You are on page 1of 17

Group Dynamics:

Theory, Research, and Practice


© 2023 American Psychological Association 2023, Vol. 27, No. 3, 171–187
ISSN: 1089-2699 https://doi.org/10.1037/gdn0000199

Quantifying Synchronization in Groups With Three or More Members


Using SyncCalc: The Driver-Empath Model of Group Dynamics

Stephen J. Guastello and Anthony F. Peressini


Department of Psychology, Marquette University
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Objectives: This article introduces a nonlinear dynamical systems model to quantify


synchronization among group members with regard to physiological activity or overt
behaviors. Method: The driver-empath model accommodates asymmetries in influence
among group members, separates autocorrelation effects from synchronization with other
group members in time series analysis, accommodates dynamics that could be more
complex than simple oscillators, and produces metrics at the individual, dyadic, and group
levels of analysis. Results: An illustrative example of a team of seven undergraduates in a
group decision-making task is presented with commentary for data preparation and
determining lag length of time series data. The supporting literature summarizes connec-
tions between the group-level synchronization coefficient and group-level workload,
cooperation versus competition dynamics, leadership emergence, and team performance.
Conclusions: The driver-empath model has accumulated substantial external validity for
answering research questions in group dynamics. Suggested new applications include
studying synchrony across different network configurations and possible social fault lines.

Highlights and Implications


• The driver-empath model computes an overall measure of synchrony
that is centered on the most socially receptive person in the group and is applicable
to groups of three to 16 individuals.
• The model can be used with physiological or overt behavioral observations
in time series.
• The model has demonstrated substantial external validity for solving
problems in group-level workload, cooperation versus competition dynamics,
leadership emergence, and team performance.

Keywords: Synchrony, self-organization, nonlinear dynamics, team performance,


autonomic arousal

Synchrony in group dynamics can be observed synchronized swimming, rowing, military march-
as a close mimicry in the behavior of human ing, or many of the performing arts. Synchrony of
individuals, such as their body movements, auto- physical movements such as postural sway, hand
nomic arousal, and electroencelphalogram activ- and facial expressions, or speech patterns can
ity. It is often regarded as a special case of team occur spontaneously when mutual mimicry of
coordination when a situation requires physical behavior is not a performance objective, or
movements that are exactly timed, for example, in when two people are engaged in a conversation.

This article was published Online First March 2, 2023. conflicts of interest to report.
Stephen J. Guastello https://orcid.org/0000-0003- Correspondence concerning this article should be ad-
1059-7341 dressed to Stephen J. Guastello, Department of Psychology,
The production of this article was not funded by any Marquette University, P.O. Box 1881, Milwaukee, WI 53201-
public or private agency. The authors have no financial 1881, United States. Email: Stephen.guastello@marquette.edu

171
172 GUASTELLO AND PERESSINI

The majority of studies on synchrony in individuals. The models lend themselves to


human relationships have been based on dyads hierarchical linear modeling to study group-to-
rather than larger groups (Kazi et al., 2021). individual effects and infer causal relations
Occasionally triads were analyzed as three pos- from an otherwise cross-sectional and time-
sible dyads (Jiang et al., 2015; Mønster et al., static data set.
2016). The driver-empath model of group syn- Constellations of group behavior and internal
chrony (Guastello & Peressini, 2017) was alliances, however, are emergent phenomena.
developed as a group-level metric of group or They begin with pairwise interactions among
team synchrony that avoids the potential limita- group members, and after a period of time, they
tions of alternative techniques. This article is a self-organize into patterns that are adopted by the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tutorial on the theoretical origins of the driver- members. The consolidated patterns then have a
This document is copyrighted by the American Psychological Association or one of its allied publishers.

empath model with a computational illustration top-down influence on the current members and
involving seven group members and their any new members who should join the group
autonomic arousal levels. (Kozlowski, 2015; Sawyer, 2005). Thus the static
The following sections touch briefly on some view of the person–group interaction combined
problems in group dynamics and team perfor- with its top-down perspective is not sufficient to
mance that could be resolved through the syn- illustrate or explain how dyadic alliances form,
chronization paradigm. They are followed by whether or not they stabilize as group-level re-
an exposition of the driver-empath model and gimes, or how the group as a whole reacts to
its assumptions, the SyncCalc program, and a changing environmental demands. Research ques-
supporting program in R script that processes data tions about emergent processes could benefit from
into the format used by the SyncCalc program finer-grained time series analysis.
(Peressini & Guastello, 2016, 2022). The illus- Researchers studying group actor–partner
tration is followed by an overview of the substan- interdependence have recognized the need for
tive contributions of the driver-empath model an expanded use of time series analysis. In one
to the understanding of group dynamics and illustrative application (Gillis et al., 2016),
possible new directions for further research. groups of 10 young adult males participated in
a wilderness camp therapy. Those who held more
positive impressions of the group after initial
Approaches to Group Dynamics contact were hypothesized to be more likely to
engage in more therapeutic sessions and thus
Group Actor–Partner Interdependence show greater improvement in their conditions.
Results indicated, however, that those who were
The most common approach to group dynam- more initially positive about the group were not
ics, group actor–partner interdependence models so engaged, and those who were more engaged
(Kenny & Cook, 1999), expresses the process by did not show much improvement. The research-
which one member of a dyad influences the other. ers recommended that actor–partner interdepen-
The concept produced numerous investigations dence models be organized as time series analyses
at the dyadic and group levels of analysis in order to capture time points where therapists
(Keum, 2018; Kivlighan, 2021). The dyadic mod- could redirect individuals during the course of
els consist of as a set of attributes of the target therapy.
person and a set of attributes for the partner that In another example, Kivlighan et al. (2022)
can interact and then converge to influence beha- studied self-reports of alliances with the group
viors and outcomes for the target person. Neuroti- leader among adult psychiatric participants in
cism, anxiety, conflict, and positive attitudes two types of group therapy. They investigated
are examples of attributes of the target person or the specific effects of transient (“state-like”)
partner that could influence outcomes. alliances that might be observed in particular
Group actor–partner interdependence models sessions and (“trait-like”) alliances that persist
are organized the same way and express the across sessions. The primary result was that
influence of the group on each individual members with stronger persistent alliances
(Kenny & Garcia, 2012). The nuance is that made better progress in therapy to the extent
the partner attributes are the average of measures that other group members also exhibited stronger
from all group members minus that of the target persistent alliances (p. 147). By all appearances,
DRIVER-EMPATH MODEL 173

emotional synchrony was beneficial to therapeu- several phenomena such as emotional contagion
tic outcomes. The study also drew attention to (Hatfield et al., 1993, 2009), empathy, common
the heterogeneity of alliances that could also focus of attention, (Guastello, Mirabito, &
exist within a group. Peressini, 2020; Palumbo et al., 2017; Stevens
Kozlowski (2015) made similar observations et al., 2014), temporal regularities in the activity
about the need for finer-grained time series when or environment that correspond to repetitious
studying task groups in work organizations. He action, turn-taking among the group members
suggested using biometric measures of auto- (Guastello et al., 2017; Henning et al., 2001),
nomic arousal as supplements to nonobtrusive or external timing cues (Torrents et al., 2016).
behavioral measurements as a means of studying Thus the limitations of most uses of group
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

emergent processes. The next section on auto- actor-partner interaction models modeling—its
This document is copyrighted by the American Psychological Association or one of its allied publishers.

nomic arousal, which is central to our approach, time-static and top-down framework—can be
captures the current role of biometrics in work- overcome by adopting a research strategy that
related contexts. incorporates autonomic signal interaction and
synchrony to understand the dynamics of mutual
influence in groups. The next sections explain
Autonomic Signal Interaction and Synchrony principles of synchrony that are applicable to
in Groups many types of systems. Alternative models for
operationalizing synchrony into group dynamics
The interaction (and synchrony) of autonomic research are considered next along with problems
signals has received attention in human factors associated with them. We then introduce the
engineering, which is usually focused on person– driver-empath model. Which uses autonomic
machine relationships. It developed a concern time series data produce a group-level measure
for group dynamics now that multiple people of group synchrony along with some individual
and machines can become connected into groups and dyadic metrics. Inasmuch as the applications
and networks (Guastello, in press). A parallel of the driver-empath model to date utilized ED
development is the increased use of physiological data as the basis of synchrony, special attention is
indicators of cognitive workload (Gawron, 2019; given to the preparation of ED data for analysis by
Matthews & Reinerman-Jones, 2017) and to mon- SyncCalc. The current status of the synchrony–
itor other aspects of cognitive processing and group performance relationship is detailed later in this
dynamics when groups are performing demanding article. We conclude with some speculations
tasks, perhaps during an emergency (Gorman et al., regarding how lessons learned thus far could be
2020). Cognitive workload and social interaction applied to group therapy considerations.
produce arousal, thus in principle autonomic sig-
nals contain information about both workload Nonlinear Dynamics and Synchrony
and group interactions. Studies of autonomic syn-
chrony (e.g., electrodermal [ED] response, heart The nonlinear dynamical systems theory of
rate, heart rate variability, cortical activity) and synchronization characterizes it as “an adjust-
other neurodynamic patterns are part of a broader ment of the rhythms of oscillating objects due
movement to gain an integrated understanding of to their weak interaction” (Pikovsky et al., 2001,
connections between behavioral outcomes and p. 8). Synchrony is a special case of self-organi-
collective neurocognitive activity (Gorman et al., zation: Systems in a high state of entropy tend
2020; Salas et al., 2015). to adopt a new structure that reduces the amount
The connection between team synchrony of internal entropy and improves the performance
and autonomic arousal was propelled in part by and efficiency of both living and nonliving
early reports of stronger autonomic synchrony systems (Haken, 1984; Prigogine & Stengers,
connected to better team performance in dyads 1984). Strogatz (2003) described the minimum
(Henning et al., 2001) and larger groups (Elkins requirements for synchronization as (a) two cou-
et al., 2009). In the current state of the science, pled oscillators, (b) a feedback loop between
synchrony of autonomic arousal can occur when them, and (c) a control parameter that speeds up
two or more people engage a joint activity or the oscillating process. When the oscillation speed
conversation (Guastello et al., 2006; Kazi et al., reaches a critical point, there is a sudden change
2021; Palumbo et al., 2017) through any of in system behavior, called phase lock, in which
174 GUASTELLO AND PERESSINI

mutual mimicry is observed with minimal devia- It is not uncommon for Person A to have an
tions in timing among the contributing entities. impact on Person B that is not reciprocated. As
The legendary case of Christiaan Huygens’ an example, Levenson and Gottman (1983) col-
(1,629–1,695) two pendulum clocks hanging lected ED data during structured exercises with
from a wooden beam and eventually coming to married couples in therapy. Couples exhibiting
oscillate in tandem is an example (Strogatz, 2003). two-way influence stood a better chance of
More complicated chaotic systems, such as neu- staying married than couples in which A is reac-
rological circuits, can also exhibit synchronization tive to B, but B is indifferent to A. In therapist–
(Pikovsky et al., 2001; Stefański, 2009; Whittle, client interactions, therapy sessions in which
2010), although it is less likely that they would there was a higher synchrony of body movements
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

exhibit the distinctive phase lock associated with between the client and therapist produced more
This document is copyrighted by the American Psychological Association or one of its allied publishers.

simple oscillators. Synchronization levels thus satisfactory outcomes (Ramseyer & Tschacher,
range between very tight mimicry (phase lock) 2016), but only when the client was following the
at one extreme and none at all at the other. therapist and not vice-versa (Ramseyer, 2020).
Rather than assuming that the contributing Close mimicry is only one possible expression
time series of measurements are oscillators, the of synchrony. Mimicry in emotional or arousal
driver-empath model starts with the assumption perspectives would take the form of one
that the time series originate from low- person’s arousal level positively affecting
dimensional chaotic systems. Low-dimensional another person, and the positive response is
chaos would be evident in time series data re- mutual. “They’re feeding off each other’s excite-
flecting a low level of deterministic turbulence ment.” The cross-correlation of time series can
and a fractal dimension ranging between 1.0 also be negative–positive, such that Person A
and 2.0, which is also known as the range of induces arousal in Person B, but B has a damp-
self-organized criticality (Bak, 1996). When ening effect on A (Guastello, 2016). Interactions
systems self-organize, they exhibit greater can also be negative–negative where both parties
patterning of behavioral schemata and less are keeping each other calm.
volatile or random-appearing behavior. Self-
organization is reflected in the formation of a Alternative Group Synchrony Models
social structure within the group that is influ-
enced by the demands of the task, application Three other strategies for quantifying group
of skills and tools, physical boundary conditions, synchrony should be given brief mention. The
and the particular individuals present. Impor- phase clustering method (Elkins et al., 2009;
tantly, three coupled oscillators are sufficient Richardson et al., 2012) assumes that the data
to produce chaos (Newhouse et al., 1978). generators are oscillators that would be in-phase
Thus, even if oscillators were good characteriza- under fully synchronized conditions. Deviations
tions of each individual’s behavior, they produce from synchrony is the amount of disparity
chaotic group behavior when they become cou- within each pair of time series (members of a
pled and self-organize further. group) summed over all pairs of time series.
The driver-empath model reflects two further Phase clustering, however, does not accommo-
criteria for the development of the group-level date asymmetries in influence among group
synchrony coefficient: the separation of the auto- members, nor does it separate autocorrelation
correlational and cross-correlational effects and from synchronization with others (transfer
asymmetry in influence among group members. effects). The assumption of a simple oscillator
In the former case, two time series can appear structure is not realistic in many situations.
correlated because the agents that they represent Phase clustering also ignores the possibility
are following the same path or rule—or watching that two time series can be synchronized and
the same movie in two separate theaters—and not negatively correlated; those seesaw effects are
because they are exerting any influence on each computed as dyads out of phase with each other.
other. A time series regression analysis would need The second idea that was also rationalized
to show that the second time series had an impact from phase space analysis uses the velocity of
on the first over and above the autocorrelation of movement or measurement change as the pri-
the first (Granger, 1969) in order to conclude that mary indicator of synchrony within dyads
the cross-correlation was real and not illusory. (Helm et al., 2012); synchronized dyads should
DRIVER-EMPATH MODEL 175

change at the same rate. Dyadic synchrony is 2021; Gordon et al., 2021), computes the average
measured as the multiple correlation associated amount of synchrony for each person with all
with Granger causality regression models in others in the group. Thus for a triad, three group
Equations 1–2: measures are produced and all three would be
used as predictors of social outcomes of interest.
X̂ n = β0 + β1 X n−j , (1) As such it does not produce a group-level metric
unless one computes the average of personal
averages; MdRQA also does not separate auto-
X̂ n = β0 + β1 X n−j + β2 Pn−j : (2)
correlation from transfer effects or produce
directional relationships.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

where X is the time series for the target person,


This document is copyrighted by the American Psychological Association or one of its allied publishers.

and P is the influence from the time series of a


second person to the target person. Driver-Empath Model
The multivariate extension of this technique
(Helm et al., 2018) was framed around dyads Although synchronization phenomena have
with multiple physiological measures each. not appeared in the interdependence models,
Instead of compiling the dyadic relationships their focus on the group’s effect on individuals
into a group-level metric, the procedure engages plays a role in the driver-empath model of
multilevel modeling to investigate the patterns of synchrony. The computation for the driver-
correlation between intrapersonal and interper- empath model begins with the matrix P shown
sonal patterns of correlation. Thus the method in Table 1, which is populated with autocorrela-
does indeed separate autocorrelation from trans- tions of physiological time series on the diago-
fer effects and produces directional relationships, nal. Transfer coefficients on the off-diagonal
but it does not produce a group synchrony metric. entries show the impact of one person on each
The third alternative is based on recursion of the others. P can accept positive or negative
quantification analysis (RQA), which is a graphic coefficients for matrix entries. Negative auto-
display of potentially chaotic time series. It correlations indicate oscillating functions.
detects patterns in which a value in a continuous Negative off-diagonal elements reflect dampen-
measurement falls within a small radius of itself ing effects between people rather than mutually
later in time. Recurrences are counted irrespec- exciting effects.
tive of the number of time intervals transpiring Sums of squared coefficients of the rows and
between the two points. RQA is accompanied by columns are calculated to produce driver scores
several metrics that can be used to distinguish and empath scores respectively. The person with
the patterns, such as percent recurrence and the largest row total is the driver of the group. The
percent determinism (Aks, 2011; Webber & person with the largest column total is the empath.
Zbilut, 2007). Cross-RQA performs the same In our reasoning, a driver has little or no impact
operations, but looks for repetition of values of on group synchronization, either psychologically
one time series to show up on the second or statistically, unless other group members
series (Marwan & Kurths, 2003). The multivari- are responding. From a broader systems perspec-
ate extension for groups, (MdRQA; Coco et al., tive, agents must have sensors for incoming

Table 1
Prototype Matrix P of Synchronization Coefficients for a Team With Four Members

To
From P1 P2 P3 … Pn Driver score
P 2
P1 AR1 R12 R13 … R1n
PR1i
P2 R21 AR2 R23 … R2n 2
PR2i
P3 R31 R32 AR3 … R3n R23i
… … … … … P…2
Pn Rn1 Rn2 Rn3 …PARn Rni
P 2 P 2 P 2
Empath score Ri1 Ri2 Ri3 … R2in

Note. AR = autoregression.
176 GUASTELLO AND PERESSINI

information or else the information does not trans- in time series; dependent error is non-IID error.
mit (Sulis, 2016). The connections between team Dependent error variance arises from autocorre-
members and the empath reflects the strongest lation of residuals, complex lag functions, corre-
synchronization of any one person with the others. lations of residuals with true scores, and
Once the empath in the group has been identi- sensitivity to initial conditions. In the latter
fied, the row and column for the empath are case, an external shock could enter the system
then separated from the matrix for further calcu- and induce an IID error at one point in time, but
lation. The last calculation is a single number, when the deterministic function iterates over time,
synchronization with the empath (SE), which the error also carries through and compounds over
expresses a global level of synchronization for time and becomes non-IID error.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the group. The structure of the nonlinear process


This document is copyrighted by the American Psychological Association or one of its allied publishers.

The autocorrelations (rii in Table 1) and the remains unknown, of course, until specific non-
transfer coefficients (rij) are produced by ordinary linear models are explicitly tested. A well-chosen
least squares (OLS) linear regression with nonlinear model would capture the dependent
Equations 1–2. The diagonal values would be error as part of the function. In other words,
the β1 in Equations 1–2. The off-diagonal va- what was previously regarded as non-IID error
lues would be β2 in Equation 2. is actually subsumed as part of the deterministic
There are several reasons why the linear nonlinear function (Guastello & Gregson, 2011).
dyadic model would be a good fit. The first is Thus the linear model potentially “cheats” by
topological: Two observations that are made including nonlinearities of an unclassified
close together in time would be efficiently nature (Guastello & Mirabito, 2018; Guastello
connected by a line. It is only when enough points & Peressini, 2021a). This is not to say, however,
are collected that it is possible to see a global that an otherwise well-chosen nonlinear model
nonlinear trend (Wiggins, 1988). If there is captures all the sources of non-IID error. We
enough lag time between successive pairs of return to potentially useful nonlinear dynamic
observations, a nonlinear model could produce models later in this article.
a more accurate result than a linear model Moving forward, the first step after composing
(Guastello & Gregson, 2011). The second is matrix P is to identify the empath, which is the
that Equation 1 can produce globally nonlinear person with the largest sum of squared column
trends in the form of exponential expansion entries. Second, the column of coefficients for
or exponential decay. the empath is removed from P, the empath’s
Third, mathematical simulation studies by autocorrelation is dropped, and the remaining
Stefański (2009) showed that linear models are column coefficients become a column vector V′.
sufficient for characterizing pairwise levels of Third, the empath’s row is also removed, leaving
synchrony to the extent that the agents are a square matrix M. Thus if the original group
operating under homogeneous conditions and consisted of four people, M is a 3 × 3 matrix when
the time intervals between observations are the empath is removed.
relatively short. If two chaotic generators are Fourth, take the inverse of M and calculate a
highly synchronized, the relationship between vector of weights:
them is likely to be linear because no further
translational function would be needed to predict Q = M−1 V′: (3)
one from the other. Thus the linear relationships
between two time series can be sufficient when The elements (weights) of Q are comparable in
the synchronizing functions are chaotic and both meaning to standardized regression weights
time series follow the same deterministic rule. where the empath is a (human) dependent mea-
The fourth reason is statistical: Because the sure, and the other members of the group are
linear regression models in Equations 1–2 are independent variables. The fifth and final step
OLS models, they capture unspecified amounts produces the synchronization coefficient:
of dependent error variance, which indicates that
a nonlinear process is occurring (Guastello & SE = V′Q: (4)
Gregson, 2011; Ozaki, 2012). Ordinary psycho-
metric error is known as independently and iden- Note that matrices P and M do not contain 1.0
tically distributed (IID) error when it is observed in the diagonal entries, and neither is symmetric
DRIVER-EMPATH MODEL 177

as would be the case with a standard correlation 2003). The activity was divided into four parts:
matrix. Thus the theoretical range of values for (a) Participants watched a video (5 min) that
SE is not confined between 0 and 1. Its population described the circumstances of their airplane
mean for realistic combinations of diagonal crashing in a desolate place in northern Canada.
and off-diagonal values is close to the theoretical (b) They were given a list of 15 objects that were
mean of 0.00. The actual mean from the simula- retrieved from the crash, which they were asked
tions in Guastello and Peressini (2017) is 0.112 to rank order in terms of their value for survival.
(SD = 6.968; standard error = 0.650; 95% CI (c) The group discussed their ranking and pro-
[−1.177, 1.402]). SE is approximately normally duced a new set of rankings based on whatever
distributed and does not have bounded upper or decision-making strategy the group adopted. (d)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

lower limits. They watched the remainder of the video in which


This document is copyrighted by the American Psychological Association or one of its allied publishers.

SE is not affected by matrix size. Thus if an expert explained the optimal values for the
a research result shows differences in SE based 15 objects.
on group size, the effect is a true positive and The participants wore ED sensors on their
not an artifact of the algorithm. Thus no special nondominant hands during the first three parts
adjustments in the computations are needed to of the study. One objective of the study was to
accommodate different group sizes. determine whether SE varied by part of the study:
In summary, the driver-empath model sepa- watching a video together in the same room,
rates autocorrelations from transfer effects, the individual activity in which everyone re-
accommodates asymmetric relationships as mained in the same room together, and the group
well as reciprocal or dampening effects between activity. The second objective was to determine
group members, and produces a group-level the connection between synchrony and team
metric. Thus it circumvents the problems associ- performance. Team performance was the degree
ated with the alternative models. Furthermore, if to which the group’s rankings matched those of
the time series are strongly oscillating, Equations the expert.
1–2 can still be used with a lag time equal to Each time series lasted approximately 5 min.
the cycle length. Contrary to assertions by Some preprocessing of data was required before
Gordon et al. (2021, p. 11), SE is not calculated analysis to determine the optimal lag length
as an average of dyadic relationships. It is calcu- and down-sampling rate, which turned out to
lated relative to the person with the strongest be a down-sampling rate of three observations
receptivity in the group—the empath, not the per second and a lag of two third-second units. We
driver. return to questions about down-sampling and
lag length later in this article.
Illustrative Example The numerical example that follows came from
a group of seven participants during the group
The following example was drawn from a discussion phase of the experiment. The starting
recent study (Guastello, Mirabito, & Peressini, matrix P was composed from Equations 1–2:
2020) in which 43 groups of three to seven under- (See below)
graduates participated in a group decision-making Driver and empath scores were computed
activity entitled Subarctic Survival (Lafferty, by SyncCalc. In this example, the driver was

To1 To2 To3 To4 To5 To6 To7 Driver Score


From1 0.8528 0.5446 0.5397 0.5613 0.5726 0.4879 0.4868 2.4330
From2 0.1461 0.7359 0.4360 0.4381 0.4099 0.4230 0.4141 1.4634
From3 0.0837 0.2355 0.6151 0.3105 0.3038 0.2998 0.3858 0.8682
P= From4 0.1284 0.2217 0.4051 0.5926 0.3183 0.3147 0.3776 0.9240
From5 0.0904 0.2600 0.3428 0.4175 0.7088 0.3692 0.4529 1.2114
From6 0.1166 0.2000 0.3168 0.3198 0.3121 0.5983 0.3896 0.8633
From7 0.0623 0.1480 0.2352 0.2695 0.1801 0.2198 0.5531 0.5404
Empath Score 0.7978 1.0722 1.2970 1.3038 1.3217 1.1485 1.3627
178 GUASTELLO AND PERESSINI

0.8528 0.5446 0.5397 0.5613 0.5726 0.4879


0.1461 0.7359 0.4360 0.4381 0.4099 0.4230
0.0837 0.2355 0.6151 0.3105 0.3038 0.2998
M=
0.1284 0.2217 0.4051 0.5926 0.3183 0.3147
0.0904 0.2600 0.3428 0.4175 0.7088 0.3692
0.1166 0.2000 0.3168 0.3198 0.3121 0.5983
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Person 1, and the empath was Person 7. The task, p = .051; R = .483, F(3, 37) = 3.743, p =
This document is copyrighted by the American Psychological Association or one of its allied publishers.

reduced vector of coefficients for the empath .019. The results illustrated a profile for success
was as follows: whereby better-performing teams were more
strongly focused on the information presentation,
0.4868 acted more individualistically during the individ-
0.4141 ual task, and acted more like a cohesive group
0.3858 during the group task (Guastello, Mirabito, &
V′ = Peressini, 2020, p. 95).
0.3776
0.4529
0.3896 Preparing Data for SyncCalc

In preparation for using the SyncCalc program


The reduced matrix of coefficients with the
three major steps are required with respect to
empath removed was as follows:
the particular time series data in use. (a) The
(See above)
data may need to be down-sampled depending
The vector of weights was as follows:
on the source, (b) a lag length needs to be deter-
mined that is appropriate to the application,
−0.0552
and (c) the time-series data must be processed
−0.0037 into a matrix corresponding to the model the
0.2832 researcher chooses to employ. The third step
Q=
0.1613 can be tedious, so we include an R script to
0.2649 automate it as described below.
0.2887
Down-Sampling Biometric Data
Finally, the SE coefficient for this group and
activity segment was 0.3742. Down-sampling may be required when one’s
In the broader scope of the data analysis, the data stream is too dense for subsequent analysis.
three mean values of SE were 0.36, 0.34, and 0.31 In our case, the BIOPAC MP150 samples ED
for the video, individual task, and group task data at 200 cycles per second. This rate is far too
respectively. One-sample t tests comparing dense for an effective time series analysis, par-
the samples against the population value of SE ticularly if the experimental task is particularly
indicated that a substantial level of synchrony long. Neurocognitive time series have a multi-
occurred in all three conditions, t(40) = 12.33, fractal structure, meaning that there could be
11.34, and 10.00 respectively, all p < .001 different patterns of events transpiring at differ-
(Guastello, Mirabito, & Peressini, 2020, p. 92). ent time intervals. The shortest, however, are
A repeated-measures analysis of variance indi- in the neighborhood of 300–500 ms. Sampling
cated that differences in SE were not statistically intervals that are too short could lose true infor-
significant across the three events, F(2, 66) = mation because the neurocognitive process has
2.324, p = .106. The better performing groups not been completed. Intervals that are too short
associated with higher levels of synchronization would introduce signal noise, momentary signal
during the video ( p = .090), lower synchroniza- dropouts and other sources of error, and could
tion during the individual task ( p = .004), complicate efforts to establish a meaningful lag
and higher synchronization during the group length for the time series analysis.
DRIVER-EMPATH MODEL 179

When down-sampling within BIOPAC, the 2 units and additional lags at further intervals
Transform Graph command assigns the of 2 units; the additional lags were possibly
mean value of the time interval to the interval autonomic responses to other people in the group
and presents it as one observation. This procedure who were also responding to other group mem-
is preferable to the possible alternative of corre- bers at a lag of 2 units.
lating a measurement from a too-short interval To determine optimal lag length for unfamil-
with others later in time. SyncCalc was designed iar data, we recommend proceeding in a way
for use with any biometric or behavioral data. similar to that in Guastello and Mirabito (2018):
For some types of data, down-sampling might Spreadsheets of biometric data for the groups
not be necessary at all. collected in the experiment were uploaded to
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

SPSS, and OLS autocorrelations were assessed


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Determining Lag Length for the time series at different combinations of


plausible lag lengths and down-sampling rates.
The optimal lag is context-dependent: The The combination of lags and down-sampling
overriding principle is that it takes real time rates with the largest mean autocorrelation
for an event to occur and exert its influence on would be the combination to adopt.
the next event. For instance, neurocognitive, Note that OLS autocorrelations captures the
industrial production, and macroeconomic pro- non-IID error that is actually a hallmark of an
cesses all operate on very different time scales. underlying nonlinear process (Guastello &
Social–behavioral data can operate at a scale Gregson, 2011; Ozaki, 2012). Also, it is impor-
of up to a few seconds, and possibly as many as tant to maintain consistent lag and down-
20 s, depending on the types of utterances, com- sampling choices for all persons, groups, and
plexity of the cognitive processes, and whether experimental conditions throughout the data set
movement of physical objects is involved. Inas- in order to avoid introducing artifacts related to
much as the brain’s responses to workload appear data handling. A lag of 2 is a good place to begin
to have a lower limit of 300 ms, one-third of a with ED data sampled at 1 s or less.
second is probably the shortest time interval and
lag unit that one can expect. Most biometric data Preparing Matrix for SyncCalc
collected from groups and teams have used a lag of
1 s (Kazi et al., 2021), but longer and shorter lags In order to run, SyncCalc requires a matrix P
may be optimal in some situations. of autocorrelations and transfer coefficients as
SPSS and other statistical packages can pro- input. These can be generated with any statistics
duce autocorrelations and partial autocorrelations package desired, but can be quite tedious
using autoregressive integrated moving averages and time intensive. In order to facilitate this
(Box & Jenkins, 1970). Autoregressive inte- procedure, we created a script in R (proc-
grated moving averages linear correlations players2.r) that will analyze the time series
control for dependent (non-IID) error. Partial analysis for the downsampling and lag analysis
correlations indicate the effect of particular lags and generate a matrix suitable for direct input
controlling for effects at shorter lags. They could into the SyncCalc program. The script produces
provide some insight as to the location of primary (a) detailed output for all the regression analyses,
and secondary lag effects. (b) a brief summary of that output, and (c) the
Figure 1 contains graphs of autocorrelation matrices for input to SyncCalc.
and partial autocorrelation spectra for two people To use the R routines, the data should be
in one of the groups in the Subarctic study organized on an EXCEL spreadsheet that con-
(Guastello, Mirabito, & Peressini, 2020). The tains the time marker in the first column, and
trend in Figure 1A for one of the participants is one column for each person in the group; both
typical of time series analyses: Autocorrelation is items are produced automatically in BIOPAC.
highest at a lag of 1 unit and decays afterwards. There should be no additional columns. The
The partial autocorrelation spectrum (Figure 1B) names of the measurements should appear on
indicated some small but significant ( p < .05) the top row and edited to remove nonalphanu-
secondary lags after lag 2. The trends for the meric characters. The first observations should
other person in Figure 1C and 1D were more start on the second row with no additional rows
interesting. The optimal lag appeared at lag for headings. The spreadsheet must then be saved
180 GUASTELLO AND PERESSINI

Figure 1
Time Series Analysis for Two Members of a Group Participating in the Subarctic Survival Experiment
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. A: Autocorrelation spectrum (ACF) for Person 1. B: Partial ACF for Person 1. C: ACF for Person 2. D: Partial ACF for
Person 2. Horizontal bars indicate the upper and lower limits of the 95% confidence interval. From “Time Granularity, Lag
Length, and Down-Sampling Rates for Neurocognitive Data,” by S. J. Guastello and L. Mirabito, 2018, Nonlinear Dynamics,
Psychology, and Life Sciences, 22(4), p. 460. Copyright 2018 by the Society for Chaos Theory in Psychology & Life
Sciences. Reprinted with permission of the Society for Chaos Theory in Psychology & Life Sciences. See the online article for
the color version of this figure.

as a comma separated variables file for SyncCalc other nonlinear models, which are discussed in
to access. From that point, procplayers2.r the next section of this article.
can be run as an R script in the usual way. There are two lines of the R script that analysis.
Recall that the precise matrix generated is One defines the intended lag length (scroll down
dependent upon the model one chooses for the the script file to see it):
data. In our discussion above, we focused on a LAG <- 1
linear model (Equations 1–2) in the foregoing
discussion because it is the dyadic model used The numeral 1 indicates a lag of one unit. It can
most often with SyncCalc to date. The linear be changed to 2, 3, etc., as needed. The other is
matrix will be output to a file with the suffix COLUMNS_TO_IGNORE <- 0
“-lMat.txt.” So if your data file is called
MyData.txt, SyncCalc will generate the linear The numeral 0 tells the program not to exclude
model matrix in a file called MyData- any of the columns in the spreadsheet from the
lMat.txt, and the detailed calculation will analysis. If 0 is changed to numeral 1, the program
be output to MyData-out.txt. The R script will ignore the last column in the spreadsheet. If
also produces matrices corresponding to three 0 is changed to 2, it will ignore the last two
DRIVER-EMPATH MODEL 181

columns, etc. This code can be used in studies that playing an emergency response board game against
involve separating two interacting teams that are one human opponent. Results showed that there
recorded simultaneously. was validity to all four models, although the linear
From this point on, the R script centers the model produced the greatest amount of synchrony
time series and converts millivolts to microsie- and overall accuracy, followed by the double-
mens by multiplying the data by −1,000; this exponential model, followed by a tie between
function (called tConvert) can be customized the two logistic map models. The generalizability
as needed. After that the routine constructs the of these results has not yet been determined.
matrix of autocorrelations and transfer coeffi-
cients for a linear model using the standard Invoking SyncCalc
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

built in R routine lm. (It also calculates three


This document is copyrighted by the American Psychological Association or one of its allied publishers.

other matrices for nonlinear models as described SyncCalc produces all the output shown in
in the next section.) Finally, it generates a the numerical example in a.txt file. To invoke
summary of the calculations in a file called SyncCalc, the command synccalc My-
MyData-sum.txt. Data-lMat.txt will output to the screen.
Alternatively, the command synccalc My-
Nonlinear Dynamical Dyadic Models Data-lMat.txt > out.txt produces the
output in the file out.txt. Figure 2 sum-
procplayers2.r also produces P using marizes the sequence of analytic steps for
three different nonlinear models that have described above, from raw time series data to
been offered to describe dyadic interactions, invoking SyncCalc.
with file names ending in *nl1mat.txt,
*nl2mat.txt, and *.nl3mat.txt. Research Applications of SyncCalc
They are offered as part of a work in progress
to explore synchrony in higher-dimensional Substantive research to date using the SE
chaotic systems (Guastello & Peressini, coefficient has also investigated relationships
2021a). *nl1mat.txt uses an exponential between autonomic synchrony and performance,
structure for the dyadic relationship in place perceptions of workload, cooperation and com-
of Equations 2. Two studies indicated that the petition dynamics, and leadership emergence.
double-exponential model picks up transfer In one series of studies, groups of three to eight
effects more often (Guastello et al., 2006) participants played the role of an emergency
and more often correlates with other important response team in a tabletop board game,
effects (Guastello, 2016) than the linear mod- The Creature that Ate Sheboygan (Simulation
els. The double-exponential model is computed Productions, 1979), in which they tried to subdue
through nonlinear regression, however, which one or two monsters who were trying to destroy
is a computational procedure necessitated a city.
by the use of regression weights located in For a sample of teams containing three or
exponents. four participants and one opponent, greater
*nl2mat.txt uses a logistic map model in synchrony occurred in groups of four. When
which each member of a dyad acts as a control data for teams and their opponents were analyzed
parameter that modulates the dynamics of the together, the opponent was the empath in 38
target person. Vallacher et al. (2005) proposed out of the 59 games played; in those cases the
the logistic map function for synchronized inter- opponent was more likely to win (Guastello
nal states that could reflect autonomic arousal, et al., 2018). For teams with two opponents,
moods, and motivation levels. The logistic map SE was greater when the contesting teams were
is the simplest structure that could produce fixed analyzed together rather than separately
points (producing linear time series), oscillations, (Guastello, Palmer, et al., 2020). SE did not
and chaotic time series, depending on the value predict team performance in a subsequent
of the control parameter. *.nl3mat.txt em- game, but prior wins predicted subsequent syn-
ploys an exponential variant of the logistic map. chrony. A later study reported, however, that
The four dyadic models were compared through synchrony during one three-game experimental
SyncCalc in a recent study (Guastello & Peressini, session predicted performance in a session a
2021a) using ED data from a four-person team week later, once group size was statistically
182 GUASTELLO AND PERESSINI

Figure 2
Flowchart of Analytic Steps From Raw Data to Final Results
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. ARIMA = autoregressive integrated moving averages; ANOVA = analysis of variance; OLS = ordinary least squares.
See the online article for the color version of this figure.

controlled (Guastello, Witty, et al., 2020); thus ratio is higher (Guastello & Peressini, 2021b).
the synchrony–performance relationship could The ratio is the highest empath score in the group
be operating on multiple time scales. divided by the highest driver score in the group.
In an expanded sample of 44 teams with three The empath-driver ratio is naturally lower in
to eight members, workload manipulations command-and-control environments, and higher
(one vs. two opponents, time pressure) affect when members of the group interact outside of
synchrony levels (Guastello et al., 2019). Higher the formal chain of communication. Those results
values of SE were correlated with lower ratings open up numerous possibilities for examining
of coordination demand and higher levels of connections between synchrony and different
communication demand by the team members social network configurations, social fault lines
at the start of the session; higher values of SE were within the group, or other meaningful social
also correlated with lower team dissatisfaction outcomes.
at the end of the session. SE is indifferent to the data type and not limited
There is a tendency in the synchrony literature to ED, and only requires time series data. Re-
to regard the more influential member of a dyad searchers have started using SE with heart rate
as the leader and the partner as the follower. That (Danyluck et al., 2020; Rosen, 2021) and photo-
leader–follower designation does not hold for plethysmography (van Eijndhoven et al., 2022)
groups of three or more members, however. In with team data. At present it is too early to make
two experimental tasks (Guastello, Witty, et al., further generalizations about the efficacy of the
2020), individuals’ driver scores were either various physiological measures for detecting
unrelated or negatively related to leadership char- synchrony.
acteristics. Empath scores, in contrast, were posi-
tively related to leadership characteristics. Limitations
Further research could explore conditions that
promote greater autonomic synchrony and those A simple research question is whether a group
that would dissuade it. A recent study showed has synchronized to any meaningful extent or not.
that stronger SE occurs when the empath-driver One could employ a one-sample z test using the
DRIVER-EMPATH MODEL 183

population mean and standard error cited earlier. measured by questionnaires at the end of each
The population standard error is rather large, session, experience better outcomes if the groups’
however, because it was based on a wide range average alliance with the therapist is also strong
of realistic combinations of autocorrelations (Kivlighan et al., 2022). From the vantage point
and off-diagonal entries (Guastello & Peressini, of synchrony research, we have learned that
2017). Standard errors in real experiments are movement synchrony in the client–therapist
much smaller; a more realistic test of the hypothe- dyad signals better results from individual ses-
sis can be made by collecting a few samples from a sions and in the longer term (Ramseyer &
few groups of a similar nature and using the Tschacher, 2016) so long as the therapist was
sample’s standard error in a one-sample t test as leading the client and not the other way around
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

in the example above. As a last resort, however, the (Ramseyer, 2020). Stronger autonomic syn-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

statistical significance of synchrony in a single chrony occurs when the therapist’s and client’s
time series can be assessed by collecting samples empathy levels are higher (Marci et al., 2007;
of randomly shuffled data and comparing SE from Palumbo et al., 2017). The movement synchrony
the target time series against those in the surrogate results appear similar to Jiang et al. (2015) who
sample. Random shuffling preserves the mean, found greater influence of leaders to followers,
standard deviation, and distribution of the raw measured by electroencelphalogram, when they
observations, but breaks up the serial order among decomposed triads. For larger groups, however,
the observations that is produced by deterministic driver scores were negatively correlated with
processes (Moulder et al., 2018). leadership ratings, and empath scores were posi-
SE is theoretically capable of producing some tively related (Guastello, Witty, et al., 2020).
extreme values when diagonal values of P are Based on this incomplete information, one
very low and off-diagonal values are high. This would speculate that the therapist would be
condition occurred in about 2% of possible con- both a driver and an empath when alliance with
ditions tested in the simulation study (Guastello & the therapist is strong. With weak alliances, the
Peressini, 2017). Research is in progress to coun- therapist might not be either one. SE would be a
teract this possible condition (Peressini & low value if the group members did not have
Guastello, 2021). attunement to the therapist in common. The
SE is not affected by group size up to 16 combination of weak alliances a high SE could
members. The extrapolation to larger groups has occur, however, if there was a member of the
not yet been explored. We caution interested re- group demonstrating an unusual amount of nar-
searchers, however, that the social–psychological cissistic tendency or disruptive impact on the
characteristics of larger groups, for example, group outside of what the therapist might have
those with hierarchical internal relationships, deliberately planned. It would be valuable to
could present some nuances. organize studies that connect SE to measures
of alliance with the therapist and other critical
Future Applications events in therapy that signal positive therapeutic
outcomes, such as moments of insight.
The known applications of the driver-empath Families in therapy, by definition, have rela-
model were concerned with work-related behav- tionships outside of the therapeutic context. The
ior by newly formed groups that did not have relationships can be characterized by the “five
appointed leaders. The group therapy context Rs:” responses to other family members, rules of
presents some new and interesting research engagement, roles in the family, pairwise rela-
directions for studying synchrony, considering tionships, and subjective realities (Pincus, 2001,
that there is a therapist leader, and the group is 2009). If one were to analyze the content of their
working on psychological issues germane to their interactions in therapy sessions, unhealthy family
mental health. Group therapies for clients in system would display rigid and stereotypic pat-
residential treatment bring another nuance which terns of interaction, whereas a healthy system
is that the participants might have relationships would display nuanced and flexible interactions
outside of the group that could affect their (Pincus & Guastello, 2005). One can also char-
interactions within group sessions. acterize their interaction patterns in terms of
What we have learned so far is that individuals the “three Cs:” conflict, control, and closeness,
who have stronger alliances with the therapist, which regulate the flow of information within the
184 GUASTELLO AND PERESSINI

group (Bales, 1999; Hollander & Offermann, routine is associated with poor performance in a
1990). Thus it would also be valuable to organize team of male volleyball athletes [Paper presentation].
studies that connect SE to different patterns of Rs Annual International Conference of the Society for
and Cs. Chaos Theory in Psychology & Life Sciences,
Chapel Hill, North Carolina, United States.
Elkins, A. N., Muth, E. R., Hoover, A. W., Walker,
Annotated Bibliography A. D., Carpenter, T. L., & Switzer, F. S. (2009).
Physiological compliance and team performance.
Adamatzky, A. (2005). Dynamics of crowd- Applied Ergonomics, 40(6), 997–1003. https://
minds: Patterns of irrationality in emotions, be- doi.org/10.1016/j.apergo.2009.02.002
liefs, and actions. World Scientific. Gawron, V. J. (2019). Workload measures. CRC
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

This is primarily a mathematical text that ex- Press. https://doi.org/10.1201/9780429019579


This document is copyrighted by the American Psychological Association or one of its allied publishers.

amines some patterns of diffusion of dysfunc- Gillis, H. L. L., Jr., Kivlighan, D. M., Jr., & Russell,
tional ideas through large-scale social networks K. C. (2016). Between-client and within-client
engagement and outcome in a residential wilderness
and how they synchronize and stabilize.
treatment group: An actor partner interdependence
Guastello, S. J., & Liebovitch, L. S. (2009). analysis. Psychotherapy: Theory, Research, and
Introduction to nonlinear dynamics and complex- Practice, 53(4), 413–423. https://doi.org/10.1037/
ity. In S. J. Guastello, M. Koopmans, & D. pst0000047
Pincus. (Eds.), Chaos and complexity in psychol- Gordon, I., Wallot, S., & Berson, Y. (2021). Group-
ogy: Theory of nonlinear dynamical systems (pp. level physiological synchrony and individual-level
1–40). Cambridge University Press. anxiety predict positive affective behaviors during
This chapter explains basic concepts of non- a group decision-making task. Psychophysiology,
linear dynamics, with substantial references, that 58(9), Article e13857. https://doi.org/10.1111/ps
are used in various psychological applications, yp.13857
Gorman, J. C., Grimm, D. A., Stevens, R. H., Gallo-
including group dynamics, throughout the book:
way, T., Willemsen-Dunlap, A. M., & Halpin, D. J.
attractors, bifurcations, chaos, fractals, cata- (2020). Measuring real-time team cognition during
strophes, and self-organization. team training. Human Factors, 62(5), 825–860.
https://doi.org/10.1177/0018720819852791
Software Availability Granger, C. W. J. (1969). Investigating causal rela-
tions by econometric models and cross-spectral
SyncCalc, Procplayers2.r, operation instructions, methods. Econometrica, 37(3), 424–438. https://
and test data are freely available from: https://acade doi.org/10.2307/1912791
Guastello, S. J. (2016). Physiological synchronization
mic.mu.edu/peressini/synccalc/slinks.htm
in a vigilance dual task. Nonlinear Dynamics Psy-
chology and Life Sciences, 20(1), 49–80.
References Guastello, S. J. (in press). Human factors engineering
and ergonomics: A systems approach (3rd ed.).
Aks, D. (2011). Analysis of recurrence: Overview CRC Press.
and application to eye-movement behavior. In Guastello, S. J., Correro Ii, A. N., Marra, D. E., &
S. J. Guastello & R. A. M. Gregson (Eds.), Non- Peressini, A. F. (2019). Physiological synchroniza-
linear dynamical systems analysis for the behav- tion and subjective workload in a competitive
ioral sciences using real data (pp. 231–250). emergency response task. Nonlinear Dynamics
CRC Press. Psychology and Life Sciences, 23(3), 347–376.
Bak, P. (1996). How nature works. Springer. https:// Guastello, S. J., & Gregson, R. A. M. (Eds.). (2011).
doi.org/10.1007/978-1-4757-5426-1 Nonlinear dynamical systems analysis for the
Bales, R. F. (1999). Social interaction systems: Theory behavioral sciences using real data. CRC Press.
and measurement. Transaction Publishing. Guastello, S. J., Marra, D. E., Castro, J., Equi, M., &
Box, G. E. P., & Jenkins, G. M. (1970). Time series Peressini, A. F. (2017). Turn taking, team synchro-
analysis: Forecasting and control. Holden-Day. nization, and non-stationarity in physiological time
Coco, M. I., Mønster, D., Leonardi, G., Dale, R., & series. Nonlinear Dynamics Psychology and Life
Wallot, S. (2021). Unidimensional and multidimen- Sciences, 21(3), 319–334.
sional methods for recurrence quantification analy- Guastello, S. J., Marra, D. E., Peressini, A. F., Castro,
sis with crqa. The R Journal, 13(1), 145–163. J., & Gomez, M. (2018). Autonomic synchroniza-
https://doi.org/10.32614/RJ-2021-062 tion, team coordination, participation, and perfor-
Danyluck, D., Tamminen, K., & Chen, R. (2020, July). mance. Nonlinear Dynamics Psychology and Life
Physiological synchrony during a pre-practice Sciences, 22(3), 359–394.
DRIVER-EMPATH MODEL 185

Guastello, S. J., & Mirabito, L. (2018). Time granu- Henning, R. A., Boucsein, W., & Gil, M. C. (2001).
larity, lag length, and down-sampling rates for Social-physiological compliance as a determinant
neurocognitive data. Nonlinear Dynamics Psychol- of team performance. International Journal of
ogy and Life Sciences, 22(4), 457–483. Psychophysiology, 40(3), 221–232. https://doi.org/
Guastello, S. J., Mirabito, L., & Peressini, A. F. 10.1016/S0167-8760(00)00190-2
(2020). Autonomic synchronization under three Hollander, E. P., & Offermann, L. R. (1990). Power
task conditions and its impact on team performance. and leadership in organizations: Relationships in
Nonlinear Dynamics Psychology and Life Sciences, transition. American Psychologist, 45(2), 179–189.
24(1), 79–104. https://doi.org/10.1037/0003-066X.45.2.179
Guastello, S. J., Palmer, C., Marra, D. E., & Peressini, Jiang, J., Chen, C., Dai, B., Shi, G., Ding, G., Liu, L.,
A. F. (2020). The effect of cooperation and compe- & Lu, C. (2015). Leader emergence through inter-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tition dynamics on autonomic synchrony in teams. personal neural synchronization. Proceedings of the
This document is copyrighted by the American Psychological Association or one of its allied publishers.

In K. Viol, H. Schöller, & W. Aichhorn (Eds.), Self- National Academy of Sciences of the United States
organization: A paradigm for the human sciences? of America, 112(14), 4274–4279. https://doi.org/10
(pp. 303–318). Springer. .1073/pnas.1422930112
Guastello, S. J., & Peressini, A. F. (2017). Develop- Kazi, S., Khaleghzadegan, S., Dinh, J. V., Shelhamer,
ment of a synchronization coefficient for biosocial M. J., Sapirstein, A., Goeddel, L. A., Chime, N. O.,
interactions in groups and teams. Small Group Salas, E., & Rosen, M. A. (2021). Team physiolog-
Research, 48(1), 3–33. https://doi.org/10.1177/104 ical dynamics: A critical review. Human Factors,
6496416675225 63(1), 32–65. https://doi.org/10.1177/001872081
Guastello, S. J., & Peressini, A. F. (2021a). A com- 9874160
parison of four dyadic synchronization models. Kenny, D. A., & Cook, W. (1999). Partner effects in
Nonlinear Dynamics Psychology and Life Sciences, relationship research. Conceptual issues, analytic
25(1), 19–39. difficulties, and illustrations. Personal Relation-
Guastello, S. J., & Peressini, A. F. (2021b). The ships, 6(4), 433–448. https://doi.org/10.1111/j.14
relative influence of drivers and empaths on team 75-6811.1999.tb00202.x
synchronization. Nonlinear Dynamics Psychology Kenny, D. A., & Garcia, R. (2012). Using the actor-
and Life Sciences, 25(3), 357–382. partner interdependence model to study the effects
Guastello, S. J., Pincus, D., & Gunderson, P. R. of group composition. Small Group Research,
(2006). Electrodermal arousal between participants 43(4), 468–496. https://doi.org/10.1177/10464964
in a conversation: Nonlinear dynamics and linkage 12441626
effects. Nonlinear Dynamics Psychology and Life Keum, B. T. (2018). Conceptual application of the
Sciences, 10(3), 365–399. group actor–partner interdependence model for
Guastello, S. J., Witty, B., Johnson, C., & Peressini, person–group psychological research. Translational
A. F. (2020). Autonomic synchronization, leader- Issues in Psychological Science, 4(4), 340–348.
ship emergence, and the roles of drivers and em- https://doi.org/10.1037/tps0000180
paths. Nonlinear Dynamics Psychology and Life Kivlighan, D. M., Jr. (2021). From where is the group?
Sciences, 24(4), 451–473. To what is the group?: Contributions of actor−part-
Haken, H. (1984). The science of structure: Syner- ner interdependence modeling. Group Dynamics:
getics. Van Nostrand Reinhold. Theory, Research, and Practice, 25(3), 229–237.
Hatfield, E., Rapson, R. L., & Le, Y. C. L. (2009). https://doi.org/10.1037/gdn0000164
Emotional contagion and empathy. In J. Decety & Kivlighan, D. M., III, Cohen, K., Zilcha-Mano, S.,
W. Ickes (Eds.), The social neuroscience of empa- Palgi, S., & Zilberstein, A. (2022). Examining state
thy (pp. 19–30). MIT Press. https://doi.org/10.7551/ and trait alliance in group therapy: A within-person
mitpress/9780262012973.003.0003 and between-person actor-partner interdependence
Hatfield, E., Cacioppo, J. T., & Rapson, R. L. (1993). model. Group Dynamics: Theory, Research, and
Emotional contagion. Current Directions in Psy- Practice, 26(2), 137–150. https://doi.org/10.1037/
chological Science, 3(3), 96–99. https://doi.org/10 gdn0000152
.1111/1467-8721.ep10770953 Kozlowski, S. W. J. (2015). Advancing research on
Helm, J. L., Miller, J. G., Kahle, S., Troxel, N. R., & team process dynamics: Theoretical, methodologi-
Hastings, P. D. (2018). On measuring and modeling cal, and measurement considerations. Organiza-
physiological synchrony in dyads. Multivariate tional Psychology Review, 5(4), 270–299. https://
Behavioral Research, 53(4), 521–543. https:// doi.org/10.1177/2041386614533586
doi.org/10.1080/00273171.2018.1459292 Lafferty, C. J. (2003). The subarctic survival situation.
Helm, J. L., Sbarra, D., & Ferrer, E. (2012). Assessing Human Synergetics.
cross-partner associations in physiological re- Levenson, R. W., & Gottman, J. M. (1983). Marital
sponses via coupled oscillator models. Emotion, interaction: Physiological linkage and affective
12(4), 748–762. https://doi.org/10.1037/a0025036 exchange. Journal of Personality and Social
186 GUASTELLO AND PERESSINI

Psychology, 45(3), 587–597. https://doi.org/10 sciences. Cambridge University Press. https://


.1037/0022-3514.45.3.587 doi.org/10.1017/CBO9780511755743
Marci, C. D., Ham, J., Moran, E., & Orr, S. P. (2007). Pincus, D. (2001). A framework and methodology
Physiologic correlates of perceived therapist empa- for the study of nonlinear, self-organizing family
thy and social–emotional process during psycho- dynamics. Nonlinear Dynamics Psychology and Life
therapy. Journal of Nervous and Mental Disease, Sciences, 5(2), 139–174. https://doi.org/10.1023/A:
195(2), 103–111. https://doi.org/10.1097/01.nmd 1026419517879
.0000253731.71025.fc Pincus, D. (2009). Coherence, complexity, and infor-
Marwan, N., & Kurths, J. (2003). Nonlinear analysis mation flow: Self-organizing processes and psycho-
of bivariate data with cross recurrence plots. Physics therapy. In S. J. Guastello, M. Koopmans, & D.
Letters, 302, 299–307. Pincus (Eds.), Chaos and complexity in psychology:
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Matthews, G., & Reinerman-Jones, L. E. (2017). Theory of nonlinear dynamical systems (pp. 335–
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Workload assessment: How to diagnose workload 369). Cambridge University Press.


issues and enhance performance. Human Factors Pincus, D., & Guastello, S. J. (2005). Nonlinear
and Ergonomics Society. dynamics and interpersonal correlates of verbal
Mønster, D., Håkonsson, D. D., Eskildsen, J. K., & turn taking patterns in group therapy. Small Group
Wallot, S. (2016). Physiological evidence of inter- Research, 36(6), 635–677. https://doi.org/10.1177/
personal dynamics in a cooperative production task. 1046496405280864
Physiology & Behavior, 156, 24–34. https://doi.org/ Prigogine, I., & Stengers, I. (1984). Order out of
10.1016/j.physbeh.2016.01.004 chaos: Man’s new dialog with nature. Bantam.
Moulder, R. G., Boker, S. M., Ramseyer, F., & Ramseyer, F., & Tschacher, W. (2016). Movement
Tschacher, W. (2018). Determining synchrony coordination in psychotherapy: Synchrony of hand
between behavioral time series: An application movements is associated with session outcome: A
of surrogate data generation for establishing single-case study. Nonlinear Dynamics Psychology
falsifiable null-hypotheses. Psychological Meth- and Life Sciences, 20(2), 145–166.
ods, 23(4), 757–773. https://doi.org/10.1037/me Ramseyer, F. T. (2020). Exploring the evolution of
t0000172 nonverbal synchrony in psychotherapy: The idio-
Newhouse, R., Ruelle, D., & Takens, F. (1978). graphic perspective provides a different picture.
Occurrence of strange attractors: An axiom near Psychotherapy Research, 30(5), 622–634. https://
quasi-periodic flows on Tm, m>3. Communications doi.org/10.1080/10503307.2019.1676932
in Mathematical Physics, 64(1), 35–41. https:// Richardson, M. J., Garcia, R. L., Frank, T. D., Gergor,
doi.org/10.1007/BF01940759 M., & Marsh, K. L. (2012). Measuring group
Ozaki, T. (2012). Time series modeling of neurosci- synchrony: A cluster-phase method for analyzing
ence data. CRC Press. https://doi.org/10.1201/ multivariate movement time-series. Frontiers in
b11527 Physiology, 3, Article 405. https://doi.org/10.3389/
Palumbo, R. V., Marraccini, M. E., Weyandt, L. L., fphys.2012.00405
Wilder-Smith, O., McGee, H. A., Liu, S., & Goodwin, Rosen, M. (2021, February). Cross-analog validity
M. S. (2017). Interpersonal autonomic physiology: A evidence for unobtrusive measurement strategies
systematic review of the literature. Personality and for individual and team competencies [Paper pre-
Social Psychology Review, 21(2), 99–141. https:// sentation]. Paper presented to the Bioastronautic-
doi.org/10.1177/1088868316628405 s@hopkins Inaugural Event. The Johns Hopkins
Peressini, A. F., & Guastello, S. J. (2016). SyncCalc University School of Medine, Baltimore, Maryland,
1.0 [Computer software]. Retrieved August 1, 2016 United States.
from https://academic.mu.edu/peressini/synccalc/ Salas, E., Stevens, R., Gorman, J., Cooke, N. J.,
synccalc.htm Guastello, S. J., & von Davier, A. (2015). What
Peressini, A. F., & Guastello, S. J. (2021, July). will quantitative measures of teamwork look like
Large-scale databases for nonlinear phenomena: in 10 years? Proceedings of the Human Factors
Sync Coefficient SE and variations [Paper presen- and Ergonomics Society, 59(1), 235–239. https://
tation]. 31st Annual International Conference of doi.org/10.1177/1541931215591048
the Society for Chaos Theory in Psychology & Sawyer, R. K. (2005). Social emergence: Societies as
Life Sciences, Chapel Hill, North Carolina, United complex systems. Cambridge University Press.
States. https://doi.org/10.1017/CBO9780511734892
Peressini, A. F., & Guastello, S. J. (2022). SyncCalc Simulation Productions. (1979). The creature that ate
2.1 [Computer software]. Retrieved June 20, 2022 Sheboygan: Wreak havoc with the monster of your
from https://academic.mu.edu/peressini/synccalc/ choice.
synccalc.htm Stefański, A. (2009). Determining thresholds of com-
Pikovsky, A., Rosenblum, M., & Kurths, J. (2001). plete synchronization, and application. World
Synchronization: A universal concept in nonlinear Scientific. https://doi.org/10.1142/7139
DRIVER-EMPATH MODEL 187

Stevens, R., Galloway, T., & Lamb, C. (2014). in Biological and Artificial Systems, 6(1), 35–52.
Submarine navigation team resilience: Linking https://doi.org/10.1075/is.6.1.04val
EEG and behavioral models. Proceedings of the van Eijndhoven, K. H. J., Wiltshire, T. J., Halgas, E.,
Human Factors and Ergonomics Society, 58(1), & Gevers, J. M. P. (2022, April 19). Never waste
245–249. https://doi.org/10.1177/154193121458 a good crisis: Examining team coordination break-
1051 downs during crisis situations. PsyArXiv. https://
Strogatz, S. (2003). Sync: The emerging science of doi.org/10.31234/osf.io/p4r6h
spontaneous order. Hyperion. Webber, C. L., Jr., & Zbilut, J. P. (2007). Recurrence
Sulis, W. H. (2016). Synchronization, TIGoRS, and quantification: Feature extractions from recurrence
information flow in complex systems: Dispositional plots. International Journal of Bifurcation and Chaos
celluar automata. Nonlinear Dynamics Psychology in Applied Sciences and Engineering, 17(10), 3467–
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and Life Sciences, 20(2), 293–317. 3475. https://doi.org/10.1142/S0218127407019226


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Torrents, C., Balagué, N., & Hristovski, R. (2016). Whittle, P. (2010). Neural nets and chaotic carriers
Interpersonal coordination in competitive and (2nd ed.). Imperial College Press. https://doi.org/10
cooperative performance contexts. In P. Passos, .1142/p728
K. Davids, & J. Y. Chow (Eds.), Interpersonal Wiggins, S. (1988). Global bifurcations and chaos.
coordination and performance in social systems Springer. https://doi.org/10.1007/978-1-4612-1042-9
(pp. 85–93). Routledge.
Vallacher, R. R., Nowak, A., & Zochowski, M. Received December 11, 2021
(2005). Dynamics of social coordination. Interac- Revision received October 11, 2022
tion Studies: Social Behaviour and Communication Accepted January 9, 2023 ▪

You might also like