UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND

A C T A U N I V E R S I T A T I S O U L U E N S I S

University Lecturer Tuomo Glumoff

University Lecturer Santeri Palviainen

Senior research fellow Jari Juuti

Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Planning Director Pertti Tikkanen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho

Publications Editor Kirsti Nurkkala

ISBN 978-952-62-2218-9 (Paperback)ISBN 978-952-62-2219-6 (PDF)ISSN 0355-323X (Print)ISSN 1796-2242 (Online)

U N I V E R S I TAT I S O U L U E N S I SACTAE

SCIENTIAE RERUM SOCIALIUM

E 184

AC

TAH

éctor Javier Pijeira D

íaz

OULU 2019

E 184

Héctor Javier Pijeira Díaz

ELECTRODERMAL ACTIVITY AND SYMPATHETIC AROUSAL DURING COLLABORATIVE LEARNING

UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU,FACULTY OF EDUCATION

ACTA UNIVERS ITAT I S OULUENS I SE S c i e n t i a e R e r u m S o c i a l i u m 1 8 4

HÉCTOR JAVIER PIJEIRA DÍAZ

ELECTRODERMAL ACTIVITY AND SYMPATHETIC AROUSAL DURING COLLABORATIVE LEARNING

Academic dissertation to be presented with the assent ofthe Doctoral Training Committee of Human Sciences ofthe University of Oulu for public defence in the OPauditorium (L10), Linnanmaa, on 3 May 2019, at 12 noon

UNIVERSITY OF OULU, OULU 2019

Copyright © 2019Acta Univ. Oul. E 184, 2019

Supervised byProfessor Sanna JärveläProfessor Paul A. KirschnerProfessor Hendrik Drachsler

Reviewed byProfessor Petri NokelainenAssociate Professor Daniel Spikol

ISBN 978-952-62-2218-9 (Paperback)ISBN 978-952-62-2219-6 (PDF)

ISSN 0355-323X (Printed)ISSN 1796-2242 (Online)

Cover DesignRaimo Ahonen

JUVENES PRINTTAMPERE 2019

OpponentProfessor Timo Tobias Ley

Pijeira Díaz, Héctor Javier, Electrodermal activity and sympathetic arousal duringcollaborative learning. University of Oulu Graduate School; University of Oulu, Faculty of EducationActa Univ. Oul. E 184, 2019University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland

Abstract

This dissertation investigates high school students’ individual and interpersonal physiology ofelectrodermal activity (EDA) during collaborative learning in naturalistic settings. EDA is anindex of sympathetic arousal, which is concomitant with cognitive and affective processes.

Two data collections were organized with students working collaboratively in triads. The firstone took place during the performance of a science task, and the second during two runs of a six-week advanced physics course. The data collected included EDA (measured unobtrusively usingEmpatica® E3 and E4 wristbands), performance measures (pre- and post-tests, task solutions, andcourse exam), and questionnaires on cognitive, affective, and collaborative aspects of learning.The work was reported in three articles.

The results indicate that, on average, students spent more than half (60%) of the class at a lowarousal level, possibly signaling relaxation, disengagement, or boredom. Most of the time(≈60–95% of the lesson), triad members were at a different arousal level, which might indicatethat students took turns (alternating task-doers) in executing the task or applied some division oflabor rather than truly collaborating. In terms of achievement, sympathetic arousal during theexam was a predictor of the exam grades, and pairwise directional agreement of EDA waspositively and highly correlated to the dual learning gain. Arousal contagion could have occurredin up to 41% of the high arousal intervals found. The possible arousal contagion cases took placemostly on a 1:1 basis (71.3%), indicating that interactions in a collaborative learning triad seem tooccur mainly between two members rather than among the three.

The findings provide an ecologically-valid picture of the students’ EDA responses in theclassroom, both individually and collaboratively, benefiting from the connection of arousal tocognitive and affective processes to increase the saliency of otherwise elusive phenomena.Methodologically, the study contributes to the exploration and exploitation ofpsychophysiological approaches for collaborative learning research. On a practical level, itprovides physiological indices that could be incorporated into learning analytics dashboards tosupport students’ awareness and reflection, and teachers’ pedagogical practices.

Keywords: collaborative learning, electrodermal activity, interpersonal physiology,psychophysiology, sympathetic arousal

Pijeira Díaz, Héctor Javier, Elektrodermaalinen aktiivisuus ja sympaattinenvireystila yhteisöllisen oppimisen aikana. Oulun yliopiston tutkijakoulu; Oulun yliopisto, Kasvatustieteiden tiedekuntaActa Univ. Oul. E 184, 2019Oulun yliopisto, PL 8000, 90014 Oulun yliopisto

Tiivistelmä

Tässä väitöstutkimuksessa tarkastellaan elektrodermaalista aktiivisuutta (EDA) ja tästä johdet-tua sympaattista vireystilaa ja fysiologisia indeksejä, samanaikaisesti yksilöiden ja yksilöidenvälisten kognitiivisten ja affektiivisten prosessien kanssa.

Tutkimusaineisto kerättiin yhteisöllisen oppimisen tilanteista, joissa oppilaat työskentelivätkolmen hengen ryhmissä. Ensimmäinen osa aineistosta kerättiin oppilaiden suorittaessa luon-nontieteiden alan tehtävää ja toinen kahden fysiikan syventävän kurssin aikana. Aineistoon sisäl-tyi EDA (Empatica® E3- ja E4-rannekkeista), oppimisen mittaukset (alku- ja lopputestit, tehtä-vien ratkaisut ja kurssikokeet) sekä kyselylomakkeet oppimisen kognitiivisista, affektiivisista jayhteisöllisen työskentelyn näkökulmista. Tutkimus on raportoitu kolmessa artikkelissa.

Tulokset osoittavat, että opiskelijoiden sympaattisen hermoston vireystila oli keskimäärin ylipuolet (60 %) luokkatyöskentelystä alhainen, mikä viittaa mahdolliseen rentoutumiseen, osallis-tumisen puutteeseen tai tylsistymiseen. Ryhmänjäsenet olivat suurimman osan ajasta (≈60-95 %)eri vireystilan tasoilla, mikä voi tarkoittaa, että he suorittivat tehtävää vuorotellen (tehtävän suo-rittajaa vaihdellen) tai jonkinlaista työnjakoa käyttäen, yhteisöllisen työskentelyn sijaan. Sym-paattinen vireystila kurssikokeessa ennusti kokeen arvosanoja. Lisäksi oppilasparien EDA:nsamansuuntaisuus korreloi vahvasti oppimistulosten kanssa. Yksilöiden välillä tapahtuvaa sym-paattisen vireystilan ”tarttumista” on voinut esiintyä jopa 41 prosentissa todetuista korkean vire-ystilan intervalleista. Mahdolliset ”tarttumiset” ilmenivät enimmäkseen (71,3 %) 1:1 suhteessa,mikä viittaa siihen, että vuorovaikutus yhteisöllisessä oppimisessa näyttäisi tapahtuvan pääasias-sa kahden yksilön välillä kaikkien kolmen sijaan.

Tulokset tarjoavat ekologisesti validin kuvan opiskelijoiden EDA-reaktioista luokkahuonees-sa sekä yksilöllisesti että yhteisöllisesti tarkasteltuna, selventäen samalla kuvaa sympaattisenvireystilan yhteydestä kognitiivisiin ja affektiivisiin prosesseihin. Menetelmällisesti tutkimuskartoittaa psykofysiologisen lähestymistavan mahdollisuuksia yhteisöllisen oppimisen tutkimuk-sessa. Se esittelee fysiologisia indeksejä, jotka voitaisiin visualisoida oppimisen analytiikansovelluksissa opiskelijoiden tietoisuuden ja reflektion sekä opettajien pedagogisten käytäntöjentukemiseksi.

Asiasanat: elektrodermaalinen aktiivisuus, ihmistenvälinen fysiologia, psykofysiologia,sympaattinen vireystila, yhteisöllinen oppiminen

To my lifetime supervisors and indefatigable supporters, my parents

8

9

Acknowledgements It was the early spring of 2015 when I started this journey to the Ph.D. Along the

way, I have felt privileged to count on the supervision of eminent professors such

as Sanna Järvelä (University of Oulu), Paul A. Kirschner (Open University of the

Netherlands), and Hendrik Drachsler (Goethe University Frankfurt), to whom I am

profoundly grateful. I am especially indebted to Sanna Järvelä, my principal

supervisor, for her support in so many ways throughout these years. This Ph.D.

would simply not have been possible without Sanna’s involvement and leadership,

and her innate capacity to motivate and inspire people with her kindness, result-

orientation, and passion for excellence. Thank you, Sanna, for being always

available when I needed it, no matter how busy you were; for all the expert guidance

of a natural leader like you, and for embracing multidisciplinarity to advance the

learning sciences. To Paul A. Kirschner I am grateful for the insightful scientific

discussions, for his honest commentary on my work, and for his efforts in shaping

my scientific writing skills with a focus on structure, clarity, line of argument, and

coherence. From his keen remarks as an experienced editor, I have learned much

of what I now about academic writing formalities and the APA style, for example.

Thank you, Paul, also for always coming to my conference presentations despite

your tight agenda as a sought-after expert in those events, and for the feedback

afterwards. To Hendrik Drachsler I am thankful for his dedication to getting me

started with the Ph.D. endeavor, for helping me to pave the way for independent

work to advance learning analytics, for introducing me to valuable tools and

research methods, and in general, for the much-appreciated guidance and

opportunities along these four years.

I would like to express my gratitude to Professor Timo Tobias Ley (Tallinn

University), an internationally recognized expert in learning analytics and

educational innovation, who is highly experienced in multidisciplinary work, for

accepting to act as opponent at the public defense of this dissertation. For their

meticulous work as pre-examiners of this dissertation, through which they provided

me with insightful, actionable, and critically constructive feedback, I want to thank

Engineering Pedagogy Professor Petri Nokelainen (Tampere University of

Technology) and Associate Professor Daniel Spikol (Malmö University). Similarly,

I am deeply grateful to the Academy of Finland post-doctoral research fellows Piia

Näykki and Sonja Lutovac, for their elaborate, highly-detailed, and thorough

commentary on a previous version of this dissertation, as part of the corresponding

pre-defense seminar at the Faculty of Education.

10

During my doctoral studies, annual meetings with my follow-up group helped

me to evaluate my progress and to keep my research plan up-to-date. I would like

to show my appreciation to the chair of this group, Adjunct Professor Arto Hautala

(Cardiovascular Research Group, Oulu University Hospital), as well as to the other

two members: Dr. Miki Kallio (Unit for Strategy and Science Policy) and Dr. Arttu

Mykkänen (Learning and Educational Technology research group; LET). I am

thankful to this multidisciplinary group for their diligent work in this role,

constructive feedback, experiences shared, and useful suggestions.

The University of Oulu Graduate School (UniOGS) does a great job to, as

indicated in their goal statement, “provide the framework and conditions for high-

quality, research-driven doctoral education.” I would like to acknowledge their

work in helping us doctoral candidates to acquire general and transferable skills

through the several courses they organize on ethics, scientific communication,

intellectual property rights, and university pedagogy, just to name a few. Special

thanks to Anthony Heape and Titta Kallio-Seppä for their guidance on procedures

and paperwork.

The final revision of this dissertation in terms of (APA) style, formatting, and

conventions was carried out by the series editor, University Lecturer Veli-Matti

Ulvinen. I would like to show my appreciation for his rigorousness, precision,

guidance, quick reaction, and clear and didactic suggestions.

Although the work on a dissertation is independent, my Ph.D. studies have

benefited from the frequent scientific activities organized at the LET research group,

including, but not limited to, reading circles, writing Wednesdays, seminars,

presentations by renowned international experts, workshops, strategy days, and

theoretical discussions. For that, and for an unparalleled work atmosphere, I thank

each and every one at the LET team. You have been welcoming and supportive. I

am lucky to have you all as colleagues. Special thanks to Márta Sobocinski, Eetu

Haataja, and my lunch partner in crime Muhterem Dindar. Together with Márta and

Eetu, I started the immersion in the promising world of psychophysiology and

interpersonal physiology. We became avid readers on the topic and were often

sharing and discussing what we learned and what it meant for our respective

research interests. Theories, methods, findings, unsolved questions, challenges...

Our discussions were illuminating on both the possibilities and the difficulties. In

addition, I thank Márta, the first person I talked to after my arrival at Oulu Airport,

for her generosity and invaluable practical support for this adventurer to settle down

in his new home and workplace. To Muhterem, I am grateful for countless affable

discussions over lunch on the nature and practicalities of science and scientific

11

lifestyle, as well as on more mundane topics; and most importantly, for his

friendship.

In the first year of my Ph.D. studies, I had a special collaborator from the

Center of Wireless Communications, M.Sc. Abdul Moiz. Together, we set up our

own instance of the Open edX platform on an Ubuntu Server to be used in the

second data collection of this dissertation, explored possibilities for real-time

biosensor-data collection, and designed and populated a MySQL database, among

other engineering tasks. During the months we were closely working together, we

faced technical challenges, made design decisions, and in general, worked hard to

get the system up-and-running. It was a practical and hands-on work. I would like

to thank Moiz for his commitment to the project, hard work, contagious motivation,

for fruitful discussions in design meetings, and for so many productive

disagreements which led to the best solution for the project. Beyond work, his

dependable friendship made a difference and is one of the highlights of my stay in

Oulu. Thank you, Moiz.

This dissertation has been supported by the Academy of Finland via the SLAM

project (grant number 275440), and the Faculty of Education. I express my

gratitude to both institutions. In addition, I consider myself fortunate to have been

part of the Faculty of Education these years, a pleasant and supportive working

environment, packed with friendly people. I want to thank the Faculty

administration and technical staff for their work to support research. I would also

like to acknowledge my two long-term office mates, Markus Packalen and Laura

Palmgren-Neuvonen. It was a pleasure sharing office with them, and our short

conversations on science, everyday life, sports, languages, Finnish culture, travel,

and so on, provided an excellent break from periods of intense concentration. I

definitely enjoyed my time at KTK328, my research home for most of the Ph.D.

studies.

My first scientific research experience comes from the time when I was

working on my master’s thesis under the supervision of Associate Professor Pedro

J. Muñoz Merino (Universidad Carlos III de Madrid). The work on learning

analytics visualizations from a computer science perspective paved the way for my

discovery of the learning sciences and the research field of technology-enhanced

learning. I am thankful to Pedro for his exemplary work ethics, which inspired me,

and for introducing me to this opportunity for doing my Ph.D. studies in a

multidisciplinary, international environment.

Scientific writing is a skill that develops over time, and (in)formative,

professional commentary directly on our writing equip us with valuable knowledge

12

of formalities, formatting, conventions, word choice, and clarity, among other

aspects. In this sense, I would like to show my appreciation for the work of

Scribendi’s editor 1081.

A Ph.D. journey is a roller-coaster ride spanning practically all the spectrum of

mental states and emotions: excitement, confusion, enjoyment, frustration,

engagement, anxiety, flow, disgust, hope, disappointment, pride, fear, surprise,

gratitude, and the list goes on. Therefore, beyond the academic setting, friends and

family play an indispensable role in helping us regulate those emotions, keeping

the motivation high, putting things in perspective, and strengthening our resilience.

In Oulu, I have found an amazing gang from the four corners of the earth, who, in

one way or another, have contributed to this journey. There are so many of them

that it is virtually impossible to name them all, but huge thanks to each and every

one of them for their kindness, time, company, and generosity. In addition to those

aforementioned, I would like to offer my special appreciation to my YOK18 partner

in crime and close friend Uzair Khan, Laura Jalkanen (to whom I am also indebted

for the initial version of the Finnish abstract of this dissertation), Sonia Saher, Lisi

Herrera, Onel Alcaraz López, and Ilaria Gabbatore. Friendship blossoms anywhere.

Naturally, due to geographical reasons, I have spent more time during the Ph.D.

with my friends in Oulu. However, my gratitude also goes to my dear friends from

Spain and Cuba, and family therein. Special mention to Alejandro Reyes

Bascuñana, Marielena García López, Guillermo López Lagomasino, Carmen

Gutiérrez Rodríguez, and Lucía Hidalgo Sánchez. In Oulu, volleyball has been my

favorite hobby during these Ph.D. years. Playing volleyball several times a week

has proved an effective strategy to maintaining a positive attitude and my signature

smile on my face, especially on moments of setback and despair. Accordingly, I

would like to acknowledge the host of people with whom I have played volleyball

through these years. I am glad I discovered this healthy hobby early enough in my

Ph.D. studies.

Last but not least, my whole life would not be enough to thank the two persons

this dissertation is dedicated to, my parents. You have stimulated my passion for

learning since my early days. You have supported me in every possible way, and

you have given your all for my education, not only in terms of knowledge, but also

in values and principles. I owe you everything. Thanks for the gift of life, the

unconditional love, and for being there through thick and thin. This is for you both.

March 14th, 2019 Héctor J. Pijeira Díaz

13

Abbreviations EDA electrodermal activity

GSR galvanic skin response

MSLQ motivated strategies for learning questionnaire

PCI physiological coupling index

PGR psychogalvanic response

RQ research question

SCL skin conductance level

SCR skin conductance response

14

15

List of original publications This thesis is based on the following publications, which are referred throughout

the text by their Roman numerals:

I Pijeira-Díaz, H. J., Drachsler, H., Järvelä, S., & Kirschner, P. A. (2016). Investigating collaborative learning success with physiological coupling indices based on electrodermal activity. In Proceedings of the Sixth International Conference on Learning Analytics & Knowledge (pp. 64–73). New York, NY: ACM. https://doi.org/10.1145/2883851.2883897

II Pijeira-Díaz, H. J., Drachsler, H., Kirschner, P. A., & Järvelä, S. (2018). Profiling sympathetic arousal in a physics course: How active are students? Journal of Computer Assisted Learning, 34(4), 397–408. https://doi.org/10.1111/jcal.12271

III Pijeira-Díaz, H. J., Drachsler, H., Järvelä, S., & Kirschner, P. A. (2019). Sympathetic arousal commonalities and arousal contagion during collaborative learning: How attuned are triad members? Computers in Human Behavior, 92(March), 188–197. https://doi.org/10.1016/j.chb.2018.11.008

The articles are co-authored with the three supervisors of this dissertation, who

offered guidance during the process. The author of this dissertation, first author in

all three publications, was responsible for empirical data collection, data analysis,

theoretical grounding and writing.

16

17

Contents Abstract

Tiivistelmä

Acknowledgements 9 Abbreviations 13 List of original publications 15 Contents 17 1 Introduction 19 2 Theoretical framework 27

2.1 Collaborative learning ............................................................................. 27 2.2 Electrodermal activity ............................................................................. 29 2.3 Arousal .................................................................................................... 33

2.3.1 Cognition, affect, and arousal ....................................................... 34 2.3.2 Arousal and performance ............................................................. 38 2.3.3 Arousal in the classroom .............................................................. 39

2.4 Interpersonal physiology ......................................................................... 40 2.4.1 Previous work on interpersonal physiology in

collaborative learning ................................................................... 42 2.4.2 Arousal contagion ......................................................................... 43 2.4.3 Physiological coupling indices ..................................................... 44

2.5 Research gap ........................................................................................... 46 3 Aims 49 4 Methods 51

4.1 Participants .............................................................................................. 51 4.2 Materials ................................................................................................. 52

4.2.1 The Empatica® E3 and E4 wristbands ......................................... 52 4.2.2 Performance instruments .............................................................. 54 4.2.3 Online learning environments ...................................................... 54 4.2.4 Questionnaires .............................................................................. 55

4.3 Procedure ................................................................................................ 56 4.3.1 Data collection I ........................................................................... 56 4.3.2 Data collection II .......................................................................... 57

4.4 Data analysis ........................................................................................... 58 4.4.1 Electrodermal activity .................................................................. 58 4.4.2 Performance .................................................................................. 63 4.4.3 Questionnaires .............................................................................. 64

18

4.5 Research evaluation ................................................................................ 64 4.6 Ethics ....................................................................................................... 67

5 Overview of the articles 69 5.1 Article I: Investigating collaborative learning success with

physiological coupling indices based on electrodermal activity ............. 69 5.2 Article II: Profiling sympathetic arousal in a physics course:

How active are students? ......................................................................... 70 5.3 Article III: Sympathetic arousal commonalities and arousal

contagion during collaborative learning: How attuned are triad

members? ................................................................................................ 71 6 Main findings and discussion 73

6.1 Triad arousal levels in collaborative learning .......................................... 73 6.2 Arousal during exams and achievement .................................................. 75 6.3 Pairwise directional agreement and dual learning gain ........................... 76 6.4 Within-triad high arousal contagion in collaborative learning ................ 77

7 Conclusions 79 List of references 83 Original publications 99

19

1 Introduction Successful collaborative learning endeavors have long been promoted and pursued

by educational institutions at national and international levels as a way to attend to

students’ academic needs and future employability as well as to the practical

consideration of sharing scarce resources (e.g., computers and laboratory

equipment; M. Baker, 2015). Collaborative learning has been used in instruction to

a greater or lesser extent since at least the second half of the 19th century (Johnson

& Johnson, 2002), as it is widely believed that individual outcomes (e.g., learning

gain) will be favored by mutual intellectual engagement (Kuhn, 2015).

Collaborative learning has been actively and extensively studied since it entered

the educational research agenda in the 1970s (M. Baker, 2015), spurred by

awareness of societal relevance and embraced by a vibrant community of learning

scientists (Hoadley, 2018). Apart from the desired higher order thinking, an early

review (Slavin, 1980) reported other benefits of collaborative learning such as the

development of prosocial behavior. Furthermore, efficient collaboration is among

those skills beyond domain knowledge that are permanently in high demand in

today’s extremely specialized labor market. Specialization translates into

distributed expertise that, when synergistically combined, constitutes a competitive

advantage for organizations. Thus, organizations need expert collaborators or team

players to thrive in a highly competitive global ecosystem. Such orientation to

collaboration is expected to have already been developed as students advance

through their academic years. Therefore, the development of collaborative learning

has been embraced and promoted by educational stakeholders including

practitioners, organizational administrators, and policy-makers.

Shaped by theoretical and/or technological advances, as efforts to capitalize on

the affordances of collaborative learning, a variety of methods has been applied

across the decades in schools such as jigsaw (Aronson, Blaney, Sikes, & Snapp,

1978; Aronson & Patnoe, 2011), structured academic controversy, reciprocal

teaching, student teams achievement division, and others cited by Koschmann

(1996), such as expeditionary learning, group investigation, problem-based

learning, and project-based learning, as well as several forms of small-group

learning (A. L. Brown, 1992; Slavin, 1980; Webb, 1991). Based on societal needs,

it is no surprise not only that collaborative learning pedagogies have thrived, but

also that they are increasingly emphasized in educational systems through the

design of curricula and instruction (OECD, 2017).

20

Before going any further, it is important to acknowledge that the terms

collaborative and cooperative learning have sometimes been used interchangeably,

while at other times an explicit distinction is made between them (M. Baker, 2015;

Dillenbourg, Baker, Blaye, & O’Malley, 1996). This dissertation follows the

distinction posed by Dillenbourg (1999), in which cooperation is seen as labor

division (i.e., partners split the work, solve sub-tasks individually, and then

assemble the partial results into the final output), while in collaboration, partners

do the work together. Aligned with that distinction, this dissertation adopts the most

widely used definition of collaboration (Sanna Järvelä et al., 2015), that of

Roschelle and Teasley (1995, p. 70, emphasis added), which states that

“collaboration is a coordinated, synchronous activity that is the result of a

continued attempt to construct and maintain a shared conception of a problem.” It

is acknowledged also that in real-life scenarios, collaboration and cooperation are

hardly separable, given the nature of group work, which often requires a

combination of the two (Jeong & Hmelo-Silver, 2016). Nonetheless, the focus here

is on the joint work of learners and, therefore, on collaborative learning, understood

as previously defined.

In terms of learning, the process of collaboration is as important as the outcome,

since the aim is often not only that students reach a correct problem solution, but

also that they acquire the skills to tackle such problems more efficiently in the

future, whether together or alone, in virtue of the appropriation of a co-elaborated,

deeper task-domain conceptual understanding (M. Baker, 2015). The success of the

collaborative endeavor, however, cannot be taken for granted. Neutral and negative

effects of collaborative learning have also been reported in the scientific literature

(P. A. Kirschner, Sweller, Kirschner, & Zambrano, 2018; Kuhn, 2015; Mullins,

Rummel, & Spada, 2011). On the negative side, for example, it has been found that

low achievers progressively become passive when collaborating with high

achievers (Dillenbourg et al., 1996), that groups sometimes engage in destructive

interactions (Cohen, 1994), including interpersonal aggression (M. Baker, 2015)

and blaming each other for making mistakes (Miyake & Kirschner, 2014), and that

collaboration may even lead to a decline in thinking quality due to overconfidence

produced by the group interaction (Kuhn, 2015). The complexity of the outcomes

of collaborative learning, ranging from the positive and highly desirable to the

negative and detrimental for learning, has naturally attracted the interest of

educational researchers, who want to understand the phenomenon so that,

ultimately, collaboration can be properly structured and adequate interventions can

be designed.

21

Three paradigms have been used to study collaborative learning: the effect

paradigm, the conditions paradigm, and the interactions paradigm (see Dillenbourg

et al., 1996). In the effect paradigm, the question is whether collaborative learning

is more efficient than learning alone. Positive outcomes have largely dominated the

results, but some of the negative effects previously mentioned are stable and well

documented (Dillenbourg et al., 1996). The variability in the findings led

researchers to investigate under what conditions collaborative learning is efficient,

which is the question the conditions paradigm tries to answer. When it comes to the

quality of problem solutions and learning outcomes, whether working together is

more effective than working alone depends indeed on factors such as the task-

complexity, group size, differences in prior knowledge, age, team-member

familiarity, and social skills (M. Baker, 2015; P. A. Kirschner et al., 2018). The

many independent variables in this paradigm (i.e., the conditions) were shown not

to have simple effects on the learning outcomes but instead to interact with each

other in a complex, virtually inextricable way (Dillenbourg et al., 1996). Research

then moved to explore intermediate variables that describe the interactions among

group members to understand sources of variability in collaborative outcomes

(Barron, 2000). The interactions paradigm, with higher granularity than its

predecessors in terms of unit of analysis, is concerned with the questions of which

interactions occur under which conditions and what effects these interactions do

have (Dillenbourg et al., 1996). It was found that the simple frequency of

interaction does not predict individual achievement in collaborative learning

(Cohen, 1994), pointing to the need of characterizing the interactions.

Theories of learning (e.g., socio-cultural and situated cognition theories)

relocating the attention from the individual to the activity structures in which

learning occurs, highlighted the role of discourse in conceptual development

(Barron, 2000). As an object of study, communication provides a window into the

cognitive and social processes related to collaborative skills, such as grounding,

mutual goal establishment, progress toward goals, and sharing perspectives (Kuhn,

2015). Types of verbal interactions of interest as learning mechanisms were, for

example, negotiation, considered an indicator of joint involvement in task solutions,

and argumentation, seen as a possible means for resolving socio-cognitive conflict

(Beers, Boshuizen, Kirschner, & Gijselaers, 2007; Dillenbourg et al., 1996). The

study of productive verbal communication has made extensive use of conversation

analysis (Sacks, 1972), a methodology from the sociological research tradition

which continues to be an important tool to analyze and make detailed meaning of

interactions in collaborative learning (Stahl, Koschmann, & Suthers, 2014). Audio

22

and video recordings, together with the development of specialized software, have

enabled code and count analysis of content data, which accounts for much of the

field’s analytic efforts (Jeong, Hmelo-Silver, & Yu, 2014). However, those analyses

are time-consuming, very expensive microgenetic methods and do not scale up well

(Wise & Schwarz, 2017). Steps toward automation of content analysis have been

taken using natural language processing technologies (Blikstein & Worsley, 2016;

Kelly, Thompson, & Yeoman, 2015), but analyses of verbal data are often complex

(Chi, 1997) for both human coders and machines. In addition to the cost and

limitations of content analysis, and although verbal communication has

understandably taken the lion’s share of process analysis focus in collaborative

learning, researchers acknowledge that, to properly understand the collaborative

process, it is also necessary to take into account other features integral to the

collaboration process (e.g., nonverbal communication; M. Baker, 2015).

Conversational processes alone do not explain the effects observed (Dillenbourg et

al., 1996).

The complex processes of collaboration present a challenge for consistent,

accurate, and reliable measurement across individuals and across user populations

(OECD, 2017). A wide range of data sources has been used to different extents to

study collaborative learning, with increased reliance on questionnaires, interviews,

and observation (Jeong et al., 2014). However, process-oriented approaches,

encouraged by the interactions paradigm, have made patent the need for new tools

for analyzing and modelling interactions (Stahl et al., 2014). Collaborative learning

is a process that occurs over time, but quantitative research has traditionally focused

mostly on relationships between relatively static input and outcome variables (Wise

& Schwarz, 2017). There is room for development, so the field needs to explore a

larger repertoire of analytic strategies and data modalities (Jeong et al., 2014; Vogel

& Weinberger, 2018), especially to address the need for real-time assessment

measures (S. Dawson & Siemens, 2014).

In this context, novel technologies, innovative developments, and applications

of established technologies are opening up attractive research opportunities, with

wearable sensors being a prominent example. Wearable sensors are increasingly

part of everyday life as standalone devices or are embedded into accessories, such

as smartwatches, mostly for sports and fitness activity tracking (Swan, 2012). But

the affordances of wearable sensors for continuous and unobtrusive measurement

of physiological responses and human behavior have found research applications

in such areas as healthcare (e.g., Koskimäki et al., 2017; Pentland, 2004) and

marketing (e.g., Bettiga, Lamberti, & Noci, 2017), among other fields. In learning

23

research, the affordances to study cognitive-affective dimensions have also been

recognized (Schneider, Börner, van Rosmalen, & Specht, 2015), as well as the

“unprecedented insight into the minute-by-minute development” of social

interactions (Blikstein & Worsley, 2016, p. 222). Wearable sensors, such as

necklace microphones, can be used as sociometric badges (Kim, Mcfee, Olguin,

Waber, & Pentland, 2012) to track the structural features of verbal communication,

including speech overlapping, distribution, and contributions from individual team

members. The biosensor type of wearables provide access to physiological

responses, whose capability to make salient psychological cognitive and affective

processes is the target of study of the field of psychophysiology (Cacioppo &

Tassinary, 1990b).

Fig. 1. Nervous system simplification.

One of the most widely used physiological responses in psychophysiology is

electrodermal activity (EDA), which refers to the variation in the electrical

conductivity of the skin (M. E. Dawson, Schell, & Filion, 2017; Kreibig, 2010).

The electrodermal system has a particularity. This is shown in the simplified

scheme of the different branches of the nervous system displayed in Fig. 1,

representing how the electrodermal system is innervated only by the sympathetic

nervous system, as compared to other systems, such as the cardiovascular and the

respiratory systems, which have both sympathetic and parasympathetic innervation.

Furthermore, the electrodermal system is the only one in the human body with this

feature of exclusive sympathetic innervation, which is why EDA is considered a

pure marker of sympathetic activation (i.e., sympathetic arousal; Boucsein, 2012).

Arousal refers to the degree of physiological activation and responsiveness

triggered by an event, object, or situation during a person’s interaction with the

24

environment (De Lecea, Carter, & Adamantidis, 2012; Juvina, Larue, & Hough,

2018). The psychophysiological value of arousal lies in the indirect access it

provides to cognitive and affective processes (Mandler, 1984). Arousal increases

with the cognitive process of attention (Raskin, 1973; Sharot & Phelps, 2004),

which has been explained through evidence pointing to arousal and attention as

sharing the same neural mechanisms (Critchley, 2002). Arousal is also the

activation dimension of emotions according to the circumplex model of affect (see

Fig. 2 in Sub-chapter 2.3.1), which conceives emotions as states characterized by

arousal and valence (i.e., the degree of [un]pleasantness) components (Russell,

1980).

The opportunities are obvious, but wearable biosensors have not yet become

popular in learning research (Blikstein & Worsley, 2016). Reasons that might

prevent the use of physiological data can be related, first, to the technological

equipment needed to collect them (e.g., the equipment can be expensive, or

sometimes intrusive), and second, to the substantial domain-specific knowledge

base needed to analyze and interpret them (Palumbo et al., 2017).

This dissertation utilizes the Empatica® E3 and E4 multi-sensor wristbands

(Empatica Inc., Cambridge, MA, USA), devices specifically designed for research

as opposed to consumer-oriented devices in the market with lower accuracy

(Garbarino, Lai, Tognetti, Picard, & Bender, 2014), to measure EDA and calculate

a number of derived indices of sympathetic arousal and interpersonal physiology,

in collaborative learning situations in classroom and classroom-like environments.

The coordination and synchronicity characteristics of collaborative learning

processes (Roschelle & Teasley, 1995) are also hypothesized to be reflected in

mutual influence and shared physiological states among collaborators (Palumbo et

al., 2017). Accordingly, there is a need to investigate the characteristics of

physiological group responses, such as homogenous activation levels and

simultaneous or sequential physiological changes (Knierim, Rissler, Dorner,

Maedche, & Weinhardt, 2018). Moreover, physiological metrics of collaboration

can be considered universal, as they are objective indices (Chanel, Kivikangas, &

Ravaja, 2012; Henning, Armstead, & Ferris, 2009; Montague, Xu, & Chiou, 2014;

Picard, Fedor, & Ayzenberg, 2016; Swan, 2012) that are independent of language

and culture.

Through three scientific articles, this dissertation explores EDA and derivative

measures, such as sympathetic arousal and physiological coupling indices (PCIs),

concomitant with individual and interpersonal cognitive and affective processes,

respectively, during collaborative learning in a naturalistic setting. Article I studies

25

PCIs (i.e., group measures attending to different parameters) based on EDA, in

relation to subjective and performance-related measures of collaboration. Article II

provides a profile of the degree of physiological activation (i.e., arousal) in the

classroom, together with the relationship of arousal to achievement on the exam.

Finally, Article III explores common arousal levels and possible arousal contagion

in classroom groups.

The structure of this dissertation follows. After this introduction, Chapter 2

lays out the theoretical background for the work, which involves collaborative

learning research, EDA, arousal, and interpersonal physiology. The theory chapter

ends with a sub-chapter reflecting on the research gap from different fields, and on

the combination of need and opportunity motivating this work. Chapter 3 concisely

presents the aims of the dissertation, which shape the methodology detailed in

Chapter 4. The latter concludes with two sub-chapters addressing the research

evaluation and ethical considerations. Chapter 5 provides an overview of the three

articles on which this dissertation is based, connecting how each article contributes

to the overall aims. Also aligned with the aims, the main findings are presented and

discussed in Chapter 6. Finally, concluding remarks including significance of the

work, implications, applications, and future work, are presented in Chapter 7.

26

27

2 Theoretical framework With the integration of different theoretical frameworks, namely, the theories

behind collaborative learning, arousal theory from general psychology, and EDA

from the field of physiology, in addition to computer science approaches to

physiological data processing, this work shows the commitment and will to keep

advancing the interdisciplinarity that has characterized the learning sciences since

their early days (Hoadley, 2018), which is a staple of collaborative learning

research as well (Stahl et al., 2014).

2.1 Collaborative learning

Three leading theoretical frameworks for collaborative learning have been the

socio-constructivist approach, which builds on the work of Jean Piaget and his

followers; the socio-cultural approach, rooted in the work of Vygotsky; and the

shared or situated cognition theory (see Dillenbourg et al., 1996). The socio-

constructivist approach considers that learning is the result of disagreement

between two points of view in social interaction, when such a socio-cognitive

conflict leads to the coordination of points of view, resulting in an enhanced

understanding at the individual level (Doise & Mugny, 1984). In the socio-cultural

approach, communication rather than disagreement is seen as the mechanism of

learning, with inter-psychological processes happening in the interaction and then

being assimilated at the intra-psychological level (Wertsch, 1991). According to the

socio-cultural tradition, communication goes iteratively from a dialogue with peers

to an internal dialogue with oneself, the reflection part, when the new knowledge

is then internalized (Vygotsky, 1978). The two theories acknowledge the role of

both the group and the individual in the process of learning and the bidirectional

influence between both roles (Butterworth, 1982).

Collaboration is considered to be conducive to higher order thinking skills

through effective communication (Cohen, 1994). Verbal interactions activate

several cognitive processes, including perception, comprehension, information

processing, representation, and anticipation, among others (Ahn et al., 2018).

According to Bruer (1994), dialogue might lead to deeper cognition because it

involves both the comprehension and production of language, the latter more

cognitively demanding than the former. Bruer sees dialogue and discussion as

vehicles of learning, which not only lead to higher order processing, but also allow

students to share and distribute the cognitive load so that no individual working

28

memory is required to hold all the information needed to solve a problem (F.

Kirschner, Paas, & Kirschner, 2011). Apart from alleviating cognitive load, Bruer

reasons that collective working memory renders encoded information more

accessible, since the same knowledge may have a different representation in each

participant’s long-term memory, and a wider repertoire of cues are then available

to the group to trigger information retrieval from individual memories. In other

words, collective working memory reduces cognitive load, facilitates positive

interdependence, and increases information availability through redundancy (van

Bruggen, Kirschner, & Jochems, 2002).

Collaborative processes thought to be beneficial for learning include, but are

not limited to, knowledge co-construction (Jeong & Hmelo-Silver, 2016),

transactivity or building on each other’s reasoning (Weinberger, Stegmann, &

Fischer, 2007), and argumentation (M. Baker, 2015). Other support processes are,

for example, the identification of shared knowledge (Roschelle & Teasley, 1995),

the establishment of common ground (Reiter-Palmon, Sinha, Gevers, Odobez, &

Volpe, 2017), mutual modelling of the collaborative partner’s knowledge state

(Dillenbourg, 1999), and coordinating joint efforts (Järvenoja, Järvelä, &

Malmberg, 2015). All the previous processes implicitly highlight the notion of

synchronicity and simultaneity presumed in collaborative learning, as opposed to

cooperative learning, which is characterized by individually solving sub-tasks

through division of labor (Dillenbourg, 1999). Joint and/or together are the

keywords that prominent researchers in the field consistently use to define

collaborative learning or associated processes. Such descriptions include (emphasis

added in all cases) “working together” (Cohen, 1994), “mutual engagement of

participants in a coordinated effort to solve the problem together” (Roschelle &

Teasley, 1995), the notion of “joint attention as social cognition” (Tomasello, 1995),

partners doing the work together (Dillenbourg, 1999), “a jointly produced activity”

(Enyedy & Stevens, 2014), “an active and joint process” (M. Baker, 2015), and

“working together toward a shared learning goal” (Jeong & Hmelo-Silver, 2016),

among others. Accordingly, shared cognition is regarded as a critical team state

(Reiter-Palmon et al., 2017).

As students learn collaboratively, they often create a joint product, artifact, or

solution to the task at hand, and upon a justified and successful process, individual

domain knowledge is expected to be enhanced (i.e., learning gains; Rummel, 2018).

In this dissertation, the joint product and the pairwise learning gain (i.e.,

aggregation of the individual learning gains) are referred to as the collaborative

learning product and dual learning gain, respectively.

29

Collaboration can only occur if the group members strive for building and

maintaining a shared understanding of the task and its solutions (Barron, 2000). In

other words, the students’ attitudes toward collaboration, that is, their motivation

to constructively engage in collaborative work, become important factors in

predicting the process success (Cohen, 1994). In this dissertation, the term

collaborative will is used to designate the students’ attitudes toward collaboration.

2.2 Electrodermal activity

EDA refers to “electrical changes across the skin in areas of the body that are

psychologically responsive” (Roy, Boucsein, Fowles, & Gruzelier, 1993, p. v).

With a history of more than a century, a number of features have contributed to

make EDA one of the most widely used response systems in the trajectory of

psychophysiology (M. E. Dawson et al., 2017; Kreibig, 2010), the discipline

concerned with the measurement of physiological changes in order to increase

understanding of psychologically relevant phenomena (Cacioppo & Tassinary,

1990b; Lyytinen, 1984). A listing of such features follows:

1. EDA is easy to measure with a properly spaced pair of electrodes (Boucsein,

2012). So much so, that in its early decades, it was common for workers in this

field to use equipment built in their own laboratories, until a commercially

manufactured apparatus became more available (Venables & Christie, 1973).

2. The techniques used to record EDA are completely harmless and risk-free (M.

E. Dawson et al., 2017).

3. Electrodermal changes can be easily detected by eye (G. E. Schwartz &

Shapiro, 1973). Today, advanced sophisticated algorithms based on the

mathematical operation of deconvolution (Benedek & Kaernbach, 2010b)

allow for the offline biosignal processing of continuous EDA recordings, but

before the invention of the computer, the simple quantification of responses in

stimulus-response experimental paradigms proved convenient (G. E. Schwartz

& Shapiro, 1973).

4. EDA is considered a remarkably sensitive measure because it changes with low

levels of psychological and affective states and may vary with the intensity of

the state (Thorson, West, & Mendes, 2018).

5. EDA is relatively inexpensive in terms of the hardware and software required,

especially when compared to eye tracking, brain-computer interfaces and facial

30

expression recognition (Gonzalez-Sanchez, Baydogan, Chavez-Echeagaray,

Atkinson, & Burleson, 2017).

6. The exclusive innervation of the sweat glands by the sympathetic nervous

system provides a relatively direct and undiluted representation of sympathetic

activity (M. E. Dawson et al., 2017). The sympathetic nervous system is the

control system for the so-called fight-or-flight response, which reflects a state

of activation, thus, the use of EDA to index arousal. Conversely, the

parasympathetic nervous system is responsible for bringing the body back to

homeostasis, the state of equilibrium corresponding to the so-called rest-and-

digest response, thus counteracting arousal. In other words, EDA is a clean

measure of sympathetic arousal without parasympathetic interference.

These features of EDA, together with the vast amount of research that EDA has

been the object of, have led to its long-term consideration as a well-validated,

widely accepted, and perhaps the most readily accessible and simplest indicator of

sympathetic arousal (Boucsein, 2012; Critchley, 2002; Wang, 1957). Altogether,

EDA has been regarded, not surprisingly, as a valuable addition to the toolkit of

scientists, especially for the those tackling psychophysiological problems (G. E.

Schwartz & Shapiro, 1973).

EDA history dates back to the 19th century. By 1840, it was generally accepted

that the body’s electrical characteristics provided a basis for diagnosis and therapy,

as part of the view of vital processes as electrical (E. Neumann & Blanton, 1970).

Vigouroux (1879), a French electrotherapist and electro-diagnostician, was the first

to describe the changes in the electrical resistance of the skin. Féré (1888), usually

credited with the discovery of the electrodermal reflex, demonstrated that rapid

fluctuations of skin resistance could occur in response to emotional stimulation

(Edelberg, 1972b). The report by Féré appears to be the first attempt to use EDA to

index a psychological construct (Raskin, 1973). Since then, the amount of work on

EDA has been extensively reviewed to summarize and update the understanding of

the electrodermal mechanisms, as such, and their applications to

psychophysiological research (Boucsein, 1992, 2012; Critchley, 2002; M. E.

Dawson, Schell, & Filion, 1990, 2007; M. E. Dawson et al., 2017; Edelberg, 1967;

E. Neumann & Blanton, 1970; Venables & Christie, 1980; Venables & Martin,

1967). There have been periods of development and periods of stagnation, the latter

owing, at least in part, to the use of inadequate or poorly understood techniques

(Venables & Christie, 1973).

31

It is important to be aware of different terminologies that have been used to

refer to the EDA phenomenon. These terminologies not only can be found in

today’s scientific literature, but also in classic and still highly relevant work on the

topic. Although the formal history of EDA dates back to 1879 as previously

indicated, according to Boucsein (2012, p. 2), the term EDA was first introduced

by Johnson and Lubin in 1966 “as a common term for all electrical phenomena in

skin, including all active and passive electrical properties which can be traced back

to the skin and its appendages.” Other well used terms, at the time, were

psychogalvanic response (PGR) and galvanic skin response or reflex (GSR). PGR

was coined by Veraguth in 1908 when, oblivious to the three previous decades of

work on EDA, he believed that he had discovered a new reflex (E. Neumann &

Blanton, 1970). As for the GSR, its utilization has been long advised against due to

ambiguity in its early use on the one hand, and to the findings that multiplicity and

complexity of the EDA phenomena (e.g., spontaneous responses and

psychologically elicited changes) cannot be explained simply as a galvanic element,

on the other hand (Boucsein, 2012; Venables & Christie, 1973). Accordingly, EDA

is now the preferred term over GSR and PGR, for variations in electrical

conductance of the skin, including phasic changes that result from sympathetic

neuronal activity (Critchley, 2002; M. E. Dawson et al., 2017). EDA is therefore

the term used in this dissertation. Notwithstanding, it has to be mentioned that some

authors, though acknowledging that GSR is now less often used, consider it “still

accepted” (M. S. Schwartz, Collura, Kamiya, & Schwartz, 2017).

The physiological basis of EDA has been ascribed to muscular, vascular, and

secretory mechanisms, causing a debate which mounting evidence settled several

decades ago in favor of the secretory theory (Venables & Christie, 1973). Today, it

is known that the eccrine sweat glands, uniquely innervated by the sympathetic

nervous system, are the primary determinants of EDA (Boucsein, 2012; Critchley,

2002). The eccrine sweat glands are coiled, tubular structures that extend from the

epidermis to the lower dermis and secrete water, electrolytes, and mucin (i.e., sweat,

basically; Critchley, 2002). Eccrine and apocrine are the two forms of sweat glands

in the human body, but the former are of primary interest in psychophysiology

because, although their primary function is thermoregulation, all eccrine glands are

believed to be involved in psychological sweating (M. E. Dawson et al., 2017). The

distribution of eccrine glands, however, is not homogeneous over the entire skin.

Their highest density (approximately 400/mm2) occurs in palmar and plantar (or

sole) regions (Critchley, 2002). This means, on the one hand, that EDA measures

in different body parts are not comparable, and on the other hand, that the palm and

32

the sole are the best areas to record EDA. However, in ecologically valid studies,

the placing of electrodes on such regions interferes with the subjects’ activities and

is therefore not practical. Consequently, alternatives have been explored and more

practical locations for daily life measurements such as the wrist have been found

viable (Poh, Swenson, & Picard, 2010; van Dooren, de Vries, & Janssen, 2012).

This dissertation employs non-dominant wrist measurements of EDA via an

unobtrusive watch-like research-quality off-the-shelf wristband (see Garbarino et

al., 2014).

The EDA signal is the combination of a tonic, slow-varying component and a

phasic, fast-changing component superimposed on the tonic level (Boucsein, 2012).

The tonic component accounts for the level of skin conductance in the absence of

any particular discrete environmental event or external stimuli, and it is

denominated as the skin conductance level (SCL), after a suggestion by Venables

and Martin (1967). The phasic component, of wider use in this dissertation,

resembles peaks, any of which are referred to as skin conductance response (SCR),

a term coined by Burch and Greiner in 1958 according to Edelberg (1967, p. 44)

and endorsed by the Society for Psychophysiological Research in 1967 (Venables

& Christie, 1973). SCRs rise steeply to their peak and decline slowly, following an

exponential-like decay, to the baseline (Benedek & Kaernbach, 2010b). The

succession of SCRs usually results in a superposition of subsequent SCRs, as one

SCR arises on top of the declining trail of the preceding one (Boucsein, 2012).

Today, advanced deconvolution-based algorithms are available to decompose

superimposed SCRs in a way that enhances their quantification as compared to

previous historical approaches (Benedek & Kaernbach, 2010b).

It was not an interest in the phenomenon itself which led to the high volume

and rapid development of research on EDA, but the possibility of applications to

medical diagnosis and to studies of mental processes of interest in a wide variety

of fields (E. Neumann & Blanton, 1970). For example, Poh et al. (2010, pp. 1243–

1244) cite potential clinical applications of EDA which have been researched, such

as screening for cystic fibrosis, classification of depressive illnesses, prediction of

functional outcome in schizophrenia, discrimination between healthy and psychotic

patients, characterization of sympathetic arousal in autism, early diagnosis of

diabetic neuropathy, and providing biofeedback in treating epileptic seizures.

Edelberg (1967) cites clinical explorations for dermatological research and

neurological diagnosis. Although the clinical applications are many, those outside

medical settings (i.e., of interest in psychophysiological research) are not fewer.

EDA has been largely explored in relation to cognitive and affective processes,

33

including, but not limited to, attention (Critchley, 2002; Thorson et al., 2018),

cognitive load measurement (Boucsein & Backs, 2000; Shi, Ruiz, Taib, Choi, &

Chen, 2007), memory retention (Critchley, Eccles, & Garfinkel, 2013), stress

detection (Visnovcova, Mestanik, Gala, Mestanikova, & Tonhajzerova, 2016),

engagement (Gillies et al., 2016), decision-making and judgment (Finger &

Murphy, 2011), self-efficacy (McQuiggan, Mott, & Lester, 2008), goal orientation

(Edelberg, 1972a), the state of flow (Knierim et al., 2018), and the study of the

social psychology of human trust (G. E. Schwartz & Shapiro, 1973). Some of those

processes in the cognitive and affective domains have led to incorporating EDA

into intelligent tutoring systems (Cooper et al., 2009; Strauss et al., 2005) and

intelligent interruption management or interruption notification mechanisms

(Goyal & Fussell, 2017; Knierim et al., 2018).

Yet, as diverse as the interpretations might seem in the cognitive and affective

domains, they all share a common denominator: arousal. Whether explicit or

implicit, the notion of arousal, in particular sympathetic arousal, is present

whenever EDA is being used for psychophysiological purposes, since EDA and

arousal are inextricably linked. A reminder of the underlying rationale follows. The

skin, an organ whose activity is electrically traceable (i.e., EDA), is the only organ

of the entire human body exclusively innervated by the sympathetic nervous system,

which is the controller of the so-called “fight-or-flight” response (i.e., preparing the

body for action, whether physical or intellectual). Consequently, EDA enables

access to sympathetic activity. In other words, in psychophysiological research,

above all, EDA is an index of sympathetic arousal (Boucsein, 2012; Critchley, 2002;

Wang, 1957).

2.3 Arousal

The interest in the study of arousal arises from its role in cognitive and affective

processes (see Sub-chapter 2.3.1) to the extent that it has an impact on performance

(see Sub-chapter 2.3.2). The construct of arousal is at the center of the connection

of body and mind (i.e., cognitive-affective function; Hanoch & Vitouch, 2004;

Mandler, 1984). The notion of arousal was developed hand in hand with the early

work on EDA. According to Neumann and Blanton (1970), Féré (1888) initially

studied the EDA phenomenon as an arousal indicator. He demonstrated that rapid

fluctuations of skin electrical resistance could occur in response to emotional

stimulation (Edelberg, 1967). Thereafter came “a long period of neglect” (Raskin,

1973, p. 126), until further work developed the arousal theory (e.g., Duffy, 1962;

34

Hebb, 1955). Although arousal is a term of widespread use in psychology, other

terms have been used as synonyms in the scientific literature, including activation,

energy mobilization, excitation, energy, tension, and activity (Duffy, 1962; Russell

& Barrett, 1999).

While arousal measures the degree of physiological activation, and it would

therefore be recorded more objectively and precisely through physiological

responses, it has to be noted that different scales have been developed to measure

it (Bettiga et al., 2017). Such scales might be convenient when biosensors are not

available, but research has shown that there is often a mismatch between the two

measures (i.e., self-reported scales vs. biosensors). Accordingly, Reisenzein (1983)

proposes making a distinction between physiological arousal per se and perceived

arousal. Arousal in this dissertation is measured physiologically, as indexed

through EDA.

Physiological arousal, in turn, is a complex response which manifests through

a variety of central (i.e., electrocortical), motor (i.e., muscle tension), autonomic

(e.g., sweating, heart rate, and blood pressure), and endocrine (i.e., hormonal)

responses (American Psychiatric Association, 2000). For example, an increase in

muscle tension usually goes hand in hand with an increased SCL (Frijda, 1986).

However, it should not be assumed that the arousal response in each of the different

participating systems is necessarily equivalent, or even related. No correlation has

been found between electrocortical and autonomic arousal (Frijda, 1986), and the

heart has been reported to slow down even as other indices (e.g., pupil size) suggest

an increase of arousal (Kahneman, 1973). In general, the different indices of

physiological arousal, whether electrocortical, biochemical, electrodermal,

cardiovascular, or skeletal motor, have been shown to correlate poorly, and the term

directional fractionation has been chosen to refer to the different patterning (Lacey,

1967). Basically, directional fractionation means that different systems involved in

the arousal response show a different pattern, which might even be in different

directions (i.e., one response increasing while the other is decreasing). Accordingly,

separate types of arousal have been proposed (e.g., sympathetic and electrocortical),

which should not be mixed (Frijda, 1986). Sympathetic arousal as indexed by EDA

is the type that is the focus of this work.

2.3.1 Cognition, affect, and arousal

Arousal has a bidirectional relationship with cognitive and affective processes

(Mandler, 1982). Cognitive and affective drivers have a reflection in arousal levels

35

(Critchley, 2002; Poh et al., 2010), but, in turn, changes in arousal inform cognition

(Critchley et al., 2013; Critchley & Garfinkel, 2018) and affect (Reisenzein, 1983;

Schachter & Singer, 1962). It is known, for example, that people rely partly on

information from their arousal states in judging their capabilities (Bandura, 1982),

and that arousal has an effect on human behavior (Damasio, Tranel, & Damasio,

1991). Thus, the psychology and physiology of arousal have a reciprocal influence

on one another. Parenthetically, arousal is also influenced by physical activity and

by the intake of stimulant and depressant substances (e.g., caffeine; Hanoch &

Vitouch, 2004; Russell & Barrett, 1999), but, obviously, the cognitive and affective

drivers are the main interest in psychophysiology, as well as to this work where the

focus is on classrooms and classroom-like spaces as students learn science.

For the cognitive part, there is a close interdependence between arousal and

attention (Kahneman, 1973; Raskin, 1973; Robbins, 1997; Sharot & Phelps, 2004),

which naturally extends to other cognitive processes that presuppose attention, such

as memory and decision-making (Garfinkel, Critchley, & Pollatos, 2017). Arousal

assists in the mental sharpening that enhances attention to and the processing of

potentially important information (Critchley et al., 2013). Empirically, arousal has

been shown to increase with attention in relation to eye-catching, engaging stimuli,

and attention-demanding tasks (Poh et al., 2010). Such a relationship might be

explained by indications from clinical studies that a common neural substrate may

be shared by arousal and attention mechanisms (Critchley, 2002). By modulating

the selectivity of attention and then increasing attentional time on arousing stimuli,

arousal can delay disengagement and influence memory encoding (Fox, Russo,

Bowles, & Dutton, 2001). In addition, arousal at the time of memory encoding is

theorized to facilitate memory retrieval, more so for individuals with enhanced

capabilities to be aware of their arousal and internal bodily states in general, known

as interoceptive awareness (Garfinkel et al., 2017). Emotionally arousing incidents

and stimuli have been shown to be much better remembered than those without

emotional relevance (Marchewka et al., 2016). The representation of bodily

changes in arousal may mediate the feeling of knowing ascribed to some familiar

memories (Garfinkel et al., 2017). In terms of decision-making, interoceptive

signals are said to shape rapid decision-making in stressful contexts (Critchley et

al., 2013; Critchley & Garfinkel, 2018). The role of arousal in emotions, discussed

next, has been at the center of its consideration as a well-suited tool in decision-

making and judgment research (Finger & Murphy, 2011).

Arousal and valence are the two dimensions of emotions according to the

classic circumplex model of affect (Russell, 1980). The model is represented

36

graphically in rectangular coordinate axes as shown in Fig. 2. The arousal axis

(vertical) represents a continuum from deactivation to activation (bottom-up),

whereas the valence axis (horizontal) represents a continuum from

negative/unpleasant to positive/pleasant (left-right). The notion of arousal as an

essential component of emotions, however, is much older than the model. The

construct of arousal was born together with EDA research through the realization

that fluctuations of EDA could occur in response to emotional stimulation

(Edelberg, 1967; Féré, 1888; E. Neumann & Blanton, 1970). Later, when the theory

of arousal or activation (Duffy, 1962) was further developed, the “general level of

energy mobilization” was included as “a major aspect of what has been historically

labeled as emotion” (Raskin, 1973, p. 126). Around the same time, Schachter and

Singer (1962) formulated their classical cognition-arousal theory of emotion, also

known as cognitive-physiological theory of emotion, two-factor theory of emotion,

or simply the Schachter theory of emotional states (Reisenzein, 1983). According

to the theory, an emotional state is the result of the cognitive appraisal (i.e., labelling)

of the state of arousal caused by the emotional stimulus. In addition, arousal

feedback can have an intensifying effect on emotional states, and this arousal-

emotion relationship is mediated, in part, by causal attributions regarding the

source of arousal (Reisenzein, 1983; Schachter & Singer, 1962).

Building on the circumplex model of affect and particularizing for academic

emotions, the quality of students’ information processing (i.e., cognition) and the

types of strategies unfolded by them have been mapped to the four quadrants of the

rectangular coordinate axes formed by arousal and valence (Pekrun, Goetz, Titz, &

Perry, 2002), as shown in Fig. 2. The quadrants have been labelled according to the

two arousal regions divided by the valence axis (“activating” for the top and

“deactivating” for the bottom), and the two valence regions divided by the arousal

axis (“positive” for the left and “negative” for the right). Thus, starting from the

top right and moving clockwise, the four quadrants are positive activating, positive

deactivating, negative deactivating, and negative activating. In the activating region,

positive activating emotions have been shown to elicit flexible, creative learning

strategies, such as elaboration, organization, and critical evaluation, whereas

negative activating emotions, on the other hand, seem to evoke more rigid strategies,

such as simple rehearsal and use of algorithmic procedures (Pekrun et al., 2002).

In the deactivating region, a shallower, superficial processing of information has

been found to predominate for deactivating emotions, whether positive or negative.

37

Fig. 2. Arousal-valence affective space and associated cognitive processing (from Article II, reproduced with permission from John Wiley and Sons).

Such a model of characterizing emotions with arousal and valence dimensions

might prove more useful than distinguishing specific emotions, since the latter

might be affected by insufficient discriminant validity owing to the overlapping

traits in different emotions (Russell & Barrett, 1999). Moreover, the arousal

dimension might be more relevant than the valence dimension when it comes to

learning. Activating states, whether positive (e.g., enjoyment of learning and pride

of success) or negative (e.g., anxiety, frustration, and confusion) seem preferable.

For example, it has been argued that negative activating states, which tend to

receive more attention than negative deactivating states (e.g., boredom), could be

productive for learning if they are properly regulated, and that they might not even

need remediation (R. Baker, D’Mello, Rodrigo, & Graesser, 2010). Conversely, in

the negative deactivating area, for example, boredom may impair learning by

leading to behavioral or mental avoidance strategies when students face tasks that

surpass their capabilities or are not engaging (Pekrun et al., 2002). The case of

boredom is of particular interest because it has been reported as more persistent

than other negative states, including both activating and deactivating ones (R.

Baker et al., 2010).

38

The cognitive processes and affective states closely related to arousal here

discussed, obviously, have an impact on performance too, leading to a transitive

relation between arousal and performance.

2.3.2 Arousal and performance

Known for over a century, the relationship between arousal and performance is

dictated by the so-called Yerkes-Dodson law (Yerkes & Dodson, 1908) and

described by an inverted-U function elaborated by Hebb (1955), as illustrated in

Fig. 3. When the arousal level is low, so is performance. As arousal rises due to

increased attention, interest, and engagement, performance increases up to a level

of optimal arousal for performance at the top of the inverted-U. If arousal keeps

rising to too high of a level because of excessive stress or strong anxiety,

performance is impaired. Such impairment has been explained through a reduction

of the processing capacity available for cognitive evaluation (Mandler, 1982;

Sanbonmatsu & Kardes, 1988). In the same vein, students with too high levels of

test anxiety perform worse on tests, and their overall academic achievement is

lower (Aritzeta et al., 2017). Test anxiety, located in the high arousal, negative

valence quadrant of the circumplex model of affect (see Fig. 2), can reduce working

memory resources and has been shown as the academic emotion most often

reported by the students (Pekrun et al., 2002).

The relationship between performance and arousal has been taken advantage

of, for example, to explore the development of interruption management systems.

Sometimes, external agents (e.g., a teacher) need to give a team either feedback or

additional information required to complete a task, while the task is being executed.

Such an interruption, although relevant, may result in being disruptive to the work.

However, a significant increase has been found in both task-performance and

reported collaboration experience when the information was presented during

arousal acceleration, compared with random interruption (Goyal & Fussell, 2017).

Studies of arousal in relation to performance and to the cognitive processes and

affective states behind performance have been carried out in a variety of settings

(Hanoch & Vitouch, 2004). Next, some related work is presented in the particular

setting of the classroom, including during exams.

39

Fig. 3. Yerkes-Dodson law (Hebb, 1955; Yerkes & Dodson, 1908).

2.3.3 Arousal in the classroom

Though studies of arousal during classroom instruction are rare, some exceptions

follow. Arroyo et al. (2009) used raw EDA from high school students taking their

regular mathematics class over a period of five days, to determine their affective

states in combination with self-reports, a camera for facial expression recognition,

a pressure mouse, and a posture analysis seat. More recently, Gillies et al. (2016)

reported a case study on the in-class arousal of sixth graders during a unit of their

science curriculum, in relation to their beliefs and engagement. They determined

arousal through EDA by averaging the signal within time bins. Calderón (2016)

tracked cardiovascular measures of arousal during two science education course

presentations from two student teachers. However, to the best of our knowledge,

arousal has not been measured throughout an entire course. Furthermore, studies

from some course sections which employed EDA to determine arousal have not

used the frequency of SCRs to operationalize the level of arousal.

When it comes to formal learning situations, exams have received much more

attention than everyday classroom lessons, given the special interest on the relation

between arousal and performance. Exams are challenging situations often

involving stress and anxiety because they have not only psychological implications

40

for the students but also practical consequences for their progress in academic life,

as well as their future. Real exams may produce different responses than

challenging laboratory situations because individuals evaluate exams more

seriously and are under more pressure to do well (Spangler, 1997). Learning

scientists have long been interested in studying students’ responses to exams

(particularly test anxiety) and the implication they have for their performance

(Pekrun et al., 2002). The physiological responses in exam situations have been

studied from the cardiovascular system, the endocrine system, and the immune

system perspectives (Spangler, 1997). However, no exam studies have included

pure measures of sympathetic activity (Spangler, 1997), as would be the case in

EDA measures. Moreover, wristband EDA measures are less invasive and

unobtrusive than the endocrine and immune system measurements involving saliva.

Furthermore, the analysis of saliva to assess those physiological responses requires

the use of medical laboratory equipment, while, for the wristband EDA measures,

once the recording is done, the analysis is carried out using computer software.

As discussed in Chapter 1 and Sub-chapter 2.1, collaborative learning

pedagogies are increasingly encouraged and used in the classroom for academic

(e.g., stimulating higher order cognitive processes) and practical (e.g., sharing of

resources) reasons, among others. It was also discussed that collaboration is not

always effective, which has driven research to explore a variety of measures in an

attempt to quantify and qualify collaborative processes. Arousal refers to an

individual degree of physiological activation and has been a central construct in

psychophysiology, the inference of psychological states from physiological

responses, at the individual level. But group processes obviously require group

approaches. The socio-cultural theory of learning postulates that inter-

psychological processes at the group level are then assimilated by the individual at

the intra-psychological level (Dillenbourg et al., 1996). Although the field of

psychophysiology targets the individual, inter-psychological processes already

provided a motivation several decades ago to also explore related inter-

physiological processes, giving birth to the sister discipline of social

psychophysiology or interpersonal physiology.

2.4 Interpersonal physiology

Once physiological technology developed to a point where empirical studies were

easy to carry out, physiological approaches to the analysis of small-group

interaction naturally became of particular interest to social psychologists and

41

psychophysiologists (G. E. Schwartz & Shapiro, 1973). From the marriage of social

psychology and psychophysiology, the field of interpersonal physiology (DiMascio,

Boyd, & Greenblatt, 1957) was born. Beginning in the 1950s, social scientists

started collecting physiological data from two or more people in interpersonal

interactions to explore commonalities and interdependence between their

physiological states (Thorson et al., 2018). Among the array of physiological

signals, EDA (and by extension sympathetic arousal) enjoyed privileged attention

as “a prime reflector of the social-biological nature of human behavior” (G. E.

Schwartz & Shapiro, 1973, p. 412). Ascribed connections of interpersonal

physiology to early development, language learning, group engagement, social

connection, and group membership might have been responsible for a continuous

interest in such an approach and a body of related literature which has kept growing

(Delaherche et al., 2012; Palumbo et al., 2017).

Interpersonal physiology refers to “the study of the reciprocal influence of

human physiological systems and social systems” (Kaplan & Bloom, 1960, p. 128).

In dyads, the stability and influence model (Thorson et al., 2018) considers how a

person’s physiology at a certain point in time is predicted by both his/her own

physiology at a prior time point (i.e., the stability effect) and by the other dyad

member’s physiology at the prior time point (i.e., the influence effect). This model

easily scales up to groups of any size. The influence is not explained by physiology

itself, as such, but is associated with the psychosocial processes which occur

between or among group members (e.g., one person’s heart rate influencing a

partner’s heart rate via a verbal outburst of anger; Thorson et al., 2018).

Whether driven by a shared environment, coordinated behaviors, matched

responses to a third variable, interpersonal processes, or other factors, the

mechanisms behind interpersonal physiology are still unclear (Field, 2012;

Palumbo et al., 2017). Evidence from the past decades seems to point, at least in

part, to the mirror system, composed of mirror neurons in several brain regions

(Rizzolatti & Fabbri-Destro, 2008). The first demonstration of a mirror system in

humans dates to 1995, when a study using magnetic stimulation found that “there

is a system matching action observation and execution” (Fadiga, Fogassi, Pavesi,

& Rizzolatti, 1995, p. 2608). Currently, the mirror system is discussed as the

evolutionary basis for observational learning (Field, 2012) and might play a role in

interpersonal physiology (Karvonen, Kykyri, Kaartinen, Penttonen, & Seikkula,

2016; Rizzolatti & Arbib, 1998).

The variety of methodologies and approaches from different disciplines are

reflected in the number of conceptualizations used in the study of interpersonal

42

physiology, including but not limited to attunement (Field, 2012; Karvonen et al.,

2016), compliance (Chanel et al., 2012; Elkins et al., 2009; Henning, Boucsein, &

Gil, 2001), concordance (Slovák, Tennent, Reeves, & Fitzpatrick, 2014), co-

regulation (Sbarra & Hazan, 2008), coupling (Chanel, Bétrancourt, Pun, Cereghetti,

& Molinari, 2013; Singer et al., 2016), covariation (Kaplan, Burch, Bloom, &

Edelberg, 1963), entrainment (Dikker et al., 2017), influence (Thorson et al., 2018),

linkage (Simo Järvelä, Kivikangas, Katsyri, & Ravaja, 2014), physiological

markers of togetherness (Noy, Levit-Binun, & Golland, 2015), and synchrony

(Dich, Reilly, & Schneider, 2018; Gillies et al., 2016; Liu, Zhou, Palumbo, & Wang,

2016). Occasionally, these terms imply conceptual differences in what is being

measured or are used to refer to specific analytic techniques or theoretical

approaches, but a systematic review of the literature shows that this is inconsistent

(Palumbo et al., 2017).

Different types of interpersonal relationships have been the focus of studies on

interpersonal physiology, such as parent-child, therapist-client, individuals in a

couple, artist-audience, conductor-choir, friends, teammates, groups of strangers,

and so forth (Palumbo et al., 2017). Physiological interdependence has been found

to be greater, for example, between individuals who are in a romantic relationship

(Sbarra & Hazan, 2008), in psychotherapists who are more empathetic to their

clients (Delaherche et al., 2012), and in game players who report greater rapport

with each other (Noy et al., 2015). However, when it comes to students in a

collaborative learning setting, the use of physiological data has received little

attention, as evidenced by only two papers found in a recent systematic review of

the literature including all publication dates through November 2015 (Palumbo et

al., 2017).

2.4.1 Previous work on interpersonal physiology in collaborative learning

Early examples are found in the work of Kaplan and associates in the 1960s, which

explored sympathetic arousal in relation to the affective orientation of group

members in either dyads or four-person groups of medical and nursing students

(Kaplan, 1967; Kaplan, Burch, Bedner, & Trenda, 1965; Kaplan et al., 1963). Four

decades later, sympathetic arousal commonalities were investigated in dyads of

college students in terms of nonlinear dynamics and linkage effects (Guastello,

Pincus, & Gunderson, 2006). Physiological commonalities in emotion management

and convergence during collaborative processes have been studied at the dyad level

43

using a variety of physiological responses and measures, such as

electrocardiography, respiration, EDA, temperature, and eye-tracking (Chanel et al.,

2013). More recently, physiological commonalities in relation to cognitive and

affective states (e.g., mental workload and emotional valence) during a pair-

programming task in a classroom environment have been studied using

electrocardiography and EDA (Ahonen et al., 2016; Ahonen, Cowley, Hellas, &

Puolamäki, 2018). All these previous studies focused on physiological responses

from the autonomic nervous system, which can be measured more cheaply, quickly,

and unobtrusively than those of the central nervous system (see Fig. 1 in Chapter

1) and are more easily interpreted (Novak, Mihelj, & Munih, 2012). Nonetheless,

using portable technology in a high school classroom during 11 lessons,

commonalities in electroencephalogram signals have been suggested to predict

group engagement and social dynamics (Dikker et al., 2017). None of the studies

focused on triads as the unit of analysis.

The dyadic nature of some of the relationships studied (e.g., therapist-client

and couples), together with the greater availability of pairwise physiological

indices (discussed in Sub-chapter 2.4.3), might be among the reasons why

interpersonal physiology has predominantly focused on dyads as units of analysis

(Delaherche et al., 2012). However, the general dynamics of triads and larger

groups are different than those of dyads. For example, dyads cannot form coalitions

and have newcomers or old-timers, nor can there be majority/minority influence in

dyads. Also, negotiation and conflict (i.e., argument) are more complicated in triads

and larger groups (Reiter-Palmon et al., 2017). When it comes to learning, dyads

are often considered to be peer learning, with real collaborative learning beginning

with three or more participants. Therefore, the focus of this work is on triads.

2.4.2 Arousal contagion

It is generally recognized in social psychology that the emotional responses of one

person may elicit emotional responses from another (Berger, 1962; Reeck, Ames,

& Ochsner, 2016). The effect, which has been conceptualized as emotional

contagion, is driven by feedback from facial and vocal expressions, postures, and

behaviors (Hatfield, Cacioppo, & Rapson, 1994). Emotional contagion can also

occur when changes in one’s visceral state during emotion may be mirrored in an

observer, inducing a corresponding representation (Critchley et al., 2013). The

emotion “caught” by the influenced person can be either similar or complementary,

the latter sometimes called counter-contagion (Hatfield et al., 1994). Based on

44

whether the emotional contagion leads to a similar or complementary valence (i.e.,

pleasant or unpleasant), for example, Berger (1962) provided “a simple, yet

convenient definitional framework” for the social emotions of empathy, envy, and

sadism (G. E. Schwartz & Shapiro, 1973, p. 391). According to the framework,

empathy is described as the social emotion where the valence is similar (whether

pleasant or unpleasant) in influencer and influenced, while in envy, a pleasant state

in the influencer causes an unpleasant state in the influenced person, and the other

way around in sadism, which elicits a pleasant affect in the influenced based on an

unpleasant state of the influencer. Rudimentary or primitive emotional contagion is

described as relatively automatic, unintentional, uncontrollable, and largely

inaccessible to awareness (Hatfield et al., 1994).

A particular case of interpersonal physiology explored in this dissertation is

that of arousal contagion, inspired by the notion of emotional contagion, given that

arousal is the activation dimension of emotion according to the circumplex model

of affect (Russell, 1980). Arousal contagion helps studying the same phenomena in

groups (e.g., to assess collaboration or leadership), without the need to distinguish

specific emotions, which could result in insufficient discriminant validity, as there

are overlapping traits in different emotions (Pekrun et al., 2002). The motivating

affordance is that, as a degree of physiological activation, arousal is objectively

measurable and theoretically relevant for the introspection of cognitive and

affective states.

2.4.3 Physiological coupling indices

The theoretical relevance of interpersonal physiology for the study of group

processes, including collaborative learning, faces the innate challenge of defining

and delineating variables of theoretical significance and practical relevance (G. E.

Schwartz & Shapiro, 1973), given the variety of physiological signals, the number

of features extractable from each at the individual level, and the panoply of

aggregation approaches for group level analyses (Delaherche et al., 2012; Kreibig,

2010; Palumbo et al., 2017). The individual features of physiological signals can

be descriptive statistics (e.g., mean, median, variance, standard deviation, different

percentiles, and interquartile range), which can be applied to all signals, or signal-

specific, which depend on the particular signal at hand. For example, studied

features from EDA include, but are not limited to, SCL, SCR amplitude, SCR

latency, SCR rise time, SCR half-recovery time, SCR peaks per minute, and SCR

area under the curve (Boucsein, 2012; M. E. Dawson et al., 2017). As an example

45

of other bodily systems, tens of signal-specific features have been extracted from

respiratory and cardiovascular and signals (Kreibig, 2010). In an attempt to

estimate cognitive load, Ferreira and associates (2014) employed up to 128 features

from four signals. These examples illustrate the variety of options available at the

individual level.

The aggregation of individual features at the group level, or the direct

computation of group features from particular parameters (e.g., difference in

amplitude, rate of change, direction, and linear relationship) of the individual

signals, translates into a variety of indices which are called PCIs in this dissertation.

Basically, PCIs are indices of interpersonal physiology.

Building on the indices proposed by Elkins et al. (2009), this dissertation

explores as PCIs in relation to some collaborative learning aspects (see Article I),

the indices of signal matching, instantaneous derivative matching, directional

agreement, Pearson’s correlation coefficient, and a transformation of it. Except

from directional agreement, those indices can only be computed on a pairwise basis.

Signal matching measures the difference in the amplitude of two signals.

Instantaneous derivative matching accounts for the difference in the rate of change,

as determined by the first derivative of two signals. Directional agreement

measures the percent of time that an arbitrary number of signals is changing in the

same direction (i.e., upwards, constant, or downwards), making it a simple yet

convenient index for the study of the interpersonal physiology for more than two

persons.

Signal matching, instantaneous derivative matching, and directional agreement,

applied to cardiovascular and EDA measures, have been explored in connection to

team performance, joint activity, and shared trust in technology (Elkins et al., 2009;

Montague et al., 2014). Much more often employed in interpersonal physiology has

been Pearson’s correlation, a standard and popular statistical tool, which in this case

determines the strength of the linear relationship between the physiological signals

from two individuals. Although no conclusive results have been obtained, cross

correlation has been found to predict task completion time (Henning et al., 2001)

to be higher in a conflicting situation than when playing cooperatively in digital

gaming (Chanel et al., 2012), to indicate team performance after controlling for

task/technology conditions (Montague et al., 2014), and to relate to self-reported

social presence in digital gaming (Simo Järvelä et al., 2014), among others. Even

half a century ago, correlations of the frequency of electrodermal SCRs over time

between two individuals were presumed to relate to rapport (G. E. Schwartz &

Shapiro, 1973). However, correlations, although simple, powerful, and easy to

46

interpret, might not be well-suited for continuously measured physiological signals

because data typically show sequential dependency (i.e., autocorrelations) and non-

stationarity over time (i.e., changing mean and variance), which violates general

linear model assumptions of data independence and stationarity (Palumbo et al.,

2017). In an attempt to overcome such limitations, some transformations, such as

the Fisher’s z-transform, and sophisticated statistical techniques, such as dynamical

correlation, have been proposed. The Fisher’s z-transform consists of applying the

inverse hyperbolic tangent to the correlation coefficient, in order to stabilize the

variance and obtain a normally distributed index (Chanel et al., 2013; Simo Järvelä

et al., 2014). Dynamic correlation is a nonparametric approach that places no

assumption on homogeneity of the sample, although it has limitations of its own

and complicates interpretation (Liu et al., 2016).

Although not applicable to the EDA signal, other PCIs found in the

interpersonal physiology literature are, for example, dynamic common activation

and weighted coherence. Dynamic common activation reflects both the strength

and variation of functional magnetic resonance imaging activation across subjects

(Singer et al., 2016). Weighted coherence (Porges et al., 1980) is a frequency

domain PCI which quantifies the similarities of two individuals’ physiological

responses on a specified frequency band regardless of phase differences (Montague

et al., 2014). Weighted coherence is calculated over a range of frequencies rather

than at only a specific frequency, which makes it suitable to capture the inherent

variability in frequency of interpersonal physiological indices during active

behaviors (Henning et al., 2009). Naturally, being a frequency domain index, it is

mostly useful for periodic physiological signals, such as heart beat and respiration

rate (Chanel et al., 2013). EDA, the subject of this dissertation, is not a periodic

signal. Therefore, weighted coherence was not considered among the PCIs.

In sum, PCIs can play a role in studying social interactions (e.g., as in

collaborative learning and team processes) by providing an objective measure, but

research in this direction needs further explorations to keep enhancing our

understanding on the affordances of interpersonal physiological processes to index

interpersonal psychological processes related to cognition and affect and,

ultimately, to learning.

2.5 Research gap

Collaborative learning is extensively researched (M. Baker, 2015; Jeong & Hmelo-

Silver, 2016). Yet, the field is missing continuous measures with real-time potential,

47

as collaboration is a process whose evolution in time is of interest and a determinant

of success (Dillenbourg et al., 1996; Jeong et al., 2014). These measures have been

also referred to as online measures, meaning that they are taken as the process

unfolds, and not before or after. In addition, the field is on the lookout for measures

complementing subjective data as coming from self-reports (Wise & Schwarz,

2017).

EDA has a long history of research, over a century old (Boucsein, 2012).

However, it has not been until recently that it has become easily and unobtrusively

measurable in the wild, outside laboratory settings, thanks to the increasing

availability and technological quality of wearable biosensors (Garbarino et al.,

2014; Torniainen, Cowley, Henelius, Lukander, & Pakarinen, 2015). Yet, students

remain an understudied population, as clinical studies take the lion share of EDA

research (M. E. Dawson et al., 2017; Poh et al., 2010).

In psychology, arousal is a commonly studied construct due to its connection

to cognition and affect, and the natural psychological interest in these domains

(Kuppens, Tuerlinckx, Russell, & Barrett, 2013). Yet, we know little about arousal

in the classroom and its patterns. In the few studies tackling arousal in the formal

academic setting of the classroom (e.g., Arroyo et al., 2009; Gillies et al., 2016),

the affective component steals the show, while it is rare to find interpretations or

foci on the realm of cognitive processes such as attention. A little more studied have

been arousal or physiological measures in exams situations (Spangler, 1997;

Spangler, Pekrun, Kramer, & Hofmann, 2002). However, measures of arousal

based on the frequency of SCRs have not yet been tested during exams in

connection to students’ performance.

Interpersonal physiology has long been exploring the commonalities and

influence between subjects with certain relationships, predominantly parent-child,

therapist-client, and individuals in a couple (Palumbo et al., 2017). When it comes

to individuals collaborating, workplace performance is a more frequent target than

collaborative learning in schools (Elkins et al., 2009; Henning et al., 2001). In terms

of group size, dyads have received much of the attention, while other common

group sizes in collaborative learning such as triads have been neglected (Reiter-

Palmon et al., 2017; Thorson et al., 2018).

In this context, this dissertation has found a research gap to address, matching

multidisciplinarity, interest, opportunity, and relevance. This work contributes to

the particular case of collaborative learning by investigating PCIs, arousal levels,

and arousal contagion, as such, and in relation to outcomes as learning gain and

academic achievement.

48

49

3 Aims This dissertation pursues the exploration of EDA and derivative measures, such as

sympathetic arousal and PCIs, concomitant with individual and interpersonal

cognitive and affective processes, during collaborative learning in a naturalistic

setting. To that extent, the main aims of this dissertation are to:

1. investigate the interpersonal physiology of EDA in connection to collaborative

learning processes (addressed in Article I: research question [RQ] 1, RQ2;

Article III: RQ1, RQ2),

2. profile sympathetic arousal in the classroom at both the individual and

interpersonal levels in collaborative learning (addressed in Article II: RQ1,

RQ2, RQ3; Article III),

3. examine both the individual and interpersonal physiology of EDA in relation

to academic achievement (addressed in Article I: RQ2; Article II: RQ4), and

4. explore sympathetic arousal contagion among group members as a result of the

cognitive-affective social interactions in collaborative learning (addressed in

Article III: RQ3, RQ4).

In order to meet the aims, two data collections were organized, one in a classroom-

like research environment of the University of Oulu (see Article I) and another

during two consecutive runs of a regular course in a real classroom (see Articles II

and III).

50

51

4 Methods Two research designs were conceived for this dissertation, having ecological

validity in mind. To each design, a data collection process was associated. There

were similarities and differences between the designs. The participants in the

studies were Finnish high school students. A classroom-like environment and a real

classroom were used for collecting the data in naturalistic locations. The topics

students were dealing with during the data collections were part of their curriculum

and the tasks were either embedded in or regular part of their coursework. The

materials involved a biosensor wristband (Empatica®) to record EDA,

performance instruments (e.g., knowledge tests, course exams), online learning

environments, and questionnaires. Specialized computer software was used for the

analysis of the relevant variables according to the aims. In the light of the

reproducibility principle, this chapter presents the research methodology in detail,

an overview of which is listed in Table 1.

Table 1. Method overview.

Aspect Data collection I Data collection II

Participants N = 24; high school students N = 24; high school students

Location Classroom-like Classroom, classroom-like

Topic Nutrition Advanced Physics

Duration 2 h 15 min 2 x (75 min, 3 x week, 6 weeks)

EDA sensor Empatica® E3 Empatica® E4

Performance measures Pre- & post-test, joint

solution

Course exam

Online learning environment weSPOT Open edX

Questionnaire(s) Motivated Strategies for

Learning Questionnaire

(MSLQ)

MSLQ and a questionnaire adapted

from Schmitz and Skinner (1993)

EDA analysis software EDA Explorer Ledalab

Variables PCIs, collaborative will,

collaborative learning

product, dual learning gain

Sympathetic arousal: level, direction,

and contagion; exam grades

4.1 Participants

The participants in the two data collections for this research were high school

students from the University of Oulu’s Teacher Training School.

52

Participants (N = 24) in the first data collection (see Article I) were 13 males

(54%) and 11 females (46%) aged 16 to 19 years (M = 17.3; SD = 0.62).

Participants (N = 24) in the second data collection (see Articles II and III) were

randomly selected from those enrolled in the regular, elective Advanced Physics

course, which runs twice during the spring term. Twelve students were recruited in

each course run as participants. The gender distribution was six females (25%) and

18 males (75%) aged 16 to 17 years (M = 16.5; SD = 0.5). Their attendance average

was 10.9/12 (SD = 1.3) for the first course run and 10.5/12 (SD = 1.3) for the

second. They were high achievers in the preceding physics course, obtaining an

average grade of 9.0 out of 10 (SD = 0.6).

4.2 Materials

The materials for this study can be divided into four groups: the EDA sensors,

performance instruments, online learning environments, and questionnaires.

4.2.1 The Empatica® E3 and E4 wristbands

Essential tools in the data collection for this dissertation were the Empatica® E3

and E4 (Empatica Inc., Cambridge, MA, USA) biosensor wristbands (see Fig. 4 for

pictures of the E4; the E3 is similar), since physiological data, particularly EDA, is

at the core of this research. The Empatica® E3 was used for the data collection I

(see Article I), and the E4 for the data collection II (see Articles II and III). The

Empatica® E3 and E4 wristbands are equipped with four sensors: a

photoplethysmograph to measure blood volume pulse, an EDA sensor, a 3-axis

accelerometer, and an infrared thermopile to measure peripheral skin temperature.

The EDA sensor, the one used in the studies of this dissertation, has a sampling

frequency of 4 Hz (i.e., it takes four measurements per second). The Empatica® E3

and E4 EDA sensor utilizes the exosomatic method to measure EDA, which

consists of applying a small external current by means of two electrodes and

determining the skin electrical conductance through Ohm’s law (Edelberg, 1967).

The exosomatic method is usually credited to an 1888 study by the French

neurologist Charles Féré, but was already described by the French electrotherapist

and electrodiagnostician Romain Vigouroux in an 1879 study (Edelberg, 1967).

The units for the skin electrical conductance are microSiemens (µS). On a historical

note, resistance, the electrical inverse of conductance, was also used in the early

days of EDA research. However, in 1970, the editor of the journal

53

Psychophysiology proposed the elimination of resistance measures and the use of

conductance only when measuring exosomatic activity, based on the peripheral

mechanisms of the skin (Venables & Christie, 1973, p. 6).

The Empatica® E3 and E4 have two modes of operation: recording in internal

memory and real-time streaming via Bluetooth®. The former was deemed more

suitable for the studies of this dissertation, which used offline analysis of the data.

Data recorded in internal memory are stored in a comma-separated values file for

each sensor and are accessible via a USB connection. The EDA file consists of a

first row with the recording starting time, a second row with the sampling frequency

(4 Hz), and the continuous EDA measures in the subsequent rows. The Bluetooth®

streaming affords applications for learning beyond research, such as for learning

analytic dashboards, intelligent tutoring systems, and implementation of alerts in

critical moments, among others. Such potential applications are facilitated by a

software development kit, including a mobile programming interface that enables

the developing of apps to leverage real-time data from the multisensor wristband.

Fig. 4. Front and back view of the Empatica® E4 watch-like biosensor wristband (photos by courtesy of Empatica Inc.).

Although the two data collections involved a larger number of participants, for the

studies of this dissertation, participants were limited by the biosensor wristbands

available. At the time of the first data collection, our laboratory was equipped with

six Empatica® E3 wristbands, and the number rose to 12 Empatica® E4s for the

second data collection, in a purchase deal that involved returning the E3s. The

Empatica® E3 and E4 are research devices, more accurate and precise than

54

consumer-oriented biosensor wristbands available in the market, but also several

times more expensive, which limited the number that our laboratory was able to

acquire, as has happened in other studies also (e.g., Koskimäki et al., 2017).

4.2.2 Performance instruments

The performance measures in the first data collection included a pre-test, a post-

test, and a collaborative learning product (i.e., the groups’ joint solution to the task),

while in the second data collection, a real-life advanced physics course exam was

used. The pre- and post-tests were identical: multiple-choice tests consisting of ten

questions with four options each. An example of a question is: “Fiber intake is

important because it: a) allows for consistent energy supply, b) maintains bowel

health, c) lowers cholesterol levels, and d) contains calcium.” In total, out of the 40

choices, 22 were correct. Shared spreadsheets using Google Sheets technology

were provided for students to collaboratively report on their joint solution to the

task. The physics course exam consisted of three sections (A, B, and C) with three

questions each. Section A was theoretical and focused on the definition, connection,

and explanation of concepts and phenomena (e.g., resonance, interference, and

Doppler effect). Sections B and C had a more practical character and focused on

calculations and graphical representations. The students had to choose and answer

two questions from each section, except from section C, where only one question

was required to be answered. Thus, students had to answer in total five questions

on the exam.

4.2.3 Online learning environments

Two online learning environments were used in this study: weSPOT

(Mikroyannidis et al., 2013) and Open edX (https://open.edx.org/) for the first and

second data collections, respectively. WeSPOT stands for working environment

with social and personal open tools for inquiry-based learning. Accordingly,

weSPOT scripts collaborative learning in compliance with an inquiry-based

learning paradigm (Edelson, Gordin, & Pea, 1999) consisting of five steps: 1) plan

the design method, 2) set the criteria, 3) collect the data, 4) discuss the findings,

and 5) communicate the results. Open edX is an open source software which

powers the edX platform (https://www.edx.org/) for so-called massive open online

courses (MOOCs). It was chosen because of its customization affordances since it

is open source and utilized to create a dedicated online course following the

55

traditional course structure as provided by the teachers (e.g., course content,

additional support learning material, simulations, and videos).

4.2.4 Questionnaires

The MSLQ (Pintrich, Smith, Garcia, & McKeachie, 1991, 1993) was used in both

data collections to assess the learning regulation profiles of the students, which

served as criterion for the formation of groups as heterogeneously as possible. The

MSLQ is one of the most frequently used inventories for that purpose (Bjork,

Dunlosky, & Kornell, 2013; Panadero, 2017). The questionnaire consists of 81

items organized in two sections, motivation and learning strategies, which contain

6 and 9 scales, respectively. Each item is rated on a Likert scale from 1 (not at all

true of me) to 7 (very true of me).

The peer learning scale of the MSLQ—comprising items 34, 45, and 50—was

also utilized to assess the student’s predisposition to collaborate in a learning task

(i.e., collaborative will; see Article I). The items of this scale account for the

students’ propensity to explain learning material to group members, work together

with others to complete an assignment, and discuss the task in the group.

A questionnaire (see Table 1 of Article II) to assess students’ perception of

cognitive, affective, motivational, and collaborative aspects on a lesson basis was

adapted from the one employed by Schmitz and Skinner (1993) for an experiment

in a similar setting, a naturalistic study with data collected in the classroom over

four months. The questionnaire consisted of four items asking students to rate their

task domain knowledge, mood, motivation, and group functioning, on a scale of 1

to 10. The simplicity principle prevailed in choosing and tailoring this

questionnaire given the limited time available for these measurements in the

naturalistic setting of our study: a real high school course, packed with content and

coursework, including both paper-and-pencil tasks and hands-on experiments.

Conveniently, the questionnaire was embedded in the dedicated course created on

the Open edX platform. In this way (i.e., simplicity and easiness), and since the

students were expected to complete the questionnaire in every single lesson, we

minimized the impact on the measurement environment (Winne & Perry, 2000).

While larger inventories are considered to be more reliable, rating scales based on

fewer items are acknowledged to be economical and useful under specific

circumstances (Pekrun et al., 2002).

56

4.3 Procedure

Two data collections were organized targeting EDA specifically, but also

performance measures and questionnaires as data sources. Two important issues

involving EDA recording were considered in both data collections: room

temperature and biosensor wristband adjustment. Room temperature for EDA

measures is recommended to be 23 °C (Boucsein, 2012). No adjustment was

required in this regard since the temperature in the locations for both data

collections was controlled by a thermostat, whose standard setting coincided with

the recommended temperature. Also, according to the good practices of recording

EDA, students were instructed to wear the biosensor wristband on their non-

dominant hand in order to reduce the effect of movement artifacts and to adjust the

wristband strap neither loosely nor tightly so that the electrodes made proper but

not excessive contact with the skin, which would result in a pressure artifact

(Edelberg, 1967).

In both data collections, students worked in triads that were formed based on

the principles of gender diversity (Bear & Woolley, 2011) and heterogeneity of

learning regulation profiles (i.e., according to the MSLQ scores), for the sake of

between-triad comparability.

4.3.1 Data collection I

Two weeks before the collaborative task, the students were asked to answer the pre-

test and the MSLQ at their school. The collaborative task was carried out in a

classroom-like research infrastructure of the University of Oulu (i.e., the Learning

and Interaction Observation Forum; http://www.oulu.fi/leaf-eng/). The experiment

was organized in four sessions with 12 different participants each (N = 48),

although, for the purpose of this dissertation, only participants wearing the

biosensor wristband were considered (N = 24), that is, six participants in each

session. The sessions took place in the morning and the afternoon of two

consecutive days for a total of four sessions. The duration of the sessions was

2 h 15 min. The first 20 min were for preparation and instructions, followed by

1 h 40 min to work on the actual task and 15 min for the post-test. In the preparation

phase, students were introduced to the materials, particularly the biosensor

wristband, the weSPOT environment, and the collaborative learning task.

Students sat at individual tables clustered in four groups of three. Each table

was supplied with a digital tablet to support the task execution. The students in two

57

groups were each provided with an Empatica® E3 biosensor wristband to measure

EDA.

The task consisted of the design of adequate, healthy breakfasts for at least one

of several cases corresponding to people with different nutritional requirements,

such as a marathon runner in training, a teenager, a patient with a heart condition,

a diabetic person, and someone wanting to lose weight. Relevant variables of the

target hypothetical case, needed to carry out the task, were provided, such as age,

height, weight, and lifestyle details or health issues with implications for the

breakfast characteristics. The task was designed to fit in the students’ science

curriculum by being aligned with their current knowledge and the content and goals

of their science studies.

On the tablet provided to each student, a shortcut to two Google-Docs® files

was available: a document and a spreadsheet. The document detailed the nutritional

needs of marathon runners. The spreadsheet had a template with rows for food

items and columns for the weight (in g) of the different nutrients contained by the

food item. The template included several examples for illustration purposes. Upon

task completion, students delivered their solution through the weSPOT

environment. Finally, the students were asked to answer the post-test.

4.3.2 Data collection II

Ecological validity was at the center of the second data collection. Therefore, the

classroom was the chosen setting. Using Empatica® E4 biosensor wristbands, EDA

was measured in 12 students simultaneously in each of two consecutive runs of an

elective, advanced physics course (including the exam), as the course is offered

twice during the spring term. The course consists of 18 lessons, plus the exam.

There are lessons three times a week, lasting 75 min each. The students put the

wristband on at the beginning of each lesson and took it off at the end. Topics of

the lessons include, among others, wave formation and interference, reflection and

refraction, sound waves, and geometrical optics. In general, the lessons begin with

a theoretical part (i.e., the teacher’s explanation of the topic) followed by a

collaborative practical activity with paper-and-pencil problems, or hands-on

experiments, and instructions and support material in the Open edX environment.

In the practical activity, students put the learned physical theories, laws, and

principles to the test in groups of three. The groups remained the same for the

duration of the course. The students accessed Open edX during class on the tablets

assigned to them by the school as a technological support to their studies. At the

58

end of each lesson, students were asked to answer the questionnaire adapted from

Schmitz and Skinner (1993), which was embedded in the Open edX environment.

4.4 Data analysis

Aligned with the materials and data collected, the analysis targeted three data

modalities. First, the EDA data collected by the biosensor wristbands were

processed to extract different relevant indices in accordance with the aims. Second,

the variety of performance instruments were assessed. Third, questionnaire data

were prepared for correlation analysis. Next, the analysis of these three data

modalities is detailed.

4.4.1 Electrodermal activity

Two approaches are used to analyze the EDA data, one utilizing the EDA signal as

such (see Article I), and the other based on sympathetic arousal, a derivative

measure from the EDA signal (see Articles II and III). For efficient use of the

storage capacity in internal memory, the biosensor wristband by design records the

data in a comma-separated values file containing a single column, where the first

row displays the starting time in UNIX format (i.e., the number of seconds elapsed

from 00:00:00 UTC on January 1st, 1970), the second row stores the sampling

frequency in Hz, and the actual EDA signal values (in μS) start from the third row

on. Using the initial time and the sampling frequency (number of samples taken

each second), together with the position of each value, it is straightforward to obtain

the time-value pairs needed for synchronization and analysis. Therefore, common

to both approaches is an initial data pre-processing to transform the file provided

by the biosensor wristband to a data format where each EDA value appears next to

its corresponding timestamp. Next, the specificities of each approach are discussed.

Approach I: Electrodermal activity signal based

In this approach, used in Article I, the EDA signal constitutes the pillar of the

analysis. The EDA signal is susceptible to pressure and movement artifacts

(Edelberg, 1967). Pressure artifacts are linked to wristband maladjustment when it

is put on too tight. To reduce pressure artifacts, students were instructed to wear the

wristband not too tight. As for reducing movement artifacts, as explained earlier,

students were asked to wear the biosensor wristband in their nondominant hand.

59

Notwithstanding, it is necessary to identify possible artifacts in the data to be able

to discard them in the analysis phase for the sake of the reliability of the results.

The web-based tool EDA Explorer (Taylor et al., 2015) was used to automatically

detect artifacts. EDA Explorer employs a machine learning classifier to determine

on a 5-second epoch basis whether it corresponds to noise or a clean signal.

Reportedly, the accuracy of such a binary classifier surpasses 95% (Taylor et al.,

2015). The outcome of applying EDA Explorer to the EDA signal recorded in the

first data collection was that 490,880 from 547,848 data points (i.e., 90% of the

signal) were classified as clean by the algorithm. Consequently, those were the data

selected for the next analysis steps, while the other 10%, labelled as noise, was

discarded at this point.

The PCIs to be computed from the clean EDA look at different parameters of

the signals. Some of these PCIs are robust in respect to individual differences in the

SCRs (e.g., directional agreement is immune to anything related to amplitude as it

only considers the direction), while others are highly sensitive to it (e.g., the signal

matching precisely accounts for amplitude differences). Therefore, especially for

the sensitive cases, it is necessary to normalize the signal so as to make it

comparable across group members. The individual differences are caused, for

example, by the thickness of skin (Braithwaite, Watson, Jones, & Rowe, 2015),

especially the thickness of the corneum (M. E. Dawson et al., 2017), which is the

outermost layer of the epidermis. Normalization for comparability purposes was

then accomplished by calculating the t-scores (i.e., subtracting the sample mean

from every value and then dividing by the sample standard deviation). Both sample

M and SD were computed on an individual basis.

The final step before computation of the PCIs is the synchronization in each

group of the clean, normalized data for all group members. Since each individual

signal has a timestamp for each data point, the synchronization process can be

basically seen as the alignment of the timestamps at the group level. Although, the

synchronization step was carried out after normalization, the reader should note

that both processes are totally independent and, as such, can be executed in any

order without affecting the final result.

The clean, normalized, and synchronized data are the base for the computation

of the PCIs. Since, by definition, most of the PCIs are computed pairwise, with

directional agreement being an exception as it easily extends to groups of any size,

the PCIs were obtained pairwise for all triad member combinations (AB, AC, and

BC). The PCIs used are signal matching, instantaneous derivative matching,

directional agreement, Pearson’s correlation coefficient, and Fisher’s z-transform.

60

Signal matching consists first of a between signal differentiation and then of

an aggregation. This is, instantaneous absolute differences in the two students’

signals were computed and then aggregated as the mean of the entire collaborative

learning session.

Instantaneous derivative matching compares the slopes of the two students’

signals. Mathematically, this is expressed in Eq. (1) as:

∑ | − − − |, (1)

where T is the time in number of samples of the signal, a and b refer to the signal

of each student, respectively, t refers to the current data point, and t+1 to the next

data point.

Directional agreement refers to the percentage of the time in which the signals

(EDA in this case) of the students are going in the same direction, that is,

simultaneously rising, falling, or staying constant. For the computation of

directional agreement, first it was determined whether there was instantaneous

agreement or not, and then the percentage of agreement was obtained.

Pearson’s correlation coefficient needs no further explanation as it is a standard

and widely used measure in statistical analysis. Fisher’s z-transform is a

transformation of Pearson’s correlation coefficient—by applying the inverse

hyperbolic tangent—so as to obtain a normally distributed index.

Approach II: Sympathetic arousal based

In this approach, followed in Articles II and III, EDA is not used as such, but as the

means to index sympathetic arousal, the degree of physiological activation, in this

case from the sympathetic nervous system. Since arousal was to be indexed by the

frequency of SCRs, the MATLAB-based Ledalab (version 3.4.9;

http://www.ledalab.de/) software was used for the detection of SCRs. It allows for

the separation of superimposed responses using continuous decomposition analysis

by means of nonnegative deconvolution (Benedek & Kaernbach, 2010b). Thus,

Ledalab helps to address the traditional bias of counting the SCRs without

accounting for those in a superimposed response, a classical method known as

trough-to-peak (Boucsein, 2012). Ledalab is also the tool for the analysis of the

EDA signal recommended by the E3 and E4 biosensor wristband manufacturer

(Empatica Inc., 2018).

61

The Ledalab algorithm assumes raw, unfiltered data as the input for

decomposition analysis and feature extraction of the EDA signal (Benedek &

Kaernbach, 2010a). Therefore, no signal preprocessing, such as cleaning or

filtering, was performed on the EDA signal as obtained from the wristband sensor.

In determining which change in conductance will be considered an SCR, an

amplitude threshold needs to be specified as parameter. For this dissertation,

0.05 μS have been chosen for the threshold as it is a long-time standard owing to

the maximum sensitivity of earlier EDA sensors (Braithwaite et al., 2015; Venables

& Christie, 1973, 1980).

By means of Ledalab, the SCR onsets (i.e., the starting point of the peaks) were

determined from the EDA recordings. The onsets were used to compute the SCR

frequency on a minute basis, to be used as a measure of arousal in peaks/min (ppm).

The calculation of SCR frequency followed a moving window (width 1 min; step

250 ms) approach, so that arousal at each instant would be the count of SCR onsets

in the previous minute. Arousal was thus computed every 250 ms, the maximum

temporal resolution available as it is the sampling interval of the Empatica® E3

and E4 EDA sensors. The frequency approach has the advantage of not having to

standardize the data for comparison across individuals, as amplitude measures are

not considered. The frequency on a minute basis and the total count of SCRs or

arousal episodes were both used in the analysis to answer different RQs, according

to their scope and target.

The SCR frequency was also used as the criterion to categorize arousal into

low, medium, and high levels, based on the literature. Typically, a frequency of 1–

3 ppm occurs at rest (M. E. Dawson et al., 2017, p. 225), and as frequency increases

with the arousal level, values higher than 20 ppm are interpreted as high arousal

(Boucsein, 2012, p. 222). Accordingly, frequencies of up to 3 ppm were labelled as

low, from 20 ppm and up were labelled as high, and anything in between was

labelled as medium. The incidence of each arousal level was determined for every

student in terms of percentage relative to the time of each lesson. The persistence

of each occurrence of an arousal level for every student was computed as the

continuous duration of the occurrence (i.e., without switches of level in between).

Notwithstanding the precautions taken in instructing students on the correct

placement of the biosensor wristband, its occasional maladjustment resulted in the

measurement of basically noise rather than an actual EDA signal, due to inadequate

electrode-skin contact. Therefore, an important part of the analysis consisted in the

removal of those cases for the sake of the reliability of the results. A technique for

automatic assistance to such a purpose was developed for this dissertation, building

62

on the typical frequency of SCRs at rest (1–3 ppm). Essentially, based on the

minimum of 1 ppm as the worst case, the students’ EDA recordings where the

number of SCR onsets was less than 1 ppm times the lesson duration in min were

not considered in the analysis. That was the case for 137 (33%) of the EDA

recording sessions during the course lessons. Due to students being absent for some

lessons, 46 (11%) of the expected recording sessions were missing. Taking together

the missing sessions and those discarded, 237 (56%) were available for further

analysis at the individual level (see Article II). As for the group level (see Article

III), it was necessary to identify the cases where valid EDA data were available for

all members of a triad in a certain lesson, since the unit of analysis in this context

was the triad. Eighteen such cases (corresponding to 18 x 3 = 54 recordings) were

found, meaning that in the final sample for group level analysis, 7 of 8 triads and

15 of 35 lessons were represented. Fortunately, in the exam, there were no absent

students, and no EDA data had to be discarded as noisy.

In the previous approach (see Approach I: Electrodermal activity signal based)

one of the PCIs used was directional agreement, computed directly on the EDA

data. For Article III, directional agreement was selected among the PCIs for its easy

and direct calculation for groups of any size, since the other indices can only be

computed pairwise, and also since it produced the better results in the analysis for

Article I. In addition, since the focus of Article III was on sympathetic arousal in

collaborative learning, directional agreement was applied to the arousal signal

derived from the EDA signal, rather than to the EDA signal as such as in Approach

I. Moreover, Approach I applied directional agreement to the whole session, as this

approach uses a moving window allowing for a continuous measure of directional

agreement for every instant during the lesson after the second initial minute. The

latter is because a first minute is necessary for the initialization of the arousal

measure, and a second minute is necessary for the first computation of directional

agreement based on those initial arousal values. Previously, directional agreement

has been calculated immediately from data points (e.g., Elkins et al., 2009;

Montague et al., 2014), but it was noticed from the data that sometimes there are

instantaneous fluctuations that do not actually represent the direction when the

adjacent samples are considered. For example, there can be an instantaneous small

increase in one value in respect to the previous one, but this could actually be a

tendency to decrease if we look at the neighboring data. To overcome this issue, it

was decided to base the direction calculation on one-minute trend lines rather than

on two consecutive values which are 250 ms apart. Trend lines were determined

for one-minute windows every 250 ms (moving window), and the sign slope of the

63

trend line was used as the measure of direction for that instant, namely, upward (+1),

constant (0), or downward (–1). This means that the arousal direction at each

instant was the sign of the slope of the trend line involving the previous minute of

data. Trend lines were obtained using MATLAB’s built-in polyfit function. Finally,

again using a moving window of 1 min with a step of 250 ms, the arousal

directional agreement for each triad at every particular instant was computed as the

percentage of the previous min in which all triad members’ arousal was going in

the same direction.

Another phenomenon targeted in the analysis was high arousal contagion (see

Article III). To that extent, the duration of each high arousal interval was marked at

the beginning of the interval. Then, for each interval, it was determined if, within

it, some other triad member(s) reached the level of activation of high arousal

(influencer point of view or potential arousal contagion), and, if so, at which latency

(i.e., time elapsed from the interval start to a triad peer’s high arousal interval start).

It was also determined how many triad peers could have caused it via arousal

contagion (influenced point of view), if any.

4.4.2 Performance

The pre- and post-tests were graded from 0 to 40 by adding the sum of options

chosen or left unchosen correctly (scoring 1) or incorrectly (scoring 0). The

individual learning gain was calculated by subtracting pre-test from post-tests

scores. The dual learning gain was aggregated as the sum of the individual learning

gains.

The collaborative learning product was measured as the grade of the groups’

joint solution to the task of designing appropriate breakfast menus for people with

different needs. The solution was graded from 4 to 15 by a research assistant,

according to the variety of the breakfast (up to 3 points; e.g., 1 if there were <3

ingredients), accuracy (up to 3 points; e.g., 2 if there was realistic information about

the macronutrients and fiber, but the solution contained small mistakes in the

calculations), adequacy in terms of matching the needs of macronutrients (up to 3

points) and energy (up to 3 points), attention to all the special needs of the particular

person (up to 2 points) of the different cases, and number of breakfast menus

designed (1 point for more than one).

The results of both the dual learning gain and the collaborative learning product

were correlated (see Article I) to five PCIs: signal matching, instantaneous

derivative matching, directional agreement, Pearson’s correlation coefficient, and

64

Fisher’s z-transform. The purpose was to investigate the predictive power of such

indices in terms of the collaborative learning outcomes.

The exam of the advanced physics course (data collection II) was graded by

the teacher. In this sense, it constituted an authentic measure as it is precisely how

the students’ achievement is determined in real life. In light of the Yerkes-Dodson

law, this performance measure was correlated to the students’ number of

sympathetic arousal episodes (indicating sympathetic activation) during the exam

(see Article II).

4.4.3 Questionnaires

The MSLQ scales are scored by averaging the ratings of the items belonging to

them (Pintrich et al., 1991). Consequently, the collaborative will, as measured by

the peer learning scale, ranges from 1 to 7, similar to the individual items. The

collaborative will was correlated to the five PCIs (see Article I): signal matching,

instantaneous derivative matching, directional agreement, Pearson’s correlation

coefficient, and Fisher’s z-transform. The purpose was to study the relation between

such objective indices and the students’ subjective ratings as to their propensity to

engage in collaborative behavior.

The ratings of the questionnaire adapted from Schmitz and Skinner (1993)

were used directly for correlation with the in-class arousal, as measured by the

count of SCRs or sympathetic arousal episodes. The questionnaire measured

students’ perception of cognitive, affective, motivational, and collaborative aspects

on a lesson basis. Unfortunately, in some lessons, students did not complete the

questionnaire as required. From the 237 cases where valid student EDA data were

available, 153 had the corresponding answered questionnaires. Hence, correlation

analyses between the questionnaires and EDA data were performed on this subset

of 153 cases.

4.5 Research evaluation

Since EDA is at the core of this research, its evaluation as a physiological response

and as a signal is paramount for reliability, together with the validity of its use as a

proxy for sympathetic arousal.

Research utilizing EDA in a variety of disciplines has a solid tradition spanning

over a century, with the large number of studies stimulating periodic and

comprehensive reviews over the decades (Boucsein, 1992, 2012; Critchley, 2002;

65

M. E. Dawson et al., 1990, 2007, 2017; Edelberg, 1967; E. Neumann & Blanton,

1970; Venables & Christie, 1980). To date, EDA remains one of the most common

data sources in the field of psychophysiology (Boucsein, 2012; Cacioppo &

Tassinary, 1990a; Kreibig, 2010), that is, the area of study of psychological

processes based on physiological signals. The results of the extensive research EDA

has been the subject of have enabled its consideration in the scientific community

as a well-validated, widely accepted measure of sympathetic arousal (Critchley,

2002). In addition, it is regarded as perhaps the most readily accessible and simplest

indicator of sympathetic arousal (E. Neumann & Blanton, 1970), in connection to

the exclusive characteristic of the electrodermal system as being uniquely

innervated by the sympathetic nervous system (Boucsein, 2012). In the early years

of EDA, there were discrepancies as to whether the EDA phenomenon was

produced by vasomotor activity or sweat gland activity, but its interpretation as a

measure of arousal has undisputedly accompanied EDA since its very discovery

(Edelberg, 1967). Building on such a rich and long tradition, it is therefore

considered valid to use EDA in this dissertation to index sympathetic arousal.

No less important is the ecological validity (see A. L. Brown, 1992) of the data

collections of this dissertation for the interpretation of the results. In this regard, it

is a strength of this research that the study for Articles II and III took place during

a real course in an actual classroom, while the study for Article I took place in a

classroom-like research environment, with a task integrated in a high school

science curriculum. Not only the locations contributed to the ecological validity,

but also the unobtrusive, seamless nature of the comfortable, watch-like, and

scientific-quality of the wearable biosensor used. Although it has to be

acknowledged that the palms and soles constitute the preferred skin sections for

recording EDA (Boucsein et al., 2012), as they contain the highest density of

eccrine sweat glands (Critchley, 2002; Edelberg, 1967), studies have shown that

the wrist is a viable alternative to them (Poh et al., 2010; van Dooren et al., 2012),

while also being more practical for daily life measurements. The EDA data

collections also complied with the recommendations of recording the signal from

the nondominant hand to minimize motion artifacts (Boucsein et al., 2012;

Edelberg, 1967), and under a constant temperature of 23 °C to minimize

thermoregulatory interference (Boucsein, 2012).

Technological advances contribute to the accessibility and reliability of

physiological measures of interest in psychological research (Cacioppo & Tassinary,

1990a; Swan, 2012), including applications to learning (Schneider et al., 2015). In

this context, the Empatica® E3 and E4 biosensor wristbands have been designed

66

purposefully for scientific research, while being practically unobtrusive (Garbarino

et al., 2014). Their EDA sensor has been validated (Empatica Inc., 2017) against

the Procomp® Infiniti SA9309M EDA sensor (Thought Technology Ltd., Montreal,

Canada). Successful validation has been carried out also for other wristband sensors

(e.g., McCarthy, Pradhan, Redpath, & Adler, 2016). However, despite the quality

of the sensor, part of the data were discarded (see Sub-chapter 4.4.1) as a result of

concerns regarding noisy recordings possibly due to wristband maladjustment. The

analysis for the first data collection used the EDA Explorer software to detect noise

within the EDA signal, with a reported accuracy of 95% (Taylor et al., 2015). In the

case of the second data collection, the algorithms powering the Ledalab software

used are a recent development in EDA signal processing based on mathematical

and physiological modeling (Boucsein, 2012). Ledalab is the tool to analyze EDA

that is currently recommended by the Empatica® E3 and E4 manufacturer

(Empatica Inc., 2018). The amount of data discarded is comparable to that of other

studies also using biosensors in a science high school classroom (e.g., Arroyo et al.,

2009). Further technological developments are needed to systematically maximize

the amount of valid data. Nevertheless, the dimensions of the second data collection

involving two entire runs of a physics course still allowed us to study the

phenomenon of interest at the individual and interpersonal levels in collaborative

learning. The several-month data collection together with the two course runs also

helped to compensate for the number of participants that was limited by the number

of biosensor wristbands available.

It is known that gender plays a role in collaboration (Bear & Woolley, 2011),

and that same gender structure is preferable when comparing group measures based

on EDA (Boucsein et al., 2012). Unfortunately, due to the predominantly male

gender of the students enrolled in the course for the second data collection, it was

not possible to compose all triads with the same gender structure, having six triads

with one female and two males, and two triads composed of three males.

Nonetheless, clear effects of gender in psychophysiological studies have been

elusive (Cacioppo & Tassinary, 1990a). However, research using

electrocardiography and/or EDA measures has reported largely comparable

responses in males and females, resulting in gender having only a small effect (S.

A. Neumann & Waldstein, 2001), and in other cases has deliberately excluded

gender from analysis after initial examination of results, due to an inconsistent

effect on the variables (Simo Järvelä et al., 2014). Therefore, the reportedly small

and inconsistent effect of gender mitigates this limitation.

67

Finally, the fact that the second data collection was conducted with a student

population of high achievers must be considered when it comes to external validity

or generalizability of the results to a more heterogeneous sample.

4.6 Ethics

This dissertation follows the ethical guidelines and principles of research in the

humanities and social and behavioral sciences of the Finnish Advisory Board on

Research Integrity (2009; 2012). Such principles involve the three areas of

respecting the autonomy of research subjects, avoiding harm, and insuring privacy

and data protection.

Since physiological data were to be collected on the participants while learning,

an ethical statement at the project level was requested to, and the corresponding

approval obtained from, the Ethics Committee of Human Sciences of the University

of Oulu.

Participation in the study was voluntary, and refusing to participate had no

implication at all for the students’ grades. An informed written consent was

obtained from the students, who could revoke it at any time during the respective

studies.

Written information was provided to the students and their parents, including

the principal researcher’s contact information, the research topic, the method of

collecting data and the estimated time required, the purpose for which data would

be collected, how it would be archived for secondary use, and the voluntary nature

of participation (Finnish Advisory Board on Research Ethics, 2009, p. 7). In

addition to the written information, the principal researcher of the project, together

with several project members, detailed the information orally to the students prior

to the data collection and answered their questions.

Since the first data collection took place in a classroom-like laboratory outside

the school, the task was coordinated with the science teachers so that it was part of

their course work, meaning it did not represent extra effort for the students. In the

second data collection, the measurements were taken in their regular classroom

following a minimum intervention method.

The personal data collected were those strictly necessary to analyze the data

and report on the research. For the data analysis phase, data were anonymized by

replacing students’ names with numeric identifiers. Such anonymization also

allowed for protection of privacy in any potential research publications that would

68

report the findings. To ensure confidentiality, the digital library data were stored

and made available only to authorized research team members

During the studies, pictures were taken to support research dissemination by

helping stakeholders to better understand the design, setting, environment, material,

cognitive-affective states, and so on. Due to the sensitivity of this information, their

informed written consent was also requested. However, even in those cases where

consent was obtained, using pictures with faces was avoided unless it was necessary

to make a point about a particular relevant cognitive-affective state.

69

5 Overview of the articles This dissertation involves three published scientific articles, which address the four

aims pursued. In investigating psychological processes associated with learning

through the proxy of a concomitant physiological signal, the articles follow

multidisciplinary approaches at the intersection of the learning sciences,

psychophysiology, and computer science. Articles I and III focus on the

interpersonal level within triads, and Article II examines sympathetic arousal in the

classroom at the individual level. The connection of the articles to the dissertation

aims and to the data collections is summarized in Table 2.

Table 2. Overview of relevant aspects about the articles.

Article Aim Data collection Individual/Interpersonal

I 1, 3 I Interpersonal

II 2, 3 II Individual

III 1, 2, 4 II Interpersonal

5.1 Article I: Investigating collaborative learning success with physiological coupling indices based on electrodermal activity

This article investigates the interpersonal physiology of EDA in connection to

collaborative learning processes (Aim 1) and to students’ academic achievement

(Aim 3). The article is based on the first data collection of the dissertation (see Sub-

chapter 4.3.1). Participants (N = 24) were high school students tasked with

designing a healthy breakfast for an athlete training for a marathon. The task was

aligned with their science curriculum and fit into their coursework. Students

worked in groups of three to solve the task.

Interpersonal physiology was operationalized through PCIs involving

parameters such as the amplitude (signal matching), rate of change (instantaneous

derivative matching), direction (directional agreement), linear relationship

(Pearson’s correlation coefficient), and weighted correlation (Fisher’s z-transform).

Three collaborative learning measures were explored in relation to the PCIs (Aim

1): 1) the collaborative will, that is, the predisposition to collaborate in a learning

task, measured using the peer learning scale of the MSLQ (Pintrich et al., 1993);

2) the collaborative learning product, that is, the task solution produced by the

students in collaboration; and 3) the dual learning gain, that is, the aggregate of the

individual learning gains as determined by the difference between pre- and post-

70

test scores. The collaborative learning product and the dual learning gain were used

as indicators of academic achievement.

Directional agreement, the percentage of the time that the EDA of the group

members were going simultaneously in the same direction (i.e., upwards,

downwards, or constant), was found to be the best predictor (r = .7) of the dual

learning gain (Aim 3). In other words, the more the team was physiologically

coupled in terms of the direction of their EDA, the more they learned together.

Directional agreement was also the only physiological coupling index found in the

literature that is readily applicable to teams of any size, while the other indices can

only be applied pairwise by definition, meaning that for teams with more than two

group members, the indices have to be computed for all the possible pairs and then

aggregated. The latter is not only more laborious, but also complicates

operationalization and interpretation, since there are several competing approaches

for aggregation. As a physiological coupling index, directional agreement emerged,

thus, as being simpler and more convenient with more explanatory power than

signal matching, instantaneous derivative matching, Pearson’s correlation, and the

Fisher’s z-transform, when considering measures for teams of more than two

individuals.

Instantaneous derivative matching, which accounts for the similarities in the

rate of change of the EDA signal, was found to have a moderate correlation to the

collaborative learning product (r = .59; Aim 3) and to the collaborative will (r = .5;

Aim 1). This is to say that teams for which the rate of change of the EDA signal of

their members was closer produced a better task solution, and the teams with

individuals more prone to collaborate showed a more similar rate of EDA change.

5.2 Article II: Profiling sympathetic arousal in a physics course: How active are students?

This study builds on the notion that learning is an active process (Bjork et al., 2013;

Hebb, 1955) involving cognitive and emotional processes and that arousal is a

physiological component of those processes (Critchley, 2002; Poh et al., 2010).

Learning activations are examined from an EDA perspective. The approach chosen

is necessarily exploratory due to the lack of studies characterizing sympathetic

arousal in the naturalistic setting of an actual class, as opposed to in a laboratory.

Sympathetic arousal is profiled in the classroom (Aim 2)—in terms of incidence

and persistence of low, medium, and high arousal levels—during two consecutive

runs of an elective advanced high school physics course, which is offered twice to

71

the students during the spring term. Incidence was described in terms of percentage

relative to the time of each lesson, which was 75 min. Persistence was

operationalized as the continuous duration of a certain arousal level (i.e., without

switches of level in between). Participants in the data collection (see Sub-chapter

4.3.2) were high school students (N = 24) who were randomly selected from the

course population. In the course exam, the relation between arousal and

performance was also explored (Aim 3) in the light of the well-known Yerkes-

Dodson law (Yerkes & Dodson, 1908) as theoretical background.

EDA was measured unobtrusively during the course lessons and exam via the

Empatica® E4 wristband to index arousal by the frequency of SCRs, which

resemble peaks in the signal. Based on the peaks/min (ppm), the level of arousal

was categorized into low, medium and high (Boucsein, 2012, p. 222; M. E. Dawson

et al., 2017, p. 225). There were at least six changes of arousal level within the

lessons, with an average of 48, representing a change every 1.5 min. High arousal

was the only level that did not occur for each student in every lesson. On average,

low arousal was the level of highest incidence (60% of the lesson) and longest

persistence, lasting on average three times longer than medium arousal and two

times longer than high arousal level occurrences.

In general, this picture of pervasive low arousal in the classroom is aligned

with a similar finding from a study involving three computer-based learning

environments (R. Baker et al., 2010), where boredom (i.e., a low arousal state) was

found to be the most persistent state, in contrast to others in the high arousal area,

such as confusion, frustration, and engaged concentration. This might be

considered, at best, to be neutral and, at worst, detrimental to learning.

During the course exam, arousal was positively and highly correlated (r = .66)

with achievement (Aim 3) as measured by the students’ grades; the more active the

students in terms of sympathetic arousal, the higher the grades and vice versa. The

study adds to the extant literature by providing evidence supporting the left half of

the Yerkes-Dodson curve (i.e., non-detrimental arousal) during an authentic exam

situation, from a sympathetic arousal perspective as indexed by EDA.

5.3 Article III: Sympathetic arousal commonalities and arousal contagion during collaborative learning: How attuned are triad members?

This paper explored sympathetic arousal during collaborative learning in the

naturalistic setting of the classroom (Aims 1 & 2) by studying, first, the

72

commonalities in the direction and level (i.e., low, medium, high) of arousal in a

triad, and, second, the phenomenon of arousal contagion (Aim 4) inspired by the

notion of emotional contagion (Hatfield et al., 1994). The study benefited from the

affordances of the Empatica® E4 biosensor wristband for unobtrusive and

continuous measurement of EDA to index sympathetic arousal. EDA was measured

in 24 high school students working in triads during the lessons of two runs of an

elective advanced physics course (see Sub-chapter 4.3.2).

Most of the time (≈60–95% of the lesson) the triad members were at different

arousal levels, and, when they were at the same level, it was mainly the low arousal

(or deactivated) level. The low percentage of simultaneous medium or high arousal

suggests that, when the triad is doing something, the work is being led or conducted

by one or two of the triad members who have been termed task-doers or the people

who are leading the problem-solving at the moment (Miyake & Kirschner, 2014).

Meanwhile, the other member(s) is/are less involved or engaged. Given the

synchronicity and coordination presupposed in collaborative learning as a jointly

produced activity (Enyedy & Stevens, 2014; Roschelle & Teasley, 1995), where

joint attention is paramount (Tomasello, 1995), and given the close relation

between arousal and attention (Sharot & Phelps, 2004), the results provide evidence

for the claim that only certain specific phases of group work are truly collaborative

(M. Baker, 2015).

Possible within-triad arousal contagion cases (Aim 4) took place mostly on a

1:1 basis (71.3%) and with a latency from within a few seconds up to 10 min, but

usually within 1 min. The predominance of arousal contagion on a 1:1 basis

suggests that the strongest interactions, including influence, occur mostly at a pair

level, even if the team is larger, a triad in this case. In addition, the study illustrates

the potential of physiological measures to enrich and strengthen research on

collaborative learning by exploring a new form of data using a computational

approach with high granularity to characterize the process, which is significantly

less labor-intensive than traditional approaches, such as conversation and video

analyses, and offers affordances which are all welcome and needed in the field

(Wise & Schwarz, 2017).

73

6 Main findings and discussion In connection to the dissertation aims, the main findings are presented and

discussed in this chapter. Results linked to Aims 1 and 2 are presented and

discussed in Sub-chapter 6.1. The relation of academic achievement to the

individual and interpersonal physiology of EDA (Aim 3) is examined in Sub-

chapters 6.2 (individual) and 6.3 (interpersonal). Finally, the sympathetic arousal

contagion among group members as a result of the cognitive-affective social

interactions in collaborative learning (Aim 4) is discussed in Sub-chapter 6.4.

6.1 Triad arousal levels in collaborative learning

In the naturalistic space for formal learning, the classroom, low arousal had the

highest incidence among activation levels (i.e., low, medium, and high). On average,

students spent 60% of each lesson at a low activation level, accounting for more

than half of the class. These results are conspicuous in the light of, on the one hand,

the interest that is reasonable to assume in the students, as the advanced physics

course subject of study is elective, and on the other hand, the intricate

interconnection of interest, attention, engagement, and arousal. In addition,

episodes of low arousal were the most persistent, averaging three times longer than

those of medium arousal and two times longer than high arousal level occurrences.

Taken together, these findings provide a profile of sympathetic arousal in the

classroom at the individual level (Aim 1).

This picture of pervasive low arousal in the classroom, in general, is aligned

with a similar finding from a study involving three computer-based learning

environments (R. Baker et al., 2010), where boredom (i.e., a low arousal state) was

found to be the most persistent state, in contrast to others in the high arousal area,

such as confusion, frustration, and engaged concentration. Therefore, in both digital

and physical learning environments, low arousal levels seem to be the standard.

Unfortunately, students at a low arousal level are considered, in general, to be

disengaged (Pekrun et al., 2002), and at best, to be accumulating shallow facts in

comfortable learning environments without challenges, which does not yield deep

learning (D’Mello & Graesser, 2012). It has to be acknowledged that, as much as

moments of relaxation facilitate disengagement, low arousal might serve the

recovery purposes necessary in between intense periods of cognitive efforts and

could also reinforce motivation for the next stage of learning (Pekrun et al., 2002).

Consequently, low arousal plays a role in learning, but, ideally, the proportion of

74

low arousal should be small in comparison with states of activation (i.e., medium

and high arousal), as learning is most effective when it is active (Bjork et al., 2013;

Hebb, 1955).

The predominance of low arousal (i.e., low activity or inactive students) in

class, even when students themselves choose to enroll in the course, which

presupposes interest, might support the view that, too often, school tasks and

practices are inappropriate for the intended learning goals (Scardamalia & Bereiter,

2014), thus failing to engage even interested learners. Moreover, it supports the

claim that more attention should be paid and resources dedicated to tackling the

problem of bored students (or inactive ones, in general), than dealing with confused

or frustrated students (R. Baker et al., 2010), because the latter both exist less often

and have shorter persistence on average.

The previous result (i.e., predominance of low arousal in the classroom) has an

individual level scope. At the group level, this dissertation explored the

commonality in the arousal levels of triad members during collaborative learning

(Aim 1 and group level part of Aim 2). The triad members were at different arousal

levels most of the time (≈60–95% of the lesson). When they happened to be at the

same level, it was mainly the low arousal (or deactivated) level, while <4% of the

lesson the triad members were simultaneously at the high arousal level. As far as

the synchronicity and coordination presupposed in collaborative learning, this

result is aligned with findings of systematic inequalities in participation among

group members (Cohen, 1994), infrequency of true collaboration (Barron, 2003),

and claims that, most likely, only specific phases of group work are collaborative

(M. Baker, 2015). In the same vein, Dillenbourg (1996) proposes to refer to precise

categories of interactions rather than to collaboration.

The use of interactive activities, such as communication about and

coordination of activities (e.g., transactive activities; see P. A. Kirschner et al., 2018)

during the learning process, is ascribed to positively affect the learning outcome

(Vogel & Weinberger, 2018). However, the difference in the triad members’ arousal

levels might indicate a division of labor, or that the students are taking turns in

doing the collaborative work (i.e., alternating task-doers). Some spontaneous

division of labor may occur in collaboration, in which there are two roles, task-doer

and observer, corresponding to task execution and task monitoring, respectively

(Miyake, 1986). The distribution of roles depends on the nature of the task and may

change frequently (Dillenbourg et al., 1996). In any case, division of labor is a

feature associated with cooperation rather than collaboration (Dillenbourg, 1999).

Although cooperation and collaboration are hardly separable in practice when it

75

comes to group work (Jeong & Hmelo-Silver, 2016), the finding of triad members

being mostly at different arousal levels points to cooperation taking the lion’s share

of group work.

When given a group task, it has been shown that students tend to approach it

as individual work, unless there is some reason for the group to interact (Cohen,

1994). Hertz-Lazarowitz (1989) distinguishes between simple (low) and complex

(high) collaborative group tasks. According to her, the former tends to be

characterized by interactions among the means and the product, while in the latter,

students interact about the process to do the work together, involving discussions

on planning and decision-making, which she argues lead to more high-level

elaboration and, consequently, to more effective learning. Collaboration needs to

be structured beyond physical proximity to others, discussing material with other

students, helping other students, or sharing materials with other students (Johnson

& Johnson, 2002). The importance of coordinated and cohesive participation has

been repeatedly emphasized by collaborative learning research (e.g., Barron, 2000;

Fransen, Weinberger, & Kirschner, 2013; Kreijns, Kirschner, & Jochems, 2003).

6.2 Arousal during exams and achievement

The relation between individual physiology and academic achievement is part of

the third aim of this dissertation. A high, positive, and statistically significant

Pearson’s correlation (r = .66; p = .02) was found between the number of

sympathetic arousal episodes of the students during the course exam, as indexed by

the count of SCRs, and their achievement, as measured by their exam grades. The

more active the students, the higher the grades and vice versa. This result is aligned

with the well-known Yerkes-Dodson law (Hebb, 1955; Yerkes & Dodson, 1908)

and places our observation space on the left half of the inverted-U curve that

characterizes performance as a function of arousal according to the law. The

inverted-U shape describes how performance improves with arousal but only up to

a point from where any further increase in arousal (e.g., excessive stress or strong

anxiety) impairs performance (see Fig. 3). Because the participants were high

achievers in the previous physics course, it is reasonable to presuppose that they

would have high physics self-efficacy (Bandura, 1982) or confidence in their

capacity to succeed in the exam. Therefore, stress and anxiety at the level that

hinders performance was neither expected nor observed. The Yerkes-Dodson law

has been previously tested across a variety of cognitive tasks in laboratory settings

(Sanbonmatsu & Kardes, 1988). Arousal during exams has been explored in

76

relation to academic emotional reactions (especially test anxiety), using a variety

of physiological responses, namely, those from the cardiovascular, endocrine, and

immune systems, but not from the perspective of pure sympathetic arousal

(Spangler, 1997). This study adds to the extant literature by providing evidence

supporting the left half the Yerkes-Dodson curve (i.e., non-detrimental arousal)

during an authentic exam situation, from a sympathetic arousal perspective as

indexed by EDA.

6.3 Pairwise directional agreement and dual learning gain

The second part of aim three of this dissertation concerns the relation between

interpersonal physiology and academic achievement. Directional agreement

measures the percentage of the time that the signals of a number of individuals are

going in the same direction (i.e., upwards, downwards, or constant). During the

collaborative task of designing a healthy breakfast for a marathon athlete, EDA-

based directional agreement was found to have a strong correlation (r = .7) to the

dual learning gain, the aggregation of individual learning gains, which were

determined by means of pre- and post-tests. In other words, at the pairwise level,

students whose sympathetic activations as indexed by their EDA signals were more

attuned in terms of direction learned better together. The commonalities in EDA

direction might reflect a process of joint attention, as sympathetic activation

(arousal) has been shown to increase with attention (Critchley, 2002). Joint

attention, in turn, might be a result of the mutual engagement and coordinated effort

to solve the problem together, which defines collaborative learning (Roschelle &

Teasley, 1995).

Directional agreement explained almost half (r2 = .49) of the variation in the

dual learning gain. Of the PCIs explored (i.e., signal matching, instantaneous

derivative matching, directional agreement, linear correlation, and Fisher’s z-

transform), directional agreement proved to be the best predictor of this indicator,

which is in agreement with a former study showing directional agreement as the

most sensitive PCI to differences in team performance (Elkins et al., 2009).

This is a useful result, considering, on one hand, that directional agreement is

simpler than the other PCIs studied and, on the other hand, that it is the only one

which scales up best (i.e., it can be calculated directly for groups of any size, as

compared to pairwise only calculations of the other PCIs).

77

6.4 Within-triad high arousal contagion in collaborative learning

The fourth aim of the dissertation pursues the exploration of sympathetic arousal

contagion among group members during collaborative learning. The cognitive-

affective social interactions which take place in collaborative learning act as a

mechanism through which emotional contagion (Hatfield et al., 1994) might occur

(Thorson et al., 2018). Being that arousal is one of the two dimensions of emotion

in the circumplex model of affect (Russell, 1980), it motivated the notion of arousal

contagion explored in this dissertation, derived from the emotional contagion

phenomenon. The focus is on high arousal contagion as the most intense activation

level.

At the individual level, 635 high arousal intervals occurred, of which 263 (41%)

might have caused or have been caused by arousal contagion, meaning that they

took place at the time that some other triad member was experiencing a high arousal

interval. In other words, a triad member in high arousal causes a teammate (or two)

to reach that level of activation in 4 of 10 intervals of high arousal. It is worth noting

that this ratio serves as an upper boundary, since the other triad member(s) might

have come to the high arousal level on his or her own through mechanisms other

than arousal contagion. Nonetheless, to assume that, most likely, triad members

might be influencing one another appears reasonable, given that, although not

impossible, it is difficult to isolate oneself in a collaborative learning setting

(Miyake & Kirschner, 2014).

The possible arousal contagion cases took place mostly (71.3%) on a 1:1 basis,

and 1:2 contagion cases accounted for about 15%, while cases where two triad

members in high arousal could have brought the third member to that activation

level (i.e., 2:1) through arousal contagion represented almost 14%. The fact that the

majority of arousal contagion cases happened on a 1:1 basis suggests that, although

there are three students working (learning in this case) together, the interaction

seems to be mostly between two of the members, not necessarily always the same

two, but between two rather than among the three of them. Although a priori it

would seem more probable that two triad members in high arousal provoke the third

to be highly active, compared to one provoking the other two to reach a high arousal

state, the results showed that arousal contagion happened quite similarly on a 1:2

basis as on a 2:1 basis, the former being only 1% higher than the latter.

In terms of latency, cases of possible arousal contagion were found to occur

from within a few seconds up to close to 10 min. Most of the possible high arousal

contagion happened within one min, and from then on, the chances of contagion

78

tended to decrease with time. Provided that arousal contagion occurred, its latency

seemed to be, therefore, relatively short. This might suggest that arousal contagion

is most likely to happen if the high arousal cues are made available early by the

triad member(s) at that activation level. Longer latency values (e.g., from 5 to

10 min) might be explained in several ways. On one hand, no visible cue might be

available to indicate that another triad member is at a high arousal level (Thorson

et al., 2018). On the other hand, the cue might be visible, but the appraisal could

differ at first (e.g., the to be influenced triad member does not see a need to become

active at that moment according to the situation or his/her own motivation to

participate), and the situation might be reappraised a few min later (e.g., due to a

persistent cue or a situation change; Koole, 2009).

The directionality of physiological influence (i.e., who influences whom) has

been used as indicator of group leadership (Chanel & Muhl, 2015). To that extent,

the triad members leading the contagion were examined. Evidence of all possible

profiles was found. Thus, in some cases, there was a predominant member of the

triad from whom the arousal contagion arises or two predominant members, while

in other cases, it is quite evenly originated from all three members. This result

aligns with the reported distinct dynamics of collaborative learning unfolding in

different triads (see Enyedy & Stevens, 2014; Miyake & Kirschner, 2014). In fact,

the importance of developing ways to study interaction quality over time is

underscored by reports that suggest large variations in the ways that group

interactions can unfold (Barron, 2000).

79

7 Conclusions Collaborative skills have been largely taken for granted, but research shows that

they are not an automatic consequence of group work and that training is necessary

for effective collaborative groups (Cohen, 1994). Collaborative learning has been

shown to be ineffective and even counterproductive if it is not properly

implemented (M. Baker, 2015; P. A. Kirschner et al., 2018; Kuhn, 2015; Mullins et

al., 2011). Consequently, researchers on educational psychology, teachers, and

policy-makers, have long been interested in the study of collaborative learning and

the devising of metrics to assist in distinguishing effective from ineffective

collaboration, so that these results could be used in training. Wise and Schwarz

(2017) use the term collaborative learning analytics. Few instruments in

educational psychology capture joint action, and there is a need to provide accounts

of interactions that capture the dynamic interplay in collaborative learning (Barron,

2000). Although the extent to which the evaluation of learning can be based only

on group processes is debatable, some phenomena may emerge only in learners’

interactions, such as co-constructed knowledge or social skills (Vogel &

Weinberger, 2018).

Interactions are reflected in a variety of modalities, including verbal and non-

verbal (Enyedy & Stevens, 2014). In a dominantly descriptive design, studies of

collaborative learning have mainly involved conversation analyses and surveys,

including questionnaires (e.g., open-ended, multiple-choice, and Likert scales) and

interviews (Jeong et al., 2014). However, those analyses are time-consuming, very

expensive microgenetic methods, and do not scale up well (Wise & Schwarz, 2017),

which has led to the consideration of alternative data modalities to support the

characterization of interactions. One such modality is physiological data, which has

a tradition in the study of cognitive-affective processes in psychophysiological

research, although it is rarely employed in educational psychology. Physiological

data might assist in the instantaneous characterization of interactions according to

parameters, such as commonalities and influence among peers, and could help to

identify fluctuations in attention and coordination, features of interaction that are

not often described in research on collaborative learning processes (Barron, 2000).

One of the most popular and widely studied physiological signals in

psychophysiology is EDA, regarded as one of the simplest and most readily

accessible indicators of sympathetic arousal. In turn, sympathetic arousal is an

important psychological construct at the intersection of mind and body (Mandler,

80

1984), connected to cognitive processes, especially attention (Raskin, 1973), and

to affective processes through the circumplex model of affect (Russell, 1980).

This dissertation was conducted in the light of the gaps identified in the

literature, the need for scalable ways to characterize interactions in collaborative

learning, the theoretical frameworks of collaborative learning research, EDA,

sympathetic arousal and interpersonal physiology, and the affordances of a

wristband biosensor for an unobtrusive, continuous recording of EDA in

collaborative learning settings. The overall aim was the exploration of EDA and

derivative measures, such as sympathetic arousal and PCIs, concomitant with

individual and interpersonal cognitive and affective processes, respectively, during

collaborative learning in a naturalistic setting.

At the individual level, students spent more than half of the class at a low

arousal level, possibly signaling disengagement. At the group level, triad members

were most of the time at a different arousal level, and when the levels coincided, it

was mainly at the low arousal level. The difference might indicate that students

took turns (alternating task-doers) in executing the task, or applied some division

of labor, rather than truly collaborating. In terms of achievement, two physiological

indices were found to relate to learning gains and exam grades. In the first data

collection, pairwise EDA-indexed directional agreement was positively and highly

correlated to dual learning gain. The more the EDA signals moved in the same

direction, the better students learned together. In the second data collection, arousal

during the exam, as indexed by the total count of SCRs (peaks in the EDA signal),

was a predictor of the students’ exam grades. The more active they were, the better

their achievement. Arousal is also known to impair performance when the level is

too high, for example, due to excessive stress or strong anxiety (Yerkes-Dodson

law), but those levels were not reached in our sample, probably because the students

were high achievers in the preceding level course of the same subject. In terms of

influence from a physiological perspective, 41% of the high arousal intervals that

were found could have caused or have been caused by arousal contagion, a

construct derived in this dissertation from the phenomenon of emotional contagion

and the classical consideration of arousal as the activation dimension of emotion.

Furthermore, the possible arousal contagion cases were found to take place mostly

(71.3%) on a 1:1 basis, indicating that interactions in the collaborative learning

triad seem to occur mostly between two of the three members, not necessarily

always the same two, but between two rather than among the three of them.

From the methodological point of view, two significant contributions were

made by this dissertation. First, the scheme for classifying sympathetic arousal

81

levels into low, medium and high was obtained from the combination of different

literature sources and results. Second, and also based on results reported in the

literature, the criteria were established to disregard as noise those EDA recordings

where the number of peaks in the signal (SCRs) is lower than the minimum

expected theoretically.

A panoply of data sources and their corresponding methodological approaches

have been used in collaborative learning research. This dissertation has delved into

EDA, drawing from and building on the psychophysiological research tradition, at

the individual and interpersonal levels, and applied to the study of collaborative

learning. Importantly, the data were collected in ecologically valid settings, either

classrooms or classroom-like environments.

This work is part of the quest to advance learning research by enriching

traditional subjective data sources (e.g., questionnaires and interviews) with

objective online measures of learning processes (Azevedo, 2015; Boekaerts &

Corno, 2005; Winne, 2010). The current state of the art makes it imperative to focus

on the physiological signal at this stage as a first integration step, but it is my hope

that the studies comprising this dissertation will serve as the basis for including

physiological signals as a regular practice in data collection and experiment designs

in the learning sciences. As many other researchers in the field (e.g., Azevedo, 2015;

Boekaerts & Corno, 2005; Sanna Järvelä, Hadwin, Malmberg, & Miller, 2018;

Panadero, Klug, & Järvelä, 2015), I believe that triangulation of different data

modalities, including physiological data, has the potential to provide stronger

evidence and lead us as a community to draw sounder conclusions.

Several directions could be considered for future work. First, more

physiological signals can be explored as long as it makes theoretical sense. Second,

other indices could be developed or reused from the literature. It is important to

keep it simple for interpretability and practical usefulness. Previous research

reported that the simplest measures used were shown to be the most sensitive ones,

which has led to the claim that physiological indices, both at individual and

interpersonal levels, should be straightforward and uncomplicated (Elkins et al.,

2009), which extends to other measures as well. Third, the study of the combination

of indices from different signals and/or data modalities (e.g., physiological, self-

reported, and observation), since the eyes of science are on multimodal approaches,

should be pursued so that the limitations of individual modalities can be overcome,

leading to higher validity of the conclusions drawn. Physiological data, as opposed

to self-reports, is objective, but cannot be interpreted without knowledge of the

context. In this sense, qualitative data are needed to further characterize and

82

understand the variations in PCIs and sympathetic arousal at individual and

interpersonal levels, in terms of, for example, why a triad reaches full directional

agreement in one lesson but not in another, and what situations cause triad members

to sustain full arousal directional agreement during several min. Fourth, although

here a general profile of the class and relations of the physiological measures to

academic achievement were pursued, the methodology can be applied to the study

of specific events of interest (e.g., directional agreement or arousal levels within a

response window).

The physiological indices used in this dissertation could be fed back to the

students via a learning analytics dashboard to support their reflection upon and

awareness of the learning process. Both retrospective and real-time applications

(e.g., alarms when certain thresholds are surpassed) are possible. It is known that

individuals vary in their ability to perceive their autonomic physiological state,

known as interoceptive awareness, which they need to inform cognitions and

behaviors (Critchley et al., 2013). For example, people rely partly on information

from their physiological state in judging their capabilities (Bandura, 1982). At the

collaboration level, it has been argued that groups should become aware of their

interpersonal processes and take time to discuss how they are doing as a group

(Cohen, 1994). For both individual and interpersonal level purposes, the

incorporation of physiological measures would enrich and provide a new

dimension to learning analytics dashboards, which to date largely focus on log data

(Schwendimann et al., 2017). From the teacher’s perspective, such a dashboard

could assist in group composition (Ahonen et al., 2018) and identify those activities

which cause more students to be active and engaged. Similarly, it could help in

distinguishing those tasks leading to actual group work rather than to alternating

task doers or division of labor. In sum, the findings of this study have applications

to learning dashboards, prompts (e.g., in intelligent tutoring systems or online

learning environments), and in general, to the development of tools for better

collaboration.

83

List of references Ahn, S., Cho, H., Kwon, M., Kim, K., Kwon, H., Kim, B. S., … Jun, S. C. (2018). Interbrain

phase synchronization during turn-taking verbal interaction – A hyperscanning study using simultaneous EEG/MEG. Human Brain Mapping, 39(1), 171–188. https://doi.org/10.1002/hbm.23834

Ahonen, L., Cowley, B., Torniainen, J., Ukkonen, A., Vihavainen, A., & Puolamäki, K. (2016). Cognitive collaboration found in cardiac physiology: Study in classroom environment. PLoS One, 11(7), 1–16. https://doi.org/10.1371/journal.pone.0159178

Ahonen, L., Cowley, B. U., Hellas, A., & Puolamäki, K. (2018). Biosignals reflect pair-dynamics in collaborative work: EDA and ECG study of pair-programming in a classroom environment. Scientific Reports, 8(1), 1–16. https://doi.org/10.1038/s41598-018-21518-3

Aritzeta, A., Soroa, G., Balluerka, N., Muela, A., Gorostiaga, A., & Aliri, J. (2017). Reducing anxiety and improving academic performance through a biofeedback relaxation training program. Applied Psychophysiology and Biofeedback, 42(3), 193–202. https://doi.org/10.1007/s10484-017-9367-z

Aronson, E., Blaney, N., Sikes, J., & Snapp, M. (1978). The jigsaw classroom. Beverly Hills, CA: SAGE.

Aronson, E., & Patnoe, S. (2011). Cooperation in the classroom: The jigsaw method (3rd ed.). London, UK: Pinter & Martin.

Arroyo, I., Cooper, D. G., Burleson, W., Woolf, B. P., Muldner, K., & Christopherson, R. (2009). Emotion sensors go to school. In Proceedings of the Conference on Artificial Intelligence in Education (pp. 17–24). Amsterdam, Netherlands: IOS Press.

Azevedo, R. (2015). Defining and measuring engagement and learning in science: Conceptual, theoretical, methodological, and analytical issues. Educational Psychologist, 50(1), 84–94. https://doi.org/10.1080/00461520.2015.1004069

Baker, M. (2015). Collaboration in collaborative learning. Interaction Studies, 16(3), 451–473. https://doi.org/10.1075/is.16.3.05bak

Baker, R., D’Mello, S. K., Rodrigo, M. T., & Graesser, A. C. (2010). Better to be frustrated than bored: The incidence and persistence of affect during interactions with three different computer-based learning environments. International Journal of Human-Computer Studies, 68(4), 223–241. https://doi.org/10.1016/j.ijhcs.2009.12.003

Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist, 37(2), 122–147. https://doi.org/10.1037/0003-066X.37.2.122

Barron, B. (2000). Achieving coordination in collaborative problem-solving groups. Journal of the Learning Sciences, 9(4), 403–436. https://doi.org/10.1207/S15327809JLS0904

Barron, B. (2003). When smart groups fail. Journal of the Learning Sciences, 12(3), 307–359. https://doi.org/10.1207/S15327809JLS1203_1

Bear, J. B., & Woolley, A. W. (2011). The role of gender in team collaboration and performance. Interdisciplinary Science Reviews, 36(2), 146–153. https://doi.org/10.1179/030801811X13013181961473

Beers, P. J., Boshuizen, H. P. A., Kirschner, P. A., & Gijselaers, W. H. (2007). The analysis

84

of negotiation of common ground in CSCL. Learning and Instruction, 17(4), 427–435. https://doi.org/10.1016/J.LEARNINSTRUC.2007.04.002

Benedek, M., & Kaernbach, C. (2010a). A continuous measure of phasic electrodermal activity. Journal of Neuroscience Methods, 190(1), 80–91. https://doi.org/10.1016/j.jneumeth.2010.04.028

Benedek, M., & Kaernbach, C. (2010b). Decomposition of skin conductance data by means of nonnegative deconvolution. Psychophysiology, 47(4), 647–658. https://doi.org/10.1111/j.1469-8986.2009.00972.x

Berger, S. M. (1962). Conditioning through vicarious instigation. Psychological Review, 69(5), 450–466. https://doi.org/10.1037/h0046466

Bettiga, D., Lamberti, L., & Noci, G. (2017). Do mind and body agree? Unconscious versus conscious arousal in product attitude formation. Journal of Business Research, 75(June), 108–117. https://doi.org/10.1016/j.jbusres.2017.02.008

Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-regulated learning: Beliefs, techniques, and illusions. Annual Review of Psychology, 64, 417–444. https://doi.org/10.1146/annurev-psych-113011-143823

Blikstein, P., & Worsley, M. (2016). Multimodal learning analytics and education data mining: using computational technologies to measure complex learning tasks. Journal of Learning Analytics, 3(2), 220–238. https://doi.org/http://dx.doi.org/10.18608/jla.2016.32.11

Boekaerts, M., & Corno, L. (2005). Self regulation in the classroom: A perspective on assessment and intervention. Applied Psychology, 54(2), 199–231. https://doi.org/10.1111/j.1464-0597.2005.00205.x

Boucsein, W. (1992). Electrodermal activity. New York, NY: Plenum Press. Boucsein, W. (2012). Electrodermal activity (2nd ed.). New York, NY: Springer. Boucsein, W., & Backs, R. W. (2000). Engineering psychophysiology as a discipline:

Historical and theoretical aspects. In R. W. Backs & W. Boucsein (Eds.), Engineering psychophysiology. Issues and applications (pp. 3–30). Mahwah, NJ: Erlbaum.

Boucsein, W., Fowles, D. C., Grimnes, S., Ben-Shakhar, G., Roth, W. T., Dawson, M. E., & Filion, D. L. (2012). Publication recommendations for electrodermal measurements. Psychophysiology, 49(8), 1017–1034. https://doi.org/10.1111/j.1469-8986.2012.01384.x

Braithwaite, J. J., Watson, D. G., Jones, R., & Rowe, M. (2015). A guide for analysing electrodermal activity (EDA) & skin conductance responses (SCRs) for psychological experiments. Birmingham, United Kingdom. Retrieved from https://www.birmingham.ac.uk/Documents/college-les/psych/saal/guide-electrodermal-activity.pdf

Brown, A. L. (1992). Design experiments: Theoretical and methodological challenges in creating complex interventions in classroom settings. Journal of the Learning Sciences, 2(2), 141–178.

Bruer, J. T. (1994). Classroom problems, school culture, and cognitive research. In K. McGilly (Ed.), Classroom lessons: Integrating cognitive theory and classroom practice (pp. 273–290). Cambridge, MA: MIT Press.

85

Butterworth, G. (1982). A brief account of the conflict between the individual & the social in models of cognitive growth. In G. Butterworth & P. Light (Eds.), Social cognition: Studies of the development of understanding (pp. 3–16). Brighton, United Kingdom: Harvester Press.

Cacioppo, J. T., & Tassinary, L. G. (1990a). Inferring psychological significance from physiological signals. American Psychologist, 45(1), 16–28. https://doi.org/10.1037/0003-066X.45.1.16

Cacioppo, J. T., & Tassinary, L. G. (1990b). Psychophysiology and psychophysiological inference. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of psychophysiology: Physical, social, and inferential elements (pp. 3–33). New York, NY: Cambridge University Press.

Calderón, O. (2016). Oximetry: A reflective tool for the detection of physiological expression of emotions in a science education classroom. Cultural Studies of Science Education, 11(3), 653–667. https://doi.org/10.1007/s11422-016-9731-y

Chanel, G., Bétrancourt, M., Pun, T., Cereghetti, D., & Molinari, G. (2013). Assessment of computer-supported collaborative processes using interpersonal physiological and eye-movement coupling. In Humaine Association Conference on Affective Computing and Intelligent Interaction (pp. 116–122). Geneva, Switzerland: IEEE. https://doi.org/10.1109/ACII.2013.26

Chanel, G., Kivikangas, J. M., & Ravaja, N. (2012). Physiological compliance for social gaming analysis: Cooperative versus competitive play. Interacting with Computers, 24(4), 306–316. https://doi.org/10.1016/j.intcom.2012.04.012

Chanel, G., & Muhl, C. (2015). Connecting brains and bodies: Applying physiological computing to support social interaction. Interacting with Computers, 27(5), 534–550. https://doi.org/10.1093/iwc/iwv013

Chi, M. T. H. (1997). Quantifying qualitative analyses of verbal data: A practical guide. Journal of the Learning Sciences, 6(3), 271–315. https://doi.org/10.1207/s15327809jls0603_1

Cohen, E. G. (1994). Restructuring the classroom: Conditions for productive small groups. Review of Educational Research, 64(1), 1–35. https://doi.org/10.3102/00346543064001001

Cooper, D. G., Arroyo, I., Woolf, B. P., Muldner, K., Burleson, W., & Christopherson, R. (2009). Sensors model student self concept in the classroom. In Proceedings of the 17th International Conference on User Modeling, Adaptation, and Personalization (pp. 30–41). Berlin, Germany: Springer. https://doi.org/10.1007/978-3-642-02247-0_6

Critchley, H. D. (2002). Electrodermal responses: What happens in the brain. Neuroscientist, 8(2), 132–142. https://doi.org/10.1177/107385840200800209

Critchley, H. D., Eccles, J., & Garfinkel, S. N. (2013). Interaction between cognition, emotion, and the autonomic nervous system. In R. M. Buijs & D. F. Swaab (Eds.), Handbook of Clinical Neurology (Vol. 117, pp. 59–77). Amsterdam, Netherlands: Elsevier. https://doi.org/10.1016/B978-0-444-53491-0.00006-7

Critchley, H. D., & Garfinkel, S. N. (2018). The influence of physiological signals on

86

cognition. Current Opinion in Behavioral Sciences, 19(6), 13–18. https://doi.org/10.1016/j.cobeha.2017.08.014

D’Mello, S. K., & Graesser, A. (2012). Dynamics of affective states during complex learning. Learning and Instruction, 22(2), 145–157. https://doi.org/10.1016/J.LEARNINSTRUC.2011.10.001

Damasio, A. R., Tranel, D., & Damasio, H. C. (1991). Somatic markers and the guidance of behavior: Theory and preliminary testing. In H. S. Levin, H. M. Eisenberg, & A. L. Benton (Eds.), Frontal Lobe Function and Dysfunction. New York, NY: Oxford University Press.

Dawson, M. E., Schell, A. M., & Filion, D. L. (1990). The electrodermal system. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of psychophysiology: Physical, social, and inferential elements (pp. 295–324). New York, NY: Cambridge University Press.

Dawson, M. E., Schell, A. M., & Filion, D. L. (2007). The electrodermal system. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.), Handbook of psychophysiology (3rd ed., pp. 159–181). Cambridge, United Kingdom: Cambridge University Press.

Dawson, M. E., Schell, A. M., & Filion, D. L. (2017). The electrodermal system. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.), Handbook of psychophysiology (4th ed., pp. 217–243). Cambridge, United Kingdom: Cambridge University Press.

Dawson, S., & Siemens, G. (2014). Analytics to literacies: The development of a learning analytics framework for multiliteracies assessment. The International Review of Research in Open and Distance Learning, 15(4), 284–305.

De Lecea, L., Carter, M. E., & Adamantidis, A. (2012). Shining light on wakefulness and arousal. Biological Psychiatry, 79(12), 1046–1052. https://doi.org/10.1016/j.biopsych.2012.01.032

Delaherche, E., Chetouani, M., Mahdhaoui, A., Saint-Georges, C., Viaux, S., & Cohen, D. (2012). Interpersonal synchrony: A survey of evaluation methods across disciplines. IEEE Transactions on Affective Computing, 3(3), 349–365. https://doi.org/10.1109/T-AFFC.2012.12

Dich, Y., Reilly, J., & Schneider, B. (2018). Using physiological synchrony as an indicator of collaboration quality, task performance and learning. In C. P. Rosé, R. Martínez-Maldonado, H. U. Hoppe, R. Luckin, M. Mavrikis, K. Porayska-Pomsta, … B. du Boulay (Eds.), International Conference on Artificial Intelligence in Education (pp. 98–110). London, United Kingdom: Springer. https://doi.org/10.1007/978-3-319-93843-1

Dikker, S., Wan, L., Davidesco, I., Kaggen, L., Oostrik, M., McClintock, J., … Poeppel, D. (2017). Brain-to-brain synchrony tracks real-world dynamic group interactions in the classroom. Current Biology, 27(9), 1375–1380. https://doi.org/10.1016/j.cub.2017.04.002

Dillenbourg, P. (1999). What do you mean by “collaborative learning”? In P. Dillenbourg (Ed.), Collaborative learning: Cognitive and computational approaches (pp. 1–19). Oxford, UK: Elsevier.

Dillenbourg, P., Baker, M., Blaye, A., & O’Malley, C. (1996). The evolution of research on collaborative learning. In P. Reimann & H. Spada (Eds.), Learning in humans and

87

machine: Towards an interdisciplinary learning science (pp. 189–211). Oxford, United Kingdom: Pergamon. https://doi.org/10.1007/978-1-4020-9827-7_1

DiMascio, A., Boyd, R. W., & Greenblatt, M. (1957). Physiological correlates of tension and antagonism during psychotherapy: A study of interpersonal physiology. Psychosomatic Medicine, 19(2), 99–104. https://doi.org/10.1097/00006842-195703000-00002

Doise, W., & Mugny, G. (1984). The social development of the intellect. Oxford, United Kingdom: Pergamon.

Duffy, E. (1962). Activation and behavior. New York, NY: Wiley. Edelberg, R. (1967). Electrical properties of the skin. In C. C. Brown (Ed.), Methods in

psychophysiology (pp. 1–53). Baltimore, MD: Williams & Wilkins. Edelberg, R. (1972a). Electrodermal recovery rate, goal-orientation, and aversion.

Psychophysiology, 9(5), 512–520. https://doi.org/10.1111/j.1469-8986.1972.tb01805.x

Edelberg, R. (1972b). The electrodermal system. In N. S. Greenfield & R. A. Sternbach (Eds.), Handbook of Psychophysiology. New York, NY: Holt.

Edelson, D. C., Gordin, D. N., & Pea, R. D. (1999). Addressing the challenges of inquiry-based learning through technology and curriculum design. Journal of the Learning Sciences, 8(3–4), 391–450. https://doi.org/10.1080/10508406.1999.9672075

Elkins, A. N., Muth, E. R., Hoover, A. W., Walker, A. D., Carpenter, T. L., & Switzer, F. S. (2009). Physiological compliance and team performance. Applied Ergonomics, 40(6), 997–1003. https://doi.org/10.1016/j.apergo.2009.02.002

Empatica Inc. (2017). Have you done comparative studies or validation on electrodermal activity sensor? Retrieved May 28, 2018, from https://support.empatica.com/hc/en-us/articles/203005295-Have-you-done-comparative-studies-or-validation-on-Electrodermal-activity-sensor-

Empatica Inc. (2018). Recommended tools for signal processing and data analysis. Retrieved March 1, 2018, from https://support.empatica.com/hc/en-us/articles/202872739-Recommended-tools-for-signal-processing-and-data-analysis

Enyedy, N., & Stevens, R. (2014). Analyzing collaboration. In R. K. Sawyer (Ed.), The Cambridge handbook of the learning sciences (2nd ed., pp. 191–212). Cambridge, United Kingdom: Cambridge University Press. https://doi.org/10.1017/CBO9781139519526.013

Fadiga, L., Fogassi, L., Pavesi, G., & Rizzolatti, G. (1995). Motor facilitation during action observation: A magnetic stimulation study. Journal of Neurophysiology, 73(6), 2608–2611.

Féré, C. (1888). Note sur les modifications de la résistance électrique sous l’influence des excitations sensorielles et des émotions [Note on the modification of electric resistance under the influence of sensory excitement and emotions]. Comptes Rendus Société de Biologie, 5, 217–219.

Ferreira, E., Ferreira, D., Kim, S., Siirtola, P., Röning, J., Forlizzi, J. F., & Dey, A. K. (2014). Assessing real-time cognitive load based on psycho-physiological measures for younger and older adults. In Computational Intelligence, Cognitive Algorithms, Mind,

88

and Brain (CCMB), 2014 IEEE Symposium on (pp. 39–48). Orlando, FL: IEEE. https://doi.org/10.1109/CCMB.2014.7020692

Field, T. (2012). Relationships as regulators. Psychology, 3(6), 467–479. https://doi.org/http://dx.doi.org/10.4236/psych.2012.36066

Finger, B., & Murphy, R. O. (2011). Using skin conductance in judgment and decision making research. In M. Schulte-Mecklenbeck, A. Kuehberger, & R. Ranyard (Eds.), A handbook of process tracing methods for decision research (pp. 163–184). New York, NY: Psychology Press.

Finnish Advisory Board on Research Ethics. (2009). Ethical principles of research in the humanities and social and behavioural sciences and proposals for ethical review. Helsinki. Retrieved from http://www.tenk.fi/sites/tenk.fi/files/ethicalprinciples.pdf

Finnish Advisory Board on Research Integrity. (2012). Responsible conduct of research and procedures for handling allegations of misconduct in Finland. Helsinki. Retrieved from http://www.tenk.fi/sites/tenk.fi/files/HTK_ohje_2012.pdf

Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do threatening stimuli draw or hold visual attention in subclinical anxiety? Journal of Experimental Psychology: General, 130(4), 681–700.

Fransen, J., Weinberger, A., & Kirschner, P. A. (2013). Team effectiveness and team development in CSCL. Educational Psychologist, 48(1), 9–24. https://doi.org/10.1080/00461520.2012.747947

Frijda, N. H. (1986). The emotions. Cambridge, United Kingdom: Cambridge University Press.

Garbarino, M., Lai, M., Tognetti, S., Picard, R. W., & Bender, D. (2014). Empatica E3 – A wearable wireless multi-sensor device for real-time computerized biofeedback and data acquisition. In International Conference on Wireless Mobile Communication and Healthcare (pp. 39–42). Athens, Greece: IEEE. https://doi.org/10.4108/icst.mobihealth.2014.257418

Garfinkel, S. N., Critchley, H. D., & Pollatos, O. (2017). The interoceptive system: Implications for cognition, emotion and health. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.), Handbook of psychophysiology (4th ed., pp. 427–443). Cambridge, United Kingdom: Cambridge University Press.

Gillies, R. M., Carroll, A., Cunnington, R., Rafter, M., Palghat, K., Bednark, J., & Bourgeois, A. (2016). Multimodal representations during an inquiry problem-solving activity in a year 6 science class: A case study investigating cooperation, physiological arousal and belief states. Australian Journal of Education, 60(2), 111–127. https://doi.org/10.1177/0004944116650701

Gonzalez-Sanchez, J., Baydogan, M., Chavez-Echeagaray, M. E., Atkinson, R. K., & Burleson, W. (2017). Affect measurement: A roadmap through approaches, technologies, and data analysis. In M. Jeon (Ed.), Emotions and affect in human factors and human-computer interaction (pp. 255–288). London, United Kingdom: Academic Press. https://doi.org/10.1016/B978-0-12-801851-4.00011-2

Goyal, N., & Fussell, S. R. (2017). Intelligent interruption management using electro dermal activity based physiological sensor for collaborative sensemaking. In Proceedings of

89

the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. New York, NY: ACM. https://doi.org/10.1145/3130917

Guastello, S. J., Pincus, D., & Gunderson, P. R. (2006). Electrodermal arousal between participants in a conversation: Nonlinear dynamics and linkage effects. Nonlinear Dynamics, Psychology, and Life Sciences, 10(3), 365–399.

Hanoch, Y., & Vitouch, O. (2004). When less is more: Information, emotional arousal and the ecological reframing of the Yerkes-Dodson law. Theory & Psychology, 14(4), 427–452. https://doi.org/10.1177/0959354304044918

Hatfield, E., Cacioppo, J. T., & Rapson, R. L. (1994). Emotional contagion. New York, NY: Cambridge University Press.

Hebb, D. O. (1955). Drives and the CNS (conceptual nervous system). Psychological Review, 62(4), 243–254. https://doi.org/10.1037/h0041823

Henning, R. A., Armstead, A. G., & Ferris, J. K. (2009). Social psychophysiological compliance in a four-person research team. Applied Ergonomics, 40(6), 1004–1010. https://doi.org/10.1016/j.apergo.2009.04.009

Henning, R. A., Boucsein, W., & Gil, M. C. (2001). Social-physiological compliance as a determinant of team performance. International Journal of Psychophysiology, 40(3), 221–232. https://doi.org/10.1016/S0167-8760(00)00190-2

Hertz-Lazarowitz, R. (1989). Cooperation and helping in the classroom: A contextual approach. International Journal of Educational Research, 13(1), 113–119. https://doi.org/10.1016/0883-0355(89)90020-7

Hoadley, C. (2018). A short history of the learning sciences. In F. Fischer, C. E. Hmelo-Silver, S. R. Goldman, & P. Reimann (Eds.), International handbook of the learning sciences (pp. 11–23). New York, NY: Routledge.

Järvelä, S. [Sanna], Hadwin, A. F., Malmberg, J., & Miller, M. (2018). Contemporary perspectives of regulated learning in collaboration. In F. Fischer, C. E. Hmelo-Silver, S. R. Goldman, & P. Reimann (Eds.), International handbook of the learning sciences (pp. 127–136). New York, NY: Routledge.

Järvelä, S. [Sanna], Kirschner, P. A., Panadero, E., Malmberg, J., Phielix, C., Jaspers, J., … Järvenoja, H. (2015). Enhancing socially shared regulation in collaborative learning groups: Designing for CSCL regulation tools. Educational Technology Research and Development, 63(1), 125–142. https://doi.org/10.1007/s11423-014-9358-1

Järvelä, S. [Simo], Kivikangas, J. M., Katsyri, J., & Ravaja, N. (2014). Physiological linkage of dyadic gaming experience. Simulation & Gaming, 45(1), 24–40. https://doi.org/10.1177/1046878113513080

Järvenoja, H., Järvelä, S., & Malmberg, J. (2015). Understanding regulated learning in situative and contextual frameworks. Educational Psychologist, 50(3), 204–219. https://doi.org/10.1080/00461520.2015.1075400

Jeong, H., & Hmelo-Silver, C. E. (2016). Seven affordances of computer-supported collaborative learning: How to support collaborative learning? How can technologies help? Educational Psychologist, 51(2), 247–265.https://doi.org/10.1080/00461520.2016.1158654

Jeong, H., Hmelo-Silver, C. E., & Yu, Y. (2014). An examination of CSCL methodological

90

practices and the influence of theoretical frameworks 2005–2009. International Journal of Computer-Supported Collaborative Learning, 9(3), 305–334. https://doi.org/10.1007/s11412-014-9198-3

Johnson, D. W., & Johnson, R. T. (2002). Learning together and alone: Overview and meta�analysis. Asia Pacific Journal of Education, 22(1), 95–105. https://doi.org/10.1080/0218879020220110

Juvina, I., Larue, O., & Hough, A. (2018). Modeling valuation and core affect in a cognitive architecture: The impact of valence and arousal on memory and decision-making. Cognitive Systems Research, 48(May), 4–24. https://doi.org/10.1016/j.cogsys.2017.06.002

Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-Hall. Kaplan, H. B. (1967). Physiological correlates (GSR) of affect in small groups. Journal of

Psychosomatic Research, 11(2), 173–179. https://doi.org/10.1016/0022-3999(67)90004-9

Kaplan, H. B., & Bloom, S. W. (1960). The use of sociological and social-psychological concepts in physiological research: A review of selected experimental studies. Journal of Nervous and Mental Disease, 131(2), 128–134. https://doi.org/10.1097/00005053-196008000-00006

Kaplan, H. B., Burch, N. R., Bedner, T. D., & Trenda, J. D. (1965). Physiologic (GSR) activity and perceptions of social behavior in positive, negative and neutral pairs. The Journal of Nervous and Mental Disease, 140(6), 457–463.

Kaplan, H. B., Burch, N. R., Bloom, S. W., & Edelberg, R. (1963). Affective orientation and physiological activity (GSR) in small peer groups. Psychosomatic Medicine, 25(3), 245–252.

Karvonen, A., Kykyri, V. L., Kaartinen, J., Penttonen, M., & Seikkula, J. (2016). Sympathetic nervous system synchrony in couple therapy. Journal of Marital and Family Therapy, 42(3), 383–395. https://doi.org/10.1111/jmft.12152

Kelly, N., Thompson, K., & Yeoman, P. (2015). Theory-led design of instruments and representations in learning analytics: Developing a novel tool for orchestration of online collaborative learning. Journal of Learning Analytics, 2(2), 14–43. https://doi.org/10.18608/jla.2015.22.4

Kim, T., Mcfee, E., Olguin, D. O., Waber, B., & Pentland, A. (2012). Sociometric badges: Using sensor technology to capture new forms of collaboration. Journal of Organizational Behavior, 33(3), 412–427. https://doi.org/10.1002/job.1776

Kirschner, F., Paas, F., & Kirschner, P. A. (2011). Task complexity as a driver for collaborative learning efficiency: The collective working-memory effect. Applied Cognitive Psychology, 25(4), 615–624. https://doi.org/10.1002/acp.1730

Kirschner, P. A., Sweller, J., Kirschner, F., & Zambrano, J. (2018). From cognitive load theory to collaborative cognitive load theory. International Journal of Computer-Supported Collaborative Learning, 13(2), 213–233. https://doi.org/10.1007/s11412-018-9277-y

Knierim, M. T., Rissler, R., Dorner, V., Maedche, A., & Weinhardt, C. (2018). The psychophysiology of flow: A systematic review of peripheral nervous system features.

91

In F. Davis, R. Riedl, J. vom Brocke, P.-M. Léger, & A. B. Randolph (Eds.), Information Systems and Neuroscience: Gmunden Retreat on NeuroIS 2017. Gmunden, Austria: Springer. https://doi.org/10.1007/978-3-319-67431-5_13

Koole, S. L. (2009). The psychology of emotion regulation: An integrative review. Cognition & Emotion, 23(1), 4–41. https://doi.org/10.1080/02699930802619031

Koschmann, T. (1996). Paradigm shifts and instructional technology: An introduction. In T. Koschmann (Ed.), CSCL: Theory and practice of an emerging paradigm (pp. 1–23). Mahwah, NJ: Erlbaum.

Koskimäki, H., Mönttinen, H., Siirtola, P., Huttunen, H.-L., Halonen, R., & Röning, J. (2017). Early detection of migraine attacks based on wearable sensors – Experiences of data collection using Empatica E4. In Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers (pp. 506–511). New York, NY: ACM. https://doi.org/10.1145/3123024.3124434

Kreibig, S. D. (2010). Autonomic nervous system activity in emotion: A review. Biological Psychology, 84(3), 394–421. https://doi.org/10.1016/j.biopsycho.2010.03.010

Kreijns, K., Kirschner, P. A., & Jochems, W. (2003). Identifying the pitfalls for social interaction in computer-supported collaborative learning environments: A review of the research. Computers in Human Behavior, 19(3), 335–353. https://doi.org/10.1016/S0747-5632(02)00057-2

Kuhn, D. (2015). Thinking together and alone. Educational Researcher, 44(1), 46–53. https://doi.org/10.3102/0013189X15569530

Kuppens, P., Tuerlinckx, F., Russell, J. A., & Barrett, L. F. (2013). The relation between valence and arousal in subjective experience. Psychological Bulletin, 139(4), 917–940. https://doi.org/10.1037/a0030811

Lacey, J. I. (1967). Somatic response patterning and stress: Some revisions of activation theory. In M. H. Appley & R. Trumbull (Eds.), Psychological stress: Issues in research (pp. 14–42). New York, NY: Appleton-Century-Crofts.

Liu, S., Zhou, Y., Palumbo, R. V., & Wang, J.-L. (2016). Dynamical correlation: A new method for quantifying synchrony with multivariate intensive longitudinal data. Psychological Methods, 21(3), 291–308. https://doi.org/10.1037/met0000071

Lyytinen, H. (1984). The psychophysiology of anticipation and arousal. Jyväskylä, Finland: Jyväskylä Studies in Education, Psychology and Social Research.

Mandler, G. (1982). Mind and emotion (2nd ed.). Malabar, FL: Krieger. Mandler, G. (1984). Mind and body: Psychology of emotion and stress. New York, NY:

Norton & Company. Marchewka, A., Wypych, M., Moslehi, A., Riegel, M., Michalowski, J. M., & Jednoróg, K.

(2016). Arousal rather than basic emotions influence long-term recognition memory in humans. Frontiers in Behavioral Neuroscience, 10. https://doi.org/10.3389/fnbeh.2016.00198

McCarthy, C., Pradhan, N., Redpath, C., & Adler, A. (2016). Validation of the Empatica E4 wristband. In IEEE EMBS International Student Conference. Ottawa, ON: IEEE. https://doi.org/10.1109/EMBSISC.2016.7508621

92

McQuiggan, S. W., Mott, B. W., & Lester, J. C. (2008). Modeling self-efficacy in intelligent tutoring systems: An inductive approach. User Modeling and User-Adapted Interaction, 18(1–2), 81–123. https://doi.org/10.1007/s11257-007-9040-y

Mikroyannidis, A., Okada, A., Scott, P., Rusman, E., Specht, M., Stefanov, K., … Chaimala, F. (2013). weSPOT: A personal and social approach to inquiry-based learning. Journal of Universal Computer Science, 19(14), 2093–2111. https://doi.org/10.3217/jucs-019-14-2093

Miyake, N. (1986). Constructive interaction and the iterative process of understanding. Cognitive Science, 10(2), 151–177. https://doi.org/10.1207/s15516709cog1002_2

Miyake, N., & Kirschner, P. A. (2014). The social and interactive dimensions of collaborative learning. In R. K. Sawyer (Ed.), The Cambridge handbook of the learning sciences (2nd ed., pp. 418–438). Cambridge, United Kingdom: Cambridge University Press. https://doi.org/10.1017/CBO9781139519526.026

Montague, E., Xu, J., & Chiou, E. (2014). Shared experiences of technology and trust: An experimental study of physiological compliance between active and passive users in technology-mediated collaborative encounters. IEEE Transactions on Human-Machine Systems, 44(5), 614–624. https://doi.org/10.1109/THMS.2014.2325859

Mullins, D., Rummel, N., & Spada, H. (2011). Are two heads always better than one? Differential effects of collaboration on students’ computer-supported learning in mathematics. International Journal of Computer-Supported Collaborative Learning, 6(3), 421–443. https://doi.org/10.1007/s11412-011-9122-z

Neumann, E., & Blanton, R. (1970). The early history of electrodermal research. Psychophysiology, 6(4), 453–475. https://doi.org/10.1111/j.1469-8986.1970.tb01755.x

Neumann, S. A., & Waldstein, S. R. (2001). Similar patterns of cardiovascular response during emotional activation as a function of affective valence and arousal and gender. Journal of Psychosomatic Research, 50(5), 245–253. https://doi.org/10.1016/S0022-3999(01)00198-2

Novak, D., Mihelj, M., & Munih, M. (2012). A survey of methods for data fusion and system adaptation using autonomic nervous system responses in physiological computing. Interacting with Computers, 24(3), 154–172. https://doi.org/10.1016/j.intcom.2012.04.003

Noy, L., Levit-Binun, N., & Golland, Y. (2015). Being in the zone: Physiological markers of togetherness in joint improvisation. Frontiers in Human Neuroscience, 9, 1–14. https://doi.org/10.3389/fnhum.2015.00187

OECD. (2017). PISA 2015 collaborative problem-solving framework. In PISA 2015 assessment and analytical framework: Science, reading, mathematic, financial literacy and collaborative problem solving (pp. 131–188). Paris, France: OECD Publishing. https://doi.org/http://dx.doi.org/10.1787/9789264281820-8-en

Palumbo, R. V., Marraccini, M. E., Weyandt, L. L., Wilder-Smith, O., McGee, H. A., Liu, S., & Goodwin, M. S. (2017). Interpersonal autonomic physiology: A systematic review of the literature. Personality and Social Psychology Review, 21(2), 99–141. https://doi.org/10.1177/1088868316628405

93

Panadero, E. (2017). A review of self-regulated learning: Six models and four directions for research. Frontiers in Psychology, 8, 1–28. https://doi.org/10.3389/fpsyg.2017.00422

Panadero, E., Klug, J., & Järvelä, S. (2015). Third wave of measurement in the self-regulated learning field: When measurement and intervention come hand in hand. Scandinavian Journal of Educational Research, 60(6), 1–13. https://doi.org/10.1080/00313831.2015.1066436

Pekrun, R., Goetz, T., Titz, W., & Perry, R. P. (2002). Academic emotions in students’ self-regulated learning and achievement: A program of qualitative and quantitative research. Educational Psychologist, 37(2), 91–105. https://doi.org/10.1207/S15326985EP3702

Pentland, A. (2004). Healthwear: Medical technology becomes wearable. Computer, 37(5), 42–49. https://doi.org/10.1109/MC.2004.1297238

Picard, R. W., Fedor, S., & Ayzenberg, Y. (2016). Multiple arousal theory and daily-life electrodermal activity asymmetry. Emotion Review, 8(1), 62–75. https://doi.org/10.1177/1754073914565517

Pintrich, P. R., Smith, D. A. F., Garcia, T., & McKeachie, W. J. (1991). A manual for the use of the Motivated Strategies for Learning Questionnaire (MSLQ). Technical Report No. 91-B-004. Ann Arbor, MI.

Pintrich, P. R., Smith, D. A. F., Garcia, T., & McKeachie, W. J. (1993). Reliability and predictive validity of the motivated strategies for learning questionnaire (MSLQ). Educational and Psychological Measurement, 53(3), 801–813. https://doi.org/10.1177/0013164493053003024

Poh, M. Z., Swenson, N. C., & Picard, R. W. (2010). A wearable sensor for unobtrusive, long-term assesment of electrodermal activity. IEEE Transactions on Biomedical Engineering, 57(5), 1243–1252. https://doi.org/10.1109/tbme.2009.2038487

Porges, S. W., Bohrer, R., Cheung, M., Drasgow, F., McCabe, P., & Keren, G. (1980). New time-series statistic for detecting rhythmic co-occurrence in the frequency domain: The weighted coherence and its application to psychophysiological research. Psychological Bulletin, 88(3), 580–587.

Raskin, D. C. (1973). Attention and arousal. In W. F. Prokasy & D. C. Raskin (Eds.), Electrodermal activity in psychological research (pp. 125–155). New York, NY: Academic Press.

Reeck, C., Ames, D. R., & Ochsner, K. N. (2016). The social regulation of emotion: An integrative, cross-disciplinary model. Trends in Cognitive Sciences, 20(1), 47–63. https://doi.org/10.1016/j.tics.2015.09.003

Reisenzein, R. (1983). The Schachter theory of emotion: Two decades later. Psychological Bulletin, 94(2), 239–264. https://doi.org/10.1037/0033-2909.94.2.239

Reiter-Palmon, R., Sinha, T., Gevers, J., Odobez, J.-M., & Volpe, G. (2017). Theories and models of teams and groups. Small Group Research, 48(5), 1–24. https://doi.org/10.1177/1046496417722841

Rizzolatti, G., & Arbib, M. A. (1998). Language within our grasp. Trends in Neurosciences, 21, 188–194. https://doi.org/10.1016/S0166-2236(98)01260-0

Rizzolatti, G., & Fabbri-Destro, M. (2008). The mirror system and its role in social cognition.

94

Current Opinion in Neurobiology, 18(2), 179–184. https://doi.org/10.1016/j.conb.2008.08.001

Robbins, T. W. (1997). Arousal systems and attentional processes. Biological Psychology, 45(1–3), 57–71. https://doi.org/10.1016/S0301-0511(96)05222-2

Roschelle, J., & Teasley, S. D. (1995). The construction of shared knowledge in collaborative problem solving. In C. O’Malley (Ed.), Computer supported collaborative learning (pp. 69–97). Berlin, Germany: Springer. https://doi.org/10.1007/978-3-642-85098-1_5

Roy, J.-C., Boucsein, W., Fowles, D. C., & Gruzelier, J. H. (Eds.). (1993). Progress in electrodermal research. New York, NY: Plenum Press.

Rummel, N. (2018). One framework to rule them all? Carrying forward the conversation started by Wise and Schwarz. International Journal of Computer-Supported Collaborative Learning, 13(1), 123–129. https://doi.org/10.1007/s11412-018-9273-2

Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161–1178. https://doi.org/10.1037/h0077714

Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes, and other things called emotion: Dissecting the elephant. Journal of Personality and Social Psychology, 76(5), 805–819. https://doi.org/10.1037/0022-3514.76.5.805

Sacks, H. (1972). An initial investigation of the usability of conversational data for doing sociology. In D. Sudnow (Ed.), Studies in social interaction (pp. 31–74). New York, NY: Free Press.

Sanbonmatsu, D. M., & Kardes, F. R. (1988). The effects of physiological arousal on information processing and persuasion. Journal of Consumer Research, 15(3), 379–385. https://doi.org/10.1086/209175

Sbarra, D. A., & Hazan, C. (2008). Coregulation, dysregulation, self-regulation: An integrative analysis and empirical agenda for understanding adult attachment, separation, loss, and recovery. Personality and Social Psychology Review, 12(2), 141–167. https://doi.org/10.1177/1088868308315702

Schachter, S., & Singer, J. (1962). Cognitive, social, and physiological determinants of emotional state. Psychological Review, 69(5), 379–399. https://doi.org/10.1037/h0046234

Schmitz, B., & Skinner, E. (1993). Perceived control, effort, and academic performance: Interindividual, intraindividual, and multivariate time-series analyses. Journal of Personality and Social Psychology, 64(6), 1010–1028. https://doi.org/10.1037//0022-3514.64.6.1010

Schneider, J., Börner, D., van Rosmalen, P., & Specht, M. (2015). Augmenting the senses: A review on sensor-based learning support. Sensors, 15(2), 4097–4133. https://doi.org/10.3390/s150204097

Schwartz, G. E., & Shapiro, D. (1973). Social psychophysiology. In W. F. Prokasy & D. C. Raskin (Eds.), Electrodermal activity in psychological research (pp. 377–416). New York, NY: Academic Press.

Schwartz, M. S., Collura, T. F., Kamiya, J., & Schwartz, N. M. (2017). The history and definitions of biofeedback and applied psychophysiology. In M. S. Schwartz & F.

95

Andrasik (Eds.), Biofeedback: A practitioner’s guide (4th ed., pp. 3–23). New York, NY: Guilford.

Schwendimann, B. A., Rodriguez-Triana, M., Vozniuk, A., Prieto, L., Boroujeni, M., Holzer, A., … Dillenbourg, P. (2017). Perceiving learning at a glance: A systematic literature review of learning dashboard research. IEEE Transactions on Learning Technologies, 10(1), 30–41. https://doi.org/10.1109/TLT.2016.2599522

Sharot, T., & Phelps, E. A. (2004). How arousal modulates memory: Disentangling the effects of attention and retention. Cognitive, Affective, & Behavioral Neuroscience, 4(3), 294–306. https://doi.org/10.3758/CABN.4.3.294

Shi, Y., Ruiz, N., Taib, R., Choi, E. H. C., & Chen, F. (2007). Galvanic skin response (GSR) as an index of cognitive load. In CHI ’07 Extended Abstracts on Human Factors in Computing Systems (pp. 2651–2656). New York, NY: ACM. https://doi.org/10.1145/1240866.1241057

Singer, N., Jacobi, N., Lin, T., Raz, G., Shpigelman, L., Gilam, G., … Hendler, T. (2016). Common modulation of limbic network activation underlies the unfolding of musical emotions and its temporal attributes. NeuroImage, 141(November), 517–529. https://doi.org/10.1016/j.neuroimage.2016.07.002

Slavin, R. E. (1980). Cooperative learning. Review of Educational Research, 50(2), 315–342. https://doi.org/10.3102/00346543050002315

Slovák, P., Tennent, P., Reeves, S., & Fitzpatrick, G. (2014). Exploring skin conductance synchronisation in everyday interactions. In Proceedings of the 8th Nordic Conference on Human-Computer Interaction (pp. 511–520). New York, NY: ACM. https://doi.org/10.1145/2639189.2639206

Spangler, G. (1997). Psychological and physiological responses during an exam and their relation to personality characteristics. Psychoneuroendocrinology, 22(6), 423–441. https://doi.org/10.1016/S0306-4530(97)00040-1

Spangler, G., Pekrun, R., Kramer, K., & Hofmann, H. (2002). Students’ emotions, physiological reactions, and coping in academic exams. Anxiety, Stress and Coping, 15(4), 413–432. https://doi.org/10.1080/1061580021000056555

Stahl, G., Koschmann, T., & Suthers, D. (2014). Computer-supported collaborative learning. In R. K. Sawyer (Ed.), The Cambridge handbook of the learning sciences (2nd ed., pp. 479–500). Cambridge, United Kingdom: Cambridge University Press. https://doi.org/10.1017/CBO9781139519526.029

Strauss, M., Reynolds, C., Hughes, S., Park, K., McDarby, G., & Picard, R. W. (2005). The HandWave Bluetooth skin conductance sensor. In International Conference on Affective Computing and Intelligent Interaction (pp. 699–706). Berlin, Germany: Springer. https://doi.org/10.1007/11573548_90

Swan, M. (2012). Sensor mania! The Internet of things, wearable computing, objective metrics, and the quantified self 2.0. Journal of Sensor and Actuator Networks, 1(3), 217–253. https://doi.org/10.3390/jsan1030217

Taylor, S., Jaques, N., Chen, W., Fedor, S., Sano, A., & Picard, R. W. (2015). Automatic identification of artifacts in electrodermal activity data. In 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) (pp.

96

1934–1937). Milan, Italy: IEEE. https://doi.org/10.1109/EMBC.2015.7318762 Thorson, K. R., West, T. V., & Mendes, W. B. (2018). Measuring physiological influence

in dyads: A guide to designing, implementing, and analyzing dyadic physiological studies. Psychological Methods, 23(4), 595–616. https://doi.org/10.1037/met0000166

Tomasello, M. (1995). Joint attention as social cognition. In C. Moore & P. J. Dunham (Eds.), Joint attention: Its origins and role in development (pp. 103–130). Hillsdale, NJ: Erlbaum.

Torniainen, J., Cowley, B., Henelius, A., Lukander, K., & Pakarinen, S. (2015). Feasibility of an electrodermal activity ring prototype as a research tool. In 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) (pp. 6433–6436). Milan, Italy: IEEE. https://doi.org/10.1109/EMBC.2015.7319865

van Bruggen, J. M., Kirschner, P. A., & Jochems, W. (2002). External representation of argumentation in CSCL and the management of cognitive load. Learning and Instruction, 12(1), 121–138. https://doi.org/10.1016/S0959-4752(01)00019-6

van Dooren, M., de Vries, J. J. G., & Janssen, J. H. (2012). Emotional sweating across the body: Comparing 16 different skin conductance measurement locations. Physiology and Behavior, 106(2), 298–304. https://doi.org/10.1016/j.physbeh.2012.01.020

Venables, P. H., & Christie, M. J. (1973). Mechanisms, instrumentation, recording techniques, and quantification of responses. In W. F. Prokasy & D. C. Raskin (Eds.), Electrodermal activity in psychological research (pp. 1–124). New York, NY: Academic Press.

Venables, P. H., & Christie, M. J. (1980). Electrodermal activity. Techniques in Psychophysiology, 54(3).

Venables, P. H., & Martin, I. (1967). Skin resistance and skin potential. In P. H. Venables & I. Martin (Eds.), A manual of psychophysiological methods. Amsterdam, Netherlands: North-Holland.

Vigouroux, R. (1879). Sur le rôle de la résistance électrique des tissues dans l’électro-diagnostic [On the role of electrical resistance of tissue in electro-diagnosis]. Comptes Rendus Société de Biologie, 31, 336–339.

Visnovcova, Z., Mestanik, M., Gala, M., Mestanikova, A., & Tonhajzerova, I. (2016). The complexity of electrodermal activity is altered in mental cognitive stressors. Computers in Biology and Medicine, 79(December), 123–129. https://doi.org/10.1016/j.compbiomed.2016.10.014

Vogel, F., & Weinberger, A. (2018). Quantifying qualities of collaborative learning processes. In F. Fischer, C. E. Hmelo-Silver, S. R. Goldman, & P. Reimann (Eds.), International handbook of the learning sciences (pp. 500–510). New York, NY: Routledge.

Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard University Press.

Wang, G. H. (1957). The galvanic skin reflex: A review of old and recent works from a physiologic point of view. Part I. American Journal of Physical Medicine, 36, 295–320.

97

Webb, N. M. (1991). Task-related verbal interaction and mathematics learning in small groups. Journal for Research in Mathematics Education, 22(5), 366–389. https://doi.org/10.2307/749186

Weinberger, A., Stegmann, K., & Fischer, F. (2007). Knowledge convergence in collaborative learning: Concepts and assessment. Learning and Instruction, 17(4), 416–426. https://doi.org/10.1016/j.learninstruc.2007.03.007

Wertsch, J. V. (1991). A sociocultural approach to socially shared cognition. In L. B. Resnick, J. M. Levine, & S. D. Teasley (Eds.), Perspectives on socially shared cognition (pp. 85–100). Washington, DC: American Psychological Association. https://doi.org/10.1037/10096-004

Winne, P. H. (2010). Improving measurements of self-regulated learning. Educational Psychologist, 45(4), 267–276. https://doi.org/10.1080/00461520.2010.517150

Winne, P. H., & Perry, N. E. (2000). Measuring self-regulated learning. In M. Boekaerts, P. R. Pintrich, & M. Zeidner (Eds.), Handbook of self-regulation (pp. 531–566). San Diego, CA: Academic Press. https://doi.org/10.1016/B978-012109890-2/50045-7

Wise, A. F., & Schwarz, B. B. (2017). Visions of CSCL: Eight provocations for the future of the field. International Journal of Computer-Supported Collaborative Learning, 12(4), 423–467. https://doi.org/10.1007/s11412-017-9267-5

Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength of stimulus to rapidity of habit-formation. Journal of Comparative Neurology and Psychology, 18(5), 459–482. https://doi.org/10.1037/h0073415

98

99

Original publications I Pijeira-Díaz, H. J., Drachsler, H., Järvelä, S., & Kirschner, P. A. (2016). Investigating

collaborative learning success with physiological coupling indices based on electrodermal activity. In Proceedings of the Sixth International Conference on Learning Analytics & Knowledge (pp. 64–73). New York, NY: ACM. https://doi.org/10.1145/2883851.2883897

II Pijeira-Díaz, H. J., Drachsler, H., Kirschner, P. A., & Järvelä, S. (2018). Profiling sympathetic arousal in a physics course: How active are students? Journal of Computer Assisted Learning, 34(4), 397–408. https://doi.org/10.1111/jcal.12271

III Pijeira-Díaz, H. J., Drachsler, H., Järvelä, S., & Kirschner, P. A. (2019). Sympathetic arousal commonalities and arousal contagion during collaborative learning: How attuned are triad members? Computers in Human Behavior, 92(March), 188–197. https://doi.org/10.1016/j.chb.2018.11.008

Copyright of Article I held by the author. Article II reprinted with permission from

John Wiley and Sons. Permission is not required from the publisher of Article III

(i.e., Elsevier), as the author retains the right to include the article in a thesis or

dissertation, provided it is not published commercially.

Original publications are not included in the electronic version of the dissertation.

100

A C T A U N I V E R S I T A T I S O U L U E N S I S

Book orders:Granum: Virtual book storehttp://granum.uta.fi/granum/

S E R I E S E S C I E N T I A E R E R U M S O C I A L I U M

168. Oikarinen, Juho (2016) Technology-enhanced statistics learning experiment : acase study at upper secondary level

169. Knihtilä, Aila (2016) Eetoksena hyvyys ja kauneus : taiteilija ja musiikkikasvattajaAnna-Liisa Antila-Kaljusen toimintaelämäkerta

170. Suša, Rene (2016) Social cartographies of internationalization of higher educationin Canada : a study of exceptionalist tendencies and articulations

171. Lassila, Erkki T. (2017) Tensions in the relationships : exploring Japanesebeginning teachers’ stories

172. Rönkä, Anna Reetta (2017) Experiences of loneliness from childhood to youngadulthood : study of the Northern Finland Birth Cohort 1986

173. Irjala, Marja (2017) Osallinen, syrjässä, marginaalissa, onnellinen? : tutkimusoppisopimuskoulutuksen erityisopiskelijoista Suomessa ja Saksassa

174. Kurki, Kristiina (2017) Young children’s emotion and behaviour regulation insocio-emotionally challenging situations

175. Lujala, Elise (2018) Pohjoisen varhaiskasvatuksen puolesta : Oulunlastentarhanopettajakoulutus vuosina 1967–1995

176. Mertaniemi, Seija (2018) Seitsemäsluokkalaisten kerrontaa kirjoittamisestaanopetusta eheyttävässä moniaineprojektissa

177. Pihkala, Suvi (2018) Touchable matters : reconfiguring sustainable change throughparticipatory design, education, and everyday engagement for non-violence

178. Mertala, Pekka (2018) Two worlds collide? : mapping the third space of ICTintegration in early childhood education

179. Juutinen, Jaana (2018) Inside or outside? : small stories about the politics ofbelonging in preschools

180. Välitalo, Riku (2018) The Philosophical Classroom : balancing educationalpurposes

181. Välimäki, Anna-Leena (2018) Kun työ tulee hymyillen kotiin : suomalaisenperhepäivähoidon 50 vuoden tarina 1960-luvulta 2010-luvulle

182. Tuomisto, Timo (2018) Kansanopistopedagogiikka kolmessa kristillisessäkansanopistossa

183. Uematsu-Ervasti, Kiyoko (2019) Global perspectives in teacher education : acomparative study of the perceptions of Finnish and Japanese student teachers

UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND

A C T A U N I V E R S I T A T I S O U L U E N S I S

University Lecturer Tuomo Glumoff

University Lecturer Santeri Palviainen

Senior research fellow Jari Juuti

Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Planning Director Pertti Tikkanen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho

Publications Editor Kirsti Nurkkala

ISBN 978-952-62-2218-9 (Paperback)ISBN 978-952-62-2219-6 (PDF)ISSN 0355-323X (Print)ISSN 1796-2242 (Online)

U N I V E R S I TAT I S O U L U E N S I SACTAE

SCIENTIAE RERUM SOCIALIUM

E 184

AC

TAH

éctor Javier Pijeira D

íaz

OULU 2019

E 184

Héctor Javier Pijeira Díaz

ELECTRODERMAL ACTIVITY AND SYMPATHETIC AROUSAL DURING COLLABORATIVE LEARNING

UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU,FACULTY OF EDUCATION