SOCIETY, ORGANIZATIONS

AND THE BRAIN: BUILDING

TOWARDS A UNIFIED

COGNITIVE NEUROSCIENCE

PERSPECTIVE

EDITED BY : Carl Senior, Nick Lee and Sven Braeutigam

PUBLISHED IN : Frontiers in Human Neuroscience

1 June 2015 | Society, Organizations and the BrainFrontiers in Human Neuroscience

Frontiers Copyright Statement

All content included on this site,

such as text, graphics, logos, button

icons, images, video/audio clips,

downloads, data compilations and

software, is the property of or is

licensed to Frontiers Media SA

(“Frontiers”) or its licensees and/or

subcontractors. The copyright in the

text of individual articles is the property

of their respective authors, subject to

a license granted to Frontiers.

The compilation of articles constituting

this e-book, wherever published,

as well as the compilation of all other

content on this site, is the exclusive

property of Frontiers. For the

conditions for downloading and

copying of e-books from Frontiers’

website, please see the Terms for

Website Use. If purchasing Frontiers

e-books from other websites

or sources, the conditions of the

website concerned apply.

Images and graphics not forming part

of user-contributed materials may

not be downloaded or copied

without permission.

Individual articles may be downloaded

and reproduced in accordance

with the principles of the CC-BY

licence subject to any copyright or

other notices. They may not be

re-sold as an e-book.

As author or other contributor you

grant a CC-BY licence to others to

reproduce your articles, including any

graphics and third-party materials

supplied by you, in accordance with

the Conditions for Website Use and

subject to any copyright notices which

you include in connection with your

articles and materials.

All copyright, and all rights therein,

are protected by national and

international copyright laws.

The above represents a summary

only. For the full conditions see the

Conditions for Authors and the

Conditions for Website Use.

ISSN 1664-8714

ISBN 978-2-88919-580-0

DOI 10.3389/978-2-88919-580-0

About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering

approach to the world of academia, radically improving the way scholarly research

is managed. The grand vision of Frontiers is a world where all people have an equal

opportunity to seek, share and generate knowledge. Frontiers provides immediate and

permanent online open access to all its publications, but this alone is not enough to

realize our grand goals.

Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online

journals, promising a paradigm shift from the current review, selection and dissemination

processes in academic publishing. All Frontiers journals are driven by researchers for

researchers; therefore, they constitute a service to the scholarly community. At the same

time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing

system, initially addressing speciﬁc communities of scholars, and gradually climbing up to

broader public understanding, thus serving the interests of the lay society, too.

Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative

interactions between authors and review editors, who include some of the world’s best

academicians. Research must be certiﬁed by peers before entering a stream of knowledge

that may eventually reach the public - and shape society; therefore, Frontiers only applies

the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding

research, evaluated with no bias from both the academic and social point of view.

By applying the most advanced information technologies, Frontiers is catapulting scholarly

publishing into a new generation.

What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series:

they are collections of at least ten articles, all centered on a particular subject. With their

unique mix of varied contributions from Original Research to Review Articles, Frontiers

Research Topics unify the most inﬂuential researchers, the latest key ﬁndings and historical

advances in a hot research area! Find out more on how to host your own Frontiers

Research Topic or contribute to one as an author by contacting the Frontiers Editorial

Ofﬁce: researchtopics@frontiersin.org

2 June 2015 | Society, Organizations and the BrainFrontiers in Human Neuroscience

This e-book brings together scholars in both the neurosciences and organizational sciences

who have adopted various approaches to study the cognitive mechanisms mediating the

social behavior that we see within organizations. Such an approach has been termed by

ourselves, and others, as ‘organisational cognitive neuroscience’. In recent years there has

been a veritable increase in studies that have explored the cognitive mechanisms driving such

behaviours, and much progress has been made in understanding the neural underpinnings of

processes such as ﬁnancial exchange, risk awareness and even leadership. However, while these

studies are informative and add to our understanding of human cognition they fall short of

providing evidence-based recommendations for practice. Speciﬁcally, we address the broader

issue of how the neuroscientiﬁc study of such core social behaviors can be used to improve

the very way that we work. To address these gaps in our understanding the chapters in this

book serve as a platform that allows scholars in both the neurosciences and the organizational

sciences to highlight the work that spans across these two ﬁelds.

The consolidation of these two ﬁelds also serves to highlight the utility of a uniﬁed and

singular organizational cognitive neuroscience. This is a fundamentally important outcome of

the book as the application of neuroscience to address economically relevant behaviours has

seen a variety of ﬁelds evolve in their own right, such as neuromarketing, neuroeconomics

and so forth. The use of neuro-scientiﬁc technologies,in particular fMRI, has indeed led

to a bewildering and somewhat suffocating proliferation of new approaches, however, the

speed of such developments demands that we must proceed carefully with such ventures or

risk some fundamental mistakes. The book that you now hold will consolidates these new

neuroscience based approaches and in doing so highlight the importance of this approach

in helping us to understand human social behaviour in general. Taken together the chapters

provide a framework for scholars within the neurosciences who wish to explore the further

the opportunities that the study of organisational behaviour may provide.

Citation: Senior, C., Lee, N., Braeutigam, S., eds. (2015). Society, Organizations and the Brain:

Building Towards a Uniﬁed Cognitive Neuroscience Perspective. Lausanne: Frontiers Media.

doi: 10.3389/978-2-88919-580-0

SOCIETY, ORGANIZATIONS AND THE BRAIN:

BUILDING TOWARDS A UNIFIED COGNITIVE

NEUROSCIENCE PERSPECTIVE

Topic Editors:

Carl Senior, Aston University, UK

Nick Lee, Loughbourgh University, UK

Sven Braeutigam, Oxford University, UK

3 June 2015 | Society, Organizations and the BrainFrontiers in Human Neuroscience

Table of Contents

05 Society, organizations and the brain: building toward a unified cognitive

neuroscience perspective

Carl Senior, Nick Lee and Sven Braeutigam

09 The relationship between self-report of depression and media usage

Martin Block, Daniel B. Stern, Kalyan Raman, Sang Lee, Jim Carey, Ashlee A.

Humphreys, Frank Mulhern, Bobby Calder, Don Schultz, Charles N. Rudick, Anne J.

Blood and Hans C. Breiter

19 Antagonistic neural networks underlying differentiated leadership roles

Richard E. Boyatzis, Kylie Rochford and Anthony I. Jack

34 Redefining neuromarketing as an integrated science of influence

Hans C. Breiter, Martin Block, Anne J. Blood, Bobby Calder, Laura Chamberlain, Nick

Lee, Sherri Livengood, Frank J. Mulhern, Kalyan Raman, Don Schultz, Daniel B.

Stern, Vijay Viswanathan and Fengqing (Zoe) Zhang

41 Age-related striatal BOLD changes without changes in behavioral loss aversion

Vijay Viswanathan, Sang Lee, Jodi M. Gilman, Byoung Woo Kim, Nick Lee, Laura

Chamberlain, Sherri L. Livengood, Kalyan Raman, Myung Joo Lee, Jake Kuster,

Daniel B. Stern, Bobby Calder, Frank J. Mulhern, Anne J. Blood and Hans C. Breiter

53 On the interpretation of synchronization in EEG hyperscanning studies: a

cautionary note

Adrian P. Burgess

70 Operationalizing interdisciplinary research—a model of co-production in

organizational cognitive neuroscience

Michael J. R. Butler

73 Dehumanization in organizational settings: some scientific and ethical

considerations

Kalina Christoff

78 Cognitive requirements of competing neuro-behavioral decision systems: some

implications of temporal horizon for managerial behavior in organizations

Gordon R. Foxall

95 The marketing firm and consumer choice: implications of bilateral contingency

for levels of analysis in organizational neuroscience

Gordon R. Foxall

109 Near-infrared spectroscopy (NIRS) as a new tool for neuroeconomic research

Isabella M. Kopton and Peter Kenning

122 Ideology in organizational cognitive neuroscience studies and other misleading

claims

Dirk Lindebaum

4 June 2015 | Society, Organizations and the BrainFrontiers in Human Neuroscience

125 The evolution of leader–follower reciprocity: the theory of service-for-prestige

Michael E. Price and Mark Van Vugt

142 Recommendations for sex/gender neuroimaging research: key principles and

implications for research design, analysis, and interpretation

Gina Rippon, Rebecca Jordan-Young, Anelis Kaiser and Cordelia Fine

155 Using evolutionary theory to enhance the brain imaging paradigm

Gad Saad and Gil Greengross

158 A sociogenomic perspective on neuroscience in organizational behavior

Seth M. Spain and P. D. Harms

173 A face for all seasons: searching for context-specific leadership traits and

discovering a general preference for perceived health

Brian R. Spisak, Nancy M. Blaker, Carmen E. Lefevre, Fhionna R. Moore and Kleis

F. B. Krebbers

182 A case for neuroscience in mathematics education

Ana Susac and Sven Braeutigam

185 The role of attachment styles in regulating the effects of dopamine on the

behavior of salespersons

Willem Verbeke, Richard P. Bagozzi and Wouter E. van den Berg

198 A comment on the service-for-prestige theory of leadership

Christopher R. von Rueden

200 Interdisciplinary research is the key

David A. Waldman

203 Consumer neuroscience to inform consumers—physiological methods to

identify attitude formation related to over-consumption and environmental

damage

Peter Walla, Monika Koller and Julia L. Meier

EDITORIAL

published: 19 May 2015

doi: 10.3389/fnhum.2015.00289

Frontiers in Human Neuroscience | www.frontiersin.org May 2015 | Volume 9 | Article 289

Edited and reviewed by:

Srikantan S. Nagarajan,

University of California, San Francisco,

USA

*Correspondence:

Carl Senior,

c.senior@aston.ac.uk;

Nick Lee,

n.lee@lboro.ac.uk;

Sven Braeutigam,

sven.braeutigam@psych.ox.ac.uk

Received: 25 March 2015

Accepted: 04 May 2015

Published: 19 May 2015

Citation:

Senior C, Lee N and Braeutigam S

(2015) Society, organizations and the

brain: building toward a uniﬁed

cognitive neuroscience perspective.

Front. Hum. Neurosci. 9:289.

doi: 10.3389/fnhum.2015.00289

Society, organizations and the brain:

building toward a uniﬁed cognitive

neuroscience perspective

Carl Senior

, Nick Lee

and Sven Braeutigam

School of Life & Health Sciences, Aston University, Birmingham, UK,

School of Business and Economics, Loughborough

University, Loughborough, UK,

Oxford Centre for Human Brain Activity, Oxford University, Oxford, UK

Keywords: organizational cognitive neuroscience, functional brain imaging, neuromarketing, neuroeconomics

The Oxford English Dictionary contains the following entry for the word “postal” as:

• adjective relating to or carried out by post.

– PHRASES go postal US informal go mad, especially from stress. With reference to cases in

which postal employees have run amok and shot colleagues.

Even a superﬁcial knowledge of recent events may lead to the conclusion that the contemporary

organization is perhaps not an easy thing to manage in a way that guarantees both economic

and social prosperity. As such, it seems to be part of the modern human condition to be at least

somewhat unhappy, stressed, or otherwise negatively impacted by either organizational life itself,

or the impact of organizations on today’s society. Fortunately, however, worst-case scenarios—as

implied by the OED above—are very rare.

It does not come as a surprise then, that researchers have expended considerable eﬀorts on

exploring and understanding the formation, management, and ethical sustentation of organizations

of all kinds and sizes, from bleeding-edge venture enterprises operating in break-ne ck markets to

perhaps non-competitive, non-proﬁt charities. Drawing from an interest in the negative eﬀects

workplaces can have on individuals, some of us published a clarion call, raising questions about

how a better understanding of our biological systems could inform an understanding of the social

behavior that we manifest within organizations (Butler and Senior, 2007a,b). The critical question

here is how the organization and the individual interact and inﬂuence each other, given that it that

organizations are designed as they are by t he very same species which will work in them, and equally

important how cognitive neuroscience in particular can help to unravel such mechanisms.

Scholars have indeed begun to explore the neuroscience of organizational behavior. These eﬀorts

go under the names of Organizational Neuroscience and Organizational Cognitive Neuroscience,

terms that refer to cross-disciplinary perspectives on organizational research, which take as their

foci of study the cognitive mechanisms that drive human behaviors in response to organizational

manifestations (Senior et al., 2008, 2011; Becker et al., 2011; Lee et al., 2012a). Such approaches seem

to have some merit in the study of the eﬀects of organizational life on human beings, and also on

how one can mitigate the more deleterious eﬀects that appear inherent to such contexts. However,

even with such rich empirical intercourse there remains an opportunity to examine further the

current state of the art research endeavors that span the biological and organizational domains to

inform our understanding of the type of social behavior that most of us will carry out most days for

most of our lives.

The articles contained within this research topic do just that, and go beyond merely explicating

further the possible mechanisms that drive “social behavior that occurs within organizational

manifestations” (Senior et al., 2011, p. 2) but ensure that such an understanding actually informs

Senior et al. Building toward a uniﬁed cognitive neuroscience

our knowledge of a socially rele vant and species speciﬁc social

behavior. In the call for papers we chose not to restrict the

nature of articles, but to ensure that all submissions could inform

our wider understanding social behavior in this applied context

(Waldman, 2013). The resulting submissions can be loosely

grouped into four main clusters–(a) general management, (b)

leadership, (c) neuromarketing science, and (d) papers that have

made speciﬁc recommendations for subsequent work.

To fully realize the potential for the impact of these articles,

it is important to ﬁrst reﬂect upon the industrial revolution and

how it showed that complex products could most proﬁtably be

made by breaking them up into small spe cialized, repetitive t asks.

As far back as the early 20th Century, with the emergence of

“scientiﬁc management” (e.g., Taylor, 1911), and the principles of

Fordism, the place of humans in this workﬂow was also treated

as a mechanistic process, to be designed in such a way as to

maximize eﬃciency and minimize defects. In such a context,

one could be forgiven for wondering whether working in such

organizations was what humans were ideally suited to. Even so, it

is undeniable that humans are the only species to have organized

itself into abstract organizations (i.e., not solely related to survival

or socialization), suggesting that perhaps something about this

ability does confer a collective advantage, if not an individual

one. In such a context, one would be forgiven for fearing that

the application of cognitive neuroscientiﬁc technology to helping

us understand more about our behaviors within the workplace

may drive the onset of what might become a neo-scientiﬁc

management; one that sees the data from workers as merely a

mechanism to maximize eﬃciency and minimize defects. Yet the

articles contained in this research topic show that this is far from

the case and, rather than driving biological reductionism, the

articles collectively demonstrate the signiﬁcant impact that such

approaches can bring to helping us understand human behavior.

In a novel approach to addressing a signiﬁcant question, Block

et al. (2014) carried out a large-scale interrogation of an existing

database on media behavior and found a signiﬁcant relationship

between media usage e.g., Internet, television and other social

media, and self-reports of depression. Christoﬀ (2014) takes

exploration of the relationship between organizational settings

and mental-health a stage further, and argues that a discourse

exploring the role emotions play in organizational decision-

making is needed. In light of the fact that in modern

organizations, so many of us place such heavy emphasis on such

media outlets when enacting our working roles, considering the

possible eﬀect that they may have on mental health ensures that

we consider the welfare of the individual workers of paramount

importance (see also Senior and Lee, 2013 for further discussion

here).

Taken together, the work by Spain and Harms (2014) and

Verbeke et al. (2014), converges on a greater understanding of

individual behavior within an organization at a genetic level.

This socio-economic approach is t hen examined further with the

submission by Foxall (2014a), who suggests a model for eﬀective

managerial behavior; that is, the function of competitive neural

systems. The notion of dual systems operating in competition to

drive eﬀective managerial behavior was examined further with

work by Boyatzis et al. (2014) who carried out an fMRI study

identifying ant

agonistic neural systems responsible for diﬀerent

types of leadership behaviors.

Such work continues to inform our understanding of how

social cognitive neuros cience (Ochsner and Lieberman, 2001) can

adv

ance organizational research—a project essentially started by

our e arlier work (e.g., Lee et al., 2012a). In particular, and possibly

as the result of serendipitous collaboration, neuroscientiﬁc

measuring tools such as functional magnetic resonance imaging

(fMRI) and magnetoencephalography (MEG) have been applied

to a number of organizational research questions (e.g., McClure

et al., 2004; Braeutigam, 2005; Deppe et al., 2005). Such

approaches have given the world terms as “neuroeconomics”

(Braeutigam, 2005) and “neuromarketing” (Breiter et al., 2015b),

and have inspired some considerable controversy in the scientiﬁc

press (e.g., Nature Neuroscience July 2004). Such debate is

healthy and as is shown by Butler (2014), Lindebaum (2014),

and Waldman (2013) helps to drive consolidation of theory and

clariﬁcation of approaches.

This, then, is the foundation of Organizational Cognitive

Neuroscience (OCN), which as an approach brings together

diversity in research approaches that use neuroscientiﬁc theories

and methods to examine organizational research issues (Senior

et al., 2011; Lee et al., 2012a). Indeed, the beneﬁts of an OCN

approach are exempliﬁed by Foxall (2014b),Walla et al. (2014)

and Breiter et al. (2015a), who each describe how the study of

exchanges within a market scenario can provide insig hts more

general human behavior, which in turn would lead to a more

“integrated science of inﬂuence” (Breiter et al., 2015b p. 1). These

scholars highlight both theoretical and methodological advances

within mainstream cognitive neurosciences and the implications

for a greater understanding of human behavior when market

exchanges are speciﬁcally investigated. Such methodological

advances are explored further with work by Kopton and Kenning

(2014) and Burgess (2013) who, among other things, develop

novel statistical approaches for the analysis of hyper scanning

data—which looks likely to be a crucial technique in exploring

the sort of interactions so central to organizational life.

That said, such work clearly shows that theoretical

advancement is not dependent on simply grafting advanced

measurement tools (such as fMRI) on to existing theories, as

implied by many early uses terms such as “neuromarketing”

(and here we recognize that Breiter et al. (2015b) clearly deﬁne a

more scientiﬁcally-rigorous useage of neuromarketing). Instead,

the OCN approach explicitly recognizes that it is the interaction

between cognitive neuroscience and organizational research as

distinct ﬁelds of research which is critical—incorporating not

just new methods, but also new theoretical explanations. In such

a way, the ﬁeld can lead to advances in both its parent disciplines

(Lee et al., 2012b).

We have previously conceptualized OCN as an approach that

considers human behavior made in response to organizational

manifestations (e.g., products, advertisements) as a set of

theoretical layers, e ach building upon the last to add more

context-speciﬁc theory (Lee et al., 2012b). At the most abstract

level, the behavior of individuals and groups at the intersection

between t he organization and the human is considered. Yet

such behavior is a subset of human social behavior in general.

Frontiers in Human Neuroscience | www.frontiersin.org 2 May 2015 | Volume 9 | Article 289

Senior et al. Building toward a uniﬁed cognitive neuroscience

Therefore it is an additional layer of theory that can be added

upon social psychology. In turn, social psychology is founded

on theories of cognitive psychology, which also impact directly

on many of our responses to organizational manifestations such

as advertisements and products. At an even more basic level,

are the lower-level brain systems and structures that drive such

cognitions, analysis here could be termed the neural level of

analysis. To facilitate investigations across the various layers

of analysis that are diagnostic of the organizational cognitive

neuroscience approach, Rippon et al. (2014) provide a set of

recommendations that could be adopted when studying the

eﬀects of gender on particular task.

The organizational, social, and neural levels that are described

above have been the focus of existing OCN theory (e.g., Lee

and Chamberlain, 2007). Yet, at a more fundamental level one

can also describe the adaptive forces t hat have shaped our brain

physiology in an evolutionary advantageous manner (Saad and

Greengross, 2014). Knowledge of the evolutionary adaptations

that may mediate our behavior at the social and ultimately

organizational level is essential to complete the explanation

of why we behave in the way we do, and also critical in

understanding the potential negative (and positive) inﬂuence of

organizational life on human beings.

To move back to the example of “scientiﬁc management”

previously alluded to; an understanding of whether the ability

to focus on repetitive small tasks may have conferred an

evolutionary advantage in the past (which therefore would have

led to a predilection for this ability in humans) may then

lead to greater understanding of whether scientiﬁc management

principles are likely to be beneﬁcial to employees. Importantly,

this is quite apart from the logical principles of the approach,

which may indeed suggest that it may be the most eﬃcient

manner with which to produce a complex product with

minimum defe cts. Indeed, the key social processes (wit hin

organizations) that humans have a predilection toward are

discussed subsequently.

Such an idea has been developed further with t he work

by Saad and Greengross (2014), who go so far as to say

that an understanding of evolutionary theory is of paramount

importance when using cognitive neuroscientiﬁc technology to

explore organizationally-relevant behaviors. However, it is with

the work by Spisak et al. (2014) and Price and Van Vugt (2014)

where the importance of studying adaptive behaviors and the role

that they may play in facilitating eﬀective organizational is made

crystal clear (See also von Rueden, 2014). Developing this further,

Susac and Braeutigam (2014) describe how an understanding

of the neural substrates underpinning mathematical cognition

may in fact facilitate the ability for mathematical re asoning—

which itself has implic ations for the subsequent design of eﬀective

education.

Here it is clear t hat it is not possible to fully understand a

given organizationally-relevant behavior by ignoring the various

interweaved layers of theory introduced above, Focusing on the

neural level—without taking into account the more fundamental

evolutionary level, or e ven the more abstract organizational

and social levels—is likely to result in important explanatory

contextual factors being overlooked. OCN explicitly recognizes

the symbiotic relationship between the layers of theory and in

doing so develops more rigorous testable hypotheses, and ties this

to advances in research methods that can more accurately test

these hypotheses. The studies noted above develop existing OCN

theory (e.g., Butler and Senior, 2007a) to show in more depth the

evolutionary processes that may impact on our organizationally-

relevant actions. The focus here is on how the neural and

evolutionary levels interact, and the question of whether such

adaptations actually can inﬂuence our behavior within, and our

response to, organizations and their manifestations.

As noted above, organizations that are designed around

the social processes that humans have a predilection for are

likely to operate more eﬃciently. Yet we should not consider

the application of neuroscience to understanding organizational

behavior as a means merely to make such organizations more

eﬃcient. In spite of the working environment being constantly

in ﬂux, the central concept of organizational behavior has and

will always remain the same. Most of us are likely to spend a

major proportion of our lives in a work-related environment.

One may argue thus that organizational cognitive neuroscience

is an approach by which to understand the cognitive signature of

our own species-speciﬁc social behavior.

We would like to dedicate this research topic to the many

reviewers who considered the submitted papers in such a timely

fashion–without them this collection would not have happened.

References

Becker, W. J., Cropanzano, R., and Sanfey, A. G. (2011). Organizational

neuroscience: taking organizational theory inside the neural black box.

J. Manag. 37, 933–961. doi: 10.1177/0149 2 0 6 3 1 1 3 9 8 9 5 5

Block, M., Stern, D. B., Raman, K., Lee, S., Carey, J., Humphreys, A. A., et al.

(2014). The relationship between self-report of depression and media usage.

Front. Hum. Neurosci. 8:712. doi: 10.3 3 8 9 /fnhum.2 0 1 4 .0 0 7 1 2

Boyatzis, R. E., Rochford, K., and Jack, A. I. (2014). Antagonistic neural networks

underlying diﬀerentiated leadership roles. Front. Hum. Neurosci. 8:114. doi:

10.3389/fnhum.2014.00114

Braeutigam, S. (2005). Neuroeconomics – from neural systems to economic

behaviour. Brain Res. Bull. 67, 355–360.doi: 10.10 1 6 /j .brainresbull.2005.

06.009

Breiter, H. C., Block, M., Blood, A. J., Calder, B., Chamberlain, L., Lee, N., et al.

(2015b). R edeﬁning neuromarketing as an integrated science of inﬂuence.

Front. Hum. Neurosci. 8:1073. doi: 10 .3 3 8 9 /fnhum.2 0 1 4 .0 1 0 7 3

Breiter, H. C., Viswanathan, V., Lee, S., Gilman, J. M., Kim, B. W., Lee, N., et al.

(2015a). Age-related striatal BOLD changes without changes in behavioral loss

aversion. Front. Hum. Neurosci. 9:176. doi: 10.3389 /fnhum.2 0 1 5 .0 0 1 7 6

Burgess, A. P. (2013). On the interpretation of synchronization in EEG

hyperscanning studies: a cautionary note. Front. Hum. Neurosci. 7:881. doi:

10.3389/fnhum.2013.00881

Butler, M. J. R. (2014). Operationalizing interdisciplinary research—a model of

co-production in organizational cognitive neuroscience. Front. Hum. Neurosci.

7:720. doi: 10.3389/fnhum.2013.00720

Butler, M. J. R., and Senior, C. (2007a). Towards an organizational cognitive

neuroscience. Ann. N.Y. Acad. Sci. 111 8 , 1–17. doi: 10.1196/annals.1412.009

Frontiers in Human Neuroscience | www.frontiersin.org 3 May 2015 | Volume 9 | Article 289

Senior et al. Building toward a uniﬁed cognitive neuroscience

Butler, M. J. R., and Senior, C. (2007b). Research possibilities for organizational

cognitive neuroscience. Ann. N.Y. Acad. Sci. 1118, 206–210. doi:

10.1196/annals.1412.010

Christoﬀ, K. (2014). Dehumanization in organizational settings: some

scientiﬁc and ethical considerations. Front. Hum. Neurosci. 8:748. doi:

10.3389/fnhum.2014.00748

Deppe, M., Schwindt, W., Kugel, H., Plassman, H., and Kenning, P. (2 0 0 5 ).

Non-linear responses within the medial prefrontal cortex reveal when speciﬁc

implicit information inﬂuences economic decision making. J. Neuroimaging 15,

171–182. doi: 10.1177/1051228405275074

Foxall, G. R. (2014a). Cognitive requirements of competing neuro-

behavioral decision systems: some implications of temporal horizon for

managerial behavior in organizations. Front. Hum. Neurosci. 8:184. doi:

10.3389/fnhum.2014.00184

Foxall, G. R. (2014b). The marketing ﬁrm and consumer choice: implications

of bilateral contingency for levels of analysis in organizational neuroscience.

Front. Hum. Neurosci. 8:472. doi: 10.3 3 8 9 /fnhum.2 0 1 4 .0 0 4 7 2

Kopton, I. M., and Kenning, P. (2014). Near-infrared spectroscopy (NIRS) as

a new tool for neuroeconomic research. Front. Hum. Neurosci. 8:549. doi:

10.3389/fnhum.2014.00549

Lee, N., Senior, C., and Butler, M. J. R. (2012a). The domain of organisational

cognitive neuroscience: Theoretical and empirical challenges. J. Manag. 38,

921–931. doi: 10.1177/0149206312439471

Lee, N., Senior, C., and Butler, M. J. R. (2012b). Leadership research and

cognitive neuroscience: the state of this union. Leadersh. Q. 23, 213–218. doi:

10.1016/j.leaqua.2011.08.001

Lee, N., and Chamberlain, L. (2007). Neuroimaging and psychophysiological

measurement in organizational research: an agenda for organizational cognitive

neuroscience. Ann. N.Y. Acad. Sci. 111 8 , 18–43.doi: 10.1196/annals.1412.003

Lindebaum, D. (2014). Ideology in organizational cognitive neuroscience

studies and other misleading claims. Front. Hum. Neurosci. 7:834. doi:

10.3389/fnhum.2013.00834

McClure, S. M., Li, J., Tomlin, D., Cypert, K. S., Montague, L. M., and Montague,

P. R. (2004). Neural correlates of behavioral preference for culturally familiar

drinks. Neuron44, 379–387. doi: 10.1016/j.neuron.2004.09.019

Ochsner, K. N., and Lieberman, M. D. (2001). The emergence of social cognitive

neuroscience. Am. Psychol. 56, 717–734. doi: 10.1037/0 0 0 3 - 0 6 6 X.5 6 .9 .7 1 7

Price, M. E., and Van Vugt, M. (2014). The evolution of leader–follower

reciprocity: the theory of service-for-prestige. Front. Hum. Neurosci. 8:363. doi:

10.3389/fnhum.2014.00363

Rippon, G., Jordan-Young, R., Kaiser, A., and Fine, C. (2014). Recommendations

for sex/gender neuroimaging research: key principles and implications for

research design, analysis, and interpretation. Front. Hum. Neurosci. 8:650. doi:

10.3389/fnhum.2014.00650

Saad, G., and Greengross, G. (2014). Using evolutionary theory to

enhance the brain imaging paradigm. Front. Hum. Neurosci. 8:452. doi:

10.3389/fnhum.2014.00452

Senior, C., and Lee, N. (2013). The state of the art in organisational cognitive

neuroscience: the therapeutic gap and possible implications for clinical practice.

Front. Hum. Neurosci. 7:808. doi: 10.3 3 8 9 /fnhum.2 0 1 3 .0 0 8 0 8

Senior, C., Butler, M. J. R., and Lee, N. (2008). Fear and loathing in the work place.

Am. J. Bioeth. 8, 20–21. doi: 10.1080/152651608021 8 0 0 8 3

Senior, C., Lee, N., and Butler, M. J. R. (2011). Organizational cognitive

neuroscience. Org. Sci. 22, 804–815. doi: 10.1287/orsc.1100.0532

Spain, S. M., and Harms, P. D. (201 4 ). A sociogenomic perspective on

neuroscience in organizational behavior. Front. Hum. Neurosci. 8:84. doi:

10.3389/fnhum.2014.00084

Spisak, B. R., Blaker, N. M., Lefevre, C. E., Moore, F. R., and Krebbers, K. F. B.

(2014). A face for all seasons: searching for context-speciﬁc leadership traits and

discovering a general preference for perceived health. Front. Hum. Neurosci.

8:792. doi: 10.3389/fnhum.2014.00792

Susac, A., and Braeutigam, S. (2014). A case for neuroscience in mathematics

education. Front. Hum. Neurosci. 8:314. doi: 10.3389/fnhum.2014.00314

Taylor, F. W. (1911). The Principles of Scientiﬁc Management. New York, NY:

Harper & Row.

Verbeke, W., Bagozzi, R . P., and van den Berg, W. E. (2014). The role of

attachment styles in regulating the eﬀects of dopamine on the behavior

of salespersons. Front. Hum. Neurosci. 8:32. doi: 10 .3 3 8 9 /fnhum.2 0 1 4 .

00032

von Rueden, C. R. (2014). A comment on the service-for-prestige theory of

leadership. Front. Hum. Neurosci. 8:412. doi: 10.3389/fnhum.2014.00412

Waldman, D. A. (20 1 3 ). Interdisciplinary research is the key. Front. Hum.

Neurosci. 7:562. doi: 10.3389/fnhum.2013 .0 0 5 6 2

Walla, P., Koller, M., and Meier, J. L. (2014). Consumer neuroscience to inform

consumers—physiological methods to identify attitude formation related to

over-consumption and environmental damage. Front. Hum. Neurosci. 8:304.

doi: 10.3389/fnhum.2014.003 0 4

Conﬂict of Interest Statement: The authors declare that the research was

conducted in the absence of any commercial or ﬁnancial relationships that could

be construed as a potential conﬂict of interest.

distributed under the terms of the Creative Commons Attribution License (CC BY).

The use, distribution or reproduction in other forums is permitted, provided the

original author(s) or licensor are credited and that the original publication in this

journal is cited, in accordance with accepted academic practice. No use, distribution

or reproduction is permitted which does not comply with these terms.

Frontiers in Human Neuroscience | www.frontiersin.org 4 May 2015 | Volume 9 | Article 289

ORIGINAL RESEARCH ARTICLE

published: 12 September 2014

doi: 10.3389/fnhum.2014.00712

The relationship between self-report of depression and

media usage

Martin Block

1,2

†

,DanielB.Stern

2,3†

, Kalyan Raman

1,2‡

, Sang Lee

2,4‡

,JimCarey

1,2‡

Ashlee A. Humphreys

1,2‡

, Frank Mulhern

1,2‡

, Bobby Calder

2,5‡

, Don Schultz

1,2‡

, Charles N. Rudick

2,6†

Anne J. Blood

2,4,7†

and Hans C. Breiter

2,3,4,7†

Medill Integrated Marketing Communications, Northwestern University, Evanston, IL, USA

Applied Neuromarketing Consortium, Medill, Kellogg, and Feinberg Schools, Northwestern University, Evanston, IL, USA

Department of Psychiatry and Behavioral Science, Warren Wright Adolescent Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA

Laboratory of Neuroimaging and Genetics, Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA

Department of Marketing, Kellogg School of Management, Northwestern University, Evanston, IL, USA

Department of Urology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA

Mood and Motor Control Laboratory, Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA

Edited by:

Sven Braeutigam, University of

Oxford, UK

Reviewed by:

Christian Lambert, St. George’s

University of London, UK

Jessica Clare Scaife, Oxford

Univeristy, UK

*Correspondence:

Martin Block, Northwestern

University, Medill Integrated

Marketing Communications, MTC

3-123, 1845 Sheridan Road,

Evanston, IL 60208, USA

e-mail: mp-block@northwestern.edu

†,‡

Authors made equal contribu-

tions, corresponding to First (

†)or

Second (

‡) authorship.

Depression is a debilitating condition that adversely affects many aspects of a person’s life

and general health. Earlier work has supported the idea that there may be a relationship

between the use of certain media and depression. In this study, we tested if self-report of

depression (SRD), which is not a clinically based diagnosis, was associated with increased

internet, television, and social media usage by using data collected in the Media Behavior

and Inﬂuence Study (MBIS) database (N = 19,776 subjects). We further assessed the

relationship of demographic variables to this association. These analyses found that SRD

rates were in the range of published rates of clinically diagnosed major depression. It

found that those who tended to use more media also tended to be more depressed,

and that segmentation of SRD subjects was weighted toward internet and television

usage, which was not the case with non-SRD subjects, who were segmented along

social media use. This study found that those who have suffered either economic or

physical life setbacks are orders of magnitude more likely to be depressed, even without

disproportionately high levels of media use. However, among those that have suffered

major life setbacks, high media users—particularly television watchers—were even more

likely to report experiencing depression, which suggests that these effects were not just

due to individuals having more time for media consumption. These ﬁndings provide an

example of how Big Data can be used for medical and mental health research, helping

to elucidate issues not traditionally tested in the ﬁelds of psychiatry or experimental

psychology.

Keywords: depression, big data, marketing communications, media use

INTRODUCTION

Depression is known to affect many kinds of human behavior,

and is quite common. As of 2005, the lifetime prevalence of

major depressive disorder in the US population was reported to be

16.5% (Kessler et al., 2005a), with 6.7% prevalence in a 12-month

period, 30.4% of which were severe (or 2.0% of the U.S. popula-

tion) (Kessler et al., 2005b). Given the prevalence of depression,

there is interest from a neuromarketing perspective in how it

may be related to patterns of media consumption. Such issues

are of fundamental concern for mechanisms of behavior change

research and psychology (e.g., Morgenstern et al., 2013).

There is a developing literature e valuating the relationship

between v arious types of media use and psychiatric conditions.

For instance, one study found a high positive correlation between

internet addiction and depression among university students

(Orsal et al., 2012). Another study found that adults with major

depressive disorder spent excessive amounts of leisure time on

the computer, while those with dysthymia, panic disorder, and

agoraphobia spent more time watching television than the con-

trol group or those with other disorders (de Wit et al., 2011).

However, results have not always been consistent, particularly in

thedomainofsocialmediause.Arecentpaperfailedtoﬁnd

any association between social network use and depression in

older adolescents (Jelenchick et al., 2013), while other studies have

found positive associations between Facebook use and depression

in high school students (Pantic et al., 2012), and Facebook use

and a lack of subjective well-being in young adults (Kross et al.,

2013). Given the heterogeneity across previous studies, and the

rapid evolution of media formats over the past decade, we used a

large consumer database (>19,000 subjects) to assess the relation-

ship between self-reported depression (SRD) and media usage,

taking into account demographic information which may impact

the incidence of SRD such as employment status and disability.

We used SRD since major depression cannot be diagnosed with

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

HUMAN NEUROSCIENCE

Block et al. The relationship between self-report of depression and media usage

big data surveys, and compared the rate of SRD to published

incidence data on the diagnosis of major depression.

This study differed from previous studies in the following

ways. (1) The sample size of the dataset was substantially larger

than any previous study e v aluating the relationship between

media use and depression. (2) We evaluated the link between

depression and multiple domains of media use, whereas most

previous studies have focused primarily on single domains. For

example, recent work with a smaller database has suggested there

is an increase in digital media usage in “depressed” adolescents

(Primack et al., 2009), but this study did not investigate its rela-

tionship to different subcomponents of media, such as social

media, internet, and television.

Our analysis started with descriptive and bivariate statisti-

cal analyses. These were followed by omnibus approaches to

assess general effects given the number of variables describing

media usage: (a) Chi-squared Automatic Interaction Detection or

CHAID tree analysis (Kass, 1980; Biggs et al., 1991)(aformof

recursive partitioning; Zhang and Singer, 1999) and (b) discrim-

inant analysis.

MATERIALS AND METHODS

DATA ACQUISITION

The dataset was derived from the Media Behavior and Inﬂuence

Study (MBIS), a syndicated online study of American adult (i.e.,

>18 years of age) consumers, conducted twice yearly since 2002

by BIGinsight of Columbus, Ohio. The current wave of 19,776

participants was completed in December, 2012. Using a dou-

ble opt-in methodology, each MBIS study was balanced to meet

demographic criteria established by the US census. MBIS data

has been used by a variety of well-known, commercial market-

ing organizations. Variables of interest included depression by

gender, age, employment status, marital status, race and ethnic-

ity, income, education, measures of isolation, and internet, TV

and social media use. These variables were selected because they

have been variables of interest in previous depression studies,

andhavebeenshowntohavepredictivevalue(e.g.,Catalano and

Dooley, 1977; Wilkowska-Chmielewska et al., 2013). Media usage

for internet, television, and social media are based on yes/no

responses to several day-parts of variable hour durations for

a typical weekday (see Supplementary Materials). These blocks

of time were shorter for typical waking hours and longer for

overnight and weekend periods. Block length was used to weight

media usage probability dur ing the calculation of total hours of

consumption (i.e., divided by the number of hours in each block

of time). Average hour exposure probabilities were calculated for

a 24 h period, and minutes per day were estimated by multi-

plying the result by 1440. Internet, TV, and social media usage

were hence composite variables created as probabilities of num-

ber of minutes daily usage, derived from data indicating w hether

or not subjects used the respective services in seven discrete

variable-hour blocks.

DATA ANALYSIS

Three types of analysis were performed with this data. First,

we performed a descriptive statistical analysis, inclusive of cor-

relations between depression and media consumption variables

to facilitate interpretation of the subsequent analyses. Second,

weusedtheresultstoinformatypeofrecursivepartitioning

(Morgan and Sonquist, 1963; Friedman, 1977; Breiman et al.,

1984; Gordon and Olshen, 1984; Quinlan, 1986; Mingers, 1989),

namely CHAID tree analysis (Kass, 1980; Biggs et al., 1991).

Third, we performed a multivariate discriminate analysis. Given

the descriptive statistical analyses were standard, these are not fur-

ther discussed herein. In all analyses below excepting the CHAID

analysis, fewer than 50 total comparisons were made; to correct

for multiple comparisons we used a Bonferroni correction for 50

comparisons, requiring a p < 0.001 to be considered a signiﬁcant

result.

CHAID tree analysis

We performed two recursive partitioning analyses, one focused

on SRD and the second on a variable not of interest, namely non-

SRD, to act as a control for SRD results. Our working hypothesis

was that the control analysis of non-SRD subjects would not repli-

cate or provide an opponent (i.e., completely non-overlapping)

set of nodes to the analysis of SRD subjects.

Construction of statistical CHAID trees (SPSS tree) evaluated

the interaction among a number of predictor variables of SRD,

and separately non-SRD. Typically, such schemes are deﬁned in

terms of demographic variables such as age and gender; however

we have also included occupation, education, marital status and

media use. Splitting criteria included minimum parent node size

of 100 and child node size of 50, and a p-value threshold of 0.05.

These splitting criteria were used for both CHAID analyses.

Discriminant analyses

Discriminant analysis was used to conduct a multivariate analy-

sis of variance for the hy pothesis that people who self-reported

having depression would differ signiﬁcantly from non-SRD sub-

jects on a linear combination of eleven variables: income, internet

usage, TV usage, social media usage, education, age, living in top

10 metropolitan area (MSA), gender, having children, employ-

ment status, and disability. The discriminant analysis was run

using SPSS defaults, resulting in the canonical linear discriminant

analysis. Depression was the binary dependent variable entered

in the “group” dialog. The discriminating variables were entered

together (i.e., not stepwise) in the variables subcommand. The

discriminating variables income, internet usage, TV usage, social

media usage, and education, all took on continuous values in the

range from 0 to 1. “Living in top 10 MSA,” gender, employment

status and disability were binary categorical variables while hav-

ing children was ordinal. Overall, the data were complete with no

missing values (i.e., every subject had every data point).

RESULTS

DESCRIPTIVE AND CORRELATION ANALYSES

Geographic and temporal patterns

The MBIS study shows little to no geographic pattern for SRD

(Figure 1).

The data does show that SRD among all adults in the USA (18

and over) has grown from 11.2% in 2009 to 12.1% in December,

2012, with a linear trend (r

= 0.246) (Figure 2). It is interest-

ing that the r ate rises to 15.2% in June 2012 (similar to rates in

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

FIGURE 1 | Rates of Self-Reported Depression by State, December

2012. This infographic characterizes rates of self-reported depression by

state, with darker states showing greater rates of depression. The image

demonstrates few patterns in depression by geography, with perhaps the

exception that state with large metropolitan areas tend to show somewhat

less depression.

the 2005 MBIS data where the depressive rate was reported to be

14.9%), then drops to 12.1% in December 2012, which is con-

sistent with a previous study that used the emotive content of

tweets to show a similar annual pattern of decreased depression

over Christmas (Dodds et al., 2011).

Depression demographics

Gender. RatesofSRDinthecurrentstudywavewerenearlyiden-

tical by gender as shown in Tab le 1, with males slightly lower at

11.8%, compared to females at 12.3%.

Age and marital status. Bivariate analysis suggested an inverse

linear association of SRD with age, which is consistent with pre-

viously reported studies (Henderson et al., 1998). Individuals who

were married were also different than those who were unmar-

ried as shown in Ta ble 1, with married respondents representing

a large portion of the sample (42.5%), and reporting a lower

SRD rate of 9.5%. The highest rate of SRD was from those in

same sex unions, at 22.2%. Those living with an unmarried part-

ner, divorced or separated, or single (never married) reported

rates between 14.1% and 15.5%, while those that were widowed

reported rates (12.4%) nearly the same as the overall average.

Race and ethnicity. Tab le 1 showed lower rates among Hispanics

(10.9%), and lower yet among African-Americans (8.7%) and

Asians (7.9%) as compared to Multi-ethnic individuals and

Caucasians. SRD was highest among Caucasians (13.6%), who

represented more than half (58.4%) of the sample studied.

Income and education. Both income and education (Ta b le 1),

also demonstrated a strong inverse linear association with SRD,

similar to age (statistics not provided given omnibus analyses to

follow). Non-high school graduates self-reported a 21.7% depres-

sion rate compared to those with post college study or degree at

8.8%. The overall average income was $62,800, with those report-

ing depression indicating an average of $49,000. Occupation

levels showed similar effects, as shown in Tab le 1, with those dis-

abled (unable to work) reporting a 42.7% depression rate. Other

high reporting categories included the unemployed at 18.8%,

and students at 13.0%. The lowest category was professional and

management at 8.2%.

Health and lifestyle characteristics. SRD was also related to

the reporting of other health conditions as shown in Tabl e 1 .

Generally, those reporting depression were likely to say they

had other health-related conditions, such as anxiety (54.8%).

Other conditions more prevalent in SRC subjects included:

back pain (42.7%), overweight (37.6%), acid reﬂux (30.5%),

headaches/migraines (29.6%), insomnia/difﬁculty sleeping

(27.5%) (Tabl e 1 ).

Isolation. Residents of states with large urban areas and those

living in the top 10 metropolitan statistical areas (MSAs), have

lower rates of SRD. The top 10 MSAs include Los Angeles, New

York, Chicago, San Francisco, Philadelphia, Washington, Boston,

Detroit, Phoenix and Houston. This suggests that residents of

rural areas tend to report higher rates of depression.

Media use. Overall there were low but signiﬁcant positive lin-

ear correlations between SRD and media consumption. In these

descriptive analyses, the three most consumed media were televi-

sion, on average 129 min per day per adult (18+), the internet,

on average 143 min per day, and social media, on average 83 min

perday.Thebivariateassociation(r)betweenSRDandtelevision

consumption was 0.089, surﬁng the internet was 0.089 and social

media was 0.063 (all p < 0.001).

Media usage quintiles, a method commonly used in the media

industry, were created using the composite media usage variables

described above, and showed higher rates of depression among

the most active users of media. Figure 3 shows that for the h igh-

est 20% of telev ision users (quintile 5, 289 min per day) the SRD

rate was 16.9%. The SRD rate among the highest internet users

(327 min per day) was also 16.9%. SRD was slightly lower among

the highest social media users (279 min per day) at 15.4%. The

patterns among all three media categories were the same: higher

consumption of any form of media was associated with higher

rates of reported depression.

It should be noted that there was some co-linearity between the

three media categories. The correlation of television and internet

consumption was moderate at 0.495, slightly higher for inter-

net and social media at 0.510, but lower for television and social

media at 0.247. All of these correlations were signiﬁcant (p <

0.001), raising the possibility of simultaneous consumption.

CHAID TREE ANALYSIS

The analyses reported above were limited to bivariate correla-

tions. To better understand how multiple variables for media

consumption and other demographics/activities related to SRD,

a multivariate segmentation scheme was employed based on

recursive partitioning (Morgan and Sonquist, 1963; Friedman,

1977; Breiman et al., 1984; Gordon and Olshen, 1984; Quinlan,

1986; Mingers, 1989). The ﬁrst CHAID tree (Kass, 1980; Biggs

et al., 1991)(Figure 4) shows the interaction among the predictor

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

FIGURE 2 | Trends in MBIS Rates of Self-Reported Depression in

Adults (18+), June 2009–December 2012. This chart reports the rate

of self-reported depression every 6 months, beginning in June 2009

and ending in December 2012, ﬁtted with a linear trend line. All data

was collected the same way by BIGinsight of Ohio as part of the

MBIS study.

variables on the rate of SRD (the target variable). The second

CHAID tree (Figure 5) shows the interaction among the predic-

tor variables and those who did not self-report being depressed.

The ﬁrst analysis on depressed individuals generated 22 terminal

nodes, while the second on non-depressed subjects generated 21.

The trees (Figures 4, 5) were pruned to include only 8 and 10

terminal nodes where the depression rate was 15% or more and

the non-depression rate was 87% or higher, respectively. The tree

nodes showed the variable used to create the node, the depression

rate, and the percent of all adults that the node represented. In

Figure 4, those that were unemployed, for example, were 6.0% of

the sample and reported a depression rate of 18.8%. Note that

media-related nodes were shown in white and other variables

shown in blue/gray.

In the analysis of SRD subjects, the CHAID tree segments

(Figure 4) that were the basis for understanding the relation-

ship of depression to media and other variables were as follows.

In general, factors such as disability, unemployment and lower

incomes were associated with higher rates of SRD. Media con-

sumption tended to signiﬁcantly leverage the rate attributable to

these characteristics. Six nodes of interest are brieﬂy described.

The node with the highest depression rate (47.3%) was being dis-

abled(1)andinthetoptwoTVconsumptionquintiles.Thiswas

compared to being disabled and in the bottom three TV quin-

tiles w ith a somewhat lower depression rate (35.2%). The next

highest depression node (30.7%) consisted of (2) those who were

unemployed, in the top internet quintile, and had less than a col-

lege education. Those in a professional or managerial occupation

that made $30,000 or less, and were in the highest TV quintile

(3), reported a depression rate of 26.9%. (4) Female students or

homemakers older than 34, reported a 20.0% depression rate.

(5) Those in other occupations, including workers, sales, military

and retired, that make less than $42,500 and were in the high-

est social media quintile, reported a depression rate of 19.1%.

(6) Male students or homemakers in the highest two internet

quintiles reported a depression rate of 17.4%.

In the analysis of non-depressed individuals (non-SRD), the

CHAID tree segments (Figure 5)thatbestexplainedtherelation-

ships between media use, demographic variables, and non-SRD,

described ten nodes. The node with the highest non-depression

rate (96.2%) was being professional (1) with a salary of $150,000

and more in the lowest social media quintiles. This was com-

pared to being professional (2) with a salary of $150,000 and

more in the highest social media quintile (89.1%). The next

highest non-depression node (3) consisted of those in other occu-

pations, including workers, sales, military and retired, making

$100,000 and more (94.3%). This was contrasted with (4) those in

other occupations making $50,000 to $100,000 and in the lowest

social media quintiles (91.4%), and (5) those in other occupa-

tions making $50,000 to $100,000 and in the highest social media

quintiles (87.7%). Those who were professional making $75,000

to $100,000 and in the lowest social media quintile had a non-

depression rate of 94.2% (6), whereas those who were professional

making $75,000 to $100,000 and in the highest social media quin-

tile had a non-depression rate of 89.1% (7). Professionals making

$35,000 to $75,000 and male gender (8) were higher (91.9%) than

those professionals making $35,000 to $75,000 and female gender

(9).Lastly,studentswhoweremaleinthelowestweb-radioquin-

tile (10) had a non-depression rate of 90.8%. In general, being a

student or employed with a high income were most closely associ-

ated with not being depressed, particularly when combined with

varying levels of social media use.

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

Table 1 | Depression by demographics, December 2012.

% All adults Adults with Index

depression

Age (average years)* 45.443.896.3

Male 48.311.897.5

Female 51.712.3 101.7

Income (000)* 62.849.078.0

Have children 29.130.2103.9

Live in top 10 MSA 24.710.082.6

Married 42.59.578.5

Living with unmarried partner 7.215.5128.1

Divorced or separated 10.215.4128.3

Widowed 3.012.4102.5

Single, never married 25.714.1116.5

Same sex union 0.522.2183.5

Have not graduated high

school

1.521.7179.3

Graduated high school 16.813.1108.3

Technical school or vocational

training

5.71

8114.0

1–3 years of college (did not

graduate)

20.215.1124.8

Associates or professional

degree

8.913.2109.1

Bachelor’s degree 22.59.477.7

Post college study or degree 13.58.872.7

Business Owner 4.211.796.7

Professional/managerial 25.58.267.8

Salesperson 3.611.595.0

Factory worker/laborer/driver 3.39.679.3

Clerical or service worker 9.511.998.3

Homemaker 3.614.7121.5

Student, high school or

college

8.413.0107.4

Military 0.711.695.9

Retired 13.710.88

Unemployed 5.518.8155.4

Disabled (unable to work) 2.042.7 352.9

Obsessive-compulsive

disorder (OCD)

2.19.8 458.1

Anxiety 12.954.8 425.8

Dyslexia 0.82.6 334.5

Fibromyalgia 2.37.4 327.5

Insomnia/difﬁculty sleeping 8.427.5 325.8

Restless leg syndrome(RLS) 4.111.6 279.8

Irritable Bowel Syndrome

(IBS)/crohn’s disease

2.46.3 262.6

Chronic bronchitis/COPD 2.97.3 252.5

Sleep apnea 6.616.4 248.2

Heartburn/indigestion 9.922.7 230.3

Headaches/migraines 14.029.6211.5

Back pain 21.542.71

8.4

Acid reﬂux 15.830.5192.9

Heart disease 3.25.9185.9

Hearing impairment 4.38.0184.6

(Continued)

Table 1 | Continued

% All adults Adults with Index

depression

Overweight 21.237.6177.1

Arthritis 15.527.3176.4

Asthma 9.717.0174.8

Vision impairment 15.024.8165.2

Enlarged prostate/Benign

Prostatic Hyperplasia (BPH)

2.23.6162.6

Diabetes 9.315.0162.1

Osteoporosis 2.54.1161.6

High cholesterol 18.829.3155.9

Black 18.08.771.9

Asian 3.07.965.3

Multi 0.816.9139.7

Native 0.415.5128.1

White 58.413.6112.4

Other 0.59.981.8

Hispanic 18.91

990.1

The three columns of numeric values represent (i) the percentage of all adults

in the survey with the given attribute (excepting income and age, which lists

the survey mean), (ii) the percentage of subjects having the given attribute that

self-reported depression, and (iii) the index of the given attribute as it relates to

depression, which is calculated as the percentage of those depressed, divided

by the total number depressed multiplied by 100. Rows on the left of the table

are clustered around demographics, relationship status, education, and work

identiﬁcation. Rows on the right are clustered by illnesses separate from depres-

sion, and with race/ethnicity (Note: These do not follow NIH deﬁnitions of race

and ethnicity).

It is important to note that the CHAID analysis with non-

SRD did not replicate the analysis with SRD. Furthermore, there

was a segmentation observed between these analyses which was

distinct, in that the types of media use that segmented the SRD

subjects was not the same as that which segmented the non-SRD

subjects. The terminal nodes of the two analyses were different

along dimensions of occupation, income, and media use.

DISCRIMINANT ANALYSIS

The results of the discriminant analysis revealed that, other

than disability and income, the three single best predictors of

depression in this model were increased use of television, the

internet, and social media (Tabl e 2 ). The overall Chi-squared

test of the discriminant model was signiﬁcant (Wilk’s λ = 0.945,

Chi-square = 922.117, df = 9, Canonical correlation = 0.235,

p < 0.001). The st ructure matrix demonstrated the weights of the

discriminating variables, an indication of their importance for

predicting depression: being disabled (0.760), income (−0.519),

internet consumption (0.399), television consumption (0.368),

social media use (0.278), level of education ( −0.255), being

unemployed (0.223), age (−0.170), top 10 MSA (−0.142), female

gender (0.062), and having children (0.010).

DISCUSSION

The primary ﬁnding of this study is that those who tend to

use more media in general, also tend to have more self-reports

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

FIGURE 3 | MBIS Rates of Depression by Media Use Quintiles,

December 2012. This chart demonstrates the percent of subjects with

depression in each media quintile. Quintiles were determined by ordering

subjects based on estimated minutes of a given media consumed; the ﬁrst

1/5 used the least of a given media and comprised the 0–20% quintile, the

second ﬁfth used more than the ﬁrst 1/5 (but less than the third 1/5) and

comprised the 21–40% quintile, and so on. Quintiles were computed for

each type of media use of interest and graphed side by side. The graph

depicts a clear trend associating increased media usage with increased

rates of depression.

of depression. We found a current incidence of SRD at 12.1%

which is slightly less than reports of lifetime clinical depression

and more than the 12 month incidence of diagnoses of major

depression. However, the picture is far more nuanced than sim-

ple description of descriptive statistics and bivariate correlations

between media use and depression. For instance, the CHAID tree

analysis with SRD subjects (along with the discriminant anal-

ysis) shows that those who have suffered either economic or

physical life setbacks are orders of magnitude more likely to be

depressed, even without disproportionately high levels of media

use (37.2%). However, among those that have suffered major life

setbacks, hig h media users—particularly television watchers—

were even more likely to report experiencing depression (47.3%

in the highest two quintiles, as compared with 35.2% in the

lower three quintiles), which suggests that these effects were not

just due to individuals having more time for media consump-

tion. These effects were not observed with the control analysis

in non-SRD subjects. That the economically disadvantaged are

signiﬁcantly more likely to experience depression is well sup-

ported by research in social psychology, which suggests that

lower-income groups feel a sense of disempowerment (Henry,

2005; Stephens et al., 2007). The lack of ﬁnancial and temporal

resources they experience can lead to feelings of a lack of control

over one’s life and an inability to act efﬁcaciously in the world,

which is thought to be a basic human need. Support ing this inter-

pretation, the CHAID analysis of the non-SRD subjects showed

that high earners that use less social media tend to be signiﬁcantly

less depressed.

Life challenges may not be the only experiences related to

depression. As noted with our descriptive statistical analysis, per-

sistent environmental factors such as isolation can also contribute

to the prevalence of a psychological experience. Generally, isola-

tion is a known correlate of depression symptomology, and our

data suggest that residents of rural areas tend to report higher

rates of depression. Within the context of isolation, one can dis-

tinguish between physical and non-physical isolation; and within

non-physical isolation one can look at social and emotional iso-

lation. These various subclasses of isolation ﬁnd ample support

in the literature. Weiss (1973) ﬁrst distinguished the two types of

non-physical isolation—social and emotional—which have sub-

sequently been empirically shown as distinct (DiTommaso and

Spinner, 1997). Although conceptually distinct, the various typ es

of isolation interact. Physical isolation has been shown to affect

social and emotional isolation, especially for the elderly (Dugan

and Kivett, 1994) and adolescents (Brage et al., 1993). We mea-

sured social isolation through the proxy of living alone and

physical isolation through the proxy of place of residence, ﬁnd-

ing that both correlate to rates of self-reported depression. For

instance, those living in more populated cities (top 10 MSAs) tend

to report lower rates of depression.

In addition to the current state of depression, the data we

analyzed reveals that SRD has been in a state of ﬂux over the

past decade. At the beginning of this time frame, the rates we

observed were low compared to 2005 MBIS data where the

depressive r ate was reported to be 14.9%; a ﬁgure consistent with

a co-occurring 2005 study wherein a lifetime prevalence rate of

16.5% for major depressive disorder was reported (Kessler et al.,

2005a,b). Interestingly, the 2005 MBIS data and the Kessler et al.

(2005a,b) data show remarkable concordance despite differences

in inclusion criteria (exclusion of non-English speakers in the

Kessler et al., 2005a,b studies), the use of a structured clinical

interview vs. self-report data, and overall subject demographics.

This ﬂux in reported incidence of depression over the past decade

is further supported by the MBIS data showing self-reported

depression has been on the rise in adults (18+) over the last 4

years in the United States.

It is worth considering the demographics of individuals (e.g.,

gender) reporting SRD in the context of a ﬂux in depression rates

over time. As recently as 5 years ago, females were more likely to

report being depressed (i.e., SRD). However, in the most recent

MBIS study, the data shows SRD to be similarly associated with

both genders, with males reporting only a slightly lower rate of

depression. This is different than the rates reported by Primack

et al. (2009) where females were shown to be signiﬁcantly more

likely to be depressed, as was also observed in the December 2005

MBIS data. In comparison to prior big data reports, there appears

to be a narrowing in the gap of reported depression in females and

males, which could potentially reﬂect a change in the likelihood of

genders to self-report. One factor that has remained constant was

that depression is inversely related to age, with those younger than

24 reporting the highest rate, and older mar ried persons reporting

the lowest.

There are several important limitations to this study that are

worth mentioning. First, the data used was self-reported depres-

sion, which does not necessarily reﬂect whether the subject has

ever received a clinical diagnosis of depression. The subjective

phenotypes of those who have a clinical diagnosis of major

depression versus those that self-report depression could skew

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

FIGURE 4 | Pruned CHAID Tree Characterizing SRD, December 2012. The

pruned CHAID tree shows groups of subjects wherein rates of depression

were greater than 15%. These nodes represent only a subset of all nodes

generated by the CHAID tree. Of particular interest for this paper are nodes

that are white (instead of blue); these nodes have been highlighted because

they are partially deﬁned by the presence of a media use quintile.

the data in a number of different ways. For instance, it has been

observed that those who have been diagnosed with depression are

sometimes reticent to share their diagnosis. Alternatively, there is

a multiplicity of reasons to think that subjects without depression

may report being depressed. The balance of these considerations

leaves uncertainty in the true sample parameters, although the

percentage of subjects with SRD in this study was quite similar

to rates of depression found in previous studies.

Second, the variables computed for amount of television,

internet, and social media use are not direct measures. These

variables are composite variables computed from self-reports

of whether or not subjects used those various media during

discrete variable-hour-length blocks. This can introduce inter-

subject variability along a number of dimensions. For instance,

some subjects may report “yes” for one of the intervals based

on an hour’s worth of use, while others may respond the same

based on several hours’ worth of use. The probabilities computed

represent just that, a probability of time spent using a given media

relative to other subjects.

Third, the analyses done cannot speak to a causal relation-

ship between media consumption and depression, or to any

directionality between the observed associations. We think the

likeliest explanation is that these two variables form a com-

plex bi-directional relationship with autocatalytic properties. An

alternative explanation is that depression and increased media

use are a byproduct of a third confounding factor. It should also

be noted that the direction of causality between depression and

media use could also vary across individuals (i.e., whether media

usage helps to ameliorate depression or whether it contributes

to it). Whatever the exact relationship between depression and

increased media use, it is clear that the two are closely associated.

Fourth, it is important to acknowledge the potential con-

founds of concurrent medical illness on assessing associations

with SRD. In the literature on major depression, hypotheses

have been raised that depression in association with a med-

ical illness does not necessarily reﬂect the same structural

and functional circuitry alterations seen in depression with

strong familial heritability (e.g., see Cloninger, 2002; Breiter

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

FIGURE 5 | Pruned CHAID Tree Characterizing Non-SRD, December 2012.

The pruned CHAID tree shows groups of subjects wherein rates of

non-depression were greater than 87%. These nodes represent only a subset

of all nodes generated by the CHAID tree. Of particular interest for this paper

are nodes that are white (instead of blue); these nodes have been highlighted

because they are partially deﬁned by the presence of a media use quintile.

and Gasic, 2004; Breiter et al., 2006). There is a strong pos-

sibility of biological subtypes in depression (e.g., see Blood

et al., 2010), meaning depression comorbid with other ill-

nesses may reﬂect a directionality with media that is distinct

from other putative depressive subtypes. Depression in associ-

ation with another medical (e.g., se vere coronary artery dis-

ease) or psychiatric condition (e.g., OCD, generalized anxiet y,

or body image disorders) may have a complex directional rela-

tionship with these other conditions, and there is published

evidence that TV viewing itself is associated with anxiety and

body image issues (e.g., see Thompson and Heinberg, 2002;

de Wit et al., 2011), potentially leading to the self-reported

depression. These issues also relate to the potential for drug

and alcohol to confound effects with SRD; this data set did

not contain such information, so future work is needed to

assess the relationship of drug and alcohol effects on SRD and

media use.

This information can help to form hypotheses to test in future

studies of relevance to psychology. One such hypothesis could

relate to the directionality of the relationship between SRD and

media, to determine if any media use acts as feedback to exacer-

bate symptoms. Another hypothesis might attempt to relate the

relationship to existing social psychological constructs such as the

“empty self ” hypothesis. Cushman (1990) developed the “empty

self” hypothesis to describe those who feel depressed and may be

likely to engage in impulsive or excessive consumption behavior

in order to “ﬁll up” a perceived deﬁciency in the self (see also

Ahuvia, 2005). In the context of media usage, the “empty self”

might be expected to show increased consumption of media asso-

ciated with increased SRD; such behavior might be indicative of

a subtype of major depression. A third possibility, is a hypothesis

that increased media use by SRD subjects acts as indicator of the

illness, much like a biomarker. Such hypotheses point to further

opportunities for use of big data with psychological research.

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

Table 2 | Structure matrix of discriminant analysis predicting

depression, December 2012.

Canonical correlation 0.235 Chi-square 922.117

Wilk’s lambda 0.945 signiﬁcance <0.001

Disabled 0.760

Income −0.519

Internet usage 0.399

TV usage 0.368

Social media usage 0.278

Education −0.255

Unemployed 0.223

Age −0.170

Living in top 10 MSA −0.142

Female 0.062

Having children 0.010

This table reports the structure matrix of the discriminant analysis. The nine pre-

dictive variables used in the discriminant analysis are reported in the left column,

while a measure of that variable’s predictive importance in the discriminant func-

tion is listed on the right. A variable’s importance is determined by its magnitude,

while its relationship to depression is determined by its valence. Negative num-

bers describe an inverse relationship: for example, higher income is predictive

of less depression.

In summary, the data reveal that there is a consistent pat-

tern of results that link self-reported depression with increased

media use, even when taking into account other variables, such

as disability and unemployment. This media use was focused

more on internet use and TV exposure, for those making self-

reports of depression. The rate of SRD was between two standard

indices used in published reports of clinically diagnosed major

depression, namely the lifetime prevalence, and recent 12 month

incidence of major depression. These observations suggest the

current ﬁndings with big data may have relevance to the literature

focused on the clinical diagnosis of depression.

SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found

online at: http://www.frontiersin.org/HumanNeuroscience/10.

3389/fnhum.2014.00712/abstract

REFERENCES

Ahuvia, A. C. (2005). Beyond the extended self: loved objects and consumers’

identity narratives. J. Consum. Res. 32, 171–184. doi: 10.1086/429607

Biggs, D., De Ville, B., and Suen, E. (1991). A method of choosing multi-way

partitions for classiﬁcation and decision trees. J. Appl. Stat. 18, 49–62. doi:

10.1080/02664769100000005

Blood, A. J., Iosifescu, D. V., Makris, N., Perlis, R. H., Kennedy, D. N., Dougherty,

D. D., et al. (2010). Phenotype genotype project on addiction and mood

disorders. Microstructural abnormalities in subcortical reward circuitry of

subjects with major depressive disorder. PLoS ONE 5:e13945. doi: 10.1371/jour-

nal.pone.0013945

Brage, D., Meredith, W., and Woodward, J. (1993). Correlates of loneliness among

midwestern adolescents. Adolescence 28, 685–693.

Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. (1984). Classiﬁcation

and Regression Trees. The Wadsworth statistics/probability series. New York, NY:

Chapman and Hall/CRC.

Breiter, H. C., and Gasic, G. P. A. (2004). “General circuitry of reward/aversion and

its implications for neuropsychiatric illness,” in The Cognitive Neurosciences,ed

M. Gazzaniga (Cambridge, MA: MIT Press), 1043–1065.

Breiter, H. C., Gasic, G. P., and Makris, N. (2006). “Imaging the neural systems

for motivated behavior and their dysfunction in neuropsychiatric illness,” in

Complex Systems Science in Biomedicine, eds T. S. Deisboeck, and J. Y. Kresh

(New York, NY: Springer Verlag), 763–810. doi: 10.1007/978-0-387-33532-2_33

Catalano, R., and Dooley, C. D. (1977). Economic predictors of depressed mood

and stressful life events in a metropolitan community. J. Health Soc. Behav. 18,

292–307.

Cloninger, C. R. (2002). The discovery of susceptibility genes for mental disorders.

Proc. Natl. Acad. Sci. U.S.A. 99, 13365–13367. doi: 10.1073/pnas.222532599

Cushman, P. (1990). Why the self is empty. Toward a historically situated psychol-

ogy.AmPsychol.45, 599–611. doi: 10.1037/0003-066X.45.5.599

de Wit, L., van Straten, A., Lamers, F., Cuijpers, P., and Penninx, B. (2011).

Are sedentary television watching and computer use behaviors associated

with anxiety and depressive disorders? Psychiatry Res.186, 239–243. doi:

10.1016/j.psychres.2010.07.003

DiTommaso, E., and Spinner, B. (1997). Social and emotional loneliness: a reex-

amination of Weiss’ typology of loneliness. Pers. Individ. Dif. 22, 417–427. doi:

10.1016/S0191-8869(96)00204-8

Dodds, P. S., Harris, K. D., Kloumann, I. M., Bliss, C. A., and Danforth, C. M.

(2011). Temporal patterns of happiness and information in a global social

network: hedonometrics and twitter. PLoS ONE 6:e26752. doi: 10.1371/jour-

nal.pone.0026752

Dugan, E., and Kivett, V. R. (1994). The importance of emotional and social isola-

tion to loneliness among very old rural adults. Gerontologist 34, 340–346. doi:

10.1093/geront/34.3.340

Friedman, J. H. (1977). A recursive partitioning decision rule for nonparametric

classiﬁcation. IEEE Trans. Comput. 26, 404–407. doi: 10.1109/TC.1977.1674849

Gordon, L., and Olshen, R. A. (1984). Almost surely consistent nonparamet-

ric regression from recursive partitioning schemes. J. Multivariate Anal.15,

147–163. doi: 10.1016/0047-259X(84)90022-8

Henderson, A. S., Jorm, A. F., Korten, A. E., Jacomb, P., Christensen, H., and

Rodgers, B. (1998). Symptoms of depression and anxiety during adult life:

evidence for a decline in prevalence with age. Psychol. Med. 28, 1321–1328.

Henry, P. C. (2005). Social class, market situation, and consumers’ metaphors of

(dis)empowerment. J. Consum. Res. 31, 766–778. doi: 10.1086/426610

Jelenchick, L. A., Eickhoff, J. C., and Moreno, M. A. (2013). “Facebook depression?”

social networking site use and depression in older adolescents. J. Adolesc. Health.

52, 128–130. doi: 10.1016/j.jadohealth.2012.05.008

Kass, G. V. (1980). An exploratory technique for investigating large quantities of

categorical data. Appl. Stat. 29, 119–127. doi: 10.2307/2986296

Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., and Walters,

E. E. (2005a). Lifetime prevalence and age-of-onset distributions of DSM-IV

disorders in the National Comorbidity Survey Replication. Arch.Gen.Psychiatry

62, 593–602. doi: 10.1001/archpsyc.62.6.593

Kessler, R. C., Chiu, W. T., Demler, O., Merikangas, K. R., and Walters, E. E.

(2005b). Prevalence, severity, and comorbidity of 12-month DSM-IV disor-

ders

in the National Comorbidity Survey Replication. Arch. Gen. Psychiatry 62,

617–627. doi: 10.1001/archpsyc.62.6.617

Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D. S., Lin, N., et al. (2013).

Facebook use predicts declines in subjective well-being in young adults. PLoS

ONE 8:e69841. doi: 10.1371/journal.pone.0069841

Mingers, J. (1989). An empirical comparison of pruning methods for decisiontree

induction. Mach. Learn. 4, 227–243. doi: 10.1023/A:1022604100933

Morgan, J. N., and Sonquist, J. A. (1963). Problems in the analysis of

survey data and a proposal. J. Am. Stat. Assoc. 58, 415–434. doi:

10.1080/01621459.1963.10500855

Morgenstern, J., Naqvi, N. H., Debellis, R., and Breiter, H. C. (2013). The contri-

butions of cognitive neuroscience and neuroimaging to understanding mecha-

nisms of behavior change in addiction. Psychol. Addict. Behav. 27, 336–350. doi:

10.1037/a0032435

Orsal, O., Orsal, O., Unsal, A., and Ozlap, S. (2012). Evaluation of internet addiction

and depression among university students. Procedia Soc. Behav. Sci. 82, 445–454.

doi: 10.1016/j.sbspro.2013.06.291

Pantic, I., Damjanovic, A., Todorovic, J., Topalovic, D., Bojovic-Jovic, D., Ristic,

S., et al. (2012). Association between online social networking and depression

in high school students: behavioral physiology viewpoint. Psychiatr. Danub. 24,

90–93.

Primack, B. A., Swanier, B., Georgiopoulos, A. M., Land, S. R., and Fine, M.

J. (2009). Association between media use in adolescence and depression in

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

Block et al. The relationship between self-report of depression and media usage

young adulthood: a longitudinal study. Arch. Gen. Psychiatry 66, 181–188. doi:

10.1001/archgenpsychiatry.2008.532

Quinlan, J. R. (1986). Induction of decision trees. Mach. Learn. 1, 81–106. doi:

10.1007/BF00116251

Stephens, N. M., Markus, H. R., and Townsend, S. S. (2007). Choice as an act

of meaning: the case of social class. J.Pers.Soc.Psychol.93, 814–830. doi:

10.1037/0022-3514.93.5.814

Thompson, J. K. and Heinberg, L. J. (2002). The media’s inﬂuence on body image

disturbance and eating disorders: we’ve reviled them, now can we rehabilitate

them? J. Soc. Issues 55, 339–353. doi: 10.1111/0022-4537.00119

Weiss, E. B. (1973). Marketing to the New Society. Chicago, IL: Crain

Communications.

Wilkowska-Chmielewska, J., Szelenberger, W., and Wojnar, M. (2013).

Age-dependent symptomatology of depression in hospitalized patients

and its implications for DSM-5. J. Affect. Disord. 150, 142–145. doi:

10.1016/j.jad.2012.12.012

Zhang, H. P., and Singer, B. (1999). Recursive Partitioning in the Health Sciences.

New York, NY: Springer. doi: 10.1007/978-1-4757-3027-2

Conﬂict of Interest Statement: The authors declare that the research was con-

ducted in the absence of any commercial or ﬁnancial relationships that could be

construed as a potential conﬂict of interest.

Received: 17 April 2014; accepted: 26 August 2014; published online: 12 September

2014.

Citation: Block M, Stern DB, Raman K, Lee S, Carey J, Humphreys AA, Mulhern F,

Calder B, Schultz D, Rudick CN, Blood AJ and Breiter HC (2014) The relationship

between self-report of depression and media usage. Front. Hum. Neurosci. 8:712. doi:

10.3389/fnhum.2014.00712

This article was submitted to the journal Frontiers in Human Neuroscience.

Schultz, Rudick, Blood and Breiter. This is an open-access ar ticle distributed under the

terms of the Creative Commons Attribution License (CC BY). The use, distribution or

reproduction in other forums is permitted, provided the original author(s) or licensor

are credited and that the original publication in this journal is cited, in accordance with

accepted academic practice. No use, distribution or reproduction is per mitted which

does not comply with these terms.

Frontiers in Human Neuroscience www.frontiersin.org September 2014 | Volume 8 | Article 712

REVIEW ARTICLE

published: 04 March 2014

doi: 10.3389/fnhum.2014.00114

Antagonistic neural networks underlying differentiated

leadership roles

Richard E. Boyatzis

1,2

*, Kylie Rochford

and Anthony I. Jack

Department of Cognitive Science, Case Western Reserve University, Cleveland, OH, USA

Department of Organizational Behavior, Case Western Reserve University, Cleveland, OH, USA

Edited by:

Carl Senior, Aston University, UK

Reviewed by:

Carl Senior, Aston University, UK

Sven Braeutigam, Oxford Centre for

Human Brain Activity, University of

Oxford, UK

*Correspondence:

Richard E. Boyatzis, Department of

Organizational Behavior, Case

Western Reserve University, 10900

Euclid Avenue, Cleveland, OH 44106,

USA

e-mail: richard.boyatzis@case.edu

The emergence of two distinct leadership roles, the task leader and the socio-emotional

leader, has been documented in the leadership literature since the 1950s. Recent research

in neuroscience suggests that the division between task-oriented and socio-emotional-

oriented roles derives from a fundamental feature of our neurobiology: an antagonistic

relationship between two large-scale cortical networks – the t ask-positive network (TPN)

and the default mode network (DMN). Neural activity in TPN tends to inhibit activity in

the DMN, and vice versa. The TPN is important for problem solving, focusing of attention,

making decisions, and control of action. The DMN plays a central role in emotional self-

awareness, social cognition, and ethical decision making. It is also strongly linked to

creativity and openness to new ideas. Because activation of the TPN tends to suppress

activity in the DMN, an over-emphasis on task-oriented leadership may prove deleterious to

social and emotional aspects of leadership. Similarly, an overemphasis on the DMN would

result in difﬁculty focusing attention, making decisions, and solving known problems. In

this paper, we will review major streams of theory and research on leadership roles in the

context of recent ﬁndings from neuroscience and psychology. We conclude by suggesting

that emerging research challenges the assumption that role differentiation is both natural

and necessary, in particular when openness to new ideas, people, emotions, and ethical

concerns are important to success.

Keywords: leadership, roles, neural networks and behavior, DMN,TPN, anti-correlated networks, opposing domains

hypothesis

INTRODUCTION

The emergence of two distinct leadership roles, the task leader and

the socio-emotional leader, has been documented in the leadership

literature since the 1950s (Bales, 1958). The separation of these

roles has been seen as pragmatic in applications and is accepted

as a ﬁnding in behavioral research. However, in practical applica-

tions, it leaves an organization with a challenge that can detract

from leadership development, succession planning, and organi-

zational ﬂexibility. In this article, we will show that the division

between these two types of leadership roles lies far deeper than has

traditionally been thought. Recent research in neuroscience sug-

gests that the division between task-oriented and socio-emotional

leadership roles derives from a fundamental feature of our neu-

robiology: an antagonistic relationship between two large-scale

cortical networks that is present in every individual.

Neural activity in the task-positive network (TPN) tends to

inhibit activity in the default mode network (DMN; Raichle et al.,

2001; Fransson, 2005; Raichle and Snyder, 2007; Buckner et al.,

2008), and vice versa (Uddin et al., 2009; Jack et al., 2012). The

TPN is activated during a broad range of non-social tasks (Fox

et al., 2005; Buckner et al., 2008; Uddin et al., 2009; Andrews-

Hanna, 2012), and is thought to be important for problem solving,

focusing of attention, making decisions, and control of action –

in other words for getting things done. However, activation of the

TPN also has a deleterious effect on other cognitive functions that

are essential to leadership: it suppresses activity in the DMN.

The DMN plays a central role in emotional self-awareness

(Ochsner et al., 2005; Schilbach et al., 2008), social cognition

(Schilbach et al., 2008; Jack et al., 2012; Mars et al., 2012), and

ethical decision making (Koenigs et al., 2007; Bzdok et al., 2012;

Jack et al., in press). It is also strongly linked to creativity and

insightful problem solving (Subramaniam et al., 2009; Takeuchi

et al., 2011). The antagonistic relationship between the TPN

and DMN creates a fundamental neural constraint on cogni-

tion that is highly relevant to the different roles and capabilities

that effective leaders must astutely juggle and deploy. An impor-

tant consequence of this constraint is that an over-emphasis on

task-oriented leadership can prove deleterious to an organiza-

tion: in particular when openness to new ideas, people, emotions,

and ethical concerns are important to success. On the other

hand, the over emphasis on relationship oriented leadership may

prove deleterious to focus and the execution of clearly deﬁned

goals.

We are not the ﬁrst to note the striking correspondence between

the functions of brain networks and leadership. In a recent book,

social neuroscientist Matthew Lieberman distinguishes between

the social brain and the business brain and doesnote the antagonis-

tic relationship between the TPN and the DMN (Lieberman, 2013,

p. 257). In this paper, we extend Lieberman’s preliminary work by

providing a more nuanced account of the antagonistic relationship

between the TPN and the DMN and their corresponding task and

relationship leadership roles. We then extend Lieberman’s work by

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

theorizing strategies that will allow leaders to effectively navigate

this fundamental cognitive constraint.

OPPOSING NEURAL DOMAINS

When discussing the functional anatomy of the brain, it is

important to appreciate that the literature is less clear than is

often acknowledged. The cognitive characterization of the ten-

sion between the DMN and the TPN that guides this inquiry is

different in important respects from the view that is frequently

stated as accepted and uncontroversial (including in the context

of leadership, e.g., Waytz and Mason, 2013). Nonetheless, as we

will brieﬂy review, the characterization we offer is better sup-

ported by the scientiﬁc evidence. A number of inconsistencies

in the literature facilitate misunderstanding and over-conﬁdence:

ﬁrst, anatomical labels are not always used consistently and some-

times fail to distinguish areas which careful evidence reveals have

quite distinct functional roles (Kubit and Jack, 2013). Second,

researchers working with different types of cognitive tasks often

form quite different views about the primary functional role

of a region or a set of regions (i.e., a network). In particular,

it is now well acknowledged that researchers have not always

been careful to critically evaluate the evidence supporting their

inferences about the functional role of brain areas, and some-

times fail to consider evidence about function that derives from

a broader view of the literature (Henson, 2006; Poldrack, 2011).

Third, networks can be deﬁned according to a number of different

criteria.

The issue of how networks can be deﬁned is important to clar-

ify for our purposes. Networks of regions are most often deﬁned

(and g iven a label denoting a provisional functional role) either

because: (1) they are frequently found to be activated by a class

of cognitive tasks (Corbetta et al., 1998; Duncan and Owen, 2000;

Van Overwalle, 2009); or (2) the regions demonstrate strong posi-

tive resting state connectivity w ith each othe r – a ﬁnding which is

commonly interpreted as indicating a deg ree of functional coher-

ence amongst the regions in the network (Fox et al., 2005; Cohen

et al., 2008; Vincent et al., 2008; Van Dijk et al., 2010; Yeo et al.,

2011). These two criteria are thought to be complementary and

broadly consistent with each other, although they do not always

yield identical results (Laird et al., 2013).

A third quite different principle that can be used to deﬁne a

network is through the tendency of a set of regions to be deac-

tivated (i.e., less active than when the participant is at rest) by

a class of cognitive tasks (Shulman et al., 1997b). This is how

the DMN was originally deﬁned (Raichle et al., 2001). Regions

involved in the DMN have also be deﬁned: (a) on the basis of pos-

itive resting connectivity between regions in the network, and (b)

on the basis of negative correlation (“anti-correlation”) with other

regions as revealed by resting functional connectivity. These three

ways of deﬁning the DMN (deactivation, positive correlation, and

anti-correlation) are broadly consistent and regarded as comple-

mentary (Fox et al., 2005; Raichle and Snyder, 2007; Buckner et al.,

2008; Uddin et al., 2009; Andrews-Hanna, 2012).

The deﬁnition we take as primary for both the DMN and the

TPN is their anti-correlation. This follows the original deﬁnition

of these networks (Fox et al., 2005). An illustration of the two

networks, deﬁned in this way, can be seen in Figure 1. Figure 1

also illustrates a recent state-of-the-art division of the entire cor-

tex into seven networks based on functional coupling between

regions, i.e., positive resting connectivity (Yeo et al., 2011). The

mapping between Yeo et al.’s (2011) positive connectivity maps

and the anti-correlated networks is illustrated in greater detail in

Figures 2 and 3, for the DMN and TPN, respectively. This reveals

the need for some revision of early characterizations of the anti-

correlated networks. The broad characterization of the DMN has

remained quite stable, and is reasonably consistent using positive

and negative correlation criteria (Figure 2). However, this is not

the case for the TPN (Figure 3). It was originally thought (Fox

et al., 2005) that the DMN was anti-correlated primarily with the

dorsal attention network (Corbetta et al., 1998). However, sub-

sequent work using more data-driven methods (Fox et al., 2005;

Chang and Glover, 2009; Chai et al., 2012; Jack et al., 2012) has

revealed that the TPN overlaps par ts of both the dorsal atten-

tion network and the fronto-parietal control network (Vincent

et al., 2008

). Figur

e 3 also illust

rates clear overlap between the

TPN and the resting state network which Yeo et al. (2011) identify

with the ventral attention network (Fox et al., 2006). The ven-

tral attention network is recruited during a var iety of demanding

attention tasks, however, its precise functional characterization

is still subject to debate. It has been characterized as involved

primarily in the reorienting of attention, however, more recent

analysis suggests that the network as identiﬁed by Yeo et al. (2011)

can be better characterized as being involved in detecting and

responding to task-relevant stimuli (Kubit and Jack, 2013). Hence,

particularly for the TPN, it is important to note there is inconsis-

tency between the criteria for deﬁning networks. It appears that

DMN is in tension with a set of regions (i.e., the TPN) that lie

within a number of different cortical networks that can be dis-

tinguished and deﬁned by distinct proﬁles of positive functional

coupling.

THE DEFAULT MODE NETWORK

Until quite recently, the function of the DMN was regarded as

largely mysterious. In a comprehensive review, Buckner et al.

(2008) noted that “a unique challenge for understanding the func-

tions of the brain’s default network is that the system is most

active in passive settings and during tasks that direct attention

away from external stimuli.” While these and other researchers

acknowledged the difﬁculty in deﬁnitively identifying a function

for the DMN, nonetheless a consensus view began to emerge: that

the tension between the TPN and DMN could be best explained by

a distinction between two kinds of attention, namely externally vs.

internally directed cognitive processing. T his view was encouraged

by some inﬂuential early work (Gusnard et al., 2001), by the mis-

taken view that the DMN was anti-correlated with regions whose

primary function was external attention (Fox et al., 2005), and by

the paucity of evidence that any externally oriented task could pro-

duce activation of the DMN above resting le vels (but see Iacoboni

et al., 2004).

It is clear that the DMN is reliably engaged by some tasks that

involve attention to internal stimuli. Notable examples include:

autobiographical recall and prospection (imagining future events;

Buckner and Carroll, 2007; Spreng et al., 2009; Spreng and Grady,

2010), self-related processing (Mitchell et al., 2006), and emotion

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

FIGURE 1

Different criteria for deﬁning brain networks. Functional

networks shown on the left hemisphere of the PALS atlas (top – lateral

view above, bottom – medial view). Left panels show anti-correlated

networks derived from resting state data. The TPN is shown in blue, the

DMN in orange/yellow. Darker colors correspond to anti-correlated

networks using a global (whole brain) regressor, brighter colors without a

global regressor (i.e., most strongly anticorrelated areas). Maps derived

from data and methods described in Jack et al. (2012). Right panels show

the seven network stable solution derived by Yeo et al. (2011) on the basis

of positive correlations in resting state data. Area shown in black

corresponds to a cross-section through sub-cortical structures, not

mapped by Yeo et al. (2011).

FIGURE 2

The default mode network (DMN). Strong overlap in the DMN

representation based on anti-correlations and positive correlations in resting

state data. Left panels show just the DMN derived from anti-correlations in

orange/yellow. Right panels show networks derived by Yeo et al. (2011) with

substantial overlap. Borders of anti-correlated regions are carried over from

the left panels. Key to Yeo et al. (2011) networks is shown in the middle of the

ﬁgure. Labels denote best current understanding of the primary functions of

key parts of the DMN.

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

FIGURE 3

The task-positive network (TPN). The TPN represent ation based

on anti-correlations corresponds to multiple networks deﬁned by positive

correlations. Left panels show just the TPN derived from anti-correlations in

blue. Right panels show networks derived by Yeo et al. (2011) with subst antial

overlap. Borders of anti-correlated regions are carried over from the left

panels. Key to Yeo et al. (2011) networks is shown in the middle of the ﬁgure.

Labels denote best current understanding of the primary functions of key

parts of the TPN.

self-regulation (Ochsner et al., 2005). However, the characteriza-

tion of the DMN as being primarily involved in internal processing

glosses over other highly reliable ﬁndings which associate DMN

activation with externally oriented tasks. These types of externally

oriented tasks fall into two broad categories, which correspond to

a fractionation of the DMN which can be observed both through

meta-analysis of activation ﬁndings (Andrews-Hanna et al., 2010)

and by resting connectivity analysis (Uddin et al., 2009). More

dorsal parts of the midline structures of the DMN, medial parietal

and dorsal medial prefrontal cortex (dmPFC) are reliably associ-

ated with thinking about the mental states of others, including

both their emotional and cognitive (e.g., belief) states (Amodio

and Frith, 2006; Schilbach et al., 2008; Van Overwalle, 2009).

In contrast, ventral medial prefrontal cortex (vmPFC) is associ-

ated with representing the value of external objects (Grabenhorst

and Rolls, 2011). Moral decision making tasks, which are again

predominantly external in focus, are associated with activity in

both dorsal and ventral parts of the DMN (Greene et al., 2004;

Bzdok et al., 2012; Koenigs et al., 2012; Jack, in press). Since these

ﬁndings appear to be inconsistent with the commonly held view

that the tension between the TPN and DMN reﬂects a tension

between external and internal attention, we will temporarily set

aside that view and return to it in a later section. Instead, the

characterization of the DMN we offer here will be guided by

a broad view of tasks that are positively associated with DMN

engagement.

The DMN may be seen as having two primary circuits. The

ﬁrst is the dorsal portions of the midline structures and the right

temporo-parietal junction (rTPJ), which are most clearly asso-

ciated w ith mentalizing, that is, thinking about our own and

other’s mental states (Ochsner et al., 2005; Amodio and Frith,

2006; Schilbach et al., 2008; Van Overwalle, 2009). It is impor-

tant to note that the anatomy of this mentalizing circuit is distinct

from other regions that make distinct contributions to social

cognition. These include: a set of regions primarily involved in

perceptual processing of social stimuli such as faces and bodies

(Wiggett et al., 2009) and the mirror neuron network, which is

involved in both executing and observing actions, and thought to

underlie our ability to mimic the actions of others. In contrast,

the mentalizing system is thought to be involved not in emo-

tional contagion, but in the cognitive representation of emotion

(Lindquist et al., 2012).

The second of the primary DMN circuits is the more ven-

tral portions of the midline structures, which are associated w ith

self-related processing, autobiographical memory and prospec-

tion, cognitive representation of emotion, representation of

value/reward, emotion self-regulation, and autonomic process-

ing. We endorse the characterization of these processes offered

byarecentreview(Roy et al., 2012), which concludes: “The

vmPFC is not necessary for affective responses per se, but is

critical when affective responses are shaped by conceptual infor-

mation about speciﬁc outcomes. The vmPFC thus functions as

a hub that links concepts with brainstem systems capable of

coordinating organism-wide emotional behavior, a process we

describe in terms of the generation of affective meaning.” The

severe consequence of poor function in this circuit has long

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

been known through the highly inﬂuential work of Damasio

(1994) looking at patients with vmPFC damage. These patients

exhibit severely impaired social, moral, and decision making

behavior, cannot hold down a job, and tend to be shunned

by family members, even though they often have high IQs

(Anderson et al., 1999). More recent work with moral decision

making tasks indicates that they also tend to favor a course of

action, which appears to promote the best over all outcome, even

when that involves denying individual rights and directly harming

others. In other words, they are more utilitarian in their thinking

(Koenigs et al., 2007).

With regard to leadership roles, the DMN is the basis for rela-

tional roles in which the leader makes sense of their own and

other people’s emotions and helps to construct a sense of purpose

or vision for the group. Given that both subsystems of the DMN

are typically deactivated by tasks that activate the TPN, it is trou-

bling to imagine the potential consequences of placing too strong

an emphasis on adopting leadership roles that recruit the TPN.

THE TASK-POSITIVE NETWORK

The broadly deﬁned functions of the TPN are agreed in the lit-

erature, although some debate persists about the best way to

characterize the function of the ﬁner grained functionally cou-

pled networks which it overlaps (Kubit and Jack, 2013). The

TPN is activated, and the DMN deactivated, by wide variety

of non-social tasks including those involving focused attention,

working memory, language, logical reasoning, mathematical rea-

soning, and causal/mechanical reasoning (Shulman et al., 1997a,b;

Duncan and Owen, 2000; Fox et al., 2005; Owen et al., 2005; Van

Overwalle, 2011). While the TPN includes some regions of the

brain associated with social processes such as the mirror neu-

ron network, it is distinct from the mentalizing network of the

DMN both in its location and in the types of tasks that activate

it (Van Overwalle and Baetens, 2009; Van Overwalle, 2011; Jack

et al., 2012). The TPN includes parts of the dorsal attention sys-

tem (Fox et al., 2005), the fronto-parietal control network (Vincent

et al., 2008), and the ventral attention network (Fox et al., 2006;

Kubit and Jack, 2013). Using more relaxed criteria, it can also

be seen to overlap the somatomotor network (Figure 3). These

networks are broadly associated with focus on, and execution of,

well-deﬁned tasks that are non-social in nature. Jack (in press)

deﬁnes the TPN as “analytical–empirical–critical reasoning, such

as mechanical reasoning.” Given our current understanding of

the TPN, leadership roles associated with this network are those

focused on ﬁnancial planning, metrics, forecasting, problem solv-

ing, as well as strategic social engagement for the purpose of task

achievement.

THE OPPOSING DOMAINS HYPOTHESIS

As discussed, it is generally accepted in the cognitive neuroscience

literature that the DMN and the TPN are anti-correlated (Greicius

et al., 2003; Fox e t al., 2005; Fransson, 2005; Golland et al., 2007;

Tian et al., 2007; Buckner et al., 2008; Jack et al., 2013b). It is also

broadly agreed that a variety of cognitively demanding non-social

tasks tend to activate the TPN and deactivate the DMN. There has

been less agreement about how best to cognitively characterize the

tension between these networks.

The opposing domains hypothesis predicts that activation of

the DMN or the TPN is a result of the type of thinking a per-

son engages in to order to complete a given task, regardless of

whether the task is externally or internally oriented (e.g., relates

to perceived stimuli vs. to information recalled from memory).

Personality factors play a role in the deployment of these net-

works, particularly for tasks where the most productive strategy is

unclear or ambiguous. Nonetheless, it is thought that all neurotyp-

ical individuals are capable of ﬂexibly deploying these networks,

hence tasks which have a clear affordance for one type of pro-

cessing over the other will tend to engage the relevant network.

When a person engages in a cognitively demanding non-social

task, they will tend to deﬁne their role as having an analytic focus.

In this circumstance, both the opposing domains hypothesis and

other accounts predict the TPN will tend to be activated and the

DMN deactivated. When a person engages in a task that involves

mentalizing and/or thinking about affective meaning, and as a

result, deﬁnes their role as social and/or relational, the oppos-

ing domains hypothesis predicts the DMN will be activated and

the TPN deactivated. This prediction, which entails that an indi-

vidual’s analytic abilities are suppressed when they are empath-

ically engaged with people, is unique to the opposing domains

account.

Jack et al. (2012) used functional magnetic resonance imaging

(fMRI) to record brain activity when participants were engaged

in social vs. mechanical/analytic tasks. The social tasks required

participants to answer questions about the beliefs and attitudes of

the characters in emotionally and morally laden text passages or

video clips. The mechanical tasks required participants to com-

plete science puzzles, presented either as written passages, or as

video clips taken from the Video Encyclopedia of Physics. A rest

condition was included, in which participants were only asked to

stare at a ﬁxation point, for the purpose of establishing a resting

baseline against which both activations and deactivations could be

observed.

The ﬁndings showed that the neural activation during the

social tasks, speciﬁcally the activation of the rTPJ, medial pari-

etal/posterior cingulate and the medial prefrontal cortex, was

accompanied by the deactivation of the neural networks respon-

sible for mechanical reasoning, speciﬁcally, the superior frontal

sulcus, lateral prefrontal cortex, and the intraparietal sulcus. Con-

trols were put in place for task and perceptual demands to rule out

the alternate hypothesis that the TPN vs. DMN dichotomy can be

best accounted for by internal vs. external attention.

Jack et al. (2012) conclude that the anti-correlation between

the TPN and DMN “reﬂects a powerful human tendency to dif-

ferentiate between conscious persons and inanimate objects in

both our attitudes and modes of interaction” (p. 396). Further

work (Jack et al., 2013b) looking at humanizing vs. objectifying

(where people tend to be viewed as objects) narratives about peo-

ple provides additional support for the view that these networks

are not driven by the surface characteristics of the stimuli, but

rather reﬂect ﬂexible cognitive stances which can be deployed

depending on the attitudes and role adopted by the individual.

The implications of these ﬁndings for leadership and organiza-

tional behavior are vast. The dualit y of task and relationship;

inanimate and animate; and social and non-social can be found in

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

the personalit y, motivation, group dynamics, socialization, con-

ﬂict, trust, decision making, mindfulness, and moral reasoning

literatures.

A further wrinkle to this account is important to note. While

it is very well documented that the TPN and DMN tend to be

antagonistic, both at rest and during the performance of tasks, the

dichotomy between these networks is not absolute. They can be

simultaneously activated and work in cooperation with each other

(Fornito et al., 2012; Meyer et al., 2012). Both theory and observa-

tion indicates that this occurs when participants are engaged in a

type of social reasoning that is highly instrumental, and lacking in

genuine empathy. For instance, this pattern is seen more in indi-

viduals who are prone to Machiavellian thinking (Bagozzi et al.,

2013). It is also the neural signature that is seen when people are

animalistically dehumanized, rather than objectiﬁed (Jack et al.,

2013b).

This recruitment of the TPN during social cognition is likely

due to a person thinking critically, strategically, or mechanically

about people. Hence, it does not appear to be the case that we

can be both genuinely empathetic and analytic at the same time.

Instead, our view is that when these networks work co-operatively

with each other (Fornito et al., 2012), they realize a different type

of cognitive processing from either genuine empathy or pure ana-

lytic reasoning. When we engage the analytic network alongside

the DMN, this corresponds to a mode of social interaction that

is alienating to others. This is reﬂected in everyday language.

When we say that people are being “manipulative” or “calculat-

ing,” the metaphor may make sense because they are using TPN

brain regions involved in ﬁne motor control and/or mathemat-

ical reasoning to operate on social representations (Meyer et al.,

2012). Although this cognitive mode involves a sense of social

or emotional distance, it is clearly useful to think “strategically”

or “politically” in this way on o ccasion. This is reﬂected in the

observation that strategic social reasoning plays a crucial role in

leadership research. For example, Boyatzis and Goleman (2007)

make the distinction between emotional intelligence (EI) and

social intelligence (SI). EI is deﬁned as the ability of a person

to recognize, understand, and use emotional information about

oneself or others (Boyatzis, 2006). We expect these competencies

will tend to recruit both the DMN and the TPN in tandem. By con-

trast, SI competencies are non-task focused, for example, acting

with compassion, and would therefore primarily recruit the DMN

(Boyatzis and Goleman, 2007). Interestingly, Boyatzis et al. (2006)

predicted that SI competencies are linked to the parasympathetic

nervous system, while focusing on tasks and instrumental use of

EI is linked to the sympathetic nervous system. This would suggest

that EI competencieswould relate tothe TPN.Research is currently

underway to explore what overlap exists between the divisions of

the autonomic nervous system and the TPN and DMN. While the

evidence is far from conclusive, early ﬁndings based on recent

meta-analyses (see, for example, Beissner et al., 2013) indicate

that the TPN is more closely linked to the sympathetic ner-

vous system while the DMN is closer to the parasympathetic

nervous system. The labels in Figures 2 and 3 indicate regions

which show the closest correspondence between the DMN and

TPN and regions implicated in autonomic nervous function

(Eisenberger and Cole, 2012).

OPPOSING ROLES OF LEADERSHIP

LEADERSHIP ROLES

A role is a constellation of behavior and expectations enacted in a

social situation. Mead (1934) claimed that roles were embedded in

the situation and therefore were constantly emerging and evolving.

As a sociologist and helping to formulate the interactionist theory

of roles, Mead (1934) emphasized the expectations of others in

the forming and reforming of roles. In this way, roles are distinct

from style. Leadership style describes a dispositional comportment

of a person as a leader that can vary by situation and demand.

Leadership style contributes to role enactment as Biddle (1986)

explained from a functionalist theory of roles.

In leadership, a role provides a guide to one’s behavior and

helps new role occupants become socialized into how they could

or should act in a way consistent with the norms and culture of

an organization. A leadership role of a person emerges in response

to the expectations of others around the leader, especially the fol-

lowers, and the leader’s own style/s, competencies, and values

(Boyatzis, 1982). This process of role emergence as a result of the

interaction between expectations and behavior has been explained

by cognitive role theory (LeMay, 1999).

There appears to be two distinct components of the enactment

of leadership roles that parallel the distinction between the TPN

and the DMN: task achievement and relationship development.

We will place the evolution of this duality of leadership roles in

historical context.

FROM TAYLOR TO BASS APPROACHES TO LEADERSHIP ROLES

The centrality of tasks to a leader’s role was evident in Cowley’s

(1928) deﬁnition of a leader: “an individual who is moving in a

particular direction and who succeeds in inducing others to follow

him...A leader then, is a person who is going somewhere, who has

a motive, who has a program.” (p. 145–146). In short, without a

task, leaders do not exist.

Fredrick Taylor’s scientiﬁc management movement empha-

sized the task aspect of leadership. Followers were viewed as

“machines” to be used efﬁciently by the leader. In other words,

by treating humans as inanimate objects subject to analyti-

cal analysis, the role of a leader in the scientiﬁc management

era was largely reliant on recruitment of the TPN rather than

the DMN (Jack et al., 2013b) and therefore discouraged the

leader from fostering more meaningful social–emotional con-

nections with followers. The introduction of the “human” side

of leadership in the late 1930s gave rise to a multitude of

leadership frameworks that reﬂected the scientiﬁc management

(task) – human relations (relationship) duality. These frameworks

include the autocratic (task)-democratic (relationship) contin-

uum (Tannenbaum and Schmidt, 1958), the Ohio State Leader-

ship Studies on structure (task) and relationship (Shartle, 1950;

Fleishman, 1953; Halpin and Winer, 1957), and the Michigan

Leadership Studies (Katz et al., 1950) of a production orientation

and an employee orientation.

ATTEMPTING TO INTEGRATE TASK AND RELATIONSHIP ROLES FROM

THE 1950s TO THE 1970s

When symbolic interaction theory transferred from social to orga-

nizational psychology, leadership theory changed to require both

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

relationship building and task attainment. This development par-

allels the duality of the neural opposing domains: relational

leadership competencies are facilitated by the DMN, while the

task competencies are facilitated by the TPN. Bales (1958) was

instrumental in this theoretical shift in the leadership literature

with his claim that a task-oriented group requires two types of

leaders to maximize effectiveness; one leader to attend to the

task functions of the group (inst rumental leader) and another

to attend to the emotional needs of the group (expressive leader).

The suggestion that two leaders are required in a group stood

in stark contrast to the idea that all leadership roles could be

carried out by one individual. Given that we now know that

the neural networks underlying these two roles are antagonis-

tic, having two leaders in a group presents what initially appears

to be a reasonable strategy for relieving the task–relationship

tension.

By the 1960s, a consensus began to form that building relation-

ships with followers was a fundamental component of leadership

and a critical ingredient in task achievement. Leadership theorists

become interested in developing strategies that would enable lead-

ers to integrate these two fundamental leadership roles. The early

Ohio State Leadership Studies (Fleishman, 1953) conﬁrmed the

conceptual distinction between task-oriented and relationship-

oriented leadership and developed a measurement tool for each.

Following from this work, task-oriented leadership has been iden-

tiﬁed as being concerned with production (Blake and Mouton,

1964), goal achievement (Cartwright and Zander, 1960; Bowers

and Seashore, 1966), labor allocation, enforcement of sanctions

(McGregor, 1960), and creating a context and structure for fol-

lowers (Shartle, 1950; Fleishman, 1953; Halpin and Winer, 1957).

At the individual level, task-oriented leaders were said to have a

high need for achievement (McClelland, 1961; Wofford, 1970), to

be achievement oriented (Indvik, 1986), and to be cold and aloof

to signal their preference for psychological distance from follow-

ers (Blau and Scott, 1962). Interestingly, these descriptions are not

only consistent with the role of the TPN discussed earlier, but also

allude to the suppression of the DMN – in particular, they ﬁt well

with our observation that the socially distancing effect of dehu-

manizing is associated with a shift from DMN to TPN engagement

(Jack et al., 2013b).

Conversely, relationship-oriented leaders were described as

being focused on follower well-being (Hemphill, 1950), con-

cerned with developing and maintaining relationships (Hersey

and Blanchard, 1969), to be more democratic as opposed to

autocratic (Bass, 1990), and as placing value on friendships,

open communication, and mutual trust. These psychological fea-

tures ﬁt a proﬁle of DMN activation and TPN deactivation –

the neural signature associated with thinking about the expe-

riences of others (Jack et al., 2012) and humanizing (Jack et al.,

2013b). Individuals who are humanized experience positive emo-

tions, whereas those who are dehumanized experience negative

emotions (Bastian and Haslam, 2011). Correspondingly, research

has shown that relationship-oriented leaders are associated with

higher job satisfaction and lower turnover in organizations (Bass,

1990; Yukl, 2006).

Yukl (1999) provides the most recent taxonomy of task and rela-

tion oriented leader roles. He also distinguished a third dimension

in his taxonomy, labeled “change orientation” which includes a

mix of task and relationship oriented roles. However, in critiquing

other’s work, Yukl (1999, 2008) notes the limitations of the task–

relationships distinction, speciﬁcally, the inconsistent empirical

results and the lack of evidence that it is indeed possible to be

high in both task and relationship competencies and that being

so is related to leadership effectiveness (Yukl, 1999). Research con-

ducted since adds support toYukl’s critique; speciﬁcally, Kaiserand

Kaplan (unpublished, as cited in Bass and Bass, 2008 ) found that

in a sample of managers from a consulting ﬁrm, 46% were highly

task oriented, 19% were highly relationship oriented, and only 6%

showed ﬂexibility between the two dimensions. The remaining

29% were low in both task and relationship orientation.

The distinction between task-oriented roles vs. relationship-

oriented roles presented in Yukl’s taxonomy appears to closely

parallel the distinction made between social and mechanical tasks

used to test the opposing domains hypothesis (Jack et al., 2012).

Speciﬁcally, the task-oriented tasks are, by and large, mechanical

and analytical in nature and would therefore activate the leader’s

TPN, while the relationship oriented tasks are generally social in

nature and would therefore activate the leader’s DMN.

The anti-correlation between these two networks suggests that

when a leader is focused on a task-oriented role, their ability

and desire to attend to the relationship needs of their followers

is diminished. This is not to say that leaders do not have the ability

to be both highly analytical and to build effective relationships. In

the absence of any task, it is known that the human brain naturally

cycles between activation of the TPN and activation of the DMN,

activating each several times in a period of a minute (Fox et al.,

2005). Hence, it is clearly possible to switch between networks and

possibly roles.

However, the evidence from neuroscience does suggest that

leaders

cannot sim

ultaneously attend to these distinct roles, hence

conﬂicts are likely to arise if leaders get “stuck” in one role,

which decreases their ability to switch between the two networks,

resulting in the prolonged suppression of one of the networks

and associated roles. This trade-off has been noted by a num-

ber of scholars. Most recently Yukl (2008) notes that “efforts to

improve one performance determinant may have an adverse effect

on another performance determinant...when leaders are preoccu-

pied with responding to external threats (task), there is less time for

people-oriented concerns such as being supportive and developing

member skills” (pp. 711–714). The opposing domains hypothesis

provides an alternative explanation for why this trade-off occurs.

Rather than, or at least in addition to, a simple lack of time

and overall cognitive resources, the opposing domains hypoth-

esis suggests that the suppression of the opposing neural network

further exacerbates the “adverse effect on another performance

determinant.”

The antagonistic relationship between the TPN and DMN

may also help us to understand why the type of role a leader

adopts moderates the relationship between leadership role (task

vs. relationship), task performance and leadership effectiveness

respectively. Speciﬁcally, Burke (1965) found that a group’s

performance of a coding task was completed more effectively

under a production-oriented (task) leader while the shared

decision-making (task) was carried out more effectively under a

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

relationship-oriented leader. The opposing domains hypothesis

offers two explanations for this ﬁnding – the ﬁrst related to the

ability of the followers to perform the task, the second related to

the leadership role that the followers’ perceived to be effective.

First, followers performed more effectively on the analyti-

cal task when the leadership role matched the neural network

activated as a result of the analytical nature of the task (the

TPN). When a relationship-oriented leadership role was used,

the leader’s behavior and focus on relationships may have acti-

vated the DMN in followers, which we now know to be associated

with lapses in attention and performance errors (McKiernan et al.,

2003; Mason et al., 2007; Fassbender et al., 2009; Pyka et al., 2009),

hence the lower performance on the analytical task. Conversely,

the shared decision task, which required interpersonal interaction

and therefore would have activated the DMN, was performed more

effectively under a relationship oriented leader. A relationship-

oriented leader in this task would have allowed for the follower’s

DMN to be dominant, which is the network required to perform

this task effectively. A task-oriented leader would have activated

the follower’s TPN, which would suppress the DMN resulting in

a reduction of human and ethical insight (Jack et al., 2013b, in

press) and consequently lower task performance.

Second, when completing the coding tasks the followers were

predominately engaging their TPN, hence the task-oriented lead-

ership role may have been perceived by followers to be more

effective because it was more psychologically resonant. Similarly

the relationship-oriented leadership role may have been preferred

by followers engaged in the moral decision task because of the

resonant emphasis on DMN engagement for both the task and the

leadership role.

Further evidence of the interaction between task type and

follower preference for a task vs. relationship-oriented leader is

offered by Weed et al. (1976). This study found that the only type

of task that was a signiﬁcant moderator between leadership role

and task performance was the difﬁcult–ambiguous task. While this

study was focused on the difﬁculty and ambiguity of the task rather

than the mechanical or social nature of the task, the authors note

that the difﬁcult–ambiguous task was the only task that involved

ethical/moral problems and therefore “the signiﬁcant ﬁnding s on

this task may be partially attributable to the types of skills required

to deal with these problems” (p. 65).

Finally, Hersey and Blanchard (1969) developed the Tri-

Dimensional Leadership Model that sought to link task and

relationship oriented leadership roles to follower perceptions of

leadership effectiveness. The Tri-Dimensional Leadership Model

suggests that leaders who are highly task oriented will be per-

ceived as “effective” because they know what they want and are

able to impose this to accomplish a task without causing resent-

ment. A highly task-oriented leader will be perceived as ineffective

when followers’ perceive the leader has no conﬁdence in others, is

unpleasant, or only shows interest in short-run output. In other

words, a highly task-oriented leader will only be seen as effec-

tive if they are aware of, and attend to, relationships – rather

than dehumanizing their followers – while accomplishing the

task.

Similarly, a leader with a high relationship-orientation will be

seen as effective by followers when they perceive the leader has

implicit trust in them and is primarily concerned with develop-

ing their talents – behaviors associated with humanizing and the

DMN. They will be seen as ineffective when followers perceive

the leader to be passive and showing little care about the task at

hand – behaviors associated sustained activity of the DMN that

suppresses the TPN (Mason et al., 2007; Jack et al., 2012). These

ﬁndings extend the hypotheses derived from the Burke (1965)

study with the suggestion that regardless of a leader’s preferred

role (task-oriented or relationship oriented), the leader must be

able to switch ﬂuidly between the opposing neural networks (the

TPN and the DMN) in order to be perceived as effective by their

followers. Given this, it seems that rather than identifying the

individual difference variables and situational variables that are

associated with each st yle in isolation, it may be more important

to identify variables that correspond to the ability of a leader to

switch between the TPN and the DMN in order to maximize leader

effectiveness.

DEALING WITH THE DIALECTICAL TENSION

The previous section has shown that the distinction between task

and relationship roles is evident throughout the leadership liter-

ature. In this section, we explore potential strategies to resolve,

or at least minimize the consequences of the tension that leaders

face in developing their roles to attend to task requirements and

attending to relationships as a result of the antagonism between

the TPN and the DMN.

NEURAL DISPOSITION: MATCHING THE LEADER TO THE SITUATION

As with hormonal disposition, there is some evidence to suggest

that humans have a natural disposition toward either analytical–

mechanical reasoning and therefore the TPN or social–relational

reasoning and therefore the DMN (Jack et al., 2012). While neural

disposition is a relatively nascent area, hormonal dispositions are

well documented in the literature (see Insel, 1997; Schulkin, 1999).

For example, people with higher unconscious power motives are

known to have higher resting levels of epinephrine secretion, while

people with a higher unconscious Need for Achievement appear

to have higher resting levels of vasopressin secretion (Boyatzis and

Sala, 2004; Boyatzis et al., 2006). At the behav ioral level, neural

and hormonal dispositions may underlie certain personality char-

acteristics, learning styles, and perhaps preferred leadership roles

(Boyatzis and Sala, 2004).

Considering the opposing domains model, one way to address

the tension between the task and relationship leadership roles

could be to match a leader’s predisposition toward either task

or relationship to the type of task they perceive as their primary

role and responsibility. In many ways, Bales’ (1958) suggestion of

the need for two leaders was an attempt to do this. The utility of

matching a leader’s natural inclination toward either a task or rela-

tionship role to the situation has found some empirical support

in the literature. For example, Slatter (1955) found that when task

demands are high, being liked does not contribute to leadership

effectiveness and social or relational skills are not highly valued,

whereas in therapy groups, sensitiv ity training groups, and social

clubs socio-emotional skills were important. However, the practi-

cal results from dividing the leadership roles have been less than

ideal.

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

Like the old CEO (chief executive ofﬁcer) and COO (chief

operating ofﬁcer) split, father and mother role split, CO (com-

manding ofﬁcer) and XO (executive ofﬁcer) split in the military,

this leadership role differentiation is possible but appears to be less

effective in the long-term. Attempts to have two people occupy

co-chair, or co-CEO roles have been, at best, confusing. Usu-

ally leaders in these roles cannot sustain true status and power

equalization over time. In addition, others around them may not

allow it and their expectations and preferences push for more

status differentiation, not less. Over time, one role dominates

the other, and as a result one neural network dominates brain

activation (even as an organizational norm) leading the organi-

zation down a narrow path of either problem focused decisions

with little openness to new ideas or events occurring in the larger

environment (like market shifts) OR an organization preoccupied

with environmental changes and meeting the various desires of

employees that has difﬁculty executing a strategy consistently over

time.

Differentiating the leadership role effectively allows a leader

to spend the majority of their time in one of the two neural net-

works and reduces the need to cycle between them. From the point

of view of sustained leadership effectiveness, ﬂuid cycling at rest

between the two networks is associated with good mental health

and higher IQ, whereas a lessening of the cycling between net-

works is associated w ith a variety of mental disorders (Broyd et al.,

2009; Andrews-Hanna et al., 2010; Anticevic et al.,2012; Whitﬁeld-

Gabrieli and Ford, 2012). While studies of the long-term effects of

privileg ing the engagement of one cognitive mode over the other

in task performance have yet to be done, it is plausible that a

more balanced approach is associated with better long-term men-

tal health and performance, whereas over-privileging one network

for sustained periods leads to mental exhaustion and burn-out –

two detrimental effects that are often discussed in the leader-

ship literature. Therefore, role differentiation, while presenting

an easier short-term strategy for an organization to accomplish

a balance, may prove far less productive if the individual’s role

remains constant over time.

These considerations suggest a more effective approach would

be to train and develop leaders so that they not only possess a

high level of competency in enacting both the task and relation-

ship leadership roles, but also have the ability to switch ﬂuidly

between them along an awareness of appropriate cues and con-

texts for doing so (French and Jack, in press), which also requires

a perceptual facility with perceiving when each is needed or more

appropriate. The a ctions to invoke a change of role are within a

leader’s purview. For example, a leader w itnessing a competitor

take some of their clients or market share could see this as a need

for analytic investigation – is there is pricing difference? Are there

differences in transportation costs or delivery speed?

Similarly, a leader who is heavily into a task role (and TPN

activation) may decide that each day he/she will coach another

person to help them develop. In a within subjects study com-

paring a method of coaching with the positive emotional attractor

(PEA; i.e., coaching with compassion) vs. the more t ypical method

of coaching someone to ﬁx them – to the negative emotional

attractor (NEA; i.e., coaching for compliance), components of the

DMN were signiﬁcantly activated in the PEA condition more than

the NEA condition (Jack et al., 2013a). Since this neural network

allows a person to be more perceptually and cognitively open to

new ideas, it may be the process needed to help people become

open to switching and engaging both domains with ﬂuidity. By

deciding to coach one person each day (formally or informally

over coffee or a lunch), and to commit to doing it about the per-

son’s dreams, vision or values, the leader commits to switching

into the relationship role at least once a day and activating the

DMN in both himself/herself and the other person. Over time

and practice, the switching between the task and relationship roles

would likely become easier and help the leader develop more cues

as to when one role is more appropriate or possible than the other.

Not quite as simple as changing hats or shoes, but it could become

as convenient.

NEURAL RESOURCE EFFICIENCY

The opposing domains hypothesis could be framed as presenting

a form of “trade-off” between adopting roles favoring task-

related leadership activities and therefore activating the TPN and

suppressing the DMN and adopting roles favoring relationship

building activities and therefore a ctivating the DMN and suppress-

ing the TPN. In this framing, neurological activation is essentially

a form of resource that leaders distribute between the task and

relationship roles to attain or increase their effectiveness. When a

leader expends more neural resources attending to relationships,

they consequently have fewer resources to invest in the task role

and vice versa. It may be possible that we can change the rate

or efﬁciency of the leader’s neural resources, thereby minimizing

the extent of activation required to successfully complete the task.

Because the TPN and the DMN are antagonistic, minimizing the

degree of activation in one network also serves to proportionally

decrease the suppression of the opposing network (McKiernan

et al., 2003; Pyka et al., 2009).

Task-positive network competencies

We expect that leaders with a high level of cognitive intelligence

(g) will show greater ability in analytical or mechanical type tasks.

They are also more likely to perceive the need for an analytic

leadership role focusing on the TPN and to adopt this role in

an organization. When faced with an analytical or mechanical

task, we expect leaders with high levels of cognitive intelligence

will require less “effort” or neurological resources to successfully

complete an analytical or mechanical task than leaders with lower

levels of cognitive intelligence (Graham et al., 2010). Similarly, we

expect that when faced with an analytical task requiring the str ate-

gic use of emotional information, leaders with high EI w ill require

less cognitive effort to engage followers than leaders with low EI.

Leaders with higher levels of EI can access the emotional infor-

mation more easily than those with low EI, even if it is in service

of an analytic task. This helps to clarify that emotional labor can

be in ser vice of either a task demand or an emotional and social

demand.

A leader who shows a high level of familiar ity or specialized

competence in a given task should also require less neurological

resources than leaders with a lower level of familiarity or special-

ized competence. A high level of task competence will result in

the leader experiencing the task with less intensity resulting in a

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

lower level of activation in the TPN. By the work appearing more

routine to a leader using an appropriate role, it may allow a leader

to switch roles more easily. Holding competence, experiences, and

intelligence constant, more difﬁcult tasks will require more neu-

rological resources than easy tasks (McKiernan et al., 2003; Mason

et al., 2007; Pyka et al., 2009).

Default mode network competencies

In similar vein to the reasoning discussed above, when faced with a

social or relational task, we expect leaders with high SI (as opposed

to EI – see Boyatzisand Goleman,2007) to require less neurological

resources to successfully complete a relational task relative to those

leaders with low SI. We would also expect that leaders who possess

high levelsof empathetic concern and compassion tobe ina similar

position. These leaders are also more likely to perceive the need

for and adopt a relational leadership role.

SWITCHING BETWEEN THE TPN AND DMN

While the idea that a task can be classiﬁed as either analytical or

social is useful for theoretical purposes, as with most dichotomies,

the distinction is rarely so clean-cut in practice. In reality, and par-

ticularly in the context of leadership, all tasks have a relational

component and an analytical component. Leadership almost

always requires consideration of both analytical tasks (TPN) and

relationships (DMN), therefore we suspect that the greater ability

a leader has to switch between these two modes of reasoning the

more effective they will be as a leader. We suspect that minimiz-

ing the suppression of the opposing network will make it easier,

faster, and less costly for the leader to switch between the two

networks. For example, we have already argued that leaders with

certain social competencies require less cognitive effort to com-

plete a social task than those without these competencies, resulting

in less activation of the DMN and less suppression of the TPN.

This reduction in the difference or gap between the two networks

should make switching between the two networks faster and less

costly.

Along with the ability to switch between the two opposing net-

works, which we argued may require a reduction in intensity of

the dominant network, we also expect the ability to appropriately

time the shift from task to relationship to be important both in

terms of minimizing disruption for followers and for maximizing

the effectiveness of the shift. For example, knowing that activa-

tion of the DMN suppresses our analytical–mechanical reasoning

abilities (Jack et al., 2012), and that activation of the DMN dur-

ing analytic tasks is associated with mind-wandering and lapses

of attention (Mason et al., 2007; Fassbender et al., 2009), it would

appear unwise for a leader to engage followers in activities requir-

ing social or relational reasoning at a time when it is important or

urgent to maintain analytic focus. Similarly, in situations requir-

ing social or relational reasoning, for example, during the group’s

formation period, introducing task-focused activities may inhibit

relationship development in the group.

The timing of the switch may also have implications for the

timing of feedback. Based on the opposing domains hypothesis,

we would expect that the closer the time period between action

and feedback, the more congruence there should be between the

type of feedback and the type of task. For example, if feedback is

being given while a person is performing an analytical task, the

feedback given should be analytical or technical in nature because

this is the type of feedback is consistent with the neural network

within which the receiver is engaged. If the leader wishes to give

feedback that requires the follower to engage in social or rela-

tional reasoning, he/she would be better to wait until the receiver

has disengaged from the analytical task. The same can be said

for g iving task-related feedback in an emotionally charged situa-

tion. Gottman et al. (2002) documented something every married

or partnered person should know. When your spouse or part-

ner is angry and yelling at you, this is not the time to reply by

analyzing the situation in emotional distant terms. It does not

help and in fact, as Gottman et al. (2002) document, inﬂames the

situation.

Finally, given that the decision to switch from one cognitive

domain to another requires reasoning about the emotional state

of self and others, we would expect that leaders with greater DMN

abilities are more likely to be able to successfully time the shifts

than leaders who lack such abilities.

In sum, by using the evidence of these antagonistic neural net-

works and recent research on activating each of them, we can

hypothesize that that most effective leadership requires a combi-

nation of three facilitating factors: (a) a decrease in the switching

time (or cycle time) between these networks; (b) t raining people

to high levels of competence in enacting the roles requiring each

network, so decreasing the cognitive effort required and hence the

degree of deactivation of the opposing network; and (c) training

leaders to recognize and perceive contexts and cues which require

a switch between modes, so they do not remain “stuck in set” and

apply an ineffective cognitive strategy for the task at hand (e.g., by

privileg ing analytic thinking when faced with an ethical decision,

or intuitive thinking when faced with a logical task where creativity

is not helpful).

To make this happen, we conjecture that people would have to

be trained in multiple techniques that invokea tippingpoint. These

techniques would function somewhat like cognitive behavioral

therapy, helping leaders to identifyexternal (e.g., follower thoughts

and emotional state) and internal (e.g., own thoughts and emo-

tional state) cues and respond appropriately. In particular, people

may need to develop the ability to calm the system and lower the

intensity when there is a danger they are becoming “stuck in set” –

whether that involves preoccupation with social/emotional con-

cerns (over-privileging the DMN) or an overly narrow task focus

(over-privileging the TPN).

This emphasis contrasts with current beliefs that to motivate

we must increase energy and incentives (whether positive or neg-

ative). While that approach may preclude the dualistic swing

between these two cognitive modes, and the occasionally con-

ﬂicting considerations raised by each, it also presents a larger

danger: losing sight of important insights, either emotional or

analytic. Additionally, the ability to sense the optimal timing and

context for switches may also be a critical component in under-

standing leadership effectiveness. For example, in the context of

learning a new task, interruptions can be extremely costly, not

only in terms of task outcomes (errors) but also in the ability to

“pick up where you left off.” Once a person has gained a higher

level of mastery, interruptions w ill be less costly and following the

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

interruption, the person will be able to re-engage with the task

faster.

Prior research on the intensity of emotions suggests that in

order to move a person from a negative emotional state (NEA) to

a positive emotional state (PEA), the intensity of the emotion must

be reduced to reach a tipping point (Boyatzis, 2013). It seems rea-

sonable to suggest that a similar principle mayexist when switching

between the TPN and the DMN. Boyatzis (2013) argues that when

people are in the PEA they are “more perceptually open and accu-

rate in perceptions of others” (p. 1978; see also Fredrickson and

Branigan, 2005; Boyatzis, 2006), which is consistent with the work

of the DMN in allowing individuals to engage in reasoning about

the emotions of others. In contrast, the NEA is said to be linked to

human survival, particularly to defend against threats. NEA also

balances “unchecked optimism” through suppressing the DMN,

which has been linked to poor investment decisions (Gibson and

Sanbonmatsu, 2004) – an analytical task that would require acti-

vation of the TPN rather than the DMN. The opposing domains

model suggests the NEA’s ability to balance unchecked optimism is

achieved through both activating the TPN required for analytical

reasoning and suppressing the DMN, which is largely responsible

for the overly optimistic state.

The link between the PEA–NEA and the DMN–TPN is also

reﬂected in Fiedler’s (1986) cognitive resources theory. Fiedler and

McGuire (1987, as cited in Bass and Bass, 2008) found that under

non-stressful conditions, leaders with ﬂuid intelligence (IQ) per-

form better than leaders with crystallized intelligence (experience);

however under stressful conditions, leaders with crystallized intel-

ligence performed better. Cognitive resource theory posits that

the reason for this distinction is that under stressful conditions,

a leader with ﬂuid intelligence will rely on intellectual solutions

to a task even when such solutions are inappropriate. In other

words, under stressful conditions, a leader is “stuck” in the TPN

and also in the NEA due to the stress condition, which limits their

ability to switch into the DMN and the PEA, which enables them

to explore more creative and non-intellectual solutions. Leaders

with crystallized intelligence (intelligence based on past experi-

ence and learning) are likely to experience the same situation with

less intensity, thus these leaders will be: (1) closer to the NEA–PEA

tipping point; and (2) more able to switch between the TPN and

the DMN.

FURTHER CONSIDERATIONS ABOUT MAPPING BRAIN NETWORKS

ONTO LEADERSHIP ROLES

In this article, we have focused on mapping a duality which has

long been noted at the behavioral level, between different lead-

ership roles, onto a duality in the brain, highlighted by recent

research in neuroscience. There appears to be a ver y promising

mapping between these two domains, which suggests a fundamen-

tal neurophysiological basis for the observed duality in leadership

roles. At the same time, we acknowledge that there is a con-

siderable distance between neurophysiological observations and

leadership behavior. Hence, more research is needed to ﬁrmly

establish the links we highlight, and to elaborate and extend the

model sketched here. Our main goal has been to highlight these

links as a very promising avenue for further research. With an

eye to this future research, in this section, we respond to three

speciﬁc questions that naturally arise in response to our proposed

mapping.

First, the DMN has been found to exist in many species

(Mantini et al., 2011). Further, in humans its function serves as

basic index of level of consciousness, even in non-communicative

brain-damaged patients (Vanhaudenhuyse et al., 2010). Given that

this network appears to have such basic functions, it may seem

surprising that we are suggesting it plays a key role in effective

leadership, which would appear a hig her level function. However,

we regard this as consistent with our account, which is based on the

view that the default network originally evolved to play a basic role

in visceral awareness and emotion self-regulation, and that these

functions expanded and evolved so that in the human this cortex

additionally supports complex representations of value and the

mental states of others (Jack et al., 2012, 2013b, in press). This ﬁts

with evidence that the default network is considerably expanded

in the human compared to other species, even when its size is

considered relative to overall cortical volume – which is massively

expanded in the human compared to other primates (Jack et al.,

in press; Figure 1).

More broadly, our rev iew indicates that we see the default

network as critically involved in self-management, in particular

mindfulness, motivation, and affective meaning. We see relational

aspects of leadership as an extension of these functions which

arises through coupling of them with our capacity to metalize. In

summar y, we suggest it quite natural to view the relational aspects

of effective leadership as a skillful cognitive blending of our basic

capacities for emotional self-regulation and social cognition. This

view sits very well with what is known about the function and

evolution of the DMN, and is not contradicted by ﬁndings which

indicate the DMN plays a role in more basic functions.

Second, we admit and welcome the possibility that there are

likely to be additional mechanisms, be yond the DMN/TPN dual-

ity we highlight, which are critical for understanding leadership.

For example, we have highlighted a duality which places two much

discussed systems involved in social cognition in opposition: the

mirror neuron system lies within the TPN, and hence there is

a general tendency for it to be deactivated w hen the metaliz-

ing system of the DMN is activated, and vice-versa. The mir ror

neuron system is thought to play a role in “emotional conta-

gion” (Hatﬁeld et al., 1994)–akeymechanism used by leadership

scholars to explain the transfer of emotion between leader and

follower (Bono and Ilies, 2006; Johnson, 2008; see also Sy et al.,

2005) and from followers to the leader (Dasborough et al., 2009).

Additionally, parts of both the mirror neuron network and the

DMN were activated in older executives when remembering spe-

ciﬁc moments with resonant vs. dissonant leaders in their past

(Boyatzis et al., 2012). In summary, it is not our claim that the

DMN/TPN dichotomy highlighted here represents an exhaustive

description of the neural processes involved in effective leader-

ship. We look forward to future research that may clarify different

types of neural interaction. In particular, we acknowledge the

importance of looking at ways in the DMN and TPN work coop-

eratively in addition to our focus here on the“default”tendency for

them towork competitively (Fornito et al., 2012). Such cooperative

interactions between the networks need not always be anti-

social in effect, although we document evidence above that some

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

modes of cooperation are associated with a greater sense of social

distance.

Finally, the methodological concern might be raised that our

analysis depends on reverse inference (Poldrack, 2011). That is,

since br ain imaging evidence is essentially correlational in data,

it is not clear that the DMN and TPN are essential to the spec-

iﬁed roles in effective leadership. We acknowledge this concern,

which applies to all neuroimaging data. One impor tant way to

mitigate faulty inferences of this type is to conduct broad anal-

yses of the literature in order to justify the claimed association

between a speciﬁc brain area and a speciﬁc function (Poldrack,

2011). Another important step is to conduct critical tests of the

hypothesized account against other accounts.

We have taken both these steps, using both meta-analysis (Jack

et al., in press) and critical tests of our theory (Jack et al., 2012,

2013b) to justify our view that the DMN vs. TPN dichotomy

reﬂects a tension between empathetic vs. analytic reasoning, as

opposed to the more broadly stated view that it reﬂects a ten-

sion between internal vs. external attention. Nonetheless, we

agree that further testing is wanted. Ideal tests would involve

directly up or down-regulating one of the networks, or mod-

ifying the ability to switch between them, and then assessing

the impact on naturalistic leadership behavior. As noted, there

is already evidence suggesting that patients with vmPFC dam-

age perform poorly in relational leadership roles (Anderson et al.,

1999), however it would be ideal to manipulate neural process-

ing and study effects w ithin individuals. While this is challenging

to do directly, some more indirect tests along these lines may

be possible. For instance, oxytocin administration appears to

up-regulate DMN function (Bethlehem et al., 2013), hence we

would predict oxytocin administr ation should improve perfor-

mance in relational leadership roles and deteriorate performance

in task-oriented roles. Alternatively, it has been shown that

different forms of meditation tend to either increase (focused

meditation) or decrease (non-dual awareness meditation) anti-

correlations between the DMN and TPN (Josipovic et al., 2012).

The former should increase leadership ﬂexibility (i.e., the abil-

ity to switch between different roles and perform well in both),

the latter decrease it. Exploring these and other potential manip-

ulations of neural processing is an important area for further

research.

CONCLUSION

The emergence of two distinct leadership roles, the task-oriented

leader and the relationship-oriented leader, has been documented

in the leadership literature since the 1950s ( Bales, 1958). Recent

discoveries in neuroscience that the TPN, which allows us to focus

on problem-solving and analytic work, is antagonistic with the

DMN, which allows a person to be socially engaged and open to

new ideas, creates a dialectical tension that reverberates through-

out the leadership role literature and raises questions as to how

leaders can effectively fulﬁll both roles.

RESEARCH, THEORY, AND PRACTICAL IMPLICATIONS

This paper has identiﬁed a key pattern in the leadership literature

and linked this pattern to cutting edge research in the neuro-

science domain. In doing so, we have raised a number of questions

regarding our treatment of the task and relationship distinction in

the literature to date, par ticularly the assumption that leaders are

able to attend to both leadership roles simultaneously. Addition-

ally, we havebeen ableto add further explanation to some historical

ﬁndings attempting to understand the interaction between task,

leadership roles, and leadership effectiveness. Finally, we sug-

gested an array of conceptual implications that might extend our

current conceptualization and operationalization of leadership

effectiveness.

From a practical standpoint, this paper suggests that developing

a leader’s analytical and relational abilities may be an important

way to offset the costly effects of the antagonistic relationship

between the TPN and the DMN. Increasing a leader’s abilities

in each role should facilitate the ability faster and more ﬂuidly

between the task and relationship roles by reducing the cogni-

tive effort, and consequently the differential activation between

the TPN and the DMN, required to perform effectively in each

respective role. We belie ve that the ability to switch between

these networks and corresponding leadership roles may be a key

component of leadership effectiveness.

Additionally, knowing that engagement in analytical tasks

inhibits our ability to engage in social or relational reasoning

and vice versa may have important implications for organiza-

tions in terms of how they structure and order tasks that have

analytical and relational components. For example, when giving

feedback, managers should consider if the feedback is analytical

or task related in nature or interpersonal in nature and time the

delivery of that feedback accordingly. The same may be said for

the ordering of meeting agendas and the time and structure of

performance review meetings.

While this paper has focused speciﬁcally on the implications

of the opposing domains hypothesis for leadership roles, we

believe that the distinction and antagonistic relationship between

analytical–mechanical reasoning and social reasoning exists in

many other areas in the organizational behavior domain. These

areas include, but are not limited to, leadership styles, conﬂict

management, trust, and moral reasoning. For example, distinc-

tions in the literature have been made between cognitive and

relational trust (Lewis and Weigert, 1985); cognitive conﬂict and

affective conﬂict (Jehn, 1997; De Dreu and Weingart, 2003);

empathy and dehumanization (Haslam, 2006).

Relevant to the leadership domain speciﬁcally, further test-

ing is needed in order to understand if individual difference

variables play a role in facilitating the ability to switch between

the two networks. Research is currently underway to address

this question by surfacing the opposing domains at the behav-

ioral level and linking individual difference variables and abilities

to the speed at which an individual is able to switch between

tasks requiring analytical and tasks requiring relational reason-

ing. Once we have more information on this we will be able

to target these variables in leadership development training pro-

grams. Additionally, manipulation of situational characteristics

within each type of task, for example, task difﬁculty and rou-

tineness for analytical tasks, and prior relationship quality for

social tasks, will allow us to isolate the key situational variables at

play. Finally, further research on the link between hor monal sys-

tems and neurological systems would allow us to understand how

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

tipping points in hormonal systems inﬂuence neurological tipping

points.

REFERENCES

Amodio, D. M., and Frith, C. D. (2006). Meeting of minds: the medial frontal

cortex and social cognition. Nat. Rev. Neurosci. 7, 268–277. doi: 10.1038/

nrn1884

Anderson, S. W., Bechara, A., Damasio, H., Tranel, D., and Damasio, A. (1999).

Impairment of social and moral behavior related to early damage in human

prefrontal cortex. Nat. Am. 2, 1032–1037.

Andrews-Hanna, J. R. (2012). The brain’s default network and its adaptive role

in internal mentation. Neuroscientist 18, 251–270. doi: 10.1177/10738584114

03316

Andrews-Hanna, J. R., Reidler, J. S., Sepulcre, J., Poulin, R., and Buckner, R. L.

(2010). Functional-anatomic fractionation of the brain’s default network. Neuron

65, 550–562. doi: 10.1016/j.neuron.2010.02.005

Anticevic, A., Cole, M. W., Murray, J. D., Corlett, P. R., Wang, X. J., and Krystal, J. H.

(2012). The role of default network deactivation in cognition and disease. Trends

Cogn. Sci. 16, 584–592. doi: 10.1016/j.tics.2012.10.008

Bagozzi, R. P., Verbeke, W. J., Dietvorst, R. C., Belschak, F. D., van den Berg, W.

E., and Rietdijk, W. J. (2013). Theor y of mind and empathic explanations of

Machiavellianism: a neuroscience perspective. J. Manag. 39, 1760–1798. doi:

10.1177/0149206312471393

Bales, R. F. (1958). “Task roles and social roles in problem-solving groups,” in

Readings in Social Psychology, eds E. Maccoby, T. Newcomb, and E. Hartley (New

York, NY: Rinehart & Winston), 437–447.

Bass, B. M. (1990). From transactional to transformational leadership: learning to

share the vision. Organ. Dyn. 18, 19–31. doi: 10.1016/0090-2616(90)90061-S

Bass, B. M., and Bass, R. (2008). The Bass Handbook of Leadership: Theory, Research,

and Managerial Applications. New York, NY: Simon and Schuster.

Bastian, B., and Haslam, N. (2011). Experiencing dehumanization: cognitive and

emotional effects of ever yday dehumanization. Basic Appl. Soc. Psychol. 33, 295–

303. doi: 10.1080/01973533.2011.614132

Beissner, F., Meissner, K., Bar, K., and Napadow, V. (2013). The autonomic brain:

an activation likelihood estimation meta-analysis for central processing of auto-

nomic function. J. Neurosci. 33, 10503–10511. doi: 10.1523/JNEUROSCI.1103-

13.2013

Bethlehem, R. A., van Honk, J., Auyeung, B., and Baron-Cohen, S.

(2013). Oxytocin, brain physiology, and functional connectivity: a review of

intranasal oxytocin fMRI studies. Psychoneuroendocrinology 38, 962–974. doi:

10.1016/j.psyneuen.2012.10.011

Biddle, B. J. (1986). Recent developments in role theory. Annu. Rev. Soc. 12, 67–92.

doi: 10.1146/annurev.so.12.080186.000435

Blake, R., and Mouton, J. (1964). The Managerial Grid. Houston, TX: Gulf.

Blau, P., and Scott, W. (1962). Formal Organizations. San Francisco, CA: Chandler.

Bono, J. E., and Ilies, R. (2006). Charisma, positive emotions and mood contagion.

Leadersh. Q. 17, 317–334. doi: 10.1016/j.leaqua.2006.04.008

Bowers, D. G., and Seashore, S. E. (1966). Predicting organizational effective-

ness with a four-factor theory of leadership. Adm. Sci. Q. 11, 238–263. doi:

10.2307/2391247

Boyatzis, R. E. (1982). The Competent Manager: A Model for Effective Performance.

New York, NY: John Wiley.

Boyatzis, R. E. (2006). Using tipping points of emotional intelligence and cognitive

competenciesto predictﬁnancial performance of leaders. Psicothema 18,124–131.

Boyatzis, R. E. (2013).“When pulling to the negative emotional att ractor is too much

or not enough to inspire and sustain outstanding leadership,” in The Fulﬁlling

Workplace: The Organization’s Role in Achieving Individual and Organizational

Health, eds. R. J. Burke and C. L. Cooper (Burlington, TV: Ashgate Publishing

Ltd), 139–150.

Boyatzis, R. E., and Goleman, D. (2007). Emotional and social competency

inventory. Boston: Hay Group.

Boyatzis, R. E., Passarelli, A. M., Koenig, K., Lowe, M., Mathew, B., Stoller, J.

K., et al. (2012). Examination of the neural substrates activated in memories of

experiences with resonant and dissonant leaders. Leadersh. Q. 23, 259–272. doi:

10.1016/j.leaqua.2011.08.003

Boyatzis, R. E., and Sala, F. (2004). “The emotional competence inventory (ECI),” in

Measuring Emotional Intelligence: Common Ground and Controversy, ed. G. Geher

(Hauppauge, NY: Nova Science Publishers), 147–180.

yatzis,

R. E., Smith, M. L., and Blaize, N. (2006). Developing sustainable leaders

through coaching and compassion. Acad. Manag. Learn. Educ. 5, 8–25. doi:

10.5465/AMLE.2006.20388381

Broyd, S. J., Demanuele, C., Debener, S., Helps, K., James, C. J., and

Sonuga-Barke, E. J. (2009). Default-mode brain dysfunction in mental dis-

orders: a systematic review. Neurosci. Biobehav. Rev. 33, 279–296. doi:

10.1016/j.neubiorev.2008.09.002

Buckner, R., and Carroll, D. (2007). Self-projection and the brain. Trends Cogn. Sci.

11, 49–57. doi: 10.1016/j.tics.2006.11.004

Buckner, R. L., Andrews-Hanna, J. R., and Schacter, D. L. (2008). The brain’s default

network. Ann. N. Y. Acad. Sci. 1124, 1–38. doi: 10.1196/annals.1440.011

Burke, W. W. (1965). Leadership behavior as a function of the leader, the follower,

and the situation. J. Pers. 33, 60–81. doi: 10.1111/j.1467-6494.1965.tb01372.x

Bzdok, D., Schilbach, L., Vogeley, K., Schneider, K., Laird, A. R., Langner, A., et al.

(2012). Parsing the neural correlates of moral cognition: ALE meta-analysis on

morality, theory of mind, and empathy. Brain Struct. Funct. 217, 783–796. doi:

10.1007/s00429-012-0380-y

Cartwright, D., and Zander, A. (1960). Group Dynamics: Research and Theory.

Evanston, IL: Row Peterson.

Chai, X. J., Castañón, A. N., Öngür, D., and Whitﬁeld-Gabrieli, S. (2012). Anticor-

relations in resting state networks without global signal regression. Neuroimage

59, 1420–1428. doi: 10.1016/j.neuroimage.2011.08.048

Chang, C., and Glover, G. H. (2009). Effects of model-based physiological noise cor-

rection on default mode network anti-correlations and correlations. Neuroimage

47, 1448–1459. doi: 10.1016/j.neuroimage.2009.05.012

Cohen, A. L., Fair, D. A., Dosenbach, N. U., Miezin, F. M., Dierker, D., Van

Essen, D. C., et al. (2008). Deﬁning functional areas in individual human

brains using resting functional connectivity MRI. Neuroimage 41, 45–57. doi:

10.1016/j.neuroimage.2008.01.066

Corbetta, M., Akbudak, E., Conturo, T. E., Snyder, A. Z., Ollinger, J. M., Drury, H.

A., et al. (1998). A common network of functional areas for attention and eye

movements. Neuron 21, 761–773. doi: 10.1016/S0896-6273(00)80593-0

Cowley, W. (1928). Three distinctions in the study of leaders. J. Abnorm. Soc. Psychol.

23, 144–157. doi: 10.1037/h0073661

Damasio, A. R. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain.

New York: Putnam.

Dasborough, M., Ashkanasy, N., Tee, E., and Tse, H. (2009). What goes around

comes around: how meso-level negative emotional contagion can ultimately

determine organizational attitudes toward leaders. Leadersh. Q. 20, 571–585.

doi: 10.1016/j.leaqua.2009.04.009

De Dreu, C., and Weingart, L. (2003). Task versus relationship conﬂict, team per-

formance and team member satisfaction: a meta-analysis. J. Appl. Psychol. 88,

741–749. doi: 10.1037/0021-9010.88.4.741

Duncan, J., and Owen, A. M. (2000). Common regions of the human frontal

lobe recruited by diverse cognitive demands. Trends Neurosci. 23, 475–483. doi:

10.1016/S0166-2236(00)01633-7

Eisenberger, N. I., and Cole, S. W. (2012). Social neuroscience and health: neuro-

physiological mechanisms linking social ties with physical health. Nat. Neurosci.

15, 669–674. doi: 10.1038/nn.3086

Fassbender, C., Zhang, H., Buzy, W. M., Cortes, C. R., Mizuiri, D., Beckett, L., et al.

(2009). A lack of default network suppression is linked to increased distractibility

in ADHD. Brain Res. 1273, 114–128. doi: 10.1016/j.brainres.2009.02.070

Fiedler, F. E. (1986). The contribution of cognitive resources and leader behav-

ior to organizational performance. J. Appl. Soc. Psychol. 16, 532–548. doi:

10.1111/j.1559-1816.1986.tb01157.x

Fiedler, M. E., and McGuire, M. A. (1987). Paper, Third Army Leadership Conference.

Kansas City, MO.

Fleishman, E. A. (1953). Leadership climate, human relations training, and

supervisory behavior. Pers. Psychol. 6, 205–222. doi: 10.1111/j.1744-6570.1953.

tb01040.x

Fornito, A., Harrison, B. J., Zalesky, A., and Simons, J. S. (2012). Com-

petitive and cooperative dynamics of large-scale brain functional networks

supporting recollection. Proc. Natl. Acad. Sci. U.S.A. 109, 12788–12793. doi:

10.1073/pnas.1204185109

Fox, M. D., Corbetta, M., Snyder, A. Z., Vincent, J. L., and Raichle, M. E.

(2006).

Spontaneous neur

onal activ ity distinguishes human dorsal and ventral

attention systems. Proc. Natl. Acad. Sci. U.S.A. 103, 10046–10051. doi:

10.1073/pnas.0604187103

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., and

Raichle, M. E. (2005). The human brain is intrinsically organized into dynamic,

anticorrelated functional networks. Proc. Natl. Acad. Sci. U.S.A. 102, 9673–9678.

doi: 10.1073/pnas.0504136102

Fransson, P. (2005). Spontaneous low-frequency BOLD signal ﬂuctuations: an fMRI

investigation of the resting-state default mode of brain function hypothesis. Hum.

Brain Mapp. 26, 15–29. doi: 10.1002/hbm.20113

Fredrickson, B. L., and Branigan, C. (2005). Positive emotions broaden the scope

of attention and thought-action repertoires. Cogn. Emot. 19, 313–332. doi:

10.1080/02699930441000238

French, S. E., and Jack, A. I. (in press). “Dehumanizing the enemy: the intersection

of neuroethics and military ethics,” in The Responsibility to Protect: Alternative

Perspectives, ed. D. Whetham (Martinus Nijhoff), due in pr int April 2014.

Gibson, B., and Sanbonmatsu, D. M. (2004). Optimism, pessimism, and gam-

bling: the downside of optimism. Pers. Soc. Psychol. Bull. 30, 149–160. doi:

10.1177/0146167203259929

Golland, Y., Bentin, S., Gelbard, H., Benjamini, Y., Heller, R., Nir, Y., et al.

(2007). Extrinsic and intrinsic systems in the posterior cortex of the human

brain revealed during natural sensory stimulation. Cereb. Cortex 17, 766–777.

doi: 10.1093/cercor/bhk030

Gottman, J. M., Murray, J. D., Swanson, C. C., Tyson, R., and Swanson, K. R. (2002).

The Mathematics of Marriage: Dynamic Non-linear Models. Cambridge, MA: MIT

Press.

Grabenhorst, F., and Rolls, E. T. (2011). Value, pleasure and choice in the ventral

prefrontal cortex. Trends Cogn. Sci. 15, 56–67. doi: 10.1016/j.tics.2010.12.004

Graham, S., Jiang, J., Manning, V., Nejad, A. B., Zhisheng, K., Salleh, S. R., et al.

(2010). IQ-related fMRI differences during cognitive set shifting. Cereb. Cortex

20, 641–649. doi: 10.1093/cercor/bhp130

Greene, J. D., Nystrom, L. E., Engell, A. D., Darley, J. M., and Cohen, J. D. (2004).

The neural bases of cognitive conﬂict and control in moral judgment. Neuron 44,

389–400. doi: 10.1016/j.neuron.2004.09.027

Greicius, M. D., Krasnow, B., Reiss, A. L., and Menon, V. (2003). Functional con-

nectivity in the resting brain: a network analysis of the default mode hypothesis.

Proc. Natl. Acad. Sci. U.S.A. 100, 253–258. doi: 10.1073/pnas.0135058100

Gusnard, D. A., Akbudak, E., Shulman, G. L., and Raichle, M. E. (2001).

Medial prefrontal cortex and self-referential mental activity: relation to a default

mode of brain function. Proc. Natl. Acad. Sci. U.S.A. 98, 4259–4264. doi:

10.1073/pnas.071043098

Halpin, A., and Winer, B. (1957). “A factorial study of the leader behavior descrip-

tions,” in Leader Behavior: Its Description and Measurement, eds R. Stodgill and

A. Coons (Columbus, OH: Bureau of Business Research, Ohio State University),

39–51.

Haslam, N. (2006). Dehumanization: an integrative review. Pers. Soc. Psychol. Rev.

10, 252–264. doi: 10.1207/s15327957pspr1003_4

Hatﬁeld, E., Cacioppo, J., and Rapson, R. (1994). Emotional Contagion.NewYork,

NY: Cambridge University Press.

Hemphill, J. K. (1950). Relations between the size of the group and the behavior

of “superior” leaders. J. Soc. Psychol. 32, 11–22. doi: 10.1080/00224545.1950.

9919026

Henson, R. (2006). Forward inference using functional neuroimaging: dissociations

versus associations. Trends Cogn. Sci. 10, 64–69. doi: 10.1016/j.tics.2005.12.005

Hersey, P., and Blanchard, K. (1969). Management of Organizational Be havior:

Utilizing Human Resources. Englewood Cliffs, NJ: Prentice-Hall.

Iacoboni, M., Lieberman, M. D., Knowlton, B. J., Molnar-Szakacs, I., Moritz, M.,

Throop, C. J., et al. (2004). Watching social interactions produces dorsome-

dial prefrontal and medial parietal BOLD fMRI signal increases compared to

a resting baseline. Neuroimage 21, 1167–1173. doi: 10.1016/j.neuroimage.2003.

11.013

Indvik, J. (1986). Path-goal theory of leadership: a meta-analysis. Acad. Manag.

Proc. 1986, 189–192. doi: 10.5465/AMBPP.1986.4980581

Insel, T. R. (1997). A neurobiological basis of social attachment. Am. J. Psychiatry

154, 726–735.

Jack, A., Boyatzis, R. E., Khawaja, M., Passarelli, A. M., and Leckie, R. (2013a).

Visioning in the brain: an fMRI study of inspirational coaching and mentoring.

Soc. Neurosci. 8, 369–384. doi: 10.1080/17470919.2013.808259

Jack, A. I., Dawson, A. J., and Norr, M. (2013b). Seeing human: distinct and overlap-

ping neural signatures associated with two forms of dehumanization. Neuroimage

79, 313–328. doi: 10.1016/j.neuroimage.2013.04.109

ack,

A. I. (in press). “A scientiﬁc case for conceptual dualism: the problem

of consciousness and the opposing domains hypothesis,” in Oxford Studies in

Experime ntal Philosophy, Vol. 1, eds J. Knobe and S. Nichols (Oxford: Oxford

University Press).

Jack, A. I., Dawson, A., Begany, K., Leckie, R. L., Barry, K., Ciccia, A., et al.

(2012). fMRI reveals reciprocal inhibition between social and physical cog-

nitive domains. Neuroimage 66C, 385–401. doi: 10.1016/j.neuroimage.2012.

10.061

Jack, A. I., Robbins, P. A., Friedman, J. P., and Meyers, C. D. (in press). “More than a

feeling: counterintuitive effects of compassion on moral judgment,” in Advances

in Experimental Philosophy of Mind, eds J. Sytsma (Continuum).

Jehn, K. (1997). “Affective and cognitive conﬂict in work groups: increasing

performance through value-based intragroup conﬂict,” in Using Conﬂict in

Organizations, eds C. De Dreu and E. Van de Vliert (London: Sage), 87–100.

Johnson, S. K. (2008). I second that emotion: effects of emotional contagion and

affect at work on leader and follower outcomes. Leadersh. Q. 19, 1–19. doi:

10.1016/j.leaqua.2007.12.001

Josipovic, Z., Dinstein, I., Weber, J., and Heeger, D. J. (2012). Inﬂuence of medita-

tion on anti-correlated networks in the brain. Front. Hum. Neurosci. 5:183. doi:

10.3389/fnhum.2011.00183

Katz, D., Maccoby, N., and Morse, N. C. (1950). Productivity, Supervision, and

Morale in an Ofﬁce Situation. Oxford: Institute for Social Research, University of

Michigan.

Koenigs, M., Kruepke, M., Zeier, J., and Newman, J. P. (2012). Utilitarian

moral judgment in psychopathy. Soc. Cogn. Affect. Neurosci. 7, 708–714. doi:

10.1093/scan/nsr048

Koenigs, M., Young, L., Adolphs, R., Tranel, D., Cushman, F., Hauser, M., et al.

(2007). Damage to the prefrontal cortex increases utilitarian moral judgements.

Nature 446, 908–911. doi: 10.1038/nature05631

Kubit, B., and Jack, A. I. (2013). Rethinking the role of the rTPJ in

attention and social cognition in light of the opposing domains hypoth-

esis: ﬁndings from an ALE-based meta-analysis and resting-state func-

tional connectivity. Front. Hum. Neurosci. 7:323. doi: 10.3389/fnhum.

2013.00323

Laird, A. R., Eickhoff, S. B., Rottschy, C., Bzdok, D., Ray, K. L., and Fox,

P. T. (2013). Networks of task co-activations. Neuroimage 80, 505–514. doi:

10.1016/j.neuroimage.2013.04.073

LeMay, R. A. (1999). “Roles, identities, and expectancies: positive contributions to

normalization and social role valorization,”in A Quar ter-century of Normalization

and Social Role Valorization: Evolution and Impact, eds R. J. Flynn and R. A. Lemay

(Ottawa, ON: University of Ottawa Press), 219–240.

Lewis, J. D., and Weigert, A. (1985). Trust as a social reality. Soc. Forces 63, 967–985.

Lieberman, M. D. (2013). Social: Why are Brains are Wired to Connect.NewYork,

NY: Crown Publishers.

Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., and Barrett, L. F. (2012).

The brain basis of emotion: a meta-analytic review. Behav. Brain Sci. 35, 121–143.

doi: 10.1017/S0140525X11000446

Mantini, D., Gerits, A., Nelissen, K., Durand, J. B., Joly, O., Simone, L., et al. (2011).

Default mode of brain function in monkeys. J. Neurosci. 31, 12954–12962. doi:

10.1523/JNEUROSCI.2318-11.2011

Mars, R. B., Neubert, F. X., Noonan, M. P., Sallet, J., Toni, I., and Rushworth, M. F.

(2012). On the relationship between the “default mode network” and the “social

brain”. Front. Hum. Neurosci. 6:189: doi: 10.3389/fnhum.2012.00189.

Mason, M. F., Norton, M. I., Van Horn, J. D., Wegner, D. M., Grafton, S. T., and

Macrae, C. N. (2007). Wandering minds: the default network and stimulus-

independent thought. Science 315, 393–395. doi: 10.1126/science.1131295

McClelland, D. (1961). The Achieving Society. Princeton, NJ: Van Nostrand. doi:

10.1037/14359-000

McGregor, D. (1960). The Human Side of Enterprise.NewYork,NY:McGrawHill.

McKiernan, K. A., Kaufman, J. N., Kucera-Thompson, J., and Binder, J. R.

(2003). A parametric manipulation of factors affecting task-induced deac-

tivation in functional neuroimaging. J. Cogn. Neurosci. 15, 394–408. doi:

10.1162/089892903321593117

Mead, G. H. (1934). Mind, Self, and Society. Chicago: University of Chicago Press.

Meyer, M. L., Spunt, R. P., Berkman, E. T., Taylor, S. E., and Lieberman, M. D.

(2012).

idence for social working memory from a par ametric functional MRI

study. Proc. Natl. Acad. Sci. U.S.A. 109, 1883–1888. doi: 10.1073/pnas.11210

77109

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

Boyatzis et al. Leadership roles and opposing domains

Mitchell, J. P., Macrae, N. C., and Banaji, M. R. (2006). Dissociable medial prefrontal

contributions to judgments of similar and dissimilar others. Neuron 50, 655–663.

doi: 10.1016/j.neuron.2006.03.040

Ochsner, K. N., Beer, J. S., Robertson, E. R., Cooper, J. C., Gabrieli, J. D. E., Kihsltrom,

J. F., et al. (2005). The neural correlates of direct and reﬂected self-knowledge.

Neuroimage 28, 797–814. doi: 10.1016/j.neuroimage.2005.06.069

Owen, A. M., McMillan, K. M., Laird, A. R., and Bullmore, E. (2005). N-back work-

ing memory paradigm: a meta-analysis of normative functional neuroimaging

studies. Hum. Brain Mapp. 25, 46–59. doi: 10.1002/hbm.20131

Poldrack, R. A. (2011). Inferring mental states from neuroimaging data:

from reverse inference to large-scale decoding. Neuron 72, 692–697. doi:

10.1016/j.neuron.2011.11.001

Pyka, M., Beckmann, C. F., Schöning, S., Hauke, S., Heider, D., Kugel, H., et al.

(2009). Impact of working memory load on fMRI resting state pattern in subse-

quent resting phases. PLoS ONE 4:e7198. doi: 10.1371/journal.pone.0007198.

Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A., and

Shulman, G. L. (2001). A default mode of brain function. Proc. Natl. Acad. Sci.

U.S.A. 98, 676–682. doi: 10.1073/pnas.98.2.676

Raichle, M. E., and, Snyder, A. Z. (2007). A default mode of brain func-

tion: a brief history of an evolving idea. Neuroimage 37, 1083–1090. doi:

10.1016/j.neuroimage.2007.02.041

Roy, M., Shohamy, D., and Wager, T. D. (2012). Ventromedial prefrontal-subcortical

systems and the generation of affective meaning. Trends Cogn. Sci. 16, 147–156.

doi: 10.1016/j.tics.2012.01.005

Schilbach, L., Eickhoff, B., Rotarska-Jagiela, A., Fink, G. R., and Vogeley, K. (2008).

Minds at rest? Social cognition as the default mode of cognizing and its putative

relationship to the “default system” of the brain. Conscious. Cogn. 17, 457–467.

doi: 10.1016/j.concog.2008.03.013

Schulkin, J. (1999). The Neuroendocrine Regulation of Behavior. Cambridge:

Cambridge University Press.

Shartle, C. (1950). “Studies of leadership by interdisciplinary methods,” in Leader-

ship in American Education, ed. A. G. Grace (Chicago, IL: University of Chicago

Press).

Shulman, G. L., Corbetta, M., Buckner, R. L., Fiez, J. A., Miezin, F. M., Raichle, M.

E., et al. (1997a). Common blood ﬂow changes across visual tasks: I. Increases

in subcortical structures and cerebellum but not in nonvisual cortex. J. Cogn.

Neurosci. 9, 624–647. doi: 10.1162/jocn.1997.9.5.624

Shulman, G. L., Corbetta, M., Buckner, R. L., Fiez, J. A., Miezin, F. M., Raichle, M.

E., et al. (1997b). Common blood ﬂow changes across visual tasks: II. Decreases

in cerebral cortex. J. Cogn. Neurosci. 9, 648–663. doi: 10.1162/jocn.1997.9.5.648

Slatter, W. L. (1955). Why not a leadership training program? J. Dairy Sci. 38,

1408–1409. doi: 10.3168/jds.S0022-0302(55)95131-6

Spreng, R. N., and Grady, C. L. (2010). Patterns of brain activity supporting

autobiographical memory, prospection, and theory of mind, and their rela-

tionship to the default mode network. J. Cogn. Neurosci. 22, 1112–1123. doi:

10.1162/jocn.2009.21282

Spreng, R. N., Mar, R. A., and Kim, A. S. (2009). The common neural basis of

autobiographical memory, prospection, navigation, theory of mind, and the

default mode: a quantitative meta-analysis. J. Cogn. Neurosci. 21, 489–510. doi:

10.1162/jocn.2008.21029

Subramaniam, K., Kounios, J., Parrish, T. B., and Jung-Beeman, M. (2009). A brain

mechanism for facilitation of insight by positive affect. J. Cogn. Neurosci. 21,

415–432. doi: 10.1162/jocn.2009.21057

Sy, T., Côté, S., and Saavedra, R. (2005). The contagious leader: impact of the leader’s

mood on the mood of group members, group affective tone, and group processes.

J. Appl. Psychol. 90, 295–305. doi: 10.1037/0021-9010.90.2.295

Takeuchi, H., Taki, Y., Hashizume, H., Sassa, Y., Nagase, T., Nouchi, R.,

et al. (2011). Failing to deactivate: the association between brain activity dur-

ing a working memory task and creativity. Neuroimage 55, 681–687. doi:

10.1016/j.neuroimage.2010.11.052

Tannenbaum, R., and Schmidt, W. (1958). How to choose a leadership pattern.

Harv. Bus. Rev. 36, 95–101.

Tian, L., Jiang, T., Liu, Y., Yu, C., Wang, K., Zhou, Y., et al. (2007). The relationship

within and between the extrinsic and intrinsic systems indicated by resting

state correlational patterns of sensory cortices. Neuroimage 36, 684–690. doi:

10.1016/j.neuroimage.2007.03.044

Uddin, L. Q., Kelly, A. M., Biswal, B. B., Xavier Castellanos, F., and Milham, M.

P. (2009). Functional connectivity of default mode network components: cor-

relation, anticorrelation, and causality. Hum. Brain Mapp. 30, 625–637. doi:

10.1002/hbm.20531

an D

ijk, K. R., Hedden, T., Venkataraman, A., Evans, K. C., Lazar, S. W., and

Buckner, R. L. (2010). Intrinsic functional connectivity as a tool for human

connectomics: theory, properties, and optimization. J. Neurophysiol. 103, 297–

321. doi: 10.1152/jn.00783.2009

Vanhaudenhuyse, A., Noirhomme, Q., Tshibanda, L. J. F., Bruno, M. A., Boveroux,

P., Schnakers, C., et al. (2010). Default network connectivity reﬂects the level of

consciousness in non-communicative brain-damaged patients. Brain 133, 161–

171. doi: 10.1093/brain/awp313

Van Overwalle, F. (2009). Social cognition and the brain: a meta-analysis. Hum.

Brain Mapp. 30, 829–858. doi: 10.1002/hbm.20547

Van Overwalle, F. (2011). A dissociation between social mentalizing and general

reasoning. Neuroimage 54, 1589–1599. doi: 10.1016/j.neuroimage.2010.09.043

Van Overwalle, F., and Baetens, K. (2009). Understanding others’ actions and goals

by mirror and mentalizing systems: a meta-analysis. Neuroimage 48, 564–584.

doi: 10.1016/j.neuroimage.2009.06.009

Vincent, J. L., Kahn, I., Snyder, A. Z., Raichle, M. E., and Buckner, R. L.

(2008). Evidence for a frontoparietal control system revealed by intrinsic

functional connectivity. J. Neurophysiol. 100, 3328–3342. doi: 10.1152/jn.903

55.2008

Waytz, A., and Mason, M. F. (2013). Your brain at work. Harv. Bus. Rev. 91, 102.

Weed, S. E., Mitchell, T. R., and Mofﬁtt, W. (1976). Leadership style, subordinate

personality, and task type as predictors of performance and satisfaction with

supervision. J. Appl. Psychol. 61, 58–66. doi: 10.1037/0021-9010.61.1.58

Whitﬁeld-Gabrieli, S., and Ford, J. M. (2012). Default mode network activity

and connectivity in psychopathology. Annu. Rev. Clin. Psychol. 8, 49–76. doi:

10.1146/annurev-clinpsy-032511-143049

Wiggett, A. J., Pritchard, I. C., and Downing, P. E. (2009). Animate and inanimate

objects in human visual cortex: evidence for task-independent category effects.

Neuropsychologia 47, 3111–3117. doi: 10.1016/j.neuropsychologia.2009.07.008

Wofford, J. C. (1970). Factor analysis of managerial behavior variables. J. Appl.

Psychol. 54, 169–173. doi: 10.1037/h0029058

Yeo, B. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead,

M., et al. (2011). The organization of the human cerebral cortex estimated

by intrinsic functional connectivity. J. Neurophysiol. 106, 1125–1165. doi:

10.1152/jn.00338.2011

Yukl, G. (1999). An evaluative essay on current conceptions of effective leadership.

Eur. J. Work Organ. Psychol. 8, 33–48. doi: 10.1080/135943299398429

Yukl, G. (2006). Leadership in Organizations, 6th Edn. Upper Saddle River, NJ:

Prentice-Hall.

Yukl, G. (2008). How leaders inﬂuence organizational effectiveness. Leadersh. Q. 19,

708–722. doi: 10.1016/j.leaqua.2008.09.008

Conflict of Interest Statement: The authors declare that the research was conducted

in the absence of any commercial or ﬁnancial relationships that could be construed

as a potential conﬂict of interest.

Received: 20 December 2013; paper pending published: 23 January 2014; accepted: 17

February 2014; published online: 04 March 2014.

Citation: Boyatzis RE, Rochford K and Jack AI (2014) Antagonistic neural net-

works underlying differentiated leadership roles. Front. Hum. Neurosci. 8:114. doi:

10.3389/fnhum.2014.00114

This ar ticle was submitted to the journal Frontiers in Human Neuroscience.

under the terms of the Creative Commons Attribution License (CC BY). The use, dis-

tribution or reproduction in other forums is permitted, provided the original author(s)

or licensor are credited and that the orig inal publication in this journal is cited, in

accordance with accepted academic practice. No use, distribution or reproduction is

permitted which does not comply w ith these terms.

Frontiers in Human Neuroscience www.frontiersin.org March 2014

Volume 8

Article 114

HUMAN NEUROSCIENCE

PERSPECTIVE ARTICLE

published: 12 February 2015

doi: 10.3389/fnhum.2014.01073

Redeﬁning neuromarketing as an integrated science of

inﬂuence

Hans C. Breiter

1,2,3

†

, Martin Block

3,4†

, Anne J. Blood

2,3†

, Bobby Calder

3,5†

, Laura Chamberlain

3,6†

Nick Lee

3,7†

, Sherri Livengood

1,3†

, Frank J. Mulhern

3,4†

, Kalyan Raman

1,3,4†

, Don Schultz

3,4†

Daniel B. Stern

1,3†

, Vijay Viswanathan

3,4†

and Fengqing (Zoe) Zhang

3,8,9†

Warren Wright Adolescent Center, Department of Psychiatry and Behavioral Science, Northwestern University Feinberg School of Medicine, Chicago, IL, USA

Mood and Motor Control Laboratory or Laboratory of Neuroimaging and Genetics, Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA

Applied Neuromarketing Consortium, Medill, Kellogg, and Feinberg Schools, Northwestern University, Evanston, IL, USA

Medill Integrated Marketing Communications, Northwestern University, Evanston, IL, USA

Department of Marketing, Kellogg School of Management, Northwestern University, Evanston, IL, USA

Aston Business School, Birmingham, UK

School of Business and Economics, Loughborough University, Leicestershire, UK

Department of Statistics, Northwestern University, Evanston, IL, USA

Department of Psychology, Drexel University, Philadelphia, PA, USA

Edited by:

Sven Braeutigam, University of

Oxford, UK

Reviewed by:

Arun Bokde, Trinity College Dublin,

Ireland

Giovanni Vecchiato, Sapienza

University of Rome, Italy

*Correspondence:

Hans C. Breiter, Warren Wright

Adolescent Center, Department of

Psychiatry and Behavioral Science,

Northwestern University Feinberg

School of Medicine, 710 N. Lake

Shore Dr., Abbott Hall 1301,

Chicago, IL 60611, USA

e-mail: h-breiter@northwestern.edu

†

These authors have contributed

equally to this work.

Multiple transformative forces target marketing, many of which derive from new

technologies that allow us to sample thinking in real time (i.e., brain imaging), or to

look at large aggregations of decisions (i.e., big data). There has been an inclination to

refer to the intersection of these technologies with the general topic of marketing as

“neuromarketing”. There has not been a serious effort to frame neuromarketing, which

is the goal of this paper. Neuromarketing can be compared to neuroeconomics, wherein

neuroeconomics is generally focused on how individuals make “choices”, and represent

distributions of choices. Neuromarketing, in contrast, focuses on how a distribution of

choices can be shifted or “inﬂuenced”, which can occur at multiple “scales” of behavior

(e.g., individual, group, or market/society). Given inﬂuence can affect choice through

many cognitive modalities, and not just that of valuation of choice options, a science of

inﬂuence also implies a need to develop a model of cognitive function integrating attention,

memory, and reward/aversion function. The paper concludes with a brief description of

three domains of neuromarketing application for studying inﬂuence, and their caveats.

Keywords: neuromarketing, neuroeconomics, marketing communications, neuroimaging, scaling, inﬂuence, choice

INTRODUCTION

Marketing has been dominated for over a century by models

that assume a rational process of persuasion, which follows a

sequence from awareness through purchase that consumers can

consciously articulate. While this approach ﬁts with traditional

research methodologies, it hasn’t always explained or predicted

purchase behavior. Recent developments suggest that a new per-

spective may be emerging. In particular, marketers have sought

to integrate ideas about non-rational and rational processes

(Kahneman, 2011), and ideas related to social neuroscience vs.

individual decision-making (Lee et al., 2007; Senior and Lee,

2008), as well as using methods and technologies aligned with

neuroscience (Ioannides et al., 2000; Braeutigam, 2005; Vecchiato

et al., 2011; Plassmann et al., 2012). Some have been quick to

label—not always in a complimentary manner—such develop-

ments as “neuromarketing” (e.g., Laybourne and Lewis, 2005

See Brain scam? (2004). Nat. Neurosci. 7, 683. doi:10.1038/nn0704-683;

and Neuromarketing: beyond branding (2004). The Lancet Neurol. 3, 71.

doi:10.1016/S1474-4422(03)00643-4.

To date, neuromarketing has lacked a solid theoretical

framework. As such, the term “neuromarketing” itself runs

the risk of confusing more fundamental scientiﬁc research

with commercial applications (Lee et al., 2007; Javor et al.,

2013). In this paper, we seek to extend existing work (e.g.,

Fugate, 2007; Hubert and Kenning, 2008; Senior and Lee, 2008;

Wilson et al., 2008; Fisher et al., 2010), to clarify a frame-

work for neuromarketing as an integrated science of inﬂuence.

We start by contrasting neuroeconomics to neuromarketing.

We then consider the concept of inﬂuence across individuals,

groups, communities and markets, along with its dependency

on an integrated model of mental function, along with some

key—often unrecognized—caveats that must be considered by

neuromarketing researchers.

INFLUENCE VS. CHOICE

It is helpful to compare neuromarketing to neuroeconomics,

with which it may appear to overlap. Neuroeconomics tends

to focus on individual and group choice, or judgment and

decision-making in the context of consumables or markets

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

FIGURE 1 | (A) Neuroeconomics focuses on the model of choice, which is

centered on how we assess reward/aversion. This ﬂow diagram shows four

steps involved in making a choice. For the second step, there are several

theories that have been proposed to model valuation of choices. Matching

theory and alliesthesia (hedonic deﬁcit theory) are two theories heavily

used in neuroscience. Prospect and portfolio theory are used in economics.

All four theories have been used in neuroeconomics. New to the set of

valuation theories is relative preference theory (RPT) that is the only

valuation theory meeting Feynman criteria for lawfulness, using an

information variable, or actually scaling from group to individual behavior.

Because of this scaling across group and individual behavior, and the fact it

can be framed as a power law, RPT actually encodes the fundamental

features of the other four theories, and can be used to ground them or even

derive them. (B) In contrast to economics and neuroeconomics with their

focus on choice, marketing is focused on “inﬂuence”, which looks at how

distributions of choice behavior can be shifted or altered. This diagram

sketches one potential model for the effect of inﬂuence on behavior.

Inﬂuence can be considered the difference in gradients for preference

inside a person (or organism) and outside a person. These gradients of

preference might be schematized by RPT. They would be ﬁltered and

processed by valuation functions mentioned in panel (A), which include

alliesthesia or hedonic deﬁcit theory regarding what is in deﬁcit for an

individual, along with matching, prospect theory, and the variance mean

approach to portfolio theory. This processing would facilitate integration of

the gradient inputs and determine what goal-objects or events become the

focus of behavior, along with providing the intensity for it. Other cognitive

functions such as memory are critical to this processing and evaluation of

relative costs/beneﬁts to prospective behavior; together they give behavior

its direction and intensity. Behavior, in turn, feeds back onto these internal

and external gradients of preference as experienced utility of expressed

behavior.

(Figure 1A; Camerer, 2008). This focus on choice is dis-

tinct from the focus in neuromarketing on the issue of how

individuals and groups might be shifted from one pattern of

decisions to another pattern, or to change their distribution of

choices.

Much neuromarketing research to this point has been focused

on optimal methods to shift the distribution of choices (e.g.,

Ambler et al., 2004; McClure et al., 2004; Ohme et al., 2009; Santos

et al., 2011). The use of “neuro” as a preﬁx has thus followed a

similar rationale to that of neuroeconomics, whereby study of the

neural processes provides a tool for describing behavioral change

that was not available by the study of behavior alone (Ariely and

Berns, 2010).

Such a view of neuromarketing ignores the broader perspective

on what might be called “inﬂuence”, which is the primary issue

involved with marketing, advertising, engineering design, teach-

ing, or behavioral change in medicine. All of these categories of

“inﬂuence” focus on how to get people to engage in a behavior

preferred by a corporation, government, trade-group, culture or

other entity. From an ethical perspective, discussions regarding

consumer rights, for example privacy, are key when considering

the inﬂuencing of behavior by interest groups, and neuromarket-

ing research has been a subject of some interest in this regard (e.g.,

Murphy et al., 2008; Wilson et al., 2008). Nonetheless, inﬂuence

doesn’t just shift the distribution of choices to one favored by

the interest group in question, but balances between internal

and external forces on behavior. Inﬂuence might be considered

a balance between gradients of preference within an individual

or group that inﬂuence events outside of them, and gradients of

preference outside the individual or group that inﬂuence events

inside of them.

Such gradients of preference could be schematized by patterns

of approach and avoidance decisions (i.e., the distribution of

choice) as described by relative preference theory (RPT; Breiter

and Kim, 2008; Kim et al., 2010), an empirically-based account

of reward/aversion resembling prospect theory (Kahneman and

Tversky, 1979; Breiter et al., 2001) but grounded in informa-

tion theory (Shannon and Weaver, 1949) to account for pat-

terns in decisions that can be connected to reward/aversion

circuitry and genetic polymorphisms (e.g., Perlis et al., 2008;

Gasic et al., 2009). Using RPT, internal and external gradients

of preference would involve variables quantifying (a) the pat-

tern of approach decisions; and (b) the pattern of avoidance

decisions. In the case of internal preferences, these would char-

acterize the individual, whereas in the case of external prefer-

ences these might characterize a group of people external to

the individual (or a preference gradient from just one other

external person). RPT allows individual and group preferences

to be readily characterized in a quantitative, lawful fashion

that scales between individual and group. The integration of

internal (e.g., individual) and external (e.g., group) gradients

of preference would then be given direction in distinct deci-

sion/planning/problem solving situations by the processes brieﬂy

schematized in Figures 1A,B. Gradients of preference given direc-

tion by hedonic deﬁcit theory (i.e., alliesthesia; Cabanac, 1971;

Paulus, 2007) and other valuation processes (necessary for incor-

porating probabilities related to goal-objects, Kahneman and

Tversky, 1979; relative valuations across goal-objects, Herrnstein,

1961; and variance in valuation, Markowitz, 1952) would consti-

tute the combined intrinsic and extrinsic motivation described

by Deci and Ryan (1985), leading to behavior, which in turn

feeds back into gradients of preferences based on the experi-

enced utility in individuals involved (see Kahneman et al., 1997).

Such a schema is shown in Figure 1B as one of many pos-

sibilities for how internal and external gradients are balanced

through their effects on behavior, and can shift distributions of

choice.

In considering the balance between internal and external pref-

erence, cognitive processes thought to be separate from that of the

valuation of options come into play, such as perception, attention,

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

FIGURE 2 | (A) This schema describes an engineering-based behavioral

science (EBS) model of psychological domains that can be integrated in

accordance with existing non- engineering-based models of emotion. Unlike

other frameworks, EBS evaluates mathematical, law-like relationships

between cognitive domains such as (i) reward/aversion processing, (ii)

attention, (iii) memory, and (iv) perception, rather than associative

relationships based purely on statistics. There are a number of modern

theoretical constructs for emotion, including two examples shown from work

by (B) Barrett and (C) Gross, and they tend to include the components we

suggest integrating through EBS. (D) Individuals, groups, and/or

societies/markets can exert “inﬂuence” to shift distributions of choice

behavior in others. This expression of “inﬂuence” can be exerted within scale

and across scale (e.g., by a group on an individual). Neuromarketing uses the

valuation aspect of neuroeconomics (i.e., reward/aversion processing) and

tries to integrate it with other behavioral science and neuroscience

constructs, such as attention, memory and perception, which are all

components of the EBS model. It tries to do this at the level of the individual,

the group, and society, which are each different scales of measure.

and memory (Ioannides et al., 2000; Ariely and Berns, 2010). At

this time, no theoretical schema and little empirical data exist for

how these theoretically independent cognitive processes interact,

but cognitive processes for perception of outside stimuli exerting

inﬂuence, attention to their features, and memory for comparison

of such features to prior percepts are necessary operations for

processing “inﬂuence”. Exerting “inﬂuence” to change another’s

behavior, or being the subject of outside “inﬂuence” to change

your own behavior thus need to be considered in a much broader

construct of mental operations (see Figure 2A). One must also

recognize that this ensemble of operations (i.e., perception, atten-

tion, memory, reward/aversion processing) have been extensively

theorized to be core processes for emotion (Breiter and Gasic,

2004; Barrett, 2006; Gross, 2009, 2013; Kuppens et al., 2013;

Lindquist et al., 2013). Speciﬁc examples of this are schematized

in Figures 2B,C for the models of Barrett (2006) and Gross (2009,

2013).

The balance between internal and external forces on behav-

ior (e.g., respectively, internal emotional experience (or inter-

nal preference gradient) vs. emotional expression by entities

outside the individual (or external preference gradient)) must

also be apparent at the neural level of measurement, given

that “brain and mind are one”, a fundamental hypothesis of

neuroscience (e.g., Breiter and Rosen, 1999; Breiter and Gasic,

2004; Breiter et al., 2006). This view of neuromarketing thus has

as its focus an understanding of the balance between internal

and external preferences (emotional experiences), on individ-

uals and groups. Neural measures of one individual or inter-

acting individuals (e.g., as with joint trust games; King-Casas

et al., 2005; Tomlin et al., 2006) can be made in parallel

with behavioral ones to conﬁrm that observations made at the

behavioral level affect those at other levels of spatiotemporal orga-

nization, or actually scale across levels of spatiotemporal orga-

nization (i.e., group behavior, individual behavior, distributed

neural groups, neural group, etc.). Inﬂuence can thus be thought

of as being present across multiple spatiotemporal levels of

measurement, from group measures to measures of individual

behavior to neural groups, etc. The issue of scaling might be

considered as a “layering of inﬂuence” and warrants further

discussion.

LAYERS OF INFLUENCE AND COMMUNITIES AFFECTED

Scaling is rarely discussed in experimental psychology and other

behavioral disciplines, and was not formally introduced into

behavioral science and neuroscience until the 1990s by Sutton

and Breiter (1994). In its adaptation to biology and behavior,

scaling refers to how measures made at one level of spatiotemporal

organization, relate in a principled, lawful manner to measures at

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

other spatiotemporal levels of organization (Sutton and Breiter,

1994; Perelson et al., 2006; Savage and West, 2006). This rela-

tionship does not represent a statistical one where a certain

amount of the variance at one level of measure can predict the

variance at another level, or how some information at one layer of

organization can specify some extent of information at another

layer (Adami, 2004; Szostak, 2004). Instead, it is causal (i.e.,

mechanistic) in that the same patterns of behavior measured at

one layer are also measured at a neighboring level, and there is

a necessary relationship (Sutton and Breiter, 1994; Breiter and

Gasic, 2004; Breiter et al., 2006). Given that it has not been a

major topic in behavior science or neuroscience, few biological

measures have yet been shown to scale. One behavior that does

show scaling is that of circadian rhythms, which show measures

that scale from behavior to distributed groups of neurons to

individual neural groups to cells and molecular biology. The other

is approach/avoidance behavior described by RPT, scaling from

group behavior to individual behavior, and potentially to other

scales (Breiter and Kim, 2008; Kim et al., 2010). To date, few

behavioral constructs outside of RPT have been tested to Feynman

criteria for lawfulness, which includes scaling (Feynman et al.,

1963).

Scaling has become an important metaphor/analogy in con-

sidering the statistical association of measures made at one

spatiotemporal scale vs. another, as with the Research Domain

Criteria project (RDoC) sponsored out of the National Institutes

of Health (NIMH; Insel et al., 2010; Morris and Cuthbert, 2012;

Cuthbert and Insel, 2013). The RDoC project and projects spon-

sored out of the National Institutes of Health Connectome Project

(Van Essen et al., 2012; Barch et al., 2013) focus on measures at

different spatiotemporal scales that can predict some degree of

variance in each other. Both the RDoC and Connectome projects

are directly modeled after the Phenotype Genotype Project in

Addiction and Mood Disorders (PGP),

which successfully dis-

covered measures that scale across levels (i.e., RPT; Breiter and

Kim, 2008; Kim et al., 2010).

While scaling is a sine qua non of classical science across lev-

els of spatiotemporal organization, constructs that have become

fundamental to more contemporary approaches to science have

also become active considerations in neuroscience, in particular

the issue of uncertainty (Kahneman and Tversky, 1979; Knill and

Pouget, 2004; Gallistel and King, 2009; Kim et al., 2010; Vilares

and Kording, 2011). Scaling and uncertainty are of interest with

regard to neuromarketing, in that there is a common intuition

that inﬂuence occurs between individuals, between individuals

and a group, and between groups. As schematized in Figure 2D,

inﬂuence is thought to occur in the interaction between individ-

uals, who are embedded within groups, so that they affect their

respective groups, and the larger framework (e.g., society, market)

in which that group exists. This embedding of individuals/groups

can be directly analogized to the embedding of networks (Sutton

and Breiter, 1994).

Lawler, A. “White house stirs interest in brain-imaging initiative”, Science,

News, August 2, 2002; Abbott A. “Addicted”, Nature’s Senior European

Correspondent, Nature, News Feature, October 31, 2002; and http://pgp.

mgh.harvard.edu

This model of inﬂuence across scales of organization (e.g.,

individual, group, society/market) also relates to issues of

uncertainty due to information loss in the communication

between individuals/groups, or the uncertainty related to impre-

cision in the interpretation of communicated emotions (e.g.,

Figures 2B–D). Characterizing inﬂuence by scaling and uncer-

tainty has some appeal, but begs the issue of what inﬂuence is in

this context. When one considers inﬂuence in this model, it comes

across as resembling a ﬁeld of sorts, with a gradient of effects

as two entities wielding inﬂuence come in greater proximity

to each other (Figure 1B). To date, there has been no formal

deﬁnition of inﬂuence, either through axiomatic derivation, or

through iterative modeling (Banks and Tran, 2011) of behavior

data to show a speciﬁc mathematical formulation of a pattern

in a graph. Such work is clearly needed, and likely will depend

on the cognitive processes identiﬁed to underlie this “ﬁeld”

of inﬂuence, such as those involved with emotion, discussed

previously.

One might consider inﬂuence, and its potential scaling and

effects of uncertainty, as a product of human psychology and

the sub-processes underlying human information processing.

Such considerations point to the importance of having a com-

plete model of mental functioning, which is as yet lacking.

Neuromarketing investigations can have a major input into the

development of this integrated model, if they are conducted

from a consistent and coherent theoretical base as discussed

herein.

BASING INFLUENCE ON AN INTEGRATED MODEL OF

MENTAL FUNCTION

At this time, we have no uniﬁed model of the mind, which

shows how sub-processes such as attention, memory and

reward/aversion processing are integrated and function concur-

rently for decision-making, planning, and problem solving. When

one opens any cognitive science/biological psychology textbook,

one ﬁnds chapters on information theory, perception, attention,

decision-making, etc., but nothing integrating them. Even the use

of the term information theory—although considered the basis of

cognitive science—was never used in its mathematical framework

in cognitive science until approximately 4–6 years ago (Breiter

and Kim, 2008; Tononi, 2008; Gallistel and King, 2009; Kim et al.,

2010). For the most part, marketers have relied on thinking about

judgment and decision-making in terms of cognitive biases and

mental functions involved with choice.

Recently, attention has been given to the building of such

an integrated model of mental processing, starting with efforts

to look at the input end of cognitive function, and to consider

how quantitative descriptions of processes for reward/aversion,

attention, and memory might work together. This work has

led to research (Viswanathan et al., Under review) integrating

parts of RPT (representing reward/aversion) with variables from

signal detection theory (representing attention), and combining

signal detection theory with Ebbinghaus memory functions to

unpack sub-processes mediating working memory (Reilly et al.,

Under review). This early work argues that cognitive science

constructs can be integrated, and points to the large amount of

work needed to develop a comprehensive merger of domains in

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

cognitive science, including domains such as decision-making,

planning, and problem solving, along with output of the sys-

tem in terms of motor behavior, language, and autonomic

functions.

The ultimate integration of these cognitive functions can be

analogized to a wall chart in biochemistry where all chemical

pathways underlying biological metabolism are organized. We

are a long way from having such an integrated platform for

mental operations, particularly since such integrated systems as

in biochemistry also convey mechanism and allow causal infer-

ence. In the short term, the viability of such an effort can start

with developing complete constructs for attention, memory, and

reward/aversion processing. Complete constructs for attention,

for instance, would necessitate the mathematical description of

the relationship between focused, selective, sustained, divided,

and alternating attention. The potential for such complete con-

structs to be integrated across functions (i.e., perception, atten-

tion, memory, and reward/aversion processing) would then be

a necessary second step in testing the viability of developing a

general model of the mind.

Development of such a general model of mental function

would allow us to theorize and empirically test what set of

functions together respond to inﬂuence from another individ-

ual/organism, and exert inﬂuence on individuals/organisms out-

side of the person. With an integration of, at minimum, the

functions thought to comprise emotion and memory thereof,

cognitive psychology would likely be able to begin deﬁning a

quantitative model of inﬂuence. Such an effort will also depend

on parallel assessments of the integrated cognitive model through

approaches that (1) assess how well the integrated cognitive con-

struct ﬁts with neuroscience measures; (2) determine if impor-

tant features of the construct can be derived axiomatically (an

approach used extensively in traditional economics); and (3) test

if the integrated cognitive construct facilitates the analysis of large

data sets of human consumption and media use (referred to

as “big data”), which should show features of human cognitive

function.

Even so, efforts devoted towards developing neuromarketing

as a science of inﬂuence, and towards a general model of mental

function must remain cognizant of the risks inherent in such

research, particularly given the persuasiveness of brain imaging

(Roskies, 2008). Such risks are well covered in other founda-

tional literature (e.g., Senior et al., 2011), but it is worth noting

here that the subtractive and reverse inferential methodologies—

predicated on observing speciﬁc brain region activity associated

with speciﬁc tasks—are unable to conclusively conﬁrm either

the necessity of that speciﬁc region for that speciﬁc task, nor

the lack of involvement of other regions, particularly in complex

tasks (Friston et al., 1996; Poldrack, 2006, 2008). Confounds can

also arise in neuroscientiﬁc studies of behavioral change (i.e.,

inﬂuence). It is a mistake to assume that one wants changes

in both behavior and brain signal to interpret the effect of any

inﬂuence. Such circumstances only lend themselves to interpre-

tation when there is a parametric variation in both variables,

which in turn can lead to a power problem. Rather, it is usu-

ally preferable for variables in either behavior or neuroimaging

change, assuming the (often unrecognized) issue that baseline

or comparative conditions remain unchanged also. Similarly,

one must control for hormonal and demographic variables,

which have been shown to inﬂuence key neuroimaging variables

(Goldstein et al., 2005, 2010; Breiter et al., 2006). A ﬁnal caveat

is that we still do not understand the processes by which dis-

tributed groups of cells “process information” (e.g., Freeman,

2001). The functional domains of biochemistry, molecular biol-

ogy, and genetics are quite distinct from those we hypothesize

for behavior (e.g., attention, memory, reward/aversion process-

ing, etc.), and how distributed neural groups produce functional

domains and interact is far from understood. As such, all neural

signals must be looked at as providing ancillary support for

measures made at other spatiotemporal scales (e.g., behavior or

genetics).

There also remain key issues in the use of “big data”

approaches to neuromarketing. In particular, the high-

dimensionality and huge size of data sets in this context can

lead to inferential problems of their own—particularly spurious

correlations, noise accumulation, and incidental homogeneity

(e.g., Fan and Fan, 2008; Fan et al., 2013). The often uncontrolled

and naturalistic collection of such data sets also has the potential

to raise issues of public interest regarding the ethics of social

research (e.g., Kramer et al., 2014). That said, as long as

researchers approach their work in light of such caveats, big

data provides opportunities for neuromarketing as a science

of inﬂuence, in particular due to (i) its cohort sizes; (ii) its

attention to demographic and socio-economic variables; and (iii)

its broad array of variables that can be aligned to neuroscience

variables.

SUMMARY

This manuscript provides a theoretical framework for neuro-

marketing based on the process of inﬂuence, and how it shifts

distributions of choice across many scales of measurement,

from individual to group/market and society. As opposed to

issues of choice, issues of inﬂuence encompass a broader array

of behavioral science domains, pointing to the importance of

developing a rigorous quantitative model of mental function,

which can provide testable hypotheses for how distributions

of choice are shifted across scale and within scale (i.e., from

individuals to groups/market to society and back again). How-

ever, a tremendous amount of work is needed to get to this

point, and this work will need to meet the highest of aca-

demic standards if it is to change standards of practice and have

real relevance for the marketing community and those involved

with inﬂuence or behavior change, whether that be in educa-

tion, medicine, business, marketing communications, design, or

political policy.

ACKNOWLEDGMENTS

This work was supported by grants to Hans C. Breiter (#14118,

026002, 026104, 027804) from the NIDA, Bethesda, MD, USA

and grants (DABK39-03-0098 and DABK39-03-C-0098; The Phe-

notype Genotype Project in Addiction and Mood Disorder)

from the Ofﬁce of National Drug Control Policy—Counterdrug

Technology Assessment Center, Washington, DC, USA. Fur-

ther support was provided to Hans C. Breiter by the Warren

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

Wright Adolescent Center at Northwestern Memorial Hospital

and Northwestern University, Chicago, IL, USA. Support was also

provided by a grant to Anne J. Blood (#052368) from NINDS,

Bethesda, MD, USA. The funders had no role in study design, data

collection and analysis, decision to publish, or preparation of the

manuscript. Lastly, the authors wish to thank Charles N Rudick,

PhD for his critical commentary on the manuscript.

REFERENCES

Adami, C. (2004). Information theory in molecular biology. Phys. Life Rev. 1, 3–22.

doi: 10.1016/j.plrev.2004.01.002

Ambler, T., Braeutigam, S., Stins, J., Rose, S., and Swithenby, S. (2004). Salience

and choice: neural correlates of shopping decisions. Psychol. Mark. 21, 247–261.

doi: 10.1002/mar.20004

Ariely, D., and Berns, G. S. (2010). Neuromarketing: the hope and hype of

neuroimaging in business. Nat. Rev. Neurosci. 11, 284–292. doi: 10.1038/

nrn2795

Banks, H. T., and Tran, H. T. (2011). Mathematical and Experimental Modeling of

Physical and Biological Processes. Boca Raton, FL: CRC Press.

Barch, D. M., Burgess, G. C., Harms, M. P., Petersen, S. E., Schlaggar, B. L., Corbetta,

M., et al. (2013). Function in the human connectome: task-fMRI and individual

differences in behavior. Neuroimage 80, 169–189. doi: 10.1016/j.neuroimage.

2013.05.033

Barrett, L. F. (2006). Solving the emotion paradox: categorization and the expe-

rience of emotion. Pers. Soc. Psychol. Rev. 10, 20–46. doi: 10.1207/s1532

7957pspr1001_2

Braeutigam, S. (2005). Neuroeconomics—From neural systems to economic

behaviour. Brain Res. Bull. 67, 355–360. doi: 10.1016/j.brainresbull.2005.

06.009

Breiter, H. C., Aharon, I., Kahneman, D., Dale, A., and Shizgal, P. (2001). Functional

imaging of neural responses to expectancy and experience of monetary gains

and losses. Neuron 30, 619–639. doi: 10.1016/s0896-6273(01)00303-8

Breiter, H., and Gasic, G. (2004). “A general circuitry processing reward/aversion

information and its implications for neuropsychiatric illness,” in The Cognitive

Neurosciences III, ed M. Gazzaniga (Cambridge: MIT Press), 1043–1065.

Breiter, H. C., Gasic, G. P., and Makris, N. (2006). “Imaging the neural systems

for motivated behavior and their dysfunction in neuropsychiatric illness,” in

Complex Systems Science in Biomedicine, eds T. S. Deisboeck and J. Y. Kresh (New

York, NY: Springer), 763–810.

Breiter, H., and Kim, B. (2008). Recurrent and robust patterns underlying human

relative preference and associations with brain circuitry plus genetics in (U

Minn, Institute of Mathematics and its Applications). Available online at:

www.ima.umn.edu/2007-2008/W4.21-25.08/abstracts.html

Breiter, H. C., and Rosen, B. R. (1999). Functional magnetic resonance imaging of

brain reward circuitry in the human. Ann. N Y Acad. Sci. 877, 523–547. doi: 10.

1111/j.1749-6632.1999.tb09287.x

Cabanac, M. (1971). Physiological role of pleasure. Science 173, 1103–1107. doi: 10.

1126/science.173.4002.1103

Camerer, C. F. (2008). Neuroeconomics: opening the gray box. Neuron 60, 416–419.

doi: 10.1016/j.neuron.2008.10.027

Cuthbert, B. N., and Insel, T. R. (2013). Toward the future of psychiatric diagnosis:

the seven pillars of RDoC. BMC Med. 11:126. doi: 10.1186/1741-7015-11-126

Deci, E. L., and Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in

Human Behavior. New York and London: Plenum, p. 371.

Fan, J., and Fan, Y. (2008). High-dimensional classiﬁcation using features annealed

independence rules. Ann. Stat. 36, 2605–2637. doi: 10.1214/07-aos504

Fan, J., Han, F., and Liu, H. (2013). Challenges of big data analysis. arXiv:1308.1479

[stat]. Available online at: http://arxiv.org/abs/1308.1479. Accessed on April 30,

2014.

Feynman, R. P., Leighton, R. B., and Sands, M. (1963). The Feynman Lectures

on Physics, BD 1: Mainly Mechanics, Radiation and Heat. Redwood City, CA:

Addison-Wesley.

Fisher, C. E., Chin, L., and Klitzman, R. (2010). Deﬁning neuromarketing: prac-

tices and professional challenges. Harv. Rev. Psychiatry 18, 230–237. doi: 10.

3109/10673229.2010.496623

Freeman, W. (2001). How Brains Make Up Their Minds. New York: Columbia

University Press.

Friston, K. J., Price, C. J., Fletcher, P., Moore, C., Frackowiak, R. S., and Dolan, R. J.

(1996). The trouble with cognitive subtraction. Neuroimage 4, 97–104. doi: 10.

1006/nimg.1996.0033

Fugate, D. L. (2007). Neuromarketing: a layman’s look at neuroscience and its

potential application to marketing practice. J. Consum. Mark. 24, 385–394.

doi: 10.1108/07363760710834807

Gallistel, C. R., and King, A. P. (2009). Memory and the Computational Brain: Why

Cognitive Science Will Transform Neuroscience. UK: John Wiley & Sons.

Gasic, G. P., Smoller, J. W., Perlis, R. H., Sun, M., Lee, S., Kim, B. W., et al.

(2009). BDNF, relative preference and reward circuitry responses to emotional

communication. Am. J. Med. Genet. B Neuropsychiatr. Genet. 150B, 762–781.

doi: 10.1002/ajmg.b.30944

Goldstein, J. M., Jerram, M., Abbs, B., Whitﬁeld-Gabrieli, S., and Makris, N.

(2010). Sex differences in stress response circuitry activation dependent on

female hormonal cycle. J. Neurosci. 30, 431–438. doi: 10.1523/jneurosci.3021-09.

2010

Goldstein, J. M., Jerram, M., Poldrack, R., Ahern, T., Kennedy, D. N., Seidman,

L. J., et al. (2005). Hormonal cycle modulates arousal circuitry in women using

functional magnetic resonance imaging. J. Neurosci. 25, 9309–9316. doi: 10.

1523/jneurosci.2239-05.2005

Gross, J. J. (2009). Handbook of Emotion Regulation. New York, NY: Guilford Press.

Gross, J. J. (2013). Emotion regulation: taking stock and moving forward. Emotion

13, 359–365. doi: 10.1037/a0032135

Herrnstein, R. J. (1961). Relative and absolute strength of response as a function of

frequency of reinforcement. J. Exp. Anal. Behav. 4, 267–272. doi: 10.1901/jeab.

1961.4-267

Hubert, M., and Kenning, P. (2008). A current overview of consumer neuroscience.

J. Consum. Behav. 7, 272–292. doi: 10.1002/cb.251

Insel, T., Cuthbert, B., Garvey, M., Heinssen, R., Pine, D. S., Quinn, K., et al. (2010).

Research domain criteria (RDoC): toward a new classiﬁcation framework for

research on mental disorders. Am. J. Psychiatry 167, 748–751. doi: 10.1176/appi.

ajp.2010.09091379

Ioannides, A. A., Liu, L., Theoﬁlou, D., Dammers, J., Burne, T., Ambler, T.,

et al. (2000). Real time processing of affective and cognitive stimuli in the

human brain extracted from MEG signals. Brain Topogr. 13, 11–19. doi: 10.

1023/A:1007878001388

Javor, A., Koller, M., Lee, N., Chamberlain, L., and Ransmayr, G. (2013). Neu-

romarketing and consumer neuroscience: contributions to neurology. BMC

Neurol. 13:13. doi: 10.1186/1471-2377-13-13

Kahneman, D. (2011). Thinking, Fast and Slow. New York, NY: Macmillan.

Kahneman, D., and Tversky, A. (1979). Prospect theory: an analysis of decision

under risk. Econometrica. 47, 263–292. doi: 10.2307/1914185

Kahneman, D., Wakker, P. P., and Sarin, R. (1997). Back to bentham? Explorations

of experienced utility. Q. J. Econ. 112, 375–406. doi: 10.1162/0033553975

55235

Kim, B. W., Kennedy, D. N., Lehár, J., Lee, M. J., Blood, A. J., Lee, S., et al.

(2010). Recurrent, robust and scalable patterns underlie human approach and

avoidance. PLoS One 5:e10613. doi: 10.1371/journal.pone.0010613

King-Casas, B., Tomlin, D., Anen, C., Camerer, C. F., Quartz, S. R., and Montague,

P. R. (2005). Getting to know you: reputation and trust in a two-person

economic exchange. Science 308, 78–83. doi: 10.1126/science.1108062

Knill, D. C., and Pouget, A. (2004). The Bayesian brain: the role of uncertainty

in neural coding and computation. Trends Neurosci. 27, 712–719. doi: 10.1016/

j.tins.2004.10.007

Kramer, A., Guillory, J., and Hancock, J. (2014). Experimental evidence of massive-

scale emotional contagion through social networks. Proc. Natl. Acad. Sci. U S A

111, 8788–8790. doi: 10.1073/pnas.1320040111

Kuppens, P., Tuerlinckx, F., Russell, J. A., and Barrett, L. F. (2013). The relation

between valence and arousal in subjective experience. Psychol. Bull. 139, 917–

940. doi: 10.1037/a0030811

Laybourne, P., and Lewis, D. (2005). Neuromarketing: the future of consumer

research. Admap 13, 28–30.

Lee, N., Broderick, A. J., and Chamberlain, L. (2007). What is “neuromarketing”?

A discussion and agenda for future research. Int. J. Psychophysiol. 63, 199–204.

doi: 10.1016/j.ijpsycho.2006.03.007

Lindquist, K. A., Siegel, E. H., Quigley, K. S., and Barrett, L. F. (2013). The hundred-

year emotion war: Are emotions natural kinds or psychological constructions?

Comment on Lench, Flores, and Bench (2011). Psychol. Bull. 139, 255–263.

doi: 10.1037/a0029038

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

Breiter et al. Redeﬁning neuromarketing as an integrated science of inﬂuence

Markowitz, H. (1952). Portfolio selection

∗

. J. Finance 7, 77–91. doi: 10.1111/j.1540-

6261.1952.tb01525.x

McClure, S. M., Li, J., Tomlin, D., Cypert, K. S., Montague, L. M., and Montague,

P. R. (2004). Neural correlates of behavioral preference for culturally familiar

drinks. Neuron 44, 379–387. doi: 10.1016/j.neuron.2004.09.019

Morris, S. E., and Cuthbert, B. N. (2012). Research domain criteria: cognitive

systems, neural circuits and dimensions of behavior. Dialogues Clin. Neurosci.

14, 29–37.

Murphy, E. R., Illes, J., and Reiner, P. B. (2008). Neuroethics of neuromarketing. J.

Consum. Behav. 7, 293–302. doi: 10.1002/cb.252

Ohme, R., Reykowska, D., Wiener, D., and Choromanska, A. (2009). Analysis

of neurophysiological reactions to advertising stimuli by means of EEG and

galvanic skin response measures. J. Neurosci. Psychol. Econ. 2, 21–31. doi: 10.

1037/a0015462

Paulus, M. P. (2007). Decision-making dysfunctions in psychiatry—altered home-

ostatic processing? Science 318, 602–606. doi: 10.1126/science.1142997

Perelson, A. S., Bragg, J. G., and Wiegel, F. W. (2006). “The complexity of the

immune system: scaling laws,” in Complex Systems Science in Biomedicine, eds

T. S. Deisboeck and J. Y. Kresh (New York: Springer), 451–459.

Perlis, R. H., Holt, D. J., Smoller, J. W., Blood, A. J., Lee, S., Kim, B. W., et al.

(2008). Association of a polymorphism near CREB1 with differential aversion

processing in the insula of healthy participants. Arch. Gen. Psychiatry 65, 882–

892. doi: 10.1001/archgenpsychiatry.2008.3

Plassmann, H., Ramsøy, T. Z., and Milosavljevic, M. (2012). Branding the brain:

a critical review and outlook. J. Consum. Psychol. 22, 18–36. doi: 10.1016/

j.jcps.2011.11.010

Poldrack, R. A. (2006). Can cognitive processes be inferred from neuroimaging

data? Trends Cogn. Sci. 10, 59–63. doi: 10.1016/j.tics.2005.12.004

Poldrack, R. A. (2008). The role of fMRI in cognitive neuroscience: where do we

stand? Curr. Opin. Neurobiol. 18, 223–227. doi: 10.1016/j.conb.2008.07.006

Roskies, A. L. (2008). Neuroimaging and inferential distance. Neuroethics 1, 19–30.

doi: 10.1007/s12152-007-9003-3

Santos, J. P., Seixas, D., Brandão, S., and Moutinho, L. (2011). Investigating the

role of the ventromedial prefrontal cortex in the assessment of brands. Front.

Neurosci. 5:77. doi: 10.3389/fnins.2011.00077

Savage, V. M., and West, G. B. (2006). “Biological scaling and physiological time,”

in Complex Systems Science in Biomedicine, eds T. S. Deisboeck and Y. J. Kresh

(New York: Springer), 141–163.

Senior, C., and Lee, N. (2008). Editorial: a manifesto for neuromarketing science. J.

Consum. Behav. 7, 263–271. doi: 10.1002/cb.250

Senior, C., Lee, N., and Butler, M. (2011). Organizational cognitive neuroscience.

Organ. Science 22, 804–815. doi: 10.1287/orsc.1100.0532

Shannon, C. E., and Weaver, W. (1949). The Mathematical Theory of Communica-

tion Volume 1. Illinois: University of Illinois Press.

Sutton, J., and Breiter, H. (1994). Neural scale invariance: an integrative model with

implications for neuropathology. World Congr. Neural Netw. 4, 667–672.

Szostak, R. (2004). Classifying Science: Phenomena, Data, Theory, Method, Practice

(Vol. 7). Dordrecht: Springer.

Tomlin, D., Kayali, M. A., King-Casas, B., Anen, C., Camerer, C. F., Quartz, S. R.,

et al. (2006). Agent-speciﬁc responses in the cingulate cortex during economic

exchanges. Science 312, 1047–1050. doi: 10.1126/science.1125596

Tononi, G. (2008). Consciousness as integrated information: a provisional mani-

festo. Biol. Bull. 215, 216–242. doi: 10.2307/25470707

Van Essen, D. C., Ugurbil, K., Auerbach, E., Barch, D., Behrens, T. E.,

Bucholz, R., et al. (2012). The human connectome project: a data acqui-

sition perspective. Neuroimage 62, 2222–2231. doi: 10.1016/j.neuroimage.

2012.02.018

Vecchiato, G., Astolﬁ, L., De Vico Fallani, F., Toppi, J., Aloise, F., Bez, F., et al. (2011).

On the use of EEG or MEG brain imaging tools in neuromarketing research.

Comput. Intell. Neurosci. 2011:643489. doi: 10.1155/2011/643489

Vilares, I., and Kording, K. (2011). Bayesian models: the structure of the world,

uncertainty, behavior and the brain. Ann. N Y Acad. Sci. 1224, 22–39. doi: 10.

1111/j.1749-6632.2011.05965.x

Wilson, R. M., Gaines, J., and Hill, R. P. (2008). Neuromarketing and consumer free

will. J. Consum. Aff. 42, 389–410. doi: 10.1111/j.1745-6606.2008.00114.x

Conﬂict of Interest Statement: The authors declare that the research was conducted

in the absence of any commercial or ﬁnancial relationships that could be construed

as a potential conﬂict of interest.

Received: 01 May 2014; accepted: 29 December 2014; published online: 12 February

2015.

Citation: Breiter HC, Block M, Blood AJ, Calder B, Chamberlain L, Lee N, Livengood

S, Mulhern FJ, Raman K, Schultz D, Stern DB, Viswanathan V and Zhang FZ

(2015) Redeﬁning neuromarketing as an integrated science of inﬂuence. Front. Hum.

Neurosci. 8:1073. doi: 10.3389/fnhum.2014.01073

This article was submitted to the journal Frontiers in Human Neuroscience.

Mulhern, Raman, Schultz, Stern, Viswanathan and Zhang. This is an open-access

article distributed under the terms of the Creative Commons Attribution License (CC

BY). The use, distribution and reproduction in other forums is permitted, provided

the original author(s) or licensor are credited and that the original publication in this

journal is cited, in accordance with accepted academic practice. No use, distribution or

reproduction is permitted which does not comply with these terms.

Frontiers in Human Neuroscience www.frontiersin.org February 2015 | Volume 8 | Article 1073 |

ORIGINAL RESEARCH

published: 30 April 2015

doi: 10.3389/fnhum.2015.00176

Age-related striatal BOLD changes

without changes in behavioral loss

aversion

Vijay Viswanathan

1,2†

, Sang Lee

3,4,5‡

, Jodi M. Gilman

3‡

, Byoung Woo Kim

2,3,4,5‡

, Nick

Lee

2,6‡

, Laura Chamberlain

2,6‡

, Sherri L. Livengood

2,4‡

, Kalyan Raman

1,2,4,7$

, Myung Joo

Lee

2,3,4,5$

, Jake Kuster

3,5$

, Daniel B. Stern

2,4$

and Bobby Calder

2,7‡

, Frank J. Mulhern

1,2‡

Anne J. Blood

2,3,5†

, Hans C. Breiter

2,3,4,5

†

Medill Integrated Marketing Communications, Northwestern University, Evanston, IL, USA,

Applied Neuromarketing

Consortium: Northwestern University, Wayne State University, University of Michigan, Loughborough University School of

Business and Economics (UK) and Massachusetts General Hospital/Harvard University, Chicago, IL, USA,

Mood and Motor

Control Laboratory or Laboratory of Neuroimaging and Genetics, Department of Psychiatry, Massachusetts General Hospital,

Boston, MA, USA,

Warren Wright Adolescent Center, Department of Psychiatry and Behavioral Science, Northwestern

University Feinberg School of Medicine, Chicago, IL, USA,

Northwestern University and Massachusetts General Hospital

Phenotype Genotype Project in Addiction and Mood Disorders, Chicago, IL, USA,

Marketing Group, Aston Business

School, Birmingham, UK,

Department of Marketing, Kellogg School of Management, Northwestern University, Evanston,

IL, USA

Edited by:

Sven Braeutigam,

University of Oxford, UK

Reviewed by:

Martin P. Paulus,

University of California San Diego,

USA

Dave J. Hayes,

University of Toronto, Canada

*Correspondence:

Hans C. Breiter,

Warren Wright Adolescent Center,

Department of Psychiatry and

Behavioral Science, Northwestern

University Feinberg School of

Medicine, 710 N. Lake Shore Dr.,

Abbott Hall 1301, Chicago, IL 60611,

USA

h-breiter@northwestern.edu

†

Joint ﬁrst authorship.

‡

Joint second authorship.

Joint third authorship.

Received: 23 April 2014

Accepted: 15 March 2015

Published: 30 April 2015

Citation:

Viswanathan V, Lee S, Gilman JM,

Kim BW, Lee N, Chamberlain L,

Livengood SL, Raman K, Lee MJ,

Kuster J, Stern DB and Calder B,

Mulhern FJ, Blood AJ, Breiter HC

(2015) Age-related striatal BOLD

changes without changes in

behavioral loss aversion.

Front. Hum. Neurosci. 9:176.

doi: 10.3389/fnhum.2015.00176

Loss aversion (LA), the idea that negative valuations have a higher psychological

impact than positive ones, is considered an important variable in consumer research.

The literature on aging and behavior suggests older individuals may show more LA,

although it is not clear if this is an effect of aging in general (as in the continuum

from age 20 and 50 years), or of the state of older age (e.g., past age 65 years).

We also have not yet identiﬁed the potential biological effects of aging on the neural

processing of LA. In the current study we used a cohort of subjects with a 30 year

range of ages, and performed whole brain functional MRI (fMRI) to examine the ventral

striatum/nucleus accumbens (VS/NAc) response during a passive viewing of affective

faces with model-based fMRI analysis incorporating behavioral data from a validated

approach/avoidance task with the same stimuli. Our a priori focus on the VS/NAc

was based on (1) the VS/NAc being a central region for reward/aversion processing;

(2) its activation to both positive and negative stimuli; (3) its reported involvement with

tracking LA. LA from approach/avoidance to affective faces showed excellent ﬁdelity

to published measures of LA. Imaging results were then compared to the behavioral

measure of LA using the same affective faces. Although there was no relationship

between age and LA, we observed increasing neural differential sensitivity (NDS) of the

VS/NAc to avoidance responses (negative valuations) relative to approach responses

(positive valuations) with increasing age. These ﬁndings suggest that a central region

for reward/aversion processing changes with age, and may require more activation

to produce the same LA behavior as in younger individuals, consistent with the idea

of neural efﬁciency observed with high IQ individuals showing less brain activation to

complete the same task.

Keywords: loss aversion, aging, nucleus accumbens, reward, fMRI, neurocompensation

Frontiers in Human Neuroscience | www.frontiersin.org April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

Introduction

Age is among the most commonly used variables in marketing

and consumer research. While age is a deceptively simple

variable, the underlying construct of biological age and how it

relates to behavior is not always clear. One age effect supported

by a number of social psychology studies is that older adults

put more weight on avoiding potential negative outcomes,

as evidenced by an aversion to change (Botwinick, 1978)

and nostalgia for early experience (Schindler and Holbrook,

2003). Aging research points to an association of age with

making less risky decisions (Johnson and Busemeyer, 2010),

and suggests that older individuals generally avoid losses to

a greater extent than younger individuals (e.g., Heckhausen,

1997). A fundamental way to quantify this perspective is with

the concept of ‘‘loss aversion’’ (LA), in which negative stimuli

have a disproportionate psychological impact relative to positive

ones (Kahneman and Tversky, 1979), and can be defined

mathematically by the ratio of valuation of monetary losses

relative to valuation of gains (Tversky and Kahneman, 1991),

or in more general terms, as the ratio of avoidance to approach

measures (Abdellaoui et al., 2007). LA has become an important

variable in consumer research (Ariely et al., 2005; Paraschiv

and L’Haridon,

2008) and is consistent with the observation

that older individuals are more focused on goals pertaining to

maintenance and regulation of loss (Ebner et al., 2006). Cole et al.

(2008) suggest, based on ‘‘regulatory focus theory’’ (Avnet and

Higgins, 2006), that older individuals would be more prevention-

focused i.e., avoid losses, than promotion-focused i.e., pursuit

of gains.

Neuroscience studies have examined the biological basis

for age-related changes in cognitive function (Hedden and

Gabrieli, 2004; Mohr et al., 2010), which might affect biases in

decision-making such as LA. For instance, Raz (2000) found

a steady decline in the prefrontal cortex (PFC) structures

starting from the age of 20 along with a decline in the

striatal volume over the lifespan of an individual. In many

studies, the biology of age-related changes in the brain goes

in the same direction as behavior (Good et al., 2001), as for

instance in the domain of episodic memory, where older adults

have demonstrated decreased activation of various sites in the

left and right prefrontal cortices correlating with decreased

performance on the task relative to their younger counterparts

(Grady et al., 1995, 1999; Cabeza et al., 1997; Madden et al.,

1999; Grady and Craik, 2000; Reuter-Lorenz, 2002; Stebbins

et al., 2002). An alternate outcome is also possible, wherein

alterations in brain activity are not associated with an alteration

of behavior, namely, increasing amounts of activation are needed

to produce the same behavior (e.g., neurocompensation; Cabeza

et al., 2002; Park and Reuter-Lorenz, 2009; Daselaar et al.,

2015).

These neuroimaging and behavioral studies thus suggest at

least two potential hypotheses regarding LA, its underlying

neural substrate, and aging: (1) LA behavior may parallel

changes in neural processing, specifically, LA behavior may

increase with age along with increased activation in tissue

required to process it; or (2) LA behavior may increase more

slowly than the compensatory activity in tissue processing

it (i.e., there may be small differences in LA behavior

with age, and strong brain activity differences during its

processing). This latter possibility finds support from an

early functional MRI (fMRI) study that reported decreases in

either performance or IQ were associated with increased brain

activation during cognitive function (Seidman et al., 1998),

and the observation of potentially compensatory activity in

older individuals (Meunier et al., 2014). Two recent studies of

LA specifically support option (2) above, in that they report

LA behavior does not change between young adults and old

adults (Li et al., 2013) or between adolescents and young

adults (Barkley-Levenson et al., 2013). One of these studies

also evaluated neural processing of LA, and found differences

between adolescents and young adults in large decision-making

networks, further supporting option (2) (Barkley-Levenson et al.,

2013).

In the current study we sought to test these hypotheses

evaluating subject age against (1) the relative overweighting

of behavioral responses to negative vs. positive stimuli (i.e.,

LA behavior) using a validated keypress measure (Kim et al.,

2010); and (2) neural differential sensitivity (NDS; Tom et al.,

2007) within the ventral striatum/nucleus accumbens (VS/NAc)

to the same stimuli used in the behavior task. Given the

substantial involvement of the VS/NAc in motor preparation

(e.g., Florio et al., 1999) and abnormality with motor illnesses

such as Parkinsonism (e.g., Aarts et al., 2014; Payer et al., 2015),

we sought to avoid the motoric contamination inherent with

cognitive imaging studies of the VS/NAc using monetary choice

paradigms (Tom et al., 2007; Canessa et al., 2013). Since the

motor responses for the keypress task would not be separate from

reward/aversion assessments (the amount of VS/NAc activation

could just reflect how much an individual was keypressing to

approach or avoid a stimulus, or reflect the urgency of their

responses), the keypress task was done outside the MRI, and the

outcome of keypress responses used for model-based analysis

of VS/NAc signal during passive viewing of the same stimuli;

such a model-based approach to fMRI has been used with this

task (Aharon et al., 2001) and fMRI-based imaging-genetics

used with these stimuli and task before (Perlis et al., 2008;

Gasic et al., 2009), consistent with the framework for model-

based imaging discussed by others (e.g., Mittner et al., 2014;

Wang and Voss, 2014; White et al., 2014; Xu et al., 2015). The

implicit assumption in this model-based application was that the

emotional response to faces in the scanner, and the behavior

based on emotional response to the same stimuli outside of the

scanner would be related, as suggested for studies of emotion-

based processing by other investigators (Hayes and Northoff,

2012).

To examine the underlying physiological effects of age

on LA, we performed whole brain fMRI to monitor activity

within the VS/NAc, using a passive viewing paradigm with

affective faces known to evoke positive and negative valuations

(Strauss et al., 2005), given the VS/NAc is a central region

for reward/aversion processing (Breiter et al., 1997; Blood

et al., 1999; Breiter and Rosen, 1999; Hayes and Northoff,

2012), and has been shown to activate to both positive and

Frontiers in Human Neuroscience | www.frontiersin.org

April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

negative stimuli (Aharon et al., 2001; Becerra et al., 2001;

Breiter et al., 2001; Kober et al., 2008; Hayes and Northoff,

2012) and to track LA for the choice and anticipation phases

of decision making (Tom et al., 2007; Lee et al., 2012;

Canessa et al., 2013). For a behavioral index of LA, we used

the same affective faces (Ekman and Friesen, 1976) with a

keypress task performed outside the MRI that allowed the

subject multiple potential decisions: (1) to do nothing about

the default viewing time of a picture; (2) to view the picture

for longer (approach); or (3) to view the picture for shorter

(avoidance) time. The keypress data was analyzed to produce

a value function (Breiter and Kim, 2008; Kim et al., 2010)

for each subject that is analogous to a prospect theory value

function or utility curve (Kahneman and Tversky, 1979),

but unlike any other reward/aversion construct actually uses

an entropy variable representing information (Shannon and

Weaver, 1949). The slopes of the negative and positive portions

of this curve can be readily sampled to yield a measure of

LA in a general framework for LA as an overweighting of

aversion (toward negative stimuli) relative to approach (toward

positive stimuli) as discussed by Abdellaoui et al. (2007). For

the model-based fMRI analysis, we explicitly required that (i)

relative activation to negative (avoidance) stimuli vs. positive

(approach) stimuli (i.e., the NDS of avoidance vs. approach)

would occur in the VS/NAc; and (ii) that NDS-related fMRI

signal in the VS/NAc would significantly correlate with LA

behavior across subjects. If this were observed, we then sought

to evaluate if VS/NAc NDS would increase with age, and

whether or not it would parallel any relationship of LA

behavior to age, supporting either the first or second hypothesis

regarding the interaction of LA, its underlying neural substrate,

and aging.

Methods

Subjects

Healthy control subjects were aggregated for an exploratory

analysis on an available sample of subjects with complete

behavior and imaging data from three paradigms for which

LA parameters could be computed and evaluated. The

resulting sample of 17 subjects was compiled for use with

another project testing if the negative component of LA

could explain aspects of amygdala function across the three

paradigms and connect it to structural measures (Lee et al.,

2012); the current study focused just on the emotional faces

paradigm given the value function curves from keypressing

to these stimuli had never been published, nor evaluated

against NDS or age. These 17 subjects were recruited by

advertisement and were part of a larger phenotype genotype

project in addiction and mood disorder (PGP)

. Subjects

were free of any psychiatric, neurological, or medical issues

per psychiatrist-based SCID for DSM-IV diagnoses, medical

review of systems and physical evaluation including blood

chemistry. Race was determined by individual self-identification

using a standardized form (Benson and Marano, 1998), and

http://pgp.mgh.harvard.edu

handedness via the Edinburgh Handedness Inventory (Oldfield,

1971). Participating subjects were without any current or

lifetime DSM-IV Axis I disorder or major medical illness

known to influence brain structure or function, including

neurologic disease, HIV, and hepatitis C. Subjects were

scanned at normal or corrected normal vision. Women

were scanned during their mid-follicular phase based

upon self-reported menstrual history, with confirmation

at the time of scanning based on hormonal testing with a

urine assay.

Participants in the study were adults (10 males, 7 females;

5 African Americans and 12 Caucasians) between the ages

of 20 and 55 with a mean (±SE) age of 35.8 ± 2.7 years,

with no significant difference between men and women

(1,15)

= 1.81, P < 0.20). They had a mean educational history

of 15.4 ± 1.9 years, with no significant difference between men

and women (F

(1,15)

= 2.78, P < 0.12). Fifteen of subjects were

right-handed.

Experimental Paradigm and Ofﬂine Behavioral

Testing

In Scanner

Two fMRI scans were acquired (8 min 40 s each), each

consisting of 20-s blocks of the following seven experimental

conditions: angry, fearful, happy, sad, neutral expressions

(Ekman and Friesen, 1976), along with phase-scrambled stimuli

and fixation (Figure 1). During each scan, the seven conditions

(blocks) were presented in a counterbalanced order such

that no condition followed or preceded another more than

once. This produced a sequence of 25 blocks for the first

run, and 24 plus one blocks for the second run, with the

extra block in the second run being equivalent to the last

block in the first run, placed at the beginning to maintain

counterbalancing across all conditions. Each facial expression

block included standardized images of faces of eight individuals

(four males) in a pseudorandom order (Breiter et al., 1996;

Strauss et al., 2005). Each face was displayed for 200 ms

with a 300 ms interstimulus interval during which a fixation

cross was displayed, with five repetitions of each face stimulus

per block (40 faces total per block). Face stimuli (Ekman

and Friesen, 1976) were previously normalized at the MIT

Media Lab (Breiter et al., 1996). The face stimuli were

projected via a Sharp XG-2000V color LCD projector through

a collimating lens onto a hemicircular tangent screen and

viewed by the subject via a mirror affixed to the head

coil. Subjects were instructed to simply look at the faces,

keeping their eyes focused on the center of the picture at

the location of the cross-hair. After completion of scanning,

subjects performed a memory task in which they were asked to

identify faces and facial expressions they had seen during the

scanning session.

Ofﬂine Behavioral Testing

This experiment utilized a keypress task to determine each

subject’s relative preference toward the ensemble of faces

(Aharon et al., 2001; Elman et al., 2005; Strauss et al., 2005;

Levy et al., 2008; Makris et al., 2008; Perlis et al., 2008; Gasic

Frontiers in Human Neuroscience | www.frontiersin.org April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

FIGURE 1 | Schema for Experimental Paradigm Used with fMRI. At

top, the categories of facial expressions used are shown along with

baselines (scrambled faces and ﬁxation). A schema of how these stimuli

were presented as blocks is shown at bottom, along with details

relating to number of stimuli used per block. See Methods Section for

further detail.

FIGURE 2 | Keypress Paradigm. The behavioral task done outside the MRI

to minimize motor confounds to activation is schematized above, with an

example raster plot below.

et al., 2009; Yamamoto et al., 2009; Kim et al., 2010), which

had been used for passive viewing during scanning. The

separation of passive viewing and keypress response allowed

the fMRI component to be free of motoric elements, which

would otherwise confound interpretation of the fMRI results

(please see Introduction Section). The keypress procedure was

implemented with MatLab software on a PC (i.e., a personal

computer). This task captured the reward valuation attributed

to each observed face, and quantified positive and negative

preferences involving (i) decision-making regarding the valence

of behavior; and (ii) judgments that determine the magnitude of

approach and avoidance (Breiter et al., 2006; Perlis et al., 2008;

Figure 2). The objective was to determine how much effort each

subject was willing to trade for viewing each facial expression

compared to a default viewing time. Subjects were told that they

would be exposed to a series of pictures that would change every

8 s (the default valuation of 6 s + 2 s decision block; Figure 2)

if they pressed no keys. If they wanted a picture to disappear

faster, they could alternate pressing one set of keys (#3 and #4

on the button box), whereas if they wanted a picture to stay

longer on the screen, they could alternate pressing another set

of keys (#1 and #2 on the button box). Subjects had a choice to

do nothing (default condition), increase viewing time, decrease

viewing time, or a combination of the two responses (Figure 2).

A ‘‘slider’’ was displayed to the left of each picture to indicate

total viewing time. Subjects were informed that the task would

last approximately 20 min, and that this length was independent

of their behavior, as was their overall payment. The dependent

measure of interest was the amount of work, in number of

keypresses, which subjects traded for face viewtime.

Frontiers in Human Neuroscience | www.frontiersin.org 4 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

FIGURE 3 | Relative Preference Graphs. The output of the keypress

paradigm is shown for 17 subjects on the right. The variable K stands for

the mean number of keypresses made for category of picture viewed

(mean viewtime per category can also be substituted to produce

comparable results). K can be interpreted as the effort expended to

approach (+ keypresses) or avoid (keypresses) stimuli, and approximates

wanting. The variable H stands for the Shannon entropy (information or

uncertainty related to making a choice) for each category of picture. It can

be interpreted as representing memory about the uncertainty related to

making choices over this stimulus set. The approach graph is in green and

the avoidance graph is in red. On the left is shown a schema for prospect

theory (Kahneman and Tversky, 1979) to emphasize the similarity, despite

different variables, between the two theoretical frameworks of prospect

theory and relative preference theory (Breiter and Kim, 2008; Kim et al.,

2010). Both show an increased slope for negative or avoidance responses

relative to positive or approach responses.

Magnetic Resonance Imaging

All functional MR imaging was performed on a Siemens

Trio 3 Tesla MRI system using an eight-channel phased-array

receive-only RF coil. Subjects were positioned in the MRI scanner

and their heads stabilized using foam pads and adjustable paddles

fixed to the RF coil assembly. Blood oxygenation level-dependent

(BOLD) functional images were acquired using gradient-echo

EPI (TR/TE/α 2.5 s/30 ms/90

◦

, 3.125 mm × 3.125 mm × 3 mm

resolution), with slices situated parallel to the AC–PC line,

and parallel to the inside curve of the FOC to minimize signal

distortion in this region (Deichmann et al., 2003). Structural

images were acquired using a high resolution T1-weighted

MPRAGE sequence (192 sagittal slices over the full head volume,

matrix = 224 × 256, FOV = 224 × 256 mm

, thickness = 1

mm, no gap) before functional scanning. Details of the imaging

parameters and protocol have been reported previously (Perlis

et al., 2008; Gasic et al., 2009).

Data Analysis

Behavioral Data

Keypress data were checked by a relative preference theory

analysis of each subject, using previously validated procedures

(Breiter and Kim, 2008; Kim et al., 2010; Figure 3). These

procedures produce a valuation graph with variables K

and H that encode mean keypress number and Shannon

entropy (i.e., information) (Shannon and Weaver, 1949). This

valuation graph has been interpreted to relate ‘‘wanting’’ of

stimuli (Aharon et al., 2001) to the uncertainty associated

with making a choice (Kim et al., 2010). Using a local

and general definition of LA (Abdellaoui et al., 2007), we

computed the slope of the negative value/utility function

(s−) and the slope of the positive value/utility function (s+),

to produce s−/s+ (Figure 4). Specifically, s− and s+ were

computed by the integral of the curve-fit slope over the

10% of the curve closest to the inflection point or origin

(Figure 4). An absolute value of s−/s+ was then computed

for each subject. With the full dataset of these subjects we

then assessed the association of LA (i.e., |s−/s+|) with age

using linear regression; given this test was done in parallel

with another test against age (see NDS and age below),

there was a correction for multiple comparisons imposed of

p < 0.05/2 = 0.025.

Imaging Data

fMRI data were analyzed using the FSL platform (FMRIB’s

Software Library, v4.1.9)

, and followed signal processing and

http://fsl.fmrib.ox.ac.uk/fsl/fslwiki

Frontiers in Human Neuroscience | www.frontiersin.org 5 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

FIGURE 4 | Local Loss Aversion Computation. This study used a

local rather than a global assessment of loss aversion, meaning that the

ratio was computed using the slopes around the inﬂection point for

each graph, rather than the entire graph (Abdellaoui et al., 2007). In this

ﬁgure, the same cartoon for prospect theory used in Figure 3 is shown

on the left, with a red box around the inﬂection point between the

positive and negative graphs that meet at the origin. The approach KH

graphs (variables deﬁned in Figure 3 legend) and avoidance KH graphs

for the 17 subjects are shown as unﬁtted points in gray to the right.

One example subject is shown with color points as deﬁned in each

graph with a ﬁtted curve. The lowest 10% of these curves was then

used for assessing the slopes as shown in the mathematics at bottom.

Of interest, the LA value for these subjects approximates that reported

by Tversky and Kahneman (1992).

statistical analysis procedures we have detailed elsewhere (Perlis

et al., 2008; Gasic et al., 2009). Stimuli were grouped based

on keypress responses into stimuli subjects avoided (Angry,

Fearful, and Sad faces) and stimuli subjects approached by

using the keypress task to increase viewing time (Happy

faces). These two stimuli classes were contrasted in order to

determine brain areas that responded more highly to negative

(avoidance) than to positive (approach) stimuli [i.e., the β

slope for the negative activation (or PE for the –COPE) was

greater than the β slope of the positive activation (or PE

for the +COPE)]. Statistical maps of NDS to losses relative

to gains (− > +) were constructed as a group map, and

voxels selected above a whole brain correction for z-stat = 2.3

that overlapped the VS/NAc segmentation volumes from the

ICBM152 T1 template (Perlis et al., 2008; Gasic et al., 2009;

Figure 5). VS/NAc segmentation followed previously published

parameters for its boundaries (Breiter et al., 1997), using

processes that have been well validated (see Breiter et al., 1994;

Makris et al., 2004). As a second step in our model-based

fMRI analysis, we assessed the correlation of NDS [(Angry,

Fearful, and Sad > Happy)] in the VS/NAc to LA, as described

by Tom et al. (2007), using a subset of the 17 subjects

who were not statistical outliers (Figure 6). For this isolated

correlation, significant effects had to meet p < 0.05. With the

full dataset of these subjects we then assessed the regression

of NDS with age (Figure 7), and LA with age. Given two

assessments against the age variable, significant effects had to

meet p < 0.05/2 = 0.025.

Results

Behavioral Data

All 17 subjects produced keypress data with value function

graphs consistent with relative preference theory (Breiter and

Kim, 2008; Kim et al., 2010; Figures 3, 4). All graphs produced

LA computations (Figure 3), although five subjects had s-/s+

ratios that were > 2 standard deviations above or below the

cohort mean, and thus were considered outliers. The LA estimate

for remaining subjects was 2.06 + 0.36 (mean + SE) (Figure 4),

and the confidence interval of this group overlapped the LA

mean of 2.25, published by Tversky and Kahneman (1992). The

regression of LA to age showed a non-significant relationship

(p > 0.1).

Neuroimaging Data

fMRI data showed significant motion artifacts in two subjects.

In the remaining subjects, significant fMRI activation was

observed in the VS/NAc bilaterally in the majority of

Frontiers in Human Neuroscience | www.frontiersin.org 6 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

FIGURE 5 | fMRI Masks and Activation for Neural Differential Sensitivity

(NDS). The region of interest mask used for the VS/NAc is shown above in

color with red circles, with the statistics for the subtraction of negative stimuli vs.

positive stimuli in pseudocolor shown below. The segmented VS/NAc is shown

in light green in the images below, superimposed on the gray-tone ICBM152 T1

template. The light green segmentations are overlaid with the orange-red

statistical map for NDS. Overlap of segmented anatomy and statistical

activation map is outlined in black, with yellow inside the closed line.

subjects (see segmentation-based masks of the VS/NAc and

group statistical map in Figure 5). In individuals, voxels

of activation with p < 0.05, z = 1.96 that overlapped

segmentation of the VS/NAc were used to sample BOLD

signal representing NDS to losses relative to gains. Across

subjects, we found that activation to negative stimuli in left

and right NAc was significantly greater than activation to

positive stimuli (Figure 5). This signal was used for a control

analysis of the correlation of NDS relative to behavioral LA

(2,11)

= 0.64, p < 0.04). Without outliers (>2 SD from

mean NDS), we found the same relationship (Figure 6)

reported by others (Tom et al., 2007). When we assessed

the relationship of NDS in the VS/NAc to age, we observed

a significant positive correlation (F

(2,16)

= 9.01, p < 0.009)

(Figure 7).

Discussion

Synopsis

This study showed that a validated keypress paradigm that

allowed subjects to trade effort for view-time of emotional faces

(Ekman and Friesen, 1976), produced a relationship between the

mean (K, Figure 3) and pattern of keypressing (H, Figure 3)

consistent with previous reports using a beauty stimulus set,

the International Affective Picture Set, and food stimuli (Breiter

and Kim, 2008; Kim et al., 2010). This relationship from

the picture-based keypress task produced a ratio between the

slope of the avoidance value function (s-) and slope of the

approach value function (s+) as a LA measure quite close to that

reported by Tversky and Kahneman (1992), who used a monetary

decision task.

The keypress-based LA measure correlated with a measure

of NDS from the VS/NAc, in similar fashion to that reported

by others using a monetary choice task (Tom et al., 2007), and

meeting our two criteria for model-based fMRI effects. Although

correlation of LA with age was non-significant, the correlation

of NDS with age showed a significant positive relationship.

Two sides of the three-way correlation test between LA, NDS,

and age were significant, suggesting that as individuals age

their NDS also increases but their behavioral index of LA

does not. The discussion that follows will evaluate potential

hypotheses and implications of these findings, along with

important caveats.

Neurocompensation and Other Hypotheses

The relationship between age and NDS in the absence of a

significant relationship between age and behavioral LA suggests

interesting hypotheses about the neural processing of LA in

relation to age. First, Seidman et al. (1998) have shown

that subject IQ is inversely related to brain activation levels

across individuals, suggesting that the brain compensates with

greater neural activity when functional capacity is lower. Given

that aging is known to be associated with brain atrophy

(e.g., Raz et al., 2005) and cognitive decline (e.g., Kensinger

and Corkin, 2008), it is inferred that the functional capacity of

the brain declines with age. The findings in this study suggest

Frontiers in Human Neuroscience | www.frontiersin.org 7 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

FIGURE 6 | Association between LA and NDS from BOLD signal

collected by fMRI. NDS represents the difference between BOLD

signal in the left VS/NAc for negative stimuli (Angry, Fearful, and Sad

faces) minus positive stimuli (Happy faces). LA is represented as the

absolute value of s-/s+ from the relative preference graphs of each

subject as shown in Figure 3. The line of best ﬁt is shown from this

association.

FIGURE 7 | Association of NDS from BOLD signal collected by fMRI

with age. The line of best ﬁt is shown from this association.

the hypothesis that as an individual ages, the neural differential

between losses and gains increases to achieve the same level

of behavioral LA. This perspective is consistent with recent

research on aging wherein preserved cognitive function in the

context of age-related brain activity may be thought to represent

a neurocompensation mechanism (Meunier et al., 2014).

Consistent with the neural efficiency (neurocompensation)

hypothesis, age-related differences in reward processing have

led to evolutionary theories about the changing costs and need

to acquire resources over the lifespan. Namely, youth have

optimal health and a lack of resources, which may drive the

early aggressive pursuit of rewards (Spear, 2000; Somerville

and Casey, 2010); while as age progresses, biological decline

drives the need to minimize effort and protect what has been

acquired (Heckhausen, 1997; Ebner et al., 2006; Heckhausen

et al., 2010).

Consistent with these observations, age-related brand

loyalty (Lambert-Pandraud and Laurent, 2010) has been

attributed to an increased aversion to risk associated with change

(Montgomery and Wernerfelt, 1992; Erdem et al., 2006) leading

to investigations of age-related changes in decision making and

reward processing across the lifespan (Mata and Nunes, 2010;

Eppinger et al., 2011, 2012; Mata et al., 2011; Weller et al., 2011;

Paulsen et al., 2012a,b; Barkley-Levenson et al., 2013). In the

current study, we observed no significant differences in LA

behavior, but did find relatively early (i.e., range 20–55 years

old) age-related increases in NDS. The parametric increase in

NDS with age may indicate additional neural effort is required

to obtain the same behavioral outcomes as one ages. Previous

studies support this notion; for example, aging populations show

bilateral activation patterns within homologous areas of the PFC

while their younger counterparts achieve the same performance

with singular lateralized activations (Reuter-Lorenz et al., 1999;

Cabeza et al., 2002), suggesting compensatory mechanisms in

response to neural senescence (Cabeza et al., 2002; Park and

Reuter-Lorenz, 2009; Daselaar et al., 2015). In addition, recent

imaging work on cognitive function suggests that behavioral

outcomes reflect interactions between age, neural efficiency and

processing load (Cappell et al., 2010; Vallesi et al., 2011; Turner

and Spreng, 2012).

Alternative hypotheses might consider additional age-related

asymmetries in the larger decision making network that interact

with the VS/NAc in such a way that the response to LA is either

uninhibited or amplified. For example, on the other end of the

age spectrum, recent imaging work has shown asynchronous

developments of the VS/NAc and PFC parallels adolescents’

predisposition to engage in high-risk behaviors (Steinberg, 2008;

Van Leijenhorst et al., 2010; Blakemore and Robbins, 2012;

Barkley-Levenson and Galván, 2014), however the relationship

only becomes apparent when the asynchronous trajectories

of PFC and VS/NAc interact within a specific window of

development. The observed VS/NAc response to LA may reflect

a similar interaction, albeit in areas that did not reach thresholds

of activation that changed behavior in the current experiment,

either because of the age group examined, sensitivity of the

task, or sensitivity of the imaging paradigm. For example, given

the limitations of imaging relatively small and deep subcortical

structures, particularly in small sample sizes, additional areas

may also contribute to age related changes in processing LA,

such as functional differences between the dorsal and ventral

striatum, where activations in the dorsal striatum may have

been sub-threshold in our data, or may only be associated with

anticipation paradigms (Tom et al., 2007) related to LA.

Limitations

The differences in the relationship between age and neural

responses vs. age and behavior suggests future work might

benefit from examining VS/NAc response against other regions

implicated in decision making, which appear to contribute to

Frontiers in Human Neuroscience | www.frontiersin.org 8 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

LA processing (e.g., consider the amygdala findings in Lee et al.,

2012, RO1 submission MH098867, and Canessa et al., 2013). In

addition, while the sample size for this study is comparable to

similar such work (e.g., Tom et al., 2007), much more needs

to be done to develop a better understanding of age effects.

In this study, we have a sample size of 17 subjects in the age

range of 20–55 years. Future studies might include a wider age

range to facilitate determining if there are distinct clusters of

subjects supporting one or both of the hypotheses regarding age

and neural function, and increase the number of participants to

obtain a more densely packed gradient of age distributions. With

a more comprehensive and dense parametric gradient of age, we

speculate that we would see stronger correlations with VS/NAc

activation, however we may also find non-linear functions, or a

step function based on plateaus within certain groups of ages.

The use of a model-based approach to fMRI in this study

also needs to be carefully considered in terms of its pros and

cons. Given concerns about VS/NAc involvement with motor

preparation (e.g., Florio et al., 1999) and alteration with motor

illnesses such as Parkinsonism (e.g., Aarts et al., 2014; Payer

et al., 2015), we sought to avoid motoric influence in cognitive

imaging studies of the VS/NAc as occurs by definition with

monetary choice paradigms (e.g., Tom et al., 2007; Canessa et al.,

2013). The implicit assumption in our model-based approach

was that emotional responses would drive keypress behavior,

and thus the keypress responses would reflect the assessment

of the faces being passively viewed in the scanner (see Perlis

et al., 2008; Gasic et al., 2009). Such considerations about the

use of valuation systems for processing even passively presented

stimuli have been discussed by others (e.g., Hayes and Northoff,

2012), and similar considerations have been employed in other

model-based fMRI analyses (e.g., Mittner et al., 2014; Wang and

Voss, 2014; White et al., 2014; Xu et al., 2015). In this study,

the model had two components: (i) testing if NDS would occur

in the VS/NAc; and (ii) assessing if NDS-related fMRI signal

in the VS/NAc correlated significantly with LA behavior across

subjects. It is important to note that the behavioral process

studied is one of only two behavioral models (the other being

circadian rhythms) that have been tested to Feynman criteria

for lawfulness (Kim et al., 2010). Even with such considerations,

there is always the possibility that our implicit assumption does

not hold, and the keypress behavior outside the scanner relates

to a completely different cognitive function than that occurring

in the scanner to passive viewing, such as might be involved with

the default network.

Conclusion

In this study we used a more general interpretation of LA,

as an overweighting of aversion (toward negative stimuli)

relative to approach (toward positive stimuli), and found

close concordance with prior published LA measures (Tversky

and Kahneman, 1992; Abdellaoui et al., 2007). This LA

measure correlated significantly with the differential processing

of negative outcomes relative to positive ones by the brain

consistent with other studies (Tom et al., 2007; Lee et al., 2012;

Canessa et al., 2013). The absence of LA correlation with age,

but presence of age correlation with brain differential sensitivity

(i.e., NDS) supports a neural efficiency or neurocompensation

hypothesis regarding the effects of age on the process of LA. The

data from this study may have implications for future research

using non-monetary stimuli, such as consumables, marketing

options, or market communications. Crossing this consideration

with the aging data, the results of our study suggest that

marketing communications and brands research that target older

adults might focus on the cost of neural processing in marketing

communications. Future fMRI work might specifically target an

older adult population in order to examine how brain circuits

underlying decision-making may be altered in persons over the

age of 65. Future fMRI work might also examine the interactions

of the nature of information with other variables such as amount

of information (Mata and Nunes, 2010) and uncertainty, and

thus give us a better understanding of the sub-processes that

occur when older adults make decisions.

Acknowledgments

This work was supported by grants to HCB (#14118, 026002,

026104, 027804) from the NIDA, Bethesda, MD, and grants

(DABK39-03-0098 and DABK39-03-C-0098; The Phenotype

Genotype Project in Addiction and Mood Disorder) from

the Office of National Drug Control Policy—Counterdrug

Technology Assessment Center, Washington, D.C. Further

support was provided to HCB by the Warren Wright Adolescent

Center at Northwestern Memorial Hospital and Northwestern

University, Chicago IL, USA. Support was also provided by a

grant to AJB (#052368) from NINDS, Washington, D.C., and a

grant from the Dystonia Medical Research Foundation to AJB.

The funders had no role in study design, data collection and

analysis, decision to publish, or preparation of the manuscript.

References

Aarts, E., Nusselein, A. A., Smittenaar, P., Helmich, R. C., Bloem, B. R., and Cools,

R. (2014). Greater striatal responses to medication in Parkinson’s disease

are associated with better task-switching but worse reward performance.

Neuropsychologia 62, 390–397. doi: 10.1016/j.neuropsychologia.2014.

05.023

Abdellaoui, M., Bleichrodt, H., and Paraschiv, C. (2007). Loss aversion under

prospect theory: a parameter-free measurement. Manag. Sci. 53, 1659–1674.

doi: 10.1287/mnsc.1070.0711

Aharon, I., Etcoff, N., Ariely, D., Chabris, C. F., O’Connor, E., and

Breiter, H. C. (2001). Beautiful faces have variable reward value: fMRI

and behavioral evidence. Neuron 32, 537–551. doi: 10.1016/s0896-6273(01)

00491-3

Ariely, D., Huber, J., and Wertenbroch, K. (2005). When do losses loom

larger than gains? J. Mark. Res. 42, 134–138. doi: 10.1509/jmkr.42.2.134.

62283

Avnet, T., and Higgins, E. T. (2006). How regulatory fit affects value in

consumer choices and opinions. J. Mark. Res. 43, 1–10. doi: 10.1509/jmkr.

43.1.1

Frontiers in Human Neuroscience | www.frontiersin.org 9 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

Barkley-Levenson, E., and Galván, A. (2014). Neural representation of expected

value in the adolescent brain. Proc. Natl. Acad. Sci. 111, 1646–1651. doi: 10.

1073/pnas.1319762111

Barkley-Levenson, E. E., Van Leijenhorst, L., and Galván, A. (2013). Behavioral

and neural correlates of loss aversion and risk avoidance in adolescents

and adults. Dev. Cogn. Neurosci. 3, 72–83. doi: 10.1016/j.dcn.2012.

09.007

Becerra, L., Breiter, H. C., Wise, R., Gonzalez, R. G., and Borsook, D. (2001).

Reward circuitry activation by noxious thermal stimuli. Neuron 32, 927–946.

doi: 10.1016/s0896-6273(01)00533-5

Benson, V., and Marano, M. A. (1998). Current estimates from the National Health

Interview Survey, 1995. Vital Health Stat. 10 199, 1–428.

Blakemore, S.-J., and Robbins, T. W. (2012). Decision-making in the adolescent

brain. Nat. Neurosci. 15, 1184–1191. doi: 10.1038/nn.3177

Blood, A. J., Zatorre, R. J., Bermudez, P., and Evans, A. C. (1999). Emotional

responses to pleasant and unpleasant music correlate with activity in paralimbic

brain regions. Nat. Neurosci. 2, 382–387. doi: 10.1038/7299

Botwinick, J. (1978). Aging and Behavior: A Comprehensive Integration of Research

Findings. 2nd Edn. New York: Springer Publication.

Breiter, H. C., Aharon, I., Kahneman, D., Dale, A., and Shizgal, P. (2001).

Functional imaging of neural responses to expectancy and experience

of monetary gains and losses. Neuron 30, 619–639. doi: 10.1016/s0896-

6273(01)00303-8

Breiter, H. C., Etcoff, N. L., Whalen, P. J., Kennedy, W. A., Rauch, S. L., Buckner,

R. L., et al. (1996). Response and habituation of the human amygdala during

visual processing of facial expression. Neuron 17, 875–887. doi: 10.1016/s0896-

6273(00)80219-6

Breiter, H. C., Filipek, P. A., Kennedy, D. N., Baer, L., Pitcher, D. A., Olivares,

M. J., et al. (1994). Retrocallosal white matter abnormalities in patients with

obsessive-compulsive disorder. Arch. Gen. Psychiatry 51, 663–664. doi: 10.

1001/archpsyc.1994.03950080075010

Breiter, H. C., Gasic, G. P., and Makris, N. (2006). ‘‘Imaging the neural systems

for motivated behavior and their dysfunction in neuropsychiatric illness,’’ in

Complex Systems Science in Biomedicine, eds T. S. Deisboeck and J. Y. Kresh

(New York: Springer Verlag), 763–810.

Breiter, H. C., Gollub, R. L., Weisskoff, R. M., Kennedy, D. N., Makris,

N., Berke, J. D., et al. (1997). Acute effects of cocaine on human brain

activity and emotion. Neuron 19, 591–611. doi: 10.1016/s0896-6273(00)

80374-8

Breiter, H. C., and Kim, B. W. (2008). Recurrent and robust patterns

underlying human relative preference and associations with brain circuitry plus

genetics. Design principles in biology (U Minn, institute of mathematics and

its applications). Available online at: www.ima.umn.edu/2007-2008/W4.21-

25.08/abstracts.html

Breiter, H. C., and Rosen, B. R. (1999). Functional magnetic resonance imaging of

brain reward circuitry in the human. Ann. N Y Acad. Sci. 877, 523–547. doi: 10.

1111/j.1749-6632.1999.tb09287.x

Cabeza, R., Anderson, N. D., Locantore, J. K., and McIntosh, A. R. (2002).

Aging gracefully: compensatory brain activity in high-performing older adults.

Neuroimage 17, 1394–1402. doi: 10.1006/nimg.2002.1280

Cabeza, R., Grady, C. L., Nyberg, L., McIntosh, A. R., Tulving, E., Kapur,

S., et al. (1997). Age-related differences in neural activity during memory

encoding and retrieval: a positron emission tomography study. J. Neurosci. 17,

391–400.

Canessa, N., Crespi, C., Motterlini, M., Baud-Bovy, G., Chierchia, G., Pantaleo, G.,

et al. (2013). The functional and structural neural basis of individual differences

in loss aversion. J. Neurosci. 33, 14307–14317. doi: 10.1523/JNEUROSCI.0497-

13.2013

Cappell, K. A., Gmeindl, L., and Reuter-Lorenz, P. A. (2010). Age differences in

prefontal recruitment during verbal working memory maintenance depend on

memory load. Cortex 46, 462–473. doi: 10.1016/j.cortex.2009.11.009

Cole, C., Laurent, G., Drolet, A., Ebert, J., Gutchess, A., Lambert-Pandraud, R.,

et al. (2008). Decision making and brand choice by older consumers. Mark.

Lett. 19, 355–365. doi: 10.1007/s11002-008-9058-x

Daselaar, S. M., Iyengar, V., Davis, S. W., Eklund, K., Hayes, S. M., and Cabeza,

R. E. (2015). Less wiring, more firing: low-performing older adults compensate

for impaired white matter with greater neural activity. Cereb. Cortex 25,

983–990. doi: 10.1093/cercor/bht289

Deichmann, R., Gottfried, J., Hutton, C., and Turner, R. (2003). Optimized EPI

for fMRI studies of the orbitofrontal cortex. Neuroimage 19, 430–441. doi: 10.

1016/s1053-8119(03)00073-9

Ebner, N. C., Freund, A. M., and Baltes, P. B. (2006). Developmental changes in

personal goal orientation from young to late adulthood: from striving for gains

to maintenance and prevention of losses. Psychol. Aging 21, 664–678. doi: 10.

1037/0882-7974.21.4.664

Ekman, P., and Friesen, W. V. (1976). Measuring facial movement. Environ.

Psychol. Nonverbal Behav. 1, 56–75. doi: 10.1007/bf01115465

Elman, I., Ariely, D., Mazar, N., Aharon, I., Lasko, N. B., Macklin, M. L.,

et al. (2005). Probing reward function in post-traumatic stress disorder with

beautiful facial images. Psychiatry Res. 135, 179–183. doi: 10.1016/j.psychres.

2005.04.002

Eppinger, B., Hämmerer, D., and Li, S.-C. (2011). Neuromodulation of reward-

based learning and decision making in human aging. Ann. N Y Acad. Sci. 1235,

1–17. doi: 10.1111/j.1749-6632.2011.06230.x

Eppinger, B., Nystrom, L. E., and Cohen, J. D. (2012). Reduced sensitivity to

immediate reward during decision-making in older than younger adults. PLoS

One 7:e36953. doi: 10.1371/journal.pone.0036953

Erdem, T., Swait, J., and Valenzuela, A. (2006). Brands as signals: a cross-country

validation study. J. Mark. 70, 34–49. doi: 10.1509/jmkg.2006.70.1.34

Florio, T., Capozzo, A., Nisini, A., Lupi, A., and Scarnati, E. (1999). Dopamine

denervation of specific striatal subregions differentially affects preparation and

execution of a delayed response task in the rat. Behav. Brain Res. 104, 51–62.

doi: 10.1016/s0166-4328(99)00045-5

Gasic, G. P., Smoller, J. W., Perlis, R. H., Sun, M., Lee, S., Kim, B. W., et al.

(2009). BDNF, relative preference and reward circuitry responses to emotional

communication. Am. J. Med. Genet. B Neuropsychiatr. Genet. 150B, 762–781.

doi: 10.1002/ajmg.b.30944

Good, C. D., Johnsrude, I. S., Ashburner, J., Henson, R. N., Fristen, K. J.,

and Frackowiak, R. S. (2001). A voxel-based morphometric study of ageing

in 465 normal adult human brains. Neuroimage 14, 21–36. doi: 10.1006/

nimg.2001.0786

Grady, C. L., and Craik, F. I. (2000). Changes in memory processing with age. Curr.

Opin. Neurobiol. 10, 224–231. doi: 10.1016/s0959-4388(00)00073-8

Grady, C. L., McIntosh, A. R., Horwitz, B., Maisog, J. M., Ungerleider, L. G.,

Mentis, M. J., et al. (1995). Age-related reductions in human recognition

memory due to impaired encoding. Science 269, 218–221. doi: 10.1126/science.

7618082

Grady, C. L., McIntosh, A. R., Rajah, M. N., Beig, S., and Craik, F. I. (1999). The

effects of age on the neural correlates of episodic encoding. Cereb. Cortex 9,

805–814. doi: 10.1093/cercor/9.8.805

Hayes, D. J., and Northoff, G. (2012). Common brain activations for painful

and non-painful aversive stimuli. BMC Neurosci. 13:60. doi: 10.1186/1471-

2202-13-60

Heckhausen, J. (1997). Developmental regulation across adulthood: primary and

secondary control of age-related challenges. Dev. Psychol. 33, 176–187. doi: 10.

1037/0012-1649.33.1.176

Heckhausen, J., Wrosch, C., and Schulz, R. (2010). A motivational theory of life-

span development. Psychol. Rev. 117, 32–60. doi: 10.1037/a0017668

Hedden, T., and Gabrieli, J. D. E. (2004). Insights into the ageing mind: a

view from cognitive neuroscience. Nat. Rev. Neurosci. 5, 87–96. doi: 10.1038/

nrn1323

Johnson, J. G., and Busemeyer, J. R. (2010). Decision making under risk

and uncertainty. Wiley Interdiscip. Rev. Cogn. Sci. 1, 736–749. doi: 10.

1002/wcs.76

Kahneman, D., and Tversky, A. (1979). Prospect theory: an analysis of decision

under risk. Econom. J. Econom. Soc. 47, 263–292. doi: 10.2307/1914185

Kensinger, E. A., and Corkin, S. (2008). ‘‘Amnesia: Point and Counterpoint,’’

in Learning Theory and Behavior, Vol. 1 of Learning and Memory - A

Comprehensive Reference (Vol. 4), eds R. Menzel and J. Byrne (Oxford: Elsevier

Press), 259–286.

Kim, B. W., Kennedy, D. N., Lehár, J., Lee, M. J., Blood, A. J., Lee, S.,

et al. (2010). Recurrent, robust and scalable patterns underlie human

approach and avoidance. PLoS One 5:e10613. doi: 10.1371/journal.pone.

0010613

Kober, H., Barrett, L. F., Joseph, J., Bliss-Moreau, E., Lindquist, K., and Wager,

T. D. (2008). Functional grouping and cortical-subcortical interactions in

Frontiers in Human Neuroscience | www.frontiersin.org 10 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

emotion: a meta-analysis of neuroimaging studies. Neuroimage 42, 998–1031.

doi: 10.1016/j.neuroimage.2008.03.059

Lambert-Pandraud, R., and Laurent, G. (2010). ‘‘Impact of age on brand

choice,’’ in The Aging Consumer: Perspectives from Psychology and Economics,

eds A. Drolet, N. Schwartz and C. Yoon (Routledge: Psychology Press),

191–208, 18.

Lee, M. J., Gilman, J., Kuhnen, C., Lee, S., Kuster, J., Washburn, J., et al. (2012).

‘‘Amygdala volume is related to the slope of the negative utility function,’’

in 24th Annual Convention of the Association for Psychological Science

(Chicago, IL).

Levy, B., Ariely, D., Mazar, N., Chi, W., Lukas, S., and Elman, I. (2008). Gender

differences in the motivational processing of facial beauty. Learn. Motiv. 39,

136–145. doi: 10.1016/j.lmot.2007.09.002

Li, Y., Baldassi, M., Johnson, E. J., and Weber, E. U. (2013). Complementary

cognitive capabilities, economic decision making and aging. Psychol. Aging 28,

595–613. doi: 10.1037/a0034172

Madden, D. J., Turkington, T. G., Provenzale, J. M., Denny, L. L., Hawk,

T. C., Gottlob, L. R., et al. (1999). Adult age differences in the

functional neuroanatomy of verbal recognition memory. Hum. Brain

Mapp. 7, 115–135. doi: 10.1002/(sici)1097-0193(1999)7:2<115::aid-hbm5>3.0.

co;2-n

Makris, N., Gasic, G. P., Kennedy, D. N., Hodge, S. M., Kaiser, J. R., Lee, M. J.,

et al. (2008). Cortical thickness abnormalities in cocaine addiction—a reflection

of both drug use and a pre-existing disposition to drug abuse? Neuron 60,

174–188. doi: 10.1016/j.neuron.2008.08.011

Makris, N., Gasic, G. P., Seidman, L. J., Goldstein, J. M., Gastfriend, D. R., Elman,

I., et al. (2004). Decreased absolute amygdala volume in cocaine addicts. Neuron

44, 729–740. doi: 10.1016/j.neuron.2004.10.027

Mata, R., Josef, A. K., Samanez-Larkin, G. R., and Hertwig, R. (2011). Age

differences in risky choice: a meta-analysis. Ann. N Y Acad. Sci. 1235, 18–29.

doi: 10.1111/j.1749-6632.2011.06200.x

Mata, R., and Nunes, L. (2010). When less is enough: cognitive aging, information

search and decision quality in consumer choice. Psychol. Aging 25, 289–298.

doi: 10.1037/a0017927

Meunier, D., Stamatakis, E. A., and Tyler, L. K. (2014). Age-related functional

reorganization, structural changes and preserved cognition. Neurobiol. Aging

35, 42–54. doi: 10.1016/j.neurobiolaging.2013.07.003

Mittner, M., Boekel, W., Tucker, A. M., Turner, B. M., Heathcote, A., and

Forstmann, B. U. (2014). When the brain takes a break: a model-based analysis

of mind wandering. J. Neurosci. 34, 16286–16295. doi: 10.1523/JNEUROSCI.

2062-14.2014

Mohr, P. N. C., Li, S.-C., and Heekeren, H. R. (2010). Neuroeconomics

and aging: neuromodulation of economic decision making in old age.

Neurosci. Biobehav. Rev. 34, 678–688. doi: 10.1016/j.neubiorev.2009.

05.010

Montgomery, C. A., and Wernerfelt, B. (1992). Risk reduction and umbrella

branding. J. Bus. 65, 31–50. doi: 10.1086/296556

Oldfield, R. C. (1971). The assessment and analysis of handedness: the

Edinburgh inventory. Neuropsychologia 9, 97–113. doi: 10.1016/0028-3932(71)

90067-4

Paraschiv, C., and L’Haridon, O. (2008). Aversion aux pertes: origine,

composantes et implications marketing. Rech. Appl. En Mark. 23, 67–83.

doi: 10.1177/076737010802300204

Park, D. C., and Reuter-Lorenz, P. (2009). The adaptive brain: aging and

neurocognitive scaffolding. Annu. Rev. Psychol. 60, 173–196. doi: 10.

1146/annurev.psych.59.103006.093656

Paulsen, D. J., Carter, R. M., Platt, M. L., Huettel, S. A., and Brannon, E. M.

(2012a). Neurocognitive development of risk aversion from early childhood

to adulthood. Front. Hum. Neurosci. 5:178. doi: 10.3389/fnhum.2011.

00178

Paulsen, D. J., Platt, M. L., Huettel, S. A., and Brannon, E. M. (2012b). From

risk-seeking to risk-averse: the development of economic risk preference

from childhood to adulthood. Front. Psychol. 3:313. doi: 10.3389/fpsyg.2012.

00313

Payer, D. E., Guttman, M., Kish, S. J., Tong, J., Strafella, A., Zack, M., et al. (2015).

[

C]-(+)-PHNO PET imaging of dopamine D2/3 receptors in Parkinson’s

disease with impulse control disorders. Mov. Disord. 30, 160–166. doi: 10.

1002/mds.26135

Perlis, R. H., Holt, D. J., Smoller, J. W., Blood, A. J., Lee, S., Kim, B. W., et al.

(2008). Association of a Polymorphism Near CREB1 with differential aversion

processing in the Insula of Healthy participants. Arch. Gen. Psychiatry 65,

882–892. doi: 10.1001/archgenpsychiatry.2008.3

Raz, N. (2000). ‘‘Aging of the brain and its impact on cognitive performance:

integration of structural and functional findings,’’ in The Handbook of Aging

and Cognition, 2nd Edn. eds F. I. M. Craik and T. A. Salthouse (Mahwah, NJ:

Lawrence Erlbaum Associates Publishers), 1–90.

Raz, N., Lindenberger, U., Rodrigue, K. M., Kennedy, K. M., Head, D., Williamson,

A., et al. (2005). Regional brain changes in aging healthy adults: general trends,

individual differences and modifiers. Cereb. Cortex 15, 1676–1689. doi: 10.

1093/cercor/bhi044

Reuter-Lorenz, P. A. (2002). New visions of the aging mind and brain. Trends

Cogn. Sci. 6, 394–400. doi: 10.1016/s1364-6613(02)01957-5

Reuter-Lorenz, P. A., Stanczak, L., and Miller, A. C. (1999). Neural

recruitment and cognitive aging: two hemispheres are better than one,

especially as you age. Psychol. Sci. 10, 494–500. doi: 10.1111/1467-9280.

00195

Schindler, R. M., and Holbrook, M. B. (2003). Nostalgia for early experience as

a determinant of consumer preferences. Psychol. Mark. 20, 275–302. doi: 10.

1002/mar.10074

Seidman, L. J., Breiter, H. C., Goodman, J. M., Goldstein, J. M., Woodruff,

P. W., O’Craven, K., et al. (1998). A functional magnetic resonance imaging

study of auditory vigilance with low and high information processing

demands. Neuropsychology 12, 505–518. doi: 10.1037//0894-4105.12.

4.505

Shannon, C. E., and Weaver, W. (1949). The Mathematical Theory of

Communication. Volume 1. (Illinois: University of Illinois Press), 117.

Somerville, L. H., and Casey, B. J. (2010). Developmental neurobiology of cognitive

control and motivational systems. Curr. Opin. Neurobiol. 20, 236–241. doi: 10.

1016/j.conb.2010.01.006

Spear, L. P. (2000). The adolescent brain and age-related behavioral

manifestations. Neurosci. Biobehav. Rev. 24, 417–463. doi: 10.1016/s0149-

7634(00)00014-2

Stebbins, G. T., Carrillo, M. C., Dorfman, J., Dirksen, C., Desmond, J. E., Turner,

D. A., et al. (2002). Aging effects on memory encoding in the frontal lobes.

Psychol. Aging 17, 44–55. doi: 10.1037//0882-7974.17.1.44

Steinberg, L. (2008). A social neuroscience perspective on adolescent risk-taking.

Dev. Rev. 28, 78–106. doi: 10.1016/j.dr.2007.08.002

Strauss, M., Makris, N., Aharon, I., Vangel, M., Goodman, J., Kennedy, D., et al.

(2005). fMRI of sensitization to angry faces. Neuroimage 26, 389–413. doi: 10.

1016/j.neuroimage.2005.01.053

Tom, S. M., Fox, C. R., Trepel, C., and Poldrack, R. A. (2007). The neural basis

of loss aversion in decision-making under risk. Science 315, 515–518. doi: 10.

1126/science.1134239

Turner, G. R., and Spreng, R. N. (2012). Executive functions and neurocognitive

aging: dissociable patterns of brain activity. Neurobiol. Aging 33,

826.e1–826.e13. doi: 10.1016/j.neurobiolaging.2011.06.005

Tversky, A., and Kahneman, D. (1991). Loss aversion in riskless choice: a

reference-dependent model. Q. J. Econ. 106, 1039–1061. doi: 10.2307/293

7956

Tversky, A., and Kahneman, D. (1992). Advances in prospect theory: cumulative

representation of uncertainty. J. Risk Uncertain. 5, 297–323. doi: 10.

1007/bf00122574

Vallesi, A., McIntosh, A. R., and Stuss, D. T. (2011). Overrecruitment in the aging

brain as a function of task demands: evidence for a compensatory view. J. Cogn.

Neurosci. 23, 801–815. doi: 10.1162/jocn.2010.21490

Van Leijenhorst, L., Gunther Moor, B., Op de Macks, Z. A., Rombouts,

S. A. R. B., Westenberg, P. M., and Crone, E. A. (2010). Adolescent

risky decision-making: neurocognitive development of reward and

control regions. Neuroimage 51, 345–355. doi: 10.1016/j.neuroimage.2010.

02.038

Wang, J. X., and Voss, J. L. (2014). Brain networks for exploration decisions

utilizing distinct modeled information types during contextual learning.

Neuron 82, 1171–1182. doi: 10.1016/j.neuron.2014.04.028

Weller, J. A., Levin, I. P., and Denburg, N. L. (2011). Trajectory of risky decision

making for potential gains and losses from ages 5 to 85. J. Behav. Decis. Mak.

24, 331–344. doi: 10.1002/bdm.690

Frontiers in Human Neuroscience | www.frontiersin.org 11 April 2015 | Volume 9 | Article 176

Viswanathan et al. Age and loss aversion

White, C. N., Congdon, E., Mumford, J. A., Karlsgodt, K. H., Sabb, F. W., Freimer,

N. B., et al. (2014). Decomposing decision components in the stop-signal task:

a model-based approach to individual differences in inhibitory control. J. Cogn.

Neurosci. 26, 1601–1614. doi: 10.1162/jocn_a_00567

Xu, J., Calhoun, V. D., and Potenza, M. N. (2015). The absence of task-related

increases in BOLD signal does not equate to absence of task-related brain

activation. J. Neurosci. Methods 240, 125–127. doi: 10.1016/j.jneumeth.2014.

11.002

Yamamoto, R., Ariely, D., Chi, W., Langleben, D. D., and Elman, I. (2009).

Gender differences in the motivational processing of babies are determined

by their facial attractiveness. PLoS One 4:e6042. doi: 10.1371/journal.pone.

0006042

Conﬂict of Interest Statement: The authors declare that the research was

conducted in the absence of any commercial or financial relationships that could

be construed as a potential conflict of interest.

Raman, Lee, Kuster, Stern and Calder, Mulhern, Blood, Breiter. This is an open-

access article distributed under the terms of the Creative Commons Attribution

License (CC BY). The use, distribution and reproduction in other forums is

permitted, provided the original author(s) or licensor are credited and that the

original publication in this journal is cited, in accordance with accepted academic

practice. No use, distribution or reproduction is permitted which does not comply

with these terms.

Frontiers in Human Neuroscience | www.frontiersin.org 12 April 2015 | Volume 9 | Article 176

ORIGINAL RESEARCH ARTICLE

published: 24 December 2013

doi: 10.3389/fnhum.2013.00881

On the interpretation of synchronization in EEG

hyperscanning studies: a cautionary note

Adrian P. Burgess

Aston Brain Centre, School of Life and Health Sciences, Aston University, Birmingham, UK

Edited by:

Sven Braeutigam, University of

Oxford, UK

Reviewed by:

Martin Vinck, University of

Amsterdam, Netherlands

Douglas D. Potter, University of

Dundee, UK

*Correspondence:

Adrian P. Burgess, Aston Brain

Centre, School of Life and Health

Sciences, Aston University, Aston

Triangle, Birmingham B4 7ET, UK

e-mail: a.p.burgess@aston.ac.uk

EEG Hyperscanning is a method for studying two or more individuals simultaneously with

the objective of elucidating how co-variations in their neural activity (i.e., hyperconnectivity)

are inﬂuenced by their behavioral and social interactions. The aim of this study was to

compare the performance of different hyper-connectivity measures using (i) simulated

data, where the degree of coupling could be systematically manipulated, and (ii)

individually recorded human EEG combined into pseudo-pairs of participants where no

hyper-connections could exist. With simulated data we found that each of the most widely

used measures of hyperconnectivity were biased and detected hyper-connections where

none existed. With pseudo-pairs of human data we found spurious hyper-connections

that arose because there were genuine similarities between the EEG recorded from

different people independently but under the same experimental conditions. Speciﬁcally,

there were systematic differences between experimental conditions in terms of the

rhythmicity of the EEG that were common across participants. As any imbalance between

experimental conditions in terms of stimulus presentation or movement may affect

the rhythmicity of the EEG, this problem could apply in many hyperscanning contexts.

Furthermore, as these spurious hyper-connections reﬂected real similarities between

the EEGs, they were not Type-1 errors that could be overcome by some appropriate

statistical control. However, some measures that have not previously been used in

hyperconnectivity studies, notably the circular correlation co-efﬁcient (CCorr), were less

susceptible to detecting spurious hyper-connections of this type. The reason for this

advantage in performance is discussed and the use of the CCorr as an alternative measure

of hyperconnectivity is advocated.

Keywords: electroencephalography, hyperscanning, phase synchronization, social neuroscience, inter-brain

connectivity, Phase Locking Value

INTRODUCTION

Over the last decade, the development of techniques that allow

the measurement of neural activity from two or more individu-

als simultaneously, known as hyperscanning, has been heralded

with some justiﬁcation as a promising new ﬁeld in social neu-

roscience (Dumas, 2011; Dumas et al., 2011; Sanger et al., 2011;

Babiloni and Astolﬁ, 2012; Konvalinka and Roepstorff, 2012).

Hyperscanning methods have been used in many different social

contexts but all involve the simultaneous recording of brain activ-

ity from two or more individuals with a view to determining how

co-variation in their neural activity is related to their behavioral

and social interactions and this work has resulted in multiple

claims that neural coupling between people is increased during

social interaction. In contrast, there has been little attempt to

determine how valid the methods used to measure connectivity

are in this context and this paper is one attempt to redress that

omission.

The ﬁrst true hyperscanning study was reported by Montague

et al. (2002) using two linked fMRI scanners with two individuals

playing a variant of the children’s guessing game, “handy-dandy.”

Other studies have used near-Infrared Spectroscopy (Funane

et al., 2011) and there is also a single case study demonstrating

the feasibility of hyperscanning using magnetoencephalography

(Baess et al., 2012). Most studies, however, have relied upon EEG

which, is not only more readily available than other methods but

is also better suited for use in naturalistic social settings, and these

are the focus of this paper.

The ﬁrst EEG hyperscanning study was reported by Babiloni

et al. (2006) and involved sets of four individuals playing

Tressette, a bridge-like game. Since then, there have b een 30

more EEG publications that meet the deﬁnition of hyperscanning

coming from more than 20 independent studies have claimed

increased neural coupling between people engaged in social inter-

action (Babiloni et al., 2006, 2007a,b, 2011, 2012; Flexer and

Makeig, 2007; Tognoli et al., 2007, 2011a,b; Chung et al., 2008;

Tognoli, 2008; Yun et al., 2008; Astolﬁ et al., 2009, 2010a,b,c,

2011a,b, 2012; Lindenberger et al., 2009; Dumas et al., 2010,

2012a,b; Fallani et al., 2010; Dodel et al., 2011; Lachat et al., 2012;

Naeem et al., 2012a,b; Sanger et al., 2012, 2013; Yun et al., 2012;

Kawasaki et al., 2013). The methods used to establish neural cou-

pling between people have been very consistent and nearly all

studies have used one of three methods: (i) covariance in ampli-

tude or power, (ii) Partial Directed Coherence (PDC); (Baccala

and Sameshima, 2001), and (iii) phase synchrony, mostly the

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

HUMAN NEUROSCIENCE

Burgess EEG hyperscanning: a cautionary note

Phase-Locking Value (PLV) (Lachaux et al., 1999)oravariant

thereof.

The most frequently used method for demonstrating brain-

to-brain coupling between socially interacting individuals, used

in 12 reports, has been to show that there are contiguous, or near

contiguous changes in EEG amplitude or power (Babiloni et al.,

2007b, 2011, 2012; Tognoli et al., 2007; Yun et al., 2008; Astolﬁ

et al., 2009; Dumas et al., 2012b; Lachat et al., 2012; Naeem et al.,

2012a,b; Yun et al., 2012; Kawasaki et al., 2013). In most cases,

this EEG amplitude/power has been estimated from event-related

changes or from FFT. Showing that there are co-variances in EEG

power is a weak form of association and although it is suggestive

of neural coupling, it is by no means conclusive.

The second most commonly used method has been that of

PDC which was the approach used in the very ﬁrst EEG hyper-

scanning study (Babiloni et al., 2006)andhasbeenusedinatleast

nine further studies since (Babiloni et al., 2007a,b; Astolﬁ et al.,

2010a,b,c, 2011a,b, 2012; Fallani et al., 2010). PDC is based on

multivariate autoregressive modeling and Granger Causality and

is designed to be able to show the direction of ﬂow of information

(linear) between two systems (Baccala and Sameshima, 2001). As

such, PDC seems ideally suited to role of identifying inter-brain

coupling in hyperscanning studies, at least in those cases where

when one person’s behavior is driving another’s. However, both

PDC and Granger causality are not without their critics. Friston

(2011), for example, provides a critique of the use of Granger

causality in fMRI research and, some of the limitations he men-

tions apply equally well to EEG research. It is certainly the case

that, as Konvalinka and Roepstorff (2012) have observed, the

results of PDC in hyperscanning studies have not replicated well,

but whether this is related to the use of PDC, or to some other

cause, is not clear.

The ﬁnal class of measures of brain-to-brain coupling all

involve measures of phase synchrony (Lindenberger et al., 2009;

Dumas et al., 2010, 2012a; Sanger et al., 2012, 2013; Yun et al.,

2012). The ﬁrst use of phase synchronization as a measure of

coupling with electrophysiological data was by Tass et al. (1998),

who deﬁned synchronization as occurring when



n, m



< const,

where const is some suitably small value, n and m are integers,

n, m

(t) is the phase difference, nφ

(

)

− mφ

(

)

and φ

1, 2

are

the phases of the two oscillators. The most widely used index

of phase locking adopted in hyperscanning studies has been the

Phase Locking Value (PLV) (Lachaux et al., 1999)whichisa

measure that seems well suited for capturing the rapid ﬂow of

information between people in social situations. Interestingly,

some hyperscanning studies have used PLV to characterize behav-

ioral interactions even when they have used other measures of

coupling for the EEG (e.g., Tognoli et al., 2007).

Although both PDC and PLV have been used to measure cou-

pling between cortical oscillations recorded in the EEG from two

or more different people, what they actually measure is quite dif-

ferent in each case and, for this reason, it is worth reviewing

what is meant by synchronization. The ﬁrst scientiﬁc descrip-

tion of synchronization came in 1665 from Christiaan Huygens

who wrote a letter to the Royal Society in which he described “an

odd kind of sympathy” in which the pendulums of identical clocks

mounted on the same support came to swing exactly out of phase

(i.e., anti-phase) regardless of the phase they had been in when

they had been set running (Pikovsky et al., 2001; Klarreich, 2002).

The explanation of this phenomenon is that the swing of the pen-

dulum in one clock induced small movements in the support

from which the clocks were suspended that would slightly alter the

swing of the pendulum of the second clock. At the same time, the

pendulum of the second clock would induce movements in the

support that affected the swing of the pendulum in the ﬁrst clock.

These small mutual nudges would continue to shift the phase of

each pendulum until they came to a point where the nudge from

one would exactly counterbalance the nudge from the other and

this would occur when the pendulums were precisely anti-phase.

In modern terms, the two clocks were in a system of reciprocal

negative feedback and would continue to change until the sys-

tem reached the state of minimum energy transfer between the

two. Minimum information transfer (in fact, zero energy trans-

fer) occurs in the anti-phase condition. An example of in-phase

reciprocal synchronization is shown in Figure 1A.

True synchronization then, is of interest in neuroscience

because it is a reliable marker of the ﬂow of information

between elements of a system. Simply observing a consistent

phase relationship between two oscillators (clocks, human brains

etc.), however, does not necessarily mean that they are in the

FIGURE 1 | Types of synchrony. (A) Shows “reciprocal” synchronization

whereby the pendulums of the clocks swing in phase because there is

reciprocal inﬂuence between the two; (B) shows “induced”

synchronization whereby the phase of the pendulums of both clocks are

inﬂuenced by a common external driver; (C) shows “driven”

synchronization whereby the pendulum of one clock inﬂuences the phase

of the pendulum of the other clock without any reciprocal inﬂuence; (D)

shows “coincidental” synchronization where there is no coupling between

the clocks but the pendulums remain in a ﬁxed phase relationship to each

other because they both swing at the same frequency.

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

same condition of reciprocal information exchange displayed by

Huygens’ clocks. Synchronization might also occur if both clocks

are driven by some external inﬂuence as in Figure 1B.Inhyper-

scanning experiments, this might occur if the participants simul-

taneously experience the same stimuli such as watching a movie

together, even though they are not directly interacting (Hasson

et al., 2008). Alternatively, the inﬂuence between oscillators might

be one-way with one oscillator driving another, Figure 1C,which

is exactly the type of coupling that PDC is designed to iden-

tify. Each of these types of synchronization might be of interest,

depending upon the context of the study, and it would often be of

interest to know which type of synchronization is being observed.

In practice, however, these different types of synchronization may

be difﬁcult to tell apart.

There is a fourth type of synchrony which is not really syn-

chronization at all: coincidental synchrony, Figure 1D.Thisisa

phenomenon which is generally of no interest and, in the context

of hyperscanning, has nuisance value only. Unfortunately, it is not

a rare phenomenon. Had Huygens’s clocks been too far apart to

inﬂuence each other, they would have remained in the same ﬁxed

phase relationship to each other indeﬁnitely. Over time, small dif-

ferences between the clocks would lead to a gradual shift in phase

but, at least over short periods of time, the phase difference would

be nearly constant. In general, two oscillators will show a consis-

tent phase relationship whenever they share a common frequency

of oscillation. To put this in the context of the brain, consider

two adults, each with a dominant alpha rhythm of ∼10 Hz sit-

ting in isolation in separate rooms. If we were to measure their

EEG, we could expect to see a fairly consistent phase relation-

ship between their alpha rhythms, at least over short time scales,

even though there is no communication between them. This sit-

uation is exactly the same as the example of the identical but

unconnected pendulum clocks and stems solely from the fact that

they share a common frequency of oscillation. It follows from this

that simply observing a consistent phase relationship does not

imply synchronization or information exchange or, as Pikovsky

et al. (2001) put it, “synchronous variation of two variables does not

necessarily imply synchronization.” The critical feature of synchro-

nizationisnotthattheoscillatorsaresynchronousbutthatthere

is “...adjustment of their rhythms, or appearance of phase locking

due to interaction”(Tass et al., 1998).

To put this more formally, two oscillators can be said to be syn-

chronized if deviations from the regular oscillatory cycle of one

oscillator provides information about deviations in the oscillatory

cycle of the other. Such a deﬁnition suggests that a measure of the

co-variation or correlation between oscillators might sometimes

be more useful. It is for this reason, that most hyperscanning stud-

ies do not simply measure phase coupling in the EEG between

individuals but compare the degree of coupling between differ-

ent experimental conditions. In the best studies, the experimental

conditions are identical in every way except that in one case the

participants are socially engaged and in the other they are not. In

practice, however, this level of experimental control is difﬁcult to

achieve.

The aim of this paper is to examine the performance

of currently used measures of phase synchronization in

hyperconnectivity studies (PDC and PLV) and compare them

with alternative measures including coherence (COH), the cir-

cular correlation co-efﬁcient (CCorr) and Kraskov’s Mutual

Information estimator (KMI) (Kraskov et al., 2004). Good per-

formance, in this context, is deﬁned by three qualities. First, the

measure should be unbiased and have a low root mean squared

error of estimation (RMSE). Speciﬁcally, when the true connec-

tivity, r = 0, the estimated connectivity should be zero or very

close to it. Second, the estimate of connectivity should increase

monotonically as r increases and third, the estimate of coupling

strength between two channels should be independent of the dis-

tribution of the signal in either of the constituent channels. In

particular, the estimate of coupling strength should be insensi-

tive to changes in the variance of the marginal distributions of

deviations from the expected phase in either channel.

The ﬁrst comparison included simulated time series where the

degree of connectivity could be systematically varied. The sec-

ond comparison compared EEG from individuals independently

recorded but analyzed as though they had been recorded as part of

a hyperscanning study. Because these EEG recordings were com-

pletely independent and there was no social contract between

participants, a good measure of hyperconnectivity should not

detect any synchronization between them. The ﬁrst example

of EEG data was from an event-related potential paradigm in

which data recorded around the time of the presentation of a

visual stimulus was used. This is analogous to induced synchrony

(Figure 1B) as there may be some apparent connectivity between

individuals because they share similar external stimulation. The

second example of EEG data was from two independent resting

state conditions in which there was no external stimulation to

induce synchrony.

MATERIALS AND METHODS

MEASURES

Five different methods for estimating functional hyper-

connectivitywereusedinthisstudy.

Coherence (COH)

COH is the traditional Fourier-based method of connectivity and

the Welch estimate of coherence is given by:

COH





k = 1





∗











k = 1





∗







k = 1





∗





(1)

where X

(

)

denotes FFT of the kth segment of the time series

x(t)atfrequencyf and * indicates the transpose and complex

conjugate. The analysis was performed using the MatLab function

mscohere.m. COH values range from 0 to +1.

Partial directed coherence (PDC)

The PDC from y to x is deﬁned as:

PDC











∗









(2)

where A

(f ) is an element in A(f ) which is the Fourier Transform

of the multivariate autoregressive (MVAR) model coefﬁcients,

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

A(t), of the time series ; a

(f )isyth column of A(f ). MVAR

and PDC analysis was performed using the Extended Multivariate

Autoregressive Modeling Toolbox for MatLab (Faes and Nollo,

2011). PDC values range from 0 to +1 but as PDC

x, y

= PDC

y, x

both are reported.

Phase locking value (PLV)

There is an unfortunate terminological confusion over the use of

the term “PLV” as, not only is it often referred to as the Phase

Locking Index (PLI) but, both “PLV” and “PLI” can refer to two

quite different measures that have equations of the identical form

but quite different meaning. The PLV, as orig inally deﬁned by

Lachaux et al. (1999), is estimated by:

PLV





k = 1

(

t, k

)

−ψ

(

t, k

)



(3)

where N isthenumberoftrials,φ

(t, n)

,isthephaseontrial,n

at time t, in channel φ and ψ

(t, n)

in channel ψ. The PLV

varies

between0and1where1indicatesperfectphaselockingand0

indicates no phase locking. This form of the PLV

is a measure of

the consistency of the phase difference and is related to the inter-

trial variance of the phase difference, σ

φ − ψ

, by the relationship

PLV

= 1 − σ

φ−ψ

. Because this form of the PLV

is based on the

phase difference across trials, it is only suitable for event-related

paradigms.

However, there is a variant of the Equation (3) that has been

frequently used in EEG hyperscanning studies which involves

averaging the instantaneous phase differences over time within a

single trial:

PLV





n = 1

(

t, n

)

−ψ

(

t, n

)



(4)

where T is the number of time points. This for m of the PLV is

essentially a measure of the int ra-trial consistency of the phase

differencebetweenchannels.Aswillbecomeclear,thissmall

difference between Equations (3) and (4) has important implica-

tions for the interpretation of EEG hyperscanning methods. In an

attempt to remove any ambiguity, we shall refer to the measure

deﬁned by Equation (3) as the trial-averaged PLV or PLV

and

that described by Equation (4) as the time-averaged PLV, PLV

PLV values range from 0 to +1.

The PLV is a measure of the consistency of the phase-difference

but, as noted above, simply observing that there is a consistent

phase relationship between two signals does not imply covariance

or information exchange or between them. Indeed, the PLV

can-

not distinguish between coincidental phase synchronization and

true phase synchronization. To see why phase difference is a poor

measure of information exchange, consider the variance of the

difference in the case of the bivariate normal distribution

which

is given by:

x − y

= σ

+ σ

− 2σ

ρ, (5)

Unfortunately, the equivalent equation for the bivariate von Mises distribu-

tion is not known.

where σ

is the variance and ρ is the correlation between the

two distributions x and y.Clearlyσ

x − y

can be small, indicating

strong association between the two variables, not only when ρ is

large but when σ

and σ

are small. This means that although

x − y

is related to ρ, it is a rather poor proxy for it and makes

no sense to measure correlation this way in such cases. The nat-

ural measure of correlation in this case is the Pearson Product

Moment Correlation Coefﬁcient which measures the covariance

of the deviations from the expected (i.e., mean) values of the two

variables.

Circular correlation coefﬁcient (CCorr)

The Pearson Product Moment Correlation Coefﬁcient is not suit-

able for use with circular distributions like phase but there are

several suitable candidates including the Circular Correlation

Coefﬁcient (CCorr) (Jammalamadaka and Sengupta, 2001),

CCorr is a direct parallel to the Pearson Product Moment

Correlation Coefﬁcient for circular data and is given by:

CCorr

φ, ψ



k = 1

sin



φ − φ



sin



ψ − ψ







k = 1

sin



φ −



sin



ψ −



(6)

where

φ and ψ are the mean directions for channels 1 and 2

respectively. For oscillatory signals, like the EEG, phase is approxi-

mately uniformly distributed and the population mean directions

are not deﬁned. However, in the case of uniform marginal dis-

tributions, any arbitrary direction can be deﬁned as the mean

without ill effect although for convenience, the sample mean

directions, φ and ψ werealwaysused.UnlikePLV

, the Circular

correlation, CCorr, is much more robust to coincidental synchro-

nization. The reason for this is that CCorr measures the circular

covariance of differences between the observed phase and the

expected (i.e., mean) phase. In the case of a perfect oscillator, the

frequency of oscillation will be constant and there will be no vari-

ance. In the case of a sinusoidal oscillation, knowing the frequency

of oscillation and its phase at any single time point provides a

complete description of its behavior. For imperfect oscillators, as

all real-world oscillators are, there will be small variations in phase

over time. However, knowing the phase of such an oscillator in its

recent past makes it possible to predict its phase in the near future.

In the case of two related channels, if one channel is slightly in

advance of its expected phase at a given time, then the phase

in the other channel is also likely to be advanced (for positively

correlated signals; the reverse for negatively correlated signals).

That is, the phase variance of the oscillators co-varies and this

is what CCorr measures. In the case of two unrelated channels,

the phase variance will not co-vary and the CCorr will be zero.

As the PLV measures the phase difference, which is a poor proxy

for phase covariance, it is likely to be poorer at discriminating

between related and unrelated signals. CCorr was measured using

the CircStat toolbox for MatLab (Berens, 2009). CCorr values

range from 0 to +1.

Kraskov mutual information (KMI)

The KMI is a non-parametric estimator of mutual informa-

tion (Kraskov et al., 2004) based nearest-neighbor method for

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

estimating entropy proposed by Kozachenko and Leonenko

(1987) cited in Beirlant et al. (1997). The KMI, adapted for use

with phase data, is given by:

φψ

= 

(

)

+ 

(

)

−



i = 1







(

)

+ 1



+ 



(

)

+ 1



(7)

where (.) is the digamma function, n

φ (i)

isthenumberofpoints

with



− φ



≤ ε (i)/2andn

(i)

isthenumberofpointswith



− ψ



≤

(

)

; ε(i) is the distance from observation i to its j

nearest neighbor and distances are measured with respect to the

maximum norm ε

(

)

= max



(

)

, φ

(

)



and N is the total

number of independent observations. In the simulations reported

here, j = 5 and the distances were angular distances.

For convenience, all mutual information values were trans-

formed to the r ange 0–1 using the relationship r =



1 − e

−2I

φψ

where, I

φ

, is the mutual information between φ and  and, r,

is the correlation from a bivariate normal distribution with the

same mutual information.

SIMULATIONS

The objective of the simulations was to generate time series when

the phase-coupling between the two could be systematically var-

ied. Phase distributions can be generated from the von Mises

distribution, a circular analog of the Gaussian distribution that

ranges from -π to +π. The von Mises distribution is deﬁned by

its mean direction, μ, and concentration, κ, which are analogous

to the Gaussian mean, μ, and the reciprocal of the variance,1/σ

respectively. Examples of the von Mises distribution for μ = 0

and varying values of κ are shown in Figure 2. T he von Mises

distribution can be generalized to two dimensions where phase

can be represented as a distribution on the surface of a torus

(Singh et al., 2002). The covariance between the two dimensions

of the bivariate von Mises distribution is controlled by a param-

eter λ. The joint probability density function is deﬁned by the 5

parameters (μ

, μ

, κ

,andλ) and from this it is a simple

matter of numerical integration to calculate the mutual infor-

mation between the two distributions (see Appendix). Given the

FIGURE 2 | Probability density functions of the von Mises distribution

for different values of concentration, κ.

probability density function of a 2-D von Mises distributions it is

a simple matter to generate random variables with any different

levels of mutual information and concentration (Figure 3)using

the acceptance/rejection method (Gentle, 1998).

To generate a time series with randomly varying phase shifts,

we ﬁrst generated an unwrapped and perfectly regular phase series

[0, 2π,3π ....nπ], and generated n independent samples from a

von Mises random distribution [φ

; φ

; ....φ

]andadded

the two together giving a new phase series [0 + φ

,2π + φ

3π + φ

,....nπ + φ

]. The von Mises random v ariables were

added as phase deviations to the expected regular phase series.

In this case, the aim was to simulate an alpha rhythm with a mean

frequency of f = 10 Hz sampled at = 500 Hz. The phase series

[0, 2π,3π ....nπ] corresponded to a time series of [0, 0.1, 0.2,

0.3. . . n/f ]s so the phase values for intermediate time points from

0ton/f seconds in 1/ second intervals were estimated by spline

interpolation. Finally, the pseudo-alpha rhythm was obtained by

taking the sine of the interpolated phase series. This created a

smoothly frequency-varying oscillation with constant amplitude

in which the variance of the frequency was determined by, κ,the

concentration parameter of the von Mises distribution. In these

simulations, therefore, 1/κ is a measure of the variance of the

marginal distributions of deviations from the expected phase. An

example is shown in Figure 4. It is a simple matter to general-

ize this process to the 2D cases using random variables a 2D-von

Mises Distribution and the degree of dependency can be con-

trolled by the parameter λ. For the simulations values of λ were

chosen to approximate bivariate correlations of [0, 0.2, 0.4, 0.6,

0.8] and the concentration values of, κ, were [0.25, 0.5, 1, 2, 4,

8]. One hundred samples of 100 s epochs of pseudo-alpha were

generated for analysis for each value of λ and κ.

In order to generate pseudo-alpha time series in which there

was a time-lagged dependency between channels, n + 1indepen-

dent samples were drawn from a 2D-von Mises distribution [φ

;

; φ

; ....φ

n + 1

] and added to the phase series giving two new

phase series [0 + f

(1, 1)

,2π + φ

(2, 1)

,3π + φ

(3, 1)

....nπ + φ

(n,1)

]

and [0 + φ

(2, 2)

,2π + φ

(3, 2)

,3π + φ

(4, 2)

,....nπ + φ

(n + 1, 2)

The rest of the procedure was identical to that for the zero-lagged

time series but with the result that the two pseudo-alpha time

series were maximally correlated with a lag of 100 ms but uncor-

related at lag 0. That is, one time series “caused” the other in the

Granger sense.

Hyperconnectivity analysis

COH was estimated for each pair of time series using Welch’s

method with non-overlapping Hamming Windows of 1024 ms

Equation (1). For PDC, an MVAR model was generated for each

100 s pair of time series using a model order determined by the

Akaike Information Criterion. The PDC was estimated from the

MVAR coefﬁcients following Equation (2). COH and PDC values

were averaged across each of the 100 randomizations.

Estimates of PLV

, CCorr, and KMI were derived from the

instantaneous phase of the time series. Instantaneous phase at

each time point in each time series was estimated from the Hilbert

Transform of the pseudo-alpha data using FFT with a window of

1024 ms and it was these estimates that were used for estimating

coupling strength. The Hilbert Transform produces a “real” and

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

FIGURE 3 | Scatterplots of the 2D-von Mises distribution for different values of concentration, κ, and correlation, r.

“imaginary” time series and the phase was estimated by φ

(

)

tan

−1



Imag

(

)

Real

(

)



. In all cases, the Hilbert-estimated phases were

very close to “true” phase values that had be en entered into the

simulation. The phase-series were divided into epochs of 1024 ms

and the PLV

and CCorr were estimated for each using Equations

(4) and (6) respectively. The resulting values were averaged across

all epochs and all randomizations. This procedure of estimat-

ing hyperconnectivity over short epochs and averaging follows

the methods reported in the literature (Lindenberger et al., 2009;

Dumas et al., 2010, 2012a; Sanger et al., 2012, 2013; Yun et al.,

2012) each of whom used segments of EEG of less than 800 ms.

As estimation of KMI assumes independent observations, the

instantaneous phase data was down-sampled to a rate equal to

the mean frequency of the signal i.e., 10 Hz. An estimate of KMI

was derived for each of the down-sampled segments of pseudo-

alpha phase data using E quation (7) and averaged across the 100

random samples.

Statistical analysis

Each of the measures of connectivity was evaluated in terms of

their bias and Root Mean Squared Error of Estimation (RMSE)

for the case where the true connectivity, r,waszero.Biasand

RMSE were deﬁned as: Bias =



k = 1

(

−



and RMSE =





k = 1

−

)

where r

is the true connectivity and

is the

estimate of the true connectivity. Note that for COH, PDC and

PLV

,wherethevaluesaredeﬁnedtobegreaterthanorequalto

zero, the Bias and RMSE are equal. Mutual information, by def-

inition must also be greater than or equal to zero but, the KMI

estimator can produce small negative values and for this reason,

Bias will not always be equal to RMSE.

HUMAN EEG

PARTICIPANTS

The data used for this study are a subset of a dataset that has

previously been reported on and full details of the experiment

are reported in Burgess (2012). Participants were 10 healthy

young adults (5 women, 5 men) recruited through advertisement

with a mean age of 25.4 years (SD = 5.8; range 20–40). Written

informed consent was obtained from all subjects and the experi-

ment was conducted as approved by the Riverside Research Ethics

Committee. All investigations were conducted according to the

principles expressed in the Declaration of Helsinki and data were

analyzed anonymously.

Procedure

EEG was recorded from participants at rest (60 s Eyes Open

Relaxed and 60 s Eyes Closed Relaxed) and as they were presented

with a series of faces. There were 90 trials which included the

presentation of a ﬁxation cross for 1000 ms followed by a photo-

graph of a face for the same duration. Each photograph was of the

head and shoulders of a man or woman with neutral emotional

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

FIGURE 4 | Simulation of the pseudo-alpha rhythm. (A) Shows the

unwrapped phase of a regular 10 Hz sinusoid (Black line) with random

phase deviations generated from the von Mises distribution (Red lines) and

added to the sinusoid at 100 ms interval. The phase series for the

pseudo-alpha rhythm is formed by a smooth line (Blue line) that connects

the points of the sinusoid + phase deviation. (B) Shows the times series

generated by taking the sine of the phase series in (A).

expressions, facing directly toward the participant. The inter-trial

interval consisted of a blank screen and randomly varied between

1000 and 2000 ms. All data were recorded from participants com-

pletely independently and at separate times. There was no social

interaction between any of the participants at any time during the

recording of these data.

Materials and equipment

Twenty-eight electrodes were positioned on the scalp using

an ECI electrode cap with electrodes placed according to the

International 10–20 system with an additional nine electrodes:

Oz, FC5/6, CP1/2, CP5/6 PO1/2. Horizontal electro-oculogram

(EOG) was recorded from the external canthus of each eye, and

the vertical EOG was recorded from the supra- to suborbit of

the left eye. Electrode impedances were all under 5 k. EEG and

EOG were ampliﬁed using a 32 channel Neuroscan Synapse-II

System. Signal bandpass was 0.1–100 Hz and the digital sampling

frequency was 500 Hz. Reference was to the left ear and converted

to average reference ofﬂine.

Data preparation

For the resting state, data were divided into consecutive epochs of

1024 ms. For the event-related paradigm, EEG was divided into

pre-stimulus and post-stimulus epochs each of 1024 ms dura-

tion. The pre-stimulus epochs included data from −1024 ms

to −1 ms and the post-stimulus epochs contained data from +1

to +1024 ms where zero was deﬁned as the time of stimulus onset.

For both data sets, epochs including values outside the

range −100 to +100 μV range were excluded from the analysis.

In order to facilitate the comparison between EEG recorded from

different individuals, it was convenient to ensure that each partic-

ipant contributed the same amount of data. For this reason, only

the ﬁrst 20 epochs for the resting state conditions and the ﬁrst 50

epochs for the event-related paradigm were included.

Hyperconnectivity analysis

The data from each participant was paired with e very other par-

ticipant and analyzed as if they had been recorded jointly in

a hyperscanning experiment. With 10 participants, this gave a

total of 45 pseudo-pairings, one of whom was arbitrarily nomi-

nated as participant 1 and the other as participant 2. Twenty-eight

channels of EEG were recorded for each person meaning that

there were 56 channels for each pair of participants giving a total

of 1540 possible different channel combinations. Of these, only

the 784 hyper-connections that paired data between people were

considered further.

For each pairing, EEG data were concatenated across epochs

in preparation for the hyperconnectivity analysis. For the rest-

ing state data, 20 consecutive epochs of artifact-free EEG were

joined together from each condition to form 20.48 s of data

foreachoftheeyesopenandeyesclosedconditions.Forthe

event-related data, 50 epochs of pre-stimulus and post-stimulus

EEG were concatenated separately to give two time series of

51.2 s each.

COH and PDC were estimated from these concatenated data

for each participant separately using the method described for

simulated data and hyperconnectivity estimates were the high-

est values obtained in each of the Theta (4–8 Hz), Alpha (8–12),

Beta1 (13–19 Hz), Beta2 (20–29 Hz), and Gamma (30–70 Hz) fre-

quency bands. For the phase-based measures, PLV

,CCorrand

KMI, the concatenated data were ﬁltered into the same frequency

bands using Butterworth ﬁlters and the instantaneous phase was

estimated using the Hilbert transform in the same way as for the

simulated data. PLV

and CCorr were estimated for each 1024 ms

epoch and frequency band separately and averaged. The KMI was

estimated from the same data down-sampled to 10 Hz.

Statistical analysis

For the rest conditions, connectivity in the Eyes Open and Eyes

Closed conditions were compared and for the event-related data,

connectivity in the pre-stimulus period was compared to that in

the post-stimulus per iod. The difference in connectivity between

experimental conditions was estimated for each of the 784 elec-

trode pairs and averaged across each of the 45 pseudo-pairs of

participants.

In order to determine if the differences were reliable, a ran-

domization testing procedure was used to control the Type-1

error (Holmes et al., 1996; Burgess and Gruzelier, 1997). Consider

one electrode pair; under the null hypothesis, there should be no

difference between conditions and so randomly swapping the data

between them and calculating the difference many times should

provide a good estimate of the variability in the connectivity of

that electrode pair that is due to chance. If the difference in con-

nectivity observed in the real data set is larger than 95% of the

differences observed in the randomized data sets, it is reasonable

to say that that difference is greater than might be expected by

chance.Toextendthisideatomultipleelectrodepairs,insteadof

examining the distribution of the connectivity difference at each

electrodepairinturn,thedistributionofthelargestdifferencein

connectivity across all electrode pairs for each randomization was

examined. The 95th percentile of the distribution of the maxi-

mum difference represents the value that would not be exceeded

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

at any electrode pair by chance. In this way, the family-wise

Type-1 error can be controlled to 5%.

The maximum difference across all 784 electrode pairs was

estimated for 1000 randomizations of the data. The 95th per-

centile of this distribution was used as the upper cut-off for deter-

mining statistical signiﬁcance and controlling the per-condition

comparison Type-1 error to 5%. The same process was used to

obtain a lower cut-off value.

RESULTS

SIMULATIONS

The results from the simulations showing the effects of varying

the concentration, κ, and the zero-lagged correlation, r,oneach

of the measures of connectivity are shown in Figure 5.Theﬁrst

criterion of good performance, that the measures should have low

bias and low RMSE can be addressed by examining the mean bias

and RMSE of each of the connectivity measures when r = 0and

for each value of κ (Figure 6). Note that for COH, PDC and PLV,

the bias equals the RMSE as all values are positive and greater than

0;onlyforCCorrandKMIdotheydiffer.COHdidnotmeetthe

criterion for any value of r and PDC and PLV

only came close

for low values of κ. KMI was close to the criteria for all values of

κ but, as the minimum value of KMI is zero, there was a small,

consistent bias. Only CCorr met the criteria fully.

The second and third criteria of good performance, that

the estimate of connectivity should increase monotonically as r

increases and that it should be insensitive to changes in the vari-

ance of the marginal distributions of deviations from the expected

phase (1/κ), can be considered together. Ta b le 1 shows the pro-

portion of variance in each measure of connectivity that can be

accounted for by r, κ, the interaction r by κ and error derived

from a Two-Way ANOVA of the simulation data. For all of the

measures, except PDC, r, accounted for a good proportion of

the variance but for COH and PLV, this proportion was small

compared to the proportion attributable to κ. The poor perfor-

mance of PDC is this context was unsurprising as it is designed to

identify Granger causality in which one time series leads the other,

not instantaneous associations as seen here. Never theless, the sen-

sitivity of PDC to κ meant that relatively high values of PDC were

obtained even where there was no real association between chan-

nels and it was the measure that showed the highest proportion

of error variance. The interaction, r by κ, was important only for

the PLV

where it accounted for 7.5% of the variance and was

manifest as a relatively greater inﬂuence of r, at low values of κ

(Figure 5C). The two measures that best met the criteria were

CCorr and KMI as they were both overwhelmingly inﬂuenced

by r but not κ and,ofthetwo,CCorrhadamuchsmallererror

variance.

The results from the simulations showing the effects of vary-

ing the concentration, κ, and the 100 ms-lagged correlation, r,

on each of the measures of connectivity are shown in Figure 7.

This simulation was designed to provide an example of Granger

Causality that would be well-suited for analysis by PDC. The ﬁrst

point to note is that COH was largely unaffected by the change

(compare Figures 5A and 7A) and performed badly with both sets

of data. In contrast, the PLV, CCor r, and KMI were all adversely

affected which is unsurprising as these measures are not designed

for use in this context. In the case of PLV

and KMI, less than

1% of the variance was attributable to r. For PLV

most vari-

ance was accounted for by κ whereas for KMI it was error. The

poor performance of KMI with this data set occurred because it

was estimated from the phase-series down-sampled to 10 Hz, the

same rate at which the random phase deviations were added to

the phase sequence. This meant that the simultaneous estimates

of phase were truly independent. In contrast, because the PLV

and CCorr were estimated based on intermediate points that were

spline estimates of the preceding and subsequent phase devia-

tions, each datum contained some information about the lagged

FIGURE 5 | The relationship between the true and estimated coupling

for each measure of connectivity for the zero-lagged simulated data at

different levels of concentration, κ.(A)Shows coherence, (B,C) partial

directed coherence, (D) time-averaged phase-locking value, (E) circular

correlation coefﬁcient and (F) Kraskov mutual information. Concentration

values are ¼, ½, 1, 2, 4, and 8.

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

FIGURE 6 | The bias and RMSE for each connectivity measure

estimated from the zero-lagged simulated data at different levels of

concentration, κ, when the coupling, r = 0. Blue dots i ndicate bias and

red circles RMSE. (A) Shows coherence, (B,C) partial directed coherence,

(D) time-averaged phase-locking value, (E) circular correlation coefﬁcient

and (F) Kraskov mutual information. For COH, PDC and PLV, as all values

are >0, bias = RMSE are equal. For CCorr and KMI, where values may

be ≤ 0, bias = RMSE.

FIGURE 7 | The relationship between the true and estimated coupling

for each measure of connectivity for the 100ms-lagged simulated data at

different levels of concentration, κ .(A)Shows coherence, (B,C) partial

directed coherence, (D) time-averaged phase-locking value, (E) circular

correlation coefﬁcient and (F) Kraskov mutual information. Concentration

values were ¼, ½, 1, 2, 4, and 8.

relationship between the phase series. This is the reason why

CCorr shows some sensitivity to increases in r, although much less

than for the zero-lagged data. Of course, each of these measures

would perform much better if they had been estimated a cross a

range of time lags.

As expected, PDC performed better on this simulation than

with the zero-lagged data. PDC

1, 2

showed a clear monotonic

increase with r correctly showing that channel 1 led channel 2.

Similarly, PDC

2, 1

showed a monotonic decrease with r, meaning

that the predictability of channel 1 given channel 2 diminished as

the predictability of channel 2 increased. However, in both cases,

the largest proportion of variance was attributable to κ,notr.

HUMAN STUDIES

Event-related changes in synchrony

The results of the hyperconnectivity analysis between pre-

stimulus and post-stimulus conditions, controlled for multiple

comparisons, are shown in Figure 8. As all the data were recorded

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

independently, there can have been no true synchronization

between the recordings. Nevertheless, there were a small num-

ber of signiﬁcant changes in synchronization between the pre-

and post-stimulus conditions identiﬁed by PDC and CCorr and a

very large number for the PLV

. For PDC and CCorr, the changes

in synchronizations involved both increases and decreases but

for PLV

, they were exclusively in the direction of increased

synchrony in the post-stimulus period. The estimates of mean

synchrony averaged across the pre- and post-stimulus periods

for each of the connectivity measures are shown in Figure 9.For

PDC, the estimated levels of synchrony were consistently very low

(range 0.01–0.03) and were also low for CCorr but m ore vari-

able (range 0.001–0.14). In contrast, the mean synchronization

was much higher for PLV with values ranging from 0.19 to 0.56.

As these data were from an event-related paradigm, it was also

possible to estimate the between-trial synchronization using PLV

and CCorr

. Figure 10 shows the signiﬁcant differences in PLV

and CCorr

between the pre- and post-stimulus periods. There

were no signiﬁcant differences in synchronization between condi-

tions using CCorr

but there were several using PLV

in the theta

and alpha frequency ranges. The estimates of mean synchrony

in the pre- and post-stimulus periods for PLV

and CCorr

are

shown in Figure 11. The PLV

and CCorr

were rather larger than

their time-averaged equivalents and were approximately equal

across the frequency bands (PLV

range 0.12–0.17; CCorr range

0.12–0.19).

Resting state

The results of the hyperconnectivity analysis between the eyes

open and eyes closed resting states, controlled for multiple com-

parisons, are shown in Figure 12. There were a number of

signiﬁcant differences in hyperconnectivity between eyes open

and eyes closed using PDC. Most of these indicated that neu-

ral activity at multiple sites in pseudo-pair participant 1 was

a signiﬁcantly stronger predictor of neural activity at electrode

FIGURE 9 | Mean Hyperconnectivity values for time-averaged

event-related EEG by connectivity measure and frequency band.

FIGURE 8 | Signiﬁcant changes in mean time-averaged

hyperconnectivity between pre- and post-stimulus conditions by

connectivity measure and frequency band. The rows represent the

hyperconnectivity results for each of the measures used (PDC

1, 2

,PDC

2, 1

PLV

and CCorr) and the columns represent the frequency bands (theta,

alpha, beta1, beta2, and gamma). The pairs of large circles in each cell

represent the heads of the participants in a pseudo-pair. The smaller circles

indicate the topographical location of the EEG recording electrodes. For PLV

and CCorr, lines drawn between the heads joining electrode sites indicate

that there was a signiﬁcant change in connectivity from the pre- to the

post-stimulus periods between the ﬁrst member of a pseudo-pair and the

second member. Red lines indicate an increase in connectivity from the pre-

to the post-stimulus period and blue lines indicate a decrease. For PDC

1, 2

lines connecting electrode sites between the heads show that neural activity

in the ﬁrst member of a pseudo pair was more predictive of the neural

activity of the second member of the pair in the post-stimulus period than in

the pre-stimulus period. Allocation to ﬁrst or second member of the

pseudo-pair was arbitrary.

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

FIGURE 10 | Signiﬁcant changes in mean trial-averaged

hyperconnectivity between pre- and post-stimulus conditions by

connectivity measure and frequency band. The rows represent the

hyperconnectivity results for each of the measures used (PLV

and

CCorr

) and the columns represent the frequency bands (theta, alpha,

beta1, beta2, and gamma). The pairs of large circles in each cell

represent the heads of the participants in a pseudo-pair. The smaller

circles indicate the topographical location of the EEG recording

electrodes. Lines drawn bet ween the heads joining electrode sites

indicate that there was a signiﬁcant change in connectivity from the

pre- to the post-stimulus periods between the ﬁrst member of a

pseudo-pair and the second member. Red lines indicate an increase in

connectivity from the pre- to the post-stimulus period and blue lines

indicate a decrease.

FIGURE 11 | Mean Hyperconnectivity values for trial-averaged

event-related EEG by connectivity measure and frequency band.

site FP1 in participant 2 when the eyes were closed than when

they were open. There were also two links indicating/that neu-

ral activity participant 2 drove neural activity in participant 1.

There were also a small number of hyper-connections identi-

ﬁed by CCorr, one showing signiﬁcantly lower synchronization

between the participants in the eyes closed condition in theta fre-

quency range and four showing the reverse in the alpha frequency

range. However, by far the largest numbers of signiﬁcant changes

in synchrony were identiﬁed by PLV

. In the theta frequency

range, there were multiple hyper-connections that were signiﬁ-

cantly higher in the eyes open condition than in the eyes closed

condition. In the alpha frequency range, there was an even larger

number of hyper-connections that were greater in the eyes closed

condition. The estimates of mean synchrony in the eyes open

and eyes closed conditions for each of the connectivity measures

are shown in Figure 13 As was the case with the event-related

data, mean connectivity was low for PDC (range 0.01–0.11) and

CCorr (0.001–0.06) but was very much greater for PLV

(range

0.13–0.40).

DISCUSSION

The issue of how best to measure hyperconnectivity depends in

no small part on what one is trying to measure. Many hypercon-

nectivity researchers intended to measure synchronization which,

in the Huygens sense means that two oscillators (in this case,

the EEG of two people) interact in such a way that their cycles

become synchronous. However, synchronization, as deﬁned by

the PLV, is rather different and simply means there is a consistent

phase difference between the two signals but does not necessar-

ily imply covariance between them. By this criterion, any pair of

EEG channels with a common dominant frequency would be syn-

chronized, which surely makes this deﬁnition too inclusive to be

useful. Instead, a more useful deﬁnition is that two oscillators can

be said to be synchronized if deviations from the regular oscilla-

tory cycle of one oscillator provides information about deviations

in the oscillatory cycle of the other.

By this deﬁnition, none of the commonly used measures of

connectivity fared well in the simulations. COH, PDC, and PLV

were biased measures of the co-variation between phase series

and, under a broad range of conditions provided inaccurate esti-

mates of the true hyperconnectivity. In particular, the y were each

prone to detect hyperconnectivity that didn’t exist. It is well

known that COH is a biased estimator of true coherence (Maris

et al., 2007) but using Welch’s method limits the extent of the

problem. PLV too, is a biased estimator of coupling strength and

the bias is greater when small samples of data are used, particu-

larly, as is the case with PLV

, when non-independent data points

areused(Vinck et al., 2012). To put the scale of the problem

in context, consider those simulations where the concentration

was close to the mean value seen in the human EEG recordings

and the true hyperconnectivity was zero (κ = 2, r = 0). Here the

estimated coupling strengths were 0.65, 0.19, and 0.58 for COH,

PDC, and PLV respectively.

These spur ious couplings are not solely due to the familiar

bias of the estimators. Rather, the coupling was driven by changes

in the variances of the indiv idual phase series (i.e., 1/κ of the

marginal distributions of deviations from the expected phase).

As Ta b le 1 shows, COH, PDC, and PLV

were more sensitive to

changes in the variance of the marginal distributions of deviations

Frontiers in Human Neuroscience www.frontiersin.org December 2013 | Volume 7 | Article 881

Burgess EEG hyperscanning: a cautionary note

FIGURE 12 | Signiﬁcant changes in mean time-averaged

hyperconnectivity between eyes open and eyes closed resting states by

connectivity measure and frequency band. The rows represent the

hyperconnectivity results for each of the measures used (PDC

1, 2

,PDC

2, 1

PLV

, and CCorr) and the columns represent the frequency bands (theta,