2017, Vol. 28(5) 578 –586
© The Author(s) 2017
Reprints and permissions:
Central among the 2010 Common Core State Standards
(CCSS) is proficiency in nonnarrative reading and exposi-
tory argumentive writing. Similarly, the new Next Genera-
tion Science Standards (NGSS) identify argumentation as
key among the process skills these standards feature (NGSS
Lead States, 2013), as well as central to achieving mature
epistemological understanding of the nature of science
(Sandoval, 2014). However, not articulated in the CCSS or
in the NGSS is how these proficiencies are achieved. The
writing component is a particular challenge. Students of all
ages find expository writing challenging and perform
poorly in assessments (Graham, McKeown, Kiuhara, &
Harris, 2012), and it is unclear how young writers asked to
support claims with logical reasoning and relevant evi-
dence, as the CCSS (2010) stipulate, can best learn to do so.
The dialogic approach to achieving this goal, reported
on by Kuhn and Crowell (2011), rests on the view that
dialogue supports the development of written argument
by giving it a purpose. There is now someone to commu-
nicate to—the missing interlocutor (Graff, 2003)—and a
purpose for communicating. This dialogic approach to
promoting individual argumentive competence rests on a
tradition that emphasizes the close connection between
social and individual cognition. This tradition can be con-
sidered to go back as far even as Socrates, but certainly
goes back to Vygotsky (1937/1987) and Mead (1934/1967),
who described thought as a “conversation with the gener-
alized other,” and, later, Billig (1987), who emphasized
the connection between dialogic argument and the interi-
orized individual argument he claimed occurs in thought.
Walton (2014) attributed to Grice (1975) the introduction
of dialogic theory to modern analytical philosophy and
attributed the further development of this theory to
van Eemeren and his colleagues (e.g., van Eemeren &
Grootendorst, 1992), who emphasized the need to evaluate
arguments within their conversational context. According
to Grice, an argument should be evaluated on the basis of
its collaborative value as a contribution to dialogue.
Contemporary empirical studies have supported Graff’s
(2003) claim of the benefits of dialogue, demonstrating
that scrutiny of an opposing view and use of evidence are
greater in dialogic contexts than in individual written
689248PSSXXX10.1177/0956797616689248Zavala, KuhnSolitary Discourse
Deanna Kuhn, 525 W. 120th St., New York, NY 10027
Solitary Discourse Is a Productive
Julia Zavala and Deanna Kuhn
Teachers College, Columbia University
Young adults received information regarding the platforms of two candidates for mayor of a troubled city. Half
constructed a dialogue between advocates of the candidates, and the other half wrote an essay evaluating the
candidates’ merits. Both groups then wrote a script for a TV spot favoring their preferred candidate. Results supported
our hypothesis that the dialogic task would lead to deeper, more comprehensive processing of the two positions,
and hence a richer representation of them. The TV scripts of the dialogue group included more references to city
problems, candidates’ proposed actions, and links between them, as well as more criticisms of proposed actions and
integrative judgments extending across multiple problems or proposed actions. Assessment of levels of epistemological
understanding administered to the two groups after the writing tasks revealed that the dialogic group exhibited a
lesser frequency of the absolutist position that knowledge consists of facts knowable with certainty. The potential of
imagined interaction as a substitute for actual social exchange is considered.
cognitive complexity, discourse analysis, reasoning, role taking, citizenship
Received 8/10/16; Revision accepted 12/23/16
Solitary Discourse 579
argument (Kuhn & Moore, 2015; Macagno, 2016). Middle
schoolers participating in a multiyear intervention cen-
tered around peer dialogues, Kuhn and Crowell (2011)
found, wrote superior essays, compared with a control
group who participated in a nondialogic intervention that
had equal scope but was focused on whole-class discus-
sion and essay writing.
In the study reported here, we asked whether a paral-
lel benefit might appear in the context of a much more
limited intervention, namely, a 1-hr individual session.
Instead of engaging in actual dialogue, participants were
asked to construct a written hypothetical dialogue
between two expert arguers regarding which of two may-
oral candidates who held contrasting positions was the
better candidate. A comparison nondialogic group wrote
an essay evaluating the merits of the two candidates. In
addition to comparing characteristics of the dialogues
and essays, we compared performance on a second, non-
dialogic writing task on the same topic. Our hypothesis
was that the dialogic task would lead to a deeper, more
comprehensive processing of the two positions, and
hence a richer representation of them, that would be
manifested not only in the constructed dialogue itself but
also in the second, nondialogic writing task. Finally, a
measure of epistemological understanding (asking the
respondent to account for discrepant accounts of an
event; Barzilai & Weinstock, 2015) was administered in
order for us to examine possible effects of the dialogue
task on this important competency.
Sixty undergraduate students were recruited through
notices posted around the campus of a 4-year college in a
large city in the Northeast United States. Their mean age
was 23.4 years (SD = 6.3 years, range = 17–55 years). Fifteen
percent were freshman, 12% sophomores, 30% juniors, and
42% seniors. They reported more than 20 different majors
or prospective majors. Eighty-three percent were female
and 17% male; the high percentage of females reflects the
gender distribution of the entire population of the college
(73% female). Fifty percent of the sample self-identified as
Black or African American, 38% as Hispanic or Latino, 5%
as White, and 2% as “other”; 3% did not report their race or
ethnicity. The students were compensated for their partici-
pation with a $10 Dunkin’ Donuts gift card.
The activity was conducted individually in a quiet room.
Students were allotted as much time as needed, which
ranged from 45 to 60 min. A random-number generator
was used to assign 30 students to the experimental group
and 30 to the comparison group. Upon a student’s arrival,
the activity was explained, and the student gave consent.
Following completion of the task, the student answered
a brief demographic questionnaire.
Experimental group’s task. Students in the experi-
mental group worked first on a dialogic writing task
(Kuhn, Zillmer, Crowell, & Zavala, 2013). They were
asked to construct a dialogue between two expert argu-
ers (TV commentators Chuck and Doug) regarding who
was the better candidate (Ana Cruz or Maria Diaz) for
mayor of a troubled city. A list of major problems in the
city was presented, as were lists of the actions each can-
didate proposed to take if elected. (See Table 1 for the
instructions and the information provided.)
Comparison group’s task. Students in the comparison
(essay) group received a sheet containing the same infor-
mation, but different instructions. In place of the section
introducing Chuck and Doug and the dialogic frame,
their materials had the following instruction:
Write an argumentive essay in which you consider
the merits of each of these candidates for mayor.
(Also, the word essay replaced the word script in the final
sentence of the instructions.)
TV-script task (all participants). Next, participants
from both groups were asked to write a TV script:
You’ve been asked to appear on a TV show and
make a case for either Cruz or Diaz as the best
candidate. Who would you choose and what would
you say in your 2-minute talk? Write a short script
Epistemological assessment (all participants). Fol-
lowing the writing tasks, all participants were presented
the Livia problem (Barzilai & Weinstock, 2015; Kuhn,
Pennington, & Leadbeater, 1983) as an assessment of
their epistemological thinking. They were asked to follow
along on a written copy as the experimenter read aloud
two accounts of the fictitious Fifth Livian War between
North and South Livia. One account was from a North
Livian historian, and the other was from a South Livian
historian. The participants then answered questions
regarding their understanding of the discrepancies
between the accounts. The questions were read aloud,
the student answered verbally, and their answers were
580 Zavala, Kuhn
We first applied a common coding scheme to the dia-
logues written by the dialogue-group participants and
the essays written by the essay-group participants. Our
intention was to identify ways in which the dialogues and
essays differed, to better understand any subsequent dif-
ferences between the groups in the TV scripts they wrote.
We then turned to the TV scripts themselves and the
assessment of epistemological thinking.
Differences between the dialogues and
For the purpose of identifying differences between the
dialogues and essays, we used the coding scheme in
Table 2 (from Kuhn et al., 2013). First, the constructed
dialogues and essays were segmented into idea units,
each of which expressed a single idea or assertion. Two
coders independently segmented six dialogues and eight
essays to establish interrater reliability. The percentage
agreement for identifying segments was 100%. Each idea
unit in those dialogues and essays was then classified
according to the coding scheme in Table 2. Percentage
agreement between the same two coders was 91%. Dis-
agreements were resolved through discussion. The
remaining dialogues and essays were coded by the first
The mean number of idea units differed significantly
between the dialogues and essays, t(58) = −2.72, p = .009,
d = 0.70. The dialogue group produced more idea units
(M = 9.33, SD = 4.49) than the essay group (M = 6.67,
SD = 2.96). (As normality was violated for the number of
idea units, a square-root transformation was applied;
results were similar for the nontransformed data.)
A multinomial logistic regression, via SAS/Nonlinear
Mixed (MLE), was conducted, with the frequencies of the
category types in Table 2 as dependent variables and
group (dialogue vs. essay) as an independent variable.
Unsubstantiated statements were set as the reference cat-
egory. The dependent variables were nested within each
participant; therefore, a mixed-effects random-intercept
logistic regression was fitted.
A significant group effect was found for two catego-
ries, simple comparisons and integrative comparisons.
The mean number of simple comparative statements was
0.20 (SD = 0.55) for the essays of the comparison group
and 1.30 (SD = 1.58) for the dialogues of the experimen-
tal group. Similarly, the mean number of integrative state-
ments was 0.40 (SD = 0.68) for the essays of the
comparison group and 1.67 (SD = 1.63) for the dialogues
of the experimental group. Relative to the odds of making
Table 1. Dialogic Task Presented to Participants in the Experimental Group
Ana Cruz and Maria Diaz are running for mayor of their troubled large city. Among the city’s problems are high housing costs,
teen crime, traffic, school dropout, and unemployment.
Chuck and Doug are TV commentators arguing about who is the better candidate. Write a script of what they might say. They are
both experts on the city; they are both expert arguers and evenly matched. So your script should present the most well argued
debate you can construct.
Begin your script like this:
CHUCK: Cruz should be elected mayor because she’ll do better than Diaz.
DOUG: I disagree, because xxxxxx
Then continue their argument, filling in what each one might say:
CHUCK: xxxxxx etc.
Here is some information about Cruz’ positions. She promises to:
-create job training programs
-expand city parks
-raise teachers’ pay
-open walk-in health clinics
-impose a teen curfew
-employ senior citizens in city schools
Here is some information about Diaz’ positions. She promises to:
-improve public transportation
-open more centers for senior citizens
-revise the high school curriculum
-build a new athletic stadium
-improve health care
-build more housing
(You are not required to include all the above topics in your script.)
Solitary Discourse 581
an unsubstantiated statement, the odds of making a sim-
ple comparative statement were 5.17 times higher for the
dialogue group than for the essay group, and the odds of
making an integrative comparative statement were 3.31
times higher for the dialogue group than for the essay
group. Table 3 summarizes the modeling results for each
Furthermore, the essay group was less likely than the dia-
logue group to produce any comparative statements: Four
(of 30) essay-group participants and 17 (of 30) dialogue-
group participants produced simple comparative statements,
and this difference was significant, χ2(1, N = 60) = 12.38, p <
.001, φ = .45. Similarly, 9 (of 30) essay-group participants
included at least one integrative comparative statement in
Table 2. Coding Categories and Illustrations of Statements Appearing in Participants’ Constructed Dialogues and Essays
Category Example statement
Unsubstantiated statement Cruz is better and that’s final.
Positive single evaluation: single piece of evidence to
support the arguer’s positiona
Cruz will impose a teen curfew to reduce teen crime.
Critical single evaluation: single piece of evidence to
criticize the other position
Diaz has no plan for improving employment.
Integration, arguer’s position:a integrative use of evidence
across multiple dimensions to support the arguer’s position
Diaz promises to revise the high school curriculum, which will
help the school dropout and even teen crime.
Integration, other position: integrative use of evidence
across multiple dimensions to criticize the other position
It doesn’t make sense to expand city parks to give teens a place
to hang out and then impose a curfew.
Simple comparison: comparison of the two candidates on
Diaz is improving public transportation. That resolves the traffic;
Cruz says nothing about transportation in her plan.
Integrative comparison: integrative use of evidence across
multiple dimensions to compare the positions
Reducing rent isn’t going to change anything. Diaz promises
to build more housing. Diaz will also build a new athletic
stadium. Together [this] will create jobs and reduce rent
aIn the case of a constructed dialogue, the arguer making the statement was either Chuck or Doug. In the case of an essay, the arguer was the
participant writing the essay.
Table 3. Results of the Multinomial Logistic Regression Examining the Content
Differences Between the Dialogues and Essays
Content category and
predictor Coefficient SE t(59) pOR
Positive single evaluation
Intercept 0.837 0.293 2.86 .006
Dialogue group –0.424 0.396 –1.07 .289 0.65
Critical single evaluation
Intercept –0.898 0.406 –2.21 .031
Dialogue group 0.198 0.539 0.37 .715 1.22
Integration, arguer’s position
Intercept –0.274 0.343 –0.80 .428
Dialogue group –0.128 0.491 –0.26 .796 0.88
Integration, other position
Intercept –1.814 0.562 –3.23 .002
Dialogue group 0.464 0.732 0.63 .529 1.59
Intercept –1.814 0.580 –3.13 .003
Dialogue group 1.642 0.677 2.43 .018 5.17
Intercept –1.121 0.376 –2.98 .004
Dialogue group 1.198 0.470 2.54 .014 3.31
Random intercept variance 0.924 0.584 1.58 .119
Note: OR = odds ratio.
582 Zavala, Kuhn
Table 4. Results of the Multinomial Logistic Regression Examining the TV Scripts’
Attribute and predictor Coefficient SE t(58) pOR
References to an action
Intercept 0.845 0.490 1.72 .090
Dialogue group 3.249 0.973 3.34 .002 25.76
References to a problem
Intercept 0.236 0.549 0.43 .669
Dialogue group 3.316 0.963 3.44 .001 27.55
Statements linking a
problem and action
Intercept 0.209 0.544 0.38 .703
Dialogue group 3.173 0.977 3.25 .002 23.88
Intercept –0.736 0.575 –1.28 .206
Dialogue group 3.425 1.026 3.34 .002 30.72
Intercept –1.295 0.720 –1.80 .077
Dialogue group 3.684 1.113 3.31 .002 39.81
Random intercept variance 4.594 2.778 1.65 .104
Note: OR = odds ratio.
their essays, whereas 25 (of 30) dialogue-group partici-
pants included at least one integrative comparative state-
ment in their dialogues. This difference was also significant,
χ2(1, N = 60) = 17.38, p < .001, φ = .54.
Differences between the TV scripts of
the two groups
The TV scripts were also segmented into idea units.
Although the dialogue group produced slightly more
idea units in their TV scripts (M = 3.17, SD = 1.97) than
the essay group did (M = 2.93, SD = 1.80), this difference
was not significant, t(58) = −0.48, p = .633, d = 0.13. We
next asked whether there was evidence that the content
of the TV scripts differed between the two groups, and in
particular, whether there was evidence of the hypothe-
sized richer representation of the two positions on the
part of the dialogue group. To answer this question, we
conducted a multinomial logistic regression on the num-
ber of references to one of the stated city problems, the
number of references to a candidate’s proposed actions,
the number of statements linking a problem and an
action addressing it, the number of critical statements
regarding a candidate’s position, and the number of com-
parisons of the two candidates. In these analyses, unsub-
stantiated statements were set as the reference category.
Table 4 summarizes the modeling results.
In addition to referring to fewer problems (M = 1.23,
SD = 1.43, vs. M = 2.07, SD = 1.76) and fewer actions
addressing them (M = 2.27, SD = 1.86, vs. M = 3.53, SD =
2.47), students in the essay group were more likely than
those in the dialogue group to make claims unsubstanti-
ated by any form of support. Sixty percent of the students
in the essay group, but only 20% of those in the dialogue
group, made at least one unsubstantiated claim, χ2(1, N =
60) = 10.00, p = .002, φ = −.41. The mean number of
unsubstantiated statements was 1.07 (SD = 1.20) in the
essay group, compared with 0.20 (SD = 0.41) in the dia-
logue group. Although the majority of the students in
both groups made reference to at least some problems
and some proposed actions, and a majority of the stu-
dents in the dialogue group made links between them
and made comparisons across candidates, only half of
the students in the essay group ever referred to a link
between a problem and action to address it, and fewer
than half ever made a comparison across candidates.
The odds of referring to a problem were 27.55 times
higher for the dialogue group compared with the essay
group, and the odds of referring to a candidate’s pro-
posed action were 25.76 times higher for the dialogue
group. The odds of making a link between a problem
and an action addressing it were 23.88 times higher for
the dialogue group. The odds of noting a negative attri-
bute of one of the candidate’s positions were 39.81 times
higher for the dialogue group, and the odds of compar-
ing the two candidates were 30.72 times higher for the
dialogue group. (See Table 5 for examples of scripts writ-
ten by students in the two groups.)
Solitary Discourse 583
Group differences in epistemological
Responses to the Livia problem were coded on 22 dimen-
sions (Leadbeater & Kuhn, 1989; Weinstock & Cronin,
2003) in order to assign each participant a level of epis-
temological understanding. Table 6 characterizes the six
levels of this scheme in terms of 3 major dimensions: the
nature of the accounts, why they differ, and how claims
are justified. Levels 0 through 2 are regarded as predomi-
nantly absolutist, Level 3 as multiplist, and Levels 4 and 5
as evaluativist. A participant was assigned a level from 0
to 5 on each of the 22 dimensions, and then a dominant
level—absolutist, multiplist, or evaluativist—was assigned.
Concurrent construct validity of this coding scheme
was established by Kuhn, Cheney, and Weinstock (2000),
in a study in which participants assigned to one of the
three main epistemological levels using a different instru-
ment were independently assigned to the same episte-
mological level using their responses to the Livia problem.
To establish interrater reliability in the present study, two
researchers independently coded 20% of the data. The
percentage agreement was 83% across the 22 individual
dimensions and 96% for the overall level assigned. All
disagreements were resolved through discussion.
Figure 1 shows the percentages of participants assigned
to each level as their predominant epistemological level.
Eight participants from the essay group were categorized
as absolutist, 9 as multiplist, and 13 as evaluativist. Zero
participants from the dialogue group were categorized as
absolutist, 12 as multiplist, and 18 as evaluativist. A 2
(group) × 3 (epistemological level) Fisher exact test
revealed that epistemological levels differed significantly
by group (p = .007).
In the present work we extended the claim that dialogue
has beneficial effects to a context in which dialogue was
only hypothetical and constructed by a single individual.
We begin with discussion of effects of the dialogue-
construction task on understanding and then turn to
A mature level of epistemological understanding
provides the foundation for serious discourse (Greene,
Sandoval, & Braten, 2016; Kuhn et al., 2013; Moshman,
2015). If knowledge consists of claims not open to
question—either facts that can simply be “looked up” or
opinions that must be accepted as the unquestioned
possessions of their holders (the stances reflected in less
mature epistemological positions)—discourse serves little
purpose. Nonetheless, an adult who remains largely at an
absolutist level of epistemological understanding does
not conceive of the world in a way identical to that of
the 8-year-old absolutist, who understands all knowledge
as matters of certain fact ascertainable from direct
Table 5. Examples of TV Scripts of Students in the Essay and Dialogue Groups
Essay group Dialogue group
I would select Cruz as the better candidate as mayor. She is
reliable, trustworthy, organized, well thought out, determined
and bright. She fulfills all the characteristics that a mayor
should have. Cruz can fix all of the city’s concerns in due time.
Cruz is knowledgeable and would be a great asset to this city
Cruz will be the better candidate for this city. Instead of
building more housing [Diaz’s proposal], Cruz wants to
reduce housing cost. . . . Instead of revising the high
school curriculum, raising teachers’ pay would boost
instructors’ form of teaching because they are now getting
paid for it. Expanding city parks is a far better form of
recreation than a new stadium. Overall, Cruz has a better
proposal for mayor of this city.
I would choose Cruz for mayor because she seems to have
direct solutions for the city’s troubling concerns. It seems like
she is really for the people and wants to see the city grow in a
I think Cruz should be the mayor of the city simply because
she came up with more ways to fix most of the city’s
issues. Because Diaz left out how she will confront teen
crime, one of the most important problems they are facing
as a community. Showing that her priorities are elsewhere.
I would choose Diaz because she focuses on the wider
perspective of the community and not only on the major
problem, but from a stance where everyone can benefit from.
Diaz wants better for the community and people. The choices
that she makes are exceptional. She cares for the people’s
lifestyle. This is a plan that will work for the people of this
I personally support Cruz and think she is the far better
candidate to be our next mayor. Her plans for improving
the city are well thought out and specifically address the
needs of everyone who lives here. Her plans to increase
teachers’ pay and employ seniors in public schools
will provide a larger and more satisfied staff in all of
our schools. . . . Cruz also plans to create job training
programs which will assist anyone looking for a job as
well as those teens coming out of high school and needing
a job. Opening walk-in health clinics will help those in
need receive the care and attention they need. Diaz’s plan
to simply improve healthcare is vague. Improve how? . . .
584 Zavala, Kuhn
observation or authoritative sources. Adult absolutists
have become aware that disagreement is commonplace
and not always easily resolvable. The concept of a rigid
stage progression in epistemological understanding has
given way to one in which adults, certainly, if not chil-
dren and adolescents as well, hold a loosely connected
network of epistemological ideas that span adjacent lev-
els (Hammer & Elby, 2003). As a result, a more useful
assessment scheme may be one that categorizes individu-
als with respect to the degree to which their own set of
epistemological ideas conforms to different epistemologi-
cal levels (Barzilai & Weinstock, 2015).
If so, it is reasonable to suppose that an individual’s
context or experience prominent at a given point in time
may influence the epistemological ideas that the individ-
ual endorses at that point in time. This is the model that
we believe applies to our data on young adults’ episte-
mological thinking immediately after they have con-
structed an adversarial dialogue, as opposed to after they
have written an essay. We did not assess epistemological
understanding prior to this activity and hence, of course,
cannot say with certainty that the absence of individuals
classified as absolutists in our dialogue group reflected a
change away from absolutism that some would have
exhibited prior to engaging in the dialogue task. Nor can
we say that such a change, if it did occur, would be per-
manent. Rather, our interpretation is that performing the
dialogue-construction task heightened awareness that
conflicting claims with reasonable support can exist and
that a straightforward resolution may not be apparent.
Consistent with this view is other recent evidence that
individuals may show a shift away from an absolutist
position when they are “primed” by an experience that
highlights multiple perspectives (Fisher, Knobe, Strikland,
& Keil, 2016; Kienhues, Stadtler, & Bromme, 2011).
Another recent study provides evidence that dialogue
can have beneficial cognitive effects even though it is not
participated in directly and is only observed. Chi, Kang,
and Yaghmourian (2016) reported a parallel effect when
Table 6. Characteristics of the Six Levels of Epistemological Understanding
Level Nature of accounts Why accounts differ How claims are justified
0. Realist (absolutist) Both accounts mirror reality
Difference is not recognized No justification is needed; the
accounts are interpreted as
accurately reporting what
1. Predualist (absolutist) Each account is objective, but
limited by each historian’s
lack of access to all the facts
Differences are attributed
to incomplete, but not
The accounts are combined to
produce the whole story
2. Dualist (absolutist) One account is objectively
correct, and the other is
distorted because of error
The accounts contradict each
other; one is right, and the
other is wrong
Judgment is made as to which
account is correct
3. Multiplist The accounts are subjective
makes the accounts wholly
The assertions are personal
opinions without further
4. Objective relativist
The accounts are constructed
on the basis of evidence,
with the aim of advancing
Discrepancies are attributed to
different emphases on events
The claims are evaluated in
relation to expert knowledge
5. Conceptual relativist
The accounts refer to
evidence but are framed in
Subjective context and
perspective determine the
interpretation of evidence
The claims are evaluated
in relation to contexts of
evidence, events, and the
person making the claim
Absolutist Multiplist Evaluativist
Level of Epistemological Understanding
Fig. 1. Percentage of participants in each group who were assigned to
each epistemological level.
Solitary Discourse 585
participants watched a video that was dialogic. A video
of a tutee struggling to correct misconceptions while
interacting with a tutor, Chi et al. found, was more con-
ducive to learning than was a video of the tutor alone.
In the present study, we hypothesize, constructing a
dialogue required the writer to repeatedly shift back and
forth between one perspective and the other, generating
credible arguments to support each, as well as counterar-
guments to weaken them, and rebuttals aimed at restor-
ing their strength. As a result, the writer formed a richer
representation, not only of each position and the evi-
dence bearing on it, but also of the positions in relation
to one another.
These results have both theoretical and practical impli-
cations. Theoretically, they support the close link between
interpersonal and intrapersonal forms of thought, and
practically, they support the claim that one can serve as a
bridge to the other. Close analysis of middle-school stu-
dents’ essays as they engaged deeply with a series of
topics over 2 years (Kuhn, Hemberger, & Khait, 2016a,
2016b) showed that the dialogic structure of their argu-
mentation with peers made its way into their essays.
“Others might say . . . ,” a phrase rooted in the real-life
other that dialogue provides, appeared increasingly often
in their essays. Dialogue makes an opposing position and
its accompanying arguments clear and vivid enough for a
student to represent them in an essay and address them,
but at least as important is gaining recognition of the
relevance of doing so.
Role taking is a powerful mechanism that has the
potential to substitute for actual social exchange. The
present results suggest that it can be productive even in
virtual, imagined form. One might think of imagined dia-
logue as an intermediate process between actual social
interaction and passive observation of other peoples’
interaction, which research shows may be ineffective if
the observer does not act on these observations in some
way (Jewett & Kuhn, 2016; Muldner, Lam, & Chi, 2014).
We, of course, cannot rule out additional, noncognitive
differences between the two conditions in this study. In
particular, the conditions might have differed in motiva-
tion and engagement to the extent that participants found
the dialogue task novel and hence more engaging than
the essay task. Engagement, however, is part of what is
thought to make dialogue effective, and hence its poten-
tial contribution to the cognitive outcomes we have iden-
tified need not be excluded; it would nonetheless be
desirable to further elucidate this contribution.
Although there now exists a good deal of evidence for
the positive effects of peer interaction on intellectual
development (Resnick, Asterhan, & Clarke, 2015), such
interaction may not always be feasible, and in school set-
tings especially, teachers may be disinclined to incorporate
interactive activities into their lessons. We do not mean to
suggest that the dialogue-construction activity we exam-
ined is a satisfactory substitute, or that all possible should
not be done to encourage and support authentic class-
room discourse. The virtual form of interaction we exam-
ined here may be a less desirable but still productive
substitute. In an era in which positions on an issue too
often lack the deep analysis to support them, this may be
Gretchen B. Chapman served as action editor for this article.
The authors jointly conceived the study and wrote the manu-
script. J. Zavala conducted the data collection and analysis.
We thank Bryan Keller for providing his statistical expertise.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Barzilai, S., & Weinstock, M. (2015). Measuring epistemic think-
ing within and across topics: A scenario-based approach.
Contemporary Educational Psychology, 42, 141–158.
Billig, M. (1987). Arguing and thinking: A rhetorical approach
to social psychology. Cambridge, England: Cambridge
Chi, M. T. H., Kang, S., & Yaghmourian, D. L. (2016). Why stu-
dents learn more from dialogue- than monologue- videos:
Analyses of peer interactions. Journal of the Learning
Sciences. Advance online publication. doi:10.1080/10508406.
Common Core State Standards Initiative. (2010). Read the stan-
dards. Retrieved from http://www.corestandards.org/read-
Fisher, M., Knobe, J., Strikland, B., & Keil, F. (2016). The influ-
ence of social interaction on intuitions of objectivity and
subjectivity. Cognitive Science. Advance online publication.
Graff, G. (2003). Clueless in academe: How schooling obscures
the life of the mind. New Haven, CT: Yale University Press.
Graham, S., McKeown, D., Kiuhara, S., & Harris, K. (2012). A
meta-analysis of writing instruction for students in the ele-
mentary grades. Journal of Educational Psychology, 104,
Greene, J., Sandoval, W., & Braten, I. (Eds.). (2016). Handbook
of epistemic cognition. New York, NY: Routledge.
Grice, H. P. (1975). Logic and conversation. In D. Davidson
& G. Harman (Eds.), The logic of grammar (pp. 64–75).
Encino, CA: Dickenson.
586 Zavala, Kuhn
Hammer, D., & Elby, A. (2003). Tapping epistemological
resources for learning physics. Journal of the Learning
Sciences, 12, 53–90.
Jewett, E., & Kuhn, D. (2016). Social science as a tool in
developing scientific thinking skills in underserved, low-
achieving urban students. Journal of Experimental Child
Psychology, 143, 154–161.
Kienhues, D., Stadtler, M., & Bromme, R. (2011). Dealing with
conflicting or consistent medical information on the Web:
When expert information breeds laypersons’ doubts about
experts. Learning and Instruction, 21, 193–204.
Kuhn, D., Cheney, R., & Weinstock, M. (2000). The development
of epistemological understanding. Cognitive Development,
Kuhn, D., & Crowell, A. (2011). Dialogic argumentation as
a vehicle for developing young adolescents’ thinking.
Psychological Science, 22, 545–552.
Kuhn, D., Hemberger, L., & Khait, V. (2016a). Argue with me:
Argument as a path to developing students’ thinking and
writing (2nd ed.). New York, NY: Routledge.
Kuhn, D., Hemberger, L., & Khait, V. (2016b). Tracing the
development of argumentive writing in a discourse-rich
context. Written Communication, 33, 92–121.
Kuhn, D., & Moore, W. (2015). Argument as core curriculum.
Learning: Research and Practice, 1, 66–78.
Kuhn, D., Pennington, N., & Leadbeater, B. (1983). Adult think-
ing in developmental perspective. In P. Baltes & O. Brim
(Eds.), Lifespan development and behavior (pp. 157–195).
New York, NY: Academic Press.
Kuhn, D., Zillmer, N., Crowell, A., & Zavala, J. (2013).
Developing norms of argumentation: Metacognitive, episte-
mological, and social dimensions of developing argumen-
tive competence. Cognition and Instruction, 31, 456–496.
Leadbeater, B., & Kuhn, D. (1989). Interpreting discrepant nar-
ratives: Hermeneutics and adult cognition. In J. Sinnott
(Ed.), Everyday problem solving (pp. 175–190). New York,
Macagno, F. (2016). Argument relevance and structure:
Assessing and developing students’ uses of evidence. Inter-
national Journal of Educational Research, 79, 180–194.
Mead, G. H. (1967). Mind, self, and society. Chicago, IL:
University of Chicago Press. (Original work published 1934)
Moshman, D. (2015). Epistemic cognition and development:
The psychology of justification and truth. New York, NY:
Muldner, K., Lam, R., & Chi, M. (2014). Comparing learn-
ing from observing and from human tutoring. Journal of
Educational Psychology, 106, 69–85.
NGSS Lead States. (2013). Next Generation Science Standards:
For states, by states. Retrieved from http://www.nextgen
Resnick, L., Asterhan, C., & Clarke, S. (Eds.). (2015). Social-
izing intelligence through academic talk and dialogue.
Washington, DC: American Educational Research Association.
Sandoval, W. (2014). Science education’s need for a theory of
epistemological development. Science Education, 98, 383–
van Eemeren, F., & Grootendorst, R. (1992). Argumentation,
communication and fallacies. Hillsdale, NJ: Erlbaum.
Vygotsky, L. (1987). The collected works of L.S. Vygotsky:
Pro blems of general psychology (R. Rieber & A. Carlton,
Eds.). New York, NY: Plenum. (Original work published
Walton, D. (2014). Dialog theory for critical argumentation.
Amsterdam, The Netherlands: John Benjamins.
Weinstock, M., & Cronin, M. (2003). The everyday production
of knowledge: Individual differences in epistemological
understanding and juror-reasoning skill. Applied Cognitive
Psychology, 17, 161–181.