Capturing the Temporal Dynamics of Language Style Matching in Groups and Teams

Abstract

This article presents a dynamic conceptualization for the assessment of language style matching (LSM) over time. LSM is a team’s mutual adaption of function words like pronouns, articles, or prepositions. LSM is a nonconsciously but frequently occurring communication behavior allowing researchers unobtrusive insights into teams’ internal dynamics. Building on guidelines for the alignment of construct and measurement, a dynamic conceptualization and method for LSM are introduced. Simulated examples and interactions of N = 160 individuals in 26 teams indicate that dynamic LSM allows for a truer estimation of LSM than the hitherto used static method. Implications for future application are discussed.

Keywords

temporal dynamics team processes language style matching group process tutorial

Language is the gateway to analyzing implicit team processes and thereby improving our understanding of team dynamics. Previous research assessing language processes has already advanced our understanding of various (emergent) team phenomena (e.g., cohesiveness in Gonzales et al., 2010; influence in Yilmaz & Peña, 2015; conflict in Biesen et al., 2016). One specific linguistic process that is related to various team inputs (e.g., social status, personality) and outputs (e.g., trust, cohesion) is language style matching (LSM; Niederhoffer & Pennebaker, 2002). LSM is an implicit communication behavior characterized by the mutual use and adaption of function words (e.g., pronouns, articles, or prepositions) over the course of an interaction (e.g., Ireland & Pennebaker, 2010; Müller-Frommeyer & Kauffeld, 2021; Müller-Frommeyer et al., 2019; Niederhoffer & Pennebaker, 2002). Function words are uttered nonconsciously (Segalowitz & Lane, 2004). At the same time, they are frequently used in natural language allowing for unobtrusive insights into implicit communication dynamics in team interaction. In an overarching input-process-output framework (Ilgen et al., 2005), LSM can be classified as team process (Van Swol & Kane, 2019). Investigating LSM can yield important insights into a team’s internal dynamics (Chung & Pennebaker, 2013) and advance our understanding thereof. As such, LSM is a relevant phenomenon to team researchers and practitioners.

Although research on LSM in teams has flourished over the last decade, LSM has uniformly been treated as a static concept (e.g., Gonzales et al., 2010; Heuer et al., 2020; Yilmaz & Peña, 2015) by conceptualizing and investigating it on a conversational level. However, team processes in general, and language production in particular, are inherently dynamic phenomena, meaning that they take place in social settings and develop over time (e.g., Kozlowski, 2015; Lehmann-Willenbrock & Allen, 2018; Müller-Frommeyer et al., 2019). By choosing static conceptualizations, the temporal dynamics of LSM have been neglected and its process mechanisms remain essentially unstudied. Hence, we lack knowledge of the temporal process mechanisms of LSM and their contribution to the emergence of team phenomena. However, this lack of knowledge is grounded in a conceptual (i.e., theoretical and methodological) issue. This issue ties into an on-going call for the introduction and use of adequate methods that consider the process dynamics of team phenomena and processes that contribute to it (e.g., Kozlowski, 2015; Kozlowski et al., 2013; Lehmann-Willenbrock & Allen, 2018; Luciano et al., 2018; Waller et al., 2016).

Therefore, the present paper addresses this research gap. First, we re-conceptualize LSM in teams as a dynamic process observed over time by providing a review of theoretical and empirical insights into LSM. By implementing LSM in teams into a more dynamical framework, we contribute to expanding our knowledge of the specific process dynamics of LSM in teams. Based on our new conceptualization and a review of the existing static method to assess LSM in teams (LSM-t), we then introduce a dynamic method to assess LSM in team interactions (rLSM-t). With the new conceptualization, we follow current recommendations for construct and measurement alignment (Luciano et al., 2018). We then empirically compare both metrics: First, we use three simulated examples to examine if rLSM-t is a truer estimation of LSM than LSM-t. Then, we transfer these results to real-life interactions of N = 160 individuals in 26 innovation teams.

Throughout the paper, we provide several tutorials to facilitate the application of this new method. We give a summary of concrete guidelines on the collection, preparation, and quantification of language data from team interactions. Further, we provide sample data, transcripts, and an R Script for the calculation of our newly introduced method rLSM-t in the on the open science framework (osf) for this paper (https://osf.io/n9zg6/).

The dynamic conceptualization of LSM implements LSM into the literature on (emergent) team processes. Its future application will yield important insights for researchers interested in team dynamics in social, organizational, and communication studies. Applying this method will further advance our understanding of how verbal micro-processes like LSM contribute to the emergence of team phenomena and practical implications thereof.

Language Style Matching in Teams

When assessing LSM, researchers are interested in the use of function words such as prepositions, articles, or conjunctions. Making up less than 4% of our vocabulary (in the English language; Van Gelderen, 2014), function words account for up to 60% of the words we use in everyday conversations. Being short and having almost no meaning outside the respective context, function words reflect how we say things rather than what we say (Chung & Pennebaker, 2013). Therefore, function words can be measured and compared across different social settings and conversational contexts (Gonzales et al., 2010). Their use is more automated and nonconscious than the use of content words such as nouns and verbs (Segalowitz & Lane, 2004). Linguistically, function words reflect the relationship between content words and thereby signal shared knowledge among interaction partners (Brennan & Hanna, 2009; Cannava & Bodie, 2017). Each individual’s specific pattern of function word use is called language style (Pennebaker & King, 1999; Pennebaker et al., 2003). With its specific focus on aspects of communication that are nonconsciously but frequently used, LSM has an advantage over other well-established research techniques for analyzing communication (e.g., interaction analysis, content analysis) which focus on aspects of communication that can be deliberately trained or changed. An overview of more established research techniques for communication analysis and their specific goals in comparison to LSM can be found in Table 1.

Table 1.

Overview of LSM in Comparison to Interaction and Content Analysis.

Technique	Explanation	Aim of analysis	Focus of analysis	Example reference
LSM	Assessment of (dynamic) mutual function word use.	Analyze LSM as potential indicator of internal dynamics or emergent group phenomena.	Word-level, specifically function words.	Gonzales et al. (2010), Heuer et al. (2020)
Interaction analysis	Detailed coding of specific pre-defined group behaviors.	Uncover behavior patterns and dynamics.	Ranges from single words to whole sentences, statements, emotions or non-verbal behavior.	Kauffeld and Lehmann-Willenbrock (2012), Lehmann-Willenbrock et al. (2011)
Content analysis	Coding of text and pre-defined parts of text.	Make specific replicable and valid inferences.	Depends on the aim of the analysis, can range from specific words or sentences to the tone or emotion of a text.	Arafat et al. (2020), Karl et al. (2021)

Note. This table is meant to provide a first overview of LSM in comparison to other well-established research techniques that can be applied for the analysis of team communication. However, this table is not exhaustive in the presentation of research techniques.

When interacting with each other, team members match their language styles to one another (e.g., Gonzales et al., 2010; Heuer et al., 2020; Scissors et al., 2008), called LSM. LSM in teams is defined as verbal mimicry represented by how each team member matches the other team members on specific categories of function words in a given conversation (e.g., Carmody et al., 2017; Gonzales et al., 2010; Van Swol & Carlson, 2017; Yilmaz, 2016) and can, thus, be conceived as a form of behavior coordination. Research shows that LSM is empirically linked to team inputs (e.g., personality, social status) and outcomes (e.g., performance, cohesion). For example, Muir et al. (2016) showed that social power and personality influence LSM in a way that more LSM in low-power individuals positively influences perceptions of rapport and attractiveness. Scissors et al. (2008) report higher trust in teams that show higher LSM as compared to teams with lower LSM. Gonzales et al. (2010) found that LSM positively predicted cohesiveness in face-to-face as well as computer-mediated communication, whereas LSM influenced task performance positively only in face-to-face interactions. Heuer et al. (2020) found a negative relationship between LSM and team performance and a positive relationship between LSM and social support in face-to-face team meetings. Carmody et al.’s (2017) findings deviate from the picture presented above: They found a negative relationship between LSM and trust and no relationship between LSM and cohesion and rapport in computer-mediated communication. In summary, these empirical findings illustrate the important influence of LSM in team interaction, but at the same time these findings remain ambiguous.

Taken together, assessing LSM in teams allows researchers interested in team communication to gain unobtrusive insights into team processes that can neither be applied consciously nor deliberately changed and, thus, reflect a starting point to gain a deeper understanding of how different team phenomena develop as a result of this naturally occurring team process. However, to date approaches to assess LSM in teams have uniformly treated the phenomenon as static by neglecting its dynamic changes over time from a theoretical and methodological perspective.

Toward a Dynamic Theoretical Approach to LSM

Traditionally, LSM has been explained under a variety of theoretical frameworks including, for example, behavior mimicry (Chartrand & Lakin, 2013; Ireland & Pennebaker, 2010), interpersonal synchrony (Bernieri & Rosenthal, 1991), communication accommodation theory (Giles & Coupland, 1991; Shephard et al., 2001) or interpersonal alignment (Garrod & Pickering, 2009; Menenti et al., 2012). These theoretical approaches provide explanations of the occurrence and positive effects of LSM agreeing on two basic assumptions: (1) LSM happens automatically and (2) affects the interaction positively (Müller-Frommeyer et al., 2019). However, research has shown that LSM varies depending on team inputs (Muir et al., 2016) and there is empirical evidence that LSM does not only yield positive outcomes (Heuer et al., 2020). Additionally, the theoretical approaches to LSM named above all miss a temporal dimension of coordination that considers the dynamic and emergent nature inherent in natural language production.

Therefore, in the following, we integrate different theoretical approaches based on dynamic theories to develop a comprehensive dynamic theoretical foundation for LSM for the present paper and future research. Building on this theoretical foundation, we further present a comprehensive conceptualization of LSM.

Integrating Dynamic Theories

The unifying feature of all dynamic theories relevant in the explanation of LSM is the assumption that LSM is dynamic and over time. This temporal component can take different forms which can be summarized into three general types of temporal frameworks (for more detailed information see Luciano et al., 2018): (1) Developmental models which suggest that dynamic constructs have a pre-defined life span and naturally change as they mature over time with their current state depending on their previous state. (2) Episodic models which treat dynamic constructs as dependent on a goal or end state resulting in different characteristics at different points in time. (3) Event-based models suggest that external stimuli—that is, everything that happens in the direct environment—intermit the natural emergence of the dynamic construct by activating internal processes. Dynamic constructs like LSM can be based on multiple temporal models at the same time (Luciano et al., 2018). For LSM, this interplay of temporal models could look as follows: (1) Following developmental models, the stage of team development (e.g., newly formed vs. experienced teams) might affect LSM. (2) In episodic models, the requirements, and goals of a specific task at hand might influence LSM. (3) In event-based models, LSM might be affected by external events (e.g., team feedback).

One theoretical approach to LSM which combines aspects of all three temporal models is the interpersonal synergy approach (Riley et al., 2011). Interpersonal synergy provides a dynamical framework for LSM which is inspired by dynamical systems theory and has previously been used to explain the temporal dynamics of LSM in dyadic interactions (Müller-Frommeyer et al., 2019). The above-mentioned traditional assumption of automatic coordination is challenged by the interpersonal synergy framework that conceptualizes individual communication behavior as part of a complex process in which individuals coordinate and adapt their behavior (and specific aspects thereof) to one another (i.e., event-based) depending on contextual and situational constraints (i.e., episodic) over the course of an interaction (i.e., developmental; Fusaroli et al., 2014). Additionally, empirical evidence challenges the assumption of only positive effects of LSM (Heuer et al., 2020). We therefore propose that interpersonal synergy is also a more suitable framework for LSM in teams. Transferred to LSM in teams, this means, that all members in a team coordinate their function words—that is, they use their function words depending on the other team members’ function word use—and thereby develop LSM over the course of the interaction to meet immediate common goals throughout the interaction (e.g., solving a common task; Fusaroli et al., 2014; Riley et al., 2011).

A dynamic conceptualization of LSM in teams

After theoretically integrating LSM in teams into the concept of interpersonal synergy, the deviance between the dynamic nature of the theoretical foundation of LSM and its current definition in teams as presented in the introduction is striking. The importance of a clear and unambiguous definition of constructs in line with the theoretical framework is a fundamental aspect of high research quality and has regained attention in recent years (e.g., Aguinis & Vandenberg, 2014; Luciano et al., 2018; Podsakoff et al., 2016). Our conceptualization of LSM in teams is based on the four-stage model necessary for the explication of dynamic constructs defined by Luciano et al. (2018). These four stages encompass the construct’s space (i.e., content and dimensionality), nature (i.e., property such as process or emergent state), and entity (e.g., affect, behavior and/or cognition) as well as structure (i.e., the construct’s shape and change over time) and appearance (i.e., the observable manifestation of the construct and conditions for this manifestation to occur).

Following this four-stage model, we present a new conceptualization of LSM in teams: First, when analyzing LSM in teams, we investigate each individual’s function word use (space). All research on LSM in teams is based on computerized text analyses using the software Linguistic Inquiry and Word Count (LIWC; Pennebaker et al., 2015). LIWC is dictionary-based and assigns all words in a transcript into pre-defined categories. We therefore propose that LSM analysis only includes words that are listed in the LIWC function word categories. The current versions of the English and German dictionaries comprise the seven function word categories pronouns (e.g., I, them, itself), articles (e.g., a, an, the), prepositions (e.g., to, with, above), auxiliary verbs (e.g., am, will, have), adverbs (e.g., very, really), conjunctions (e.g., and, but, whereas), and negations (e.g., no, not, never). Deviations from the LIWC categories should be an exception and well justified. All these seven function word categories are weighed equally when investigating overall function word use. Second, we define LSM as a behavioral team process (nature). Third, following the interpersonal synergy framework, we expect LSM in teams to develop and change over the course of team interactions depending on environmental constraints (e.g., tasks, goals) (structure). Thus, methods applied to investigate LSM in teams need to be able to capture this dynamic over time. Fourth, we propose that LSM manifests in verbal interaction—this includes face-to-face and virtual interactions—of at least three interlocutors (for LSM in dyadic interaction see Müller-Frommeyer et al., 2019) (appearance). In summary, we define LSM as a dynamic verbal process that describes how individuals in a team interaction coordinate their function word use with each other over the course of an interaction depending on situational and environmental constraints.

Collection, Preparation, and Quantification of Language Samples

After providing the theoretical and conceptual foundations, we now proceed with a hands-on tutorial that aims at providing helpful guidelines to researchers who want to collect, prepare, and/or quantify communication data from a team interaction. Here, we partly deviate from classic academic writing to facilitate understanding. Based on the literature on LSM, there are currently no guidelines on these rather practical steps of conducting research. However, these steps form the basis for subsequent analysis and are possibly the most time-consuming in interaction research. Therefore, ensuring the collection of high-quality data and minimizing potential errors in and time spent on data preparation and quantification is an important step.

The basis for the analysis of LSM in teams is based on natural team interactions. These can be protocols of verbal team interactions that originate from more traditional video or audio recordings of face-to-face interactions or software-aided (e.g., Skype) interactions. With the rise of virtual team interactions, written protocols of these interactions (e.g., email, chat) or timestamped discussions on social media platforms (e.g., Twitter, Facebook) have also been considered in LSM research (Gonzales et al., 2010; Heuer et al., 2020). Because of the amount of data generated when analyzing team communication (i.e., streams of words in an interaction), communication data produced by teams can be classified as a word-related stream of big data (Luciano et al., 2018). Using LIWC, these streams of words are quantified into meaningful categories (Pennebaker et al., 2015).

Data Preparation: Getting Team Interactions Ready for LSM Calculation

Before LSM can be calculated, several data preparation and data processing steps need to be fulfilled. The steps of data preparation and analyses using LIWC are similar when working with natural language samples independent of the number of speakers or the method that is used to assess LSM. Short examples of correctly prepared transcripts are available in on the osf.

Recording face-to-face interactions

When working with face-to-face team interactions, those interactions need to be recorded to extract the team’s communication using video- and/or audio recording. Before deciding on how to record a team interaction, the following aspects should be considered: What is the primary aim of your research? For dynamic LSM research, you need to be able to (1) clearly identify each speaker throughout the whole interaction, and (2) transcribe each word uttered by each team member. With an increasing number of speakers, identifying each individual solely by their voice becomes more difficult. Therefore, we advise using video recordings when working with teams, and if applicable even video recordings combined with individual audio channels for each team member.

Transcribing face-to-face interactions

Once all team interactions are recorded, they need to be transcribed word for word clearly marking each team member. Sequences of clean-cut statements are created and each of the statements is assigned to one team member. A statement starts when a first speaker says their first word, and it ends with their last word before the following speaker utters their first word. Unfortunately, speaking turns in real-life conversations are often overlapping, for example when multiple team members react to a preceding statement simultaneously. In such cases, clean-cuts need to be artificially created, either by representing the chronological order in the range of milliseconds or by determining the order to the best of your knowledge and belief (Müller-Frommeyer et al., 2019). More details on the transcription of oral language are provided in the LIWC 2007 manual. Final transcripts include all words uttered throughout the interaction organized in clean-cut statements that are clearly assigned to one team member. To process transcripts with LIWC, they can either be prepared in a text (e.g., .doc, .txt) or table (e.g., .csv, .xlsx) format. We recommend working with table formats because this allows you to include information relevant for further analyses in addition to the linguistic content uttered by each speaker (e.g., team number, speaker). An example of a correctly prepared transcript can be found in Table 2 (columns one to three).

Table 2.

Example Transcript Illustrating the Calculation of rLSM–t Scores.

Team no.	Speaker	Linguistic content	Function words (%)	rLSM	Score for
1	A	So we start? So, they’re asking for a brochure. So, what do you think could be important information that we should include in brochure?	66.67
1	B	So as soon as the students come in, the new the first and foremost thing they have to do is a registration.	72.73	.96	B
1	A	Ok.	0	.00	A
1	B	So they should be. . .	100	NA	B
1	A	Timetable registration and	33.33	.50	A
1	B	Yeah, the registration includes a timetable and registering in their courses and. . .	50	.80	B
1	A	Yes.	0	.00	A
1	B	Like enrolling in the university, you get a rendered ID-card made, student ID made. So, yeah, the foremost thing should be the first chapter, the first bullet point should be registration.	50	NA	B
1	C	But there should also be a map, like a map of the campus.	76.92	.80	C
			rLSM-t	.51

Note. Example is taken from an interaction between three team members we recorded in a research project that had the aim of identifying the relationship between language use in teams and perceptions of warmth and competence. Team no. = Number that clearly identifies the team. Speaker = Team member. Linguistic Content = transcribed statement or documentation of virtual communication. Function Words (%) = percentage of function words in the text. rLSM = rLSM score for two successive statements. Score For = rLSM score assigned to the respective team member. NA = no score was calculated due to missing values.

Editing written protocols of virtual team interaction

Sometimes, it might be the case that researchers have access to protocols of virtual team interactions (e.g., chat protocols from Skype or Microsoft Teams, or email conversations). When working with such protocols, the (tedious) step of transcription can be skipped. However, to make sure that virtual communication data is processed correctly, a series of steps needs to be conducted to bring it into the correct format. We illustrate the importance of this step with the help of an example: Imagine that you conducted an experiment with newly formed virtual teams consisting of three members each. The teams completed a decision-making task communicating only via email. In total, you have data from 42 teams who wrote between 15 and 33 emails. While conducting the experiment, you already created a folder for each team that contains the emails exchanged by this specific team.

Now, how can we best handle and prepare the data? In a first step, all emails need to be reviewed to extract the most important aspects: (1) Chronological order: To create a chronological order, you need to identify the exact time each email was sent. (2) Speaker: To assign each email to a speaker, information on the sender needs to be extracted. (3) Linguistic content: The written language produced in the specific email by the sender. All this information is then transferred to a table in a chronologically correct order so that no information is lost. The most relevant aspects from the example (chronological order, speaker, linguistic content) are also available in other forms of virtual communication (e.g., chats or forum discussions). However, when working with chats or discussions in forums, the protocols are often in the correct order and only need to be transferred into a table format.

Analyzing language styles using the software linguistic inquiry and word count

Once the team interactions are prepared following the guidelines above, they are processed using LIWC. LIWC reads all words in each transcript and compares them with a built-in dictionary. The current version of the English dictionary, for example, contains more than 18,000 entries that are assigned to one or more of 70 nonexclusive categories. Function words are part of the basic linguistic processes theme in LIWC. Depending on the individual research question, researchers can choose their level of analysis (e.g., single words, statements, or complete conversations). When assessing LSM in teams, the interest is in the dynamic coordination of function words across all statements of the team interaction—therefore statement is the level of analysis for our successive LSM calculations. Traditionally, static LSM uses whole conversations as level of analysis. LIWC reports the proportion of words in the given level of analysis (i.e., a statement) that fall into the category of interest (i.e., function words). These results are then saved and used for further analyses.

Since the introduction of the 2015 LIWC version, the LIWC dictionary contains a category called function words in addition to the seven individual function word categories (pronouns, articles, prepositions, auxiliary verbs, adverbs, conjunctions, and negations). The function word category reports the overall frequency of function words used in the level of analysis, not distinguishing between the types of function words used. We acknowledge, that this category is a great addition to the dictionary and that overall function word use might be interesting to answer specific research questions, such as investigating if overall individual language style depends on the context (e.g., Müller-Frommeyer et al., 2020). However, we clearly recommend using individual function word categories for more fine-grained insights into the dynamics of LSM.

Methodological Approaches to Calculating LSM in Teams

Past and present research on LSM in teams has used one existing methodological approach to LSM. In the following paragraphs, we first present and explain the existing static approach (Language Style Matching in teams, that is, LSM-t) and continue with the introduction of a dynamic methodological approach that overcomes the shortcomings of the hitherto used method (reciprocal Language Style Matching in teams, that is, rLSM-t). A summary of both approaches comparing conceptual and methodological aspects can be found in Table 3.

Table 3.

Conceptual Comparison of LSM-t and rLSM-t.

Dimension of comparison	Static LSM-t	Dynamic rLSM-t
Construct	Overall similarity in function words between team members (i.e., similarity in language style across a whole conversation).	Dynamic matching of function words between team members (i.e., turn-by-turn matching) across the conversation.
Theoretical basis	Various, for example, interpersonal synchrony (Bernieri & Rosenthal, 1991), behavior mimicry (Chartrand & Lakin, 2013), communication accommodation theory (Shephard et al., 2001).Shared limitation: Lack of temporal component.	Interpersonal synergy (Riley et al., 2011)
Methodological approach	Difference score (represented in equations (1) and (2)) applied to summarized statements per team member.	Difference score (represented by equations (3)–(5)) applied to sequential, temporal order in the conversation.

The Status Quo: Current Methodological Approach to LSM in Teams (LSM-t)

To date, LSM in teams was analyzed by comparing each team member’s language style with the overall language style of the remaining team members (Gonzales et al., 2010). This strategy resulted in a separate LSM score for each team member. We subsequently call this strategy LSM-t.

Prior to LSM-t calculations, the transcripts are separated by team members creating one block of text containing all statements by one member. Additionally, a second block of text is created that contains all statements by the remaining team members. For example, in a team with three members, for member A, the team block is composed of the statements uttered by members B and C; for member B, it contains statements uttered by members A and C, and so on. This step is repeated for every team member in the team under investigation. By rearranging the interactions into separate blocks of text, the temporal dynamics of the interaction are erased from the interaction—this is what makes the approach static.

In the next step, blocks of texts are analyzed using LIWC, resulting in one LIWC score for each of the seven function word categories. Based on these scores, LSM is calculated for each team member individually (Gonzales et al., 2010) using the following equation (1):

L S M_{X} (C) = 1 - \frac{| C_{X} - C_{Y} |}{C_{X} + C_{Y} + 0.0001}

(1)

where X refers to the team member whose LSM-t score is calculated, C is the LIWC function word category (e.g., prepositions) we calculate the score for, C_X is the specific LIWC score for the focal team member and C_Y is the LIWC score for the remaining team members. In the denominator, the 0.0001 is added to prevent empty data sets (Ireland & Pennebaker, 2010). LSM-t can take values between 0 and 1, with 0 representing no and 1 representing perfect LSM. This calculation is repeated for each team member and each of the seven function word categories—resulting in seven LSM-t scores per team member. To calculate overall LSM-t per team member, these scores are averaged across the seven function word categories. Once one LSM score per team member is calculated, the team LSM score is calculated following equation (2):

L S M - t_{T} (F W) = \frac{\sum_{n} L S M (F W)}{n}

(2)

where T represents the team, FW represents the average across all individual function word categories, $\sum_{n} L S M (F W)$ is the sum of averaged function word LSM scores of all n members of the team under investigation and, accordingly, n is the number of team members in the team under investigation.

This analytic strategy captures similarity in function word use across the whole conversation between all members. Taking all empirical research that used this method into account (e.g., Gonzales et al., 2010; Heuer et al., 2020; Yilmaz, 2016; Yilmaz & Peña, 2015), a similarity in function word use between team members seems highly relevant as it influences the teams’ interrelatedness and effectiveness. However, this strategy neglects the temporal dynamics of LSM. As a result, the relationship between dynamic LSM and a teams’ interrelatedness and effectiveness is still unexplored.

Approaching Temporal Dynamics: Reciprocal LSM in Teams (rLSM–t)

To adequately capture the proposed conceptual dynamics of LSM in teams, we now present a new method that captures these temporal dynamics called rLSM-t. This method is based on the reciprocal LSM (rLSM) metric introduced by Müller-Frommeyer et al. (2019) which assesses dynamic LSM in dyadic interactions. The present paper adapts this metric to the team context. In contrast to the static approach, rLSM-t focusses on successive statements in conversations in their original, chronological order. Hence, LIWC analyses and rLSM-t calculations are based on the analyses of each team members’ statement in chronological order within the interaction. In the example presented in Table 2, we provide all scores calculated in the following paragraphs to facilitate the understanding of our step-by-step explanation.

We provide an R script for the calculation of rLSM-t on the osf for this paper. This R script is accompanied by exemplary data.

Step 1: Preserving the temporal dynamics of LSM

The first step in calculating rLSM–t is based on the calculation of rLSM in dyadic interactions (see Müller-Frommeyer et al., 2019). By investigating each pair of successive statements in the interaction (e.g., statement 1 and statement 2, statement 2, and statement 3), this approach preserves the temporal dynamics of the interaction.

To extract the temporal sequence of rLSM, equation (3) is applied to LIWC results of each pair of successive statements and function word category.

r L S M_{X Y} (C) = 1 - \frac{| C_{X} - C_{Y} |}{C_{X} + C_{Y} + 0.0001}

(3)

where X and Y refer to a pair of successive statements by different individuals (e.g., person X utters a statement and person Y then reacts to it by uttering another statement). Hence, Y is defined as X+1, thereby representing the temporal order of statements. C refers to the LIWC function word category used in this analysis, whereas C_X and C_Y are the LIWC scores for the specific function word category. The resulting LSM score it then attributed to the individual who uttered statement Y. See Table 2 for another example of this process. This step is repeated for each of the seven LIWC function word categories and then averaged across all seven function word categories per pair of successive statements. Once averaged, only the overall rLSM score is relevant for further analyses (unless theoretically specified otherwise). Referring to our example, LIWC function word scores for each statement can be found in column Function Words (%) in Table 2.

The rLSM-t metric is based on the static metric to assess LSM. Accordingly, scores can take values between 0 and 1 with higher values representing higher LSM. Applying equation (3) to a complete team interaction results in a temporal sequence of rLSM scores—represented by rLSM scores for each pair of successive statements. In the example provided in Table 2, this temporal sequence is displayed in column rLSM.

Following Müller-Frommeyer et al. (2019), missing values are treated as follows: Only if a specific function word category is used in the first statement, it is considered in the successive statement. If no words in the first statement or in the pair of successive statements fall into that specific function word category—represented by a LIWC score of 0—they are replaced by missing values. Including them into the analyses leads to skewed results.

Step 2: Extracting relevant information

The temporal sequence of rLSM scores created in Step 1 allows us to extract different information (potentially) relevant for further analyses: First, the temporal sequence of rLSM scores allows us to investigate how LSM (represented by rLSM) unfolds in a team over the course of an interaction. Please refer to Knight et al. (2016) or Meinecke et al. (2020) for further information on working with time-series data in teams.

Second, each score is assigned to different parties within the team. If you work with the R script provided in this paper and follow the input instructions thoroughly, each rLSM score is automatically assigned to (1) a team represented in the column Team No. in Table 2, and (2) a specific team member as shown in column Score For in Table 2. Because each of these scores is potentially relevant depending on your individual research question, we explain the calculation, function, and meaning of each with the help of an example we introduced earlier in this paper. In this exemplary study, we worked with data from 42 newly formed virtual teams who communicated via email to complete a decision-making task. The main aim of this study was to investigate the relationship between LSM and team satisfaction. In the following, we add additional information to the basic information provided to better differentiate between the respective scores.

Team score

The team score—rLSM-t—represents overall LSM in a team considering its temporal dynamics. rLSM-t is calculated following equation (4):

r L S M - t_{n} = M (r L S M_{X Y})

(4)

where n is the team identifier (i.e., team number), and X and Y are the first and the last statement in the interaction under investigation. Please note that it is also possible to calculate rLSM-t to phases of a team interaction if you are, for example, specifically interested in LSM within the first 5 minutes of an interaction or if you want to compare LSM between different phases of the interaction. The rLSM-t score is also displayed in Table 2.

rLSM-t is particularly relevant for researchers who want to compare effects between teams—basically any team researcher interested in the temporal dynamics of function word use in a team. Concerning our example, a researcher may ask whether LSM has the same assumed effect on team satisfaction across all 42 teams in the study.

Individual score

The individual score reflects each team member’s LSM to all other team members. For example, if member B’s statement follows member A’s statement, the score is assigned to member B. In contrast to Müller-Frommeyer et al. (2019), the individual score focuses on the overall adaption of one team member to all other team members rather than the adaptation of two specific members to each other. Individual scores are represented by equation (5):

r L S M - t_{X} = M (r L S M_{X K})

(5)

where X is the individual team member and K is every other team member.

Individual scores can be used to compare LSM between individual team members, for example, whether member A (leader) shows more LSM than team members B and C. Additionally, individual scores are relevant if researchers are interested in multilevel effects to test assumed effects on the individual (within teams) and team level (between groups) (e.g., Grille et al., 2015; Schulte et al., 2015).

Comparing LSM-t and rLSM-t

Building on the theoretical and methodological outline provided above, we proceed with an application of both scores—the static LSM-t and the dynamic rLSM-t score—to illustrate how the conceptual differences manifest in the calculated scores. Considering the temporal dynamics by assessing statement-based LSM across an interaction, we expect rLSM-t values to be significantly lower than LSM-t values. We propose that the lower rLSM-t is a truer estimation of LSM (Hypothesis 1). We further expect this difference to be present for the individual as well as overall team scores.

In the next sections, we first illustrate this assumption with the help of three simulated examples that highlight the subtleties in methodological differences by varying the temporal sequences of function word use. Building on this example, we proceed with an application of both scores to real-life interactions of N = 160 individuals in 26 innovation teams.

The Simulated Example: Three Scenarios of Function Word Use

In the following, we use three simulated examples of LIWC results for short conversations of a team with three members. As function words account for almost 60% of the words we utter (Pennebaker, 2013), we used function word scores between 0 and 1 rounded to a single decimal point to facilitate understanding. These scores represent the percentage of function words used in each statement by the respective speaker. All three examples are displayed in Table 4. For this comparison, we use individual and team scores of the static LSM-t score calculated following equations (1) and (2), and the rLSM-t calculated following equations (4) and (5).

Table 4.

Simulated Examples of Function Word Use to Illustrate Differences in rLSM–t and LSM-t Values.

	Function word scores
Speaker	Example 1	Example 2	Example 3
A	.6	.6	.4
B	.6	0	.3
C	.6	.6	.1
B	.6	0	.5
A	.6	.6	.4
C	.6	0	.7
LSM-t scores
LSM-t_A	1	.57	1
LSM-t_B	1	.0002	1
LSM-t_C	1	1	1
LSM-t	1	.52	1
rLSM-t scores
rLSM-t_A	1	.0002	.89
rLSM-t_B	1	.0002	.60
rLSM-t_C	1	.0002	.61
rLSM-t	1	.0002	.70

Note. M = Mean. FW = Function Words. LSM-t = score based on equation (2). rLSM–t = score based on equation 4. A, B, and C mark the respective speakers.

Example one illustrates that LSM–t and rLSM-t scores are only identical in one very unlikely case: When all team members use the same amount of function words throughout the whole conversation. This scenario results in LSM-t/rLSM-t scores of 1. However, this scenario is highly unlikely in real-life interactions as the example in Table 4 illustrates. Whenever there is a fluctuation in the use of function words over the course of an interaction—which is always the case in natural conversations—the calculation of LSM–t and rLSM-t results in significantly different scores.

In example two, function words are only used in every other statement. To calculate LSM-t scores, first, average function word use per speaker is calculated (M_A = .6, M_B = .0, M_C = .3). In a next step, for each speaker average function word use for all other team member is calculated (M_AB = .3, M_AC = .45, M_BC = .24). Then, equation (1) is applied to these average scores and results in the LSM-t scores displayed in Table 4. To calculate rLSM-t scores, first rLSM is calculated for each pair of successive function word scores—resulting in a score of rLSM = .00002 for all pairs of successive statements. Averaging these scores results in the individual and overall rLSM-t scores displayed in Table 4. These results indicate that rLSM-t captures the temporal dynamics in function word use more appropriately than LSM-t scores.

In example three, the use of function words varies randomly across team members. When calculating LSM-t this—by chance—results in identical mean scores for each team member (M_A = .4, M_B = .4, M_C = .4) which results in identical average scores for the remaining team members (M_AB = .4, M_AC = .4, M_BC = .4) leading to individual and team LSM-t scores of 1. This score indicates perfect LSM throughout the interaction. The rLSM–t metric, on the other hand, picks up on the dynamics and results in different scores for each speaker (rLSM-t_A = .89, rLSM-t_B = .60, rLSM-t_C = .61) and a significantly lower rLSM-t team score (rLSM-t = .70).

The Real-Life Example: LSM in Real-Life Innovation Teams

To empirically confirm the differences between LSM-t and rLSM-t, we additionally present an empirical example. The data for the empirical example come from a larger research project funded by the Federal Ministry of Education and Research (Germany). In this study, we used video recordings from N = 160 researchers (M_age = 30.92, SD = 6.85, 56% male) in N = 26 innovation teams from a variety of academic disciplines (35% technical, 27% science, 15% social sciences, 15% humanities, 4% computer science, 4% economics). Team size ranged from four to ten team members (M = 6, SD = 1.41). All teams were filmed during a regular team meeting by a fellow research associate of the executing department. Team meetings were held in German. After the meetings, most team members (91%) stated that the videotaped team meeting resembled a regular team meeting. All participants provided written consent to be videotaped. All procedures of the study were approved by the institutional review boards on data security and ethics.

Measures

We followed all recommendations for transcription, data preparation, and rLSM-t/LSM-t score calculation provided in this paper. Therefore, we will only summarize the measures here.

Data preparation

Because we worked with video recordings in this study, team meetings had to be transcribed prior to rLSM-t/LSM-t calculation. We followed the recommendations provided in this paper and created clean-cut speaking turns that were assigned to a specific team member.

LSM-t

Prior to LSM-t calculation, transcripts were separated by team members to create one block for each team member containing all statements the respective team member uttered in the conversation. We additionally created one block of text for each team member that included all statements uttered by the rest of the team. For example, in a team with four team members, for team member A the second block of text included all statements by team members B, C, and D. All the created blocks of text were then analyzed using LIWC. Individual and team LSM-t scores were calculated following equations (1) and (2).

rLSM-t

For rLSM-t calculations, the temporal order of the transcripts was maintained. Each transcript was analyzed using LIWC with statements as level of analysis. To do so, statements by individual speakers were separated by two presses of the enter key whereas there was no such separation within a statement. Accordingly, we chose LIWC settings that automatically recognized this formatting and calculated the proportion of each function word category for each statement in the transcript. More information on the exact preparation of transcripts for rLSM-t calculation can be found on the osf. For the comparison to LSM-t scores, we calculated rLSM-t individual and team scores following equations (3) and (4).

Data analysis

We performed all analysis using SPSS 25. First, we performed Shapiro-Wilk tests to assess normality of the data. Results can be found in Table 5 and indicated, that the assumption of normality was violated for all individual LSM-t and rLSM-t scores as well as most team LSM-t scores. To empirically test the differences between rLSM-t and LSM-t, we performed Wilcoxon signed-rank tests for individual and team scores.

Table 5.

Results of Shapiro-Wilks Tests.

Category	rLSM-t			LSM-t
Category	D	df	p	D	df	p
Individual scores
Overall score	.95	150	<.001	.44	151	<.001
Pronoun	.93	149	<.001	.42	151	<.001
Article	.97	150	.004	.39	151	<.001
Preposition	.97	149	.003	.55	151	<.001
Auxiliary verbs	.95	150	<.001	.41	151	<.001
Adverbs	.97	146	.007	.51	151	<.001
Conjunctions	.95	149	<.001	.56	151	<.001
Negations	.90	142	<.001	.66	151	<.001
Team scores
Overall score	.98	26	.56	.78	26	<.001
Pronoun	.97	26	.73	.76	26	<.001
Article	.97	26	.12	.68	26	<.001
Preposition	.94	26	.28	.87	26	.004
Auxiliary verbs	.95	26	.79	.74	26	<.001
Adverbs	.98	26	.96	.80	26	<.001
Conjunctions	.99	26	.11	.81	26	<.001
Negations	.94	26	.93	.93	26	.063

Results

Following Heuer et al. (2020), we only included team members into our analyses that had verbal contributions in the recorded team meeting (N = 151). Overall, the sample comprised 125,829 words. On average teams used 4840 words (SD = 531.37) per meeting. Each speaker used on average 833 words (SD = 919.63) per meeting.

Difference between team scores

Wilcoxon signed-rank tests revealed a significant difference between the average rLSM-t (Mdn = 0.44) and LSM-t scores (Mdn = 0.87, z = –4.46, p < .0001, r = –.87), with rLSM-t values being significantly lower than LSM-t values. These results are consistent for each of the individual function word categories. The results can be found in Table 6. All effect sizes can be interpreted as large (Cohen, 1988).

Table 6.

Results of Wilcoxon Signed-Rank Tests.

Category	rLSM-t		LSM-t		n	z	r
Category	M (SD)	Mdn	M (SD)	Mdn	n	z	r
Team scores
Overall score	0.44 (0.06)	0.44	0.85 (0.10)	0.87	26	−4.46	−.87
Pronoun	0.53 (0.08)	0.53	0.90 (0.11)	0.95	26	−4.46	−.87
Article	0.43 (0.07)	0.43	0.90 (0.11)	0.95	26	−4.46	−.87
Preposition	0.34 (0.07)	0.35	0.86 (0.12)	0.86	26	−4.46	−.87
Auxiliary verbs	0.48 (0.07)	0.49	0.90 (0.11)	0.95	26	−4.46	−.87
Adverbs	0.35 (0.07)	0.35	0.88 (0.10)	0.91	26	−4.46	−.87
Conjunctions	0.52 (0.08)	0.52	0.66 (0.07)	0.68	26	−4.46	−.87
Negations	0.19 (0.05)	0.19	0.81 (0.13)	0.84	26	−4.46	−.87
Individual scores
Overall score	0.41 (0.13)	0.43	0.81 (0.20)	0.89	150	−10.14	−.83
Pronoun	0.50 (0.18)	0.53	0.90 (0.22)	0.96	149	−9.68	−.79
Article	0.41 (0.17)	0.42	0.90 (0.21)	0.96	150	−10.28	−.84
Preposition	0.32 (0.16)	0.33	0.85 (0.24)	0.93	149	−10.47	−.86
Auxiliary verbs	0.45 (0.17)	0.46	0.90 (0.22)	0.96	150	−9.89	−.81
Adverbs	0.35 (0.15)	0.36	0.87 (0.21)	0.94	146	−10.25	−.85
Conjunctions	0.50 (0.17)	0.51	0.66 (0.16)	0.70	149	−8.35	−.68
Negations	0.18 (0.13)	0.17	0.81 (0.25)	0.89	142	−10.19	−.86

Difference between individual scores

Individual scores show the same effects as team scores—there is a significant difference between rLSM-t (Mdn = 0.43) and LSM-t (Mdn = 0.89, z = −10.14, p < .001, r = −.83), with rLSM-t values again being significantly lower than LSM-t values. Results are the same for each of the seven individual function word categories. All effect sizes can be interpreted as large (Cohen, 1988). To conclude, Hypothesis 1 was confirmed, that is, we find rLSM-t scores to be significantly lower than LSM-t scores for individual and team scores.

Discussion

Assessing team dynamics has allowed researchers important insight into teams’ functioning. In this vein, research on LSM in teams has begun to flourish in the last years suggesting that LSM is affected by team inputs like social power or personality and affects team outputs like cohesion and performance (e.g., Gonzales et al., 2010; Heuer et al., 2020; Scissors et al., 2008). However, theoretical, conceptual, and methodological approaches to date have treated LSM in teams as static, neglecting the dynamic nature of language (style) that unfolds over the course of social interactions (Müller-Frommeyer & Kauffeld, 2021). In this paper, we integrated existing temporal frameworks to build a theoretical foundation for LSM that considers temporal dynamics in function word use. Based on this dynamic theoretical foundation, we further provided an extensive conceptualization of LSM as basis for the present paper and future work. By providing a solid theoretical and conceptional foundation for LSM, we contribute to standardizing the understanding of LSM and thus strengthen the impact of LSM in team research.

Additionally, we reviewed the existing methodological approach to LSM in team research called LSM-t based on our new theoretical and conceptual foundation. By erasing the temporal order within a conversation, this methodological approach is contrary to the previously established criteria. We, therefore, proposed a new methodological approach to LSM in teams called rLSM-t that considers these temporal dynamics. When comparing LSM-t and rLSM-t with the help of three simulated examples and data from 26 real-life team interactions, we showed, that by considering the temporal dynamics the new methodological approach—rLSM-t—is a truer estimation of LSM in teams. These findings are in line with Müller-Frommeyer et al. (2019) who found similar effects for LSM in dyadic interaction when comparing static to dynamic methods. Using LSM-t results in rather high estimates of LSM in teams leading to ceiling effects for the phenomenon under investigation where all teams show a high amount of LSM. The use of rLSM-t, on the other hand, gives us a much more detailed insight into the dynamics of LSM where situational or task-related constraints might uncover differences in LSM in teams.

These results are particularly interesting considering existing results on the effects of LSM in teams. The name of the phenomenon alone—Language Style Matching—implies that we examine a process. However, all existing research on LSM in teams (e.g., Gonzales et al., 2010; Heuer et al., 2020), used the method we identified as static in this paper. If we regard the different theoretical and conceptual assumptions behind the two methodological approaches presented in this paper, the importance of a clear distinction between the two becomes apparent. We, therefore, propose the following to differentiate between both approaches: Previous studies that have chosen the static approach rather investigated a similarity of language styles, than the process of matching. Empirical evidence from those studies clearly demonstrate that this similarity of language styles is relevant to a team’s collaboration (e.g., Gonzales et al., 2010; Heuer et al., 2020; Scissors et al., 2008). However, in future studies these previous results need to be interpreted considering our new differentiation and the stream of research should be continued under the concept of similarity.

Implications for Future Research

With the introduction of rLSM-t a new field of research emerges. Applying rLSM-t in real-life team interactions leaves us with a whole set of unanswered and potentially interesting and relevant research questions. First ideas of potential avenues for future research are outlined below.

First, the differentiation of the two methodological approaches highlights a lack of knowledge on how an actual matching of language styles over the course of an interaction relates to team outcomes. Additionally, there is no empirical evidence of how this matching is influenced by (team) inputs. These questions need to be investigated in future research. A first approach would be a replication of previous studies that include both LSM-t and rLSM-t assesses their effects on the chosen outputs.

Second, we lack knowledge if high LSM across a whole process positively influences the outcomes of team interactions. We know that the repetition of identical behaviors—that is, high LSM across the whole process of a conversation—negatively affects the team (e.g., Meinecke et al., 2018; Stachowski et al., 2009). Identifying thresholds of beneficial versus non-beneficial LSM by applying rLSM-t to team interactions would further expand our understanding of this phenomenon and allow us to derive practical implications on the use of LSM.

Third, the introduction of the new method offers the possibility to consider LSM as a phenomenon in the direct observation of emergent team phenomena (Kozlowski et al., 2013). Emergent team phenomena are defined as multilevel, meaning that they comprise (at least) two different levels of analysis—the individual level where the phenomenon originates, and the team level where it manifests (Kozlowski et al., 2013). Higher-level emergence results from the dynamic social interaction of individual team members over time and arises in their affect, cognition, behavior, or individual characteristics (Kozlowski & Klein, 2000). To further advance our understanding of emergence in teams and to be able to provide work environments that support positive relationships between LSM and favorable team outcomes, the identification of behavioral patterns and adequate methods that capture the process of emergence is of importance for researchers and practitioners alike. Assessing verbal micro-behaviors like LSM could broaden our understanding of the emergence of team phenomena in more detail. In the same vein, clarifying to what extent LSM is an emergent team phenomenon itself and how different situational and environmental factors influence its emergence is of central importance to design team-work situations.

Fourth, the newly introduced rLSM-t could be used to assess temporal trends in team interactions. This comprises, for example, a comparison of LSM (and potentially related emergent phenomena) in (1) early versus late team interaction, (2) depending on a variety of contextual factors such as the task, goal, or duration of a team interaction or (3) in newly formed versus existing teams. All of these examples will eventually help us to broaden our knowledge and understanding of temporal processes in teams.

Conclusion

This paper introduced a metric that allows researchers in psychological and communication science to assess the coordination of function words (i.e., LSM) over the course of a team interaction. Building on a theoretical and conceptual revision, we provided a tutorial on how to prepare interaction data from teams for LIWC analyses. We further reviewed the existing metric to assess LSM in teams and introduced a new metric that considers the temporal dynamics of LSM, called rLSM-t. An empirical comparison of both metrics showed that rLSM-t is a truer and lower estimation of LSM in teams. Therefore, we recommend the differentiation of both metrics and the use of rLSM-t when interested in temporal dynamics in teams.

Footnotes

Author’s Note

Lena C. Müller-Frommeyer is now affiliated to RWTH Aachen University.

Simone Kauffeld is now affiliated to Technische Universität Braunschweig, Germany.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a research grant from the German Federal Ministry of Education and Research (BMBF; No. 16FWN005, 2013-2019).

ORCID iD

Lena C. Müller-Frommeyer

Author Biographies

Lena C. Müller-Frommeyer, PhD, is a postdoctoral research associate at the RWTH Aachen University. She received her PhD from Technische Universität Braunschweig, Germany. Her research areas include the dynamic coordination of behavior such as natural language, communication, or emotion in different forms and contexts of individual, dyadic, and team interaction.

Simone Kauffeld is full professor at the Department of Industrial/Organizational and Social Psychology, Technische Universität Braunschweig, Germany. She received her PhD from the University of Kassel and held a professorship at the University of Applied Sciences, Northwestern Switzerland. Her research focuses on team and leadership interaction, career, competence, and change.

References

Aguinis

Vandenberg

R. J.

(2014). An ounce of prevention is worth a pound of cure: Improving research quality before data collection. Annual Review of Organizational Psychology and Organizational Behavior, 1(1), 569–595. https://doi.org/10.1146/annurev-orgpsych-031413-091231

Arafat

S. M. Y.

Kar

S. K.

Menon

Kaliamoorthy

Mukherjee

Alradie-Mohamed

Sharma

Marthoenis

Kabir

(2020). Panic buying: An insight from the content analysis of media reports during COVID-19 pandemic. Neurology Psychiatry and Brain Research, 37, 100–103. https://doi.org/10.1016/j.npbr.2020.07.002

Bernieri

F. J.

Rosenthal

(1991). Interpersonal coordination: Behavior matching and interactional synchrony. In Feldman

R. S.

Rimé

(Eds.), Fundamentals of nonverbal behavior: Studies in emotion & social interaction (pp. 401–432). Cambridge University Press.

Biesen

J. N.

Schooler

D. E.

Smith

D. A.

(2016). What a difference a pronoun makes: I/We versus you/me and worried couples’ perceptions of their interaction quality. Journal of Language and Social Psychology, 35(2), 180–205. https://doi.org/10.1177/0261927x15583114

Brennan

S. E.

Hanna

J. E.

(2009). Partner-specific adaptation in dialog. Topics in Cognitive Science, 1(2), 274–291. https://doi.org/10.1111/j.1756-8765.2009.01019.x

Cannava

Bodie

G. D.

(2017). Language use and style matching in supportive conversations between strangers and friends. Journal of Social and Personal Relationships, 34(4), 467–485. https://doi.org/10.1177/0265407516641222

Carmody

P. C.

Mateo

J. C.

Bowers

McCloskey

M. J.

(2017). Linguistic coordination as an unobtrusive, dynamic indicator of rapport, prosocial team processes, and performance in team communication. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 61(1), 140–144. https://doi.org/10.1177/1541931213601518

Chartrand

T. L.

Lakin

J. L.

(2013). The antecedents and consequences of human behavioral mimicry. Annual Review of Psychology, 64(1), 285–308. https://doi.org/10.1146/annurev-psych-113011-143754

Chung

Pennebaker

(2013). The psychological functions of function words. In Fiedler

(Ed.), Social communication (pp. 343–359). Psychology Press.

10.

Cohen

(1988). Statistical power analysis for the behavioral sciences. Erlbaum.

11.

Fusaroli

Rączaszek-Leonardi

Tylén

(2014). Dialog as interpersonal synergy. New Ideas in Psychology, 32, 147–157. https://doi.org/10.1016/j.newideapsych.2013.03.005

12.

Garrod

Pickering

M. J.

(2009). Joint action, interactive alignment, and dialog. Topics in Cognitive Science, 1(2), 292–304. https://doi.org/10.1111/j.1756-8765.2009.01020.x

13.

Giles

Coupland

(1991). Language contexts and consequences. Open University Press.

14.

Gonzales

A. L.

Hancock

J. T.

Pennebaker

J. W.

(2010). Language style matching as a predictor of social dynamics in small groups. Communication Research, 37(1), 3–19. https://doi.org/10.1177/0093650209351468

15.

Grille

Schulte

E.-M.

Kauffeld

(2015). Promoting shared leadership: A multilevel analysis investigating the role of prototypical team leader behavior, psychological empowerment, and fair rewards. Journal of Leadership & Organizational Studies, 22(3), 324–339. https://doi.org/10.1177/1548051815570039

16.

Heuer

Müller-Frommeyer

L. C.

Kauffeld

(2020). Language matters: The double-edged role of linguistic style matching in work groups. Small Group Research, 51(2), 208–228. https://doi.org/10.1177/1046496419874498

17.

Ilgen

D. R.

Hollenbeck

J. R.

Johnson

Jundt

(2005). Teams in organizations: From input-process-output models to IMOI models. Annual Review of Psychology, 56, 517–543. https://doi.org/10.1146/annurev.psych.56.091103.070250

18.

Ireland

M. E.

Pennebaker

J. W.

(2010). Language style matching in writing: Synchrony in essays, correspondence, and poetry. Journal of Personality and Social Psychology, 99(3), 549–571. https://doi.org/10.1037/a0020386

19.

Karl

K. A.

Peluchette

J. V.

Aghakhani

(2021). Virtual work meetings during the COVID-19 pandemic: The good, bad, and ugly. Small Group Research. Advance online publication. https://doi.org/10.1177/10464964211015286

20.

Kauffeld

Lehmann-Willenbrock

(2012). Meetings matter: Effects of team meetings on team and organizational success. Small Group Research, 43(2), 130–158. https://doi.org/10.1177/1046496411429599

21.

Knight

A. P.

Kennedy

D. M.

McComb

S. A.

(2016). Using recurrence analysis to examine group dynamics. Group Dynamics Theory Research and Practice, 20(3), 223–241. https://doi.org/10.1037/gdn0000046

22.

Kozlowski

S. W. J.

(2015). Advancing research on team process dynamics: Theoretical, methodological, and measurement considerations. Organizational Psychology Review, 5(4), 270–299. https://doi.org/10.1177/2041386614533586

23.

Kozlowski

S. W. J.

Chao

G. T.

Grand

J. A.

Braun

M. T.

Kuljanin

(2013). Advancing multilevel research design: Capturing the dynamics of emergence. Organizational Research Methods, 16(4), 581–615. https://doi.org/10.1177/1094428113493119

24.

Kozlowski

S. W. J.

Klein

K. J.

(2000). A multilevel approach to theory and research in organizations: Contextual, temporal, and emergent processes. In Klein

K. J.

Kozlowski

S. W. J.

(Eds.), Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions (pp. 3–90). Jossey Bass.

25.

Lehmann-Willenbrock

Allen

J. A.

(2018). Modeling temporal interaction dynamics in organizational settings. Journal of Business and Psychology, 33(3), 325–344. https://doi.org/10.1007/s10869-017-9506-9

26.

Lehmann-Willenbrock

Meyers

R. A.

Kauffeld

Neininger

Henschel

(2011). Verbal interaction sequences and group mood: Exploring the role of team planning communication. Small Group Research, 42(6), 639–668. https://doi.org/10.1177/1046496411398397

27.

Luciano

M. M.

Mathieu

J. E.

Park

Tannenbaum

S. I.

(2018). A fitting approach to construct and measurement alignment: The role of big data in advancing dynamic theories. Organizational Research Methods, 21(3), 592–632. https://doi.org/10.1177/1094428117728372

28.

Meinecke

A. L.

Handke

Müller-Frommeyer

Kauffeld

(2018, July). Content and structural recurrence: Do stable behavioral patterns lead to satisfied teams? [Paper presentation]. 13th Annual INGRoup Conference, Washington, DC.

29.

Meinecke

A. L.

Handke

Mueller-Frommeyer

L. C.

Kauffeld

(2020). Capturing non-linear temporally embedded processes in organizations using recurrence quantification analysis. European Journal of Work and Organizational Psychology, 29, 483–500. https://doi.org/10.1080/1359432x.2019.1658624

30.

Menenti

Pickering

M. J.

Garrod

S. C.

(2012). Toward a neural basis of interactive alignment in conversation. Frontiers in Human Neuroscience, 6, 185. https://doi.org/10.3389/fnhum.2012.00185

31.

Muir

Joinson

Cotterill

Dewdney

(2016). Characterizing the linguistic chameleon: Personal and social correlates of linguistic style accommodation. Human Communication Research, 42(3), 462–484. https://doi.org/10.1111/hcre.12083

32.

Müller-Frommeyer

L. C.

Frommeyer

N. A. M.

Kauffeld

(2019). Introducing rLSM: An integrated metric assessing temporal reciprocity in language style matching. Behavior Research Methods, 51(3), 1343–1359. https://doi.org/10.3758/s13428-018-1078-8

33.

Müller-Frommeyer

L. C.

Kauffeld

(2021). Gaining insights into organizational communication dynamics through the analysis of implicit and explicit communication. Gruppe. Interaktion. Organisation. Zeitschrift für Angewandte Organisationspsychologie (GIO), 52(1), 173–183. https://doi.org/10.1007/s11612-021-00559-9

34.

Müller-Frommeyer

L. C.

Kauffeld

Paxton

(2020). Beyond consistency: Contextual dependency of language style in monolog and conversation. Cognitive Science, 44(4), e12834. https://doi.org/10.1111/cogs.12834

35.

Niederhoffer

K. G.

Pennebaker

J. W.

(2002). Linguistic style matching in social interaction. Journal of Language and Social Psychology, 21(4), 337–360. https://doi.org/10.1177/026192702237953

36.

Pennebaker

J. W.

(2013). The secret life of pronouns: What our words say about us. Bloomsbury Press.

37.

Pennebaker

J. W.

Boyd

R. L.

Jordan

Blackburn

(2015). The development and psychometric properties of LIWC2015. University of Texas at Austin. https://repositories.lib.utexas.edu/handle/2152/31333

38.

Pennebaker

J. W.

King

L. A.

(1999). Linguistic styles: Language use as an individual difference. Journal of Personality and Social Psychology, 77(6), 1296–1312. https://doi.org/10.1037/0022-3514.77.6.1296

39.

Pennebaker

J. W.

Mehl

M. R.

Niederhoffer

K. G.

(2003). Psychological aspects of natural language use: Our words, our selves. Annual Review of Psychology, 54, 547–577. https://doi.org/10.1146/annurev.psych.54.101601.145041

40.

Podsakoff

P. M.

MacKenzie

S. B.

Podsakoff

N. P.

(2016). Recommendations for creating better concept definitions in the organizational, behavioral, and social sciences. Organizational Research Methods, 19(2), 159–203. https://doi.org/10.1177/1094428115624965

41.

Riley

M. A.

Richardson

M. J.

Shockley

Ramenzoni

V. C.

(2011). Interpersonal synergies. Frontiers in Psychology, 2, 38. https://doi.org/10.3389/fpsyg.2011.00038

42.

Schulte

E.-M.

Lehmann-Willenbrock

Kauffeld

(2015). Treat us fairly and we won’t complain: Multilevel effects of procedural justice on complaining behavior in team meetings. Psychology, 6(14), 1795–1810. https://doi.org/10.4236/psych.2015.614176

43.

Scissors

L. E.

Gill

A. J.

Gergle

(2008, November 8–12). Linguistic mimicry and trust in text-based CMC. In Begole

(Ed.), Proceedings of the 2008 ACM conference on computer supported cooperative work, San Diego, CA. ACM. https://doi.org/10.1145/1460563.1460608

44.

Segalowitz

S. J.

Lane

(2004). Perceptual fluency and lexical access for function versus content words. Behavioral and Brain Sciences, 27(2), 307–308. https://doi.org/10.1017/S0140525X04310071

45.

Shephard

Giles

LePoire

(2001). Communication accommodation theory. In Robinson

W. P.

Giles

(Eds.), The new handbook of language and social psychology (pp. 33–56). Wiley.

46.

Stachowski

A. A.

Kaplan

S. A.

Waller

M. J.

(2009). The benefits of flexible team interaction during crises. Journal of Applied Psychology, 94(6), 1536–1543. https://doi.org/10.1037/a0016903

47.

Van Gelderen

. (2014). A history of the English language (rev ed.). John Benjamins.

48.

Van Swol

L. M.

Carlson

C. L

. (2017). Language use and influence among minority, majority, and homogeneous group members. Communication Research, 44(4), 512–529. https://doi.org/10.1177/0093650215570658

49.

Van Swol

L. M.

Kane

A. A

. (2019). Language and group processes: An integrative, interdisciplinary review. Small Group Research, 50(1), 3–38. https://doi.org/10.1177/1046496418785019

50.

Waller

M. J.

Okhuysen

G. A.

Saghafian

(2016). Conceptualizing emergent states: A strategy to advance the study of group dynamics. The Academy of Management Annals, 10(1), 561–598. https://doi.org/10.5465/19416520.2016.1120958

51.

Yilmaz

(2016). What you do and how you speak matter: Behavioral and linguistic determinants of performance in virtual teams. Journal of Language and Social Psychology, 35(1), 76–97. https://doi.org/10.1177/0261927x15575772

52.

Yilmaz

Peña

(2015). How do interpersonal behaviors and social categories affect language use? The case of virtual teams. Communication Quarterly, 63(4), 427–443. https://doi.org/10.1080/01463373.2015.1058285