader-follower dynamics from the
DCRP analysis (QDCRP), and one about nonverbal interactional dominance extracted from aCRQA (LAMARD), which
provides information about the extent of asymmetries in nonverbal interactional dominance. We included the items
of the perception of interaction questionnaire as the response variable (see Table 2), while the speech
3 The models followed the structure: [RRLOS ~ (Extraversion A * Extraversion B) * Task + (1|Dyad)]. In this
example, Extraversion A and B correspond to the scores of each interacting partner. The same procedure was
employed for the other response variables and Agreeableness.
18
synchronization/interactional dominance variables and personality traits were the predictors.4 We employed the
Benjamini-Hochberg (BH, 1995) method to correct the p-values for multiple hypothesis testing.
Table 2.
First-Person Perception of the Interaction Variables (Cuperman & Ickes, 2009)
Variable
1. How much did you feel a need to communicate with the other person?
2. How much did you use the other person’s behavior as a guide for your own behavior?
3. To what degree did you attempt to take the lead in the conversation?
4. How self-conscious did you feel when you were with the other person?
5. To what degree did the interaction seem awkward, forced, and strained to you?
6. To what degree did the interaction seem smooth, natural, and relaxed to you?
7. How involving (engaging) did you find the interaction?
8. To what extent did you feel put down, patronized, or rejected by the other person?
9. To what extent did you feel accepted and respected by the other person?
10. To what extent would you like to interact more with the other person in the future?
11. How much did you enjoy your interaction with the other person?
12. To what extent did you try to accommodate to the other person by adapting your behavior to “fit in” with his/hers?
13. How comfortable did you feel around the other person?
14. How much did you like the other person?
15. How empathic and understanding was the other person?
Note: Items were Likert-style from 1 to 5 (1=not at all, 2= a little bit, 3=moderately, 4=very much, 5=extremely).
4 The models followed the structure (example): Perception of Interaction Variable 1 ~ RRLOS * (Extraversion A *
Extraversion B).
19
3. Results
Descriptive statistics for all variables are provided in Table 3. Average speech time in seconds was the
longest when participants introduced themselves (mean = 467.14 seconds, SD = 78.53), followed by self-disclosure
(mean = 456.4, SD = 84.04), and shortest during the argumentative conversation (mean = 445.92 seconds, SD =
86.94).
Using mixed-effects models we assessed whether dyadic speech synchronization, leader-follower
dynamics, and nonverbal interactional dominance differed across the three conversation phases: Introduction, Self-
Disclosure, and Argumentation. The Introduction task served as the baseline and we examined response variables
including the Recurrence Rate Global (RRglobal), Percentage of Speech across the Line of Synchrony (lag-zero)
(RRLOS), Quotient Diagonal Cross-Recurrence Profile (absolute) (QDCRP), Relative difference of anisotropic
Laminarity (LAMARD), and Relative difference of anisotropic Trapping Time (TTARD; descriptives are provided in
Table 3). Cross-recurrence plots of dyads during the three tasks are illustrated with one example in Figure 3. Mixed
effect model estimates are provided in Figure 4 and Table 4.
3.1. Interpersonal speech dynamics of speech by conversation topic
Our global measure of speech synchronization (RRglobal) differed between conversational topics (see Figure
4A), being lowest during the introduction and highest during the argumentative conversation (argumentative >
introduction, t(100)= 2.86, p< .01), which aligns with H1a, and indicates higher synchronization of turn-taking
behaviors across all lags. Similarly, imbalances in leading-follower dynamics (QDCRP) were higher during the
argumentative conversation than during introductions (argumentative > introduction, t(100)= 2.92, p< .01), in line
with H1b (Figure 4C). Hence, we observed imbalances between the interaction partners' initiative in the turn-taking
structure: one of the participants initiated or led more during dyadic conversations than the other who followed more
(see Table 4).
During self-disclosure conversations the dyad showed greater asymmetries in the average duration of
nonverbal interactional dominance episodes (or “trapping” episodes) than during introductions (TTARD self-
disclosure > introduction, t(100)= 2.43, p <.05). Similarly, during the self-disclosure and argumentative conversations
dyads showed longer “trapping” episodes (higher TTARD, t(100)= 2.14, p <.05), in line with H2c (see Figure 4E).
Specific task effects were not significant for RRLOS and LAMARD, suggesting that speech synchronization at lag-zero
20
and the overall asymmetries of nonverbal interactional dominance did not significantly differ across tasks in our
sample (see Figures 4B and 4D respectively).
Our results suggest that different conversation types like introduction, self-disclosure, and arguments can
have distinct effects on speech synchronization (through turn-taking behaviors), leader-follower dynamics, and the
duration of nonverbal interactional dominance. Argumentative conversations were characterized by more speech
synchronization and greater imbalances in leader-follower dynamics. During self-disclosure and argumentative
conversations, asymmetries in nonverbal interactional dominance lasted longer than during introductory
conversations.
Figure 3.
Cross-recurrence plots
Note: Cross-recurrence plots depict interaction dynamics in three tasks: A. Introduction, B. Self-disclosure, C.
Argument. The horizontal and vertical axes, “Time Series 1” and “Time Series 2” respectively, represent the time
series of both interacting partners. Vertical lines represent temporal influence from one partner to the other, while
horizontal lines signify reciprocal influence. Dark lines indicate matching behavior (speaking/silent); and white
spaces indicate non-matching behaviors (e.g., simultaneous talking or silence). In the introduction (A), scattered
patterns suggest exploratory interaction, with instances of one participant leading. Self-disclosure (B) shows
pronounced matching blocks, indicating one participant's stronger influence. In the argumentative task (C),
behaviors are evenly distributed, reflecting mutual temporal influence and response between participants.
21
Figure 4.
Estimated Marginal Means of topic predicting each CRQA, DCRP, and aCRQA measure
Note: * p<.05, ** p<.01. CRQA= Cross-Recurrence Quantification Analysis, DCRP= Diagonal Cross Recurrence
Profile. aCRQA= anisotropic Cross-Recurrence Quantification Analysis. RRglobal= Global recurrence rate (speech
synchronization at all-lags). RRLOS= Recurrence rate across the line of synchrony (lag-zero). QDCRP= Quotient of
Diagonal Cross Recurrence Profile (balance in leader-follower dynamics). LAMARD= Laminarity absolute relative
difference (asymmetries in nonverbal interactional dominance). TTARD= Trapping Time absolute relative difference
(duration of episodes of nonverbal interactional dominance). Conversation topic 1 (Introduction) was the baseline.
3.2. Speech Synchronization and Personality
To estimate how personality differences predicted speech dynamics we fit mixed effects models that
revealed how more Extraversion predicted global speech synchronization (RRglobal) across conversational topics. At
least one extravert in a dyad predicted lower global speech synchronization during the argumentative conversation
than during the introduction (𝛽= -.22, t(100)= -2.04, p <.05; see Table 5, Model 1). Extraverts often had higher speech
synchronization levels (RRglobal, see Table 5), which would be indicative of attunement in conversations and turn-
taking dynamics across all lags, and all conversational topics (aligned to H2a). Introverted participants showed the
most pronounced differences between conversational topics, as they had the lowest global speech synchronization
(RRglobal) during the introduction, and the highest synchronization during the arguments, see Figure 5A. Dyads with
extraverted participants synchronized smoothly in all conversation topics (high RRglobal).
22
When we focus on Agreeableness we see that high scores of at least one partner predicted reduced global
speech synchronization during the argumentative conversation compared to the introduction (RRglobal, see Table 6,
Model 1; 𝛽= -0.26, t(100)= -2.37, p <.05). Hence, higher Agreeableness predicted decreased global speech
synchronization (i.e. RRglobal values), which argued against H2c, but it may also indicate more silences or
simultaneous talking (overlap). Low Agreeableness, in contrast, predicted the highest dyadic global speech
synchronization during argumentative conversations (see Figure 5B), which supported H2e. Synchronization of
speech through the line of synchrony –at lag-zero or simultaneous (RRLOS), indicated that only Agreeableness
predicted differences during the argumentative task compared to the introduction (𝛽= -0.18, t(100)=-2.01, p <.05).
Here, lower Agreeableness scores predicted strong synchronization (RRLOS, see Table 6, Model 2), in support of
H2e, which suggests increased turn-taking dynamics, attuned and swift conversational exchanges without delay (at
lag-zero).
In summary, Extraversion significantly predicted differences in global speech synchronization across
conversational topics. Extroverted individuals often showed more speech synchronization, suggesting greater
engagement in turn-taking dynamics across various conversation topics. The differences in global speech
synchronization between conversation topics are most pronounced in low extroverted individuals, with
synchronization being lowest during introductions and highest during argumentative conversations. Agreeableness
also influenced global speech synchronization, particularly during argumentative conversations compared to
introductions. Higher Agreeableness scores predict lower global speech synchronization values, suggesting more
silences or overlapping speech, while lower scores predict higher synchronization during argumentative
conversations. Lower Agreeableness scores predict higher scores on the line of synchrony, indicating increased turn-
taking dynamics and more attuned conversational exchanges without delay. Overall, Extraversion and
Agreeableness had varying effects on speech synchronization across different conversational topics. While high
Extraversion consistently predicted higher levels of speech synchronization across topics, high Agreeableness
showed nuanced effects, with its influence depending on the specific conversational context.
3.3. Leader-Follower Dynamics and Nonverbal Interactional Dominance
When exploring the balance of leader-follower dynamics the models based on the Quotient of Diagonal
Cross Recurrence Profiles (QDCRP) exhibited a significant additive effect during the argumentative task in the models
of Extraversion (𝛽= 0.54, t(100)= 3.02, p <.01, see Table 5, Model 3) and Agreeableness (𝛽= 0.50, t(100)= 2.81, p <.01,
23
see Table 6, Model 3). During the argumentative conversation, the leader-follower imbalances strengthened (more
QDCRP), thus one of the interacting partners typically took the initiative, e.g., spoke first while the dyadic partner
followed those rhythms. There were no significant effects of Extraversion linked to leader-follower dynamics (H3a
not supported). Higher Agreeableness scores predicted more balanced interactions (lower QDCRP) during the self-
disclosure task than when introducing oneself (𝛽= 0.56, t(100)= 3.06, p<.01). Higher scores on Agreeableness
predicted larger leader-follower imbalances (high QDCRP) when self-disclosing, and more balanced conversations
during the introduction (see Figure 5D). This indicates that low Agreeableness was associated with higher
imbalances in leading-following dynamics, while higher scores on Agreeableness were associated with more
balance when introducing themselves (supporting H3b). However, during the self-disclosure task, higher
Agreeableness was associated with higher leader-follower imbalances (higher QDCRP), with more pronounced
initiating behaviors, suggesting a task-sensitive effect. This can also indicate that one person allows (leads) the other
to talk or be silent, by initiating those behaviors.
To further visualize the recurrence rates across different lags and leader-follower dynamics, Diagonal
Cross-Recurrence Profiles (DCRPs) were plotted across the conversation topics (Figure 6A), which showed
(descriptively) the highest recurrence percentage at lag-zero (line of synchrony) during the argumentative
conversation, and the lowest recurrence rates during the self-disclosure task. This could indicate that during
arguments, there would be a similarity of behavior, a strong and immediate interaction or response between the
participants. It is also possible to see that during the introduction, the immediate effect, as captured by lag-zero, does
not prominently show the asymmetries in the interaction dynamics. However, as the lags increase, asymmetries
become more apparent. This could indicate that the interaction dynamics evolve over time, and asymmetries become
more pronounced with increases in time lags.
When plotting the DCRPs considering personality traits (Figure 6B), a segmentation approach was
employed, categorizing dyads into low and high personality trait scores using a threshold of -/+ 0.5 standard
deviations (SD, see Method section). The profiles indicated that the highest RRLOS (i.e., on the line of synchrony or
at lag-zero) was observed in the dyads composed of low/high Agreeableness, which suggests a high degree of
immediate synchronization of speech. Furthermore, as time lags increased, there was a subtle trend toward leading-
follower dynamics, indicating that low/high Agreeableness dyads seemed to evolve with one individual taking the
lead over time. This pattern was also observed in low and mixed agreeable dyads, however it was rather subtle. The
24
lowest RRLOS was observed in introverted dyads, suggesting a lower level of immediate speech synchronization in
interactions between introverted individuals. This segmentation and visualization has only descriptive purposes.
When modeling the effects of personality traits on the relative difference of Laminarity (LAMARD) the
Extraversion scores of both conversational partners significantly predicted LAMARD during the self-disclosure task
(model 4, Table 5; 𝛽= -0.42, t(100)= -2.32, p <.05), such that similarities in Extraversion (both for low and high
Extraversion) were associated with lower LAMARD (Figure 7A). This suggests that the magnitude of the speech
asymmetry in nonverbal interactional dominance decreased with personality similarity (in Extraversion) when self-
disclosing, indicating symmetrical conversational dynamics or lower nonverbal interactional dominance. The
asymmetries were higher in dissimilar dyads composed of introverted or extroverted individuals, in which case, one
of the interacting partners tended to be more nonverbal interactionally dominant (in alignment with H3a). On the
other hand, the Agreeableness level of both conversational partners predicted differences in LAMARD during the
argumentative task (𝛽= 0.36, t(100)= 2.74, p <.01; see Figure 7B). In this case, differences in the Agreeableness of
both conversational partners (low/high) predicted lower asymmetries in nonverbal interactional dominance, and
similarity in the dyadic composition, especially in high-agreeable individuals predicted higher asymmetries, which
indicates higher nonverbal interactional dominance or greater influence of one interacting partner’s behavior on the
other (in opposition to H3b).
Finally, regarding Trapping Time TTaRDif, no significant effects were linked to personality traits. Only the
type of conversation, self-disclosure, and argumentative significantly explained differences in TTaRDif in both
models of Extraversion (Table 5, Model 6) and Agreeableness (Table 6, Model 6). In both cases, the average
duration of the asymmetries was expected to be higher and last longer during the self-disclosure and argumentative
conversations than the introduction.
25
Figure 5.
Effects of Extraversion on RRglobal and Effects of Agreeableness on RRglobal, RRLOS, and QDCRP
Note: The plots represent the significant effects of the models of Extraversion and Agreeableness on the variables of
speech synchronization (RRglobal, RRLOS) and nonverbal interactional dominance (QDCRP). The significant effects are,
Panel A: [T3. Argumentative (𝛽=.32, t(100)=2.98, p<.01)], and [ExtraversionA*T3.Argumentative (𝛽=-.22, t(100)=-
2.04, p<.05)]; Panel B: [T3.Argumentative (𝛽=.34, t(100)=3.23, p<.01)], and [AgreeablenessB*T3. Argumentative
(𝛽=-.26, t(100)=-2.37, p<.05)]; in Panel C, [AgreeablenessB*T3.Argumentative (𝛽=-.18, t(100)=-2.01, p<.05)]; in Panel
B, [T3.Argumentative (𝛽=.50, t(100)=2.81, p<.01)], and [AgreeablenessB*T2.Self-disclosure (𝛽= 0.56, t(100)=3.06,
p<.01)].
26
Figure 6.
DCRPs
Note: The plots represent the DCRPs by task (Panel A) and by personality combination (Panel B). The zero on the
“x” axis (dashed line) represents the line of synchrony (LOS) corresponding to lag-zero. The “y” axis indicates the
RR (percentage of recurrence rate), representing speech synchrony (or coupling) between interacting partners during
the complete 15-minute conversation. Lags to the sides of the line of synchrony line suggest that one behavior (i.e.,
speaking) is leading, and the other behavior (i.e., listening) is following after a certain time delay. This indicates a
temporal pattern where one person initiates a turn, and the other responds after a specific duration (in seconds). In
descriptive terms, in Panel A, task 3 exhibits the highest RR, and task 2, the lowest. In Panel B, the dyads composed
of a person high and low in Agreeableness represent the highest RR and therefore, the strongest coupling; while the
lowest RR is visualized in the dyads composed of two individuals with low scores on Extraversion. In panel B
regarding dyadic personality combinations: “E++” and “A++”= high/high scores; “E--” and “A--"= low/low scores;
“E+-” and “A+-”=low/high scores.
27
Figure 7.
Effects of Extraversion and Agreeableness on LAMARD
Note: The plots show the effects of Extraversion (Panel A) and Agreeableness (Panel B) on LAMARD. Panels A and B
correspond to separate models (see Tables 5 and 6). In panel A, the effect of ExtraversionA*ExtraversionB * (2)Self-
Disclosure is statistically significant (
𝛽
=-.42, t(100)=-2.32, p<.05). In panel B, the significant effects are
AgreeablenessA* (3)Argumentative (
𝛽
=.42, t(100)=2.44, p<.05), and AgreeablenessA*
AgreeablenessB*(3)Argumentative (
𝛽
=.36, t(100)=2.74, p<.01).
28
In summary, as the Quotient of the Diagonal Cross Recurrence Profile (QDCRP) indicates, the argumentative
task significantly affects the balance of interactions. During the argumentative conversations higher imbalances in
leader-follower dynamics were observed, with one partner tending to speak first while the other follows.
Conversely, the self-disclosure task predicts more balanced interactions than introductions, particularly for
individuals with higher Agreeableness scores. However, higher Agreeableness is associated with higher imbalances
in leading-following dynamics during self-disclosure, suggesting a task-sensitive effect. Extraversion and
Agreeableness exhibit differential effects on interaction dynamics across tasks. Similarities in Extraversion between
conversational partners are associated with lower speech asymmetry during self-disclosure, indicating symmetrical
conversational dynamics or lower nonverbal interactional dominance. Conversely, dissimilar dyads, especially
composed of extroverted and introverted individuals, show higher asymmetries. Agreeableness influences
asymmetries in nonverbal interactional dominance during argumentative tasks, with similarities in Agreeableness
predicting lower asymmetries, while high Agreeableness predicts higher asymmetries or greater nonverbal
interactional dominance. The DCRPs (Figure 6) visually depict the interaction dynamics, showing the highest
recurrence percentage at lag-zero during argumentative conversations and the lowest during self-disclosure tasks.
Segmentation based on personality traits reveals immediate speech synchronization in low/high Agreeableness
dyads and subtle leading-follower dynamics over time. Regarding the average duration of nonverbal interactional
dominance, personality traits do not significantly affect the duration of the influence of one system over the other
(TTARD ), but the type of conversation significantly influences it, with self-disclosure and argumentative
conversations leading to higher and longer-lasting asymmetries in nonverbal interactional dominance compared to
introductions. The findings suggest that conversational topics and personality traits, particularly Extraversion and
Agreeableness, play significant roles in shaping interaction dynamics.
3.4. Perception of the interactions (appraisals)
Finally, we modeled the effects of personality differences (Extraversion/Agreeableness separately) of both
conversational partners and our variables of speech synchronization (RRLOS, QDCRP, and LAMARD, in separate models)
on each of the appraisals reported by the participants after their conversations (see method section for more details
about each model). Table 2 contains the variables we used to assess the perception of the interactions i.e. appraisals.
We corrected the p-values for multiple-hypothesis testing (using the Benjamini-Hochberg procedure, 1995). Tables
with all models and figures with significant effects can be consulted in the supplementary materials.
29
Inclination for communication (need to communicate with partner)
Extraversion scores without interacting with other variables predicted a higher need to communicate with
the interacting partner (𝛽= 0.10, p<.05). Higher Extraversion levels of both interacting partners and more speech
synchronization (RRLOS) were predictive of increased perceived “need to communicate” (𝛽= 0.11, p<.01), indicated
by increased synchronization of speech across the line of synchrony (model 1, Table S1, Figure S1.A). In this
context, speech synchronization (RRLOS) suggests a greater reciprocity and attunement in terms of speech-silence
matching, more time spent talking, fewer pauses, silences, and interruptions. Introverted dyads predicted the lowest
inclination for communication overall. And, in the case of mixed extroverted dyads (low/high Extraversion), when
speech synchronization (RRLOS) decreased, the need for communication was observed to increase. In the case of
Agreeableness, the interaction effect between the scores on Agreeableness of both conversation partners predicted
decreases in the need for communication for low agreeable individuals in the model of LAMARD (𝛽= -0.08, p<.05)
(model 1, Table S5, Figure S1.B); whereas the presence of at least one agreeable individual in the dyad (mixed
dyads), predicted increases in the inclination for communication. There were no specific significant effects of speech
variables and Agreeableness.
Using partner’s behavior as a guide for own behavior
Asymmetries in nonverbal interactional dominance could be observed in the model of relative differences
in Laminarity (LAMARD), where lower Extraversion scores associated with more perceived behavioral adjust to the
partner cues (compared to extroverts, 𝛽= -0.14, p<.01) (model 2, Table S2, Figure S2.A). Extraversion scores (of
one interacting partner) and LAMARD showed an interaction effect, as lower scores on Extraversion associated with
higher asymmetries (higher LAMARD) and predicted more perceived behavioral use of the partner cues (𝛽= 0.27,
p<.05). There was also a three-way interaction where Extraversion scores of both interacting partners (𝛽= 0.23, p
<.05) predicted lower symmetry (lower LAMARD) in mixed dyads. The perceived alignment to the partner cues was
predicted to increase, especially among introverts, whereas during high asymmetry (high LAMARD), participants
tended to align their behavior to the partners’ behavior in mixed dyads (low/high) and extroverted dyads, which
suggests that a more pronounced nonverbal interactional dominance in such dyads was reflected in the report of one
of them using the partners’ behavior as a guide of the own behavior.
In the model of Agreeableness with Laminarity (LAMARD) as predictor, the interactive three-way effect
(𝛽=0.27, p<.05) suggested that during lower nonverbal interactional dominance (lower LAMARD) highly agreeable
30
dyads were more likely to use the partners’ behavior as a guide for the own behavior (model 2, Table S5, Figure
S2.B). In mixed agreeable dyads, however, the interactive effects suggest that agreeable individuals exhibited
initiating behaviors, whereas disagreeables tended to align their behavior to the partner cues when LAMARD was
higher (nonverbal interactional dominance asymmetries). These results suggest that highly agreeable dyads tend to
use their partners’ behavior and that is linked to symmetrical interactions (lower LAMARD). On the other hand,
agreeable individuals may tend to take the initiative when interacting with disagreeable partners, while disagreeable
individuals align their behavior with partner cues when LAMARD is higher (nonverbal interactional dominance
asymmetries).
Attempts to lead the conversation
The effect of Agreeableness of at least one conversational partner and the (im)balances in the conversation
(QDCRP) significantly predicted the perceived attempt to lead the conversation. Increases in imbalances in leading-
following dynamics (QDCRP) predicted increases in the perceived attempt to lead the conversation by low agreeable
individuals (𝛽=0.17, p<.05). There was a three-way interaction between Agreeableness scores of both
conversational partners and the balance of the interaction (QDCRP), where in dissimilar dyads (low/high), higher
imbalances (QDCRP) predicted increases in the perceived attempt to lead the conversation by low agreeable
individuals (𝛽= -0.32, p<.05) (model 3, Table S6, Figure S3).
“Smooth, natural, and relaxed” conversations
Agreeableness positively predicted the report of smooth, natural, and relaxed conversations when
considering the personality trait without interacting with other variables (p<.05) (model 6, Table S4, Figure S4.A).
However, the interaction of Agreeableness and speech synchronization (RRLOS) was negatively related to the
perception of the conversation as smooth, natural, and relaxed (𝛽= -0.35, p<.05). This effect suggests that as the
interaction between Agreeableness and speech synchronization was higher –thus more attunement of speech-silence
turns, more time spent talking, fewer pauses, silences, and interruptions– there was a corresponding decline in the
perception of the conversation as smooth, natural, and relaxed. This may indicate a trade-off effect. In other words,
while Agreeableness and speech synchronization (at lag-zero) might enhance certain aspects of a conversation (e.g.,
interpersonal attunement), they could also make it feel less relaxed and natural, which will be addressed in more
detail in the discussion section.
31
Felt accepted and respected by partner
Overall, a main effect indicated that higher Extraversion scores were associated with increases in the
perception of being accepted and respected by the interacting partner (𝛽= 0.25, p<.01); whereas lower scores on
Extraversion predicted decreases in this perception of being accepted/respected (model 9, Table S2). No effects
were found for Agreeableness and speech synchronization variables. This could imply that these factors might not
directly influence the feeling of acceptance and respect in a conversation, or their effects might be more subtle or
complex.
Desire to interact more with partner in the future
Only Extraversion and speech synchronization (RRLOS) such as speech-silence attunement, reciprocity, and
fewer silences and interruptions, increased participants’ willingness for future interactions (model 10, Table S1,
Figure S4.B). High Extraversion scores of at least one interacting partner and increases in speech synchronization
predicted a higher post-conversational desire to interact in the future (𝛽= 0.46, p<.01). This implies that both
personality traits and the dynamics of the conversation itself can influence the desire for future interactions.
Enjoyment of the interaction
High Extraversion (of at least one interacting partner) and higher speech synchronization (RRLOS) were
associated with an increased enjoyment of the interactions (𝛽= 0.28, p <.05) (model 11, Table S1). Conversely, for
introverted individuals, higher speech synchronization predicted decreases in enjoyment. The effect of both
conversational partners was significant as well (three-way interaction), but highly extroverted dyads enjoyed
conversations with a higher speech synchronization more (𝛽=0.42, p <.05) (Figure S5.A). In more dissimilar dyads
(extroverted/introverted) lower speech synchronization was associated with increased enjoyment. In the case of the
Extraversion of both interacting partners and the asymmetries in nonverbal interactional dominance (LAMARD, the
three-way interaction), lower asymmetries and high scores on Extraversion of both interacting partners predicted
increased enjoyment (𝛽= -0.25, p <.05) (model 11, Table S2, Figure S5.B). When the asymmetries (LAMARD)
increased, the enjoyment was predicted to increase as well for mixed dyads (introverted/extroverted). Furthermore,
introverted dyads were associated with decreased enjoyment during the conversation (compared to the other
participants), independent of other interpersonal speech dynamics in the conversation. In the case of balance in
leader-follower dynamics (QDCRP), a three-way interaction indicated that increased imbalances –an interacting
partner tended to temporally initiate/lead or act first, either speaking or being silent– were predictive of enjoyment in
32
extroverted dyads (𝛽=0.44, p <.05); whereas in mixed dyads (introverted/extroverted), more balanced interactions
(QDCRP) were predictive of increases in enjoyment (model 11, Table S3, Figure S5.C).
Perceived partner as likable
The main effect of Extraversion indicated that increases in this trait were positively associated with
increases in the report of liking the conversational partner (𝛽= 0.14, p<.01) (model 14, Table S1, Figure S6.A).
Similarly, the interactive effect of Extraversion scores and increases in speech synchronization (RRLOS) predicted an
increased report of liking the conversational partner (𝛽= 0.49, p<.05). On the other hand, decreased scores on
Extraversion and lower values of speech synchronization predicted increases in liking the other person. Similarly,
the three-way effect between the Extraversion scores of both partners and speech synchronization suggested that in
extraverted dyads, increases in speech synchronization were associated with liking the interacting partner to a
greater extent; whereas in mixed dyads (introverted/extroverted), decreases in speech synchronization –which
indicates the presence of silences as well– were associated with liking the other person (𝛽=0.21, p <.05) (model 14,
Table S1, Figure S6.B). These effects may indicate that introverted and extroverted individuals valued different
aspects of the conversation. Extraverts seemed to appreciate a more dynamically synchronized conversation, while
introverts may find value in moments of silence or pauses during interactions. The distinction in preferences
between introverted and extroverted individuals is particularly salient in mixed dyads.
Perceived partner as empathic and understanding
Finally, increased Extraversion scores were associated with increases in the perception of the
conversational partner as empathic and understanding (main effect, 𝛽=0.02, p <.05). And, the interaction between
Extraversion and the nonverbal interactional dominance or asymmetries in the interaction (LAMARD), indicated that
lower asymmetries and higher Extraversion scores predicted increases in the perceived empathy and understanding;
whereas lower scores in Extraversion and increases in the asymmetries, were predictive of increased perceived
empathy and understanding (𝛽=-0.31, p <.05) (model 15, Table S2, Figure S7). This can reflect differences in the
appraisals of interpersonal dynamics of speech in the conversation by personality differences as will be discussed.
For the rest of the appraisal variables, no significant effects linked to personality traits were found after correcting
for multiple hypotheses testing, but all the results can be found in the supplementary materials.
Overall, these findings highlight the relationship between personality traits, speech synchronization, and the
perception of interaction dynamics. They underscore the importance of considering individual differences in
33
communication styles and patterns when assessing the quality of social interactions.
4. Discussion
This study was conducted to achieve four goals. First, exploring the effect of high-level constraints (here:
conversational topics) on interpersonal speech synchronization, leading-following dynamics, and nonverbal
interactional dominance in dyadic conversations (1). In the next two goals we aimed to explore how the
interpersonal speech synchronization structures were related to the socially relevant personality traits of
Extraversion and Agreeableness of the interacting partners. In particular, exploring how these personality traits were
related to synchronization of speech (2), as well as leader-follower dynamics and nonverbal interactional dominance
(3). Lastly, we aimed to explore the effect of synchronization of speech, leader-follower dynamics, nonverbal
interactional dominance, and personality traits in the appraisals of the interactions reported by the conversational
partners (4). Following these goals, our key findings are discussed below, considering theoretical implications,
limitations, and future directions.
4.1. Interpersonal speech synchronization and conversation topic
We expected that different conversational topics explained differences in our variables of speech
synchronization (RRglobal, RRLOS) (H1a), leader-follower dynamics (QDCRP,) and nonverbal interactional dominance
(LAMARD, and TTARD) (H1b). We anticipated differences in speech dynamics between the self-disclosure and
argumentative conversations compared to the introduction. Conversational topics indeed predicted differences in
global speech synchronization (RRglobal), but only in terms of more synchronization during the argumentative
conversation versus the introduction part. Regarding leader-follower dynamics and nonverbal interactional
dominance, our hypothesis (H1b) about the role of conversational topics was supported in the self-disclosure
(TTARD) and argumentative conversations (QDCRP, TTARD). Differences in high-level conversational constraints i.e. the
topics were aligned with our expectations and the literature on interpersonal synchronization. Consistently, studies
indicate that dynamic properties of interpersonal interactions (interpersonal speech synchronization) are shaped by
situational constraints, or in other words, are soft-assembled (Fusaroli et al., 2014), and depend on the context where
the encounter unfolds. This interplay aligns with the perspectives of language and joint action as complex adaptive
systems (e.g., Ellis & Larsen-Freeman, 2009; Paxton & Dale, 2017; Tschacher et al., 2018; Arellano-Véliz et al.,
34
2024). Similarly, situational factors are also relevant in individual settings, affecting movement and self-organizing
dynamics (Arellano-Véliz et al., 2023).
The role of the high-level constraints (e.g., conversation topics) in speech synchronization can vary as it
flexibly adjusts to casual encounters, bonding/affiliating, or competitive goals (Paxton & Dale, 2017). Previous
research showed that patterns of nonverbal interactional dominance and balance can affect the quality of the
interactions and interacting partners tend to be sensitive to distinct cues in the conversation, which can be functional
to interpersonal goals (Reuzel et al., 2014). In the case of this study, the effect of the argumentative conversation on
global synchronization of speech was significantly higher compared to the introductory conversation, and the same
was true for leader-follower dynamics and nonverbal interactional dominance, where the asymmetries in the
conversation were larger during this conversation. Increased interpersonal synchronization during specific
conversation topics has been observed in competitive settings previously (e.g., Tschacher et al., 2018; Arellano-
Véliz et al., 2024). Other studies reported that in-phase (simultaneous) bodily synchronization decreased during
arguments (Paxton & Dale, 2013). However, as mentioned in the introduction, the concepts of functional
interpersonal synchronization and interpersonal speech synchronization involve compensatory dynamics (that do not
necessarily unfold simultaneously) and support the emergence of functions or the achievement of goals (Nowak et
al., 2017).
Likewise, we operationalized speech synchronization as the reciprocity of the interacting partners'
nonverbal behaviors, focussing on turn-taking coordination and a conversational rhythm instead of a simultaneous
performance (Reuzel et al., 2013). Reciprocity is key in argumentative conversations, where exchanges in
interpersonal dynamics will facilitate the emergence of a dynamic and swift interplay of bidirectional exchange of
arguments. Furthermore, our observation that the relative difference in trapping time (TTARD) was larger during the
self-disclosing and argumentative conversations, can indicate sustained influence by one interacting partner for an
extended period. This nonverbal interactional dominance can mean that one partner’s speech or silence episodes
tended to “trap” the other partner in longer episodes of speech or silence, and this can be indicative of providing the
other partner with the opportunity –or the affordance– to display such behavior by acting first (Worgan & Moore,
2010). Even though it can indicate more “control” in the dynamics, this can also afford and facilitate a reciprocal
interaction when self-disclosing personal information. According to previous research, therapists use their nonverbal
interactional skills to intensify the attunement with clients by leading in the use of turn-taking and driving dynamical
35
synchronization of speech (Reuzel et al., 2014). This type of behavior might have been displayed by some
interacting partners to improve the communicational rhythm when self-disclosing, which might involve longer
interpersonal speech synchronization.
Conversational topics were unrelated to speech synchronization across the line of synchrony (RRLOS) and
one of the measures for nonverbal interactional dominance, Laminarity (LAMARD), which may be due to our modest
sample size, and indicates that these findings should be replicated in future studies. In general, the variable of global
synchronization of speech is more robust as it considers all possible lags in the conversation. It gives a general
overview of the interpersonal speech synchronization and patterns across the conversation (Reuzel et al., 2014). In
this sense, some responses and exchanges might be time-sensitive and be visible at longer lags rather than
simultaneously or very close in time, as indicated by speech synchronization across the line of synchrony (RRLOS),
which represents swift dynamics.
4.2. Personality traits, speech synchronization, and nonverbal interactional dominance
We argued that the personality traits of each conversational partner interact and promote the emergence of
interpersonal speech dynamics at the dyadic level. In particular, we expected that the dimensions of Extraversion
and Agreeableness –the key social traits– would explain part of the variability in speech synchronization (H2),
leader-follower dynamics, and nonverbal interactional dominance (H3), based on previous work on interpersonal
dynamics of body motion (Arellano-Véliz et al., 2024).
Extraversion scores were associated with increased speech synchronization (RRglobal) while introverts
synchronized less (H2a). Extraverted individuals showed more synchronized communication across multiple time
lags (not just at lag-zero as indicated by RRLOS), and across all conversational topics, which illustrates that
interpersonal speech dynamics in highly extroverted individuals were more context-independent. Extroverts are
defined as highly socially, gregarious, and outgoing individuals (Costa & McCrae, 1995), for whom social
encounters are rewarding in themselves, which promotes higher intersubjective attunement (Stern, 1985/2018;
Harris et al., 2017). Introverts exhibited the lowest synchronization of speech, especially during the introductory
conversation. Introverted individuals varied in synchronization across topics, as argumentative conversations
exhibited the highest degree of synchronized communication. These results align with the social reactivity
hypothesis that extraverts get more pleasure from social interactions, and, therefore, have more drive to engage in
them, compared to introverted individuals (Lucas & Diener, 2001). Generally speaking, introverts prefer solitude
36
and tend to be more comfortable with their inner worlds, thoughts, and feelings than extroverts (Burger, 1995;
Tuovinen et al., 2020).
Regarding the composition of our dyads, we expected that when both interacting partners were extroverted
or at least one of them was extroverted, this would lead to increased speech synchronization (H2b). In this case, the
presence of at least one extroverted individual in the dyads led to increased synchronization, which might be
relevant to facilitating social interactions for introverts, as it might boost their social engagement (Tuovinen et al.,
2020). The interactive effect of both conversational partners was not significant in the models of speech
synchronization (RRglobal, and RRLOS), only the individual effect of the trait Extraversion in global speech
synchronization (RRglobal). Therefore, the presence of at least one extravert played a more relevant role in the
interpersonal dynamics.
We expected high agreeable scores to associate with dyadic coupling and increased speech
synchronization, and inverse expectation for low-agreeable (or “disagreeable”) dyads (H2c). Speech synchronization
was predicted to be slightly lower for agreeable individuals. Dyadic composition was expected to affect speech
synchronization, especially the presence of a disagreeable individual in the dyad, but no support was observed
(H2d). However, the task sensitivity effect we predicted regarding the argumentative conversation was supported
(H2e) as low Agreeableness associated with higher speech synchronization during the argumentative conversation
(RRglobal and RRLOS). We argued that this effect could be functional to goal achieving in competitive settings by low-
agreeable individuals (e.g., DeYoung, 2015), and this could be observed in interpersonal speech dynamics by highly
responsive and swift interactions in the conversation. However, these dynamics might not be positive for the
intersubjective attunement of the interacting partners, since the literature suggests that low-agreeable individuals
exhibit poor social relationships (Anderson et al., 2020), low concern for others' needs and desires, and less efficient
social information processing (e.g., DeYoung, 2010).
Concerning leader-follower dynamics and nonverbal interactional dominance, we expected that higher
scores of Extraversion would be associated with asymmetries in the speech dynamics when interacting with
introverts, possibly because of the emergence of leading (initiating) and influencing tendencies (H3a). This was
observed only with asymmetries reflected by the relative difference of Laminarity (LAMARD) during the self-
disclosing conversations, where personality similarity fostered symmetric interpersonal speech dynamics, and
personality dissimilarity accentuated asymmetries between interacting partners. It is possible that extroverted
37
individuals took a leading and initiating role within the conversation when interacting with introverted partners
allowing for longer periods of interactional attunement.
We also expected more balanced leading-following dynamics for agreeable individuals, in contrast to
disagreeable individuals, where the latter might predict larger imbalances in the interactions (H3b). Results
supported this hypothesis when looking at the diagonal cross-recurrence profiles (through QDCRP) during the
introduction. We found that low-agreeable individuals exhibited a larger imbalance in leader-follower when
introducing themselves and lower when self-disclosing. Highly agreeable individuals predicted the opposite
patterns, with lower imbalance when introducing themselves and higher when self-disclosing, which indicates that
agreeables might have taken the lead during this conversation. When looking at the asymmetries in nonverbal
interactional dominance (LAMARD), the argumentative conversation elicited higher nonverbal interactional
dominance, especially in high-agreeable individuals and mixed agreeable dyads (agreeable/disagreeable). In this
sense, as mentioned before, Agreeableness is a social trait characterized by a tendency towards cooperation,
altruism, and aligning their needs with those of others (DeYoung, 2015; Hovhannisyan & Vervaeke, 2022). These
patterns of leading behavior and nonverbal interactional dominance suggest that when self-disclosing, highly
agreeable partners tend to take the initiative more, in a prosocial way. This may be especially visible in mixed
dyads, facilitating attunement to those patterns and possibly fostering a reciprocal and cooperative conversational
dynamic (Worgan & Moore, 2010). The same could be said for the argumentative conversation, where highly
agreeable individuals may have tended to “dominate” the conversation, promoting longer episodes of attunement.
Remember that dominance in this context should not be interpreted with the negative connotation it often has. The
observed tendency might be linked to the higher metacognitive capacity in agreeable individuals, described by a
better understanding of others’ needs, intentions, and desires (DeYoung, 2010).
4.3. Perception of the interaction (appraisals), interpersonal speech dynamics, and personality traits
Higher degrees of dyadic speech synchronization (H4a) and symmetrical and balanced interactions (H4b)
were predicted to increase positive post-conversational appraisal. Extroverted (H4c) and agreeable (H4d) individuals
were expected to appraise the interaction as more positive after more synchronized interactions. The results obtained
exhibited differentiated effects regarding the dyadic constitution and interpersonal speech dynamics.
First, we observed a logical connection between the interpersonal speech dynamics in the conversation
indicated by the variables extracted from the time series analyses and the appraisals reported by the participants. In
38
the case of Extraversion, this trait consistently predicted positive perceptions of the interaction as expected. When
exploring the role of speech synchronization in the conversation, we observed that high scores of Extraversion, and
increased synchronization of speech (RRLOS), were consistently related to positive appraisals such as perceptions of
the conversation as smooth/natural/relaxed, desire to interact in the future, enjoyment, liking the conversational
partner, and an inclination to communicate. These results underscore the relevance and value of attuned and swift
conversations for highly extroverted individuals. RRLOS reflects swift turn-taking in the conversation, that is, very
attuned conversations with few silence episodes (i.e. both not speaking) and interruptions (i.e. both speaking at the
same time). In this sense, as we observed a different effect for introverts, it is possible that they valued the presence
of silences and pauses more than ongoing attunement and swift dynamics. These results align with H4a and H4c.
On the other hand, increases in nonverbal interactional dominance (i.e. increases in LAMARD) predicted
perceptions such as using the interaction partner's behavior as a guide, especially in introverts, which reflects an
attempt to align to the partners’ cues. Higher asymmetries in nonverbal interaction dominance (LAMARD) led to
increased enjoyment when mixed extroverted dyads interacted. In this sense, it may be that nonverbal interactional
dominance could have promoted a better quality in the interactions in these dyads, presumably by affording and
facilitating opportunities for sustaining interactional attunement.
The effects of interpersonal speech dynamics in positive appraisals can indicate a higher intersubjective
attunement and relatedness (Stern, 1985/2018). Previous studies on undergraduate students showed that
Extraversion predicted 4 years later their subjective well-being (Harris et al., 2017). In particular, self-reported and
peer-reported positive social experiences such as feelings of belonging and larger size of social networks are highly
relevant for these young populations. It would be relevant to consider that introverted individuals can also be
rewarded by interpersonal interactions when they interact with individuals dissimilar to them in terms of personality.
In this sense, they may benefit from a certain degree of guidance/leading and initiating behaviors (someone who
creates social opportunities for them) when sustaining social interactions (Tuovinen et al., 2020). Furthermore,
introverts might value slower conversations, with more pauses and silences as their need for social stimulation is
lower than for extroverts (DeYoung, 2015; Hovhannisyan & Vervaeke, 2022).
On the other hand, we observed that in highly agreeable dyads (similar) agreeable individuals tended to use
the partner’s behavior as a guide, which was observed also in the form of lower asymmetries in nonverbal
interaction dominance (LAMARD). Agreeableness is thought to play a significant role in fostering interpersonal
39
attunement during the conversation (cf. Anderson et al., 2020). These adjustments and attunement to the partner’s
behavior can be indicative of high dyadic and nonverbal coupling as well. In addition, contrary to our expectations
(H4d), increased speech synchronization and Agreeableness were predictive of decreased perceived naturality in the
conversations. We could argue that there may be a trade-off effect to sustaining synchronized communication –in
the form of smooth turn-taking dynamics, fewer episodes of silence, and overall more time of speech– where highly
agreeable individuals achieved this communicational rhythm without feeling natural. It can be also possible that
agreeable individuals need more time to respond (more lags) or value episodes of silence as well. Similarly, previous
studies have indicated that to sustain interpersonal synchronization, there is a trade-off effect where the self-
regulation of affect decreases as the interacting partners adapt to each other (Galbusera et al., 2019). It would be
plausible to think that the drive of Agreeableness towards altruism, cooperation, and social harmony (DeYoung,
2015) could have played a role in a greater effort to sustain these smooth interpersonal speech dynamics without
experiencing the social encounter as inherently pleasant. Furthermore, we also observed a task sensitivity in both
Extraversion and Agreeableness, that underscores the situational context in shaping interpersonal communication.
5. Conclusion
Overall, our results emphasize the dynamic interplay between the personality traits Extraversion and
Agreeableness, situational constraints (conversation topics), interpersonal speech dynamics –synchronization,
leader-follower dynamics, and nonverbal interactional dominance–, as well as the subjective experiences emerging
from social interactions. Personality traits exhibited relevance in speech dynamics and appraisals, and differences in
terms of the dyads’ constitutions were observed. Generally, we observed that extroverted individuals engaged in
more synchronized communication across various conversational topics, contrary to introverts. Besides,
interpersonal speech synchronization seemed to foster intersubjective attunement and positive appraisals in
extroverts. Increased speech synchronization and Agreeableness were predictive of decreased perceived naturality in
the conversations, suggesting a potential trade-off effect. In terms of the methods employed, we were able to
observe how situational constraints and personality traits were predictive of interpersonal speech dynamics in the
conversations and appraisals. The nonlinear time-series techniques employed exhibited a useful and robust tool for
studying interpersonal dynamics in conversations. Our results support the use of dynamical approaches to, not only
understanding interpersonal communication, but also its relation to psychological constructs.
40
6. Limitations, strengths, and future directions
It is relevant to acknowledge the limitations of our study, in particular, the modest sample size, which
needs to be considered in the generalizability of our results. In this regard, some significant effects became non-
significant after multiple testing solutions, and therefore, were not discussed. Moreover, our sample predominantly
comprised women, thus limiting the ability to draw comparisons across genders. To address this limitation and
enhance the generalizability of our findings, future research should aim to utilize larger and more diverse samples.
This will allow for the replication and extension of our results across a broader demographic spectrum.
The nonlinear time series analysis methods allowed us to capture subtle and robust interpersonal dynamics
that might have not been observed otherwise, which we consider a strength in our experimental design. In this sense,
we recognize a promising toolbox to be incorporated to a greater extent in the study of interpersonal dynamics and
personality research.
Finally, we understand that we studied a dimension of speech that did not consider the context of the
conversations, the content of the utterances, and other personality traits. Instead, we focused on the interpersonal
dynamics extracted from a specific set of interpersonal speech dynamics –turn-taking behaviors– and the most
relevant “social” personality traits. In this sense, content is highly relevant to understanding these interpersonal
dynamics so other traits can be relevant. Therefore, future studies may need to consider the synchronization of
speech and interpersonal speech dynamics at the content level. This will expand our comprehension of interpersonal
dynamics, communication, and personality traits.
7. References
Abney, D. H., Paxton, A., Dale, R., & Kello, C. T. (2014). Complexity matching in dyadic conversation. Journal of
Experimental Psychology: General, 143(6), 2304–2315. https://doi.org/10.1037/xge0000021
Allsop, J. S., Vaitkus, T., Marie, D., & Miles, L. K. (2016). Coordination and collective performance: cooperative goals boos t
interpersonal synchrony and task outcomes. Frontiers in Psychology, 07. https://doi.org/10.3389/fpsyg.2016.01462
Anderson, C., Sharps, D. L., Soto, C. J., & John, O. P. (2020). People with disagreeable personalities (selfish, combative, and
manipulative) do not have an advantage in pursuing power at work. Proceedings of the National Academy of
Sciences, 117(37), 22780-22786. https://doi.org/10.1073/pnas.2005088117
Arellano-Véliz, N.A., Cox, R.F., Jeronimus, B.F., Castillo, R.D. & Kunnen, E.S. (2023). Personality Expression in Body
Motion Dynamics: An Enactive, Embodied and Complex Systems Perspective. Available at SSRN:
https://ssrn.com/abstract=4560996 or http://dx.doi.org/10.2139/ssrn.4560996
Arellano-Véliz, N. A., Jeronimus, B. F., Kunnen, E. S., & A. Cox, R. F. (2024). The interacting partner as the immediate
environment: Personality, interpersonal dynamics, and bodily synchronization. Journal of Personality, 92(1), 180-
201. https://doi.org/10.1111/jopy.12828
Aron, A., Melinat, E., Aron, E. N., Vallone, R. D., & Bator, R. J. (1997). The experimental generation of interpersonal
closeness: A procedure and some preliminary findings. Personality and Social Psychology Bulletin, 23(4), 363–377.
https://doi.org/10.1177/0146167297234003
41
Asendorpf, J. B. (2017). Personality as a Situation: A Target-Centered Perspective on Social Situations. In D. C. Funder, J. F.
Rauthmann & R. Sherman (Eds.), The Oxford Handbook of Psychological Situations.
https://doi.org/10.1093/oxfordhb/9780190263348.013.2
Bales, R.F. (1973). The equilibrium problem in small groups. In M. Argyle (Ed.), Social Encounter. Routledge.
Bates D, Mächler M, Bolker B, Walker S (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical
Software, 67(1), 1–48. doi:10.18637/jss.v067.i01.
Beckner, C., Blythe, R. A., Bybee, J., Christiansen, M. H., Croft, W., Ellis, N. C., Holland, J. H., Jin, K., Larsen‐Freeman, D.,
& Schoenemann, T. (2009). Language is a complex adaptive system: position paper. Language Learning, 59(s1), 1–
26. https://doi.org/10.1111/j.1467-9922.2009.00533.x
Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to
Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289–300.
http://www.jstor.org/stable/2346101
Bernieri, F. J., Reznick, J. S., and Rosenthal, R. (1988). Synchrony, pseudosynchrony, and dissynchrony: measuring the
entrainment process in mother-infant interactions. Journal of Personality and Social Psychology, 54, 243–253.
https://doi.org/10.1037/0022-3514.54.2.243
Berger, J., Cohen, B. P., & Zelditch, M. (1972). Status characteristics and social interaction. American Sociological Review,
37(3), 241–255. https://doi.org/10.2307/2093465
Bleidorn, W., Schwaba, T., Zheng, A., Hopwood, C. J., Sosa, S. S., Roberts, B. W., & Briley, D. A. (2022). Personality
stability and change: A meta-analysis of longitudinal studies. Psychological Bulletin, 148(7-8), 588–619.
https://doi.org/10.1037/bul0000365
Bloomfield, L., Lane, E., Mangalam, M., & Kelty-Stephen, D. G. (2021). Perceiving and remembering speech depend on
multifractal nonlinearity in movements producing and exploring speech. Journal of the Royal Society Interface,
18(181), 20210272. https://doi.org/10.1098/rsif.2021.0272
Boersma, Paul & Weenink, David (2023). Praat: doing phonetics by computer [Computer program]. Version 6.3.17,
http://www.praat.org/
Borgatta, E. F., & Bales, R. F. (1953). Interaction of individuals in reconstituted groups. Sociometry, 16, 302–320.
https://doi.org/10.2307/2785935
Burger, J. M. (1995). Individual Differences in Preference for Solitude. Journal of Research in Personality, 29(1), 85-108.
https://doi.org/10.1006/jrpe.1995.1005
Chemero, A. (2003). An outline of a theory of affordances. Ecological Psychology, 15(2), 181–195.
https://doi.org/10.1207/S15326969ECO1502_5
Costa, P. T., & McCrae, R. R. (1995). Domains and facets: Hierarchical personality assessment using the Revised NEO
Personality Inventory. Journal of Personality Assessment, 64(1), 21–50.
https://doi.org/10.1207/s15327752jpa6401_2
Cox, R. F. A., van der Steen, S., Guevara Guerrero, M., Hoekstra, L., & van Dijk, M. (2016). Chromatic and anisotropic
cross-recurrence quantification analysis of interpersonal behavior. In C. Webber, C. Ioana, & N. Marwan (Eds.),
Recurrence Plots and Their Quantifications: Expanding Horizons: Proceedings of the 6th International Symposium
on Recurrence Plots, Grenoble, France, 17-19 June 2015 (Vol. 180, pp. 209-225). (Springer Proceedings in Physics;
Vol. 180). Springer. https://doi.org/10.1007/978-3-319-29922-8
Cuperman, R., & Ickes, W. (2009). Big Five predictors of behavior and perceptions in initial dyadic interactions: Personality
similarity helps extraverts and introverts, but hurts “disagreeables”. Journal of Personality and Social Psychology,
97(4), 667–684. https://doi.org/10.1037/a0015741
Dale, R., Kirkham, N. Z., & Richardson, D. C. (2011). The Dynamics of Reference and Shared Visual Attention. Frontiers in
Psychology, 2, 15505. https://doi.org/10.3389/fpsyg.2011.00355
DeYoung, C. G., Hirsh, J. B., Shane, M. S., Papademetris, X., Rajeevan, N., & Gray, J. R. (2010). Testing Predictions From
Personality Neuroscience: Brain Structure and the Big Five. Psychological Science, 21(6), 820.
https://doi.org/10.1177/0956797610370159
DeYoung, C. G. (2015). Cybernetic Big Five Theory. Journal of Research in Personality, 56, 33–58.
https://doi.org/10.1016/j.jrp.2014.07.004
42
Delaherche, E., Chetouani, M., Mahdaoui, A., Saint-Georges, C., Viaux, S., & Cohem, D. (2012). Interpersonal synchrony: A
survey of evaluation methods across disciplines. IEEE Transactions on Affective Computing, 3(3), 349–365.
https://doi.org/10.1109/T-AFFC.2012.12
Di Paolo, E. A., Cuffari, E. C., & De Jaegher, H. (2018). Linguistic Bodies: The Continuity between Life and Language. MIT
Press.
Ellis, N. C., & Larsen‐Freeman, D. (2009). Language as a complex adaptive system. John Wiley & Sons.
Falandays, J. B., Batzloff, B. J., Spevack, S., & Spivey, M. J. (2018). Interactionism in language: from neural networks to
bodies to dyads. Language, Cognition and Neuroscience, 35(5), 543–558.
https://doi.org/10.1080/23273798.2018.1501501
Fişek, M., Berger, J., & Norman, R. Z. (2005). Status cues and the formation of expectations. Social Science Research, 34(1),
80-102. https://doi.org/10.1016/j.ssresearch.2003.10.004
Fowler, C. A., Richardson, M. J., Marsh, K., & Shockley, K. D. (2008). Language use, coordination and the emergence of
cooperative action. In Fuchs & V. Jirsa (Eds.), Coordination: Neural, behavioral and social dynamics (pp. 161-180).
Berlin: Springer Verlag. http://dx.doi.org/10.1007/978-3-540-74479-5_13
Funder, D. C., & Sneed, C. D. (1993). Behavioral manifestations of personality: An ecological approach to judgmental
accuracy. Journal of Personality and Social Psychology, 64(3), 479–490. https://doi.org/10.1037/0022-
3514.64.3.479
Fusaroli, R., R ˛aczaszek-Leonardi, J., and Tylén, K. (2014). Dialog as interpersonal synergy. New Ideas in Psychology, 32,
147–157. https://doi.org/10.1016/j.newideapsych.2013.03.005
Galbusera, L., Finn, M. T., Tschacher, W., & Kyselo, M. (2019). Interpersonal synchrony feels good but impedes self-
regulation of affect. Scientific Reports, 9(1), 1-12. https://doi.org/10.1038/s41598-019-50960-0
Gibson, J. J. (1979). The ecological approach to visual perception. Psychology Press.
Goldberg, L. R., Sweeney, D., Merenda, P., & Hughes, J. (1998). Demographic variables and personality: The effects of
gender, age, education, and ethnic/racial status on self-descriptions of personality attributes. Personality and
Individual Differences, 24, 393–403. https://doi.org/10.1016/S0191-8869(97)00110-4
Harley, T. A. (2013). The psychology of language: From data to theory. Psychology press.
Harris, K., English, T., Harms, P. D., Gross, J. J., & Jackson, J. J. (2017). Why Are Extraverts More Satisfied? Personality,
Social Experiences, and Subjective Well–Being in College. European Journal of Personality, 31, 170-186.
https://doi.org/10.1002/per.2101
Hove, M. J., & Risen, J. L. (2009). It’s all in the timing: Interpersonal synchrony increases affiliation. Social Cognition, 27,
949–960. doi:10.1521/soco.2009.27.6.949.
Hovhannisyan, G., & Vervaeke, J. (2022). Enactivist Big Five Theory. Phenomenology and the Cognitive Sciences, 21(2),
341–375. https://doi.org/10.1007/s11097-021-09768-5
Johnson, J. A. (2014). Measuring thirty facets of the five factor model with a 120-item public domain inventory:
Development of the IPIP-NEO-120. Journal of Research in Personality, 51, 78–89.
https://doi.org/10.1016/j.jrp.2014.05.003
Kelso, J. S., Tuller, B., Vatikiotis‐Bateson, E., & Fowler, C. A. (1984). Functionally specific articulatory cooperation
following jaw perturbations during speech: Evidence for coordinative structures. Journal of Experimental
Psychology: Human Perception and Performance, 10(6), 812–832. https://doi.org/10.1037/0096-1523.10.6.812
Koole, S. L., Jager, W., Berg, V. D., Vlek, C., & Hofstee, W. (2001). On the Social Nature of Personality: Effects of
Extraversion, Agreeableness, and Feedback about Collective Resource Use on Cooperation in a Resource Dilemma.
Personality and Social Psychology Bulletin, 27(3), 289–301. https://doi.org/10.1177/0146167201273003
Kothe, C., Medine, D., Boulay, C., Grivich, M., & Stenner, T. (2019). Lab streaming layer.
https://github.com/labstreaminglayer/ App-LabRecorder
Larsen, R. J., Buss, D. M., Wismeijer, A., Song, J., & van den Berg, S.M., Jeronimus, B.F. (2020). Personality psychology:
Domains of knowledge about human nature. McGraw-Hill.
Leaper, C., & Ayres, M. M. (2007). A Meta-Analytic review of gender variations in adults’ language use: talkativeness,
affiliative speech, and assertive speech. Personality and Social Psychology Review, 11(4), 328–363.
https://doi.org/10.1177/1088868307302221
43
Li, Y., He, L., Zhuang, K., Wu, X., Sun, J., Wei, D., & Qiu, J. (2020). Linking personality types to depressive symptoms: A
prospective typology based on neuroticism, Extraversion and conscientiousness. Neuropsychologia, 136, 107289.
https://doi.org/10.1016/j.neuropsychologia.2019.107289
López Pérez, D., Leonardi, G., Niedźwiecka, A., Radkowska, A., & Tomalski, P. (2017). Combining Recurrence Analysis
and Automatic Movement Extraction from Video Recordings to Study Behavioral Coupling in Face-to-Face Parent-
Child Interactions. Frontiers in Psychology, 8, 300530. https://doi.org/10.3389/fpsyg.2017.02228
Lucas, R. E., & Diener, E. (2001). Understanding extravert’s enjoymentof social situations: The importance of pleasantness.
Journal of Personality and Social Psychology, 81, 343–356 https://doi.org/10.1037/0022-3514.81.2.343.
Lund, K., Fossali, P. B., Mazur, A., & Ollagnier-Beldame, M. (2022). Language is a complex adaptive system: explorations
and evidence. In HAL (Le Centre pour la Communication Scientifique Directe). French National Centre for
Scientific Research. https://doi.org/10.5281/zenodo.6546419
Marwan, N., Carmen Romano, M., Thiel, M., & Kurths, J. (2007). Recurrence plots for the analysis of complex systems.
Physics Reports, 438(5-6), 237-329. https://doi.org/10.1016/j.physrep.2006.11.001
Marwan, N. (2024). Cross Recurrence Plot Toolbox for MATLAB, Version 5.29 (R38), https://tocsy.pik-
potsdam.de/CRPtoolbox
McCrae, R. R., & Costa, P. T., Jr. (2008). The five-factor theory of personality. In O. P. John, R. W. Robins, & L. A. Pervin
(Eds.), Handbook of personality: Theory and research (3rd ed., pp. 159–181). The Guilford Press.
Meeker, B. F. (2020). Nonlinear models of distribution of talking in small groups. Social Science Research, 85, 102367.
https://doi.org/10.1016/j.ssresearch.2019.102367
Mischel, W., & Shoda, Y. (1995). A cognitive-affective system theory of personality: Reconceptualizing situations,
dispositions, dynamics, and invariance in personality structure. Psychological Review, 102(2), 246–268.
https://doi.org/10.1037/0033-295x.102.2.246
Nguyen, P. L. L., Syed, M., & McGue, M. (2021). Behavior genetics research on personality: Moving beyond traits to
examine characteristic adaptations. Social and Personality Psychology Compass, 15(8), e12628.
Nowak, A., Vallacher, R. R., & Żochowski, M. (2005). The emergence of personality: Dynamic foundations of individual
variation. Developmental Review, 25(3–4), 351–385. https://doi.org/10.1016/j.dr.2005.10.004
Nowak, A., Vallacher, R. R., Żochowski, M., & Rychwalska, A. (2017). Functional synchronization: the emergence of
coordinated activity in human systems. Frontiers in Psychology, 8(945). https://doi.org/10.3389/fpsyg.2017.00945
Nowak, A. K., Vallacher, R. R., Praszkier, R., Rychwalska, A., & Zochowski, M. (2020). In sync: The Emergence of
Function in Minds, Groups and Societies. Springer. https://doi.org/10.1007/978-3-030-38987-1
Oben, B., & Brône, G. (2016). Explaining interactive alignment: A multimodal and multifactorial account. Journal of
Pragmatics, 104, 32–51. https://doi.org/10.1016/j.pragma.2016.07.002
Paxton, A., & Dale, R. (2013). Argument disrupts interpersonal synchrony. Quarterly Journal of Experimental Psychology,
66(11), 2092–2102. https://doi.org/10.1080/17470218.2013.853089
Paxton, A., & Dale, R. (2017). Interpersonal movement synchrony responds to high- and Low-Level conversational
constraints. Frontiers in Psychology, 8. https://doi.org/10.3389/fpsyg.2017.01135
Peabody, D., & Goldberg, L. R. (1989). Some determinants of factor structures from personality-trait descriptors. Journal of
Personality and Social Psychology, 57, 552–567. https://psycnet.apa.org/doi/10.1037/0022-3514.57.3.552
Peperkoorn LS, Becker DV, Balliet D, Columbus S, Molho C, Van Lange PAM (2020) The prevalence of dyads in social
life. PLoS ONE 15(12): e0244188. https://doi.org/10.1371/journal.pone.0244188
Pickering, M. J., & Garrod, S. (2004). Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences, 27, 169
–190. https://10.1017/S0140525X04000056
R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. URL https://www.R-project.org/
Revelle, W., & Wilt, J. (2020). The history of dynamic approaches to personality. The Handbook of Personality Dynamics
and Processes, 3-31. https://doi.org/10.1016/B978-0-12-813995-0.00001-7
Reuzel, E., Embregts, P.J.C.M., Bosman, A.M.T. et al. (2013). Conversational Synchronization in Naturally Occurring
Settings: A Recurrence-Based Analysis of Gaze Directions and Speech Rhythms of Staff and Clients with
Intellectual Disability. Journal of Nonverbal Behavior, 37, 281–305. https://doi.org/10.1007/s10919-013-0158-9
44
Reuzel, E., Embregts, P., Bosman, A., Cox, R. F. A., Van Nieuwenhuijzen, M., & Jahoda, A. (2014). Verbal interactional
dominance and coordinative structure of speech rhythms of staff and clients with an intellectual disability. Nonlinear
Dynamics, Psychology, and Life Sciences, 18(4), 371–396. https://pubmed.ncbi.nlm.nih.gov/25196706
Richardson, D. C., and Dale, R. (2005). Looking To Understand: The Coupling Between Speakers' and Listeners' Eye
Movements and Its Relationship to Discourse Comprehension. Cognitive Science, 29(6), 1045–1060.
https://doi.org/10.1207/s15516709cog0000_29
Satchell, L., Kaaronen, R. O., & Latzman, R. D. (2021). An ecological approach to personality: Psychological traits as
drivers and consequences of active perception. Social and Personality Psychology Compass, 15(5).
https://doi.org/10.1111/spc3.12595
Scheidt, C. E., Pfänder, S., Ballati, A., Schmidt, S., & Lahmann, C. (2021). Language and Movement Synchronization in
Dyadic Psychotherapeutic Interaction – A Qualitative Review and a Proposal for a Classification. Frontiers in
Psychology, 12, 696448. https://doi.org/10.3389/fpsyg.2021.696448
Shockley, K., Santana, M.-V., & Fowler, C. A. (2003). Mutual interpersonal postural constraints are involved in cooperative
conversation. Journal of Experimental Psychology: Human Perception and Performance, 29(2), 326–332.
https://doi.org/10.1037/0096-1523.29.2.326
Shockley, K., Richardson, D. C., & Dale, R. (2009). Conversation and coordinative structures. Topics in Cognitive Science,
1(2), 305–319. https://doi.org/10.1111/j.1756-8765.2009.01021.x
Sosnowska, J., Kuppens, P., De Fruyt, F., & Hofmans, J. (2019). A dynamic systems approach to personality: The
Personality Dynamics (PersDyn) model. Personality and individual differences, 144, 11-18.
https://doi.org/10.1016/j.paid.2019.02.013
Stern, D. N. (1985/2018). The interpersonal world of the infant. In Routledge eBooks.
https://doi.org/10.4324/9780429482137
The MathWorks Inc. (2022). MATLAB version: 9.13.0 (R2022b), Natick, Massachusetts: The MathWorks Inc.
https://www.mathworks.com
Thorne, A. (1987). The press of personality: A study of conversations between introverts and extraverts. Journal of
Personality and Social Psychology, 53(4), 718–726. https://doi.org/10.1037/0022-3514.53.4.718
Thibault, P. (2004). Agency and Consciousness in Discourse: Self-other Dynamics as a Complex System. Bloomsbury
Publishing.
Thorson, K. R., Ketay, S., Roy, A. R., & Welker, K. M. (2021). Self-disclosure is associated with adrenocortical attunement
between new acquaintances. Psychoneuroendocrinology, 132, 105323.
https://doi.org/10.1016/j.psyneuen.2021.105323
Tomashin, A., Leonardi, G., & Wallot, S. (2022). Four Methods to Distinguish between Fractal Dimensions in Time Series
through Recurrence Quantification Analysis. Entropy, 24(9), 1314. https://doi.org/10.3390/e24091314
Tschacher, W., Ramseyer, F., & Koole, S. L. (2018). Sharing the Now in the Social Present: Duration of Nonverbal
Synchrony Is Linked With Personality. Journal of Personality, 86(2), 129-138. https://doi.org/10.1111/jopy.12298
Tuovinen, S., Tang, X., & Salmela-Aro, K. (2020). Introversion and Social Engagement: Scale Validation, Their Interaction,
and Positive Association With Self-Esteem. Frontiers in Psychology, 11. https://doi.org/10.3389/fpsyg.2020.590748
Van Dijk, M., Van Voorthuizen, B., & Cox, R. F. (2018). Synchronization of mother-infant feeding behavior. Infant
Behavior and Development, 52, 97-103. https://doi.org/10.1016/j.infbeh.2018.06.001
Varela, F. J., Thompson, E., & Rosch, E. (1991/2017). The Embodied Mind: Cognitive Science and Human Experience.
https://doi.org/10.7551/mitpress/9780262529365.001.0001
Wallot, S., & Leonardi, G. (2018). Analyzing Multivariate Dynamics Using Cross-Recurrence Quantification Analysis
(CRQA), Diagonal-Cross-Recurrence Profiles (DCRP), and Multidimensional Recurrence Quantification Analysis
(MdRQA) – A Tutorial in R. Frontiers in Psychology, 9, 365352. https://doi.org/10.3389/fpsyg.2018.02232
Warner, R. (1992).Sequential analysis of social interaction: Assessing internal versus social determinants of behavior.
Journal of Personality and Social Psychology, 63(1),51–60. https://doi.org/10.1037/0022-3514.63.1.51
Worgan, S. F., & Moore, R. K. (2010). Speech as the perception of affordances. Ecological Psychology, 22(4), 327–343.
https://doi.org/10.1080/10407413.2010.517125
45
Xu, T.L., De Barbaro, K., Abney, D.H., & Cox, R.F.A. (2020). Finding Structure in Time: Visualizing and Analyzing
Behavioral Time Series. Frontiers in Psychology, 11: 1457.
Zbilut, J.P. and Webber Jr., C.L. (1992) Embeddings and Delays as Derived Quantification of Recurrence Plots. Physics
Letters A, 171, 199-203. http://dx.doi.org/10.1016/0375-9601(92)90426-M
Zbilut, J. P., Giuliani, A., & Webber, C. L. (1998). Recurrence quantification analysis and principal components in the
detection of short complex signals. Physics Letters A, 237(3), 131-135. https://doi.org/10.1016/S0375-
9601(97)00843-8
46
Table 3.
Descriptive statistics
1. Introduction
2. Self-disclosure
3. Argumentative
Variable
M
SD
Mdn
Range
M
SD
Mdn
Range
M
SD
Mdn
Range
Speech
time
(seconds)
467.14
78.53
461
[306, 616]
456.4
84.04
469
[258, 617]
445.92
86.94
443.5
[230, 671]
RRglobal
0.46
0.06
0.47
[0.26, 0.54]
0.46
0.05
0.47
[0.32, 0.55]
0.47
0.06
0.49
[0.30, 0.58]
RRLOS
0.56
0.14
0.59
[0.17, 0.85]
0.55
0.14
0.56
[0.13, 0.80]
0.58
0.15
0.58
[0.10, 0.82]
QDCRP
0.02
0.02
0.02
[0, 0.05]
0.03
0.02
0.02
[0, 0.12]
0.03
0.03
0.03
[0, 0.10]
LAMa
0.93
0.03
0.94
[0.82, 0.98]
0.94
0.04
0.95
[0.81, 0.99]
0.94
0.04
0.95
[0.75, 0.99]
TTa
8.17
2.63
7.92
[4.24, 18.8]
8.74
4.46
8.07
[3.7, 46.29]
9.51
3.77
8.7
[4.55, 25.95]
LAMARD
0.02
0.02
0.02
[0, 0.11]
0.02
0.02
0.02
[0, 0.07]
0.02
0.02
0.02
[0, 0.07]
TTARD
0.15
0.13
0.09
[0.01, 0.54]
0.16
0.13
0.12
[0.11, 0.53]
0.16
0.10
0.13
[0.01, 0.44]
N = 100 participants (50 dyads). M = mean, SD = standard deviation, Mdn = median. RRglobal= Recurrence Rate
Global, RRLOS= Percentage of Speech across the line of Synchrony (lag-zero), QDCRP = Quotient Diagonal Cross-
Recurrence Profile (absolute), LAMa= anisotropic Laminarity, TTa= anisotropic Trapping Time, LAMARD= Relative
difference of anisotropic Laminarity, TTARD= Relative difference of anisotropic Trapping Time. Anisotropic
Laminarity (LAMa) and anisotropic Trapping Time (TTa) are the measures from where the relative difference of
anisotropic Laminarity (LAMARD) and Trapping Time (TTARD) were calculated respectively.
47
Table 4.
Mixed-effects models predicting CRQA, DCRP and aCRQA measures from tasks with 50 dyads (Ni) (100
participants) and 150 observations (Nt), (50i * 3 tasks)
M1. RRglobal
M2. RRLOS
M3. QDCRP
M4. LAMARD
M5. TTARD
Predictors
Estimate
B (SE)
t
Estimate
B (SE)
t
Estimate
B (SE)
t
Estimate
B (SE)
t
Estimate
B (SE)
t
Intercept
0.456
(0.008)
53.99***
0.564
(0.021)
28.69***
0.022
(0.002)
7.43***
0.017
(0.002)
8.32***
0.111
(0.017)
6.62***
Task 2
0.004
(0.007)
0.61
-0.015
(0.013)
-1.23
0.004
(0.003)
0.97
0.002
(0.002)
0.69
0.052
(0.021)
2.43*
Task 3
0.019
(0.007)
2.86**
0.014
(0.126)
1.08
0.011
(0.004)
2.92**
0.001
(0.002)
0.29
0.046
(0.021)
2.14*
Random Effects
ICC
0.70
0.80
0.70
0.27
0.19
Marg. R2/
Cond. R2
0.02/0.70
0.01/0.80
0.01/0.70
0.002/0.27
0.04/0.22
AIC
-461
-247
-693
-811
-189
Note: Significance was indicated as * p < .05. ** p < .01,***p <.001. Ni= number of participants. Nt= total; number of
observations, which was = 150 (50 dyads * 3 tasks). SE= Standard Error. Task 2 = Self-disclosure. Task 3 =
Argument. Task 1 (Introduction) was considered the baseline in the models. AIC = Akaike’s Information Criterion
(lower values indicate better fit). ICC = Intra-class Correlation Coefficient. see Table 1 for definitions. RRglobal=
Recurrence Rate Global, RRLOS= Percentage of Speech across the line of Synchrony (lag-zero), QDCRP = Quotient
Diagonal Cross-Recurrence Profile (absolute), LAMa= anisotropic Laminarity, TTa= anisotropic Trapping Time,
LAMARD= Relative difference of anisotropic Laminarity, TTARD= Relative difference of anisotropic Trapping Time.
Definitions of each measure can be found in Table 1.
48
Table 5.
Mixed-effects models predicting CRQA, DCRP, and aCRQA measures from Extraversion and task. Ni=50 dyads (100 participants); Nt=150 observations (50i*3
topics)
M1. RRglobal
M2. RRLOS
M3. QDCRP
M4. LAMARD
M5. TTARD
Predictors
B
𝛽
t
B
𝛽
t
B
𝛽
t
B
𝛽
t
B
𝛽
t
Intercept
0.45
-0.15
54.34***
0.56
-0.01
28.8***
0.02
-0.22
7.70***
0.02
-0.07
8.55***
0.11
-0.26
6.65***
Extraversion “A” (EA)
0.01
0.21
1.45
0.03
0.21
1.44
-0.00
-0.01
-0.06
0.00
0.01
0.05
0.00
0.02
0.14
Extraversion “B” (EB)
-0.01
-0.21
-1.44
-0.03
-0.18
-1.22
0.00
0.03
0.23
0.00
0.19
1.35
0.01
0.08
0.57
Task 2. Self-disclosure
0.00
0.07
0.69
-0.01
-0.10
-1.09
0.00
0.12
0.70
0.00
0.18
1.10
0.05
0.44
2.51*
Task 3. Argumentative
0.02
0.32
2.98**
0.01
0.09
1.04
0.01
0.54
3.02**
0.00
0.09
0.57
0.05
0.37
2.10*
EA * EB
0.01
0.18
1.2
0.02
0.12
0.76
-0.00
-0.13
-0.86
0.00
0.06
0.44
-0.00
-0.03
-0.19
EA * Task 2
-0.01
-0.13
-1.15
-0.01
-0.06
-0.62
0.00
0.16
0.88
0.00
0.00
0.01
0.02
0.14
0.78
EA * Task 3
-0.01
-0.22
-2.04*
0.00
0.00
0.05
-0.00
-0.14
-0.76
-0.00
-0.14
-0.85
0.00
0.00
0.01
EB * Task 2
0.01
0.20
1.77
0.02
0.12
1.28
-0.01
-0.34
-1.82
-0.00
-0.16
-0.93
-0.00
-0.03
-0.18
EB * Task 3
0.00
0.02
0.17
0.02
0.12
1.33
-0.00
-0.09
-0.46
0.00
0.14
0.81
0.00
0.04
0.21
EA*EB*Task 2
0.00
-0.04
-0.36
-0.01
-0.09
-0.91
0.01
0.36
1.82
-0.01
-0.42
-2.32*
-0.01
-0.11
-0.55
EA*EB*Task 3
0.00
-0.04
-0.30
0.00
0.03
0.28
-0.00
-0.04
-0.20
-0.00
-0.30
-1.66
0.00
0.04
0.20
ICC
0.80
0.80
0.13
0.28
0.18
Marg. R2/Cond. R2
0.063/0.725
0.050/0.805
0.106/0.218
0.094/0.346
0.063/0.235
AIC
-379.1
-175.4
-597
-713.2
-118.3
Note: B=unstandardized raw estimate; 𝛽=beta weights (standardized). * indicates p < .05. ** indicates p < .01. ***indicates p <.000. p-values were BH corrected (Benjamini &
Hochberg, 1995) FDR procedure. Task 1 (Introduction) is the baseline. Task 2 = Self-disclosure; Task 3 = Argumentative. E=Extraversion. ICC = Intra-class Correlation
Coefficient. AIC = Akaike’s Information Criterion (lower values indicate better fit). Personality traits were centered/scaled. RRglobal= Recurrence Rate Global, RRLOS = Percentage
of Speech across the line of Synchrony (lag-zero), QDCRP= Quotient Diagonal Cross-Recurrence Profile (absolute), LAMa= anisotropic Laminarity, TTa= anisotropic Trapping
Time, LAMARDf= Relative difference of anisotropic Laminarity, TTARD= Relative difference of anisotropic Trapping Time. Definitions of each measure can be found in Table 1.
49
Table 6.
Mixed-effects models predicting CRQA, DCRP and aCRQA measures from Agreeableness and task. Ni=50 dyads (100 participants); Nt=150 observations (50i *
3 topics)
M1. RRglobal
M2. RRLOS
M3. QDCRP
M4. LAMARD
M5. TTARD
Predictors
B
𝛽
t
B
𝛽
t
B
𝛽
t
B
𝛽
t
B
𝛽
t
Intercept
0.46
-0.10
55.14***
0.57
0.03
29.27***
0.02
-0.21
7.83***
0.02
-0.02
8.93***
0.11
-0.25
6.90***
Agreeableness “A” (AA)
-0.00
-0.02
-0.17
-0.01
-0.09
-0.62
-0.00
-0.04
-0.30
-0.00
-0.12
-0.81
0.00
0.03
0.18
Agreeableness “B” (AB)
-0.01
-0.09
-0.67
-0.01
-0.10
-0.73
-0.01
-0.23
-1.67
0.00
0.07
0.48
-0.00
-0.12
-0.82
Task 2. Self-disclosure
0.00
0.07
0.66
-0.02
-0.12
-1.31
0.00
0.14
0.79
0.00
0.10
0.63
0.05
0.42
2.38*
Task 3. Argumentative
0.02
0.34
3.23**
0.02
0.12
1.31
0.01
0.50
2.81**
-0.00
-0.04
-0.23
0.05
0.39
2.20*
AA * AB
-0.01
-0.09
-0.81
-0.02
-0.12
-1.01
-0.00
-0.12
-1.12
-0.00
-0.15
-1.32
-0.01
-0.10
-0.85
AA * Task 2
-0.00
-0.01
-0.11
0.00
0.02
0.20
-0.00
-0.21
-1.10
0.00
0.32
1.85
0.02
0.17
0.90
AA * Task 3
-0.01
-0.09
-0.77
-0.00
-0.02
-0.17
-0.00
-0.15
-0.80
0.00
0.42
2.44*
0.01
0.12
0.65
AB * Task 2
-0.00
-0.04
-0.36
-0.01
-0.08
-0.87
0.01
0.56
3.06**
0.01
-0.07
-0.45
0.01
0.12
0.65
AB * Task 3
-0.02
-0.26
-2.37*
-0.03
-0.18
-2.01*
0.01
0.26
1.42
-0.00
0.29
1.76
-0.03
-0.21
-1.15
AA*AB*Task 2
-0.00
-0.01
0.87
0.00
0.03
0.35
0.00
0.16
1.08
0.00
0.07
0.55
0.00
0.04
0.26
AA*AB*Task 3
-0.01
-0.14
-1.57
-0.01
-0.08
-1.12
0.00
0.16
1.06
0.00
0.36
2.74**
-0.01
-0.04
-0.29
ICC
0.70
0.79
0.13
0.29
0.16
Marg. R2/Cond. R2
0.090/0.727
0.071/0.808
0.130/0.241
0.117/0.377
0.105/0.252
AIC
-379.7
-176.4
-599.3
-716.3
-122.2
Note: B=unstandardized raw estimate; 𝛽=beta weights (standardized). * indicates p < .05. ** indicates p < .01. ***indicates p <.000. p-values were BH corrected (Benjamini &
Hochberg, 1995) FDR procedure. Task 1 (Introduction) is the baseline. Task 2 = Self-disclosure; Task 3 = Argumentative. A=Agreeableness. ICC = Intraclass Correlation
Coefficient. AIC = Akaike’s Information Criterion (lower values indicate better fit). Personality traits were centered/scaled. RRglobal= Recurrence Rate Global, RRLOS= Percentage
of Speech across the line of Synchrony (lag-zero), QDCRP = Quotient Diagonal Cross-Recurrence Profile (absolute), LAMa= anisotropic Laminarity, TTa= anisotropic Trapping
Time, LAMARD= Relative difference of anisotropic Laminarity, TTARD = Relative difference of anisotropic Trapping Time. Definitions of each measure can be found in Table 1.
