 

 

 

 

 

 

 

 

 

war

$1
5.5 l‘\ .

ibiﬁw .,.., .f §
1.6..quwﬂlwxyh ﬂﬁhwwén
tamn_i_&ml..h-: own.
. ‘ @ﬂhﬁﬁﬁ‘ﬁﬂ .
g? .53.

3‘3!

.53..“

a \v .
. t. 2‘: . _ ,
uni-‘11:“: .. antiwar: I
Sliﬁwn iii
.5

:{v

1....

1’ 1.3:. \ i)
933131?
. lelnvnnndu.t,uiﬁ
£3 25..
$.32.

x. has
.. a:
34...! E, 9......
.13... . .
if! . l

, .. 9,
V . . ‘
:....ne.....k. . . . %»4.wnqﬂ.nf}
,, , , ‘..n........h:. L.‘ .

:Zlifv

 

 

(‘j

0".
3C0?

This is to certify that the i
dissertation entitled f

A METHOD FOR IDENTIFYING INFLUENTIALS

 

WITHIN A SOCIAL NETWORK
4% presented by
>- +- 3‘
m (f) .—
< c 9
0: «5 g
£3 £2) '5 MICHAEL RYAN KOTOWSKI
;-- .C
l-‘ .2 3
i 2
“‘— has been accepted towards fulﬁllment

of the requirements for the

Doctoral degree in Communication

ﬂag or:

Major Professor’s Signature

 

21' 3m 3907

Date

MSU is an aﬂirmative-action. equal-opportunitv employer

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:lClRC/DateDue.indd-p.1

 

 

A METHOD FOR IDENTIFYING INFLUENTIALS
WITHIN A SOCIAL NETWORK

By

Michael Ryan Kotowski

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Communication

2007

ABSTRACT

A METHOD FOR IDENTIFYING INFLUENTIALS
WITHIN A SOCIAL NETWORK

By

Michael Ryan Kotowski

Developing the ability to identify those responsible for enacting the majority of
inﬂuence within a social network is an intriguing endeavor to social inﬂuence theorists
and practitioners alike. One method of identiﬁcation employs an argument that
inﬂuentials possess a speciﬁc, quantiﬁable array of individual differences predisposing
them to be superdiffusers of information. The three studies presented here reﬁne the
psychometric properties of a method of measuring the individual differences that may
predispose some people to be particularly inﬂuential; and consequently efﬁcient

purveyors of change.

ACKNOWLEDGEMENTS

The largest acknowledgement goes to my major professor, Dr. Franklin J. Boster.
His continuing mentorship has been invaluable. Approximately three years ago I wrote in
the acknowledgements of my master’s thesis that Dr. Boster’s standards and values have
made an enormous positive impact on my intellectual and personal growth. It is three
years later and not much has changed except for the size of the impact. One could not
have asked for a better mentor (or somebody with a better collection of scotch).

Dr. Sandi W. Smith has been an ideal committee member. Dr. Smith offered both
critique and encouragement with what seemed to be a natural sense for when I needed
each. Committee member, Dr. Timothy R. Levine, also deserves a sincere thank you for
always asking tough questions and providing many helpful critiques. Finally, committee
member Dr. Toby Ten Eyck continues to remind me to consider social phenomena from
multiple perspectives. Although he was the external member of my committee, his
assistance and input have always gone far beyond expectations.

I am also fortunate to have several colleagues who have made completing this
dissertation much more enjoyable. The Portuguese has been a wonderful conversational
partner, always excited to engage in candid conversation on the current state of affairs.
Trenchmouth has been an important collaborator even if the politics of his haircut are not
consonant with the politics of his vote. Spicy has played the important role of reminding
me that all work and no play makes Mike a dull boy, even if he does not always listen.

Finally, my entire family, especially Erin, deserves acknowledgement. It takes a
special person to love somebody who is completing a doctoral degree. I am still grateful

for the constant support and encouragement.

iii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. vi

LIST OF FIGURES ............................................................................... vii

INTRODUCTION .................................................................................. 1

STUDY 1 ........................................................................................... 23

Method ..................................................................................... 23

Subjects ........................................................................... 23

Procedures ........................................................................ 23

Instrumentation .................................................................. 23

Results ...................................................................................... 26

Discussion ................................................................................. 34

STUDY 2 ........................................................................................... 38

Method ..................................................................................... 38

Subjects ........................................................................... 38

Design ............................................................................. 38

Procedures ........................................................................ 39

Instrumentation ................................................................... 40

Results ........................................................................................ 41

Discussion ..................................................................................... 55

STUDY 3 ........................................................................................... 59

- Method ..................................................................................... 59

Subjects ........................................................................... 59

Design ............................................................................. 59

Procedures ......................................................................... 59

Instrumentation .................................................................. 60

Interview .......................................................................... 61

Interview Coding ................................................................ 64

Results ...................................................................................... 65

Measurement Models ............................................................ 65

Predictive Validity ............................................................... 70

Discussion ................................................................................. 74

GENERAL DISCUSSION ....................................................................... 77
APPENDICIES

Appendix A: Category Scaling Instructions ........................................... 86

Appendix B: Demographic Items ....................................................... 87

Appendix C: Connector Items ........................................................... 88

Appendix D: Persuader Items ........................................................... 89

iv

Appendix E:
Appendix F:

Appendix G:
Appendix H:

Appendix 1:
Appendix J:

Appendix K:

Appendix L:

Appendix M:
Appendix N:
Appendix 0:

Appendix P:

Appendix Q:

Appendix R:
Appendix S:

REFERENCES .......

Healthy Lifestyle Maven Items90
Need for Afﬁliation Items .............................................. 91
Perspective Taking Items .............................................. 92
Empathic Concern Items ............................................... 93
Opinion Leadership Items .............................................. 94
Healthy Lifestyle Activism Items .................................... 95
Direct Interval-Estimation Scaling Instructions ..................... 96
Social Anxiety Items ................................................... 98
Argumentativeness Items ............................................... 99
Value-Relevant Involvement Items ................................. 100
Argumentative Complexity Interview Protocol ................... 101
Phonebook Test ........................................................ 103
Healthy Lifestyle Knowledge Test ................................. 105
Interview Code Sheet ................................................. 108
Argumentative Complexity Coding Examples .................... 109
............................................................................ 122

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 1 1

Table 12

Table 13

LIST OF TABLES

Boster et al. (2006) Nomological Network Correlations ...................

Study 1 Factor Loadings and Factor Correlation Matrix... .....................

Study 1 Measure Descriptives .................................................

Study 1 Regression Models ....................................................

Study 2 CFA Fit Statistics .......................................................

Study 2 Direct Interval-Estimation Factor Loadings and

Factor Correlation Matrix... . . . . . . .. .........................................................

Study 2 Measure Descriptives .................................................

Study 2 Regression Models

Study 3 Factor Loadings and Factor Correlation Matrix. . . .

Study 3 Argumentative Complexity Second-Order
Factor Loadings.. ....................................................................................

Study 3 Healthy Lifestyle Knowledge Factor Loadings. . . . . . . . . . . . . .

Study 3 Measure Descriptives .................................................

Study 3 Predictive Validity Correlations .....................................

vi

10

28

31

34

42

46

49

.54

65

69

69

70

72

Figure 1

Figure 2

Figure 3

Figure 4
Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

LIST OF FIGURES

Hypothesized Convergent and Divergent Validity Nomological
Network for the Connector Construct ......................................... l7

Hypothesized Convergent and Divergent Validity Nomological
Network for the Persuader Construct .......................................... 18

Hypothesized Convergent and Divergent Validity Nomological

Network for the Healthy Lifestyle Maven Construct ........................ 19
Opinion Leadership by Superdiffuser Scatterplot Matrix ................... 32
Healthy Lifestyle Activism by Superdiffuser Scatterplot Matrix. . . . . 33

Convergent and Divergent Validity Nomological Network
for the Connector Construct .................................................... 50

Convergent and Divergent Validity Nomological Network
for the Persuader Construct ..................................................... 51

Convergent and Divergent Validity Nomological Network
for the Healthy Lifestyle Maven Construct .................................. 53

Predictive Validity Scatterplots ................................................ 71

vii

INTRODUCTION

Although the interpersonal inﬂuence generated by a mediated communication
campaign is often seen as a design confound between intervention and comparison
groups (V alente, 2001), when controlled by researchers, the effectiveness of campaigns
that employ interpersonal inﬂuence has been demonstrated, particularly in grassroots
campaigning (Salmon & Atkin, 2003). One reason for the effectiveness of campaigns
utilizing interpersonal inﬂuence resides in the fact that an inﬂuence agent can exert
inﬂuence adaptively, whereas the adaptive ability of a mediated message is limited. Even
though both mediated and interpersonal inﬂuence agents can target speciﬁc populations,
the interpersonal inﬂuence agent can more easily tailor appeals to the audience. More
precise tailoring is advantageous from the perspective of the functional theory of attitudes
(Katz, 1960; Shavitt, 1990) because as a message matches more closely the function
served by an attitude, the likelihood of the desired persuasive outcome increases. The
tailoring also results in messages addressing barriers to change at the individual level
rather than at the group level. According to the health belief model (Rosenstock, 1974)
this more precise attack on the target’s perceived barriers (as well as other factors such as
perceived severity and susceptibility) will result in a greater likelihood of inﬂuence.
These positions are consistent with Chaffee (1982) who suggested that although
traditional media campaigns are effective at disseminating information, interpersonal
inﬂuence is necessary for change.

Employing interpersonal inﬂuence in persuasive campaigns is, however, not
innovative. The potential for interpersonal inﬂuence has received substantial recognition

from scholars, most notably in the two-step ﬂow hypothesis (Katz, 1957; Katz &

Lazarsfeld, 1955) and Rogers’ (1995) diffusion of innovations. The two-step ﬂow
hypothesis posits that mediated campaigns exert inﬂuence upon a small subgroup of
opinion leaders that in turn exert inﬂuence upon the remainder in the social network. This
idea has been reﬁned in subsequent work (Granovetter, 1978; Valente, 1995) proposing
that the common characteristic of the subgroup is their tendency to adopt change more
quickly. Put differently, the subgroup possesses a lower threshold for inﬂuence than those
composing the majority of the network.

Understanding two-step ﬂow in terms of inﬂuence thresholds is also consistent
with Rogers’ (1995) diffusion of innovations model. According to the model, for any
social network there is within group variability of people’s thresholds for adopting an
innovation (i.e., a new technology or idea). Rogers (1995) posits that people can be
classiﬁed into the categories of innovators, early adopters, early majority, late majority,
and laggards based on their threshold. Innovators and early adopters have the lowest
thresholds for adoption and are analogous to the initial subgroup in the two-step ﬂow
hypothesis. Laggards on the other hand have the highest threshold for adoption and are
consequently the last to adopt an innovation. According to Rogers (1995), because they
are actively seeking new information a campaign exerts the most inﬂuence on the
innovators and early adopters who, in turn, exert interpersonal inﬂuence on people in the
remaining three categories.

Common to both the two-step ﬂow hypothesis and the diffusion of innovations
model are the persuasive effects that opinion leaders and innovators have on the
remainder of the social network. In fact, evaluations of health campaigns employing the

interpersonal inﬂuence of opinion leaders often reveal success (Celentano, Bond, Lyles,

Eiumtralcul, Go, Beyrer, Chiangmai, Nelson, Khamboonruang, & Vaddhanaphuti, 2000;
Earp, Eng, O’Malley, Altpeter, Rauscher, Mayne, Matthews, Lynch, & Gaquish, 2002;
Kelly, St. Lawrence, Stevenson, Hauth, Kalichman, Diaz, Brasﬁeld, Koob, & Morgan,
1992; Miller, Klotz, & Eckholdt,1998; Soumerai, McLauglin, Gurwitz, Guadagnoli,
Hauptman, Borbas, Morris, McLaughlin, Gao, Willison, Asinger, & Gobel, 1998). Thus,
it appears that the ability to identify opinion leaders within a social network is of
particular value to the inﬂuence scholar and practitioner.

The question then becomes how does one identify opinion leaders so that their
power of inﬂuence can be utilized to persuade others. Network analysis provides one
possible solution (Durland & Fredericks, 2006; Valente, 1995). Network analysis is a
technique used to map the pattern of relationships among actors in a group. By mapping
the amount of communication between actors, network analysis can by used to identify
possible opinion leaders by locating those with many communication links with others. A
sociogram (Festinger, 1949; Moreno, 1934) is a map of the social network deﬁning each 1
individual as a node within a network of interconnected nodes where the connections
represent lines of communication. By employing network analysis techniques, the density
of communication between nodes can be quantiﬁed (Festinger, 1949; Scott, 2000).
Through the examination of the density of communication within the network, it is
possible to predict which actors will emerge as opinion leaders within the social network
(V alente, 1995).

Network analysis is an expensive and time consuming method of identifying
opinion leaders, particularly for large networks, because the methodology requires

complex analyses to be carried out on a large number of observations to locate a small

number of opinion leaders. Rogers (1995) reports two more efﬁcient techniques to
identify opinion leaders. The ﬁrst method involves asking actors to report who the leaders
of the network are and the second method involves asking actors to report if they are a
leader in the network. The validity of both methods, however, rests upon the ability and
willingness of the respondent to report their own group status and the group status of
others. The accuracy of these reports, however, can be suspect (Rogers, 1995).
Consequently, as a potential solution to this problem, Rogers (1995) points out several
demographic variables that could also be used to identify opinion leaders, such as higher
exposure to the media than others in the network, extensive interpersonal relationships,
high socioeconomic status, and a proclivity for innovativeness.

Although these demographic characteristics allow opinion leaders to be
identiﬁed, they do not speak to the individual differences that explain why the opinion
leader is an efﬁcient purveyor of interpersonal inﬂuence. This fact is important for at
least three reasons. First, if those responsible for diffusing the majority of information
across a network are psychologically unique, the process of identifying them can be
reﬁned to a short psychographic inventory thus reducing the amount of resources
necessary to locate them. Second, the same individual differences could be used to offer
insight into the development of persuasive message strategies designed to inﬂuence
people with those traits in the way Palmgreen, Donohew, Lorch, Hoyle, and Stephenson
(2002) developed messages speciﬁcally tailored for high sensation seekers. Third, if these
individual differences are motives for behavior, understanding them and the affiliated
behaviors of the opinion leader can reveal ways to train people who do not have the traits

to exhibit the behaviors. Therefore, based on the work of Gladwell (2002) and Boster,

Kotowski, and Andrews (2006), this series of studies seeks to develop further and reﬁne a
method of measuring the individual characteristics predisposing people to inﬂuence
others effectively.

In a review of the literature across the disciplines of psychology, sociology,
communication, and marketing, Gladwell (2002) suggested three individual difference
characteristics — connector, salesman, and maven - that explain opinion leadership.
According to Gladwell’s (2002) conceptualization, a person need only exhibit one of the
traits strongly to be an opinion leader. If a person exhibited a substantial amount of all
three traits, however, not only would their ability to inﬂuence the network likely be ‘
greater, they would also be qualitatively different ﬁom an opinion leader. Such a person
would be a superdiffuser, not only an early adopter of new information, but also a
proactive disseminator of information across a diverse social network. Although the traits
of connector, salesman, and maven are by no means the only important traits of the
superdiffuser, they do provide an intuitive triumvirate of characteristics that one would
naturally expect to be descriptive of an effective agent of inﬂuence —— the well connected,
person adaptive, knowledgeable persuader.

One of the classic investigations of the level of connectedness among people is
the Travers and Milgram (1969; Milgram, 197 7) study of the small world problem. This
study found that information has a tendency to travel through a small handful of well
connected people serving as common connections for the majority of the social network.
Milgram and his colleagues randomly asked residents of a distant Midwestern city to
forward a package to a stranger in Boston by passing the package through friends and

acquaintances. In the end, three people accounted for 48% of the 64 packages that

eventually arrived at the target’s address. Furthermore, the median number of
intermediaries necessary to get a package from the Midwest to Boston was only 5.5.
Granted, many of the packages never made it to Boston and the impact of these failed
chains on the actual chain length had they not failed was not fully considered (e.g.,
Hunter & Shotland, 1974). Nevertheless, the small world problem provides an example
of the variance in the level of connectedness that exists in social networks.

Based on the work of Gladwell (2002) it is thought that the connectors in the
Travers and Milgram (1969) study were people who easily form close acquaintances with
others. Furthermore, they were also likely good at maintaining those acquaintances once
formed. It is possible however, to think of people who have lives rife with acquaintances
who are not opinion leaders. This situation would most likely result when the
acquaintances are all within group members. Although connectors do indeed form within
group acquaintances, a more important prerequisite for a connector than the number of
acquaintances is the number of acquaintances from different groups within the larger
social network (Granovetter, 1973; 1982).

Weak-ties connect otherwise disconnected groups and consequently reduce the
overall degree of homophily, or similarity, among people (Lazarsfeld & Merton, 1964).
This reduction occurs because the more that people interact, the more that information is
shared (Berger & Calabrese, 1975). As more information is shared, homophily within the
group increases. Given that within group members are in more frequent contact with each
other than with members of the larger social network, the level of within group
homophily is higher than the level of homophily in the larger social network. As a result,

there is a greater chance of exposure to unique information coming from a person outside

of the immediate social group than from within. Although connectors are within group
members, they have exposure to the unique information coming from the larger social
network because of their weak-tie relationships between groups. These weak—tie
acquaintances increase the chance of unique information from one group being available
to another group in the network. Hence, because of their tendency to form acquaintances
not only within a group but across groups, connectors are pivotal players in the transfer of
unique information across the entire social network.

Gladwell’s (2002) explication of the salesman construct focuses on the subtle
behavior that people may express when trying to persuade others with a particular focus
on the ability to be emotionally expressive (Friedman, Prince, Riggio, & DiMatteo,
1980). Although emotional expressiveness may increase the level of emotional contagion,
which increases similarity, which increases liking, which ﬁnally impacts persuasion, even
if the model was correct and every path coefﬁcient in the model equaled .70 the
relationship between emotional expressiveness and persuasiveness would equal a
correlation of .24. In other words, the link between Gladwell’s (2002) conceptualization
of the salesman and the intended outcome of persuasion are so far removed in the causal
process that it is unlikely to be useful. F urtherrnore, salesmen make sales whether or not
the customer has privately accepted (Festinger, 1953) the reason for complying, and short

term compliance is not the goal of the opinion leader. The goal is attitude change and
without private acceptance or the presence of the opinion leader to enforce compliance,
prolonged behavior change is unlikely to result. Therefore, the second of Gladwell’s

(2002) opinion leader traits has been redeﬁned into the persuader.

Effective opinion leaders are persuaders because of the ability to effectively
search their message repertoire for a target adapted message (for a review, see O’Keefe,
2002). In other words, persuaders are skillful inﬂuencers of others in their social network.
As an agent of inﬂuence, persuaders are able to generate highly accurate implicit theories
of mind about the target. Persuaders also understand the implications of that implicit
theory for persuasive message design. Thus, persuaders have superior person perception
on some dimensions in a manner similar to high empathics (Davis, 1980) but persuaders
utilize the accurate person perception to develop effective persuasive strategies for the
speciﬁc inﬂuence target.

This notion of skill which is central to the persuader construct likely shares much
variance, possibly converging over time, with a motivation to argue. That is, people who
enjoy arguing are more likely to develop skill and people who have skill are more likely
to enjoy arguing. Because of this relationship, persuaders are motivated to share
information and try to convince others in the social network of the inforrnation’s merit.
Furthermore, persuaders are unlikely to concede a position of contention. Persuaders are
not likely motivated due to dogmatism (Goldsmith & Goldsmith, 1980). If dogmatism
were high, the persuader would have little information to share. Rather persuaders exert
inﬂuence because of a need to share information for what is perceived to be the beneﬁt of
others. Accordingly, persuaders are proﬁcient but fair debaters. Thus, persuaders are
inﬂuentials within the social network not only because of persistent argument but also
because that argument is tailored to the receiver.

The ﬁnal construct of the inﬂuential triumvirate identiﬁed by Gladwell (2002) is

the maven. Interest in mavens ﬁrst surfaced in marketing because an understanding of

their characteristics could help market researchers understand marketplace consumption
(Feick & Price, 1987). From the perspective of information diffusion, however, mavens
are important because they ﬁnd generating expertise in a content domain an exciting
route to self-expression which is achieved through sharing the expertise with others that
are perceived to be in need of information (Price, Feick, & Guskey, 1995). As a result of
the information gathering involved in expertise generation, not only are mavens exposed
to new information before others, but they are also adept at integrating new information
into existing knowledge structures. Although Gladwell (2002) and Feick and Price (1987)
deﬁne mavens as having broad expert knowledge, the more useful criterion for the
purpose of identifying mavens for use in persuasive campaigns is expertise in a speciﬁc
content area such as automobiles, sports, technology, or health.

This study considers the maven construct in the health context. The healthy
lifestyle domain was chosen because its breadth is similar to that of the consumer
products domain where much of the work on mavens has already been conducted.
Speciﬁcally, the health domain is broader than cancer knowledge but narrower than the
entirety of health knowledge. Healthy lifestyle mavens would have broad knowledge of
healthy lifestyle topics, enjoy searching for new healthy lifestyle knowledge, enjoy
sharing this knowledge with others who were perceived to be in need, and receive
recognition by others as healthy lifestyle experts. A healthy lifestyle maven would be
asked health-related questions often, and would serve as an information resource for
others. Thus, healthy lifestyle mavens are opinion leaders because their evaluation and

sharing of new information entering a social network is often the ﬁrst heard by others.

Although Gladwell (2002) identiﬁed the constructs, a method of measuring the
extent to which people possessed the three constructs with quantiﬁable estimates of
reliability and validity was not developed until Boster et al. (2006) made an initial
attempt that offered promising results. Across two independent samples Boster et al.
(2006) tested the content validity of three measures. Each measure was composed of four
Likert self-report items developed to form unidirnensional measures of one of the
connector, persuader, and healthy lifestyle maven constructs. Employing conﬁrmatory
factor analysis (CFA) the internal consistency and parallelism of each of the measures
was found to exhibit substantial content validity in both samples. Furthermore, each
measure had ample levels of reliability estimated by Standardized Item (1 (S.I.a) z .84.
Boster et al. (2006) also examined the convergent and divergent validity of the three
measures within a nomological network (Cronbach & Meehl, 1955) of three additional
constructs — social anxiety, argumentativeness, and value relevant involvement — and
found results consistent with what would be expected given what the six measures were
purported to measure (Table 1).

Table 1

Boster et a1. (2006) Nomological Network Correlations

 

Connector Persuader Maven Anxiety Argue Value

Connector

Persuader 0.29

Maven 0.16 0.23

Anxiety -0.55 -0.20 -0.10

Argue 0.10 0.55 0.12 -0.21

Value 0.03 0.10 0.51 0.11 0.02

 

Notes. Maven = Healthy Lifestyle Maven, Anxiety = Social Anxiety, Argue = Argumentativeness, Value =
Value-Relevant Involvement, n = 189, matrix contains correlations corrected for attenuation due to
measurement error

10

Finally, Boster et al. (2006) quantiﬁed superdiffusers as people who scored at
least one standard deviation above the mean on the connector, persuader, and healthy
lifestyle maven measures simultaneously. According to this classiﬁcation method
approximately 2% of Boster et al.’s (2006) sample was classiﬁed as superdiffusers, which
is consistent with what would be expected if superdiffusers are similar to the innovators
of the diffusion of innovations literature (Rogers, 1995).

These studies build directly upon the ﬁndings of Boster et al. (2006) and offer
reﬁnements of those measures. First, although the Boster et al. (2006) data demonstrated
that the items on the three measures were internally consistent and largely parallel across
two independent samples, the small number of indicators resulted in some parallelism
misﬁt between two of the measures. Therefore, although fewer items can be
pragmatically superior, the studies reported subsequently increased the number of items
on each of the three measures from four to ten in an attempt to ﬁnd additional highly
valid indicators of these constructs.

Another outcome of these additional items is that as long as the item reliabilities
of the new items are at least equal to or greater than the means of the existing item
reliabilities, the reliability estimates of the overall measures will increase. This increase
occurs because the equivalence method of Cronbach’s a (1951) (a) deﬁnes reliability as
the mean of all possible split-half coefﬁcients. According to the formula for a, as the sum
of the item covariances increase relative to the sum of the item variances, a will increase.
That fact, in conjunction with the fact that items which are alternate indicators of the

same construct increase the sum of the item covariance faster than the sum of the item

11

variance, explains why the addition of alternate indicators of the same construct to a
measure will increase a.

Assuming that at least two of the new items for each measure will be discarded
because of internal consistency and (or) parallelism issues, that the item reliabilities of
the new items will be at least equal to the mean of the existing item reliabilities, and that
the reliability estimates provided by Boster et a1. (2006) will replicate, it is hypothesized
that data collected with the revised measures will reveal that S.I.a 1' .93 for the connector
and maven measures and that S.I.a z .91 for the persuader measure according to the
Spearman-Brown prophecy formula (Nunnally & Bernstein, 1994).

This research also attempts to improve upon the developments of Boster et al.
(2006) by modifying the method of scaling employed for the items composing each of the
measures. In addition to increasing the number of items on the measure, the reliability of
a measure can also be improved by increasing the range of item variances. The issue at
hand is essentially a range restriction problem (Nunnally & Berstein, 1994). If item
responses are not allowed to deviate from the mean item response because the range of
the response scale is restrictive, the item covariances will also be restricted, and
consequently the reliability of the measure will be attenuated. If, however, the range of
the response scale is increased, to the extent that the entire scale is used by respondents,
the item responses will be free to deviate to a greater extent ﬁ'om the mean item response.
This deviation increase will in turn increase the item covariances and ﬁnally disattenuate
the measure reliability.

One interesting technique to increase the range of the eight point categorical

Likert response scale employed in Boster et al. (2006) is the method of direct magnitude

12

estimation (Meek, Sennott-Miller, & F erketich, 1992; Stevens, 1956). The application of
direct magnitude estimation to social phenomenon evolved from research in the area of
psychophysics and involves presenting respondents with a single reference stimulus
against which the respondents compare their possession of the attribute being assessed by
the item. This comparison is accomplished by assigning the reference stimulus an
arbitrary intensity value, typically 100 points. If respondents believe they possess twice
as much of the attribute as the reference stimulus they are instructed to report 200 points.
On the contrary, if they feel they possess only half as much of the attribute as the
reference stimulus they are instructed to report 50 points. Respondents are further
instructed to use the entire range of possible multiples when making comparisons against
the reference stimulus.

Advocates of direct magnitude estimation argue that because respondents are not
limited to a set number of response options as in categorical Likert scaling, the range of
response variance is increased leading to improvements in overall measure reliability,
among other psychometric beneﬁts (e. g., Lodge, 1981). Research by Levine (1994),
however, suggests that these beneﬁts may only exist for responses that fall above the
reference stimulus and that direct magnitude scaling is detrimental to the psychometric
qualities of the measure for responses falling below the reference stimulus. Levine (1994)
employed a repeated measures technique and compared responses to crime severity items
made on a category scale versus a direct magnitude estimation scale where the magnitude
estimations were collapsed into the category scale values. An examination of Levine’s
(1994) data reveals that only 11% of the crime severity ratings were in agreement across

both types of scaling when the rating was below the value of the reference stimulus

13

whereas 52% of the ratings were in agreement when the rating was above the value of the
reference stimulus. One description of this effect is that respondents have more difﬁculty
quantifying the perceived level of an attribute relative to the reference stimulus when the
perceived level is below the reference and less difficulty when the perceived level of the
attribute is greater than the reference stimulus. This difﬁculty could result if respondents
had greater difficulty with the fractional multiplication involved in estimations below the
reference stimulus than with the whole number multiplication involved in estimations
above the reference stimulus.

- Silverrnan and Johnston (1975) present an elegant solution to this problem in the
form of direct interval-estimation. Direct interval-estimation differs from direct
magnitude estimation only in that respondents are given an interval anchored by low and
high stimuli rather than a single reference stimulus. The stimulus representing the least
amount of the attribute is assigned the lower value (e. g., zero) and the stimulus
representing the greatest amount of the attribute is assigned the higher value (e.g., 100).
Respondents are asked to consider the extent to which they possess the attribute in
question, relative to the stimuli anchoring each end of the interval. That assessment is
then quantiﬁed by the respondent as they report their location on the interval using the
values assigned to both anchors as bounds. For example, a respondent believing that they
possess a level of the attribute in the mid-point of the interval bounded by zero and 100
would report 50.

Silverrnan and Johnston (1975) reported data comparing the reliability of two
groups of 15 novice coders of two tapes of esophageal speakers which differed only in

the instructions given to the raters. One group was instructed to evaluate the quality of the

14

speaker using direct magnitude estimation scaling and the second group was instructed to
evaluate the quality of the speaker using direct interval-estimation scaling. The data
revealed that the direct interval-estimation ratings were substantially more reliable (intra-
class correlation coeﬁ‘icient = .89) than the direct magnitude estimation ratings (intra-
class correlation coeﬁicient = .53). Thus, the direct interval-estimation solution is elegant
not only because the method retains the variance maximizing beneﬁts of direct magnitude
estimation over categorical scaling formats, but it simultaneously eliminates the validity
problems identiﬁed by Levine (1994). Consequently, it is hypothesized that direct
interval-estimation scaling of item responses will result in greater reliabilities for the
measures employed in this research than will categorical Likert scaling of item responses.

By extending the nomological network (Cronbach & Meehl, 1955) tested in
Boster et al. (2006) to include measures of eight satellite constructs —— social anxiety
(Leary, 1983), need for afﬁliation (Edwards, 1959), argumentativeness (Infante &
Rancer, 1982), perspective taking (Davis, 1983), value-relevant involvement (Cho &
Boster, 2005), empathic concern (Davis, 1983), opinion leadership (Flynn, Goldsmith, &
Eastman, 1996), healthy lifestyle activism (modiﬁed from Corning & Meyers, 2002
conventional social activism) — the present research also seeks to bolster the construct
validity evidence produced by Boster et al. (2006). Cronbach and Meehl’s (1955)
approach involves hypothesizing a theoretically predicted nomological network among a
set of constructs. This method demonstrates convergent validity and divergent validity
through an a priori speciﬁcation of how the measure under consideration correlates with
other measures of theoretically related and unrelated constructs and subsequently

comparing that speciﬁcation to observed data. If the pattern of correlations matches

15

theoretical expectations then evidence is obtained consistent with construct validity. The
larger network of eight constructs provides stronger construct validity evidence because
random chance is less likely to produce a network of correlations consistent with theory.

Six of the satellite constructs were chosen so that, to the extent that the measures
exhibit concurrent validity, the data produced by two of the satellite measures would
correlate substantially with the data produced by the connector, persuader, or healthy
lifestyle maven measures and the remaining four satellite measures would correlate to a
lesser degree. Furthermore, the measures were chosen so that the satellite measures
producing the two strong correlations would be unique for each of the connector,
persuader, and healthy lifestyle maven measures.

Speciﬁcally, it is hypothesized that because connectors enjoy forming and
maintaining acquaintances across groups, the connector construct will show a substantial
negative correlation with social anxiety as measured by Leary’s (1983) interaction
anxiousness measure. Social anxiety is deﬁned as a subjective aversion to social
interactions. It is also hypothesized that the connector construct will have a substantial
positive correlation with the need for afﬁliation as measured by the Edwards personal
preference schedule (Edwards, 1959). Need for afﬁliation is deﬁned as a motivation to
draw near, cooperate with, and remain loyal to allied others. Finally, it is hypothesized
that the connector construct will correlate to a lesser degree with argumentativeness,
perspective taking, value-relevant involvement, and empathic concern because the desire
to form connections with others is unlikely to be directly causally related to those

constructs (see Figure l).

16

Afﬁliation Opinion

 

 

Ample Ample Ample
Positive Positive Negative
Argue Null Connect Null
Ample
Null Positive Null
Perspective Activism

Figure I . Hypothesized convergent and divergent validity nomological network for the
connector construct.

Notes. Connect = Connector, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve = Value-
Relevant Involvement, Afﬁliation = Need for Affiliation, Perspective = Perspective Taking, Empathy =
Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism

The persuader construct, or the extent to which a person is able to debate with and
inﬂuence others adaptively, is hypothesized to have a substantial positive correlation with
argumentativeness as measured by Infante and Rancer’s (1982) argumentativeness
measure. Argumentativeness is deﬁned as the predisposition to advocate and defend
positions on controversial issues while attacking other’s stances on those issues. The
persuader construct is also hypothesized to correlate positively with perspective taking as
measured by Davis’s (1983) perspective taking measure. Perspective taking is deﬁned as
the ability to adopt the psychological point of view of others. Social anxiety is expected
to have a positive, albeit weaker, correlation with the persuader construct given that

nervousness during social interaction likely makes it difﬁcult to persuade effectively.

l7

Because need for afﬁliation, value-relevant involvement, and empathic concern are not
likely to have direct causal relationships with the persuader construct, the correlations

between the persuader construct and those four variables are hypothesized to be weaker

 

 

(see Figure 2).
Afﬁliation Opinion
Ample
Null Positive Null
Ample
Positive
Argue Persuade Nu"
Ample
Ample Positive Null
Positive
Perspective Activism

Figure 2. Hypothesized convergent and divergent validity nomological network for the
persuader construct.

Notes. Persude = Persuader, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve = Value-
Relevant Involvement, Afﬁliation = Need for Affiliation, Perspective = Perspective Taking, Empathy =
Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism

Given that the deﬁnition of the healthy lifestyle maven involves the motivation to
develop expert healthy lifestyle knowledge along with the drive to share that information
with others who may be in need, it is hypothesized that the maven construct will have an
ample positive correlation with value-relevant involvement as measured by a modiﬁed
version of cm and Boster’s (2005) value-relevant involvement measure. Value-relevant

involvement is deﬁned as a motivational state that links knowledge to the activation of

18

attitudes central to a person’s values. The healthy lifestyle maven construct is also
hypothesized to have an ample positive correlation with empathic concern as measured
by Davis’s (1983) empathic concern measure. Empathic concern is deﬁned as the
proclivity to experience feelings Of sympathy and concern for others in need. Finally, the
healthy lifestyle maven construct is hypothesized to correlate within sampling error of
zero with social anxiety, need for afﬁliation, argumentativeness, and perspective taking

because of the lack of direct causal relationships (see Figure 3).

 

Ample
Null Positive Null
Ample
Positive
Argue Nu" Maven
Ample
Null POSItIVC Ample
Positive

 

Figure 3. Hypothesized convergent and divergent validity nomological network for the
healthy lifestyle maven construct.

Notes. Maven = Healthy Lifestyle Maven, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve
= Value-Relevant Involvement, Afﬁliation = Need for Afﬁliation, Perspective = Perspective Taking,
Empathy = Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism

The remaining two satellite constructs were chosen so that the data produced by
them would correlate substantially with the data from the connector, persuader, and

healthy lifestyle maven measures. Particularly, it is hypothesized that because the traits of

19

connector, persuader, and healthy lifestyle maven are thought to predict the extent to
which people are superdiffusers it is expected that they will also correlate substantially
with other behaviors prototypical of inﬂuentials, particularly opinion leadership as
measured by Flynn, Goldsmith, and Eastrnan’s (1996) opinion leadership measure and
healthy lifestyle activism as measured by a modiﬁed version of Coming and Myers’
(2002) conventional activism orientation measure. Opinion leadership is deﬁned as a
predisposition to exert a disproportionate amount of inﬂuence on the decisions of others
(Rogers & Cartano, 1962). Healthy lifestyle activism is deﬁned as a trait-like
predisposition to engage in low-risk behavior with the goal of advocating socially a
healthy lifestyle. Because the connector, persuader, and healthy lifestyle maven
constructs are conceptualized as unique components of the superdiffuser and not
isomorphic with opinionileadership or healthy lifestyle activism it is hypothesized that
the bivariate distributions between each of the three superdiffuser components and
opinion leadership or healthy lifestyle activism will demonstrate heteroscedasticity.
Speciﬁcally, the high end of the bivariate distribution will show more variance than the
low end. This effect results from all non — connectors, persuaders, or healthy lifestyle
mavens also not being opinion leaders or healthy life style activists, but not all
connectors, persuaders, or healthy lifestyle mavens being opinion leaders or healthy life
style activists.

In addition to reﬁning the reliability of the connector, persuader, and healthy
lifestyle maven measures and assessing their content and construct validity, there is also a
need to test the extent to which the superdiffuser measures exhibit predictive validity.

Predictive validity refers to the extent to which a measure predicts scores on a behavioral

20

criterion. Because connectedness in part refers to people’s number of acquaintances,
scores on the connector measure will correlate with the number of acquaintances people
have to the extent that the connector measure has predictive validity. Furthermore,
because persuasiveness is in part determined by argumentative adaptability, scores on the
persuader measure will correlate with people’s ability to formulate complex arguments to
the extent that the persuader measure has predictive validity. Finally, because healthy
lifestyle mavens are characterized in part by content domain expertise, scores on the
healthy lifestyle maven measure will correlate with people’s content domain knowledge
to the extent that the maven measure has predictive validity.

This research tests these claims in three parts. In Study 1 the new connector,
persuader, and healthy lifestyle maven items are combined with the Boster et al. (2006)
items in order to assess the content validity and potential reliability improvements to the
overall measures, that result from the inclusion of the new items. Study 1 also assesses
the content validity of the need for afﬁliation (Edwards, 1959), perspective taking (Davis,
1983), empathic concern (Davis, 1983), opinion leadership (Flynn et al., 1996), and
healthy lifestyle activism (Corning & Myers, 2002) measures. Study 1 does not examine
the content validity of the original social anxiety (Leary, 1983), argumentativeness
(Infante & Rancer 1982), and value-relevant involvement (Cho & Boster, 2005) measures
because Boster et al. (2006) replicated content valid measurement models for those
measures in two independent samples. Finally, because additional adjustments to the
format of the connector, persuader, and healthy lifestyle maven measure will be made in

Study 2, Study 1 refrains from examining their construct validity.

21

The ability of the content valid measurement models for the social anxiety,
argumentativeness, and value-relevant involvement measures obtained in Boster et al.
(2006) to replicate a third time will be assessed in Study 2. Study 2 also compares the
categorical Likert scaling used by Boster et al. (2006) against direct interval-estimation
scaling to determine if direct interval-estimation can improve the reliability of the
measures as hypothesized. Furthermore, Study 2 examines the extent to which the
nomological networks for the connector, persuader, and healthy lifestyle maven measures
are consistent with theory as an indicator of the measure’s construct validity.

Study 3 examines the predictive validity of the reﬁned connector, persuader, and
healthy lifestyle maven measures. The ability of the connector measure to predict the
number of acquaintances people have is tested with a phone book test. The persuader
measure’s ability to predict people’s argumentative complexity is tested with an
argumentative skill interview. Finally, the healthy lifestyle maven measure’s ability to

predict content domain knowledge is tested with a healthy lifestyle knowledge test.

22

STUDY 1

Method

Subjects. Study 1 sampled 178 students enrolled in undergraduate Communication
classes at a large Midwestern university. Of the 178 students sampled, 73% were females
and 27% were males. On average, the sample was 20.65 (s = 2.38) years old and 3.14 (s =
.73) years into their college career. When asked if they had ever held a leadership
position, 85% responded yes. These students were compensated with course credit for
their participation. Sampling was constrained to students at least 18 years old.

Procedures. Participation was solicited in a classroom setting immediately before
starting. The researcher described Study 1 as, “an examination of social opinion and
social behavior involving the completion of a questionnaire containing 73 items that ask
respondents to report the extent of their agreement with each item. ” The researcher
informed the students agreeing to participate (there were no students opting not to
participate) that 45 minutes were allocated for the completion of the questionnaire. These
subjects were also instructed to return their completed questionnaire to the researcher
before leaving. At this point, the questionnaire was administered and a brief set of
instructions regarding how to complete the items on the questionnaire was reviewed
(Appendix A) before allowing the subjects to begin.

Instrumentation. In addition to a measure containing several demographic items
(Appendix B), the questionnaire was composed of eight separate measures organized
serially. The eight measures were designed to measure the constructs of connector,
persuader, healthy lifestyle maven, need for afﬁliation, perspective taking, empathic

concern, opinion leadership, and healthy lifestyle activism. The items on each measure

23

were a Likert format with eight point categorical response scales ranging from disagree
strongly to agree strongly. The scales were coded to reﬂect that a response of agree
strongly indicated that the respondent possessed more of the construct being measured
than a response of disagree strongly. Thus, when coded, a higher number indicated more
of the construct than a lower number. It is hypothesized that the items forming each
measure ﬁt unidimensional measurement models.

The connector construct was measured by a set of 10 items designed to reﬂect that
connectors enjoy forming and maintaining relationships that serve as bridges between
social groups and people who would not otherwise know one another. Items one through
four in Appendix C were taken from an initial set of connector items written by Boster et
al. (2006). In order to create a measure that embodies the construct more fully, items ﬁve
through 10 were added to the connector measure.

The persuader construct was also measured by a set of 10 items (Appendix D).
These items, however, were designed to measure the extent to which the respondent can
inﬂuence people to adopt a particular position, employing techniques adapted to the
speciﬁc situation if perceived to be necessary. Like the connector measure, items one
through four were from Boster et al. (2006) and items ﬁve through 10 were developed to
create a measure more representative of the persuader construct.

Similar to the connector and persuader constructs, the healthy lifestyle maven
construct was assessed by a set of 10 items (Appendix E). The ﬁrst four items were taken
from Boster et al. (2006) and the remaining six were developed for this study to create a

set of 10 items more representative of the construct than the original four. All 10 items

24

were created to measure the extent to which the respondent enjoys learning about an issue
so that others can be helped by sharing the knowledge with those who are most in need.

Need for afﬁliation was measured by the nine manifest afﬁliation need items from
the Edwards personal preference schedule (Edwards, 195 9) (Appendix F). The items
purport to measure an overt diffuse secondary need consistent with Murray’s (l 93 8)
conceptualization that speciﬁes a motivation to draw near, cooperate with, and remain
loyal to allied others. This measure of affiliation has been shown to possess some degree
of predictive validity (Kirchner, Dunnette, & Mousley, 1960) and reliability, r,e,,-,e,e,, =
.77, (Mann, 1958). The Edwards personal preference schedule (Edwards, 1959),
however, is an ipsative measure of forced choice comparisons between items measuring
competing needs. For the sake of scaling consistency with the other measures in the
pretest, each of the items was presented in a Likert format.

Davis’ (1980) interpersonal reactivity index was applied to measure the construct
of perspective taking. The construct was measured by the seven item perspective taking
component of the index (Appendix G). The items on this measure focus on the
respondent’s ability to adopt the psychological point of view of others. Davis (1983)
demonstrated the construct validity of this measure showing that it correlates modestly
with other measures of interpersonal reactivity. This perspective taking measure also has
been shown to be reliable, or 2: .74 (Davis, 1983).

The interpersonal reactivity index (Davis, 1980) was also employed to measure
empathic concern. The seven item empathic concern component of the measure was used
for this purpose (Appendix H). These seven items were designed to measure the tendency

for the respondent to experience feelings of sympathy and concern for others in need. The

25

construct validity of this measure has been shown in that scores correlate with measures
of concern for others (Davis, 1983). The measure also has also been shown to be reliable,
a z .74 (Davis, 1983).

The six item Flynn et al. (1996) opinion leadership measure (Appendix I) was
used to measure opinion leadership in the form of a tendency to exert a disproportionate
amount of inﬂuence on the decisions of others. Flynn et a1. (1996) demonstrated the
predictive validity of this measure by correlating it with a series of information seeking
and sharing behaviors. This previous research also found the reliability of the measure to
be, a = .78 (Flynn et al., 1996).

The healthy lifestyle activism measure (Appendix J) was a modiﬁed version of
the conventional activism orientation measure (Corning & Myers, 2002). The items on
the original measure were designed to measure political activism. The original politically
focused measure was found to be largely construct valid, in that it correlated with other
measures of political activism, and have a reliability estimated by a to be .96 (Coming &
Meyers, 2002). The political focus of the items was changed, however, to a healthy
lifestyle focus by replacing references to politics with references to a healthy lifestyle.
The modiﬁed measure was employed to measure the anticipated tendency to engage in a
low-risk form of healthy lifestyle activism at a point in the future. The construct was
measured with a set of 10 healthy lifestyle activism items randomly selected from a larger
set of 28 activism items on the Corning and Myers (2002) measure.

Results
Preliminary review of the eight factor inter-item correlation matrix revealed

several items that failed to exhibit internal consistency, parallelism, or both (Hunter &

26

Gerbing, 1982). Those items were discarded and the measurement model was subjected
to testing with CFA. The CFA method applied in this study to test measurement models
calculated factor loadings with a centroid estimation procedure using communalities on
the diagonal (Hunter & Hamilton, 1992). Factor loadings and model speciﬁcations were
used to generate a predicted inter-item correlation matrix which was subtracted from the
observed inter-item correlation matrix. The resultant residual matrix was examined to
assess model ﬁt. To the extent that the residuals in the matrix were within sampling error
of zero, the model was said to exhibit ﬁt with the data. This process revealed several
more items exhibiting gross misﬁt. These items were also removed from the analyses and
an eight factor measurement model consistent with the data was achieved.1 Table 2
presents the factor loadings and factor correlation matrix for the model.

Evidence generated by the analyses demonstrated that the eight factor model ﬁt
the data well. For example, all factor loadings were ample, with no item possessing a
factor loading less than .51. Additionally, this model’s residual matrix was composed of
841 numbers which could produce 43 statistically signiﬁcant residuals by chance when p
< .05. The actual number of residuals for the model not within sampling error of zero was
23, with the largest at .29. Finally, the average residual in the matrix was well within
sampling error of zero, Root Mean Square Error (RMSE) = .07. Focused examination of

the nested three factor connector, persuader, and healthy lifestyle maven measurement

 

' It is worth noting that some have argued against modifying a measure by removing or adding items, under
the premise that it destroys the measure’s integrity. That argument, however, is unfounded because it in
part assumes that the items contained within the measure are the only items that could possibly measure the
construct in question. The argument also assumes that the items in the measure are more optimal or
representative of the construct being measured than the inﬁnite remainder of items not contained within the
measure. Although these assumptions are more tenable for precisely deﬁned constructs with precisely
developed measures, this is not the case for the majority of social scientiﬁc measurement. Therefore, in the
face of empirical evidence suggesting some items do not measure what they purport to measure, the
modiﬁcation of existing measures by removing items displaying evidence of invalidity is warranted.

27

model revealed that the newly included items resolved the parallelism concerns observed
between the connector and persuader measures observed in the Boster et al. (2006) study.
Only two residuals across the three factor model were not within sampling error of zero,
with the largest being .22 (four could be expected due to chance), RMSE = .09.

Table 2

Study 1 Factor Loadings and Factor Correlation Matrix

 

Item Con Per Mav Aff Persp Emp grin Act

 

Connector 1 0.62 0.28 0.12 0.32 0.15 0.13 0.06 0.06
Connector 2 0.78 0.19 0.14 0.22 0.06 0.03 0.17 0.07
Connector 3 0.85 0.26 0.17 0.30 0.20 0.14 0.13 0.08
Connector 4 0.70 0.22 0.20 0.28 0.24 0.12 0.16 0.14
Connector 6 0.77 0.14 0.20 0.18 0.08 -0.03 0.15 0.09
Connector 9 0.66 0.22 0.19 0.19 0.17 0.01 0.17 0.09

 

Persuader 1 0.23 0.72 0.18 0.15 0.10 0.04 0.12 -0.03
Persuader 3 0.23 0.81 0.14 0.16 0.03 0.03 0.14 -0.01
Persuader 4 0.31 0.79 0.22 0.22 0.13 0.11 0.10 0.01
Persuader 5 0.15 0.86 0.18 0.13 0.16 0.10 0.05 -0.05
Persuader 6 0.14 0.76 0.12 0.26 0.12 0.04 0.00 0.00
Persuader 7 0.22 0.81 0.13 0.18 0.12 0.06 0.06 -0.01
Persuader 9 0.21 0.76 0.28 0.19 0.15 0.08 0.17 0.07
Persuader 10 0.38 0.71 0.23 0.28 0.20 0.09 0.06 0.02

 

Maven 2 0.18 0.20 0.87 0.21 0.25 0.07 0.66 0.37
Maven 4 0.14 0.21 0.87 0.14 0.22 0.10 0.79‘ 0.44
Maven 5 0.27 0.26 0.77 0.24 0.27 0.15 0.55 0.39
Maven 6 0.21 0.23 0.86 0.06 0.20 0.11 0.66 0.50
Maven 8 0.14 0.18 0.83 0.24 0.29 0.07 0.62 0.43
Maven 9 0.24 0.20 0.85 0.25 0.33 0.17 0.67 0.55
Maven 10 0.20 0.12 . . Q_§Z_ 0.22 0.25 0.08 0.78 0.52

 

Afﬁliation2 0.26 0.17 0.15 ‘: (10.861 0.35 0.48 0.07 0.11
Afﬁliation3 0.31 0.17 0.20 ’j"*0;68§ 0.32 0.43 0.13 0.14
Afﬁliation6 0.26 0.13 0.14 0.56 0.35 0.38 0.08 0.13
Afﬁliation8 0.06 0.18 0.13 0.53 0.43 0.33 0.09 0.11

28

Table 2 (cont’d)

 

Item Con Per Mav Aff Persp Emp Opin Act

Perspective 1 0.24 0.10 0.21 0.43 0.72 0.45 0.13 0.24
Perspective 3 0.06 0.11 0.23 0.47 . 0.69 0.54 0.15 0.32
Perspective 6 0.18 0.20 0.12 0.34 ~ - 0.53 0.24 0.11 0.07
Perspective 7 0.05 0.00 0.21 0.15 0.60 0.32 0.17 0.29
Empathic 3 0.07 0.09 0.08 0.37 0.45 0.65 0.05 0.18
Empathic 4 0.14 0.10 0.15 0.55 0.51 0.81 0.11 0.23
Empathic 6 0.00 0.00 -0.09 0.30 0.31 0.55 -0.06 0.19
Empathic 7 0.03 0.04 0.19 0.45 0.34 7 , 0.46:2. 0.20 0.33

 

 

Opinion3 0.06 -004 0.63 0.12 0.18 0.26 f 052 0.40
Opinion4 0.20 0.15 0.64 0.12 0.23 0.03 083 0.42
OpinionS 0.11 0.09 0.57 0.12 0.13 0.04 “‘easai‘i 0.43
Qpinion6 0.22 0.14 0.65 0.07 0.12 0.02 92L 0.46
Activism2 0.03 -0.01 0.50 0.12 0.29 0.27 0.54 ' 0.79
Activism3 0.07 0.05 0.45 0.13 0.29 0.33 0.45 0.78
Activism4 0.19 0.00 0.44 0.17 0.30 0.32 0.45 0.84
ActivismS 0.11 0.04 0.30 0.15 0.26 0.24 0.30 0.66

Activism 6 0.08 -0.07 0.45 0.16 0.29 0.24 0.44 0:87

 

 

Factor
Connector
Persuader 0.30
Maven 0.24 0.24
Afﬁliation 0.34 0.25 0.23
Perspective 0.21 0.16 0.31 0.55
Empathic 0.09 0.09 0.13 0.64 0.61
Opinion 0.19 0.11 0.81 0.14 0.22 0.11
Activism 0.12 0.00 0.54 0.19 0.36 0.35 0.56

 

Notes. Con = Connector, Per = Persuader, Mav = Healthy Lifestyle Maven, Aff = Need for Afﬁ—liation,
Persp = Perspective Taking, Emp = Empathic Concern, Opin = Opinion Leadership, Act = Healthy
Lifestyle Activism, n = 178

Consequently, the eight sets of items, each with substantial evidence of content
validity were formed into eight indices by calculating the mean response across each set

of items. The connector index (connector items: 1, 2, 3, 4, 6, and 9) ranged from 1.33 to

29

8.00 and was distributed with a slight negative skew and slightly leptokurtic, M = 6.04, s
= 1.10, a = .87, Standardized Item a (SJ. a) = .87. Scores on the persuader index
(persuader items: 1, 3, 4, 5, 6, 7, 9, and 10) ranged from 1.88 to 8.00. The mean of the
persuader distribution was 5.85 (s'= 1.10), a = .92, and SJ. (1 = .92. There was a slight
negative skew to the distribution; it was also slightly leptokurtic. The distribution of the
healthy lifestyle maven index (maven items: 2, 4, 5, 6, 8, 9, and 10), ranging from 1.29 to
8.00, was normal, M = 5.00, s = 1.57. The reliability, estimated by a = .94 and SJ. a =
.94. The distribution of the need for afﬁliation index (need for afﬁliation items: 2, 3, 6,
and 8) was negatively skewed and leptokurtic (Range = 4.00 -— 8.00), M = 7.14, s = .74, a
= .72, SJ. (1 = .75. Responses on the perspective taking index (perspective taking items:
1, 3, 6, and 7), ranged from 1.50 to 8.00, and were distributed leptokurticly, M = 5.80, s =
1.07, a = .73, S]. a = .73. The empathic concern index (empathic concern items: 3, 4, 6,
and 7) was approximately normally distributed and ranged from 3.25 to 8.00. The indices
mean response was 6.35 (s = 1.02), a = .74, SJ. (1 = .75. Scores on the opinion leader
index (opinion leader items: 3, 4, 5, and 6) also approximated normality, ranging from
1.25 to 8.00. The opinion leader distribution had a mean of 4.32 (s = 1.43). The
reliability, estimated by a = .85 and 5.1. a = .85. Finally, the healthy lifestyle activism
index (activism items: 2, 3, 4, 5, and 6) also approached a normal distribution (Range =
1.00 -— 8.00), M = 3.86, s = 1.56, a = .89, S]. a = .89. Table 3 presents a summary of the
indices’ descriptives.

There were no substantial relationships between respondent sex and scores on the
connector, persuader, or healthy lifestyle maven measures. All three relationships were

within sampling error of zero. The same ﬁnding was observed for respondent age and

30

year in school. The relationship between whether the respondent reported holding a
position of leadership and the connector measure (r = .31, r Corrected for Attenuation
due to Measurement Error (r ') = .33) was not within sampling error of zero, however. A
similar relationship was observed between whether the respondent reported holding a
position of leadership and the persuader measure (r = .16, r’ = .17). Given the deﬁnitions
of the connector and persuader constructs, however, it is not surprising that they share a
relationship with leadership.

Table 3

Study 1 Measure Descriptives

 

 

Measure M s Min - Max a S]. a Skewness Kurtosis
Connector 6.04 1.10 1.33 - 8.00 0.87 0.87 -0.97 1.84
Persuader 5.85 1.10 1.88 - 8.00 0.92 0.92 -O.87 1.10
Maven 5.00 1.57 1.29 - 8.00 0.94 0.94 -0.04 -0.69 .
Affiliation 7.14 0.74 4.00 - 8.00 0.72 0.75 -1.31 2.10
Perspective 5.80 1.07 1.50 - 8.00 0.73 0.73 -0.56 1.13
Empathic 6.35 1.02 3.25 - 8.00 0.74 0.75 -O.52 -0.08
Opinion 4.32 1.43 1.25 - 8.00 0.85 0.85 -0.13 -0.51
Activism 3.86 1.56 1.00 - 8.00 0.89 0.89 0.05 -0.84

 

Notes. Maven = Healthy Lifestyle Maven, Affiliation = Need for Afﬁliation, Perspective = Perspective
Taking, Empathic = Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle

Activism 11 = 178

4

By applying Boster et al.’s (2006) quantiﬁcation of superdiffusers as people with

responses at least one standard deviation above the mean on the connector, persuader,

and healthy lifestyle maven measures, 1.7% of this study’s sample were classiﬁed as

superdiffusers. Interestingly, 15% of the sample was one standard deviation above the

mean on the opinion leader measure and 17% on the activist measure when applying the

same criteria to the opinion leader or healthy lifestyle activism measures. Although not

sufﬁcient evidence, these differential percentages are consistent with the hypothesis that

31

the superdiffuser is indeed an entity distinct from opinion leaders and activists. Although
the bivariate scatterplots presented in Figures 4 and 5 show data inconsistent with the
heteroscedasticity hypothesis presented earlier, the plots did provide further evidence
consistent with the conclusion that superdiffusers are conceptually distinct from opinion
leaders and activists. That is, each superdiffuser measure displayed unique relationships

with the opinion leader measure and the activism measure.

 

Connector

 

 

Persuader

 

 

Maven

   

 

Opinion

 

 

 

 

 

     

 

Figure 4. Opinion leadership by superdiffuser scatterplot matrix.

Notes. Maven = Healthy Lifestyle Maven, Opinion = Opinion Leadership, n = 178

32

 

Connector

 

 

Persuader

 

 

Maven

   

 

Activism

 

 

 

 

 

 

 

Figure 5. Healthy lifestyle activism by superdiffuser scatterplot matrix.
Notes. Maven = Healthy Lifestyle Maven, Activism = Healthy Lifestyle Activism, n = 178

Examining these relationships in more detail produced an even more interesting
ﬁnding. Speciﬁcally, only the healthy lifestyle maven index appears to be associated
strongly with opinion leadership and healthy lifestyle activism. The multiple regression
analysis regressing opinion leadership onto the connector, persuader, and healthy lifestyle
maven measures as well as all two-way and three-way interactions found only the healthy
lifestyle maven construct as a substantial and statistically signiﬁcant predictor. For
example, the standardized regression coefﬁcient for the connector measure was 6 = .03 (t
(170) = 0.58, ns, ,3 Corrected for Attenuation due to Measurement Error ([3 ’) = .04), for

the persuader measure was 3 = -.10 (t (170) = -1.67, ns, ,6 ’ = -.13), and for the healthy

33

lifestyle maven measure was ,8 = .74 (t (170) = 13.48, p < .05, ﬂ’ = .84). A similar effect

was observed when healthy lifestyle activism scores were regressed onto the connector,

persuader, and healthy lifestyle maven indices, as well as all two-way and three-way

interaction terms. Once again, the healthy lifestyle maven measure emerged as the only

substantial predictor (,6 = .52, t (170) = 7.72, p < .05, ,6’ = .59). Table 4 presents the

 

 

 

regression models.
Table 4
Study 1 Regression Models
Model Component B se ,8 t si g B ’
Opinion Leadership Constant 4.33 0.08 55.07 p < .05
F (7, 170) = 28.79 C 0.00 0.07 0.03 0.58 ns 0.04
p < .05 P -0.13 0.08 -0.10 -1.67 ns -0.13
Rad,- = .72 M 0.68 0.05 0.74 13.48 p < .05 0.84
C x P -0.02 0.06 -0.02 -0.31 ns -0.03
C x M 0.00 0.04 0.00 0.06 ns -0.01
P x M -0.03 0.04 -0.04 -0.72 ns -0.05
C x P x M 0.03 0.05 0.04 0.61 ns 0.04
Activism Constant 3.89 0.11 36.70 p < .05
F (7, 170) = 10.63 C 0.05 0.10 0.03 0.48 ns 0.05
p < .05 P -0.27 0.10 -0.19 -2.59 p < .05 023
R04,- = .53 M 0.52 0.07 0.52 7.72 p < .05 0.59
C x P -0.07 0.08 -0.07 -0.91 ns -0.08
C x M 0.00 0.06 0.00 0.00, ns 0.01
P x M 0.04 0.06 -0.05 -0.73 ns -0.06
CxPxM 0.12 0.06 0.15 1.86 ns 0.15

 

Notes. C = Connector, P= Persuader, M = Healthy Lifestyle Maven, components are mean centered to
reduce multicollinearity, n = 178

Discussion

The results demonstrate the presence of three internally consistent and parallel
sets of indicators, composed of items from the original measure development of Boster et

al. (2006) as well as new items, purported to measure the focal constructs of the

34

superdiffuser: connector, persuader, and healthy lifestyle maven. The measure
improvements increased the mean level of reliability across the three measures from S].
a = .84, as observed in the original measure development studies, to S]. a = .91.

The reliability of the connector measure was, however, lower than the Spearman-
Brown estimate. Even so, the new items caused the reliability to improve relative to the
reliability estimates obtained in the two studies of Boster et al. (2006). This discrepancy
between the observed increase and the Spearman-Brown predicted increase was due to
the fact that, on average, the new items were weaker indicators of the connector construct
than the original items.

The increases for the other measures all approximated the Spearman-Brown
prophecy formula estimates. It is also worth noting that the response distributions for
each of the measures moved towards normality relative to the distributions observed in
Boster et al. (2006). Given this set of ﬁndings, it is reasonable to conclude that the new
items generally improved the psychometric qualities of the connector, persuader, and
healthy lifestyle maven measures.

The CFAs performed on the Study 1 data also allowed for improvements to the
need for afﬁliation, perspective taking, empathic concern, opinion leadership, and healthy
lifestyle activism measures by assisting in the identiﬁcation and removal of items
displaying content invalidity. Although the improvements did not result in optimal

measures of all constructsz, the reﬁnement of the measures resulted in a considerable

 

2 The perspective taking and empathic concern measures contained items with disproportionate numbers of
statistically signiﬁcant residuals, indicative that these two measures were less content valid than the other
measures. Although the accumulating residuals gives some reason for concern, because RMSE for the most
severely affected matrices was not extremely poor (perspective taking by need for afﬁliation RMSE = .10
and empathic concern by opinion leadership RMSE = .11) and the quality of measurement for these two
measures was not a primary focus of the study, the measures were considered acceptable.

35

reduction in the amount of measurement error present in the assessment of each
construct. Consequently, these reﬁned measures will be employed in Study 2 because the
reduced measurement error improves the accuracy of estimating the correlations among
the constructs; an important consideration for the nomological network approach to
assessing construct validity.

Study 1 also revealed two ﬁndings that will be considered further in Study 2.
First, 1.7% of the subjects in Study 1 were classiﬁed as superdiffusers according to their
responses on the connector, persuader, and healthy lifestyle maven measures.
Interestingly, 1.7% of the subjects assessed in the two studies of Boster et al. (2006) were
superdiffusers. Based on these data it is expected that 1.7% of the subjects in Study 2’s
sample will also be superdiffusers.

The second ﬁnding that will be considered further in Study 2 is the uniqueness of
the superdiffuser from traditional conceptualizations of opinion leader or activist. Put
differently, the regression analyses conducted in Study 1 revealed that opinion leadership
and activism are not accounted for by the connector and persuader constructs. This effect
is consistent with traditional conceptualizations of opinion leadership and activism which
focus on tendencies to have expertise about new ideas or products, without mentioning
characteristics of connectedness or persuasiveness. What is more, although these
tendencies appear akin to the healthy lifestyle maven in nature and the healthy lifestyle
maven construct correlates substantially with opinion leadership and activism, the
measurement model assessment and factor correlations from Study 1 indicate that the
healthy lifestyle maven construct is not isomorphic with opinion leadership and activism.

The data from Study 2 will be examined to assess the stability of this ﬁnding.

36

In addition to exploring these issues, Study 2 will examine two main hypotheses.
First, Study 2 offers a comparison between the Likert scaling, employed up until this
point in the development of the superdiffuser measures, and direct interval-estimation
scaling. It is predicted that direct interval-estimation will offer improvements to the
measures’ reliabilities because of the possibility for increased inter-item variance and
hence increased inter-item response covariance relative to Likert scaling. Study 2 also
examines the content validity of the measures in detail by examining the hypothesized
nomological networks for the connector, persuader, and maven measures presented in
Figures 1, 2, and 3. According to the method, the extent that the pattern of observed
correlations matches the pattern of hypothesized expectations is evidence consistent with

construct validity (Cronbach & Meehl, 1955).

37

STUDY 2

Method

Subjects. The sample for this study consisted of 300 undergraduate students from
a large Midwestern university who were on average 19.93 (s = 1.50) years old and had
spent 2.64 (s = 1.23) years in college. Additionally, the sample was 62% female and 38%
male. When asked if they had held a position of leadership since beginning college, 49%
indicated yes. The students were enrolled in undergraduate Communication classes and
received course credit in exchange for their participation. Sampling was constrained to
students at least 18 years old who had not participated in Study 1.

Design. One of the goals for Study 2 was to examine the effects that response
scaling can have on the reliability, content validity, and construct validity of a measure.
Speciﬁcally, a categorical Likert scaling format was compared against a direct interval-
estimation method. Therefore, employing random assignment, 150 subjects responded to
items that were scaled categorically and 150 subjects responded to items sealed with
direct interval-estimation. Demographic differences between the subjects in each
condition were within sampling error of zero for sex and having been in a leadership
position. Subjects in the two conditions did, however, differ on two demographic
variables. Subjects in the Likert condition were older (M = 20.40, s = 1.55) than subjects
in the interval-estimation condition (M = 19.46, s = 1.29), t (298) = 5.71, p < .05, r = .32,
and subjects in the Likert condition had spent more time in college (M = 2.97, s = 1.19)
than subjects in the interval-estimation condition (M = 2.31, s = 1.18), t (298) = 4.89, p <
.05, r = .26. Because there are no major life-events that occur systematically between

19.46 years old and 20.40 years old or between the 2.31 year of college and the 2.97 year

38

of college it is unlikely that these differences, although statistically signiﬁcant, are
substantively important.

Procedures. Students were solicited for participation via a departmental subject
pool that enabled students enrolled in any of several undergraduate courses to participate.
The study was described as, “an examination of social opinion and social behavior
involving the completion of a questionnaire containing 64 items that ask respondents to
report their evaluation of each item. ” Students choosing to participate in the study after
reading the description were given an appointment time and a location at which to meet.
Approximately 25 subjects participated at each appointment time. At each time, a
researcher welcomed the subjects before instructing them to have a seat in one of the
desks that ﬁlled the room.

After the subjects were seated the researcher described the study and informed the
subjects that participation would take approximately 30 minutes. At this point the
researcher administered the study materials to the subjects. Before allowing the subjects
to begin, the researcher reviewed one of two sets of instructions depending on the
randomly assigned condition. In one-half of the sessions, the items on the questionnaire
were administered with a categorical eight point Likert scaling system. Consequently the
researcher administered the instructions found in Appendix A. In the second one-half of
the sessions, the questionnaire was administered with a direct interval-estimation scaling
system where a person completely lacking the construct in question (zero points) and a
person who was the prototype of the construct in question (100 points) were anchors. In
this condition, the researcher administered the response instructions found in Appendix

K. The researcher then answered any questions before allowing the subjects to begin.

39

Instrumentation. In addition to a short demographics measure (Appendix B), the
questionnaire was composed of 11 independent measures arranged in succession. The 11
measures were designed to measure the constructs of connector, persuader, healthy
lifestyle maven, need for afﬁliation, social anxiety, perspective taking,
argumentativeness, empathic concern, value-relevant involvement, opinion leadership,
and healthy lifestyle activism. The two versions of the questionnaire contained identical
items but different scaling formats. One version was composed of eight point categorical
response scales ranging from disagree strongly to agree strongly. The scales were coded
to reﬂect that a response of agree strongly indicated that the respondent possessed more
of the construct being measured than a response of disagree strongly. The second version
of the questionnaire contained direct interval estimation response scales ranging from
zero to 100. In either case, a higher number indicated more of the construct than a lower
number.

The reﬁned measures from Study 1 were employed to measure the connector,
persuader, healthy lifestyle maven, need for afﬁliation, perspective taking, empathic
concern, opinion leadership and conventional social activism constructs. The measures of
social anxiety, argumentativeness, and value-relevant involvement were reﬁned versions
of existing measures. The reﬁnements were made across two independent samples by
Boster et al. (2006). That study provided data consistent with unidimensional
measurement models for measures of social anxiety, argumentativeness, and value-
relevant involvement through the removal of several items lacking internal consistency,

parallelism, or both on each of the measures.

40

Speciﬁcally, social anxiety was measured with six items from Leary’s (1983)
interaction anxiousness measure (Appendix L) that measure the extent of the
respondent’s aversion to social interactions. Leary (1983) showed that this measure is
both construct valid in that it correlates substantially with other measures of social
anxiety and shyness. Leary (1983) also demonstrated that this measure is reliable, or =
.89.

Argumentativeness was measured by seven items from Infante and Rancer’s
(1982) argumentativeness measure (Appendix M) designed to measure the respondent’s
predisposition to advocate and defend positions on controversial issues while attacking
another’s stances on those issues. Infante and Rancer (1982) demonstrated that some
initial evidence the measure is a construct valid and reliable, a = .87, indicator of
argumentativeness.

Finally, a reﬁned ﬁve item version of Cho and Boster’s (2005) value-relevant
involvement measure (Appendix N) was used to measure the extent to which the healthy
lifestyle maven issue was linked to the activation of attitudes that are central to the
respondent’s values. Cho and Boster (2005) provided content and construct validity
evidence for the measure delineating it from two other types of involvement. Cho and
Boster (2005) also reported reliability as a =1 .87 across several different value domains.
Results

CFA was employed to assess the 11 factor measurement model’s ﬁt with the
Study 2 data. The Hunter and Hamilton (1992) method described in Study 1 was
employed to diagnose items failing to exhibit internal consistency, parallelism, or both

(Hunter & Gerbing, 1982) and assess model ﬁt across both scaling format conditions.

41

Seven items exhibiting one or more of these problems were identiﬁed (connector 1,
healthy lifestyle maven 2, empathic concern 6, social anxiety 6, argumentativeness 2,
opinion leader 3, activism 2) and removed from further analyses. Subsequent analyses
resulted in an 11 factor model that ﬁt well the data produced by both the categorical
Likert scaling and direct interval-estimation scaling. Examination of the nested three
factor connector, persuader, and maven measurement model also revealed characteristics
of good ﬁt for both scaling formats. Table 5 presents the ﬁt statistics for both models
across both scaling formats.

Table 5

Sim)» 2 CFA Fit Statistics

 

 

 

Model Statistic Likert DIE
11 Factor Minimum Factor Loading .51 .47
RMSE Model 0.07 0.07
# of Statistically Signiﬁcant Residuals] 29 29
Maximum Residual 0.24 0.24
3 Factor Minimum Factor Loading .52 .63
RMSE Model 0.07 0.08
# of Statistically Signiﬁcant Residualsz 2 6
Maximum Residual 0.19 0.21

 

Notes. DIE = Direct Interval-Estimation, ‘ 68 statistically signiﬁcant residuals by chance when p < .05, 2 8
statistically signiﬁcant residuals by chance when p < .05, N = 300, n = 150

Although the ﬁt statistics in Table 5 demonstrate that the method of scaling had
little impact on the overall ﬁt of the model, the factor loadings and factor correlations
may have been impacted. Therefore, the Hunter and Hamilton (1992) CFA method was
modiﬁed to perform a multiple groups CFA. The multiple groups CFA tested for
equivalence of the factor loadings and factor correlations across the Likert scaling and

direct interval-estimation scaling conditions. The procedures for this test ﬁrst involved

42

estimating the parameters of the data collected with the Likert scaling and direct interval-
estimation scaling as would be done in a single group CFA. Then, the Likert scaling
parameters were used to test the ﬁt of the direct interval-estimation scaling data and the
direct interval-estimation scaling parameters were used to test the ﬁt of the Likert scaling
data.

Speciﬁcally, the measurement model with factor loadings constrained to be equal
across the two scaling formats is tested. If the constrained factor loading model ﬁts well
(i.e., minimal error between the predicted and obtained correlation matrices), the
measurement model with factor loadings and factor correlations constrained to be equal
across the two scaling formats is tested. If the data exhibit satisfactory ﬁt given the
constraints placed upon these parameters (i.e., factor loadings and factor correlations),
then they are consistent with the hypothesis that the two response scale formats have
factor loadings and factor correlations within sampling error of each other.

The results of the multiple groups CFA indicated that the factor loadings and
factor correlations are indeed equivalent across scaling formats. The ﬁt of the constrained
factor loading, unconstrained factor correlation 11 factor measurement model was found
to ﬁt acceptably across the scaling formats. The Likert scaling to direct interval-
estimation scaling comparison produced a RMSE = .08 and a maximum residual of .27.
The direct interval-estimation scaling to Likert scaling comparison resulted in a RMSE =
.08 and a maximum residual of .27. Finally, both scaling formats demonstrated
satisfactory ﬁt with the 1 1 factor measurement model where factor loadings and factor
correlations were both constrained to be equal. In the Likert scaling to direct interval-

estimation scaling comparison, RMSE = .11 and the maximum residual was .31. In the

43

direct interval-estimation scaling to Likert scaling comparison, RMSE = .10 and the
maximum residual was .30.

Results of the multiple groups CFA for the nested three factor connector,
persuader, and healthy lifestyle maven model also were consistent with the conclusion
that the method of scaling did not cause factor loading or factor correlation differences.
Both scaling formats were found to ﬁt well with the constrained factor loading,
unconstrained factor correlation three factor measurement model (Likert scaling to direct
interval-estimation scaling: RMSE = .10, maximum residual = .28; direct interval-
estimation scaling to Likert scaling: RMSE = .09 , maximum residual = .21) and the three
factor measurement model with factor loadings and factor correlations constrained to be
equal, (Likert scaling to direct interval-estimation scaling: RMSE = .12, maximum
residual = .30; direct interval-estimation scaling to Likert scaling: RMSE = .11, maximum
residual = .25).

The psychometric properties of the scaling formats were not entirely alike,
however. Direct interval-estimation scaling did result in modestly larger estimates of
reliability than categorical Likert scaling; albeit, these differences were not statistically
signiﬁcant. The mean S.I. a = .87, s = .06, across all 11 measures when direct interval-
estimation scaling was employed as opposed to a mean S.I. a = .86, s = .06, when
categorical Likert scaling was used.

The impact of scaling format on reliability estimates was greater when
considering only the connector, persuader, and healthy lifestyle maven measures where
the mean S]. a = .92, s = .02, for direct interval-estimation scaling as opposed to a mean

S]. a = .90, s = .03, for categorical Likert scaling. AS was the case in Study 1,

44

improvements to the persuader and maven measures approximated the Spearman-Brown
predictions. Furthermore, although the reliability of the connector measure still did not
meet the Spearman-Brown estimate, the reliability and distribution characteristics were
nonetheless improved. Consequently, although the new items and scaling format
modiﬁcation did not lead to the ideal improvements predicted by Spearman-Brown, they
still lead to improvements in psychometric qualities of the measure.

Additionally, although the increase in reliability estimates was not as substantial
as that reported by Silverrnan and Johnston (1975) it is still important. The weaker effect
is most likely explained by the fact that reliability estimates in this study, unlike the
Silverrnan and Johnson (1975) study, were substantial from the onset. Direct interval-
estimation functions to increase reliability estimates through increasing the item variance
which increases item correlations. Furthermore, increases in reliability estimates resulting
from item correlation increases attenuate as the item correlation increases. This inverse
relationship means that the rate of change in the reliability estimate slows as item
correlation increases and the effect size will be correlated negatively with the measure’s
original reliability estimate. Therefore, although the difference between .92 and .90 is
numerically small, it is not insigniﬁcant. The change was consistent across the different
measures and was a 20% increase in what is possible. Therefore, given the content
validity equivalence between the direct interval-estimation scaling and categorical Likert
scaling formats and the superior reliability estimates afforded by direct interval- I
estimation scaling, the remaining analyses will be conducted on the direct interval-
estimation data. Table 6 presents the factor loadings and factor correlation matrix for the

direct interval-estimation measurement model.

45

Table 6

Study 2 Direct Interval-Estimation Factor Loadings and Factor Correlation Matrix

 

Item Con Per Ma Aff Pers Em Anx Arg Inv Opi Act

 

Con 2 0.82 0.28 0.09 0.18 0.13 0.04 -0.38 0.05 0.15 0.19 0.06
Con 3 0.78 0.29 0.05 0.35 0.17 0.17 -0.55 0.15 0.07 0.20 0.10
Con 4 0.63 0.43 0.17 0.29 0.23 0.12 -0.27 0.26 0.12 0.19 0.04
Con 6 0.86 0.34 0.17 0.29 0.19 0.15 -0.36 0.19 0.16 0.23 0.19
Con 9 0.90 0.34 0.26 0.35 0.21 0.21 -0.40 0.23 0.12 0.27 0.22

 

Per 1 0.35 0.76 0.20 0.06 0.19 0.15 -0.20 0.49 0.19 0.23 0.10
Per 3 0.25 0.80 0.03 -0.01 0.03 0.00 -0.21 0.53 0.13 0.11 0.02
Per 4 0.34 0.91 0.03 0.03 0.06 -0.01 -0.29 0.50 0.08 0.08 -0.02
Per 5 0.36 0.88 0.00 0.00 0.09 -0.01 -0.26 0.43 0.06 0.08 -0.08
Per 6 0.38 0.82 0.08 0.01 0.20 -0.01 -0.26 0.44 0.07 0.11 0.00
Per 7 0.38 0.85 0.17 0.05 0.26 0.14 -0.24 0.46 0.09 0.17 0.08
Per 9 0.38 0.86 0.09 0.07 0.11 0.02 -0.31 0.47 0.05 0.12 0.05
Per 10 0.32 0.73 0.09 0.00 0.11 0.08 -0.28 0.49 0.00 0.17 0.09

 

Ma 4 0.22 0.12 0.77 0.22 0.26 0.13 0.01 0.13 0.52 0.69 0.38
Ma 5 0.12 -0.0l 0.84 0.27 0.15 0.14 0.06 0.14 0.47 0.59 0.45
Ma 6 0.10 0.01 0.90 0.23 0.19 0.23 0.05 0.14 0.48 0.65 0.47
Ma 8 0.24 0.17 0.77 0.20 0.24 0.13 -0.05 0.25 0.49 0.65 0.46
Ma 9 0.11 0.14 0.84 0.14 0.21 0.23 0.03 0.21 0.48 0.73 0.47
Ma 10 0.12 0.09 0.86 0.26 0.22 0.20 0.03 0.17 0.59 0.78 0.45

 

Aff2 0.22 0.04 0.22 0.59 0.15 0.22 -0.25 0.14 0.14 0.22 0.24
Aff3 0.33 0.05 0.17 0.92 0.19 0.29 -O.18 0.11 0.21 0.17 0.14
Aff6 0.30 0.02 0.20 0.86 0.24 0.30 -0.17 0.14 0.22 0.23 0.15
Aff8 0.19 -0.02 0.16 0.47 0.36 0.27 -0.09 0.02 0.03 0.14 0.02

 

Pers 1 0.19 0.03 0.21 0.24 0.78 0.49 -0.18 -0.14 0.03 0.26 0.10
Pers 3 0.17 0.19 0.26 0.30 0.87 0.39 -0.19 0.03 0.13 0.31 0.14
Pers 6 0.15 0.13 0.16 0.26 0.70 0.38 -0.12 0.14 0.03 0.25 0.13
Pers 7 0.22 0.15 0.18 0.26 0.81. 0.37 -0.19 -0.03 0.10 0.29 0.04

 

Em 3 0.16 0.17 0.11 0.21 0.39 0.75 -0.04 0.07 0.10 0.15 0.15
Em 4 0.10 -0.11 0.17 0.35 0.43 0.75 0.02 -0.04 0.09 0.22 0.18
Em 7 0.12 0.05 0.17 0.26 0 29 0.64 0.01 0.03 0.00 0.17 0.05

 

Anx 1 -0.26 -0.27 0.14 -0.15 -0.08 0.07 0.80 0.05 0.21 0.18 0.14
Anx 2 -0.24 -0.10 0.04 -0.10 -0.08 0.00 0.68 0.02 0.17 0.09 0.05
Anx 3 '-0.39 -0.20 -0.11 -0.16 -0.12 -0.05 0.54 -0.14 0.00 -O.15 -0.20
Anx 4 -0.35 -0.25 -0.02 -0.24 -0.24 -0.06 0.53 -0.07 0.01 -0.16 -0.02
Anx 5 -0.30 -O.17 0.02 -O.12 -O.16 0.02 0.61 0.01 0.16 0.00 -0.02

46

Table 6 (cont’d)

 

Item Con Per Ma Aff Pers Em Anx Arg Inv Opi Act

Arg 1 0.21 0.42 0.09 0.21 0.00 0.03 -0.06 0.69 0.04 0.08 0.10
Arg 3 0.20 0.34 0.18 0.11 0.11 0.07 -0.07 0.76 0.13 0.20 0.22
Arg 4 0.20 0.55 0.12 0.10 0.03 0.00 -0.06 0.73 0.14 0.16 0.08
Arg 5 0.17 0.47 0.19 0.10 -0.04 0.01 -0.01 0.84 0.09 0.24 0.18
Arg 6 0.15 0.52 0.23 0.06 -0.03 0.00 -0.03 0.87 0.18 0.21 0.19
Arg 7 0.14 0.45 0.17 0.12 -0.06 0.02 0.03 0.87 0.16 0.25 0.20

 

Inv l 0.12 0.08 0.51 0.20 0.06 0.05 0.16 0.15 0.76 0.54 0.42
Inv 2 0.12 0.10 0.47 0.22 0.11 -0.01 0.16 0.14 0.88 0.57 0.35
Inv 3 0.15 0.10 0.57 0.17 0.10 0.05 0.12 0.14 0.93 0.66 0.46
Inv 4 0.12 0.12 0.51 0.12 0.09 0.14 0.19 0.17 0.92 0.59 0.44
Inv 5 0.15 0.05 0.55 0.19 0.04 0.15 0.11 0.06 0.80 0.62 0.41

 

Opi 4 0.20 0.10 0.66 0.18 0.27 0.17 -0.04 0.16 0.56 0.79 0.46
Opi 5 0.25 0.12 0.65 0.23 0.30 0.23 -0.04 0.17 0.60 0.91 0.52
3n 6 0.24 0.19 0.76 0.26 0.30 0.22 0.06 0.28 0.59 0.82 0.58

 

Act 3 0.06 -0.02 0.40 0.12 0.07 0.04 -0.03 0.09 0.40 0.46 0.75
Act 4 0.10 0.04 0.45 0.18 0.08 0.19 0.02 0.14 0.43 0.55 0.85
Act 5 0.20 0.10 0.43 0.14 0.18 0.21 -0.04 0.25 0.34 0.52 0.74
Act 6 0.15 0.00 0.49 0.19 0.11 0.15 -0.01 0.19 0.43 0.52 0.95

 

Factor

 

Con

Per 0.42

Mav 0.18 0.10

Aff 0.37 0.03 0.27

Persp 0.23 0.16 0.25 0.33

Emp 0.17 0.05 0.21 0.38 0.52

Anx -0.49 -0.31 0.02 -0.24 -0.22 -0.01

Arg 0.22 0.58 0.21 0.15 0.00 0.03 -0.04

Inv 0.15 0.10 0.61 0.21 0.09 0.09 0.17 0.15
Opin 0.27 0.16 0.82 0.27 0.35 0.25 -0.01 0.24 0.69
Act 0.15 0.04 0.54 0.19 0.13 0.18 -0.02 0.20 0.49 0.62

Notes. Con = Connector, Per = Persuader, Mav = Healthy Lifestyle Maven, Aff = Need for Aft-i-liation, Pers
= Perspective Taking, Emp = Empathic Concern, Anx = Social Anxiety, Arg = Argumentativeness, Inv =
Value-Relevant Involvement, Opin = Opinion Leadership, Act = Healthy Lifestyle Activism, n = 150

47

Because a ﬁtting measurement model was found, the mean response across each
measure’s items was computed to form 11 indices. The connector index ranged ﬁ'om
11.00 to 100.00 and was distributed approximately normally, M = 62.65, s = 21.34, a =
.90, S]. a = .90. Scores on the persuader index ranged from 14.00 to 100.00. The
persuader distribution had a slight negative skew, M = 68.04, s = 17.33, a = .94, and S].
a = .94. The distribution of the healthy lifestyle maven index, ranging from .67 to 100.00,
approached normality, M = 54.89, s = 24.03. The reliability, estimated by a = .93 and 8.1.
a = .93. The distribution of the need for affiliation index was negatively skewed and
leptokurtic (Range = 5.00 — 100.00), M = 84.19, s = 13.15, a = .78, S]. a = .80.
Responses on the perspective taking index, ranging from 11.15 to 100.00, were
approximately normally distributed, M = 66.76, s = 20.17, a = .86, S]. a = .87. The
empathic concern index was distributed with a Slight negative skew and ranged from
15.00 to 100.00. The indices mean response was 76.32 (s = 17.38), a = .76, SJ. (1 = .76.
Scores on the social anxiety index were distributed approximately normal between 0.00
and 100.00, M = 41.51, s = 21.95, a = .77, S]. a = .77. The argumentativeness index
ranged from 2.50 to 100.00 and was also distributed approximately normally, M = 59.42,
s = 23.70, a = .91, SJ. (1 = .91. Scores on the value-relevant involvement index ranged
from .00 to 100.00. The value-relevant involvement distribution approached normality, M
= 56.14, s = 25.02, a = .93, and S]. a = .93. Scores on the opinion leader index were
distributed approximately normal, ranging from 0.00 to 95.00. The opinion leader
distribution had a mean of 37.20 (s = 24.52). The reliability, estimated by a = .88 and 5.1.
a = .88. Finally, the healthy lifestyle activism index was distributed with a slight positive

skew (Range = 0.00 — 100.00), M = 28.69, s = 24.36, a = .88, S]. a = .89. These direct

48

interval-estimation data appear to be in the same range as those reported in the measures’

source materials. Table 7 presents a summary of the indices descriptive statistics.

Table 7

Study 2 Measure Descriptives

 

 

Measure M s Min - Max (1 S]. a Skewness Kurtosis
Connector 62.65 21.34 11.00 - 100.00 0.90 0.90 -0.38 -0.45
Persuader 68.04 17.33 14.00 - 100.00 0.94 0.94 -O.65 0.28
Maven 54.89 24.03 0.67 - 100.00 0.93 0.93 -0.24 -0.79
Affiliation 84.19 13.15 5.00 - 100.00 0.78 0.80 -1.94 8.27
Perspective 66.76 20.17 11.25 - 100.00 0.86 0.87 -0.37 -0.54
Empathic 76.32 17.38 15.00 - 100.00 0.76 0.76 -0.85 0.33
Anxiety 41.51 21.95 0.00 - 100.00 0.77 0.77 0.27 -0.35
Argue 59.42 23.70 2.50 - 100.00 0.91 0.91 -0.10 -0.69
Involvement 56.14 25.02 0.00 - 100.00 0.93 0.93 -0.40 -0.59
Opinion 37.21 24.52 0.00 - 95.00 0.88 0.88 0.31 -0.74
Activism 28.69 24.36 0.00 - 100.00 0.88 0.89 0.70 -0.33

 

Notes. Maven = Healthy Lifestyle Maven, Affiliation = Need for Afﬁliation, Perspective = Perspective
Taking, Empathic = Empathic Concern, Anxiety = Social Anxiety, Argue = Argumentativeness,
Involvement = Value-Relevant Involvement, Opinion = Opinion Leadership, Activism = Healthy Lifestyle
Activism, n = 150

Examination of the relationships between the demographic variables and the
connector, persuader, and healthy lifestyle maven measures demonstrated results similar
to those observed in Study 1. There were no substantial relationships between the
superdiffuser measures and the demographic variables, except for the relationships
between sex and the persuader measure. Men’s persuader scores were slightly greater
than women’s scores in Study 2. The mean women’s score equaled 65.54 whereas the
mean men’s score equaled 71.88 (r = .18, r’ = .19) which could be a result of the scaling
format sensitizing the measure due to the increased scaling range.

The construct validity results which are based on the factor correlations presented
as nomological networks in Figures 6 through 8 indicated that the observed networks

were consistent with the networks hypothesized, albeit with some errors. The connector

49

construct correlated substantially with social anxiety, r’ = -.49, P (-.64 _<_ p ’ S -.32) = .95,
and need for afﬁliation, r’ = .37, P (.19 S p ’ S .54) = .95, as hypothesized. Inconsistent
with the hypotheses, but not unexpected given the Study 1 regression analyses, the
connector construct failed to correlate substantially with healthy lifestyle activism, r =
.15, and had a modest, but statistically signiﬁcant, correlation with opinion leadership, r’
= .27, P (.10 S p ' _<_ .44) = .95. The connector construct also correlated modestly with
argumentativeness, r’ = .22, P (.06 S p ’ S .39) = .95, and perspective taking, r’ = .23, P
(.07 S p ’ S .41) = .95, which is inconsistent with the hypotheses but not entirely
unexpected. Finally, consistent with hypotheses, the connector construct did not correlate

substantially with value-relevant involvement, r’ = .15 , or empathic concern, r’ = .17.

 

 

Opinion
.37 .27 -.49
Argue '22 Connect '1 5
.23 . 15 . 1 7
Perspective Activism

Figure 6. Convergent and divergent validity nomological network for the connector
construct.

Notes. Connect = Connector, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve = Value-

Relevant Involvement, Afﬁliation = Need for Afﬁliation, Perspective = Perspective Taking, Empathy =
Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism, correlations
reported have been corrected for attenuation due to measurement error, n = 150

50

Although the observed persuader network was somewhat less consistent with the
hypothesized nomological network than the connector network, the results were still
informative. The persuader construct correlated with argumentativeness as hypothesized,
r’ = .58, P (.44 Sp ’ S .70) = .95. The same was not true for perspective taking, r’ = .16,
indicating that the ability to adopt another’s psychological point of view is not related to
persuasiveness. Much like the connector construct, the persuader construct also failed to
correlate considerably with opinion leadership, r’ = .16, and healthy lifestyle activism, r’

= .04. This failure, however, is consistent with the regression ﬁndings of Study 1.

 

 

Opinion
.03 .16 -.31
Argue '58 Persuade '10
.16 .04 .05

Perspective Activism

Figure 7. Convergent and divergent validity nomological network for the persuader
construct. ,

Notes. Persude = Persuader, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve = Value-
Relevant Involvement, Affiliation = Need for Afﬁliation, Perspective = Perspective Taking, Empathy =
Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism, correlations
reported have been corrected for attenuation due to measurement error, 11 = 150

Consistent with the network hypotheses, the persuader construct failed to

correlate substantially with need for afﬁliation, r’ = .03, value—relevant involvement, r’ =

51

.10, or empathic concern, r’ = .05. On the other hand, the persuader construct did
correlate modestly with social anxiety, r’ = -.31, P (-.48 S p ’ 5 -.l3) = .95. Although this
convergent correlation was hypothesized as a divergent relationship, its presence
replicates the same relationship observed in Boster et al. (2006) and is consistent with the
assertion that the expression of the persuader construct involves social interaction which
would be attenuated by traits like social anxiety or communication apprehension
(McCroskey, Beatty, Kearney, & Plax, 1985).

The healthy lifestyle maven network was, in the main, consistent with the
hypothesized nomological network. The healthy lifestyle maven construct correlated
substantially with value-relevant involvement, r’ = .61, P (.49 S p’ S .73) = .95. The
healthy lifestyle maven construct only correlated modestly, however, with empathic
concern, r’ = .21 , P (.04 S p’ S .39) = .95, indicating that the tendency to experience
feelings of concern for others in need may not be related strongly to the maven construct.
Consistent with the regression analyses in Study 1, the healthy lifestyle maven construct
correlated considerably with opinion leadership, r’ = .82, P (.74 Sp’ S .90) = .95, and
healthy lifestyle activism, r’ = .54, P (.41 Sp’ S .67) = .95. Finally, the healthy lifestyle
maven construct correlated within sampling error of zero with its hypothesized divergent
construct of social anxiety, r’ = .02, and modestly with its hypothesized divergent
constructs of need for afﬁliation, r’ = .27, P (.09 S p ’ S .44) = .95, argumentativeness, r’

= .21, P (.04 Sp’ S .37) = .95, and perspective taking, r’ = .25, P (.09 Sp’ S .42) = .95.

52

Afﬁliation Opinion

 

 

.27 .82 .02
Argue '21 Maven '61
.25 .54 .21

Perspective Activism

Figure 8. Convergent and divergent validity nomological network for the healthy lifestyle
maven construct.

Notes. Maven = Healthy Lifestyle Maven, Anxiety = Social Anxiety, Argue = Argumentativeness, Involve
= Value—Relevant Involvement, Aﬁiliation = Need for Affiliation, Perspective = Perspective Taking,
Empathy = Empathic Concern, Opinion = Opinion Leadership, Activism = Healthy Lifestyle Activism,
correlations reported have been corrected for attenuation due to measurement error, it = 150

As was the case in Study 1, the factor correlations in the nomological network
suggest that the connector and persuader constructs may contribute little more to the
ability to predict opinion leadership and healthy lifestyle activism than what can be
accounted for by the healthy lifestyle maven construct. Therefore, multiple regression
analyses were again employed, regressing opinion leadership onto the connector,
persuader, and healthy lifestyle maven measures as well as all two-way and three-way
interactions. The standardized regression coefficient for the connector measure was ,6 =
.08 (t (142) = 1.24, ns, ,6 ’ = .08), for the persuader measure was ,6 = .05 (t (142) = 0.88,
ns, ,8 ’ = .025), and for the healthy lifestyle maven measure was ﬂ = .67 (t (142) = 11.07, p

< .05, ,8’ = .75). When healthy lifestyle activism scores were regressed onto the

53

connector, persuader, and healthy lifestyle maven indices, as well as all two-way and

three-way interaction terms, a similar effect was observed. Again, the healthy lifestyle

maven measure emerged as the only major predictor (ﬂ = .47, t (142) = 5.84, p < .05, 13’:

.53). Table 8 presents the regression models.

 

 

 

Table 8
Study 2 Regression Models
Model Component B se ,6 t sig B ’
Opinion Leadership Constant 37.04 1.42 26.18 p < .05
F (7, 142) = 28.38 C 0.09 0.07 0.08 1.24 ns 0.08
p < .05 P 0.08 0.09 0.05 0.88 ns 0.05
Rad,- = .75 M 0.68 0.06 0.67 11.07 p < .05 0.75
C x P 0.00 0.00 -0.01 -0.14 ns -0.03
C x M 0.00 0.00 -0.06 ~0.93 ns -0.04
P x M 0.00 0.00 -0.01 -0.09 ns -0.02
C x P x M 0.00 0.00 0.16 2.55 p < .05 0.15
Activism Constant 28.72 1.88 15.27 p < .05
F (7, 142) = 6.89 C 0.07 0.09 0.06 0.79 nS 0.07
p < .05 P -0.05 0.11 -0.04 -0.46 ns -0.06
Rad,- = .47 M 0.48 0.08 0.47 5.84 p < .05 0.53
C x P 0.00 0.00 0.00 0.04 ns -0.01
C x M 0.00 0.00 -0.01 -0.14 nS 0.01
P x M 0.00 0.01 -0.09 -1.07 ns -0.10
C x P x M 0.00 0.00 0.07 0.82 ns 0.05

 

Notes. C = Connector, P = Persuader, M = Healthy Lifestyle Maven, components are mean centered to
reduce multicollinearity, n = 150

Finally, quantifying superdiffusers as people responding at least one standard

deviation above the mean on the connector, persuader, and healthy lifestyle maven

measures revealed that 4% of this study’s sample was superdiffusers. The deviation in the

percentage of the sample classiﬁed as superdiffusers from Study 1 to Study 2 likely

results from a sensitization of the measurement due to the increased range of the direct

54

interval-estimation scaling. A similar percentage increase was observed in Boster et al.
(2006) when that study changed the item scaling from a 5 point Likert response scale to
an 8 point Likert response scale.

Discussion

Results from Study 2 provided more evidence for the content validity of the
connector, persuader, and maven measures. The CFA results demonstrated more evidence
that the three measures are internally consistent and parallel, not just amongst each other
but across measures of other constructs as well. Furthermore, the multiple groups CFA
demonstrated that the method of scaling did not alter the factor structure of the measures.
In all, the factor structures of the items composing the connector, persuader, and healthy
lifestyle maven measures appear stable across multiple independent samples and scaling
formats.

The scaling format did, however, produce a modest change in the reliability
estimates of the measures. Speciﬁcally, direct interval-estimation increased reliability
estimates relative to eight point categorical Likert response scaling. Although the effect
was fairly consistent across measures there was one condition where the effect was not
observed. Direct interval-estimation failed to increase the reliability of measures
containing several reverse coded items. Based on a cursory review of responses, the
reason for this effect seems to result from respondent difﬁculty in correctly inverting
their response on the reverse coded items. For example, a response of 80 on a standard
coded item dictates the response of 20 on a reverse coded item of equal quality taken
from the same content domain. Respondents in this sample, however, quite consistently

offered imperfect reﬂections of their standard responses. Levine (1994) observed a

55

similar effect in his study of direct magnitude estimation where respondents had
difﬁculty responding when they fell below the reference stimulus. For measures
containing no reverse coded items, however, direct interval-estimation bolstered the
reliability estimates quite consistently. '

Interestingly, the change in the scaling format also appears to have sensitized the
measure’s ability to detect superdiffusers. Across three previous samples, 1.7% of
subjects have been classiﬁed as superdiffusers according to their responses on the
connector, persuader, and healthy lifestyle maven measures. This study’s categorical
Likert scaling condition closely approximated that percentage with 1.3% classiﬁed as
superdiffusers. The direct interval-estimation condition, however, classiﬁed 4% of the
sample as superdiffusers employing the same criteria. The sensitization results from
direct interval-estimation normalizing the response distributions of each measure, thus
lowering the mean and increasing the number of people above one standard deviation
from the mean. Consequently, in addition to increasing the reliability estimates of the
connector, persuader, and healthy lifestyle maven measures, direct interval-estimation
will make it easier to identify potential superdiffusers in the population.

The nomological network analysis in Study 2 provided additional construct
validity evidence for the connector, persuader, and healthy lifestyle maven measures. The
hypothesized convergent correlations were ample and the hypothesized divergent
correlations were small, albeit with a few exceptions. For example, the convergent
correlations between the persuader construct and perspective taking as well as between
the healthy lifestyle maven construct and empathic concern were modest. This failure

could indicate that the two variables are not conceptually related, it could also indicate

56

that at least one of the measures is not indicative of what it purports to measure. As
another example, the connector construct correlated substantially with argumentativeness
and perspective taking, which were hypothesized as divergent correlations. This failure
could also be the result of a measurement validity issue or the presence of an unmeasured
construct sharing a relationship among the three constructs.

These network failures highlight one of the limitations of the nomological
network approach or multi-trait multi-approach (Campbell & Fiske, 1959) to construct
validity assessment. There is no good method of determining how much network ﬁt is
necessary to conclude that a measure or set of measures exhibit construct validity. That
being said there were no severe misﬁts in this network. In other words, there were no
misﬁts that could not be reasonably explained by revisiting theory or of such a magnitude
that they could not potentially be attributed to sampling error. Therefore, there is
substantial evidence of construct validity for the connector, persuader, or maven
measures.

Finally, the regression analyses conducted in Study 2 replicated closely the
ﬁndings of Study 1, again indicating that what others have called opinion leadership and
healthy lifestyle activism is, in the main, the extent to which one is a healthy lifestyle
maven and that alone. Furthermore, implementation of direct interval-estimation did not
change the ﬁnding that the connector and persuader constructs contribute little to the
relationships between super diffusion and opinion leadership or super diffusion and
activism. This ﬁnding is an interesting effect which deserves attention in its own right

and suggests that a superdiffuser is something other than an opinion leader or activist.

57

Thus far, evidence for the content and construct validity of the superdiffuser
measures is ample. The extent to which self-reports on the measures correspond to actual
behavior has not yet been the focus of examination, however. Therefore, Study 3 will
undertake a predictive validity assessment by testing the extent to which behaviors
indicative of each of the three characteristics correspond to self reported responses on the
connector, persuader, and healthy lifestyle maven measures. It is hypothesized that there
will be substantial positive correlations between: 1) connector scores and the number of
acquaintances people know, 2) persuader scores and a people’s argumentative
complexity, and 3) healthy lifestyle maven scores and scores on a healthy lifestyle

knowledge test.

58

STUDY 3

Method

Subjects. The third study sampled 35 female students who were at least 18 years
old, enrolled in undergraduate Communication classes at a large Midwestern university,
and had not participated in Studies 1 or 2. These students received compensation for their
participation in the form of course credit. Subjects were on average 19.66 (s = 1.63) years
old and in the 2.31 (s = 1.11) year of college. Finally, 23% of subjects reported holding a
leadership position since beginning college.

Design. Study 3 was designed as a predictive validity test for the connector,
persuader, and healthy lifestyle maven measures. Consequently, scores on each of the
connector, persuader, and healthy lifestyle maven measures were used to predict scores
on a corresponding criterion measure. Connector scores were expected to predict the
respondent’s number of acquaintances as measured by a phonebook test, persuader scores
were expected to predict argumentative complexity, and healthy lifestyle maven scores
were expected to predict healthy lifestyle knowledge if the superdiffuser measures did
indeed possess predictive validity.

Procedures. Subjects volunteering to participate in a study described as an
investigation of communication and social behavior arrived individually at the
researcher’s laboratory for a scheduled session lasting approximately 45 minutes per
subject. Subjects were greeted by the researcher and seated at a desk. The researcher
informed the subject that participation involved the completion of two main tasks. The
ﬁrst task involved the completion of a questionnaire containing several measures of

social opinion and social behavior, which took approximately 15 minutes to complete.

59

The second task was a structured audio taped interview of the subject focusing on the
subject’s position on a healthy lifestyle topic, which took approximately 30 minutes to
complete.

After informing the subject of the tasks, the questionnaire was administered and a
brief set of instructions regarding how to complete the items on the questionnaire was
reviewed before allowing the subjects to begin. After the questionnaire was completed
and returned to the researcher, the tape recorder was started and the subjects were
interviewed on their position towards dieting according to the protocol in Appendix 0.
After completing both tasks, subjects were dismissed.

Instrumentation. In addition to several demographic items and the direct interval-
estimation scaled connector, persuader, and healthy lifestyle maven measures, the
questionnaire in Study 3 contained measures of two predictive validity criteria. The
connector criterion took the form of a phonebook test initially proposed by Gladwell
(2002) (Appendix P). By deﬁnition, the extent to which someone is a connector is in part
how many acquaintances they have. To test this premise using the phonebook test, four of
the most common last names beginning with each letter in the alphabet were selected
from a campus directory resulting in a list of 104 common last names. Subjects were
instructed to indicate every time a last name on the list was shared with an acquaintance
of theirs. AS the number of acquaintances a person had increased, the probability that the
last names of their acquaintances would appear on the list also increased. Therefore,
given that the phonebook test was an indicator of how many acquaintances a person had,

it also served as a predictive validity test for the connector measure.

60

The questionnaire also contained the healthy lifestyle maven predictive validity
criterion. Mavens, by deﬁnition, develop expert knowledge in a content domain.
Therefore, self-reported maven levels must be predictive of knowledge in the content
domain of interest. To test this assertion, a healthy lifestyle knowledge test (Appendix Q)
was developed from two modiﬁed health knowledge tests (Centers for Disease Control
and Prevention, 2005; Centers for Disease Control and Prevention, 2006). This test
contained items reﬂecting health knowledge concerns such as exercise, nutrition,
reproductive health (including sexually transmitted infections), heart disease, and cancer.
Several of these content areas (exercise, nutrition, and reproductive health) were
identiﬁed as major healthy lifestyle concerns in a pretest drawn from the same population
as this study’s sample. Each item on the healthy lifestyle knowledge test was scored as
correct or incorrect. The greater the number of correct to items was indicative of greater
health knowledge. Consequently, the extent to which the healthy lifestyle maven
construct correlated highly with healthy lifestyle knowledge test scores was indicative of
the predictive validity of the maven measure.

Interview. The persuader criterion of argumentative complexity was measured
through the subject’s interview responses. Subjects were interviewed by the researcher
for their position on the issue of dieting. Dieting was chosen because a pretest sample
indicated that dieting was a substantial health concern of the female population that was
sampled. Consequently, subjects would be likely to have existing attitudes towards
dieting. These extant attitudes were important because interview questions asked the
subjects for detailed reports of reasons for holding their position. Furthermore, given that

the interview asked subjects to report counterarguments against their position, it was also

61

important that an issue was chosen where arguments and evidence both for and against
the issue exist.

The structure of the interview was an adaptation of Kuhn’s (1991) protocol which
elicited responses indicative of several components of argumentative complexity. Kuhn
(1991) identiﬁed several components of argumentative complexity through the content
analysis of 160 interviews taken from people across four age groups (teens, 205, 408, 603)
which asked respondents to argue for their position on social problems such as crime,
failure in school, and unemployment. In addition to identifying several concepts related
to the respondents’ cognitive complexity, the analyses revealed six components particular
to argumentative complexity: causal theories, evidence, alternative causal theories,
alternative evidence, counterargument, and rebuttal.

Causal theories are chains of premises that proceed logically to the respondent’s
position on an issue. The quality of a causal theory, according to Kuhn (1991) is based on
the length of the causal theory judged by the number of premises in the causal theory and
the amount of integration between parallel causal theories. For example, it is possible that
a respondent could have multiple chains of premises leading to the same conclusion. The
extent to which the multiple chains share premises is equivalent to the amount of
integration. The more premises in the chain and the greater the amount of integration
between chains was found by Kuhn (1991) and Campo (1999) to be indicative of greater
argumentative complexity. Alternative causal theories can be described in a similar
fashion except that the respondent’s alternative causal theories argue for a position on an

issue opposing their own.

62

The third component of argumentative complexity is the evidence respondents
provided to justify their position on an issue. Kuhn (1991) categorized evidence based on
strength into three forms: genuine evidence, pseudoevidence, and nonevidence. Genuine
evidence, being the strongest of the three forms, is true evidence which can stand alone
from the respondent’s causal theory. For example, a respondent’s statistical claim of
correlation between a cause and effect would be classiﬁed as genuine evidence.
Pseudoevidence is a scenario or script illustrating how an issue might occur. A
respondent’s anecdote for why an outcome occurred would be classiﬁed as
pseudoevidence. Finally, the weakest form of evidence according to Kuhn (1991) is
nonevidence. A respondent provides nonevidence when it is implied that evidence is not
necessary for the position taken or the position itself is claimed as evidence in a circular
manner. The greater the amount of genuine evidence and pseudoevidence to nonevidence
given by the respondent is indicative of greater argumentative complexity according to
Kuhn (1991). Alternative evidence is deﬁned by the same categories only the evidence is
provided as justiﬁcation for a position on an issue opposing one’s own.

The ﬁnal two components identiﬁed by Kuhn (1991) are counterarguments and
rebuttals. Counterargument is the respondent’s ability to generate conditions that falsify
the reasons for why they hold their position on an issue. The strength of a respondent’s
counterarguments is the extent to which the counterarguments provided falsify the
respondent’s own reasons for their position on an issue. Rebuttals on the other hand are
the respondent’s attempt to rebuff counterarguments and alternative causal theories

opposing their position. To the extent that the rebuttal argues against opposing positions

63

by integrating the respondent’s causal theory, evidence, and counterargument the rebuttal
is said to be strong.

Ku1m(1991) argues that these six components are unidimensional. Therefore, it is
hypothesized that as the degree to which these six components were present in the
subject’s interview responses increased, so did the argumentative complexity of the
response. Argumentative complexity is thought to be related to persuasiveness because
research has found that people presenting more complex arguments are not only more
persuasive, but also perceived by others as more competent persuaders (Campo, 1999;
Kuhn, 1991; Shestowsky, Wegner, and Fabrigar, 1998). Therefore, it was expected that
persuasiveness scores would correlate positively with the argumentative complexity of
the subject’s interview responses.

Interview Coding. The audio tapes from all of the interviews were transcribed by
the researcher. Following transcription, three interview coders who were blind to the
study’s hypotheses were trained. Training occurred over several steps. First, over several
sessions coders read and discussed the parts of Kuhn’s (1991) text relevant to deﬁning
the six components of argumentative complexity. Once deﬁnitional agreement among the
researcher and coders was achieved, focus turned to developing the coders’ ability to
identify the components of argumentative complexity in the interview transcripts with a
high degree of intercoder reliability. Interviews were coded employing the code sheet
presented in Appendix R. Referring to the deﬁnitions of each component provided by
Kuhn (1991), coders Were instructed to code each interview for the components of

argumentative complexity using eight point Likert response scales.

64

Several interview transcripts not included in the study’s sample were used for
training. Initially, training interview transcripts were coded collectively and coders, along
with the researcher, discussed any coding discrepancies until agreement was reached.
Then, working independently, the remaining training transcripts were coded. When
discrepancies were found between coders, a discussion was held where coders offered
their perspective until agreement was reached. Substantial intercoder reliability for each
of the components in the training interview transcripts was achieved (Ebel ’s Coeﬁ‘icient
(E.C.)3 > .85) and coders began coding the Study 3 interviews.

Results

Measurement Models. The three factor connector, persuader, and healthy lifestyle
maven measurement model was tested with Hunter and Hamilton’s (1992) method of
CFA. This analysis revealed that there were no items in the model exhibiting misﬁt and it
was concluded that the data ﬁt the model well. Speciﬁcally, all factor loadings were
above .62, only two residuals were not within sampling error of zero (eight would be
expected due to chance when p < .05), and RMSE = .12, well within sampling error of
zero. Table 9 presents the factor loadings and factor correlations.

Table 9

Study 3 Factor Loadings and Factor Correlation Matrix

 

Item Connector Persuader Maven

Connector 1 0.21 -0.09
Connector 2 0.48 0.01
Connector 3 0.19 0.08
Connector 4 , 0.29 -0. 12
Connector 5 0.31 0.03

 

 

 

3 BC. is a measure of inter-coder reliability with a range of 0 to l and is equivalent to S]. a.

65

Table 9 (cont’d)

 

 

 

 

 

Item Connector Persuader Maven
Persuaderl 0.26 - . i ‘ ‘:.0_.72 -0.06
Persuader 2 0.17 . 0.62 0.00
Persuader3 0.21 ' . 031 -0.05
Persuader4 0.35 ' . .. 0.89 0.01
PersuaderS 0.37 330.379 0.06
Persuader6 0.48 , _ : ”3.0;.79 0.17
Persuader7 0.08 V V ‘ . . 0.74 0.34
Persuader 8 0.25 ' 1' i ’1 150.72 _ 0.18
Mavenl -0.06 0.19. 3(tits
Maven2 —0.17 -0.07‘ ' 073
Maven3 -0.09 0.10.,y ., _ﬁ, 0.89
Maven4 -0.04 -0.02‘ i .7 - 0:74
Maven5 -0.08 0.15' 3: . 150.86
Maven6 0.32 0.17 . 7 0.71

Factor
Connector
Persuader 0.36
Maven -0.03 0.1 1

 

Notes. Maven = Healthy Lifestyle Maven1 n = 35

Forming connector, persuader, and healthy lifestyle maven indices by taking the
mean response across all items on each factor produced measures with descriptive
statistics comparable to those observed in Study 2. Connector scores were distributed
approximately normally, ranging from 8.60 to 84.00 with a mean of 46.09 (s = 21.84).
The reliability of this index estimated by a = .91 and S]. a = .91. The persuader index
ranged from 33.13 to 87.50. The distribution of persuader scores also approximated
normality, M= 63.21, s = 13.82, a = .91, 5.1. a = .91. The distribution of the healthy

lifestyle maven index, ranging from 19.17 to 91.67, approached normality. The mean of

66

the healthy lifestyle maven index equaled 50.33 (s = 22.10). The a reliability of the
measure equaled .91 and its S]. a reliability equaled .91.

Phonebook test scores were calculated by summing the number of afﬁrmative
responses on the test items to obtain the total names shared with the respondents’
acquaintances. These scores were approximately normally distributed. The phonebook
test scores ranged from 9 through 70 and the distribution had a mean of 3 1 .46 (s = 14.21).

The intercoder reliabilities of the six argumentative complexity components were
examined before forming the second-order argumentative complexity index. Each coder
was treated as one item on a three item measure. Thus, for each of the six components of
causal theory, evidence, alternative causal theory, alternative evidence, counterargument,
and rebuttal there were three items measuring the extent to which the construct was
present in the subjects’ interviews. Six indices were created by taking the mean of the
combination of coders’ with the greatest intercoder agreement on each component.

Scores on the causal theory index (coders: 1, 2, and 3) were distributed
approximately normal, ranged from 2.00 to 6.67, and had a mean of 3.99 (s = 1.14). The
intercoder reliability of the index was estimated at .89 using E .C.. The evidence index
(coders: 1, 2, and 3) ranged from 2.33 to 6.33 and was also distributed approximately
normal. The mean of this index was 4.17 with a standard deviation of 1.19. The
intercoder reliability, estimated by E. C. = .84. The distribution of the alternative causal
theory index (coders: 2 and 3), ranging from 1.00 to 6.00, approximated normality as
well, M = 3.90, s = 1.16, E.C. = .80. The distribution of scores approached normality on
the alternative evidence index (coders: 1, 2, and 3). The alternative evidence index ranged

from 1.00 to 4.67, had a mean of 2.90 (s = .87), and EC. = .71. The counterargument

67

index (coders: 1 and 2) was approximately normally distributed and ranged from 1.00 to
5.00. The mean response of the index was 2.46 (s = 1.01), E.C. = .83. The distribution of
the rebuttal index (coders: 1 and 2) approached normality (Range = 1.50 — 7.50), M =
4.07, s = 1.48, E.C. = .80.

CFA was not employed to test the ﬁt of the ﬁrst-order, Six factor measurement
model because the internal consistency matrices of each factor were underidentiﬁed.
Based on visual examination of the correlation matrices, however, each index was likely
ﬁrst-order unidimensional. Therefore, the internal consistency of the second-order
unidimensional measurement model, positing that the six ﬁrst-order factors are functions
of argumentative complexity was tested using CFA (Hunter & Hamilton, 1992). Initial
examination of the second-order correlation matrix revealed that both evidence and
alternative evidence did not exhibit internal consistency with the other components.
Therefore, evidence and alternative evidence were discarded from the model.

Subsequent to removing the two misﬁtting components, CFA was employed to
test the second-order unidimensionality of the causal theory, alternative causal theory,
counterargument, and rebuttal components. The analysis revealed that all factor loadings
were above .48, all residuals were within sampling error of zero, and RMSE = .11, well
within sampling error of zero. Table 10 presents the factor loadings. Given the ﬁt of the
second-order model, the mean score across each of these four components was computed
to form an argumentative complexity index. The argumentative complexity index, which
was approximately normally distributed, ranged from 1.50 to 5.04. The mean of the index
equaled 3.60 (s = .89), S]. a = .73. Appendix S presents three examples of interviews

coded as low, moderate, and high argumentative complexity.

68

Table 10

Study 3 Argumentative Complexity Second-Order Factor Loadings

 

Component Arg Complexity
Causal Theory 0.65
Alternative Causal Theory 0.48
Counterargument 0.66
Rebuttal 0.77

 

Notes. Arg = Argumentative, n = 35

Responses on the healthy lifestyle knowledge test were examined to assess the
extent to which they ﬁt a unidimensional measurement model using Hunter and
Hamilton’s (1992) method of CFA. There were several items in the model exhibiting
misﬁt which were removed from subsequent analysis. CFA conducted on the remaining
items produced a model that ﬁt the data well. All factor loadings were above .33, all
residuals were within sampling error of zero, and RMSE = .08, well within sampling error
of zero. Table 11 presents the factor loadings.

Table 11

Study 3 Healthy Lifestyle Knowledge Factor Loadings

 

Item Health
Health 2 0.53
Health 6 0.46
Health 11 0.37
Health 12 0.33
Health 14 0.57
Health 15 0.49
Health 17 0.36

 

Notes. Health = Healthy Lifestyle Knowledge, n = 35

The retained items (healthy lifestyle knowledge: 2, 6, 11, 12, 14, 15, and 17) were

formed into an index by summing the number of items answered correctly. The resultant

69

distribution of the healthy lifestyle knowledge index was distributed approximately
normal with a mean score of 4.40 (s = 1.85). The reliability of the healthy lifestyle
knowledge index equaled .63 estimated by a and SJ. (1. Table 12 presents a summary of
the indices descriptive statistics.

Table 12

Stuay 3 Measure Descriptives

 

 

Measure M s Min - Max (1 SI. (1 Skewness Kurtosis
Connector 46.09 21.84 8.00 - 84.00 0.91 0.91 0.18 -094
Persuader 63.21 13.82 33.13 - 87.50 0.91 0.91 0.01 -0.69
Maven 50.33 22.10 19.17 - 91.67 0.91 0.91 0.20 .117
Phone 31.46 14.21 9.00 - 70.00 ' ' 0.74 0.43
Argue 3.60 0.89 1.50 - 5.04 0.732 0.732 -0.39 -049
Health 4.40 1.85 1.00 - 7.00 0.63 0.63 -0.31 -091

 

Notes. Maven = Healthy Lifestyle Maven, Phone = Phonebook Test, Argue = Argumentative Complexity,
Health = Healthy Lifestyle Knowledge Test, ' computation of a reliability coefficient was not possible for
this measure, 2 intercoder reliability estimated with Ebel’s Coefficient, n = 35

Predictive Validity. The bivariate scatterplots of each superdiffuser construct and
its corresponding predictive validity test are presented in Figure 9. From a cursory
observation it is apparent that the data in all three scatterplots demonstrate positive
trends. The scatterplots, however, also suggest that there are several data points carrying
substantial inﬂuence. Because of the small sample size in Study 3 the correlation
coefficient including these points would reﬂect the inﬂuence of the atypical case rather
than the general trend in the data. Given that the goal of this study is to provide an
accurate image of the conceptual relationships between the connector, persuader, and
healthy lifestyle maven constructs and their predictive validity criteria, extremely

inﬂuential cases were removed from subsequent analyses.

70

Argumentative Complexity Phonebook Test

Healthy Lifestyle Knowledge Test

 

70-

60-

50-

 

 

Connector

 

80 100

 

31

2:1

 

1

 

 

3O

40 50 60 7O

Persuader

 

 

 

 

Maven

20 40 80

Figure 9. Predictive validity scatterplots.

Notes. Maven = Healthy Lifestyle Maven, n = 35

71

80 100

Cook’s D was employed as a measure of inﬂuence. Three cases in the bivariate
connector by phonebook test distribution (75.00, 70.00), (77.00, 10.00), and (81.00,
14.00) were identiﬁed as highly inﬂuential. Each of these three cases had values of
Cook’s D greater than 3.5 standard deviations above the mean. One case in the bivariate
persuader by argumentative complexity distribution (73.13, 1.50) was identiﬁed as highly
inﬂuential. The value of Cook’s D for this case was greater than 4 standard deviations
above the mean. Finally, one case in the bivariate healthy lifestyle maven by healthy
lifestyle knowledge test distribution (91 .67, 2.00) was considered inﬂuential. Cook’s D
for this case was 5 standard deviations above the mean. In each of these cases there was a
substantial gap before reaching the case with the next lowest Cook’s D value.

The predictive validity factor correlations are presented in Table 13.4 The matrix
below the diagonal contains correlations excluding inﬂuential cases and the portion
above the diagonal contains correlations including inﬂuential cases.

Table 13

Study 3 Predictive Validity Correlations

 

Connector Persuader Maven Phone Argue Health

Connector 0.37 -0.02 0.43 0.45 0.38
Persuader 0.18 0.11 0.08 0.23 0.00
Maven -0.08 0.1 1 0.31 0.38 0.04
Phone 0.55 0.16 0.04 0.36 0.11
Argue 0.30 0.29 0.37 0.12 0.07
Health 0.48 0.01 0.24 0.30 0.16

 

Notes. Maven = Healthy Lifestyle Maven, Phone = Phonebook Test, Argue = Argumentative Complexity,
Health = Healthy Lifestyle Knowledge Test, matrix contains correlations corrected for attenuation due to
measurement error, 11 = 30 under the diagonal, n = 35 above the diagonal

 

’ It is acknowledged that all ﬁve cases identiﬁed as inﬂuential are unlikely to truly be outliers. Given the
particularly small sample size and the substantial likelihood of sampling error, however, it was decided to
present the data in Table 13 inclusive and exclusive of the cases identiﬁed as inﬂuential.

72

The predictive validity evidence for the connector measure was convincing. The
connector measure correlated most strongly with the phonebook test and the phonebook
test correlated most strongly with the connector measure, r’ = .55, P (.26 S p’ S .84) =
.95. The connector measure also correlated amply, albeit within sampling error of zero,
with argumentative complexity, r’ = .30, P (-.15 S p’ S .75) = .95. This relationship
would be expected given that Kuhn (1991) and Campo (1999) both provided evidence
indicating that argumentative complexity was in part skill; indeed, a Skill that would be
more easily developed in people who were more likely to interact with others.

The predictive validity evidence for the persuader and healthy lifestyle maven
measures were not as clear. Although argumentative complexity did not correlate most
strongly with the persuader measure, the persuader measure correlated most strongly with
argumentative complexity, r’ = .29, P (-.15 Sp’ S .74) = .95. Therefore, there was some
evidence in these data for the predictive validity of the persuader measure. The healthy
lifestyle maven measure actually correlated most strongly with argumentative
complexity, r ’ = .37, P (-.05 S p’ S .79) = .95. Both of these correlations were, however,
within sampling error of zero. Given that the argumentative complexity interview
involved the health topic of dieting, however, one might have expected healthy lifestyle
mavens to demonstrate greater argumentative complexity on this topic than non-mavens.
Additionally, even though the healthy lifestyle maven measure’s correlation with the
healthy lifestyle knowledge test was not particularly ample, r’ = .24, P (-.15 S p’ S .78) =
.95 , it was greater than the persuader by healthy lifestyle knowledge correlation, r’ = .01.
Consequently there was some evidence in the data for the predictive validity of the

maven measure. This evidence was qualiﬁed by the fact that the healthy lifestyle

73

knowledge test may have been somewhat problematic given its weak psychometric
properties and the substantial correlation between it and the connector measure, r’ = .48,
P (.05 Sp’ S .90) = .95.

Finally, as in studies 1 and 2, quantifying superdiffusers as people responding at
least one standard deviation above the mean on the connector, persuader, and maven
measures revealed that 13% of this study’s sample was superdiffusers. Although this
percentage was a substantial increase from previous studies, given the small sample Size
in Study 3 the stability of this statistic along with the other results must be assessed
cautiously.

Discussion

Study 3 provided evidence both consistent and inconsistent with the position that
scores on the connector, persuader, and healthy lifestyle maven measures have some
power to predict actual behaviors associated with the psychological characteristics that
each measure assesses. Speciﬁcally, scores on the connector measure predicted the
number of acquaintances the respondent reported having. To a lesser extent,scores on the
persuader measure predicted the respondent’s argumentative complexity and scores on
the healthy lifestyle maven measure predicted the respondent’s healthy lifestyle
knowledge. In addition to these relationships, other relationships were observed which
were not predicted but not entirely unexpected. For example, the relationship between
argumentative complexity and the connector measure as well as the relationship between
argumentative complexity and the healthy lifestyle maven measure were two of these

relationships.

74

Consequently, the Study 3 ﬁndings must be interpreted with caution. The small
sample size limits the conﬁdence one can have in the stability of these ﬁndings. In
addition to the sample limitations, there was also a measurement limitation. The healthy
lifestyle maven measure in particular needs to be subjected to a stronger test of predictive
validity. The psychometric evidence for the healthy lifestyle knowledge test (a = .65) was
the weakest of the predictive validity criterion. This weakness made it particularly
difﬁcult to interpret the predictive validity correlation matrix concerning the healthy
lifestyle knowledge test.

These limitations aside, considering the fact that connectedness is much more
than just how many people you know (the location of the people within the larger social
network is also important), the persuader construct is much more than one’s ability to
generate complex arguments (the ability of the persuader to adapt their argument to the
target is also important), and the healthy lifestyle maven construct is much more than
general content knowledge (the ability to ﬁnd and interpret the knowledge is central to
the construct) the observed predictive validity correlations were acceptable (connector by
phonebook test = .55, persuader by argumentative complexity .29, and healthy lifestyle
maven by healthy lifestyle knowledge test = .24).

Given this mixed evidence, Study 3 could be improved by identifying better
measures of the behavioral criteria and/or considering other predictive validity criteria
that tap into features of the connector, persuader, and healthy lifestyle maven constructs.
For example, the connector evidence could be improved by comparing the results of a
network analysis against the responses on the connector measure. The persuader evidence

could be improved by obtaining argumentative complexity scores across several different

75

topics instead of j ust dieting. The healthy lifestyle maven evidence could be improved by
ensuring better correspondence between the breadth of the knowledge measure and the
breadth of the healthy lifestyle maven content domain as well as improving the measure
of healthy lifestyle knowledge, possibly within the domain of health literacy (Rothschild,
2005). Finally, there is the critique that respondents may not be able to self-report
accurately on their levels of the superdiffuser characteristics. Therefore, a comparison of
self-report scores obtained with these measures against other-report scores could also

provide useful predictive validity evidence.

76

GENERAL DISCUSSION

These three studies demonstrate evidence consistent with the validity and
reliability of three individual difference measures of interpersonal persuasive efﬁcacy.
Connectors easily form acquaintances with others and consequently serve as weak-tie
bridges with members of other groups. Persuaders are target adaptive debaters without
being aggressive or overbearing and enjoy attempting to inﬂuence others to take positions
they hold. Mavens are subject matter experts and are recognized as such, being sought
out for information from others in their social network. People possessing substantial
amounts of all three of these characteristics are classiﬁed as superdiffusers. Data across
these studies indicated that an average of 3.6% of the respondents reported being well
connected, convincing, healthy lifestyle expert superdiffusers. The ability to identify
superdiffusers would be a particular boon to persuasive campaigns. Given the unique
characteristics of the superdiffuser, it is likely that their inﬂuence spreads virally
throughout the interpersonal social network. Therefore, not only are superdiffusers able
to spread information quickly through a social network, they are also able to do so
effectively because of the interpersonal mode of communication.

The uniqueness of the superdiffuser also distinguishes it from other
conceptualizations of the inﬂuential. In particular, measures of opinion leadership and
activism which typically focus on the expertise of the inﬂuencer were demonstrated in
Studies 1 and 2 to be substantially predicted by only the maven measure. According to
the CFA of the measurement models, however, it is clear that the healthy lifestyle maven
cOnstruct is diStinct from opinion leadership and activism. Furthermore, the connector

and persuader constructs, either as direct effects or non-additively, contribute little to the

77

predictability of opinion leadership and activism. Therefore, given that current
conceptualizations of the inﬂuential overlap largely with the maven construct and
overlook connectedness and persuasiveness it is expected that employment of the
superdiffuser concept will allow for more precise delineation of inﬂuentials from non-
inﬂuentials than is possible with opinion leadership or activism. Consequently, viewing
inﬂuentials as unidimensional may be misleading. Rather viewing inﬂuentials as a non-
additive combination of the connector, persuader, and maven constructs may be more
useful because different types of people ﬁll different and important network functions.

The promises of these contributions are only as good as the quality of the
connector, persuader, and maven measures, however. Therefore, the three studies
conducted aimed to make fundamental improvements to the psychometric properties of
the measures. Study 1 found content validity evidence consistent with the conclusion that
the connector, persuader, and healthy lifestyle maven measures are three distinct,
internally consistent and parallel measures. The reliabilities of these measures were also
improved through adding additional items to the original measures developed by Boster
et al. (2006). The measures improved from a mean level of S]. a = .84 as observed in the
original measure development studies, to S]. a = .91.

This content validity evidence was replicated in Study 2 and sets of items for each
measure emerged that have been consistently parallel and internally consistent.
Therefore, a ﬁve item connector measure (items: 2, 3, 4, 6, and 9), an eight item
persuader measure (items: 1, 3, 4, 5, 6, 7, 9, and 10), and a six item healthy lifestyle
maven measure (items: 4, 5, 6, 8, 9, and 10) were identiﬁed. Study 2 further increased the

mean reliability of these measures to SI. a = .92 by employing a direct interval-

78

estimation scaling format. Furthermore, although multiple groups CF A demonstrated that
the scaling format did not alter the measures’ factor structures, the direct interval-
estimation format served to sensitize the upper end of the scale to allow for more precise
measurement of respondents with high levels of the constructs. Finally, Study 2
demonstrated construct validity evidence for the connector, persuader, and healthy
lifestyle maven measures using nomological networks. The hypothesized convergent
correlations were ample, the hypothesized divergent correlations were small with a few
exceptions, and there were no misﬁts in the networks that could not be explained
reasonably by revisiting theory or of such a magnitude that they could not be attributed to
sampling error.

In addition to the existing content and construct validity evidence, Study 3
demonstrated some mixed evidence for the predictive validity of the connector,
persuader, and healthy lifestyle maven measures. Although qualiﬁed by the small sample,
Study 3 provided some evidence for the position that scores on the connector, persuader,
and healthy lifestyle maven measures are predictive of behaviors that would result from
the extent to which people possess each of the characteristics. The ﬁndings of studies
one, two, and three taken together suggest that the connector, persuader, and healthy
lifestyle maven items form three highly reliable measures that exhibit characteristics of
content, construct, and predictive validity. This result has the potential to enhance studies
of social inﬂuence by reﬁning conceptualizations of the inﬂuential and allowing these
superdiffusers to be identiﬁed efﬁciently, without costly and complex sociometric

methods or observational studies.

79

Although this series of studies makes these contributions, it also raises several
questions that subsequent research must address. One immediate question is whether the
validity evidence presented here would generalize to other samples. Although the
connector, persuader, and maven constructs are hypothesized to be basic processes. These
studies reliance on the student sample leaves the answer to this proposition unknown.
Student populations have some unique psychographic and demographic characteristics
which may impact the generalizability of the ﬁndings of these studies. Psychographically,
in an older population, the skill component of the persuader and motivation component of
the argumentative might be more highly converged than in the college student which
could lead to different factor correlations. College students might have higher levels of
certain personality characteristics than the general population. If these personality
characteristics were akin to self-monitoring or narcissism, for example, it may make their
responses more likely to be tainted with social desirability biases than other populations.
Alternatively, the college student’s youth may make it more diﬁicult for them to report
accurately their levels of the superdiffuser constructs because of less practice with
introspection than an older population.

Demograhically, college environments may also have some unusual social
network and group identiﬁcation properties such as salient institutionalized ingroup-
outgroup categorizations which are highly intertwined. Consequently, a different pattern
of results might be obtained with a non-student sample. For example, the extent to which
someone is well connected might be considered most important in the inﬂuence process
whereas the persuader and maven constructs take second stage for some groups. For other

groups the extent to which someone is a health maven might carry more persuasive

80

weight than their connectedness or persuader characteristics. AS a result, different groups
may assign more importance to some of the superdiffuser constructs than others and
consequently the inter-construct correlations might vary. Therefore, although items on the
measures are expected to be interpreted in the same way by different demographic and
psychographic groups, there are several limitations with the use of the student sample
which must be addressed in future research. In addition to considering the performance of
these measures across different populations, it will be important in the future to
investigate the best way to norm the measures and move away from using standard
deviations for the superdiffuser classiﬁcation.

This research also raises questions about the extent to which the maven concept
can extend to other content domains. In these three studies, the maven construct was
measured as healthy lifestyle mavenism. Based on the conceptual deﬁnition of the maven
it is expected that the concept can indeed transcend content domain. There is also some
evidence from the marketing literature on market mavens (Feick & Price, 1987) which
approximates the idea of the maven in this line of research adding some credibility to the
theoretical claims about the generality of the maven concept made here. When examining
the generality of the maven in other content domains, however, it will be important to
consider the breadth of the domain so that there remains conceptual distinction between
mavens and opinion leaders. When investigating these possibilities it will be important to
consider whether the maven construct can be measured in other content domains by
modifying the existing healthy lifestyle maven items and exchanging the health reference
for the desired content domain or if the development of items unique to the content

domain would be necessary.

81

Another area of interest is the persuasive efﬁcacy of those identiﬁed as
superdiffusers. Thus far, superdiffusers have been discussed as though they are effective
inﬂuencers. Although there is theoretical and empirical evidence to suggest that people
who are highly connected or persuasive or expert make effective inﬂuencers, the
evidence provided, however, has not tested the effectiveness of people who possess all
three characteristics simultaneously. AS a result it is unclear whether a person would have

to exhibit high levels of all three constructs or if some of the constructs are not necessary

 

to be inﬂuential. Consequently, it is also important for future research to identify
superdiffusers, enlist their cooperation, and conduct experiments comparing the
persuasive ability of the superdiffuser against the non-superdiffuser. In addition to this
test, it is also important to consider the effectiveness of the communication campaign
employing superdiffusers against more traditional communication campaigns.

What also remains to be addressed is how many superdiffusers are necessary to
include in a communication campaign before their collective effectiveness is seen in
terms of persuasive outcomes. What is also important is not just the number of
superdiffusers but also how they are deployed. It may be the case that an initial wave of
superdiffusers may be used to initiate change, however, the strategic placement of a
second wave of superdiffusers that serve as ampliﬁers may be necessary to maintain the
change. Gladwell (2002) discusses some research addressing this question, including the
diffusion of innovations, investigating the proportion of a population that must adopt a
position before acceptance of the position sweeps through the rest of the population. It is
possible that the use of superdiﬁusers could reduce the critical mass that is needed to

sweep a position through the population of interest. If it is the case that the use of

82

superdiffusers can reduce the number of opinion leaders needed before change sweeps a
population then considerable resources could be saved.

The experiments necessary to investigate the persuasive efﬁcacy employing
superdiffusers in communication campaigns requires that some preliminary information
be uncovered ﬁrst. In particular, in order to recruit superdiffusers to work for a campaign
it is important to know what messages will inﬂuence them to accept the position.
Although it is possible that some superdiffusers will advocate the desired position by
default, it may be necessary to gain the conformity of some. Therefore, it would be
important to know if there are some messages that would be particularly inﬂuential for
superdiffusers. Given their desire to be socially connected it could be the case that
messages employing social norms would be particularly effective at inﬂuencing the
superdiffuser. Also, given their desire for expertise, evidence based messages could also
be particularly effective. These possibilities and others remain to be investigated in future
research.

Future research would also be well served to determine if people who are
superdiffusers employ a set of persuasive appeals unique to their classiﬁcation. If highly
connected, persuasive, content domain experts do indeed employ certain messages to
their targets it might be possible to train non-superdiffusers to use those message types
and obtain persuasive outcomes Similar to those that might be observed if true
superdiffers were used. In other words, it might be the case that an inﬂuence agent only
needs to have the inﬂuence target perceive that they have the qualities of the

superdiffuser to obtain the desired persuasive outcome. What is particularly interesting

83

 

from a communication perspective is that perceptions of all three characteristics of the
superdiffuser could be created through simple verbal or nonverbal messages.

It seems clear that the lines of research arising from this instrument development
have the potential to make interesting and important contributions to the social inﬂuence
literature. It also seems clear that the store of validity and reliability evidence for these
measures is accumulating. The validity and reliability assessment of measures are,
however, never complete. Examining the validity of the connector, persuader, and maven
measures through other-report versus self-report would provide additional evidence and
test whether people are able to identify these characteristics in others. The multi-trait
multi-method approach (Campbell & Fiske, 195 9) could also be used to provide
additional validity evidence for these measures as well. In the main, however, it is the
aim of this research to instigate more research into the characteristics of a speciﬁc

category of inﬂuentials, the superdiffuser.

84

APPENDICIES

85

APPENDIX A

Category Scaling Instructions

Instructions: Please read carefully each of the following statements. After reading each
statement, indicate the extent to which you agree with the statement by placing an X
clearly inside the box which represents how you feel. Please refer to the following
example before proceeding. If you have any questions please ask the researcher. Thank
you.

 

Exam le:

I try to avoid attacking the self-concept of others when debating an issue.

 

 

Disagree l l l I I I X l I I Agree
Strongly Disagree Neutral Agree Strongly
I enjoy a warm tub bath.

Disagree | l X | l | l | I I Agree
Strongly Disagree Neutral Agree Strongly

It may be considered old-fashioned by some, but having a decent, respectable
appearance is still the mark of a gentleman and, especially, a lady.

Disagree | l I l | l I l X I Agree
Strongly Disagree Neutral Agree Strongly

 

 

 

86

 

APPENDIX B

Demographic Items

1. What is your sex?

 

2. What is your age?

 

3. What year in school are you?

 

 

4. Have you ever held a position of leadership in a group, club, association, or team?

 

87

APPENDIX C

Connector Items

1.

2.

9.

I make friends quickly when I start a new job, class, or join a new church.

I’m often the link between friends in different groups.

I often ﬁnd myself introducing people to each other.

I try to bring people I know together when I think they would ﬁnd each other
interesting.

I maintain friendly, yet casual connections with most people I meet.

I frequently ﬁnd that I am the connection between people who would not otherwise
know one another.

People would be surprised if they knew how many ﬁiends ﬁom different groups that I
have.

I enjoy joining new groups because it means that I will have the opportunity to meet
new people.

The people I know often know each other because of me.

10. I am frequently the only link between different groups of people that I know.

88

 

APPENDIX D

Persuader Items

1.

2.

8.

9.

I am good at thinking of multiple ways to explain my position on an issue.
I don’t give up easily in an argument.

I’m able to argue well for a position I believe in.

. When in a discussion, I’m able to make others see my side of the issue.

I am able to adapt my method of argument in order to persuade someone.

When my approach to an argument is not working, it is easy for me to quickly come
up with something new that does work.

I can effortlessly offer multiple perspectives on an issue that all support my position.

When I argue, I am able to change a person’s mind without them even knowing it.

More often than not, I am able to convince others of my position during an argument.

10. I am Skilled at using my read of others to successfully persuade them.

89

L-

APPENDIX E

Healthy Lifestyle Maven Items

1. I like helping people by providing them with healthy lifestyle information they ﬁnd
pertinent to their lives. A

2. If someone asked where to ﬁnd information on a healthy lifestyle, I could point him
or her to a good source.

3. People ask me for information about healthy lifestyle topics.

4. My friends think of me as a good source of information when it comes to healthy
lifestyle issues.

5. When I know something about a healthy lifestyle topic, I feel it is important to share
that information with others. i

6. I like to be aware of the most up-to-date healthy lifestyle information so I can help
others by Sharing when it is relevant.

7. My friends know they could come to me if they needed information about a healthy
lifestyle.

8. If someone asked me about a healthy lifestyle issue that I was unsure of, I would
know how to help them ﬁnd the answer.

9. Being knowledgeable enough about healthy lifestyles so that I could teach someone
else is important to me.

10. People often seek me out for answers when they have questions about a healthy

lifestyle issue.

90

APPENDIX F

Need for Aﬂiliation Items

1.

2.

I like to be loyal to ﬁiends.

I like to participate in ﬁiendly groups.

I like to do things for friends.

I like to form new friendships.

I like to make as many ﬁiends as possible.

I like to share things with ﬁiends.

I like to do things with friends rather than alone.
I like to form strong attachments.

I like to write letters to friends.

91

APPENDIX G

Perspective Taking Items

1.

2.

6.

7.

Before criticizing somebody, I try to imagine how I would feel if I were in their place.

If I am sure I’m right about something, I don’t waste much time listening to other

 

people’s arguments.R W

I sometimes try to understand my friends better by imagining how things look from i”
their perspective. s'

I believe that there are two sides to every question and try to look at them both.

I sometimes ﬁnd it difficult to see things from the “other guy’s” point of view.R
I try to look at everybody’s side of a discussion before I make a decision.

When I’m upset at someone, I usually try to “put myself in his shoes” for a while.

Note. Responses to items marked with an R were reverse coded before analysis.

92

APPENDIX H

Empathic Concern Items

1.

2.

6.

7.

When I see people being taken advantage of, I feel kind of protective towards them.
When I see people being treated unfairly, I sometimes don’t feel very much pity for
them.R
I often have tender, concerned feelings for people less fortunate than me.

I would describe myself as a pretty soﬂ-hearted person.

Sometimes I don’t feel very sorry for other people when they are having problems.R

Other people’s misfortunes do not usually disturb me a great deal.R

I am often quite touched by things that I see happen.

Note. Responses to items marked with an R were reverse coded before analysis.

93

APPENDIX I

Opinion Leadership Items

1.

2.

5.

6.

My opinion on a healthy lifestyle seems not to count with other people.R

When they choose a healthy lifestyle, other people do not turn to me for advice.R
Other people rarely come to me for advice about choosing a healthy lifestyle.R
People that I know pick a healthy lifestyle based on what I have told them.

I often persuade other people to live the healthy lifestyle that I live.

I often inﬂuence people’s opinions about a healthy lifestyle.

Note. Responses to items marked with an R were reverse coded before analysis.

94

mil—w V
i

APPENDIX J

Healthy Lifestyle Activism Items

1. I would purchase a poster, t-shirt, etc. that endorses a healthy lifestyle.

2. I would serve as an ofﬁcer in a healthy lifestyle group.

3. I would attend an informational meeting of a healthy lifestyle group.

4. I would organize a healthy lifestyle event (e.g., talk, support group, march).
5. I would campaign door-to-door for a healthy lifestyle cause.

6. I would attend a healthy lifestyle organization’s regular planning meeting.
7. I would Sign a petition for a healthy lifestyle cause.

8. I would donate money to a healthy lifestyle organization.

9. I would wear a t—shirt or button with a healthy lifestyle message.

10. I would participate in discussion groups designed to discuss healthy lifestyles.

.95

APPENDIX K

Direct Interval-Estimation Scaling Instructions

Instructions: Please read carefully each of the following statements. After reading each
statement, think of a person who you feel exhibits none of the characteristic described by
the item and assign that person 0 points. Next, think of a person who you feel is the
prototype of the characteristic described by the item and assign that person 100 points.
Consider these two people as endpoints of a range which includes everyone. Now,
comparing yourself against the two people, using 0 as a minimum and 100 as a
maximum, indicate the extent to which you exhibit the characteristic described by the
item. If you feel you the extent to which you exhibit the characteristic described by the
item is closer to the person that personifies the characteristic, write a number closer to
100 in the space provided. If you ﬂee] you the extent to which you exhibit the
characteristic described by the item is closer to the person that exhibits none of the
characteristic, write a number closer to 0 in the space provided. You may use any
number between 0 and 100 to represent how you feel relative to the two comparison
people: 5, 34, 71, 92, etc. Please refer to the following example before proceeding. If you
have any questions please ask the researcher. Thank you.

 

Exam le:

1 have read the following item and have thought of a person who I feel does not try at all
to avoid attacking the self concept of others when debating an issue and assigned that
person 0 points. I have also thought of a person who I feel tries earnestly to avoid
attacking the self concept of others when debating an issue and assigned that person 100
points. I try to avoid attacking the self-concept of others when debating an issue about
one-half way between the middle of the range and the person I thought of that personiﬁes
this characteristic. Therefore, I responded with a 75 in the space provided.

I try to avoid attacking the self-concept of others when debating an issue.

E

 

I have read the following item and have thought of a person who I feel does not enjoy a
warm tub bath at all and assigned that person 0 points. I have also thought of a person
who I feel enjoys a warm tub bath very much and assigned that person 100 points. I enjoy
a warm tub bath about one-half way between the middle of the range and the person I
thought of that exhibits none of this characteristic. Therefore, I responded with a 25 in the
space provided.

I enjoy a warm tub bath.

25

 

96

‘. ‘3

I have read the following item and have thought of a person who I feel does not believe at
all that a respectable appearance is a virtue and assigned that person 0 points. I have also
thought of a person who I feel believes very strongly that a respectable appearance is a
virtue and assigned that person 100 points. I believe that a respectable appearance is a
virtue about in the middle of the range. Therefore, I responded with a 50 in the space
provided.

It may be considered old-fashioned by some, but having a decent, respectable
appearance is still the mark of a gentleman and, especially, a lady.

50

 

97

APPENDIX L

Social Anxiety Items

1.

2.

I often feel nervous even in casual get-togethers.

I usually feel uncomfortable when I am in a group of people I don’t know.
I am probably less shy in social interactions than most people.R

I seldom feel anxious in social interactions.R -_
I often feel nervous when calling someone I don’t know very well on the telephone.

I usually feel relaxed around other people, even people who are quite different from

myself.R

Note. Responses to items marked with an R were reverse coded before analysis.

98

 

 

APPENDIX M

Argumentativeness Items

1.

2.

6.

7.

Arguing over controversial issues improves my intelligence.

I enjoy avoiding arguments.R A

I enjoy a good argument over a controversial issue.

I enjoy defending my point of view on an issue.

I do not like to miss the opportunity to argue a controversial issue.
I consider an argument an exciting intellectual challenge.

I feel refreshed and satisﬁed after an argument on a controversial issue.

Note. Responses to the item marked with an R was reverse coded before analysis.

99

 

APPENDIX N

Value-Relevant Involvement Items

1. The values that are the most important to me are what determine my stand on healthy
lifestyle topics.

2. My position on healthy lifestyle topics is based on the values with which I try to
conduct my life.

3. Healthy lifestyle topics are relevant to the core principles that guide my life.

4. My beliefs about how I Should live my life determine my position on healthy lifestyle
topics.

5. My position on healthy lifestyle topics reﬂects who I am.

100

 

APPENDIX 0

Argumentative Complexity Interview Protocol

ﬂickground Information

I. Do you believe dieting, that is eating a restrictive diet, is a healthy or unhealthy
practice?

2. How strongly do you hold your belief?

Causal Theogj

3. For what reasons do you hold that position?

4. How do you know this supports your position?

5. Just to be sure I understand, can you explain exactly how your arguments support
your position?

Evidence

6. If you were trying to convince someone else that your view is right what evidence
[verbal emphasis] would you give to try to show this?

7. Can you be very speciﬁc, and tell me some particular facts that you could. mention to
try to convince the person?

Causal Theory

8. Is there anything further you could say to help Show that what you are correct?

Background

9. Can you remember when you began to hold this view?

10. Can you remember what it was that led you to believe what you do?

Altemative Causal Theory

 

11. Suppose that someone disagreed with your view. What might they [verbal emphasis]

101

 

say to try to convince you that you were wrong?

Alteraaﬁve Evidence

12. What evidence might this person give to try to convince you that you were wrong?

Alternative Causal Theog

13. Just to be sure I understand, can you explain exactly how they would think this would
Show that you were wrong?

14. In order to support their view, what arguments might this person give?

Counterargument

15. Is there any fact or evidence which, if it were true, would Show your view to be
wrong?

16. Could someone prove that you were wrong?

16a. (Probe if yes) How?

Alternative Causal Theory

17. Imagine a person like we have been talking about whose view is very different from
yours -— what might they say is their view?

REM

18. How would you respond? What would you say to try to convince them?

19. Just to be sure I understand, can you explain exactly how this would Show the person
was wrong?

20. Would you be able to prove this person wrong?

21. (Probe if yes) How?

22. What could you say to Show that your own view is the correct one?

102

 

APPENDIX P

Phonebook Test
__ Adams _ Allen __ Anderson __ Austin
__ Baker __ Bennett _ Brown __ Burgess
__ Campbell __ Carpenter __ Clark __ Cook
__ Davenport __ Davis __ Duncan __ Dunn
__ Edwards __ Elliott _ Erickson __ Evans
_ Farr _ Flynn __ Fox __ Fritz
__ Garcia __ Gardner __ Green _ Griffin
__ Hall __ Hamilton ' _'__ Hill _ Hunt
Ireland __ Irvin __ Irwin __ Ives
_ Jackson __ Jacobs __ Johnson ____ Jones
_ Kelley __ Kennedy __ Kim __ King
Lawrence __ Lewis __ Liu __ Lyon
__ Martinez __ McCormick __ Miller __ Murphy
__ Nash __ Nelson __ Nichols __ Norris
__ O’Neil _ Olson __ Oswald ______ Owen
__ Parker _____ Peterson __ Porter __ Powell
_ Quan __ Quinlan __ Quinn __ Quinney
__ Randall __ Reed __ Roberts __ Russell
__ Sanchez __ Schmitt __ Smith __ Stewart
__ Taylor __ Thelen __ Thomas __ Torres
__ Underwood __ Unsworth _ Upham __ Urban

103

Vasquez Villarreal Vincent Voss

__ Wagner _ Weber __ Williams __ Wright
_ Xi _ Xiao _ Xie __ Xu
_ Yan _ Yoder _ Young __ Yu
_ Zacks __ Zhang __ Zimmerman __ Zucker

104

 

APPENDIX Q

Healthy Lifestyle Knowledge Test
1. What is the leading cause of cancer death for men?

a. prostate cancer
b. colorectal cancer
c. lung cancer

(1. Skin cancer

2. Major signs of stroke include:

sudden numbness or weakness of the face, arms, or legs

sudden conﬁrsion, trouble speaking, or trouble understanding others
sudden trouble seeing in one or both eyes

sudden trouble walking, dizziness, or loss of balance or coordination
sudden severe headache with no known cause

A, C, and D

all of the above

tier-999.0579:

3. Colorectal cancer can be prevented by removing precancerous polyps or growths,
which can be present in the colon for years before invasive cancer develops. Current
guidelines recommend regular screening for all adults aged:

21 years or older
30 years or older
40 years or older
50 years or older
65 years or older

.09-99‘!”

4. What is the minimum amount of moderate-intensity physical activity adults should
engage in 5 or more days of the week?

a. 20 minutes per day
b. 30 minutes per day
c. 60 minutes per day
(I. 90 minutes per day

5. How many servings of fruits and vegetables should be eaten daily for good health?
a 3-5 servings per day
b. ‘ 4-8 servings per day
c
d

5-9 servings per day
. 6-10 servings per day

105

6. When engaging in strenuous exercise in high heat, it's important to monitor yourself
and others for Signs of heat stroke. Which of the following is NOT a Sign of heat
stroke?

red, hot, and dry skin with no sweating
dizziness and confusion

heavy sweating

nausea

9-9 9‘.»

7. Adults Should get vaccinations to protect against preventable diseases. This statement
is:

a. true
b. false

8. If untreated, gonorrhea can result in what health problem(s)?

epididymitis

higher risk of getting or transmitting HIV
cancer

both A and B

all of the above

09-99”?

9. What are some of the symptoms of prostate cancer?

blood in the urine

the need to urinate frequently, especially at night
weak or interrupted urine ﬂow

pain or burning feeling while urinating

the inability to urinate

constant pain in the lower back, pelvis, or upper thighs
all of the above

"PWSI’P‘PP‘P

10. What is the leading cause of death for men?

stroke
diabetes
cancer
suicide

heart disease

.09-99‘!”

11. Birth defects and disabilities cannot be prevented.

a. ‘ true
b. false

106

12. Cervical cancer is associated with a sexually transmitted infection.

a. true
b. false

13. Breast cancer is the leading cause of death in women.

a. true
b. false

14. Pregnant women should be screened for the bacteria group B streptococcus (group B
strep) late in pregnancy to reduce the risk of passing it to their newborn.

a. true
b. false

15. Overweight and obese people are at increased risk for cancer.

a. true
b. false

16. Pregnant women should not be vaccinated because it will deﬁnitely harm the baby.

a. true
b. false

17. People who smoke are at an increased risk for reproductive health problems.

a. true
b. false

Note. Key: l-C, 2-G, 3-D, 4-B, 5-C, 6-C, 7-A, 8-D, 9-G, lO-E, ll-B, 12-A, 13-B, l4-A, 15-A, l6-B, l7-A.

107

APPENDIX R

Interview Code Sheet

Causal Theogy

1) This respondent’s causal theory is...

Very I I I I I I I I I Very
Simple Simple Neutral Complex Complex

Evidence

1) The respondent’s ability to generate evidence for their position is...

Very I I I I I I I I | Very
Weak Weak Neutral Strong Strong

Altem_ative Causal Theog

1) This respondent’s alternative causal theory is...

Very I I I I I I I I I Very
Simple Simple Neutral Complex Complex

Alternative Evidence

1) The respondent’s ability to evidence for the alternative position is...

 

Very I I I I I I I I | Very
Weak Weak Neutral Strong Strong
Counterargument

 

1) The respondent’s ability to generate a counterargument to their position is...

Very I I I I I I I I I Very
Weak Weak Neutral Strong Strong

Rebuttal

l) The respondent’s ability to generate a rebuttal to the alternative position is...

Very | I I I I I I I I Very
Weak Weak Neutral Strong Strong

108

APPENDIX S

Argumentative Complexity Coding Examples
Low

Causal Theogy

For what reasons do you hold that position?
Because it is better for your body if you are eating properly and exercising

How do you know this supports your position?
I know there are a lot of pills and stuff and I don’t really believe in all that I think
it is better if you eat healthy and exercise so that’s just what I base it on

(Probe if necessary) Just to be sure I understand, can you explain exactly how your

arguments support your position?
No

Is there anything further you could say to help Show that you are correct?
Based on what I have read in magazines that’s all I can really offer

Evidence

If you were trying to convince someone else that your view is right what evidence

[verbal emphasis] would you give to try to Show this?
Just based on what I have see people try to do with dieting and I have seen what
works and what doesn’t people that stick with eating certain things at certain
times of the day with snacks and all that and exercising regularly that’s how I
would convince them

(Probe if necessary) Can you be very speciﬁc, and tell me some particular facts that

you could mention to try to convince the person?

109

No, I can’t think of any facts
Altem_aiive Causal Theory
Suppose that someone disagreed with your view. What might they [verbal emphasis]
say to try to convince you that you were wrong?
I don’t know
(Probe if necessary) Just to be sure I understand, can you explain exactly how they
would think this would Show that you were wrong?
I don’t know, maybe their own evidence, what they think
(Probe if necessary) In order to support their view, what arguments might this person
give?
Just that what they think is more right that what I think
(Omit if alternative theory already is generated) Imagine a person like we have been
talking about whose view is very different from yours — what might they say is their
view?
That they don’t agree with dieting and that something else would be better than
dieting
Alternative Evidence
What evidence might this person give to try to convince you that you were wrong?
Just that maybe pills are better or surgery or surgery might be better than dieting
Counterargument
(Probe if not already indicated) Is there any fact or evidence which, if it were true,
would show your view to be wrong?

Perhaps but I just think it depends on the person and that they want to do to

110

maintain their own lifestyle
Could someone prove that you were wrong?
Yes
(Probe if yes) How?
Just with facts, dirty facts
Rebuttal
(Include if no alternative theory is generated) How would you respond? What would
you say to try to convince them?
Just that I know what to be true is different from what they know and whether
they could persuade me, I don’t know
(Probe if necessary) Just to be sure I understand, can you explain exactly how this
would Show the person was wrong?
No
Would you be able to prove this person wrong?
Yes
(Probe if yes) How?
If I actually had something I could show them
(Probe if not already indicated) What could you say to show that your own view is
the correct one?

I don’t know

Moderate

Causal Theory

For what reasons do you hold that position?

111

Because if you are trying to go on a diet to lose weight or to be healthier you have
to restrict your diet in some way otherwise it is not going to change so you have
to lower your caloric intake in order to do that
How do you know this supports your position?
There has been scientiﬁc evidence that says when you are at a lower weight than
being obese you are at less risk for diabetes heart disease and that sort of thing
(Probe if necessary) Just to be sure I understand, can you explain exactly how your
arguments support your position?
Because I think that when you are dieting y0u are increaSing your health and
those studies support that
Is there anything further you could say to help Show that you are correct?
I could redirect them to someone who could give them more information or try to
convince them
Evidence
If you were trying to convince someone else that your view is right what evidence
[verbal emphasis] would you give to try to show this?
If I knew anyone with speciﬁc examples of losing weight and it being healthy I
would give them those if they had a speciﬁc topic in mind like lowering blood
pressure or whatever I would direct them to a website that would do that or a
doctor that could give them more information about it
(Probe if necessary) Can you be very speciﬁc, and tell me some particular facts that
you could mention to try to convince the person?

Dieting does lower your risk for heart disease and high blood pressure

112

Altem_ative Causal Theory

Suppose that someone disagreed with your view. What might they [verbal emphasis]

say to try to convince you that you were wrong?
They could give me speciﬁc examples of people they know where dieting didn’t
help them when they tried to lose weight and they still had high blood pressure
they could probably try to give me information from Sites that they have seen that
opposes my view

Alternative Evidence

What evidence might this person give to try to convince you that you were wrong?
I don’t know because I don’t feel that way they could Show me a number of things
like a speciﬁc example that they know of

Alternative Causal Theog

(Probe if necessary) Just to be sure I understand, can you explain exactly how they

would think this would Show that you were wrong?
If they have seen examples where my theory doesn’t hold then yeah, they are
going to believe that it doesn’t work

(Probe if necessary) In order to support their view, what arguments might this person

give?
Just the argument that dieting can be bad for you if you are not still getting your
daily fruits and vegetables it can actually hurt you more than help you in some
circumstances

(Omit if alternative theory already is generated) Imagine a person like we have been

talking about whose view is very different from yours — what might they say is their

113

view?
They could tell me that they believe dieting is unhealthy because I believe dieting
is healthy
Comtermggnrent
(Probe if not already indicated) Is there any fact or evidence which, if it were true,
would show your view to be wrong?
Only if you diet in the wrong way, me I would want to do it in a healthy way but
if you were to do it in an unhealthy way then yes
Could someone prove that you were wrong?
I hope not
(Probe if yes) How?
Not applicable
Mm
(Include if no alternative theory is generated) How would you respond? What would
you say to try to convince them?
I would tell them that it is proven when you eat less fat and more lean protein and
fruits and vegetable as opposed to the opposite that you are healthier all the way
around most people would see lower blood pressure, less likely to have heart
disease because their arteries are not as clogged
(Probe if necessary) Just to be sure I understand, can you explain exactly how this
would Show the person was wrong?
It would tell them that they are wrong by doing tests that Show their blood

pressure lower they have leaner body mass

114

Would you be able to prove this person wrong?
I hope so
(Probe if yes) How?
I would Show them by studies that have been done in a healthy way how it can
beneﬁt the person in both the long run and the short run
(Probe if not already indicated) What could you say to show that your own view is
the correct one?
I think I would do the same thing I don’t think I would do anything differently
High
Causal Theogy
For what reasons do you hold that position?
Because I know so many people who eat healthy for a little while and then it all
just comes right back and sometimes even more because when you stop eating a
certain food or cut carbs out of your diet you’ll lose a lot of weight but ﬁrst of all
its not very healthy and second of all when you do start to eat it again your body
stores it. You jut need to learn how to control proportions
How do you know this supports your position?
I’ve kind of researched it, my mom is a nurse, and my friends and I discuss it all
the time
(Probe if necessary) Just to be sure I understand, can you explain exactly how your
arguments support your position?
Making your diet healthy and balanced is a better idea if you are trying to lose V

weight or trying to become a better person because it will decrease your weight if

115

that is what you are looking for or it will make it a permanent lifestyle change
instead of a spur of the moment for a couple of months I am not going to eat a
certain food but when you do eat it again your body will turn it straight to fat and
store it like the hibernation effect when a bear hibemates it hinges and turns all of
the food to fat and the same thing occurs when you stop eating something when
you start eating again your body stores it so when you don’t have it it has it that’s
why dieting for a period of time is not a good idea you Should have a diet that is a
permanent lifestyle

Is there anything ﬁrrther you could say to help Show that you are correct?
I probably could and it depends on the person if they are going for losing weight I
could give them tips about what they should do and make sure they ﬁt in exercise
into their daily things and go over what they eat everyday to see if there are
certain things they ate a lot of or less of certain things

Evidence

If you were trying to convince someone else that your view is right what evidence

[verbal emphasis] would you give to try to Show this?
I could Show websites, or books, notes from biology class and physical evidence
to prove it to them and then I could also Show examples of ﬁiends or family
members

(Probe if necessary) Can you be very speciﬁc, and tell me some particular facts that

you could mention to try to convince the person?
Becoming healthy basically needs to be a balanced diet and when you are trying

to lose weight it’s a balancing act of how many calories you take in and how

116

many you use and if you use more calories than you are eating you are going to
lose weight if you’re eating more than what you are doing and that’s a big deal
and your diet needs to be balanced and if it is not your body will store certain
things and like an example would be like if you cut carbs from your diet your
body will get smaller but as soon as you start eating them you will gain that back
because your body will immediately store it

Alternative Causal Theory

Suppose that someone disagreed with your view. What might they [verbal emphasis]

say to try to convince you that you were wrong?
They might tell me the example of someone who needs to diet for health reasons
probably that’s how I would see it the other way around and that it is better to eat
nothing than to eat bread because of the carbs in bread and I would basically be
like you need that and they would try to convince me that as long as you are
eating less it could be just eat a couple bites of your pizza and you are good for
the day and I would keep arguing but they would keep trying to tell me that
dieting is good because you’ll lose weight faster

(Probe if necessary) Just to be sure I understand, can you explain exactly how they

would think this would Show that you were wrong?
It would give concrete evidence and it is easier to look at someone who just
dropped 100 pounds and say that’s because of a diet or they went on the Atkins or
they did Weight Watchers or talk about someone who changed they lifestyle of
eating and argue that their genes made them lost weight so they would suggest

there is no way of proving that a lifestyle of eating right is better than dieting and

117

they are not going to bring up the people who gained it back after getting off their
diet
(Probe if necessary) In order to support their view, what arguments might this person
give?
People who are severely obese need to go on a diet or do something drastic in
order to help them out so they don’t get really sick or have something deadly
happen cause some people like to eat that much like that or they could try to say
that Spring break is coming up and we need to diet fast and it you just change your
lifestyle it will take too much time and you won’t see the results
(Omit if alternative theory already is generated) Imagine a person like we have been
talking about whose view is very different from yours — what might they say is their
view?
That diets are good and losing weight it healthy
Alternative Evidence
What evidence might this person give to try to convince you that you were wrong?
They would tell me about people or themselves that have lost weight because of
dieting and probably give me a lot of examples even though from my point they
won’t keep the weight off unless they change their lifestyle
Counterargument
(Probe if not already indicated) Is there any fact or evidence which, if it were true,
would Show your view to be wrong?
Like I said I think the only reason that would be wrong and if you call something

a diet and you stay with it the rest of your life its not a diet it is a lifestyle but if

118

they say I am wrong in the essence that people who are so big that they can’t ﬁt
through a doorway or can’t roll out of bed I think that those people need to go on
a diet or need to do something severe to drop the weight fast and then slowly
make it a lifestyle change
Could someone prove that you were wrong?
Yes
(Probe if yes) How?
I think they would do that by pointing out examples about like obese people that
are in the hospital that are severely obese and I think that they could try to prove
me wrong by pointing out examples of people
329%
(Include if no alternative theory is generated) How would you respond?'What would
you say to try to convince them?
I would say that yes it is good to lose weight and eat healthier things but it should
be a life change and it Should be a permanent thing that includes exercise and it is
not just what you eat for a Short period of time, a life change that will last forever
(Probe if necessary) Just to be sure I understand, can you explain exactly how this
would Show the person was wrong?
I would say, have you ever seen someone who went on a diet and did really good
at ﬁrst and lost a ton of weight and then they ventured off of it and then all of the
sudden all of the weight is back and plus some and not know why. I know several
people who went on the Atkins who lost a lot of weight and then boom it came

back or people that you know who went on Spring break and they drop a ton of

119

weight right before and then they come back and it’s all right back and their
wearing hoodies and sweats everywhere they gained it all back when you venture
off of a diet you will gain back the same if not more weight because all you did is
change your diet for a short period of time which is actually bad for your body I
have a friend that did Weight Watchers and she lost a ton of weight and now she
has gained it all back since she went off it even though she eat pretty healthy
because she restricted herself for so long because her body thought that’s how it i
would be eating forever but it wasn’t so her body decided to store everything
since she went off her diet so you whould make a diet something that is a
permanent lifestyle change

Would you be able to prove this person wrong?
Yes

(Probe if yes) How?
I could show then examples of people who have gone off their diets and how it
effects you weight and motions and tell them stories and examples to convince
them that way

(Probe if not alreaay indicated) What could you say to Show that your own view is

the correct one?
Books could be helpful to go beyond the standard evidence or my word and Show
them here is where it is in print of take them to the gym to talk to dietitians and

things like that

120

REFERENCES

121

REFERENCES

Berger, C. R. & Calabrese, R. J. (1975). Some explorations in initial interaction and
beyond: Toward a developmental theory of interpersonal communication. Human
Communication Research, 1, 99-112.

Boster, F. J ., Kotowski, M. R., & Andrews, K. R. (2006, November). Identijying
Inﬂuentials: Development of the Connector, Persuader, and Social Maven Scales.
Paper accepted for presentation at the annual meeting of the National
Communication Association, San Antonio, TX.

Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the
- multi-trait multi-method matrix. Psychological Bulletin, 56, 81-105.

Campo, M. L. (1999). Arguing for Change: Arguments by activists, non-activists, and in
the media regarding domestic partner beneﬁts. Unpublished Doctoral
Dissertation. Michigan State University, MI

Celentano, D. C., Bond, K. C., Lyles, C. M., Eiumtrakul, S., Go, V. F. L., Beyrer, C.,
Chiangmai, C., Nelson, K. E., Khamboonruang, C., & Vaddhanaphuti, C. (2000).
Preventive intervention to reduce sexually transmitted infections: A ﬁeld trial in
the royal Thai army. Archives of Internal Medicine, I 60, 535-540.

Centers for Disease Control and Prevention. (2005). Questions and Answers: Test your
knowledge about women ’s health. Retrieved January 7th, 2007, from
http://www.cdc.gov/women/test/testall.htm

Centers for Disease Control and Prevention. (2006). Questions and Answers: Test your
knowledge about men ’s health. Retrieved January 7th, 2007, from
http://www2a.cdc.gov/od/menshealth/quizprint.asp

Chaffee, S. H. (1982). Mass media and interpersonal channels: Competitive, convergent,
or complimentary? In G. Gumpert & R. Cathart (Eds), Inter/Media:
Interpersonal communication in a media world. New York, NY: Oxford
University Press.

Cho, H. & Boster, F. J. (2005). Development and validation of value-, outcome-, and
impression-relevant involvement scales. Communication Research, 32, 23 5-264.

Corning, A. F. & Myers, D. J. (2002). Individual orientation toward engagement in social
action. Political Psychology, 23, 703-729.

Cronbach, L. J. (1951). Coefﬁcient alpha and the internal structure of tests.
Psychometrika, 16, 297-334.

122

Cronbach, L. J. & Meehl, P. E. (1955). Construct validity in psychological tests.
Psychological Bulletin, 52, 281-301.

Davis, M. A. (1980). A multidimensional approach to individual differences in empathy.
JSAS Catalog of Selected Documents in Psychology, 10, 85.

Davis, M. A. (1983). Measuring individual differences in empathy: Evidence for a
multidimensional approach. Journal of Personality and Social Psychology, 44,
113-126. .

Durland, M. M. & Fredericks, K. A. (2006). Social Network Analysis in Program
Evaluation. New Directions for Evaluation, 107. Hoboken, NJ: Wiley.

Earp, J. L., Eng, E., O’Malley, M. S., Altpeter, M., Rauscher, G., Mayne, L., Matthews,
H. F., Lynch, K. S., & Gaquish, B. (2002). Increasing use of mammography
among older, rural Aﬁican American women: Results from a community trial.
American Journal of Public Health, 82, 646-654.

Ebel, R. L. (1951). Estimation of the reliability of ratings. Psychometrika, 16, 407-424.

Edwards, A. L. (1959). Edwards Personal Preference Schedule: Manual. New York,
NY: The Psychological Corporation.

Feick, L. F. & Price, L. L. (1987). The market maven: A diffuser of marketplace
information. Journal of Marketing, 51, 83-97.

Festinger, L. (1949). The analysis of sociograms using matrix algebra. Human Relations,
2, 153-158.

Festinger, L. (1953). An analysis of compliant behavior. In M. Sheriff & M. O. Wilson
(Eds), Group Relations at the Crossroads. New York, NY: Harper.

Flynn, L. R., Goldsmith, R. E., & Eastman, J. K. (1996). Opinion leaders and opinion
seekers: Two new measurement scales. Journal of the Academy of Marketing
Science, 24, 137-147.

Friedman, H. 8., Prince, L. M., Riggio, R. E., & DiMatteo M. R. (1980). Understanding
and assessing nonverbal expressiveness: The affective communication test.

Journal of Personality and Social Psychology, 39, 333-351.

Gladwell, M. (2002). The. Tipping Point: How little things can make a big diﬂerence.
New York, NY: Back Bay Books/Little, Brown, and Company.

Goldsmith, E. B. & Goldsmith, R. E. (1980). Dogmatism and conﬁdence as related
factors in evaluation of new products. Psychological Reports, 4 7, 1068-1070.

123

Granovetter, M. (1973). The strength of weak ties American Journal of Sociology, 78,
1360-1380.

Granovetter, M. (1978). Threshold models of collective behavior. American Journal of
Sociology, 83, 1420-1443.

Granovetter, M. (1982). The strength of weak ties: A network theory revisited. In P. V.
Mardsen & N. Lin (Eds), Social Structure and Network Analysis. Newbury Park,
CA: Sage.

Hunter, J. E., & Gerbing, D. W. (1982). Unidirnensional measurement, second order
factor analysis, and causal models. Research in Organizational Behavior, 4, 267-
320.

Hunter, I. E., & Hamilton, M. A. (1992). CFA — A Program in Basic to do Conﬁrmatory
Factor Analysis. East Lansing, MI: Michigan State University.

Hunter, J. E., & Shotland, R. L. (1974). Treating data collected by the “small world”
method as a Markov process. Social Forces, 52, 321-332.

Infante, D. A. & Rancer, A. S. (1982). A conceptualization and measure of
argumentativeness. Journal of Personality Assessment, 46, 72-80.

Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion
Quarterly, 24, 163-204.

Katz, E. (195 7). The two-step ﬂow of communication: An up-to-date report on an
hypothesis. Public Opinion Quarterly, 21, 61-78.

Katz, E. & Lazarsfeld, P. F. (1955). Personal Inﬂuence: The part played by people in the
ﬂow of mass communications. New York, NY: The Free Press.

Kelly, J. A., St. Lawrence, J. S., Stevenson, L. Y., Hauth, A. C., Kalichman, S. C, Diaz,
Y. E., Brasﬁeld, T. L., Koob, J. J., & Morgan, M. G. (1992). Community
AIDS/HIV risk reduction: The effects of endorsements of popular people in three
cities. American Journal of Public Health, 82, 1483-1489.

Kirchner, W. K., Dunnette, M. D., & Mousley, N. (1960). Use of the Edwards personal
preference schedule in the selection of salesmen. Personnel Psychology, 13, 421-
424.

Kuhn, D. (1991). The Skills of Argument. Cambridge, UK: Cambridge University Press.

Lazarsfeld, P. F. & Merton, R. K. (1964). Friendship as a social process: A substantive

and methodological analysis. In M. Berger, T. Abel, & C. H. Page (Eds),
Freedom and Control in Modern Society. New York, NY: Octagon.

124

Leary, M. R. (1983). Social anxiousness: The construct and its measurement. Journal of
Personality Assessment, 4 7, 66-75.

Levine, T. R. (1994). Do individuals make interval/ratio level responses to magnitude
scaled items? Journal of Social Behavior and Personality, 9, 377-386.

Lodge, M. (1981). Magnitude scaling: Quantitative measurement of opinions.
Quantitative Applications in the Social Sciences, 25. Newbury Park, CA: Sage.

Mann, J. H. (1958) Self-Ratings and the EPPS. Journal of Applied Psychology, 42, 267-
268.

McCroskey, J. C., Beatty, M. J ., Kearney, R, & Plax, T. G. (1985). The content validity
of the PRCA-24 as a measure of communication apprehension across
communication contexts. Communication Quarterly, 33, 165-173.

Meek, P. M., Sennott-Miller, L., & Ferketich, S. L. (1992). Scaling stimuli with
magnitude estimation. Research in Nursing & Health, I 5, 77-81.

Milgram, S. (1977). The Individual in a Social World: Essays and experiments. Reading,
MA: Adison-Wesley.

Miller, R. L., Klotz, D., & Eckholdt, H. M. (1998). HIV prevention with male prostitutes
and patrons of hustler bars: Replication of an HIV prevention intervention.
American Journal of Community Psychology, 26, 97-131.

Moreno, J. L. (1934). Who Shall Survive?: Foundations of sociometry, group

psychotherapy, and sociodrama. Washington, DC: Nervous and Mental Disease
Publishing.

Murray, H. A. (193 8). Explorations in Personality. New York, NY: Oxford University
Press.

Nunnally, J. C. & Bernstein, I. H. (1994). Psychometric Theory. (3rd ed.), New York, NY:
McGraw Hill.

O’Keefe, D. J. (2002). Persuasion: Theory and research. (2"d ed.), Thousand Oaks, CA:
Sage.

Palmgreen, P., Donohew, L., Lorch, E. P., Hoyle, R. H., & Stephenson, M. (2002).
Television campaigns and sensation seeking targeting of adolescent marijuana
use: A controlled time-series approach. In R. Homik (Ed.), Public Health
Communication: Evidence for behavior change. Hillsdale, NJ: Lawrence
Erlbaum.

' 125

Price, L. L., Feick, L. F., & Guskey, A. (1995). Everyday market helping behavior.
Journal of Public Policy and Marketing, 14, 255-266.

Rogers, E. M. (1995). Diﬂusion of Innovations. (4th ed.), New York, NY: The Free Press.

Rogers, E. M. & Cartano, D. G. (1962). Methods of measuring opinion leadership. Public
Opinion Quarterly, 26, 435-441.

Rosenstock, I. M. (1974). Historical origins of the health belief model. Health Education
Monographs, 2, 328-335.

Rothschild, B. (2005). Health literacy: What the issue is, what is happening, and what can
be done. Health Promotion Practice, 6, 8-11.

Salmon C. T. & Atkin, C. K. (2003). Using media campaigns for health promotion. In T.
L. Thompson, A. M. Dorsey, K. 1. Miller, & R. Parrot (Eds), Handbook of Health
Communication. Mahwah, NJ: Lawrence Erlbaum.

Scott, J. (2000). Social Network Analysis: A handbook. (2"d ed.), Thousand Oaks, CA:
Sage.

Shestowsky, D., Wegener, D. T., Fabrigar, L. R. (1998). Need for cognition and
interpersonal inﬂuence: Individual differences in impact on dyadic decisions.
Journal of Personality and Social Psychology, 74, 1317-1328.

Shavitt, S. (1990). The role of attitude objects in attitude functions. Journal of
Experimental Social Psychology, 26, 124-148. -

Silvennan, F. H. & Johnston, R. G. (1975). Direct interval-estimation: A ratio scaling
format. Perceptual and Motor Skills, 41, 464-466.

Soumerai, S. B., McLauglin, T. J., Gurwitz, J. H., Guadagnoli, E., Hauptman, P. J.,
Borbas, C., Morris, N., McLaughlin, B., Gao, X., Willison, D. J ., Asinger, R., &
Gobel, F. (1998). Effect of local medical opinion leaders on quality of care of

acute myocardial infarction. Journal of the American Medical Association, 279,
1358-1363.

Stevens, S. S. (1956). The direct estimation of sensory magnitudes - loudness. The
American Journal of Psychology, 69, 1-25.

Travers, J. & Milgram, S. (1969). An experimental study of the small world problem.
Sociometry, 32, 425-443.

Valente, ‘T. W. (1995). Network Models of the Diffusion of Innovations. Cresskill, NJ:
Hampton Press.

126

Valente, T. W. (2001). Evaluating communication campaigns. In R. E. Rice & C. K.
Atkin (Eds), Public Communication Campaigns. (3rd ed.), Thousand Oaks, CA:
Sage.

127

 

   

III IIIIIIII IIIIIIIIII