28

Dialogic Instruction
A New Frontier
S HERICE N. C LARKE
L AUREN B. R ESNICK
University of Pittsburgh

C AROLYN P ENSTEIN R OSÉ

This chapter examines the role of dialogue in classroom that individual cognition is shaped through social interac-
learning. Classroom language has long been a focus of tion, and verbal dialogue plays a special role in this process
educational inquiry. The earliest investigations sought to (Dewey, 1916; Mead, 1967; Vygotsky, 1962, 1978; Wertsch,
understand the underlying logic of teacher–student interac- 1979). These ideas advanced the notion that dialogue is a
tions in the classroom (Bellack, Kliebard, Hyman, & Smith, form of human learning.
1966). They showed that verbal exchanges between teach- As these ideas began to penetrate adjacent disciplines
ers and students constitute much of classroom activity, like sociology and linguistics, and their newly formed
and that most of these exchanges take the form of teacher subfields of ethnomethodology (Garfinkel, 1967) and
questions or comments followed by student answers (Gall, sociolinguistics (Hymes, 1977), theories of language were
1970). evolving. Sociologists and linguists were seeking theory
Interest in improving instruction led to a growth in study- that could account for the “social” and “cultural” nature
ing of the quality of teacher questions in terms of the of language and its production (Bourdieu, 1977; Hymes,
cognitive processes they elicited from students, and their 1977). New areas of scholarship in these fields, together
effects on student achievement (Bloom & Krathwohl, with new technologies of the time (i.e., video recording),
1956; Gall, 1970; Winne, 1979). Bloom’s taxonomy (and enabled indepth ethnographic studies of classroom activ-
similar frameworks) provided a way to rank the kinds of ity and language not previously possible. These studies
questions teachers ask in terms of the cognitive demands yielded insights about classroom life similar to those that
required to answer them. At the same time, work on the educational psychologists had been documenting—class-
use of technology to support instruction drew upon these room activity consisted primarily of dialogue in the form
frameworks to build the first approaches to automating of questions and answers. Taken together, these investiga-
instruction through dialogue (Carbonell, 1969, 1970; tions also revealed how most verbal interactions in class-
Stevens & Collins, 1977). rooms were highly constrained, consisting primarily of
The hypothesis at this time was that questions that elicited recitation (a teacher’s question followed by a solicitation
higher-order thinking processes would be associated with of a student’s response, followed by the teacher’s evalu-
greater student achievement. However, empirical studies ation of the student’s response) (Mehan, 1979; Sinclair,
found little evidence to support the hypothesized correlation Coulthard, & Council, 1975).
between higher-order questions and student achievement Over the last few decades a new form of instructional
(Winne, 1979). pedagogy has emerged (Alexander, 2008; Littleton &
Around the same time, the ideas of Vygotsky and col- Howe, 2010; Resnick, Asterhan, & Clarke, 2015; Resnick,
leagues were being translated and making their way West. Michaels, & O’Connor, 2010). We have come to refer to
Blending these ideas with the ideas of Dewey, Mead, Bakhtin instructional dialogue that engages discussants in inter-
and (to some extent) Piaget, some psychological theorists mental reasoning processes to support the development of
began to offer a theory of mind and development that posited intra-mental reasoning processes as dialogic instruction.

378
 Dialogic Instruction 379

Dialogic instruction is grounded in a set of domain-specific these findings as the “big effects” of dialogic instruction. In
ideas and concepts, e.g., science, mathematics, history, lit- assembling this data, we count the following as evidence of
erature, that schools teach. Within these domains, students learning:
explain, reflect upon, and elaborate on their own and their
peers’ understandings of domain concepts. They engage a. improved initial learning, in comparison with a control or ref-
in reasoned argumentation about subject matter: support- erence group, on assessments of academic skills and knowledge
ing claims with evidence, drawing logically grounded and that are developed and administered by an authority external to
justified conclusions, and negotiating disputes through the school or classroom whose effectiveness is under study (these
could be state or national examinations, or commercial standard-
argumentation. This form of dialogue has an implicit social
ized tests);
contract; discussants endeavor to collaboratively engage b. evidence of retention of academic skills and knowledge, again
in the sense-making process. This requires discussants to in comparison with a control or reference group, across several
participate in verbal activities that externalize their emer- months or years;
gent thinking, reason aloud about solutions offered, chal- c. evidence of transfer to domains of academic skill and knowledge
lenge one another’s views, and clarify misunderstandings. other than the one taught in the study;
It requires discussants to listen to, hear, and respond to their d. evidence of improvement in widely accepted forms of reasoning
peers’ contributions. or general intelligence.
Unlike recitation, the primary role of the teacher (or
computer-mediated surrogate) is to scaffold and elicit stu- In the following sections, we will discuss five cases of instructional
dents’ verbal reasoning. A wide range of participation struc- interventions that produced one or more of the aforementioned big
tures found in classroom learning are qualitatively dialogic effects on student learning.
(e.g., cooperative learning, collaborative learning, argumen-
Initial Learning
tation, reciprocal teaching, group problem solving, peer tutor-
ing, computer-supported collaborative learning). Dialogic Several studies conducted in the 1980s experimented with
instruction can be human-mediated (e.g., by a teacher or discussion-rich approaches to mathematics instruction.
facilitator) or computer-mediated (e.g., facilitated by a com- These studies show that when teachers lead students in prob-
puter agent serving as surrogate teacher) (Dyke, Adamson, lem-solving discussions that elicit collaborative reasoning,
Howley, & Rosé, 2013), in face-to-face settings or computer- students greatly outperform non-treated comparison groups
mediated settings (Stahl, Koschmann, & Suthers, 2006). on standardized tests.
Until quite recently it has appeared that direct instruction The first case, Math3, was an intervention developed by
by an instructor (or computer-based tutor) was the only relia- a master teacher and a group of scholars to teach math in
ble way to ensure high levels of learning in core school sub- an elementary school serving ethnic-minority children from
ject matters (Klahr & Nigam, 2004). But in the past couple of low-income families (Bill, Leer, Reams, & Resnick, 1992;
decades there has been growing evidence that dialogue-rich Resnick, Bill, Lesgold, & Leer, 1991). The Math3 program
instruction is worth the extra time and effort and may be able consisted of math “story problems” that engaged students in
to produce large, measurable learning gains, including reten- a process of shared mathematical reasoning that would foster
tion and transfer. fluency in complex mathematical thinking. Lessons entailed
The next sections of the chapter bring together the body whole-class discussions, followed by teamwork and more
of evidence on human- and computer-supported dialogic discussion as each team reported out. The discussions were
instruction. We consider the best evidence on learning designed to prompt students to listen to each other’s ideas
through dialogue in terms of traditional measures of school and reflect on their own and others’ reasoning, and encour-
learning, which we refer to here as academic outcomes. We age students to explicate their thinking.
then discuss research that helps to explicate some of the Bill and colleagues conducted a 2-year study of the
mechanisms though which dialogic instruction supports Math3 program. Bill taught two cohorts of students Math3 in
growth in academic outcomes. We conclude by considering almost every math class during the 2-year period. Cohort A
future directions for research and inquiry. started Math3 in the first grade; Cohort B, in second grade;
and a Control Cohort consisted of Bill’s students prior to the
intervention as a comparison group (Resnick et al., 1991).
“Big Effects” of Dialogue Growth in academic outcomes was measured using existing
While dialogic instruction is a growing area of inquiry, a sur- assessments of student competence in math and reading, in
prisingly small number of studies examine the relationship addition to measures of the dialogic processes in the lessons.
between participation in dialogue and learning outcomes. The school-administered California Achievement Test was
This small set of studies collectively provide robust evidence used to measure student growth in academic outcomes from
of the effects of dialogic instruction on valued academic the Math3 intervention.
outcomes. When considered as a whole, the evidence sug- Dramatic changes in students’ computational ability were
gests that, when teachers lead discussions of this kind, they documented. On the California Achievement Test, Cohorts
can produce steep increases in student learning. We refer to A and B performed near the top of the scale in mathematics
380 Sherice N. Clarke et al.

after Bill’s dialogic-style teaching. However, the Control through the restructuring process, helping them to reason sci-
Cohort who did not receive the story-based lessons had rel- entifically about problems. In these interventions, dialogue is
atively low achievement scores across the 3-year period of the pedagogical tool used to support students’ problem solv-
measurement. ing and scientific reasoning.
The second case of the learning effects of dialogic instruc- The study was conducted in schools serving a range of
tion is Project Challenge, a similar mathematics program populations, with two to three classes assigned to the CASE
to Math3. Project Challenge was developed by a master intervention, and two to three matched classes assigned to
teacher and scholars at Boston University for one of the low- serve as non-treated controls. Piagetian reasoning tasks were
est-performing school districts in Massachusetts. The team administered as pretests and posttests to examine cognitive
developed an intervention that would provide students with development. School science assessments were used as
complex problems, projects, and arithmetic learning through delayed posttests of science achievement, 1 and 2 years after
games (Chapin & O’Connor, 2004). the intervention. In addition, results of students’ performance
Project Challenge lessons were taught 5 days a week and on the science, math, and English portions of the national
entailed discussions of the kind in Math3, focusing on stu- General Certificate of Secondary Education (GCSE) were
dents’ explanation of mathematical thinking. The interven- examined as a delayed posttest of achievement 2 and 3 years
tion was given to four cohorts of 25 students each, beginning after CASE.
in the fourth grade and lasting 4 years.1 The Massachusetts Immediate posttests showed that CASE cohorts had no
Comprehensive Assessment System (MCAS) state assess- significant gains in science understanding over control stu-
ment was administered to measure learning outcomes of the dents. However, delayed posttests and GCSE scores showed
intervention. considerable gains in science achievement of CASE students.
After 1 year, 57% of Project Challenge students scored The most significant results were concentrated amongst two
“Advanced” or “Proficient” on the MCAS, compared with CASE cohorts. The first cohort showed modest gains on
38% in Massachusetts overall. After 3 years in Project the school-based science assessment taken 1 year after the
Challenge, 82% of students scored in the “Advanced” or CASE intervention. However, they made considerable gains
“Proficient” ranges, compared to only 40% for Massachusetts over and above the delayed posttest on the science portion of
sixth-graders overall. Project Challenge results indicate that the GCSEs taken 2 years later, scoring on average one grade
students’ competencies did not simply grow during their higher than control students. The second cohort had a similar
years in the program, but grew exponentially over the 3-year pattern. They made significant gains on the delayed posttest
period. in science taken 2 years after CASE, and retained these gains
The evidence from Math3 and Project Challenge suggests in science on the GCSEs 3 years later. The evidence from
that, when teachers use dialogic approaches in mathemat- the CASE program indicates that one could not only teach
ics instruction that engages students in a process of shared the core “Piagetian schemata” (which constitute basic sche-
mathematical reasoning, students’ general competencies in mata for scientific and mathematical reasoning), but doing so
math increase dramatically. The effects of discussion seem can have long-term results, including better performance on
to transcend the idiosyncrasies of particular discussions and state-administered exams several years later.
discussion topics. Our fourth case, a body of instruction gathered under the
name of Philosophy for Children (P4C), also documents
long-term retention effects (Lipman, 1981). The goal of
Long-term Retention
the P4C program is to develop students’ critical reasoning
A “tough” standard of effectiveness for any form of instruc- skills through collaborative philosophical inquiry (Topping
tion would call for evidence that what students learned had & Trickey, 2007a, 2007b). Using a text or another stimulus,
“staying power”—that it lasted longer than an academic a teacher guides students through dialogue and questioning
year. Perhaps the most extensive demonstration of long- in which multiple answers are possible. Like CASE, P4C
term learning effects comes from the Cognitive Acceleration encourages cognitive conflict, involves learning processes
through Science Education (CASE) program (Adey & that promote metacognitive development, and relies on the
Shayer, 1990, 1993). CASE had three core principles: social construction of knowledge.
(a) inducing cognitive conflict; (b) facilitating social construc- Topping and Trickey studied a variant of this program
tion; and (c) supporting the development of metacognition. that was implemented in a school district in Scotland. They
The CASE intervention was based on Piaget’s schemata investigated whether the program of philosophical inquiry in
of formal operations e.g., control and exclusion of varia- the classroom had a measurable effect on children’s cognitive
bles, equilibrium, ratio, and proportionality. Students (aged reasoning ability, and whether learning would be retained. In
11–14 years) were taught 30 lessons on practical problems weekly 1-hour lessons, the teachers’ aim was for the class
in science class every other week for 2 years. Lessons were to construct deeper knowledge or a better solution than any
designed to induce cognitive conflict so as to disrupt existing one child could develop alone. Six classrooms participated
knowledge structures and make them more receptive to con- in the P4C study, four as part of the intervention and two as
ceptual change. Through discussing the problems and work- controls. The control classes were presented with the same
ing to solve them, it was believed that students would be able initial story, followed by a discussion, but did not experi-
to construct schemata themselves. Teachers guide students ence weekly inquiry lessons. The Cognitive Abilities Test
 Dialogic Instruction 381

was administered before the intervention, and again after 16 on assessments in other domains (Greeno, Moore, & Smith,
months and 43 months. 1993; Koedinger, Corbett, & Perfetti, 2012). In other words,
The experimental group showed significant gains between their conceptualization of transfer would suggest that stu-
the pretest and the first posttest, in verbal, non-verbal, and dents develop general capacities through dialogic instruction
quantitative reasoning abilities, while the control group in one domain, and these capacities transfers when students
showed none. And despite no further intervention, the exper- are assessed in another domain. Nevertheless, what is par-
imental group maintained these gains for 2 more years, while ticularly notable is that these studies are amongst the few
the control group’s scores decreased over the same period. studies that provide evidence that educational practices can
Together these studies document long-term retention actually promote domain transfer.
effects for up to 3 years from engaging in joint sense-making
discussions in science and literature-cum-philosophy. These
are sizable effects of the “staying power” of learning that Reasoning Development
occurs through discussions. The findings suggest that the Our final criterion for big effects of dialogic instruction is
process of engaging in these kinds of discussions promotes the development of reasoning ability. To date, there have
robust conceptual transformation. been just a couple of studies of dialogic instruction that have
documented these effects: P4C and Talk, Reasoning and
Computers (TRAC).
Transfer
As an outcome measure of the P4C program, Topping and
Another “tough” standard of effectiveness of any instruction Trickey administered the Cognitive Abilities Test, which
would call for evidence that what students learned spread to measures the development of non-verbal and quantitative
competencies and subject matter not directly taught. Three reasoning ability. Their findings show that students in treated
of our cases document evidence of domain transfer effects: conditions who were engaging in philosophical discussion
namely, Math3, Project Challenge, and CASE interventions. of questions such as “what is truth?” increased both their
In the Math3 program, Bill and colleagues found that, non-verbal and quantitative reasoning (Topping & Trickey,
in addition to gains that intervention cohorts had in math 2007a). Students in the control classes decreased in their
achievement on the California Achievement Test, their non-verbal and quantitative reasoning ability.
scores on the reading portion of the exam not only rose, but The fifth case is TRAC. Through TRAC, Mercer and
continued to rise when they were assessed after the first and colleagues sought to examine whether reasoning could be
second year of the intervention (Bill et al., 1992; Resnick directly taught through collaborative computer-based activ-
et al., 1991). What is notable is that Bill did not teach read- ities (Mercer, Wegerif, & Dawes, 1999; Wegerif, Mercer,
ing, and the intervention involved little written work. & Dawes, 1999). The aforementioned cases have been pri-
Project Challenge produced similar results. When scores marily concerned with teaching domain concepts through
on the English portion of the MCAS were examined, they dialogue, and studying the development of domain-specific
found that Project Challenge students had steep increases in outcomes. TRAC, however, sought to examine the lan-
literacy in addition to math, over and above the performance guage of collaborative reasoning itself. In doing so, Mercer
of the students in Massachusetts receiving regular, non- and colleagues posited that if students were taught the lan-
dialogic instruction (Chapin & O’Connor, 2004). Both these guage norms of effective collaborative reasoning, what the
interventions show evidence of transfer from math instruc- researchers referred to as “exploratory talk,” then reasoning
tion to reading. Adey and Shayer found transfer results from itself could be developed.
dialogic-style science instruction in CASE when they com- In the TRAC program teachers led students through les-
pared intervention students’ results with non-treated control sons designed to raise students’ awareness of the norms of
classes (Adey & Shayer, 1993). GCSEs were administered “exploratory talk.” The study involved four target classes
2–3 years after the CASE intervention (depending on the of primary Grade 5 children, and four matched classes as
cohort). The results showed that students who participated control groups. The target classes received nine lessons of
in the CASE intervention in science showed considerable TRAC, and the control classes received business-as-usual
gains on the English and math portion of the GCSEs 2–3 instruction. To examine the impact of TRAC on target
years later. Thus, they document not only a sizable long-term groups, the researchers collected two measures of reasoning
retention effect of the program, but also domain transfer as a development using the Ravens Progressive Matrices (RPM),
consequence of collaborative scientific inquiry discussions. a test of non-verbal reasoning ability. The first was an indi-
Together, these studies provide initial evidence that vidual RPM test to measure individual growth in non-verbal
transfer is achievable through dialogic instruction. No the- reasoning, and the second a collaborative version, to test the
ory of the day could account for these findings.2 One-time development of group reasoning. They posited that if there
gains could be chance, but these data show the same transfer was a greater incidence of “exploratory talk” in group discus-
effects from three different programs, with several cohorts sions that resulted in successful solutions to RPM items, then
of students over multiple years. Greeno and others suggest they would have evidence of relationship between language
that the phenomena of domain transfer might be explained and reasoning development.
in terms of the cognitive processes engaged when students At the individual level, the authors found that children
participate in discussions and the cognitive processes elicited in both the treatment and control classes increased their
382 Sherice N. Clarke et al.

performance on the RPM from the pretest to posttest, sug- indepth look at targeted studies of discussion-rich instruction
gesting an increase in individual non-verbal reasoning ability and learning that help to shed light on some of the mecha-
for both groups as an outcome of instruction. At the group nisms that support learning.
level, they examined the difference between group dialogues
of successful and unsuccessful solutions to RPM items. They
Unpacking the Big Effects of Dialogue: Cognitive
confirmed that group dialogues that resulted in successful
and Social
solutions to RPM items included far greater exploratory
talk than unsuccessful solutions, suggesting a relationship What are the processes engaged when an individual articu-
between language and accurate reasoning. They also found lates their thinking in discussion? Are these the same pro-
that groups increased in their amount of exploratory talk from cesses that are engaged when an individual listens to peers
pretreatment group problem solving to posttreatment prob- during discussion? These are some of the questions that
lem solving, suggesting that the intervention supported the researchers have been asking in order to identify the mecha-
development of exploratory talk. Groups in control classes nisms through which talk and dialogue support learning. The
decreased in the number of essential features of exploratory big-effects studies put a compelling set of evidence on the
talk from pretest to posttest. table of the academic outcomes that dialogic instruction can
What Mercer and colleagues were setting out to exam- foster. But precisely how dialogic instruction fosters these
ine was whether cognitive processes could be activated if steep increases in outcomes is not fully explicated. For this
instruction focused solely on teaching the language associ- reason, in this section we consider targeted interventions of
ated with those processes. What is striking about their find- dialogic instruction and learning (and its close cousins) in
ings is that the RPM is a measure of non-verbal reasoning order to identify the processes engaged when individuals
ability. Increases in performance on the RPM provide direct participate in learning dialogues.
evidence of increases in reasoning, and indirect evidence
that collaborative dialogues that include argumentative talk
Talk “Moves” That Support Cognitive Processes
support increases in reasoning ability. Like Topping and
Trickey, they provide evidence of a kind that is impossible To examine the mechanisms through which dialogic instruc-
to ignore: the tethered nature of language and thought. Direct tion supports academic outcomes, researchers have looked
instruction of the linguistic forms of reasoning can support towards the discourse itself—the “processes” of dialogic
the development of reasoning itself. participation. Collaborative sense making in discussions
provides insight into the underlying cognitive work of dis-
cussants. In particular, talk moves, a class of speech act, sig-
What Produces the Big Effects of Dialogic
nify cognitive processes (Resnick et al., 2010; Searle, 1969),
Instruction?
and social positioning in talk (Veel & Christie, 1999). For
These five studies suggest that, when teachers lead students example, elaboration requires that a speaker has heard the
in structured collaborative sense-making discussions about previous utterance, understood it, and made sense of the
subject matter, students not only learn the targeted content, ideas expressed in relation to his own current understanding.
but the learning can be retained for up to 3 years, transferred The elaboration, therefore, is an expression of the difference
to other domains, and verbal and non-verbal reasoning can between a student and his peers’ understanding. Talk moves
develop. These are powerful effects for the educational have both a social function (e.g., listening and responding to
enterprise. But what is it about dialogic instruction that pro- one another) and an intellectual function (e.g., explanation,
duces these big effects? elaboration, reasoning) (Greeno, 2015).
There are a few common features of the aforementioned The use of certain talk moves (e.g., explanation, reason-
studies. First, dialogue-intensive activities are facilitated by ing, disagreement) has been shown to support growth in aca-
teachers who lead students in reasoned discussion within an demic outcomes. We consider the evidence on each of these
academic domain about domain concepts (with the exception moves, including the evidence they provide on both sides of
of Mercer et al., 1999). Second, the dialogue is purposeful. dialogue: production (speaking) and reception (listening).
In other words, the discussions are oriented around tasks While production and reception are dynamic and interchang-
that create authentic reasons to engage in discussion. Third, ing in natural dialogue, some studies have isolated the effects
teachers (or computer-mediated surrogates) elicit students’ in terms of these distinct roles.
verbalizations of their emergent thinking to scaffold the
social and linguistic norms of collaborative reasoning. Adey Explanation. Explanations are the externalization of
and Shayer’s studies suggest a fourth criterion not described knowledge or understanding (emergent or fully formed) in
in the other big-effects studies; i.e., that the discussion task is the form of a verbal utterance. Chi and colleagues’ inves-
one that induces cognitive conflict. These features describe tigations of self-explanation provide some insight into
the general conditions of dialogic instruction, but what is it the production side of explanation (Chi, Bassok, Lewis,
that occurs through verbal interaction about academic sub- Reimann, & Glaser, 1989; Chi, Slotta, & De Leeuw, 1994).
ject matter that lends itself to the robust outcomes found in Self-explanation is the process whereby individuals explain
the big-effects studies? To address this question, we take an to themselves their understanding, e.g., verbally through
 Dialogic Instruction 383

think-aloud, in text, or inner speech (Vygotsky, 1962). Chi et knowledge structures” as a result of explanations may not
al. have found that verbal self-explanations support success be immediate (Asterhan & Schwarz, 2007, 2009; Howe,
in problem solving. When the differences between students McWilliam, & Cross, 2005). When learning is measured
who were more or less successful at problem solving were immediately following collaborative problem-solving activ-
examined, they found that one of the characteristics of suc- ities involving explanation, gains have been moderate.
cessful problem solving was the greater incidence of self- However, if a delayed posttest is given a week or two fol-
explanations, and more robust explanations. Further, they lowing collaborative activities involving explanation, learn-
found that there is a positive effect of good explanations on ing gains for individuals in conditions involving explanation
learning. Accurate explanations yield higher learning gains on exceed their immediate posttests. This suggests that the rad-
measures of knowledge and understanding. ical transformation process that explanation lends itself to
Chi and colleagues (1989, 1994) suggest that there are may take time to be synthesized and integrated. This is what
several cognitive processes at work when individuals artic- Howe has referred to as an “incubation period” (Howe et al.,
ulate their understanding in the form of self-explanations. 2005).
First, explanation drives the explainer to “fill in” missing, or Examining the question of who benefits from explana-
not yet understood, parts of phenomena in order to provide a tions in terms of conceptual change, Asterhan and Schwarz
complete explanation. In other words, through explanation, (2009) have found that this kind of deep conceptual shift is
the individual attempts to construct a coherent picture by limited to individuals giving explanations, rather than those
making inferences, and in doing so, drives the second pro- receiving explanations. Receivers benefit from explanation
cess: integration. Through explanation, the explainer inte- in terms of knowledge acquisition; however, they need to
grates new knowledge with existing knowledge. Integration actively externalize their understanding for a radical trans-
may lend itself to the third process, cognitive conflict, which formation of understanding to occur.
is when an individual becomes aware of misconceptions and
has an opportunity to correct them. Transactive reasoning. Transacts, following Berkowitz
Self-explanation as typically conceptualized, however, is and Gibbs (1983), are talk moves in which the reasoning of
inherently monologic. In what ways are explanations to peers discussants is “interpenetrated.” These are moves in which a
different from self-explanations? Researchers have docu- speaker’s utterance(s) builds on her own or another’s think-
mented the same effects of self-explanation when students ing. To a certain extent, transactivity is the natural order of
explain to peers (Fuchs et al., 1997; Howe et al., 2007; King, truly collaborative dialogue: (a) attentive to one another’s
1992; King & Rosenshine, 1993; Slavin, 1987; Veenman, ideas; and (b) progressive, the reasoning of discussants
Denessen, van den Akker, & van der Rijt, 2005; Webb, 1991; develops. While there has been a lot of interest in transacts
Webb et al., 2009). When students explain domain concepts as an indicator of productive learning dialogues, just a few
to their peers in dyads or small groups, they themselves per- studies have examined the effect of transactive moves in dia-
form better on outcome measures of knowledge and under- logue on learning.
standing. Peer explanation accuracy and sophistication have Generally, researchers have found that the incidence of
also been found to yield higher learning gains for explainers transactive moves during problem-solving discussions pos-
(King & Rosenshine, 1993; Veenman et al., 2005). However, itively predicts problem-solving accuracy and learning of
Webb and colleagues have found that, while both “explain- problem-solving skills (Azmitia & Montgomery, 1993;
ers” and receivers of explanations benefit from explanations Berkowitz & Gibbs, 1983; Chinn, O’Donnell, & Jinks, 2000;
in terms of increasing outcomes, the benefits for receivers Kruger, 1992, 1993; Teasley, 1997). There are two types of
depend on their active cognitive engagement with the expla- transactive exchanges that can occur: elaborating on one’s
nations (Webb & Mastergeorge, 2003; Webb, Troper, & Fall, own reasoning (self-transacts) and elaborating on anoth-
1995). They found that it is necessary for receivers to make er’s reasoning (other-oriented transacts). Both self- and
immediate use of explanations to solve new problems (in the other-oriented transacts have been shown to support dyadic
context of mathematics instruction) in order to achieve gains problem solving and the development of reasoning (Azmitia
on posttest measures. & Montgomery, 1993; Kruger, 1992; Teasley, 1997). If
As mentioned, the explanatory process lends itself to there is a complete absence of engagement with another’s
situations of cognitive conflict, whereby an individual reasoning, however (in other words, excessive use of self-
becomes aware of a disjuncture between her existing and transacts in the context of problem-solving discussions), then
new knowledge and provides an opportunity for integration the dyad’s or group’s performance can suffer (Barron, 2003;
of new knowledge. This transformation process is known Kruger, 1992).
as conceptual change. A growing number of studies have This research suggests that, while elaboration of one’s
been investigating the ways in which dyadic or collaborative own thinking is good for individual performance, overuse
activities involving explanation support conceptual change to the extent that the transactivity is akin to a monologue
(see Chapter 18, this volume). can be harmful for group problem solving. The “sweet
Several studies have documented a positive relationship spot” for collaborative discussions appears to be a combi-
between explanations and conceptual change. However, con- nation of self-oriented and other-oriented transacts as it sup-
ceptual change in the form of a “radical transformation of ports individual cognitive work, with the added benefit of
384 Sherice N. Clarke et al.

the plurality of ideas from peers with which to reason (Chi, Second, the very nature of opposition and disequilib-
2009). Analysis of the social properties of audio data from rium gives rise to elaborations, justifications, and resolu-
collaborative sessions suggests that the balance of self and tions, which are supportive of learning outcomes (Howe &
other transacts reflects the extent to which students are listen- McWilliam, 2006). Howe and colleagues, in particular, have
ing to one another and valuing one another’s respective per- shown that it is not the sheer number of disagreements in
spectives (Gweon, Jain, Mc Donough, Raj, & Rosé, 2013). dialogues that is associated with posttest performance on
Thus, an overrepresentation of self-oriented transacts might problem-solving activities. Rather, it is the incidence of
be symptomatic of missed opportunities to reflect on the explanations, elaborations, and justifications that disagree-
partner student’s expressed reasoning. ment gives rise to (however frequent) that is associated with
A related area of inquiry has been the study of question posttest gains (Howe et al., 2007). Similar to the effects of
asking during collaborative learning. Studies have exam- explanation on conceptual change, Howe and colleagues
ined the degree to which students asked each other ques- have found that disagreement is not only predictive of learn-
tions during collaborative learning and the quality of those ing gains immediately following group collaboration, but
questions in relation to outcomes (King, 1992, 1999; King & also predictive of gains sustained well beyond the interven-
Rosenshine, 1993; Rosenshine, Meister, & Chapman, 1996). tion, as much as 18 months later (Howe, 2010).
Research has found that, when students ask each other ques- Kruger argues that learning gains from disagreements are
tions during collaborative learning discussions, groups are not simply predicated on the social structure of activities,
more effective at problem solving and individual learning e.g., group or dyadic problem solving. Rather, the benefit
outcomes increase (King, 1999; Rosenshine et al., 1996). for learning is located in the act of collaboratively wrestling
The kinds of questions that carry the most weight in terms with inferior solutions (Kruger, 1993). In this sense, dialogue
of supporting learning are those that elicit talk moves, e.g., that involves challenges, disagreements, and contradictions
eliciting explanations, elaborations, inferences. These are prompts peers to collaboratively participate in the process of
questions that in effect facilitate transactivity, by coordinat- cognitive restructuring.
ing collaborative dialogue through the negotiation for mean-
ing. Computer facilitators of collaborative learning that elicit
Social Processes
these moves have also been found to produce more learning
than found in otherwise equivalent groups that do not have Thus far, our discussion has been limited to talk moves that
that support (Adamson, Dyke, Jang, Rosé, 2014; Dyke et al., support cognitive processes in dialogue. In what ways does
2013). the inherently social nature of dialogue shape what is said,
In addition to facilitating group-level processes, question and as a consequence, opportunities for learning through dia-
asking has been shown to be supportive of self-regulated logue? Classroom dialogue is socially contingent. Utterances
learning. In particular, King (1999) found that when students are a collaborative artifact, and contingent upon the interre-
were trained to use self-regulation questions to monitor their lationship between discussants, time, and space (Goffman,
own understanding and coordinate collaborative dialogue, 1959). There is a vast body of scholarship that explores the
they were more effective at problem solving than students contingent nature of classroom dialogue (e.g., Chen & Chiu,
who were not trained. 2008; Clarke, 2015). However, we limit our review to the
subset of studies that relate the social processes of dialogue
Challenges, disagreements, contradictions. One of the to learning outcomes, as they help provide insight into the
advantages of collaborative learning in dyads, small groups big effects of dialogue.
or whole-class situations is that by virtue of the number of
participants, discussion is a multiplier of ideas and perspec- Interpersonal effects. There are several separate lines
tives. Engaging in dialogue versus working alone increases of inquiry into the interpersonal dynamics of dialogue and
the probability that individuals will be exposed to a range of their impact on learning outcomes. This includes research
ideas beyond their own. The inherent multivocality of dia- that examines stable categories, e.g., relationship status, as
logue creates the conditions for challenges, disagreements, well as research that examines more fluid categories, e.g.,
and contradictions of opinions and ideas to arise. etiquette in dialogue. Together this body of work shows the
Challenges to ideas, disagreements, and contradictions connections between linguistic features of communication,
have been shown to be productive for learning for a num- underlying social processes, and cognitive processes that
ber of reasons. First, they have been shown to induce socio- impact learning.
cognitive conflict (Kruger, 1993; Mugny & Doise, 1978). Research that has examined benefits of collaborating with
Sociocognitive conflict is induced when an individual is con- friends versus acquaintances has shown consistent support
fronted with alternative or competing perspectives that chal- for collaborating with friends (Azmitia & Montgomery,
lenge the person’s current understanding. This state of conflict 1993; Barron, 2003; Ogan, Finkelstein, Walker, Carlson, &
prompts the individual (whose understanding is challenged) Cassell, 2012). Azmitia and Montgomery have found that
to undergo a process of cognitive restructuring in order to friends not only use more transactive moves during problem
integrate the new perspective into their understanding— solving in general, but there is a greater incidence of engage-
to re-achieve a state of equilibrium. ment with one another’s reasoning during problem solving
 Dialogic Instruction 385

(other-oriented transacts). Engagement with a friend’s rea- Howley and colleagues also document that “confron-
soning has been found to be a strong predictor of prob- tational behavior in collaboration has differential effects
lem-solving accuracy (Azmitia & Montgomery, 1993). depending on the relationship status of the students” (Howley
While the purely social aspects of language may connect et al., 2013a). These findings seem to suggest that the rapport
more indirectly with learning processes than cognitive con- that exists between friends versus acquaintances allows for
structs like transactivity, there is evidence that they play an more critical interaction, and these are the kinds of verbal
important facilitative role. As mentioned previously, acoustic interactions that are supportive of learning.
style features that signal trust and respect between students Other socially relevant group composition variables, such
are associated with increased prevalence of transactivity in as ability-level groupings, have also been examined. The evi-
conversation (Gweon et al., 2013). Similarly, Howley and dence on the relative benefits of collaborating with peers of
colleagues have demonstrated that groups that show open- the same ability level are somewhat mixed (Kruger, 1992;
ness in their rhetorical strategies elicit more displays of rea- Mugny & Doise, 1978). Mixed-ability groups have been
soning from group members (Howley, Kumar, Mayfield, found to be both productive and counterproductive for collab-
Dyke, & Rosé, 2013a). orative problem solving. Mugny and Doise (1978) found that,
Some research has examined the effect that rudeness can when children were paired with children of the same ability
have on collaborative problem solving dialogue, and whether level to solve spatial reasoning problems through dialogue,
the incidence of rudeness has a global impact, e.g., ultimately their collective performance did not improve. However,
impacting achievement. Not surprisingly, several studies when children of intermediate ability level were paired
found that rudeness can have immediate negative effects with low-ability-level partners, the collaboration yielded
on the dialogue itself and ultimately a negative impact on improvements for both children. This gestalt effect however,
achievement (Asterhan, 2013; Chiu & Khoo, 2003). There was only found between the low and intermediate dyads, and
are qualitative differences in how polite disagreements ver- not with the high–low dyads. Mugny and Doise argued that
sus adversarial disagreements are taken up by discussants, in this could be explained by the relative stability of problem-
terms of the ideas expressed by the disagreement. For exam- solving strategies used by high- and intermediate-level
ple, Chiu and Khoo (2003) found that rude disagreements children. Intermediate children’s strategies however, may
tend to spawn further rude disagreements by discussants. be still emergent and therefore more malleable, allowing
In addition, rude or adversarial disagreements lend them- for more negotiation in collaboration with low-ability-
selves to collaboration breakdowns, where participants are level children. By contrast, in high–low dyads, low-ability-
no longer co-constructing, but rather talking past one another level children’s problem solving improves to the level of their
(Asterhan, 2013). Like the research on friendship and col- high-ability-level partner at posttest. Mugny and Doise sug-
laboration, there seems to be a positive relationship between gest that this may be because high-ability-level children are
rapport in dialogue and outcomes. disseminating their strategies to their low-ability-level part-
Similarly, Howley and colleagues have reported on the ners, rather than collaboratively constructing them.
effects of confrontational remarks on collaboration in a 2-day However, when Kruger examined the differences between
activity. They found that confrontational remarks from the peer (same ability level) vs. adult–child dyads (high–low
first day of collaboration were associated with shifts in social ability level), she found that peers outperformed adult–child
positioning of students between Days 1 and 2. Recipients dyads on reasoning posttests (Kruger, 1992). In addition,
of confrontational comments on Day 1 became reluctant to peer dyads contained a greater number of transactive reason-
try again on Day 2 when they reached a problem-solving ing moves during the collaborative activity than did adult–
impasse. This in turn reduced the needed practice they were child dyads.
able to get, which was associated with a corresponding
reduction in learning (Howley, Mayfield, & Rosé, 2013b). Temporal effects. How does time impact what occurs in
The dialogue produced between friends seems to be expe- dialogue? If we conceptualize class discussions as a kind of
rienced in a qualitatively different manner from dialogue that living organism, then Richard Anderson and colleagues have
appears similar in form but is experienced between acquaint- been examining the ways in which changes in the organ-
ances. In particular, Ogan and colleagues have shown that, ism result in changes in the classroom ecology over time.
while the surface features of dialogue might be the same Anderson and colleagues have been investigating what they
when comparing friends and acquaintances, their commu- refer to as the “snowball phenomenon,” which is the way
nicative function is entirely dependent upon the relationship in which social and thinking practices, argument stratagems,
between discussants (Ogan et al., 2012). For example, they emerge in discussions over time (Anderson et al., 2001).
found that cues such as “confrontations” between friends in Argument stratagems are a class of talk moves like those
peer tutoring dyads were followed by productive responses reviewed in the previous section. Anderson and colleagues
that served to keep both participants coordinated on the prob- suggest that acquiring argument stratagems means both
lem-solving activity. However, with acquaintances, the same understanding the form and function of these moves, as well
cue (confrontations) was threatening, and as a result, these as putting them to use in appropriate and meaningful ways.
cues had a negative impact on learning outcomes of these Discussions in which children are tasked with reasoning
dyads. together provide authentic opportunities for discussants to
386 Sherice N. Clarke et al.

observe their peers using argument stratagems effectively, suggest that classroom micro cultures are co-constructed
and consequently internalize their appropriate form and use by teachers and students alike (Clarke, 2015; Clarke et al.,
in argumentation. They argue that children’s observations 2013). Verbal and non-verbal signals, such as who has the
of effective use of a stratagem is what can account for the right to speak, to try out ideas, to get things wrong, or be
snowballing phenomenon that they have documented across correct, are in the ether, so to speak. Micro cultures can shape
a series of studies (Anderson et al., 2001; Li et al., 2007; participation patterns, and may account for the differences in
Lin et al., 2012; Sun, Anderson, Lin, & Morris, 2015). They participation patterns associated with gender and SES that
have found that when children participate in successive dis- studies have reported. These results suggest that support
cussions with peers, the use of particular stratagems spread for productive co-construction of micro cultures could help
across the discussants over time, gradually increasing in the to overcome obstacles that students bring into the learning
frequency of their use. This phenomenon has been docu- environment based on demographic and dispositional factors
mented in terms of social stratagems, e.g., leadership moves (see Chapter 27, this volume).
(Anderson et al., 2001; Li et al., 2007), as well as the use
of cognitive stratagems like the use of analogical reasoning
Cognitive and Social Dimensions of Dialogic
(Lin et al., 2012; Sun et al., in press).
Instruction
In short, Anderson and colleagues have documented how
the norms of collaborative reasoning and reasoning itself The above body of evidence suggests that there are several
become socially distributed over time. As individuals appro- processes at work when individuals engage in structured col-
priate moves, they provide a model for peers. This modeling laborative sense-making discussions about academic subject
and appropriation work spreads exponentially. Thus, this matter that support the development of the kind of learning
body of research shows the social and intellectual develop- described as the big-effects of dialogic instruction.
mental process that unfolds as students participate in dia- Explanation, transactive reasoning, and challenge/dis-
logue of this kind over time. This research highlights the key agreement talk moves have been shown to support learning
role that the students themselves play in bringing change to and long-term retention, as well as conceptual change and
classroom discussion practices. Clarke and colleagues have reasoning development when used in collaborative prob-
found that initiating small groups of students into dialogic lem-solving discussions. However, the extent of the benefits
discussion practices using computer support is associated for learning seems to be associated with the role that one
with corresponding positive effects on teacher uptake of dia- inhabits with respect to the ideas produced (e.g., speaker or
logic facilitation practices in classroom discussions (Clarke listener), and the extent to which one operates on those ideas.
et al., 2013). In other words, just being there in collaborative discussions
Other temporal effects that operate at the micro level, at seems to be an insufficient condition in order to benefit in
the level of utterances or a sequence of utterances, are those terms of learning. Cognitive engagement with ideas and
reported above on the effects of rudeness on collaboration active use of ideas generated to solve problems seem to be
and outcomes (Asterhan, 2013; Chiu & Khoo, 2003; Howley necessary conditions for learning.
et al., 2013b). Dialogue in which the reasoning of discussants is interpen-
etrated supports problem solving and reasoning development.
Identity effects. Identity has primarily been examined in The evidence related to effects of transactivity suggests that
terms of how relational identities impact participation in it is not enough for individuals to externalize their thinking.
dialogue. A few studies have shown that gender and socio- Rather, the thinking made public must become an ideational
economic status (SES) are correlated with participation artifact that is manipulated and developed by discussants
patterns in dialogue (Howe & McWilliam, 2001; Kelly, over time. It is engagement in the manipulation of ideas that
2008). Howe and McWilliam and others have found that lends itself to development in reasoning ability. Challenges,
the frequency and complexity of children’s talk moves were disagreements, and contradictions are moves that are bene-
associated with SES, with children from high-SES back- ficial for learning because they induce productive cognitive
grounds engaging in argumentation with greater frequency processes (e.g., explanation and reasoning, sociocognitive
and complexity than children from low-SES backgrounds conflict) that help to push individual thinking forward. These
(Christie, 1999; Howe & McWilliam, 2001). These differences are moves that are transactive by their very definition.
have been interpreted as a function of differences in lan- Engaging in successive discussions serves as a practice
guage exposure in the home, rather than that of the relative space for individuals to learn and eventually appropriate
capabilities of the children. Student dispositions, such as productive talk moves (Anderson et al., 2001). However,
high self-efficacy beliefs in their ability to contribute mean- discussants’ willingness, actual participation, and quality of
ingfully to a collaborative discussion, have been shown to contributions in discussions can be mediated by social fac-
be associated with the ways in which students are socially tors, e.g., identity, rapport, and etiquette.
and epistemically positioned in discussions (Clarke, 2015; While our review provides insight into some of the mech-
Howley, Mayfield, & Rosé, 2011; Howley et al., 2012; see anisms through which dialogic instruction supports robust
Chapter 11, this volume). learning, our discussion is limited to the current state of the
While these studies show demographic and dispositional art. We have discussed the evidence of mechanisms that
effects on participation in dialogue, Clarke and colleagues support learning, retention, and reasoning development;
 Dialogic Instruction 387

however, we have not identified targeted dialogic instruction Technology can also play an important role in creating the
interventions that help to explain the mechanisms that sup- rich learning environments in which to use dialogic instruc-
port domain transfer. Much more scholarship is needed to tion to support learning. Existing research in this area shows
understand the cognitive and social dimensions that support how computer agents can enhance the learning experiences of
robust learning through dialogue. students. A limitation of the research to date, however, is that
study of the uses of this kind of technology is short-term, e.g.,
one or two class periods. New questions are raised when such
Into the Future interventions take place over a more extended period of time,
The findings on dialogic processes and outcomes make it e.g., successive instruction over an academic year. This will
possible to conclude that certain forms of instructional dia- raise questions such as how these technologies can adapt sup-
logue support steep increases in student learning. However, port to account for student growth in knowledge, reasoning,
surveys of classroom language show that academically and collaborative skill (Adamson et al., 2014).
productive dialogue is rarely observed in low-performing Finally, and perhaps of greatest interest to educational
schools (Pauli & Reusser, 2015). Yet these are precisely the psychologists who support pre-service teacher instruction,
instructional settings that could benefit most from introduc- technology could be used to dynamically support teachers
ing rigorous forms of instruction to raise the achievement of (e.g., Chen, Clarke, & Resnick, 2014). The same kind of
students they serve in the short term, and enable greater life technologies that support students in collaborative learn-
chances in the long term. ing can be repurposed to support teachers in facilitating the
While we may want to make this form of instruction kind of discussions that support learning in real time (e.g.,
widespread practice, there is evidence that suggests scal- McLaren, Scheuer, & Miksáko, 2010).
ing this kind of practice is not trivial. Longitudinal teach-
er-training efforts highlight how challenging it is to train Notes
teachers in low-performing schools, serving children of the 1. Cohort 1 received the Project Challenge intervention for 4 years, from
greatest need, how to use dialogic instruction effectively Grade 4 through Grade 7. Cohorts 2–4 began the intervention in Year 2,
(Clarke et al., 2013). There is a need to develop approaches receiving the program for 3 years, in Grades 4 through 6.
that can leverage these insights on dialogic instruction, to 2. The search for transfer was associated with the development of educational
improve instruction at scale. psychology in the early part of the twentieth century. By and large, research-
ers did not find it, and by the 1980s, the question had changed to whether or
In meeting this goal, the research in computer-supported not thinking skills could be taught in the abstract (Resnick, 1987).
collaborative learning shows promise for helping in this
process (see Chapter 26, this volume). This is because we
know that instructional technologies can play multiple roles References
in teaching and learning. In recent years, instructional tech- Adamson, D., Dyke, G., Jang, H. J., & Rosé, C. P. (2014). Towards an agile
nologies have developed intelligent computer agents that approach to adapting dynamic collaboration support to student needs.
can play the role of a collaborative learning partner (Ogan International Journal of AI in Education 24(1), 91–121.
et al., 2012), or as a facilitator of group discussion, lever- Adey, P., & Shayer, M. (1990). Accelerating the development of formal
thinking in middle and high school students. Journal of Research in
aging insights on talk moves to dynamically support learn- Science Teaching, 27(3), 267–285.
ing (Adamson et al., 2014; Dyke et al., 2013; Kumar, Rosé, Adey, P., & Shayer, M. (1993). An exploration of long-term far-transfer
Wang, Joshi, & Robinson, 2007). effects following an extended intervention program in the high school
Technological capacities are growing rapidly and there is science curriculum. Cognition and Instruction, 11(1), 1–29.
already a substantial body of research that provides a founda- Alexander, R. J. (2008). Towards dialogic teaching: Rethinking classroom
talk. Cambridge: Dialogos.
tion for more ambitious human–computer learning-oriented Anderson, R. C., Nguyen-Jahiel, K., McNurlen, B., Archodidou, A., Kim,
interaction in the future, especially as more attention is given S.-Y., Reznitskaya, A., . . . Gilbert, L. (2001). The snowball phenome-
towards learning at scale. We discuss here just a few lines of non: Spread of ways of talking and ways of thinking across groups of
work that provide a glimpse of possibilities and challenges children. Cognition and Instruction, 19(1), 1–46.
for the future. Asterhan, C. S. C. (2013). Epistemic and interpersonal dimensions of peer
argumentation. In M. Baker, J. Andriessen, & S. Järvelä (Eds.), Affective
The first is the use of technology as a research tool for learning together: Social and emotional dimensions of collaborative
the study of dialogic instruction. Technology can enable the learning (pp. 251–271). London: Routledge.
experimental isolation of specific characteristics of discus- Asterhan, C. S. C., & Schwarz, B. B. (2007). The effects of monological and
sion, to study the social and cognitive processes of dialogic dialogical argumentation on concept learning in evolutionary theory.
instruction systematically (e.g., Adamson et al., 2014; Dyke Journal of Educational Psychology, 99(3), 626.
Asterhan, C. S. C., & Schwarz, B. B. (2009). The role of argumentation and
et al., 2013; Kumar, Ai, Beuth, & Rosé, 2010; Kumar, Beuth, explanation in conceptual change: Indications from protocol analyses of
& Rosé, 2011). In order to uncover the mechanisms through peer-to-peer dialogue. Cognitive Science, 33(3), 374–400.
which discussion impacts learning, we must isolate spe- Azmitia, M., & Montgomery, R. (1993). Friendship, transactive dialogues,
cific characteristics of discussion, possibly even at the level and the development of scientific reasoning. Social Development, 2(3),
of presence or absence of specific talk moves. This level of 202–221.
Barron, B. (2003). When smart groups fail. Journal of the Learning Sciences,
experimental control is extremely difficult in the context of 12(3), 307–359.
whole-class, teacher-led discussion; however, it is possible to Bellack, A. A., Kliebard, H. M., Hyman, R. T., & Smith, F. L. (1966). The
achieve when the facilitator is a computer agent. language of the classroom. New York: Teachers College Press.
388 Sherice N. Clarke et al.
Berkowitz, M. W., & Gibbs, J. C. (1983). Measuring the developmental Goffman, E. (1959). The presentation of self in everyday life. New York:
features of moral discussion. Merrill-Palmer Quarterly, 29(4), 399–410. Anchor Books.
Bill, V., Leer, M., Reams, L., & Resnick, L. (1992). From cupcakes to equa- Greeno, J. G. (2015). Classroom talk sequences and learning. In L. B.
tions: The structure of discourse in a primary mathematics classroom. Resnick, C. S. C. Asterhan, & S. N. Clarke (Eds.), Socializing intelli-
Verbum, 1, 2, 63–85. gence through academic talk and dialogue. Washington, DC: American
Bloom, B. S., & Krathwohl, D. R. (1956). Taxonomy of educational objec- Educational Research Association.
tives: The classification of educational goals. Handbook I: Cognitive Greeno, J. G., Moore, J. L., & Smith, D. R. (1993). Transfer of situated
domain. New York: David McKay. learning. In D. K. Detterman & R. Sternberg (Eds.), Transfer on trial:
Bourdieu, P. (1977). The economics of linguistic exchanges. Social Science Intelligence, cognition, and instruction (pp. 99–197). Westport, CT:
Information, 16, 645–668. Ablex Publishing.
Carbonell, J. (1969). On man–computer interaction: A model and some Gweon, G., Jain, M., Mc Donough, J., Raj, B., & Rosé, C. P. (2013).
related issues. IEEE Transactions on Systems Science and Cybernetics, Measuring prevalence of other-oriented transactive contributions using
5(1), 16–26. an automated measure of speech style accommodation. International
Carbonell, J. (1970). AI in CAI: An artificial intelligence approach to com- Journal of Computer Supported Collaborative Learning 8(2), 245–265.
puter-assisted instruction. IEEE Transactions on Man-Machine Systems, Howe, C. (2010). Peer dialogue and cognitive development: A two-way
11(4), 190–202. relationship? In K. Littleton & C. Howe (Eds.), Educational dialogues:
Chapin, S., & O’Connor, C. (2004). Project challenge: Identifying and Understanding and promoting productive interaction. London: Routledge.
developing talent in mathematics within low-income urban schools. Howe, C., & McWilliam, D. (2001). Peer argument in educational settings
Boston University School of Education Research Report 1, 1–6. variations due to socioeconomic status, gender, and activity context.
Chen, G., & Chiu, M. M. (2008). Online discussion processes: Effects of Journal of Language and Social Psychology, 20(1–2), 61–80.
earlier messages’ evaluations, knowledge content, social cues and per- Howe, C., & McWilliam, D. (2006). Opposition in social interaction
sonal information on later messages. Computers & Education, 50(3), amongst children: Why intellectual benefits do not mean social costs.
678–692. Social Development, 15(2), 205–231.
Chen, G., Clarke, S. N., & Resnick, L. B. (2014). An analytic tool for support- Howe, C., McWilliam, D., & Cross, G. (2005). Chance favours only the
ing teachers’ reflection on classroom talk. In J. L. Polman, E. A. Kyza, D. prepared mind: Incubation and the delayed effects of peer collaboration.
K. O’Neill, I. Tabak, W. R. Penuel, A. S. Jurow, K. O’Connor, T. Lee, & British Journal of Psychology, 96(1), 67–93.
L. D’Amico (Eds.), Learning and becoming in practice: The International Howe, C., Tolmie, A., Thurston, A., Topping, K., Christie, D., Livingston,
Conference of the Learning Sciences (ICLS) 2014 (Vol. 1, pp. 583–590). K., . . . Donaldson, C. (2007). Group work in elementary science:
Boulder, CO: International Society of the Learning Sciences. Towards organisational principles for supporting pupil learning.
Chi, M. T. (2009). Active-constructive-interactive: A conceptual framework Learning and Instruction, 17(5), 549–563.
for differentiating learning activities. Topics in Cognitive Science, 1(1), Howley, I., Adamson, D., Dyke, G., Mayfield, E., Beuth, J., & Rosé, C. P.
73–105. (2012). Group composition and intelligent dialogue tutors for impacting
Chi, M. T., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). students’ self-efficacy. In S. A. Cerri, W. J. Clancey, G. Papadourakis,
Self-explanations: How students study and use examples in learning to & K. K., Panourgia (Eds.), ITS 2012 Proceedings of the 11th
solve problems. Cognitive Science, 13(2), 145–182. International conference on Intelligent Tutoring Systems, Lecture notes
Chi, M. T., Slotta, J. D., & De Leeuw, N. (1994). From things to processes: in computer science (Vol. 7315; pp. 551–556). Chania, Crete, Greece:
A theory of conceptual change for learning science concepts. Learning Springer-Verlag.
and Instruction, 4(1), 27–43. Howley, I., Kumar, R., Mayfield, E., Dyke, G., & Rosé, C. P. (2013a). Gaining
Chinn, C. A., O’Donnell, A. M., & Jinks, T. S. (2000). The structure of insights from sociolinguistic style analysis for redesign of conversational
discourse in collaborative learning. Journal of Experimental Education, agent based support for collaborative learning. In D. Suthers, K. Lund,
69(1), 77–97. C. P. Rosé, C. Teplovs, & N. Law (Eds.). Productive multivocality in the
Chiu, M. M., & Khoo, L. (2003). Rudeness and status effects during group analysis of group interactions, edited volume. New York: Springer.
problem solving: Do they bias evaluations and reduce the likelihood of Howley, I., Mayfield, E., & Rosé, C. P. (2011). Missing something?
correct solutions? Journal of Educational Psychology, 95(3), 506. Authority in collaborative learning. In Proceedings of the 9th
Christie, F. (Ed.) (1999). Pedagogy and the shaping of consciousness: International Computer Supported Collaborative Learning Conference,
Linguistic and social processes. London: Cassell Academic. Volume 1: Long Papers, pp. 336–373. Rhodes: International Society of
Clarke, S. N. (2015). The right to speak. In L. B. Resnick, C. S. C. Asterhan, the Learning Sciences.
& S. N. Clarke (Eds.), Socializing intelligence through academic talk Howley, I., Mayfield, E., & Rosé, C. P. (2013b). Linguistic analysis meth-
and dialogue. Washington, DC: American Educational Research ods for studying small groups. In C. Hmelo-Silver, A. O’Donnell,
Association. C. Chan, & C. Chinn (Eds.), International handbook of collaborative learn-
Clarke, S. N., Chen, G., Stainton, C., Katz, S., Greeno, J. G., Resnick, L. ing. New York: Routledge.
B., . . . Rosé, C. P. (2013). The impact of CSCL beyond the online Hymes, D. (1977). Foundations in sociolinguistics: An ethnographic approach.
environment. In N. Rummel, M. Kapur, M. Nathan, & S. Puntambekar London: Tavistock Publications.
(Eds.), To see the world and a grain of sand: Learning across levels of Kelly, S. (2008). Race, social class, and student engagement in middle
space, time, and scale: CSCL 2013 (Vol. 1, pp. 105–112). Madison, WI: school English classrooms. Social Science Research, 37, 434–448.
International Society of the Learning Sciences. King, A. (1992). Facilitating elaborative learning through guided stu-
Dewey, J. (1916). Democracy and education. New York: The Free Press. dent-generated questioning. Educational Psychologist, 27(1), 111–126.
Dyke, G., Adamson, A., Howley, I., & Rosé, C. P. (2013). Enhancing King, A. (1999). Discourse patterns for mediating peer learning. In A. M.
scientific reasoning and discussion with conversational agents. IEEE O’Donnell & A. King (Eds.), Cognitive perspectives on peer learning.
Transactions on Learning Technologies 6(3), special issue on Science Mahwah, NJ: Lawrence Erlbaum.
Teaching, 240–247. King, A., & Rosenshine, B. (1993). Effects of guided cooperative ques-
Fuchs, L. S., Fuchs, D., Hamlett, C. L., Phillips, N. B., Karns, K., & Dutka, tioning on children’s knowledge construction. Journal of Experimental
S. (1997). Enhancing students’ helping behavior during peer-mediated Education, 61(2), 127–148.
instruction with conceptual mathematical explanations. The Elementary Klahr, D., & Nigam, M. (2004). The equivalence of learning paths in early
School Journal, 223–249. science instruction: Effects of direct instruction and discovery learning.
Gall, M. D. (1970). The use of questions in teaching. Review of Educational Psychological Science, 15, 661–667.
Research, 707–721. Koedinger, K. R., Corbett, A. T., & Perfetti, C. (2012). The knowledge-learn-
Garfinkel, H. (1967). Studies in ethnomethodology. Englewood Cliffs, NJ: ing-instruction framework: Bridging the science-practice chasm to
Prentice-Hall. enhance robust student learning. Cognitive Science, 36(5), 757–798.
 Dialogic Instruction 389
Kruger, A. C. (1992). The effect of peer and adult–child transactive dis- Rosenshine, B., Meister, C., & Chapman, S. (1996). Teaching students
cussions on moral reasoning. Merrill-Palmer Quarterly (1982–), to generate questions: A review of the intervention studies. Review of
191–211. Educational Research, 66(2), 181–221.
Kruger, A. C. (1993). Peer collaboration: Conflict, cooperation, or both? Searle, J. R. (1969). Speech acts: An essay in the philosophy of language
Social Development, 2(3), 165–182. (Vol. 626). Cambridge, UK: Cambridge University Press.
Kumar, R., Ai, H., Beuth, J., & Rosé, C. P. (2010). Socially-capable con- Sinclair, J. M. H., Coulthard, M., & Council, S. S. R. (1975). Towards an
versational tutors can be effective in collaborative learning situations. analysis of discourse: The English used by teachers and pupils. London:
In Proceedings of Intelligent Tutoring Systems, lecture notes in com- Oxford University Press.
puter science, Volume 6095, pp. 156–164. Pittsburgh, PA: International Slavin, R. (1987). Ability grouping and student achievement in elementary
Society of the Learning Sciences. schools: A best-evidence synthesis. Review of Educational Research,
Kumar, R., Beuth, J., & Rosé, C. P. (2011). Conversational strategies that 57(3), 293–336.
support idea generation productivity in groups. In Proceedings of the 9th Stahl, G., Koschmann, T., & Suthers, D. (2006). Computer-supported
International Computer Supported Collaborative Learning Conference, collaborative learning: An historical perspective. In R. K. Sawyer
Volume 1: Long papers, pp. 398–405. Hong Kong: International Society (Ed.), Cambridge handbook of the learning sciences (pp. 409–426).
of the Learning Sciences. Cambridge, UK: Cambridge University Press.
Kumar, R., Rosé, C. P., Wang, Y. C., Joshi, M., & Robinson, A. (2007). Stevens, A., & Collins, A. (1977). The goal structure of a Socratic tutor. In
Tutorial dialogue as adaptive collaborative learning support. In Proceedings of the National ACM Conference. New York: Association
Proceedings of the 2007 conference on Artificial Intelligence in for Computing Machinery.
Education: building technology rich learning contexts that work, pp. Sun, J., Anderson, R. C., Lin, T., & Morris, J. (2015). Social and cognitive
383–390. Amsterdam, The Netherlands: IOS Press. development during collaborative reasoning. In L. B. Resnick, C. S. C.
Li, Y., Anderson, R. C., Nguyen-Jahiel, K., Dong, T., Archodidou, A., Kim, Asterhan, & S. N. Clarke (Eds.), Socializing intelligence through aca-
I.-H., . . . Jadallah, M. (2007). Emergent leadership in children’s discus- demic talk and dialogue. Washington, DC: American Educational
sion groups. Cognition and Instruction, 25(1), 1–2. Research Association.
Lin, T.-J., Anderson, R. C., Hummel, J. E., Jadallah, M., Miller, B. W., Teasley, S. D. (1997). Talking about reasoning: How important is the peer
Nguyen-Jahiel, K., . . . Wu, X. (2012). Children’s use of analogy during in peer collaboration? In L. B. Resnick, R. Säljö, C. Pontecorvo, &
collaborative reasoning. Child Development, 83(4), 1429–1443. B. Burge (Eds.), Discourse, tools and reasoning: Essays on situated
Lipman, M. (1981). Philosophy for children. In A. I. Costa (Ed.), Developing cognition (pp. 361–384). Berlin: Springer.
minds: Programs for teaching thinking (Vol. 2, pp. 35–38). Alexandria, Topping, K. J., & Trickey, S. (2007a). Collaborative philosophical enquiry
VA: Association for Supervision and Curricular Development. for school children: Cognitive effects at 10–12 years. British Journal of
Littleton, K., & Howe, C. (2010). Educational dialogues: Understanding Educational Psychology, 77(2), 271–288.
and promoting productive interaction. London: Routledge. Topping, K. J., & Trickey, S. (2007b). Collaborative philosophical inquiry
McLaren, B. M., Scheuer, O., & Miksáko, J. (2010). Supporting collabora- for schoolchildren: Cognitive gains at 2 year follow up. British Journal
tive learning and e-discussions using artificial intelligence techniques. of Educational Psychology, 77(4), 787–796.
International Journal of Artificial Intelligence in Education, 20(1), 1–46. Veel, R., & Christie, F. (1999). Language, knowledge and authority in
Mead, G. H. (1967). Mind, self, and society: From the standpoint of a social school mathematics. In J. R. Martin, & F. Christie (Eds.), Pedagogy
behaviorist (Vol. 1): Chicago, IL: The University of Chicago Press. and the shaping of consciousness: Linguistic and social processes
Mehan, H. (1979). Learning lessons: Social organization in the classroom. (pp. 185–216). London: Continuum.
Cambridge, MA: Harvard University Press. Veenman, S., Denessen, E., van den Akker, A., & van der Rijt, J. (2005).
Mercer, N., Wegerif, R., & Dawes, L. (1999). Children’s talk and the devel- Effects of a cooperative learning program on the elaborations of students
opment of reasoning in the classroom. British Educational Research during help seeking and help giving. American Educational Research
Journal, 25(1), 95–111. Journal, 42, 115–151.
Mugny, G., & Doise, W. (1978). Socio-cognitive conflict and structure of Vygotsky, L. S. (1962). Thought and language. Cambridge, MA: MIT Press.
individual and collective performances. European Journal of Social Vygotsky, L. S. (1978). Mind in society: The development of higher psycho-
Psychology, 8(2), 181–192. logical processes. Cambridge, MA: Harvard University Press.
Ogan, A., Finkelstein, S., Walker, E., Carlson, R., & Cassell, J. (2012). Webb, N. M. (1991). Task-related verbal interaction and mathematics learn-
Rudeness and rapport: Insults and learning gains in peer tutoring. ing in small groups. Journal for Research in Mathematics Education,
Intelligent Tutoring Systems (pp. 11–21). Berlin: Springer. 366–389.
Pauli, C., & Reusser, K. (2015). Discursive cultures of learning in (every- Webb, N. M., Franke, M. L., De, T., Chan, A. G., Freund, D., Shein, P., &
day) mathematics teaching: A video-based study on mathematics Melkonian, D. K. (2009). ‘Explain to your partner’: teachers’ instruc-
teaching in German and Swiss classrooms. In L. B. Resnick, C. S. C. tional practices and students’ dialogue in small groups. Cambridge
Asterhan, & S. N. Clarke (Eds.), Socializing intelligence through Journal of Education, 39(1), 49–70.
academic talk and dialogue. Washington, DC: American Educational Webb, N. M., & Mastergeorge, A. M. (2003). The development of students’
Research Association. helping behavior and learning in peer-directed small groups. Cognition
Resnick, L. B. (1987) Education and learning to think. Washington, DC. and Instruction, 21(4), 361–428.
National Academy of Education Press. Webb, N. M., Troper, J. D., & Fall, R. (1995). Constructive activity
Resnick, L. B., Asterhan, C. S. C., & Clarke, S. N. (Eds.). (2015). Socializing and learning in collaborative small groups. Journal of Educational
intelligence through academic talk and dialogue. Washington, DC: Psychology, 87, 406–423.
American Educational Research Association. Wegerif, R., Mercer, N., & Dawes, L. (1999). From social interaction to indi-
Resnick, L. B., Bill, V., Lesgold, S., & Leer, M. (1991). Thinking in arith- vidual reasoning: an empirical investigation of a possible socio-cultural
metic class. In B. Means, C. Chelemer, & M. S. Knapp (Eds.), Teaching model of cognitive development. Learning and Instruction, 9(6), 493–516.
advanced skills to at-risk students: Views from research and practice. Wertsch, J. V. (1979). From social interaction to higher psychological pro-
San Francisco, CA: Jossey-Bass. cesses: A clarification and application of Vygotsky’s theory. Human
Resnick, L. B., Michaels, S., & O’Connor, C. (2010). How (well-structured) Development, 22(1), 1–22.
talk builds the mind. In D. Preiss & R. Sternberg (Eds.), Innovations in Winne, P. H. (1979). Experiments relating teachers’ use of higher cogni-
educational psychology: Perspectives on learning, teaching and human tive questions to student achievement. Review of Educational Research,
development (pp. 163–194). New York: Springer. 13–49.