ArticlePDF Available

Understanding Tutor Learning: Knowledge-Building and Knowledge-Telling in Peer Tutors' Explanations and Questions


Abstract and Figures

Prior research has established that peer tutors can benefit academically from their tutoring experiences. However, although tutor learning has been observed across diverse settings, the magnitude of these gains is often underwhelming. In this review, the authors consider how analyses of tutors' actual behaviors may help to account for variation in learning outcomes and how typical tutor- ing behaviors may create or undermine opportunities for learning. The authors examine two tutoring activities that are commonly hypothesized to support tutor learning: explaining and questioning. These activities are hypothesized to sup- port peer tutors' learning via reflective knowledge-building, which includes self-monitoring of comprehension, integration of new and prior knowledge, and elaboration and construction of knowledge. The review supports these hypothe- ses but also finds that peer tutors tend to exhibit a pervasive knowledge-telling bias. Peer tutors, even when trained, focus more on delivering knowledge rather than developing it. As a result, the true potential for tutor learning may rarely be achieved. The review concludes by offering recommendations for how future research can utilize tutoring process data to understand how tutors learn and perhaps develop new training methods.
Content may be subject to copyright.
Review of Educational
The online version of this article can be found at:
DOI: 10.3102/0034654307309920
Rod D. Roscoe and Michelene T. H. Chi
Knowledge-Telling in Peer Tutors' Explanations and Questions
Understanding Tutor Learning: Knowledge-Building and
Published on behalf of
American Educational Research Association
can be found at:Review of Educational ResearchAdditional services and information for Alerts:
What is This?
- Dec 3, 2007Version of Record >>
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Review of Educational Research
December 2007, Vol. 77, No. 4, pp. 534–574
DOI: 10.3102/0034654307309920
© 2007 AERA.
Understanding Tutor Learning: Knowledge-
Building and Knowledge-Telling in Peer Tutors’
Explanations and Questions
Rod D. Roscoe and Michelene T. H. Chi
University of Pittsburgh
Prior research has established that peer tutors can benefit academically from
their tutoring experiences. However, although tutor learning has been observed
across diverse settings, the magnitude of these gains is often underwhelming.
In this review, the authors consider how analyses of tutors’ actual behaviors
may help to account for variation in learning outcomes and how typical tutor-
ing behaviors may create or undermine opportunities for learning. The authors
examine two tutoring activities that are commonly hypothesized to support tutor
learning: explaining and questioning. These activities are hypothesized to sup-
port peer tutors’ learning via reflective knowledge-building, which includes
self-monitoring of comprehension, integration of new and prior knowledge, and
elaboration and construction of knowledge. The review supports these hypothe-
ses but also finds that peer tutors tend to exhibit a pervasive knowledge-telling
bias. Peer tutors, even when trained, focus more on delivering knowledge
rather than developing it. As a result, the true potential for tutor learning may
rarely be achieved. The review concludes by offering recommendations for how
future research can utilize tutoring process data to understand how tutors learn
and perhaps develop new training methods.
KEYWORDS: peer tutoring, tutor learning, knowledge-building, explaining, ques-
tioning, metacognition, learning strategies.
One of the most intriguing aspects of peer tutoring, in which students tutor other
students, is its potential to support learning for both the tutees and the tutors. The
belief that tutors benefit academically, along with lower costs and the large pool
of potential tutors, has provided a long-standing justification for peer tutoring pro-
grams (Allen & Feldman, 1976; Gartner, Kohler, & Riessman, 1971). For exam-
ple, the 17th-century philosopher Comenius argued that “the process of teaching
. . . gives a deeper insight into the subject taught” (as quoted in Wagner, 1982,
p. 31). Similar ideas were echoed two centuries later by the Swiss educator, Johann
Pestalozzi, who wrote that his students “were delighted when they knew something
they could teach to others . . . and they learned twice as well by making the
younger ones repeat their words” (as quoted in Wagner, 1982, p. 118). More
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
recently, researchers have demonstrated empirical evidence of learning gains for
tutors compared to nontutors (P. Cohen, Kulik, & Kulik, 1982), which we refer to
as the tutor learning effect.
Previous reviews have described the scope of the effect and contributed to our
knowledge of effective program design. In this review, we critically examine
research on the actual tutoring process to better understand how tutor learning
occurs. We focus on how tutors’ explaining and questioning activities during the
tutoring sessions may incorporate either reflective knowledge-building (e.g., mon-
itoring of comprehension and knowledge) or knowledge-telling (e.g., summariza-
tion with little monitoring or elaboration), which should lead to stronger or weaker
learning, respectively. Explaining and questioning are chosen because they are
ubiquitous and fundamental aspects of the tutoring process (Graesser, Person, &
Magliano, 1995) and have been specifically studied as sources of tutor learning
(King, Staffieri, & Adelgais, 1998).
Our review is presented in three major sections. We first briefly highlight evi-
dence of tutor learning across a variety of formats, students, and domains. We then
discuss hypotheses regarding how tutors learn from explaining and questioning
and evaluate these ideas by considering studies that directly analyzed tutors’
behaviors. In the final section, we consider methodological issues in the literature
and recommend directions for new research. Throughout this review, we adopt a
primarily cognitive perspective. We are interested in how peer tutors acquire and
transform their subject matter knowledge, and so we focus on the processes that
directly support such changes. Of course, peer tutoring also occurs within a larger
community of educational and social expectations. Peer tutors’ perceptions of their
tutoring role, and their motivational attitudes and beliefs, may influence how peer
tutors approach their task and, thus, their tutoring actions (Foot, Shute, Morgan, &
Barron, 1990). To address these issues in depth is beyond the scope of this article,
but aspects of these factors will be addressed in our final section.
Evidence of Tutor Learning
We define peer tutoring as the recruitment of one student to provide one-on-one
instruction for another student, accompanied by explicit assignment of participants
to “tutor” and “tutee” roles. Typically, the tutor is more expert or advanced than the
tutee, but in some variations of tutoring this knowledge gap is minimal. At the most
fundamental level, it is the instructional task and the asymmetrical tutor and tutee
roles that distinguish peer tutoring from other forms of peer learning. For example,
in peer collaboration, students might work together in small groups to solve math
problems (Webb, Troper, & Fall, 1995). Although group members may differ in
ability and are encouraged to help each other, none are explicitly assigned to tutor
their peers. In contrast, math tutors are expected to use their math knowledge to
teach key concepts and guide tutees toward correct solutions and understanding.
Topping and Ehly (2001) have defined several features that can vary across pro-
grams, such as training method, participant age, duration, and curriculum. Different
combinations of features give rise to a wide range of tutoring programs. Remarkably,
the opportunity to tutor a peer seems able to enhance learning for diverse students
working in a variety of such settings (Allen, 1983; Britz, Dixon, & McLaughlin,
1989; P. Cohen et al., 1982; Cook, Scruggs, Mastropieri, & Casto, 1986; Goodlad &
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
Hirst, 1989; Greenwood, Carta, & Hall, 1988; Mastropieri, Spencer, Scruggs, &
Talbott, 2000; Mathes & Fuchs, 1994; McMaster, Fuchs, & Fuchs, 2006; Robinson,
Schofield, & Steers-Wentzell, 2005; Rohrbeck, Ginsburg-Block, Fantuzzo, & Miller,
2003; Topping, 2005). However, despite such broad applicability, meta-analyses
tend to report effect sizes that are underwhelming in magnitude.
To further illustrate these points, we briefly consider tutor learning across three
key program features: format, demographics, and domain. Relevant mean effect
sizes are summarized in Table 1. The most common comparisons used in individ-
ual studies were between students who learned by tutoring versus studying a text
(Annis, 1983) or participating in regular classroom activities (Sharpley, Irvine, &
Sharpley, 1983). In fewer cases, tutoring was compared to an alternative treatment
such as reciprocal teaching (Klingner & Vaughn, 1996). Effect sizes reported in
the meta-analyses were calculated by taking the difference in means between tutor-
ing and nontutoring groups and dividing by the standard deviation or pooled
standard deviation. The resulting value describes group differences in standard
deviation units, and positive values indicate an advantage for tutors. As a rule of
thumb, effect sizes around .20 are considered small, effects around .50 are medium,
and effects of .80 or higher are large (J. Cohen, 1992).
Different populations were considered across the five meta-analyses. P. Cohen
et al. (1992) considered only studies using general education students, whereas
Cook et al. (1986), Mastropieri et al. (2000), and Mathes and Fuchs (1994) con-
sidered special education students. Mathes and Fuchs further focused on reading
programs. Rohrbeck et al. (2003) analyzed studies occurring in elementary school
Tutor Learning Has Been Observed Across Different Tutoring Formats
Tutoring formats can vary based on the age and knowledge gap between par-
ticipants and the nature of their roles. In cross-age tutoring, older and more
advanced students instruct younger novices. These roles typically remain “fixed”
because tutees cannot legitimately teach the more expert tutors, although role
reversals can be used as a game to support tutee engagement (Juel, 1996). Tutor
learning has been observed in cross-age programs with large and small age gaps,
such as college students tutoring first-graders (Juel, 1996) or fifth- and sixth-
graders tutoring second- and third-graders (Allen & Feldman, 1973; Sharpley
et al., 1983). Tutor learning has also been found with moderate gaps, such as high
school (Cloward, 1967; Rekrut, 1992) or middle school tutors (Jacobson et al.,
2001) with elementary school tutees.
In same-age tutoring, tutors and tutees are of a similar age or grade. Participants
are more likely to be actual peers than in cross-age settings because they attend the
same classes or interact outside of tutoring. Another issue is the relative ability of
the tutors and tutees. Cross-age tutors’ advanced age and grade creates a clear gap
between tutors and tutees. In same-age tutoring, tutors and tutees may possess
overlapping or complimentary expertise, enabling them to learn from each other.
As a result, same-age tutoring is often “reciprocal,” and students take turns tutor-
ing each other. Tutor learning has been demonstrated with both role arrangements.
Several studies using undergraduates have observed tutor learning in same-age,
fixed-role settings (Annis, 1983; Coleman, Brown, & Rivkin, 1997). Reciprocal
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Summary of reported effect sizes across format, population, and domain variables
Variables and Effect Sizes
P. Cohen, Scruggs, Mastropieri, Rohrbeck,Ginsburg-
Kulik,, and Mastropieri, Spencer, Scruggs, Mathes and Block, Fantuzzo
Category Kulik (1982) and Casto (1986) and Talbott (2000) Fuchs (1994) and Miller (2003)
Format Cross-age (.38) Fixed-role (.36) Cross-age (.38) Cross-age (.80)
Same-age (.28) Reciprocal (.63) Same-age (.32) Same-age (.47)
Fixed-role (.42)
Reciprocal (.34)
Duration <1 month (.56) <19 hours (.38)
1-5 months (.38) >19 hours (.32)
>5 months (.10)
Structure Unstructured (.32) Unstructured (.33)
Structured (.34) Structured (.30)
No training (.32)
Training (.34) Less autonomy (.30)
More autonomy (.94)
Grade Grades 1-6 (.35)
Grades 7-9 (.28) Grades 1-6 (.33)
Population General Special Special Special
education (.33) education (.59) education (.36) education (.42) <50% Minority (.51)
>50% Minority (.23)
Domain Math (.62); Math (.67) Math (.27); Reading (.26);
Reading (.21) Reading (.30) Science (.62); Social studies (.49)
Note. Effect size values are given in parentheses.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
tutoring has been studied more frequently and has shown favorable results across
many interventions (Fantuzzo, Riggio, Connelly, & Dimeff, 1989; Fuchs et al.,
1997; Greenwood, Delquadri, & Hall, 1989; King et al., 1998; Mastropieri,
Scruggs, Spencer, & Fontana, 2003; McMaster et al., 2006; Rohrbeck et al., 2003;
Topping, 2005). One caveat is that reciprocal tutoring may incorporate elements
of peer collaboration. Both participants teach and learn from each other, and this
role and task symmetry makes it difficult to determine whether gains are due to
tutoring or being tutored.
Meta-analyses (Table 1) show positive effects across formats, although the
effects seem low in magnitude overall. Comparisons also hint that all formats do
not support tutor learning equally. For example, peer tutors may learn more in
cross-age than same-age tutoring. Additional evidence surprisingly shows higher
effect sizes in shorter programs, which might imply that tutor learning benefits
occur early in the process, and nontutor controls catch up over time. Few differ-
ences were observed between programs with or without training and structure.
However, effects were larger in programs that gave students more autonomy (e.g.,
self-selected goals). We also know that training tutors to use strategies based on
constructivist theories of learning (e.g., Fuchs et al., 1997; King et al., 1998) leads
to impressive gains compared to less trained tutors. Thus, whereas the amount of
training or structure does not strongly influence tutor learning, the kind of training
or structure is critical (see also Fantuzzo, King, & Heller, 1992; Topping, 2005).
Finally, there is interesting evidence that students can benefit by teaching
computer-simulated peers that “learn” based on inputs from the student (Schwartz,
Blair, Biswas, Leelawong, & Davis, 2007). For example, in one system (Biswas,
Leelawong, Schwartz, Vye, & TAG-V, 2005), students tutor a computer program
named Betty about river ecology. If students provide inaccurate or insufficient
information, Betty cannot correctly answer the quiz questions. By working to teach
Betty a correct model and repair her mistakes, students developed a more inte-
grated knowledge of the material than those who used the same system without
a tutoring format (Biswas et al., 2005). Although this is not peer tutoring, it is
impressive that adding a peer-tutoring-like format to an intelligent tutoring system
increased students’ learning gains.
Tutor Learning Has Been Observed With Students of Diverse Backgrounds
Research on tutor learning has found benefits for tutors from all age groups,
including college (Annis, 1983; Topping, 1996), high school (Cloward, 1967;
Morgan & Toy, 1970), middle school (Jacobson et al., 2001; King et al., 1998),
and elementary school (Fuchs et al., 1997; Sharpley et al., 1983). Effect sizes have
not been reported for all grades, but effects sizes for elementary and middle school
tutors were small but positive (Table 1).
Some researchers have studied tutor learning in populations that often experi-
ence academic challenges. Researchers have shown positive outcomes for tutors
from underprivileged backgrounds and/or living in urban areas (Greenwood et al.,
1989; Jacobson et al., 2001; Rohrbeck et al., 2003). Robinson et al. (2005) noted
that African American tutors sometimes learned more from math peer tutoring
than White students. Rohrbeck et al. (2003) similarly found that programs with
greater than 50% minority enrollment showed a larger mean effect size than pro-
grams with low enrollment. Studies focusing on special education tutors with
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
learning disabilities or behavior disorders have also noted positive outcomes
(Cook et al., 1986; Fuchs, Fuchs, & Kazdan, 1999; Klingner & Vaughn, 1996;
Mastropieri et al., 2000; Mathes & Fuchs, 1994; Scruggs & Osguthorpe, 1986).
In sum, the benefits of peer tutoring appear to be highly inclusive, with the poten-
tial to support learning for a diverse sample of students.
Tutor Learning Has Been Observed Across Different Subject Matter Domains
Tutor learning has been observed in reading (Juel, 1996; Klingner & Vaughn,
1996; Mathes & Fuchs, 1994; Rekrut, 1992), math (Fuchs et al., 1997; Robinson et
al., 2005; Sharpley et al., 1983), science (Coleman et al., 1997; Fantuzzo et al., 1989;
King et al., 1998; Topping, Peter, Stephen, & Whale, 2004), social studies (Annis,
1983; Mastropieri et al., 2003), and other domains. As with comparisons of tutor-
ing format, tutor learning effect sizes seem to differ across subject matter domains.
Tutors seem able to learn no matter what subject matter they are teaching, but math
and science programs may exhibit stronger gains than reading programs.
Understanding the Tutor Learning Effect
Prior research has established that tutor learning is a real phenomenon with
broad application across different tutoring formats, diverse student populations,
and a variety of domains. However, a rough average estimate based on available
effect sizes reveals that the magnitude of the tutor learning effect, approximately
.35, is rather modest. Although there is ample evidence that tutors can learn in a
variety of settings, such outcomes are not guaranteed (P. Cohen et al., 1982;
Mathes & Fuchs, 1994; Renkl, 1995; Rohrbeck et al., 2003).
Program design and implementation can account for some of the successes or
failures of tutor learning. Potential gains might be limited for cross-age tutors who
teach material far below their level (Sprinthall & Scott, 1989) or if program dura-
tion is too short for tutors to improve their skills or knowledge (Topping & Bryce,
2004). Similarly, a lack of autonomy (Rohrbeck et al., 2003), proper training (King
et al., 1998), or suitable rewards (Fantuzzo et al., 1992) might inhibit learning.
Treatment fidelity and assessment are also crucial. Well-designed interventions
may fail if participants ignore their assigned tasks (Dufrene, Noell, Gilbertson, &
Duhon, 2005), and standardized tests may be less sensitive to learning gains than
assessments designed around program content (Cook et al., 1986; Rohbeck et al.,
2003). In sum, careful evaluations of the implementation and outcomes of peer
tutoring can reveal important conditions under which tutor learning is more or less
likely to be observed.
An alternative method for understanding tutor learning is the direct examina-
tion of peer tutors’ instructional and learning activities. This typically requires
recording and transcription of the tutoring sessions, followed by systematic cod-
ing of target behaviors and interactions. This can be time-consuming, but process-
based approaches can complement outcome-based approaches in important ways.
Process data enables detailed manipulation checks, allowing researchers to deter-
mine whether tutors followed instructions or if they modified their training in
unproductive or innovative ways. Process data can also lead to insights about
tutors’ learning-related behaviors. For example, tutors might use unexpected but
effective strategies, which educators could develop into new methods. Tutors may
also show behaviors that appear beneficial but are actually counterproductive. For
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
instance, tutors might ask many questions, but the questions are shallow or fail to
target key concepts. Thus, when studies do not show significant learning gains,
data on tutor and tutee behaviors might help researchers diagnose and respond to
the problems that hampered learning.
Such analyses can also be extended to the behaviors of control groups. Some stu-
dents naturally use self-explaining strategies when reading a text, and students in
the same classroom could collaborate to solve problems and explain difficult con-
cepts. Thus, the “failure” of tutors to surpass nontutors may sometimes reflect the
use of good strategies by controls, rather than the use of bad strategies by tutors. It
is also possible that tutors and controls could attain similar learning outcomes using
different strategies. Analyses of both tutors’ and nontutors’ learning behaviors could
help to further understand how peer tutoring differs from other learning settings and
how it creates or inhibits learning opportunities for tutors.
Process data also enables more powerful and flexible hypothesis-testing.
Whereas outcome data can test whether program manipulations facilitated learning,
process data can shed light on the underlying causes of these effects. Researchers
can explore how training methods affected target behaviors and whether target
behaviors were truly predictive of achievement. That is, researchers can determine
if learning gains were accompanied by specific strategies and if the occurrence of
those strategies were correlated with learning measures. Data can also be reanalyzed
to focus on different tutoring behaviors (i.e., feedback, questions, hints, etc.) or to
consider the same activities from different perspectives (e.g., individual inferences
versus reasoning dialogues). Analyses may reveal unexpected patterns that lead to
new hypotheses. For example, one might find that training in one set of strategies
suppresses other nontargeted behaviors, or discover differences in how target strate-
gies are used by female and male tutors.
Despite this potential of process data, research on tutoring has focused on out-
comes and implementation (Fogarty & Wang, 1982; Foot et al., 1990). When the
peer tutoring process has been studied, researchers have typically assessed only
how tutee outcomes were affected by tutor behaviors (Chi, Siler, Jeong, Yamauchi,
& Hausmann, 2001; Graesser et al., 1995). To fully understand how tutors learn,
researchers must evaluate tutor learning outcomes in conjunction with analyses of
their learning behaviors and processes. By combining approaches, we can develop
a richer knowledge of how and why tutors (and tutees) learn. In the remainder of
this review, we will consider research on the processes of tutor learning in the con-
text of explaining and questioning activities.
Processes of Tutor Learning
Analyses of expert or experienced tutors have demonstrated the complex nature
of the tutoring task. Effective tutoring can involve a variety of tactics, such as
explaining, questioning, assessment, and feedback (Lepper, Drake, & O’Donnell-
Johnson, 1997; McArthur, Stasz, & Zmuidzinas, 1990; Merrill, Reiser, Merrill, &
Landes, 1995; Shah, Evens, Michael, & Rovick, 2002; VanLehn, Siler, Murray,
Yamauchi, & Baggett, 2003). These tactics require tutors to use and apply their
subject matter knowledge and transform this knowledge in creative ways to teach
their tutee. As a result, the very factors that make tutoring cognitively demanding
may also contribute to tutor learning (Allen & Feldman, 1976; Bargh & Schul,
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
1980; J. Cohen, 1986; Foot et al., 1990; Gartner et al., 1971; King, 1998; Renkl,
1995; Topping, 2005).
Tutoring a peer has been hypothesized to help tutors metacognitively reflect upon
their own expertise and comprehension, and constructively build upon their prior
knowledge by generating inferences, integrating ideas across topics and domains,
and repairing errors. This reflective knowledge-building process is argued to result
in a better understanding. For example, Gartner et al. (1971) proposed,
In the cognitive area...the child having taught another may himself learn as
a result of a number of processes. He reviews the material; he has to organize,
prepare, illustrate the material to present it to his student; he may try to
reshape or reformulate it so as to enable his pupil to learn it and thus himself
sees it in news ways; he may need to seek out the basic character of the sub-
ject, its structure, in order to teach it better, and may thereby himself under-
stand it better. (p. 62)
However, peer tutors might adopt an alternative approach, called knowledge-
telling, defined as peer tutors “lecturing” or “stating what they already know” by
summarizing facts with little elaboration or self-monitoring. There is little reason
to expect tutors to experience significant advancement if they focus on delivering
their knowledge without ever questioning its quality or trying to improve upon it.
This distinction provides a straightforward framework for analyzing peer tutors’
behaviors in terms of their potential for learning.
In the following sections, we summarize our review methods and then consider
two hypothesized tutor learning activities: explaining and questioning. We first
explore how these activities could support tutors’ reflective knowledge-building
and improved understanding, then evaluate the plausibility and validity of these
hypotheses based on tutoring process data.
Review Methods
Target studies for this review included published research reporting quantita-
tive observations of tutor behaviors, ideally accompanied by objective measures
of tutor learning.
To identify relevant cases we searched the PsycInfo and ERIC databases using
terminology drawn from the established literature: peer tutoring, cross-age tutor-
ing, same-age tutoring, reciprocal tutoring, peer-assisted learning, learning-by-
tutoring, tutoring strategies, tutor training, explaining, questioning, scaffolding,
and so on. Our search included commonsense variants of these terms (e.g., peer
tutoring versus student tutoring) and theory-guided combinations (e.g., peer tutor-
ing and explaining).
We reviewed the titles and abstracts of thousands of references to locate peer
tutoring studies. We used the abstracts of this subset to determine if the publication
examined tutor learning and/or the tutoring process. We attempted to obtain all stud-
ies that presented both outcome and process data for peer tutors in an academic sub-
ject matter (e.g., math, reading, etc.). We also obtained process-only studies that
might usefully supplement our primary sources. We excluded studies in which stu-
dents taught each other games or how to build an object (Cooper, Ayers-Lopez, &
Marquis, 1982; Ludeke & Hartup, 1983; Verba, 1998; Wood, Wood, Ainsworth, &
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
O’Malley, 1995) and those that did not examine behaviors related to explaining or
questioning (Jason, Ferone, & Soucy, 1979; Niedermeyer, 1970).
Overall, our search yielded six references combining tutor learning data and
process data (entries 1-6 in Table 2). Only these studies can be used to assess
whether target tutor behaviors are linked to patterns of learning outcomes. Our
search also yielded 12 supplementary references focusing on the tutoring process
(entries 7-18). These studies cannot be used to directly examine how tutors learn,
but they can provide useful insights about the nature of peer tutors’ typical behav-
iors. Together, this sample of articles can be informative about what behaviors may
support or hinder tutor learning, how commonly they occur or do not occur during
tutoring. Table 2 summarizes the tutoring setting and format for data collection,
participant age and ability, the domain, and the types of data collected. Several
studies were embedded in reciprocal tutoring programs, but the actual process data
was collected in fixed-role sessions (Bentz & Fuchs, 1996; Fuchs et al., 1996,
1997; Fuchs, Fuchs, Bentz, Phillips, & Hamlett, 1994).
The overall lack of research is a weakness in the literature, and closer exami-
nation of the distribution of studies across formats, grades, and domains reveals
specific gaps. Most studies were conducted in same-age rather than cross-age set-
tings, and there were more elementary school studies than other grades. In addi-
tion, many programs focused on reciprocal tutoring, and there was a strong bias
toward math. We were not able to locate process studies in elementary school sci-
ence, middle school reading/writing, or math tutoring in higher education.
Similarly, a number of settings, such as elementary school cross-age tutoring,
higher education reading tutoring, science tutoring in high school, and so on, have
been addressed by only one publication.
An obvious consequence of this pattern is that our knowledge of how tutors
learn is incomplete and skewed toward a subset of possible environments. This is
important because previous reviews suggest that context can impact tutor learning
gains (P. Cohen et al., 1982; Rohrbeck et al., 2003), and these effects likely stem
from how the tutoring setting shapes the tutoring process. Unfortunately, there is
not enough data to assess how such design factors affect tutors’ learning behav-
iors. To further complicate matters, there is also little consistency in how peer
tutors’ behaviors are analyzed. Researchers often employ distinct coding schemes
that are not easily compared in a detailed manner. We will revisit these issues in
later sections.
The Explaining Hypothesis
Explanations are statements that convey information with the goal of making
some idea clear and comprehensible and are one of the primary ways in which
tutors teach their tutees. Tutors use verbal explanations to communicate key ideas,
principles, and relationships and to correct tutee mistakes and misconceptions.
These explanations may involve a variety of elements such as summarizing main
ideas, examples, analogies, and so on, and can be used to share known information
(e.g., a formula for calculating t tests) or make sense of new information (e.g.,
interpreting the results of a scientific study). Tutor explanations may also be
embedded in scaffolding interactions, in which tutors and tutees interact over suc-
cessive turns to gradually develop tutees’ knowledge and explanations. Not sur-
prisingly, the richness of explaining has led many to hypothesize that it contributes
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Summary of studies incorporating peer tutoring process data
Reference Format and Setting Participants Outcome Measures and Process Categories
1. Fuchs et al. (1997) Same-age, fixed-role; ES Grades 2, 3, 4; AA tutors, Comprehensive Mathematics Test; explanations,
mathematics; LD tutees questions, feedback, demonstrations
2. Fuchs, Fuchs, Kazdan, Same-age, reciprocal; reading; ES Grades 3, 4, 5; HA, AA, or Stanford Diagnostic Reading Test; error
and Allen (1999) classroom-based LD tutors in cross-ability correction, help-giving strategies, questions
3. Ginsburg-Block and Same-age, reciprocal; ES Grades 4, 5; AA tutors, Stanford Diagnostic Math Test; curriculum test;
Fantuzzo (1997) mathematics; classroom- LD tutees help and support behaviors, engagement, insults
4. Ismail and Alexander Same-age, reciprocal; physics; Teacher/researcher-designed written tests;
(2005) classroom-based HS students questions, responses/explanations
5. King et al. (1998) Same-age, reciprocal; MS Grade 7 Teacher/researcher-designed written tests;
human physiology; explanations, questions, feedback
6. Roscoe and Chi (2004) Same-age, fixed-role; human Undergraduates Researcher-designed written tests; explanations,
physiology; laboratory questions, responses, self-monitoring
7. Bentz and Fuchs (1996) Same-age, fixed-role; ES Grades 2, 3, 4; AA tutors, Prompts, helping, explanations, questions,
mathematics; classroom- LD tutees feedback
8. Chi, Siler, Jeong, Cross-age, fixed role; Nursing/biology majors; Explanations, questions, feedback, self-
Yamauchi, and human physiology; middle school tutees monitoring, scaffolding
Hausmann (2001) supplemental
9. Crahay. Hindryckx, Same-age, fixed role; ES Grades 5, 6; HA tutors, Explanations, questions, feedback,
and Lebe (2003) mathematics; supplemental LA tutees demonstrations, patterns of interaction
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
TABLE 2 (continued)
Reference Format and Setting Participants Outcome Measures and Process Categories
10. Duran and Monereo Same-age, fixed-role vs. HS Grade 10; HA tutors, Writing activities, questions, feedback,
(2005) same-age, reciprocal; LA tutees vs. same-ability scaffolding, patterns of interaction
writing; supplemental pairs
11. Fogarty and Wang Cross-age, fixed-role; math, MS students; elementary Explanations, questions, feedback, scaffolding
(1982) computers; supplemental school tutees
12. Fuchs, Fuchs, Bentz, Same-age, fixed-role; ES Grades 3, 4, 5; AA tutors, Explanations, questions, feedback, demonstration
Phillips, and Hamlett mathematics; LA tutees
(1994) classroom-based
13. Fuchs et al. (1996) Same-age, fixed-role; ES Grades 2, 3, 4; HA and Explanations
mathematics; classroom- AA tutors, LD tutees
14. Graesser and Person Cross-age, fixed role; algebra; Graduate methods tutors, Questions
(1994) research methods; undergraduate tutees;
supplemental HS algebra tutors, MS tutees
15. Graesser, Person, and Cross-age, fixed role; algebra; Graduate methods tutors, Explanations, questions, feedback, scaffolding
Magliano (1995) research methods; undergraduate tutees;
supplemental HS algebra tutors, MS tutees
16. Juel (1996) Cross-age, fixed-role; reading; College undergraduates; Scaffolding and modeling
supplemental Grade 1 tutees
17. MacDonald (1994) Same-age, fixed-role; various College undergraduates Explanations, questions, feedback, sequences of
topics; supplemental interactions
18. Topping, Campbell, Cross-age, fixed-role;
Douglas, and Smith mathematics; classroom- ES (ages 7 and 11); age 11 Mathematical words, procedural talk, strategic
(2003) based tutors, age 7 tutees talk, feedback
Note. “Classroom-based” tutoring was primarily integrated with classroom activities. “Supplemental” tutoring typically occurred outside of a classroom (e.g., after-
school tutoring or pull-out tutoring) to help tutees with assignments or areas of difficulty. “Laboratory” tutoring occurred as part of a laboratory experiment unaffili-
ated with a particular classroom or curriculum. HA = high achiever; AA = average achiever; LA = low achiever; LD = learning disability; AR = at-risk; ES = elementary
school; MS = middle school; HS = high school.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
to tutor learning (Bargh & Schul, 1980; J. Cohen, 1986; Coleman et al., 1997;
Gartner et al., 1971; King, 1998; Topping & Ehly, 2001).
When explaining, tutors must transform their prior knowledge into instructive
messages that are relevant, coherent, complete, and accurate (Coleman et al., 1997;
King, 1994; Leinhardt, 2001; Webb, 1989). However, although peer tutors may be
more advanced than their tutees, they are not likely to have expert domain knowl-
edge. True expertise takes time and much practice to develop (Chi, 2006), and so
tutors’ knowledge is probably novice-like in a number of ways. They may still pos-
sess knowledge gaps and misconceptions, and their knowledge of the material and
procedures may be largely implicit, fragmented, and poorly organized. This may be
true even though the tutor has shown good grades and test performance (Chi, 2000).
The tension between the demands of effective explaining and peer tutors’ less
than perfect knowledge may push tutors to engage in reflective knowledge-
building. For example, to generate relevant explanations that address key topics in
a meaningful way, tutors may have to think carefully about conceptual relationships
and prioritize information. Generating coherent explanations that are internally con-
sistent and follow a natural progression of ideas may require tutors to reorganize
their own disjointed mental models by forming or rearranging connections among
concepts. Thus, explaining may help tutors improve the organization and accessi-
bility of their knowledge (Bargh & Schul, 1980; King et al., 1998; Webb, 1989).
Similarly, striving to producing complete explanations that integrate key con-
cepts and principles could push peer tutors to reassess the depth and breadth of their
own prior knowledge. Generating accurate explanations may cause tutors to eval-
uate their own comprehension and ability to explain the material. Tutors may also
have to assess whether their explanations make sense and are logical. Thus,
explaining may help peer tutors to metacognitively recognize and confront their
own knowledge gaps and misconceptions. To the extent that peer tutors attempt to
repair these problems through elaboration and inferences, their understanding
should be enhanced (Coleman et al., 1997; King et al., 1998; Roscoe & Chi, 2004).
In some ways, this pattern of generating, reflecting upon, and revising expla-
nations may be analogous to self-explaining, in which learners make sense of new
information by using prior knowledge and inferences to explain it to themselves
(Chi, 2000). These self-explanations occur when learners realize they do not under-
stand or cannot solve a problem, perhaps because they are missing information
or a have a flawed mental model (Chi, 2000; Chi & VanLehn, 1991). Such
“impasses” may occur frequently for peer tutors who are trying to explain ideas
they have learned only recently or imperfectly. In these cases, tutors may have to
work out a correct explanation for themselves before (or while) conveying this
knowledge to the tutee.
Research on self-explaining has established the efficacy of this strategy (see
Chi, 2000; McNamara, 2004; Roy & Chi, 2005, for reviews). Analyses of self-
explaining while studying worked-out examples have shown that successful learn-
ers check their comprehension, generate inferences linking underlying principles
and operations for each step, explain relationships between steps, and anticipate
future steps (Chi, Bassok, Lewis, Reimann, & Glaser, 1989; Renkl, 1997). Less
successful learners monitor their comprehension rarely, focus on repetition and
paraphrasing, and do not integrate and apply domain principles. Researchers have
also studied self-explaining in learning from multiple representations (e.g., text and
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
diagrams). Multiple representations can facilitate learning by providing extra infor-
mation and supporting complementary reasoning processes, but learning in this
setting requires integration of information across representations (Ainsworth,
1999; Schnotz & Bannert, 2003). Self-explaining may facilitate this by supporting
inferences that combine information within and across representations in a coher-
ent mental model, and may make implicit links between representations explicit
(Butcher, 2006; Roy & Chi, 2005).
Peer tutors’ explanations may often be embedded in examples and multiple rep-
resentations. For example, chemistry tutors might walk their tutee through exam-
ples of combustion reactions to discuss how the reagents interact and energy is
released. In discussing the examples, the tutor may use formulas to calculate
energy, the periodic table, and drawings of molecular bonds to represent the reac-
tion processes in different forms. To maximize the usefulness of these tools, peer
tutors may need to deconstruct the examples and explain underlying principles and
relationships between steps. The tutors must also explain how information across
the different representations is related and complementary and explain their uses
and constraints.
An important limitation on these processes may be peer tutors’ effort and abil-
ity in monitoring their own understanding (King, 1998). A long history of research
shows that this activity can be challenging (Brown, Bransford, Ferrara, &
Campione, 1983; Glenburg, Wilkinson, & Epstein, 1982; Hacker, 1998; Maki,
Shields, Wheeler, & Zacchilli, 2005). Learners have trouble recognizing compre-
hension failures, assessing what they know or do not know, predicting test perfor-
mance, and so on. If peer tutors are unable or unwilling to evaluate the quality of
their own explanations, then reflective knowledge-building may be forestalled.
In sum, explaining plausibly provides rich opportunities for tutors to engage
in reflective knowledge-building processes that support learning. As tutors work
to produce quality explanations, they may recognize flaws in their own expertise,
leading them to reorganize their knowledge and generate inferences to repair
errors. In addition, common contexts for explaining, such as working through
examples and manipulating different representations, may further support produc-
tive sense-making behaviors. However, tutors’ lack of metacognitive awareness
may constrain their learning.
Evidence Regarding Explaining in Peer Tutoring
Do peer tutors’ explanations predict learning? To what extent do tutors actually
engage in reflective knowledge-building versus knowledge-telling? We address
these questions by reviewing studies that analyzed both tutor learning and behav-
ior and several studies that explored tutors’ explanations without learning data. For
clarity, we group studies roughly by grade level and domain (see also Table 2).
Elementary school tutors. The majority of process data on peer tutoring have been
collected in elementary school math settings. Much of this work has been done by
Fuchs and colleagues, who have assessed a variety of tutoring strategies. In this
research, reciprocal peer tutors are taught to give corrective feedback when tutees
make mistakes and to gradually reduce help over time to encourage tutee participa-
tion. The learning outcomes of tutors using this feedback strategy are then compared
to untrained tutors, or tutors trained in novel strategies hypothesized to make tutoring
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
more effective. Process data are collected in separate generalization sessions in which
representative students from each condition tutor a peer while being recorded.
Tutoring roles in these sessions are usually fixed instead of reciprocal.
Only one study in this paradigm presented both outcome and process data
related to explanations. Fuchs et al. (1997) trained second- through fourth-grade
peer tutors (same-age, learning-disabled tutees) to give “conceptual explanations”
that incorporated examples, discussed the meaning of concepts and strategies, and
contained questions about problem-solving steps. Scores on the Comprehensive
Mathematics Test showed that math gains were significantly higher for tutors
trained to give conceptual explanations than tutors trained only to give corrective
feedback. Generalization sessions further showed that explanation tutors offered
significantly more knowledge-building “conceptual explanations,” whereas feed-
back tutors produced knowledge-telling “lectures” that described concepts with-
out probing for tutee understanding or participation. However, some variability
was observed. Less trained peer tutors did generate some conceptual explanations,
at about equal proportion to their lecture-like explanations.
Several process-only studies provide additional information about elementary
school math peer tutors’ typical explaining behaviors. Bentz and Fuchs (1996) taught
tutors to use help-giving strategies. Tutors were average-ability second- through
fourth-graders who taught same-age, learning-disabled peers. Generalization ses-
sions showed that tutors trained in help-giving strategies offered more help than
tutors trained only to give corrective feedback. In terms explaining, there was also a
nonsignificant trend for feedback tutors to simply solve problems for tutees, point
out errors, or give procedural help on how to solve the problems. In contrast, help-
giving tutors were more likely to give knowledge-building conceptual explanations
about the meaning or necessity of certain operations. However, some conceptual
explaining was observed in both conditions.
Fuchs et al. (1994) examined the behaviors exhibited by peer tutors trained to give
corrective feedback versus untrained tutors. The tutors were average-ability third-
through fifth-graders working with lower ability, same-age peers. Generalization ses-
sions indicated that untrained tutors gave didactic, knowledge-telling explanations
in which they showed tutees how to solve the problem instead of letting tutees do the
work. Untrained tutors did produce a few knowledge-building explanations. In con-
trast, trained tutors were much more likely to use explanations and questions to stim-
ulate knowledge-building discussions and tutee participation. Tutees who worked
with a trained tutor also solved more problems correctly during the session.
The influence of the peer tutors’ prior ability was evaluated by Fuchs et al.
(1996), which analyzed the explanations of trained second- through fourth-grade
tutors who scored “high” versus “average” on a standardized math Computation
and Applications Test. Tutees were learning-disabled same-age peers. Analysis
of tutors’ explanations revealed that both high and average ability tutors relied
on “nonelaborated” (i.e., no help or just the answer) and “procedural” (i.e., restated
the problem or used manipulatives) knowledge-telling explanations. However,
high-ability tutors also generated some knowledge-building explanations that
“bridged procedural and conceptual” information (i.e., used real-life examples,
rephrased problems to emphasize missing features, checked answers for sensibil-
ity) or were “primarily conceptual” (i.e., explained underlying concepts).
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
In sum, the results reported by Fuchs et al. (1997) show that reciprocal peer
tutors who generated knowledge-building explanations scored higher on posttests
than tutors who generated knowledge-telling explanations. However, evidence
from that study and several process-only studies suggests that these behaviors did
not often occur naturally. Many tutors focused on knowledge-telling unless they
received explanation strategy training, although some untrained and less trained
tutors were able to engage in knowledge-building spontaneously.
Researchers using alternative methods have obtained comparable results.
Crahay, Hindryckx, and Lebe (2003) examined whether same-age, fixed-role peer
tutors (fifth- and sixth-graders) with minimal training utilized “retroactive” versus
“proactive” tactics. Retroactive tactics were those in which tutors encouraged tutee
participation and built upon tutees’ statements, errors, and questions. This knowledge-
building category included assessing tutee understanding and debating, explaining,
and synthesizing ideas. Proactive tactics were those in which the tutor demon-
strated steps or guided tutee actions in a procedural, knowledge-telling manner,
with little tutee problem solving or discussion of principles. Proactive tactics
accounted for 69% of tutors’ actions. Tutors told their tutees what to do and how
to do it and “rarely explained the rationale of the approaches and computations
they prescribed” (p. 208). Retroactive tactics accounted for only 16% of tutors’
behaviors. Argument, explanation, and synthesis statements made up 8% of
tutors’ behaviors. Thus, in this process study of untrained tutors, we again see a
strong predominance of knowledge-telling with a few spontaneous instances of
Ginsburg-Block and Fantuzzo (1997) compared learning outcomes for students
trained to complete math drills in reciprocal dyads versus individually. Reciprocal,
same-age tutors scored significantly higher than nontutors on a researcher-
constructed Curriculum-Based Computation Test. To understand how tutoring
experience influenced students’ behaviors, students from both conditions were
asked to participate in a recorded tutoring session. Peer tutors’ explanations were
analyzed together with several “help” behaviors, including “task presentation, ver-
bal or nonverbal feedback, providing answers, [and] guided questioning” (p. 137).
Frequencies of help behaviors did not differ across conditions; nor were they sig-
nificantly correlated with posttest scores. One reason for this pattern may be that
the coding category combined elements of knowledge-building (e.g., explaining)
and knowledge-telling (e.g., giving answers). Separate coding of activities may
have revealed more distinct relationships with learning.
One process-only study examined math cross-age tutoring (Topping, Campbell,
Douglas, & Smith, 2003), in which 11-year-olds tutored 7-year-olds. Explanations
were not directly coded, but researchers analyzed whether dialogue was “proce-
dural” (i.e., discussing steps to take or turn-taking) versus “strategic” (i.e., verbal-
izing one’s thoughts, or discussing the reasons and strategies underlying moves).
The cross-age tutors initially generated more procedural dialogue, which signifi-
cantly decreased as a result of the program intervention. Strategic dialogue also
significantly increased.
Middle and high school tutors. With elementary school tutors, we observed that
peer tutors relied on knowledge-telling explanations, which was somewhat offset by
tutor training. Although one might hope that older students with more educational
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
experience would be more sophisticated tutors, this does not appear to be the case.
The following studies show that middle- and high-school-age tutors were just as
prone to knowledge-telling as younger tutors. This research also further highlights
how training in reflective knowledge-building strategies improves, but does not elim-
inate, this trend.
One process-only study examined the behaviors of volunteer middle school
math and computer literacy cross-age tutors who taught elementary school tutees
(Fogarty & Wang, 1982). The authors found that the “instructional process that
occurred...was not characterized by a sophisticated or variable repertoire of
instructional behaviors” (Fogarty & Wang, 1982, p. 460). The most common
behavior (29.6%) was simply giving confirmation for correct responses, or noting
incorrect responses, without explaining. Tutors also frequently gave procedural
explanations consisting of directions for solving problems (27.7%). Explanations
that incorporated knowledge-building were less common; tutors did not often elab-
orate with examples or links to related concepts (7.6%) or provide conceptual
explanations (5.7%).
King et al. (1998) examined tutoring behaviors and learning with middle school
same-age, reciprocal tutors studying human physiology. Three groups were trained
to take turns giving explanations that made conceptual connections and discussed
underlying causes and mechanisms. Two of these groups were further taught to
generate questions to scaffold each others’ learning. This included “review” ques-
tions to discuss basic concepts and prior knowledge and “thinking” questions
requiring inferences and integration. Sequenced-inquiry tutors were taught to ask
these questions in a specific sequence (i.e., review, probing and hinting questions,
and thinking questions). Inquiry-only tutors were not taught this sequence.
Learning was assessed through teacher- and researcher-constructed exams testing
tutors’ ability to integrate and generate ideas. Immediate and delayed posttests
showed that the sequenced-inquiry tutors generally scored higher than inquiry-only
tutors, who in turn outperformed unprompted tutors.
The reciprocal tutors’ behaviors were consistent with test scores. All tutors ini-
tially generated knowledge-telling statements that restated facts from the materi-
als. Even after training, knowledge-telling statements were still more common than
knowledge-building statements, but significant group differences were found.
Sequenced-inquiry tutors generated more explanations that integrated, applied, and
generated knowledge, whereas unprompted tutors produced more knowledge-
telling statements. Inquiry-only tutors fell in the middle. Unfortunately, many
tutors reverted back to knowledge-telling in transfer sessions when prompt cards
were removed.
Only one study focused exclusively on tutor learning with high school tutors
(Ismail & Alexander, 2005), using methods adapted from King et al. (1998) for use
in a physics classroom. Scores on teacher and researcher designed exams indicated
that all three groups were equivalent before the intervention. However, posttests
showed that sequenced-inquiry tutors showed better understanding of the material
than inquiry-only and unprompted tutors, who were equivalent to each other. After
4 weeks, tests found that sequenced-inquiry and inquiry-only tutors were equiva-
lent in their recall of the material, and both remembered more than unprompted
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
Process data was collected in sessions before and after training, and with or with-
out prompt cards. The reciprocal tutors’ explanations were coded as low-level (i.e.,
knowledge-telling of facts and definitions) or high-level (i.e., knowledge-building
with connections between ideas and explaining underlying causes and principles).
All tutors initially produced few knowledge-building explanations, and unprompted
and inquiry-only tutors tended to produce the most knowledge-telling explanations.
After training, sequenced-inquiry tutors generated significantly more knowledge-
building explanations and less knowledge-telling explanations, whereas other tutors
continued to engage in more knowledge-telling than knowledge-building. This was
often true regardless of whether prompt cards were used. After 1 month, the groups
no longer differed significantly, but the overall quality of explaining was still
improved for all tutors.
Higher education tutors. The same behavioral patterns observed with primary
and secondary grade tutors are also seen with college tutors. For example, Graesser
and Person (1994) and Graesser et al. (1995) examined a variety of tactics used by
student cross-age tutors. Their data were drawn from graduate student research
methods tutors (undergraduate tutees; 54 one-hour sessions) and high school alge-
bra tutors (middle school tutees; 22 one-hour sessions). Tutors had no formal tutor
training in these domains but were able to successfully tutor their pupils in the
study. Both samples were similar in their patterns of behaviors, and were collapsed
into one data set. Tutor learning was not assessed in these studies.
To account for the efficacy of the inexperienced cross-age tutors, Graesser
et al. (1995) examined whether their tutors naturally used pedagogical moves sim-
ilar to expert tutors, such as facilitating active student learning, using real-world
examples, or Socratic questioning. Evidence for these tactics was sparse. The stu-
dent tutors were highly directive, setting the agenda for the sessions and topics of
discussion, and tutors rarely employed sophisticated tutoring techniques in a delib-
erate manner. The tutors covered a fairly consistent set of topics, examples, and
questions with each tutee. Moreover, these examples were not challenging cases
requiring deep reasoning, but concrete examples to demonstrate ideas or formulas.
Tutors tended to borrow these examples directly from textbooks or exams.
Within the confines of the tutors’ semiscripted lessons, however, scaffolding
dialogues occurred in which explanations were further developed. Graesser et al.
(1995) noted a five-step dialogue frame: tutor asks a question, tutee answers the
question, tutor gives feedback, tutor and tutee elaborate upon the tutee’s answer,
and tutor evaluates the revised answer. The fourth step is potentially rich in terms
of knowledge-building, because the tutors may deviate from their script to revise
or generate new explanations. Many of the tactics observed in this phase were
fairly shallow (e.g., prompting the tutee for key words, interrupting tutees to insert
correct answers, or summarizing main ideas), but the discussions often stretched
over multiple turns, with tutors and tutees gradually developing better explanations
in small steps. In addition, tutors offered thoughtful explanations in response to
tutee errors about 25% of the time. Thus, although the cross-age tutors rarely uti-
lized expert-like tutoring strategies, the interactive scaffolding process may have
naturally offered tutors the chance to gradually refine their explanations.
A similar portrait of tutoring was reported by Chi et al. (2001), in which nurs-
ing/biology majors taught middle school students about the circulatory system.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
These cross-age tutors were knowledgeable of the domain but had little tutoring
experience. Tutors were instructed to tutor in whatever manner they preferred to
prepare their tutees to answer recall and comprehension questions. Assessment
tests showed that tutees improved significantly in their ability to articulate core
domain knowledge and draw an accurate model of the circulatory system.
During the sessions, more than half of the tutors’ coded behaviors consisted of
didactic explanations. Regression analyses showed that these contributed mainly
to shallow tutee learning. The main tutor activity linked to deep tutee learning was
scaffolding, in which the tutor supported tutees’ constructive activities by giving
hints, decomposing and modeling skills, asking questions, and giving examples.
As noted above, these kinds of scaffolding moves might provide knowledge-
building opportunities for tutors. However, tutors’ questions and scaffolding
combined accounted for only 15% of their statements. Tutors made overt self-
monitoring statements only 2% of the time.
The influence of tutor-tutee interactions on tutors’ explanations was further
explored by Roscoe and Chi (2004), which combined process and outcome data.
Tutors were undergraduates with low knowledge about the topic of the human eye
and received no training. The peer tutors first learned about the eye by studying a
text, and then either taught this material to another student (instructional dialogue)
or generated a videotaped explanation that could be later used by a peer to learn
the material (instructional monologue). Thus, in the dialogue condition, tutees
could influence tutors’ explanations via questions and comments. Posttest com-
parisons revealed that dialogue tutors were better able to define key terms and cor-
rectly answer integration and application questions. Thus, interactions with a tutee
seemed to support stronger tutor learning.
Analyses of peer tutors’ explanations showed that all tutors mainly generated
knowledge-telling explanations in which they paraphrased and summarized text
facts. These explanations were not significantly correlated with posttest scores.
However, dialogue tutors engaged in significantly more reflective knowledge-
building than monologue tutors. When reviewing concepts, dialogue tutors built
upon prior knowledge by generating new inferences and elaborations. They also
sometimes made sense of confusing concepts by thinking aloud in a manner akin
to self-explaining. These elaborated reviewing and sense-making activities were
highly metacognitive (i.e., contained overt self-monitoring statements) and signif-
icantly and positively correlated with measures of fact recall and comprehension.
This was the only study that examined tutors’ overt metacognitive behaviors in
conjunction with learning outcomes.
One study researched the explanations of cross-age reading tutors (Juel, 1996).
Low-achieving student-athletes participated in a course in which they worked on
improving their reading skills while also tutoring at-risk elementary school chil-
dren. Tutoring activities during the sessions included reading books with the chil-
dren, writing short stories and journals, and practicing word and letter sounds. A
number of standardized measures were used to assess the tutee performance (e.g.,
Iowa reading comprehension and listening comprehension subtests and Wide
Range Achievement Test). These measures showed that about half of the tutees
made gains in reading comprehension, decoding, word recognition, and spelling,
but half gained much less. These subgroups were labeled as “successful” and “less
successful” tutoring dyads, respectively.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
The author hypothesized that successful tutoring dyads would show more fre-
quent scaffolding by the tutors. Their results indicated that the successful tutors
indeed gave significantly more scaffolded instruction during discussions of tutees’
reflective journals, letter-sound instruction, story writing, and reading compre-
hension. The amount of scaffolding was positively correlated with tutee reading
and listening comprehension scores. Less successful tutors spent less time scaf-
folding, suggesting that they tutored in a knowledge-telling fashion.
Finally, a study by MacDonald (1994) examined the behaviors of four college
tutors who each taught a different subject (chemistry, economics, writing, and criti-
cal thinking) to another undergraduate. Analyses focused on five peer tutoring behav-
iors: initiation of dialogue, replies to an initiation, evaluation of responses, addition
of information, and markers (i.e., continuers). Of primary interest is the “addition”
category, which described an utterance that “clarifies, illustrates, extends, or elabo-
rates the current topic” (p. 4). For tutors, this category accounted for only 30% of
their behaviors. However, analyses did not clearly distinguish between types of
explanations that could be categorized as knowledge-telling or knowledge-building.
Discussion. Few studies have quantitatively evaluated both tutor learning out-
comes and explaining behaviors, but this research shows that tutors’ explanations,
when they incorporate elements of reflective knowledge-building, do support more
effective tutor learning. Tutors trained to give explanations that include new exam-
ples, discuss underlying principles and applications, connect ideas, or otherwise
elaborate upon the source materials scored higher than less trained tutors on mea-
sures of comprehension, transfer, and delayed recall (Fuchs et al., 1997; Ismail &
Alexander, 2005; King et al., 1998). Similarly, Roscoe and Chi (2004) found that
some untrained tutors spontaneously monitored their own comprehension and gen-
erated knowledge-building explanations when interacting with a peer, which were
positively correlated with factual recall and comprehension. Only one study did
not replicate these patterns (Ginsburg-Block & Fantuzzo, 1997). However, this is
likely due to the combination of distinct behaviors (e.g., explaining and question-
ing) into a single category.
Despite the benefits of knowledge-building, peer tutors appeared to rely most
heavily on knowledge-telling explanations. Tutors tended to give explanations that
simply revealed answers, summarized facts, or described procedures with little elab-
oration or construction of knowledge. This knowledge-telling bias was demonstrated
by tutors across age, domains, and formats (Bentz & Fuchs, 1996; Chi et al., 2001;
Crahay et al., 2003; Fogarty & Wang, 1982; Fuchs et al., 1994, 1996; Graesser et al.,
1995; Roscoe & Chi, 2004). Knowledge-telling explanations were linked to signifi-
cantly lower test scores than knowledge-building explanations (Fuchs et al., 1997;
Ismail & Alexander, 2005; King et al., 1998; Roscoe & Chi, 2004).
Another interesting finding was the variability shown in tutors’ explaining activ-
ities (Bentz & Fuchs, 1996; Fuchs et al., 1994, 1997; King et al., 1998; Ismail &
Alexander, 2005). Even with training in knowledge-building strategies, knowledge-
telling often occurred more frequently than knowledge-building. Thus, although
training positively influenced tutors’ behaviors and learning, it was not the only
determinant of tutors’ activities. Some tutors continued to adopt a knowledge-telling
bias even after extensive training, and King et al. (1998) found that the knowledge-
telling bias returned when training cues were removed. However, this variability is
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
also seen in the other direction. Some untrained tutors occasionally engaged in
spontaneous knowledge-building (Crahay et al., 2003; Fogarty & Wang, 1982;
Graesser et al., 1995; Roscoe & Chi, 2004), naturally revising and improving their
explanations and knowledge.
Knowledge-telling may have some benefits. Tutors’ factual recall might
improve by rehearsing the material, and knowledge-telling could be a necessary
precursor to knowledge-building. Tutors may need to define basic concepts before
delving into deeper topics, and tutors’ initial knowledge-telling explanations may
provide the catalyst for subsequent self-monitoring and repair. Tutors’ explana-
tions can also expose tutees to information they might otherwise lack (VanLehn
et al., 2007), and Chi et al. (2001) found that tutors’ explanations aided shallow
tutee learning. Thus, although an obvious goal is to increase the overall amount of
reflective knowledge-building, the total elimination of knowledge-telling is prob-
ably not desirable. It is not clear at this time what an optimal ratio of knowledge-
telling to knowledge-building would be.
In sum, tutors benefit by generating reflective knowledge-building explanations,
but the pervasive knowledge-telling bias limits the potential for tutor learning.
Although programs have made strides in overcoming this bias, there is room for
improvement. An important challenge for future research is to better understand
those tutors who naturally engage in reflective knowledge-building. What factors
prompt tutors to adopt productive learning behaviors? One important element, tutors’
self-monitoring, has been largely neglected in the literature. However, we have seen
evidence that asking and answering questions might support tutor learning and
reflective knowledge-building (King et al., 1998; Roscoe & Chi, 2004). In the next
sections, we consider the impact questioning activities may have on tutor learning.
The Questioning Hypothesis
Questioning, which refers to both asking and answering questions, is another
ubiquitous tutoring activity. Tutors ask questions to introduce topics and to guide
and assess tutee thinking. Similarly, tutors must respond to tutee requests for infor-
mation or clarification and to tutees’ expressed confusion. Although these two
activities are distinct, both involve making inquiries of the material that may lead
to learning. Researchers have hypothesized that asking (Fantuzzo et al., 1989; King
et al., 1998) and answering questions (Bargh & Schul, 1980; J. Cohen, 1986; Foot
et al., 1990; King et al., 1998; Roscoe & Chi, 2004) should support tutor learning.
We first consider the value of the questions tutors ask to support tutee learning.
Tutors can make use of a variety of question types, such as review questions to
introduce topics and activate prior knowledge and comprehension questions to
assess tutee understanding. Tutors also use questions to provide subtle hints, such
as asking, “What about the numerator?” instead of simply telling the tutee that the
answer is incomplete. Tutors may stimulate deeper reasoning by asking questions
about underlying principles or hypothetical situations (e.g., “What would happen
if the denominator was zero?”).
Peer tutors’ questions are different from the information-seeking questions that
are the usual focus of student questioning research (van der Meij, 1998). Student
questioning generally follows a pattern in which the student recognizes a contradic-
tion, a lack of information, or otherwise experiences a metacognitive sense of “per-
plexity” (Graesser & McMahen, 1993; van der Meij, 1994). The student then puts
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
this problem into words, articulating a question to elicit an answer. This formulation
process can require organizing and integrating concepts as well as higher order rea-
soning. Questions may also be self-directed, in which the student draws upon his or
her own knowledge to answer the question (Gavelek & Raphael, 1985; Wong, 1985).
After receiving or generating an answer, the asker must evaluate the quality of the
response and integrate the new information into their existing knowledge.
Research on student questioning, especially self-questioning, has shown posi-
tive outcomes for this process, most often when the answers require some reflec-
tive knowledge-building (Dillon, 1998; Gavelek & Raphael, 1985; King, 1994;
Rosenshine, Meister, & Chapman, 1996; Wong, 1985). Most student questions
focus on basic facts and are often posed as yes-or-no verification questions; the stu-
dent simply requests the confirmation of an assertion she or he already believes is
correct. However, questioning is most beneficial when students ask deeper ques-
tions that require integration of new and prior knowledge, reorganization of men-
tal models, generation of inferences, and metacognitive self-monitoring. For
example, King (1992) has developed a technique in which students use question
stems to formulate reasoning questions for themselves or a partner, such as, “What
is a new example of . . . ?” or “What do you think causes . . . ?” (p. 113). King
found that generating and answering questions based on these stems, either alone
or with a collaborative partner, resulted in higher exam scores.
Tutors do not ask true information-seeking questions (Graesser et al., 1995)
because they already know the answers. However, peer tutors may benefit from con-
structing questions to help tutees think deeply about the material. For example,
tutors may devise questions that contrast concepts (“How are these two kinds of
problems different from each other?”), apply concepts (“Could you use Newton’s
Third Law to solve this problem?”), unpack causal relationships (“What would hap-
pen if this force were zero?”), and so on. To generate integration and reasoning
questions, tutors may have to also reflect upon the fundamental ideas, relationships,
and principles needed to produce a correct answer. Thus, question-asking may help
tutors further reinforce and organize their own understanding (Fantuzzo et al., 1989;
King et al., 1998).
It is also important to remember that peer tutors’ domain knowledge is likely to
contain gaps and misconceptions because they have only recently or imperfectly
learned the material. In some cases, tutors may believe they know the answer to a
question and discover that they are wrong. This may often happen when tutors try
to ask deeper questions that go beyond the source materials. In this case, the tutor
has inadvertently posed an information-seeking question. To the extent that tutors
attempt to identify or construct a correct answer to such questions, similar to self-
questioning (Wong, 1985), they may experience learning benefits.
It is not clear how often peer tutors pose deep questions, or questions they can-
not answer, and so it is uncertain how much tutor question-asking contributes to
tutor learning. A potentially more powerful learning and knowledge-building
opportunity may arise from responding to tutee questions. Tutors may initially rely
on favorite scripted or well-understood explanations and examples, but persistent
tutee confusion may force tutors to generate a revised or novel explanation to help
their pupil understand. Similarly, tutee questions might indicate areas where
tutors’ knowledge is flawed. Tutees might become confused because the tutors’
explanation is contradictory or incomplete. Thus, tutees’ questions may provide
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
a metacognitive cue, or source of cognitive conflict, leading peer tutors to revise
their own understanding and knowledge deficits to provide better explanations for
their tutees (Bargh & Schul, 1980; J. Cohen, 1986; Doise, 1990; Foot et al., 1990;
King et al., 1998; Roscoe & Chi, 2004).
Answering tutee questions in this manner may be analogous to responding to
questions in other settings, such as adjunct questions embedded in text. Research
on adjunct questions has examined a number of factors such as the location, amount,
and type of embedded questions (Hamaker, 1986). In terms of question type, such
research has found that questions prompting for elaboration, inferences, and logi-
cal reasoning lead to significantly stronger learning outcomes than shallower fac-
tual questions (Andre, 1979; Hamaker, 1986). For example, a recent study by
Peverly and Wood (2001) found that inference and main-idea questions inserted
into text improved reading comprehension for high school students with reading
Other work has examined embedded questions as scaffolding prompts in
computer-based learning environments (Azevedo & Hadwin, 2005; Craig, Sullins,
Witherspoon, & Gholson, 2006; Davis, 2003). Davis (2003) has used prompts such
as “Pieces of evidence we didn’t understand very well included...”or “In think-
ing about how these ideas all fit together, we’re confused about...”to stimulate
collaborative partners’ knowledge-building and science learning. Their results
show that this computer prompting improved students’ comprehension of the
material. Prompted students also monitored themselves more effectively, gener-
ated more principle-based explanations, and integrated new and prior knowledge.
As with questions embedded in text or computer materials, the nature and qual-
ity of tutees’ questions may impact tutors’ explanations. Shallow questions about
basic facts (“How many valves does the heart have?”) may not elicit reflective or
constructive responses from tutors. Peer tutors most likely have a firm grasp of the
material at this low level, and so these questions offer little challenge or need for
knowledge-building. In contrast, questions that tap deeper reasoning and applica-
tion (“What would happen if we didn’t have valves?”) may provide more oppor-
tunities for reflective knowledge-building. These questions are probably more
likely to touch upon disorganized, fragmented, or misconceived areas of the tutors’
subject matter knowledge (King, 1998; Roscoe & Chi, 2004).
Another issue is the obstacle of getting tutees to ask questions at all. Research
on question-asking in classrooms has shown that students often ask few questions
and even fewer questions that request meaningful answers (Good, Slavings, Harel,
& Emerson, 1987). This reluctance to ask questions may stem from students’ dif-
ficulty with identifying and expressing their confusions clearly, or from fear of
social embarrassment by admitting ignorance (Graesser & McMahen, 1993;
Graesser & Person, 1994; van der Meij, 1994). These same factors may also affect
tutees in the tutoring sessions, limiting the number or quality of questions asked.
In sum, two related questioning activities in peer tutoring, asking and answer-
ing questions, could support tutors’ engagement in reflective knowledge-building
and knowledge-learning. Formulating questions may help tutors reinforce impor-
tant concepts and relationships. Questions asked by tutees may further prompt
tutors to generate or revise their explanations and knowledge. One implication is
that responding to tutee questions, especially deeper questions, may help tutors
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
overcome the knowledge-telling bias. However, the overall frequency or quality
of questions may be an important obstacle.
Evidence Regarding Questioning in Peer Tutoring
Do tutor or tutee questions predict tutor learning? To what extent do peer tutors
or tutees ask questions that support reflective knowledge-building or knowledge-
telling? We address these questions in the same manner as explaining, with stud-
ies grouped roughly by grade level and learning domain (see also Table 2). Due to
the sparseness of research, the following sections consider both tutor and tutee
questioning together. Many studies addressed questioning at a very coarse grain
size and did not distinguish between different kinds of questions asked or how
tutors responded to tutee questions.
Elementary school tutors. In addition to examining tutors’ explanations, Fuchs
and colleagues have also considered questions asked by reciprocal peer tutors.
Fuchs et al. (1997) found that tutors trained to give knowledge-building explana-
tions attained higher test scores and asked more “procedural” questions than less
trained tutors. These were questions that focused on problem-solving steps and
encouraged tutees to participate. It is not clear whether these were shallow or deep
questions. In addition, comprehension-gauging questions were sometimes embed-
ded in tutors’ conceptual explanations. Process-only studies by Fuchs et al. (1994)
and Bentz and Fuchs (1996) similarly observed that trained tutors asked more ques-
tions than less trained or untrained tutors. These studies suggest that training can
increase the rate of tutor and tutee questioning, but the effects on different kinds of
questions are not clear.
A broader array of questions were examined by Crahay et al. (2003). Several
questions were fairly shallow in nature: asking tutees to read/write problem infor-
mation, perform a computation, or give an answer. These accounted for about 12%
of the same-age math tutors’ behaviors. In a few cases, tutors prompted tutees to
formulate an approach or strategy for solving the problem (6.3%). Such questions
could be productive if tutors reflected upon the underlying purpose and fitness of
different problem-solving strategies. Tutors also occasionally (3.6%) prompted
tutees to express their understanding or confusion, which might be helpful if the
tutors simultaneously thought carefully about likely misconceptions or sources of
confusion. Overall, questions with potential for deep reasoning were not common
in untrained tutoring. The tutees also rarely initiated discussions, expressed mis-
understanding, or asked questions of the tutor.
Interesting data on age effects have been obtained in reading tutoring. Fuchs et
al. (1999) examined outcomes for second- through fourth-grade reciprocal-reading
tutors (high-ability tutors paired with low-ability students) trained to use help-
giving strategies. All tutors were taught to summarize main ideas and make predic-
tions about the text. Tutors in the help-giving condition were also taught to use
questions to help tutees attend to key characters and events and choose important
ideas for summarizing. Scores on the comprehension subtest of the Stanford
Diagnostic Reading Test showed that help-giving tutors in the fourth grade scored
higher than their less trained counterparts. However, for second- and third-graders,
the less trained tutors outperformed help-giving tutors.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
Process data examined only the percentage of tutee errors corrected by tutors
using the helping strategies. Individual question types or strategies were not coded.
For second- and third-graders, use of the strategies was low overall: 0% for less
trained tutors and 39% for trained tutors. For the fourth-graders, strategy use was
higher: 39% for less trained tutors and 97% for trained tutors. Thus, it seemed that
younger reciprocal tutors were less able to implement the questioning strategies, and
may actually have found it counterproductive to take on this additional task. Older
tutors were able to implement and benefit from the strategies more effectively.
Middle and high school tutors. Middle school and high school tutors also tended
to ask few questions or questions that tapped deeper understanding. Fogarty and
Wang (1982) found that about 21% of their cross-age tutors’ behaviors consisted
of asking questions, and about 14% of tutees’ behaviors consisted of questions. No
data was presented on what kinds of questions were asked or how tutors responded
to tutee questions.
King et al. (1998) and Ismail and Alexander (2005) directly tested impact of
questioning on tutoring and tutor learning. In both studies, same-age, reciprocal
tutors were taught to give knowledge-building explanations. In addition, some
tutors were further trained to ask review questions to activate prior knowledge and
thinking questions to stimulate critical thinking. Written measures of comprehen-
sion and recall showed that questioning significantly improved tutors’ learning out-
comes, especially when tutors (i.e., sequenced-inquiry tutors) were trained to ask
questions in a specific sequence: review questions followed by thinking questions.
The quality of the reciprocal tutors’ explanations also improved, and in both stud-
ies the sequenced-inquiry tutors produced more knowledge-building explanations
than the other reciprocal tutors.
Analyses of the reciprocal tutors’ questioning behaviors showed a related pattern.
King et al. (1998) found that all tutors asked very few questions before training. After
training, sequenced-inquiry tutors asked significantly more comprehension, think-
ing, hinting, and probing questions than inquiry-only and unprompted tutors. In light
of tutors’ explaining and learning patterns, these results suggest that asking and
answering deep questions helped to stimulate more knowledge-building explana-
tions. No analyses were presented to directly show how participants responded to
each kind of question.
Ismail and Alexander (2005) similarly demonstrated low levels of questioning
at pretest, although reciprocal tutors asked some review questions about basic
facts. After training, the frequency of thinking questions increased for all tutors,
but sequenced-inquiry tutors asked thinking questions significantly more often
than inquiry-only and unprompted tutors. Again, these results suggest that deeper
thinking questions elicited knowledge-building explanations that contributed
to learning. Indeed, Ismail and Alexander (2005) reported a positive correlation
(r = .52) between the frequency of thinking questions and high-level explanations.
In one of the few studies to address writing tutoring, Duran and Monereo (2005)
compared same-age, fixed-role, and reciprocal tutoring with high school tutors.
Tutors worked together to write a narrative short story and a review of a book, film,
or music album. Analyses of tutors’ behaviors focused on the distribution of “col-
laborative” versus “tutorial” behaviors. Collaborative behaviors included asking
questions and building on partners’ statements via “splicing.” Tutorial behaviors
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
included hints and guiding tutees through problem-solving steps and incorrect
answers. Similar to Graesser et al. (1995), Duran and Monereo (2005) found that
82% of the tutoring activities occurred during this cooperative dialogue step. Forty-
seven percent of coded messages were splicing statements, 24% were guiding
statements, and 10% were hints, but only 4% of tutors’ actions were questions.
Some differences were found across fixed-role and reciprocal tutoring: Fixed-role
tutors engaged in more guided instruction (31%) than reciprocal tutors (18%) and
engaged in less splicing (41%) than reciprocal tutors (52%). These results suggest
that reciprocal tutoring may have supported more collaborative interactions.
However, as with studies in math and science tutoring, question-asking was infre-
quent overall.
The authors (Duran & Monereo, 2005) also presented a limited sequential analy-
sis of tutoring behaviors. In terms of questions, tutee questions (which accounted
for only 12% of their activities) tended to elicit hints and guiding statements from
the tutor, which in turn initiated cycles of additional hints, splicing, and guidance
from tutors and more tutee questions. Thus, as Graesser et al. (1995) noted, the col-
laborative step of tutorial dialogues may involve shallow tutor moves but might still
help tutors engage the material more deeply as they scaffold tutee learning. Duran
and Monereo’s (2005) analysis further suggests that tutee questions are an impor-
tant component of getting these productive discussions started. It is not clear whether
different types of questions were more or less effective in this process.
Higher education tutors. Several studies conducted with older tutors corrobo-
rate the trends with lower grades. The most comprehensive coding of questions
asked during tutoring was reported by Graesser and Person (1994). Approximately
16 question types were identified, divided into short-answer, long-answer, and
deep-reasoning categories. Shallow questions included verification (“That equals
9, right?”), completion (“What was the subject population in that study?”), quan-
tification (“How many levels does that variable have?”), and other questions about
basic facts. Long-answer questions involved conceptual information and some rea-
soning, such as definition (“What is an independent variable?”) and example ques-
tions (“What is an example of a descriptive statistic?”). Deep-reasoning questions
included questions requiring reasoning about cause-and-effect relationships, the
purpose of solution steps, and contradictions. Examples include consequence
(“What would happen if we doubled the slope?”) and procedure questions (“How
would you counterbalance those conditions?”).
Graesser and Person (1994) examined question-asking for graduate school
research methods tutors (undergraduate tutees) and high school algebra tutors
(middle school tutees). They found that their cross-age tutors asked about 100
questions per hour, and the most frequent were verification (45%), completion
(10%), quantification (9%), and procedure (9%) questions. Thus, yes-or-no ques-
tions asking for confirmation of an assertion were the most common kind of tutor
question. Overall, deep-reasoning questions accounted for only one fifth of tutors’
questions. The tutees asked only 25 questions per hour and also tended to ask ver-
ification (22%) questions most often, followed by procedural (21%), completion
(11%), and quantification (11%) questions. One third of tutees questions were
deep-reasoning questions. However, because tutees asked only 25 questions per
hour, only 8 deep questions were asked in a typical session.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
Cross-age tutors in Chi et al. (2001) asked an average of 63 questions during a
typical 2-hour session. Relative to all tutor behaviors, 6% of tutors’ statements
were questions about the subject matter and 4% were questions intended to assess
tutee comprehension. Tutees asked about 7.8 questions per hour, accounting for
only 7% of their behaviors.
A few studies have tried to examine how peer tutors respond to tutee questions.
MacDonald (1994) considered sequences of several behaviors. For our purposes,
“initiations” roughly correspond to questions, and we can treat MacDonald’s
“replies” as shallow, text-based responses to questions and “additions” as more
explanatory answers. Analyses showed that less than 25% of tutors’ statements and
20% of tutees’ statements were questions. Tutee questions received shallow replies
about 55% of the time and tutor additions in less 2% of cases. However, when tutees
offered explanations, tutors followed with explanations (27%) more often than shal-
low replies (<2%), and tutor explanations elicited tutee explanations. Thus, once an
explanatory discussion began, it tended to continue.
Only one study has directly examined how tutee questions influenced peer
tutors’ reflective knowledge-building activities (Roscoe & Chi, 2004). Peer tutor-
ing dialogues were coded to identify tutee questions and their subsequent tutor
responses. Each content-relevant question asked by tutees was coded as “deep” or
“shallow” depending on whether it contained or required a meaningful inference.
Tutors’ responses were similarly coded as deep or shallow, and as “metacognitive”
or “nonmetacognitive” depending on whether they contained overt self-monitoring
statements (e.g., “I’m not sure what that means”). Across all sessions, tutees asked
a total of 152 shallow versus 88 deep questions. Only 14% of shallow questions
received deep responses, whereas 41% of deep questions received deep responses.
Similarly, 21% of shallow questions elicited self-monitoring, but 48% of deep
questions prompted tutors to overtly evaluate their own thinking. Thus, the nature
of tutees’ questions had a meaningful impact on peer tutors’ reflective knowledge-
Discussion. A few process-outcome studies offer evidence that questioning
activities can support tutor learning and reflective knowledge-building. Reciprocal
tutors trained to ask and answer integration, application, and reasoning questions
outperformed less trained tutors on measures of comprehension and recall (Ismail &
Alexander, 2005; King et al., 1998). These tutors also generated more knowledge-
building explanations, which suggests that the questions may have elicited these
higher level explanations. Similarly, tutors trained to use question strategies to sup-
port reading comprehension outperformed less trained tutors, but this was true only
for fourth-graders, and not younger second- and third-grade tutors (Fuchs et al.,
1999). However, because all these studies involved reciprocal tutoring roles, one
cannot determine whether learning was most supported by asking or answering
questions when being the tutor or tutee.
Process-only studies examining sequences of tutor-tutee behavior provide sug-
gestive evidence that responding to questions could be helpful for tutors. Duran
and Monereo (2005) showed that tutee questions could help to initiate collabora-
tive dialogues in which tutors and tutees worked together to improve answers and
explanations (see also Graesser et al., 1995; MacDonald, 1994). Question quality
was an important factor. Roscoe and Chi (2004) found that shallow tutee questions
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
elicited shallow tutor responses, but deeper questions helped to support tutors’
reflective knowledge-building explanations. Thus, in some circumstances, tutee
questions can help to overcome the knowledge-telling bias. No studies have
focused on the specific or unique benefits of tutor question-asking.
Another important observation is that tutor and tutee questions seem to consti-
tute a relatively small proportion of their respective behaviors. Various studies
show a rate of questions ranging from a low 4% to about 20% for both tutors and
their tutees (Chi et al., 2001; Crahay et al., 2003; Duran & Monereo, 2005; Fogarty
& Wang, 1982; MacDonald, 1994). In addition, the questions that are asked are
mainly shallow rather than deep. Tutors tended to ask questions about basic facts,
definitions, or computations more frequently than questions requiring deeper rea-
soning and reflection (Crahay et al., 2003; Graesser et al., 1995; King et al., 1998).
Similar patterns were shown for tutee questions (Graesser et al., 1995; Roscoe &
Chi, 2004).
In sum, questioning has strong potential for tutor learning, but tutor and tutee
questions tend to prompt for knowledge-telling rather than knowledge-building.
As a consequence, most questions in peer tutoring may only serve to reinforce
rather than overcome the knowledge-telling bias. A few studies have shown that
training in elaborated questioning strategies can sometimes improve the number
and quality of questions during the tutoring sessions (Fuchs et al., 1994, 1997,
1999; Ismail & Alexander, 2005; King et al., 1998). However, more fine-grained
research on how different kinds of questions support learning, and how tutors ask
or respond to such questions, will be needed to further explore these findings.
Summary and Directions for Future Research
The belief that students benefit academically by tutoring has provided a long-
standing justification for the use of peer tutoring programs. In support of this
assumption, research has shown that there is a replicable tutor learning effect asso-
ciated with tutoring in various formats, student populations, and domains.
However, we also find that the magnitude of these gains are somewhat low. Based
on prior reviews (P. Cohen et al., 1982; Cook et al., 1986; Mathes & Fuchs, 1994;
Rohrbeck et al., 2003), we roughly estimated an effect size for tutor learning of
about .35.
Efforts to improve tutor learning have traditionally focused on program design,
leading to the development of a number of successful interventions. Less research
has examined tutors’ actual behaviors. As a result, the literature says more about
what interventions work than why. For this reason, studies such as those by Fuchs
et al. (1997), King et al., (1998), Ismail and Alexander (2005), and Roscoe and Chi
(2004) are critical. This research combines quantitative measures of learning with
process analyses, allowing us to determine with greater clarity how tutors learn.
Our review of the literature showed straightforward conclusions about tutor
learning in regards to explaining and questioning. Peer tutors benefit from explain-
ing when they integrate concepts and principles and generate new ideas through
inferences and reasoning. Tutors also need to monitor their own understanding and
repair confusions when they arise. Deeper reasoning questions asked by tutees and
tutors can help to support this reflective knowledge-building. Unfortunately, peer
tutors primarily summarize information with little elaboration or reflection, and
students’ questions tend to ask about basic facts. Even when trained, some peer
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
tutors seem to focus on delivering rather than developing their knowledge, often
resulting in disappointing tutor learning gains.
We also see interesting variability in peer tutors’ knowledge-telling and
knowledge-building behaviors. Tutors who receive extensive training in particu-
lar strategies nonetheless find ways to modify or undermine these techniques in
unexpected ways. Similarly, untrained tutors typically show a knowledge-telling
bias but can sometimes engage in impressive reflective knowledge-building activ-
ity spontaneously. Such variability may partly explain the wide range of learning
outcomes seen within and across tutoring programs. Peer tutors can approach their
tutoring tasks in dramatically different ways, and these choices create or inhibit a
variety of learning opportunities. For educators, researchers, and parents, these
findings underscore how process data can lead to important realizations about the
limits and potential of tutor learning, providing stronger foundations for the devel-
opment of effective future tutoring programs.
Recommendations for Future Research
Available research shows that tutors can benefit from explicit training in
knowledge-building strategies. For example, tutors can be taught about the qual-
ities of good explanations and questions and appropriate timing of these instruc-
tional activities. The works of Fuchs and King describe such training methods in
detail and demonstrate their effectiveness. In keeping with the theme of our
review, our concluding sections focus on how tutor learning and process research
might be improved and how these data might uncover reasons why peer tutors
engage in knowledge-building or show a knowledge-telling bias. In turn, these
data might inform new training methods that support the occurrence of knowl-
edge-building. We first offer general recommendations for future research, and
then briefly discuss several specific areas for further study and development.
Expanding and improving the research. Evidence from meta-analyses suggests
that tutor learning is not uniform across program design variables such as format
and domain, but the underlying reasons for these differences are not well under-
stood. In light of Topping’s (2005) taxonomy of peer learning design features, it is
clear that there is a huge variety of possible tutoring environments. The current lit-
erature is unprepared to say how complex interactions between these features
affect the process of tutor learning. We thus encourage researchers to conduct
process-outcome evaluations as a standard component of peer tutoring research.
A key direction for this research may be studies that analyze and compare
behaviors and learning across different program variables, such as tutoring format
and structure. For example, Duran and Monereo (2005) found that reciprocal tutor-
ing supported more collaborative interactions than fixed-role tutoring. Reciprocal
tutors were more likely to build off of each others’ statements, whereas fixed-role
tutors took a more directive role, providing hints and guidance to scaffold their
partner’s learning. Fogarty and Wang (1982) found that middle school cross-age
tutors of older elementary school children were more responsive and interactive in
their tutoring style than tutors who worked with younger children. More research
of this kind may show us how opportunities for tutor learning differ across pro-
grams and could help educators decide what forms of tutoring would be best to
implement in their circumstances.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
A related recommendation is to also collect process data on the behaviors of stu-
dents in nontutoring control groups. Peer tutors’ behaviors may be more or less
productive, and this is equally true of students studying alone (Renkl, 1997), work-
ing in cooperative groups (Webb et al., 1995), and so on. When tutor learning is
measured against nontutoring control conditions, process data from both groups
could be helpful in several ways. First, these data may be needed to argue that
tutors learn for the reasons they are hypothesized to learn. That is, when tutors out-
perform controls, we would hope to find that their learning strategies are in some
way superior to those of controls (e.g., higher proportion of knowledge-building
statements). These data may be even more critical when tutors and controls per-
form at the same level. Analyses could help to diagnose whether tutors failed to
learn due to a strong knowledge-telling bias, or if controls learned particularly well
because they spontaneously self-explained. A very interesting finding would be if
both groups learned equally well but utilized different knowledge-building strate-
gies. The more we can discover about learning opportunities and constraints unique
to peer tutoring, the better able we will be to design tutoring environments that
specifically support tutor learning.
Finally, in addition to an expanded base of process-outcome research, it will
also be crucial to simultaneously strengthen coding methods. We observed several
problems related to the level of analysis and inclusiveness of many coding
schemes. For example, some studies used broad categories that may have com-
bined distinct behaviors (Ginsburg-Block & Fantuzzo, 1997). As we have seen,
explaining and questioning (or other tutoring behaviors) may support learning in
distinct ways. Collapsing these behaviors into a single category might obscure their
specific or unique contributions to learning. In other studies, key behaviors were
coded separately, but different forms of those activities were not teased apart
(Duran & Monereo, 2005; Fogarty & Wang, 1982; Fuchs et al., 1994, 1997). Our
review demonstrated that explanations and questions can take on distinct reflec-
tive knowledge-building and knowledge-telling qualities, with a major influence
on learning. Future work may thus need to employ more fine-grained analyses of
target behaviors. Graesser and Person (1994) is a good example of how questions
can be coded in a meaningfully detailed manner. Similarly, Fuchs et al. (1996)
divided explanations into two forms of knowledge-telling (i.e., nonelaborated and
procedural) and knowledge-building (i.e., procedural-conceptual and conceptual).
These recommendations are challenging to implement but may be facilitated by
building upon coding techniques used in prior studies (Crahay et al., 2003; Duran
& Monereo, 2005; Fuchs et al., 1997; Graesser & Person, 1994; King et al., 1998).
Replications or extensions of previous methods would help us to compare findings
across studies. Researchers are also urged to develop new coding schemes that cap-
ture reflective knowledge-building and knowledge-telling aspects of more tutoring
activities. It should be noted that many tutoring programs collect learning data for
large numbers of students across multiple classrooms, but process data can be drawn
from a smaller set of “representative students” from each location. Studies by Fuchs
and colleagues are good examples of sample size management and feasibility.
Another possibility for expanding research might be to use teachable agents
(Schwartz et al., 2007) to test hypotheses about specific tutor behaviors. We
have noted that tutors’ and tutees’ behaviors can be highly variable. The use of
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
computer-simulated tutees might allow us to exert more control over the tutee side
of the interaction. For example, simulated tutees could be programmed to ask only
shallow questions or only deep questions, to make few mistakes or frequent mis-
takes, to have high prior knowledge or low prior knowledge, and so on. Thus, we
could assess more carefully how tutors respond to particular kinds of tutee inter-
action. One current limitation on this approach is that most teachable agent pro-
grams do not support spoken dialogue, which is the canonical medium for human
tutoring. However, research on spoken dialogue in tutoring systems is developing
rapidly (Litman et al., 2006).
Contexts for explaining and questioning. One way to apply finer-grained analy-
ses to the study of tutor learning is to look more carefully at specific instructional
activities, such as using multiple representations or correcting tutee errors. Instead
of treating all explanations as one kind, it may be beneficial to tease apart patterns
of knowledge-building and knowledge-telling associated with different activities.
For instance, we earlier described links between explaining and learning from
worked-out examples. With examples, students learn by explaining underlying prin-
ciples, solutions steps, and operations, which parallels the ways that tutors might
use examples in their explanations. However, Graesser et al. (1995) observed that
tutors borrowed examples directly from textbooks and reused the same examples
across students. Examples involving deeper reasoning and real-world applications
were not often selected.
Another interesting context is scaffolding, which we earlier defined as successive
tutor-tutee interactions that gradually develop the tutees’ knowledge over multiple
turns. Although knowledge-telling explanations and questions are predominant, low-
level activities might give rise to knowledge-building after several exchanges (Chi
et al., 2001; Duran & Monereo, 2005; Graesser et al., 1995; VanLehn et al., 2007).
Although little research examined how tutors or tutees responded to each others’
questions, a few analyses have shown that tutors’ scaffolding moves can stimulate
deeper thinking from tutees (Chi et al., 2001), and deeper tutee questions can elicit
more elaborated responses from tutors (King et al., 1998; Roscoe & Chi, 2004).
Analyses of tutors’ explanations, questions, and learning in specific contexts
could have immediate implications for tutor training. By identifying how peer
tutors utilize or squander learning opportunities inherent to each context, we can
develop tutor training methods that specifically target desirable behaviors and limit
less productive behaviors. For example, peer tutors may often avoid deeper exam-
ples and may not explain how key example steps are driven by underlying princi-
ples and operations. Thus, tutors’ reflective knowledge-building and learning may
be improved if tutors are trained in concrete strategies for selecting, decomposing,
and reasoning with examples with their tutees.
Underlying cognitive skills. Considering the intellectual demands of tutoring,
cognitive and developmental variables will likely influence tutors’ learning. Fuchs
et al. (1994) found that tutors with higher prior achievement offered more sophis-
ticated explanations than students of average prior achievement. This finding sug-
gests that a certain level of background knowledge, or particular cognitive skills,
may be needed to fully engage in reflective knowledge-building. However, more
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
research is needed to identify what teachable skills have the most influence on
tutors’ instructional behaviors and learning. With this knowledge, we could design
training methods that simultaneously introduce peer tutors to desired tutoring
strategies and support the development of the underlying skills needed to use those
strategies effectively.
One concrete example may be metacognitive monitoring skill, the ability to reflect
upon and regulate one’s own learning progress (Hacker, 1998; King, 1998). To repair
or revise their explanations, peer tutors must first be able to detect any mistakes or
contradictions that arise. For example, tutors may have trouble answering questions,
causing them to reconsider the quality of their own knowledge. Only one study has
examined tutors’ overt self-monitoring statements embedded in their explanations
and responses to questions (Roscoe & Chi, 2004). Their results suggested that self-
monitoring contributed to knowledge-building and tutor learning and also demon-
strated the feasibility of analyzing tutors’ overt metacognitive actions.
However, many students find error detection and comprehension monitoring
very challenging (Glenburg et al., 1982; Hacker, 1998; Maki et al., 2005), overes-
timating how well they know or remember information. Students also have diffi-
culty choosing appropriate strategies for overcoming comprehension failures. It
seems very likely that peer tutors will also have such problems. Future research
could build upon the results of Roscoe and Chi (2004) by testing additional aspects
of self-monitoring, such as the valence (i.e., “I know this” versus “I don’t know this”)
and accuracy of tutors’ self-evaluations. One could also examine how tutors respond
to these evaluations, to determine whether knowledge-telling or knowledge-
building are more likely to occur after particular kinds of self-judgments.
Peer tutors may benefit from learning specific strategies for monitoring their
own understanding and choosing self-regulation strategies for addressing errors
and misconceptions. Research has shown that self-monitoring strategies are learn-
able with instruction and practice and benefit learners in nontutoring settings
(Bielaczyc, Pirolli, & Brown, 1995; Ghatala, 1986; Palincsar & Brown, 1984).
Thus, incorporating such methods into peer tutoring training may be a potentially
strong method for supporting the “reflective” component of reflective knowledge-
building and, thus, significantly enhancing tutor learning.
As a final consideration, it must be noted that any training based on cognitive
skills and strategies will need to take age into consideration. Not surprisingly, sev-
eral studies have shown that the tutors’ age plays a role in their behaviors (Fuchs
et al., 1999; Ismail & Alexander, 2005; King et al., 1998; Ludeke & Hartup, 1983;
Wood et al., 1995). A difference of just one or two grade levels had an impact on
tutors’ ability to implement helping strategies in Fuchs et al. (1999). Similarly,
high school tutors in Ismail and Alexander (2005) were better able to use recipro-
cal questioning strategies unaided than were middle school tutors in King et al.
(1998). It may be necessary for researchers to directly compare how peer tutors
from different age groups implement similar strategies. These data could help us
identify aspects of peer tutoring that are more challenging to younger tutors, and
then modify strategies to ensure age-appropriateness.
Motivational beliefs. In addition to cognitive and developmental issues, tutors’
motivational beliefs may also influence their tutoring activities. Self-regulated
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
learning theory proposes that learners’ academic engagement is guided by their
self-perceptions and attitudes, and positive correlations have been observed
between self-reported strategy use, effort, interest, and self-efficacy (Pintrich,
2003). It makes intuitive sense that tutors’ feelings of capability and curiosity may
affect how they tutor. High-efficacy tutors (i.e., those who feel they can understand
and teach the material) may feel capable of generating deeper elaborations and
repairing their own errors and thus more readily do so when opportunities arise.
Low-efficacy tutors may skip over difficult topics to avoid failure and embarrass-
ment. Similarly, high-interest tutors (i.e., those who enjoy tutoring or the domain)
may find it rewarding to think deeply about the material and share these ideas with
tutees. Low-interest tutors may focus on knowledge-telling because they perceive
little enjoyment in talking about difficult or boring concepts.
Motivational variables are more typically studied as outcomes of tutoring rather
than predictors. Researchers have hypothesized that tutoring improves tutors’ self-
efficacy and interest (Benware & Deci, 1984; J. Cohen, 1986; Fantuzzo et al.,
1989, 1992; Gartner et al., 1971) but have found mixed results. Benware and Deci
(1984) reported that studying with a teaching expectancy increased enjoyment of
the material relative to studying for a test, but Renkl (1995) found a decrease in
interest and along with an increase in anxiety. Similarly, Fantuzzo et al. (1992,
1989) observed that reciprocal tutors increased in self-efficacy while decreasing in
anxiety, but Rittschof and Griffin (2001) were unable to replicate that finding.
P. Cohen et al.’s (1982) meta-analysis noted only a moderate positive effect of
tutoring on tutors’ attitude toward the material (ES = .42) and a small effect on self-
concept (ES = .12).
Few studies have examined the impact of motivation on tutors’ behavior.
Medway and Baron (1977) found that tutors who felt more capable of helping
tutees solve problems (i.e., higher efficacy) exerted more effort toward tutoring.
Gabarino (1975) observed that offering tutors a reward based on tutees’ perfor-
mance undermined tutors’ interest in tutoring, leading them to be more critical of
their tutees, feel less positive about tutoring, and be less effective instructors.
Tutors who were not offered a reward reported more enjoyment, were more sup-
portive and patient, and their tutees learned more. Thus, tutors’ attitudes and self-
perceptions seemed to affect their tutoring behaviors. These results are paralleled
by research on “teaching efficacy” (Tschannen-Moran & Hoy, 2001) showing that
classroom teachers’ sense of efficacy in teaching is associated with more persis-
tence and openness to innovation in the classroom.
The examination of tutors’ motivation as a predictor of behavior and learning
may help to disentangle prior mixed results regarding motivational outcomes of
tutoring. These analyses might also reveal a need for “motivation training” of peer
tutors in addition to strategy training. For example, it is likely that novice peer
tutors will feel anxiety about their ability to explain challenging concepts, or being
responsible for another student’s learning. Such worries might be offset by incor-
porating “efficacy-building” exercises into the training phase of the program. Peer
tutors could be given opportunities to practice tutoring on a low-stakes topic, such
as how to play a game or about a favorite hobby. Early tutoring successes may
increase tutors’ perceptions of their own capabilities and give them the confidence
to generate their own ideas during tutoring and tackle deeper topics. Thus, just as
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
it is critical to support positive motivational and emotional engagement for tutees
(Lepper et al., 1997), and indeed all learners (e.g., Meyer & Turner, 2006), we may
also need to attend to tutors’ own affective experiences.
Perceptions of tutoring. Peer tutors’ perceptions of the tutoring role may also
influence their choice of tutoring behaviors (Allen, 1983; Foot et al., 1990; Robinson
et al., 2005). Peer tutors are cast in the role of “instructor” and are expected to know
the subject matter and guide tutees toward a more accurate understanding. Some
tutors might interpret this to mean they should convey the material without embell-
ishment, leading them to develop a knowledge-telling bias. Others may believe that
elaboration and self-improvement are vital to promote tutee learning, leading them
to engage in reflective knowledge-building. In a review of children’s perceptions of
teaching (Foot et al., 1990), students focused more on social support (e.g., friendli-
ness, humor, and praise) than instructional factors (e.g., explain clearly and provide
variety) and did not include learning-by-teaching as a part of the tutoring role. Thus,
some tutors may not perceive the potential or need for tutor learning.
To our knowledge, no study has sought to elicit peer tutors’ beliefs about tutor-
ing and link these perceptions to their tutoring behaviors or learning outcomes, but
some related work has been done. Bierman and Furman (1981) manipulated the
recruitment rationale given to tutors. Some tutors believed they were recruited based
on expertise, whereas others thought they were selected based on appearance, at ran-
dom, or were given no reason. Tutors given a concrete reason (i.e., expertise or
appearance) showed a more positive self-concept and attitude than the other condi-
tions. In collaborative learning, Webb, Nemer, and Ing (2006) have shown that
peers emulated their teachers’ instructional style. Teachers presented calculations
and answers rather than elaborated explanations and did not encourage students to
verbalize their thoughts and ask questions. When peers collaborated, help-givers
similarly focused on giving unelaborated answers, doing the work for their partners,
and rarely checking partners’ understanding.
Research on peer tutors’ role perceptions could help us to discover whether stu-
dents have beliefs about tutoring that inhibit reflective knowledge-building. Such
perceptions might be challenged during tutor training. Peer tutors may need to be
encouraged to generate their own examples and analogies, because they believe
that they should not digress from the textbook. Another idea may be to explicitly
teach tutors that they are able to learn from tutoring. Many students might feel that
“good tutors already know everything” and that “only bad tutors make mistakes.”
By teaching peer tutors that learning from tutoring is possible and desirable, they
may be more willing to confront and repair their own knowledge gaps and mis-
conceptions. Peer tutors might even be informed that tutor learning is a required
component of good tutoring and that they should actively work to improve their
own understanding. Tutors could then model good learning strategies for their
tutees, in addition to teaching them the necessary concepts.
Tutor learning is a meaningful phenomenon that further validates peer tutoring
as a worthwhile educational intervention. Commonplace tutoring activities, such
as explaining and questioning, provide many opportunities for tutors to engage in
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
reflective knowledge-building. Unfortunately, tutors do not always take advantage
of these opportunities. By analyzing peer tutors’ actual behaviors and their con-
nection to learning outcomes, we may further extend the impact and effectiveness
of peer tutoring programs. Future research should also examine how the tutoring
and learning process is shaped by program features and student characteristics.
This knowledge can then be used to develop tutor training strategies that target
such factors. Research of this nature may lead to immediate and exciting applica-
tions for educators and students.
The authors would like to gratefully acknowledge the assistance and helpful feed-
back of our Learning Research and Development Center colleagues, in particular Kurt
VanLehn, Kevin Crowley, Marguerite Roy, and Kirsten Butcher. Sincere thanks are
also due to the Pittsburgh Science of Learning Center (NSF SBE-0354420, www. for their support of the first author during the completion of this work,
and to our anonymous reviewers, whose insightful comments were invaluable.
Ainsworth, S. (1999). The functions of multiple representations. Computers and
Education, 33, 131–152.
Allen, V. (1983). Impact of the role of tutor on behavior and self-perceptions. In
J. Levine & M. Wang (Eds.), Teacher and student perception: Implications for
learning (pp. 367–389). Hillsdale, NJ: Erlbaum.
Allen, V., & Feldman, R. (1973). Learning through tutoring: Low-achieving children
as tutors. Journal of Experimental Education, 42(1), 1–5.
Allen, V., & Feldman, R. (1976). Studies on the role of tutor. In V. Allen (Ed.),
Children as teachers: Theory and research on tutoring (pp. 113–129). New York:
Academic Press.
Andre, T. (1979). Does answering higher-level questions while reading facilitate pro-
ductive learning? Review of Educational Research, 49(2), 280–318.
Annis, L. (1983). The processes and effects of peer tutoring. Human Learning, 2,
Azevedo, R., & Hadwin, A. (2005). Scaffolding self-regulated learning and metacog-
nition: Implications for the design of computer-based scaffolds. Instructional
Science, 33, 367–379.
Bargh, J., & Schul, Y. (1980). On the cognitive benefits of teaching. Journal of
Educational Psychology, 72, 593–604.
Bentz, J., & Fuchs, L. (1996). Improving peers’ helping behavior to students with learn-
ing disabilities during mathematics peer tutoring. Learning Disability Quarterly, 19,
Benware, C., & Deci, E. (1984). Quality of learning with an active versus passive moti-
vational set. American Educational Research Journal, 21(4), 755–765.
Bielacyzc, K., Pirolli, P., & Brown, A. (1995). Training in self-explanation and self-
regulation strategies: Investigating the effects of knowledge acquisition activities on
problem-solving. Cognition and Instruction, 13(2), 221–252.
Bierman, K., & Furman, W. (1981). Effects of role and assignment rationale on attitudes
formed during peer tutoring. Journal of Educational Psychology, 73(1), 33–40.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
Biswas, G., Leelawong, K., Schwartz, D., Vye, N., & TAG-V. (2005). Learning by
teaching: A new agent paradigm for educational software. Applied Artificial
Intelligence, 19, 363–392.
Britz, M., Dixon, J., & McLaughlin, T. (1989). The effects of peer tutoring on mathe-
matics performance: A recent review. B. C. Journal of Special Education, 13(1),
Brown, A., Bransford, J., Ferrara, R., & Campione, J. (1983). Learning, remembering,
and understanding. In J. Flavell & E. Markman (Eds.), Handbook of child psychol-
ogy: Vol. 3. Cognitive development (4th ed., pp. 420–494). New York: Wiley.
Butcher, K. (2006). Learning from text with diagrams: Promoting mental model devel-
opment and inference generation. Journal of Educational Psychology, 98, 182-197.
Chi, M. (2000). Self-explaining expository texts: The dual process of generating infer-
ences and repairing mental models. In R. Glaser (Ed.), Advances in instructional psy-
chology (pp. 161–238). Mahwah, NJ: Erlbaum.
Chi, M. (2006). Two approaches to the study of experts’ characteristics. In N. Charness,
P. Feltovich, & R. Hoffman (Eds.), Cambridge handbook of expertise and expert
performance. Cambridge: Cambridge University Press.
Chi, M., Bassok, M., Lewis, M., Reimann, P., & Glaser, R. (1989). Self-explanations:
How students study and use examples in learning to solve problems. Cognitive
Science, 13, 145–182.
Chi, M., Siler, S., Jeong, H., Yamauchi, T., & Hausmann, R. (2001). Learning from
human tutoring. Cognitive Science, 25, 471–533.
Chi, M., & VanLehn, K. (1991). The content of physics self-explanations. Journal of
the Learning Sciences, 1(1), 69–105.
Cloward, R. (1967). Studies in tutoring. Journal of Experimental Education, 36(1),
Cohen, J. (1986). Theoretical considerations of peer tutoring. Psychology in the
Schools, 23, 175–186.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155–159.
Cohen, P., Kulik, J., & Kulik, C. (1982). Educational outcomes of tutoring: A meta-
analysis of findings. American Educational Research Journal, 19(2), 237–248.
Coleman, E., Brown, A., & Rivkin, I. (1997). The effect of instructional explanations
on formal learning from scientific texts. Journal of the Learning Sciences, 6(4),
Cook, S., Scruggs, T., Mastropieri, M., & Casto, G. (1986). Handicapped students as
tutors. Journal of Special Education, 19(4), 483–492.
Cooper, C., Ayers-Lopez, S., & Marquis, A. (1982). Children’s discourse during peer
learning in experimental and naturalistic situations. Discourse Processes, 5,
Crahay, M., Hindryckx, G., & Lebe, M. (2003). Analysis of peer interaction among
children in a tutoring situation pertaining to mathematical problems of the multiply-
ing type. In E. de Corte, L. Verschaffel, N. Entwhistle, & J. van Merrienboer (Eds.),
Powerful learning environments: Unravelling basic components and dimensions
(pp. 197–211). New York: Elsevier.
Craig, S., Sullins, J., Witherspoon, A., & Gholson, B. (2006). Deep-level reasoning
questions effect: The role of dialog and deep-level reasoning questions during vic-
arious learning. Cognition and Instruction, 24(4), 565–591.
Davis, E. (2003). Prompting middle school science students for productive reflection:
Generic and directed prompts. Journal of the Learning Sciences, 12(1), 91–142.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
Dillon, J. (1998). Theory and practice of student questioning. In S. Karabenick (Ed.),
Strategic help-seeking: Implications for learning and teaching (pp. 171-193).
Mahwah, NJ: Earlbaum.
Doise, W. (1990). The development of individual competencies through social inter-
action. In H. Foot, M. Morgan, & R. Shute (Eds.), Children helping children (pp. 43-
64). New York: John Wiley.
Dufrene, B., Noell, G., Gilbertson, D., & Duhan, G. (2005). Monitoring implementa-
tion of reciprocal peer tutoring: Identifying and intervening with students who do
not maintain accurate implementation. School Psychology Review, 34(1), 74–86.
Duran, D., & Monereo, C. (2005). Styles and sequences of cooperative interaction in
fixed and reciprocal peer tutoring. Learning and Instruction, 15, 179–199.
Fantuzzo, J., King, J., & Heller, L. (1992). Effects of reciprocal peer tutoring on math-
ematics and school adjustment: A componential analysis. Journal of Educational
Psychology, 84, 331–339.
Fantuzzo, J., Riggio, R., Connelly, S., & Dimeff, L. (1989). Effects of reciprocal peer
tutoring on academic achievement and psychological adjustment: A componential
analysis. Journal of Educational Psychology, 81(2), 173–177.
Fogarty, J., & Wang, M. (1982). An investigation of the cross-age peer tutoring
process: Some implications for instructional design and motivation. Elementary
School Journal, 82(5), 450–469.
Foot, H., Shute, R., Morgan, M., & Barron, A. (1990). Theoretical issues in peer tutor-
ing. In H. Foot, M. Morgan, & R. Shute (Eds.), Children helping children (pp. 65–92).
New York: Wiley.
Fuchs, L., Fuchs, D., Bentz, J., Phillips, N., & Hamlett, C. (1994). The nature of students’
interactions during peer tutoring with and without prior training and experience.
American Educational Research Journal, 31, 75–103.
Fuchs, L., Fuchs, D., Hamlett, C., Phillips, N., Karns, K., & Dutka, S. (1997). Enhancing
students’ helping behavior during peer-mediated instruction with conceptual mathe-
matical explanations. Elementary School Journal, 97(3), 223–249.
Fuchs, L., Fuchs, D., Karns, K., Hamlett, C., Dutka, S., & Katzaroff, M. (1996). The
relation between student ability and the quality and effectiveness of explanations.
American Educational Research Journal, 33(3), 631–664.
Fuchs, L., Fuchs, D., & Kazdan, S. (1999). Effects of peer-assisted learning strategies on
high school students with serious reading problems. Remedial and Special Education,
20(5), 309-318.
Fuchs, L., Fuchs, D., Kazdan, S., & Allen, S. (1999). Effects of peer-assisted learning
strategies in reading with and without training in elaborated help-giving. Elementary
School Journal, 99(3), 201–219.
Gabarino, J. (1975). The impact of anticipated reward upon cross-age tutoring. Journal
of Personality and Social Psychology, 32(3), 421–428.
Gartner, A., Kohler, M., & Riessman, F. (1971). Children teach children: Learning by
teaching. New York: Harper & Row.
Gavelek, J., & Raphael, T. (1985). Metacognition, instruction, and the role of question-
ing activities. In D. L. Forrest-Pressley, G. E. MacKinnon, & T. G. Walker (Eds.),
Metacognition, cognition, and human performance (pp. 103–136). New York:
Academic Press.
Ghatala, E. (1986). Strategy-monitoring training enables young learners to select effec-
tive strategies. Educational Psychologist, 21(1&2), 43–54.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
Ginsburg-Block, M., & Fantuzzo, J. (1997). Reciprocal peer tutoring: An analysis of
“teacher” and “student” interactions as a function of training and experience. School
Psychology Quarterly, 12(2), 134–149.
Glenburg, A., Wilkinson, A., & Epstein, W. (1982). The illusion of knowing: Failure
in the self-assessment of comprehension. Memory and Cognition, 10(6), 597–602.
Good, T., Slavings, R., Harel, K., & Emerson, H. (1987). Student passivity: A study of
question asking in K-12 classrooms. Sociology of Education, 60(3), 181–199.
Goodlad, S., & Hirst, B. (1989). Peer tutoring: A guide to learning by teaching. New
York: Nichols Publishing.
Graesser, A., & McMahen, C. (1993). Anomalous information triggers questions when
adults solve quantitative problems and comprehend stories. Journal of Educational
Psychology, 85(1), 136–151.
Graesser, A., & Person, N. (1994). Question asking during tutoring. American
Educational Research Journal, 31(1), 104–137.
Graesser, A., Person, N., & Magliano, J. (1995). Collaborative dialogue patterns in nat-
uralistic one-to-one tutoring. Applied Cognitive Psychology, 9, 495–522.
Greenwood, C., Carta, J., & Hall, R. (1988). The use of peer tutoring strategies in class-
room management and educational instruction. School Psychology Review, 17(2),
Greenwood, C., Delquadri, J., & Hall, R. (1989). Longitudinal effects of classwide peer
tutoring. Journal of Educational Psychology, 81(3), 371–383.
Hacker, D. (1998). Self-regulated comprehension during normal reading. In D. Hacker,
J. Dunlosky, & A. Graesser (Eds.), Metacognition in educational theory and prac-
tice (pp. 165–191). Mahwah, NJ: Erlbaum.
Hamaker, C. (1986). The effects of adjunct question on prose learning. Review of
Educational Research, 56(2), 212–242.
Ismail, H., & Alexander, J. (2005). Learning within scripted and non-scripted peer-
tutoring session: The Malaysian context. Journal of Educational Research, 99(2),
Jacobson, J., Thrope, L., Fisher, D., Lapp, D., Frey, N., & Flood, J. (2001). Cross-age
tutoring: A literacy improvement approach for struggling adolescent readers. Journal
of Adolescent and Adult Literacy, 44(6), 528–536.
Jason, L., Ferone, L., & Soucy, G. (1979). Teaching peer-tutoring behaviors in first-
and third-grade classrooms. Psychology in the Schools, 16(2), 261–269.
Juel, C. (1996). Learning to learn from effective tutors. In L. Schauble & R. Glaser
(Eds.), Innovations in learning: New environments for education (pp. 49–74).
Mahwah, NJ: Erlbaum.
King, A. (1992). Comparison of self-questioning, summarizing, and note-taking review
as strategies for learning from lectures. American Educational Research Journal,
29(2), 303–323.
King, A. (1994). Guiding knowledge construction in the classroom: Effects of teach-
ing children how to question and how to explain. American Educational Research
, 31(2), 338–368.
King, A. (1998). Transactive peer tutoring: Distributing cognition and metacognition.
Educational Psychology Review, 10(1), 57–74.
King, A., Staffieri, A., & Adelgais, A. (1998). Mutual peer tutoring: Effects of structur-
ing tutorial interaction to scaffold peer learning. Journal of Educational Psychology,
90(1), 134–152.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
Klingner, J., & Vaughn, S. (1996). Reciprocal teaching of reading comprehension
strategies for students with learning disabilities who use English as a second lan-
guage. Elementary School Journal, 96(3), 275–293.
Leinhardt, G. (2001). Instructional explanations: A commonplace for teaching and
location for contrast. In V. Richardson (Ed.), Handbook of research on teaching (4th
ed., pp. 333–357). Washington, DC: AERA.
Lepper, M., Drake, M., & O’Donnell-Johnson, T. (1997). Scaffolding techniques of
expert human tutors. In K. Hogan & M. Pressley (Eds.), Scaffolding student learning:
Instructional approaches and issues (pp. 108–144). Cambridge, MA: Brookline.
Litman, D., Rose, C., Forbes-Riley, K., VanLehn, K., Bhembe, D., & Silliman, S. (2006).
Spoken versus typed human and computer dialogue tutoring. International Journal of
Artificial Intelligence in Education, 16, 145–170.
Ludeke, R., & Hartup, W. (1983). Teaching behaviors of 9- and 11-year old girls in
mixed-age and same-age dyads. Journal of Educational Psychology, 75(6), 908–914.
MacDonald, R. (1994). An analysis of verbal interaction in college tutorials. Journal
of Developmental Education, 15(1), 2–12.
Maki, R., Shields, M., Wheeler, A., & Zacchilli, T. (2005). Individual differences
in absolute and relative metacomprehension accuracy. Journal of Educational
Psychology, 97(4), 723–731.
Mastropieri, M., Scruggs, T., Spencer, V., & Fontana, J. (2003). Promoting success in
high school world history: Peer tutoring versus guided notes. Learning Disabilities
Research and Practice, 18(1), 52–65.
Mastropieri, M., Spencer, V., Scruggs, T., & Talbott, E. (2000). Students with disabil-
ities as tutors: An updated research synthesis. In T. E. Scruggs & M. A. Mastropieri
(Eds.), Educational interventions: Advances in learning and behavioral disabilities
(Vol. 14, pp. 247–279). Stamford, CT: JAI.
Mathes, P., & Fuchs, L. (1994). The efficacy of peer tutoring in reading strategies for
students with mild disabilities: A best-evidence synthesis. School Psychology
Review, 23(1), 59–80.
McArthur, D., Stasz, C., & Zmuidzinas, M. (1990). Tutoring techniques in algebra.
Cognition and Instruction, 7(3), 197–244.
McMaster, K., Fuchs, D., & Fuchs, L. (2006). Research on peer-assisted learning
strategies: The promise and limitation of peer-mediated instruction. Reading and
Research Quarterly, 22, 5–25.
McNamara, D. (2004). SERT: Self-explanation reading training. Discourse Processes,
38(1), 1–30.
Medway, F., & Baron, R. (1977). Locus of control and tutors’ instructional style.
Contemporary Educational Psychology, 2, 298–310.
Merrill, D., Reiser, B., Merrill, S., & Landes, S. (1995). Tutoring: Guided learning by
doing. Cognition and Instruction, 13(3), 315–372.
Meyer, D., & Turner, J. (2006). Re-conceptualizing emotion and motivation to learn in
classroom contexts. Educational Psychology Review, 18, 377–390.
Morgan, R., & Toy, T. (1970). Learning by teaching: A student-to-student compen-
satory tutoring program in a rural school system and its relevance to the educational
cooperative. Psychological Record, 20, 159–169.
Niedermeyer, F. (1970). Effects of training on the instructional behaviors of student
tutors. Journal of Educational Research, 64(3), 119–123.
Palincsar, A., & Brown, A. (1984). Reciprocal teaching of comprehension-fostering
and comprehension-monitoring activities. Cognition and Instruction, 1(2), 117–175.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
Peverly, S., & Wood, R. (2001). The effects of adjunct questions and feedback on
improving the reading comprehension skills of learning disabled adolescents.
Contemporary Educational Psychology, 26, 25–43.
Pintrich, P. (2003). A motivational science perspective on the role of student motiva-
tion in learning and teaching contexts. Journal of Educational Psychology, 95(4),
Rekrut, D. (1992, April). Teaching to learn: Cross-age tutoring to enhance strategy
acquisition. Paper presented at the 73rd annual meeting of the American Educational
Research Association, San Francisco.
Renkl, A. (1995). Learning for later teaching: An exploration of mediational links
between teaching expectancy and learning results. Learning and Instruction, 5,
Renkl, A. (1997). Learning from worked out examples: A study on individual differ-
ences. Cognitive Science, 21(1), 1–29.
Rittschof, K., & Griffin, B. (2001). Reciprocal peer tutoring: Re-examining the value
of a cooperative learning technique to college students and instructors. Educational
Psychology, 21(3), 313–319.
Robinson, D., Schofield, J., & Steers-Wentzell, K. (2005). Peer and cross-age tutoring
in math: Outcomes and their design implications. Educational Psychology Review,
17(4), 327–362.
Rohrbeck, C., Ginsburg-Block, M., Fantuzzo, J., & Miller, T. (2003). Peer-assisted
learning interventions with elementary school students: A meta-analytic review.
Journal of Educational Psychology, 95(2), 240–257.
Roscoe, R., & Chi, M. (2004, August). The influence of the tutee in learning by peer
tutoring. Presented at the 26th annual conference of the Cognitive Science Society,
Rosenshine, B., Meister, C., & Chapman, S. (1996). Teaching students to generate
questions: A review of the intervention studies. Review of Educational Research,
66(2), 181–221.
Roy, M., & Chi, M. (2005). The self-explanation principle in multimedia learning. In
R. Mayer (Ed.), Cambridge handbook of multimedia learning (pp. 271–286).
Cambridge: Cambridge University Press.
Schnotz, W., & Bannert, M. (2003). Construction and inference in learning from mul-
tiple representation. Learning and Instruction, 13, 141–156.
Schwartz, D., Blair, K., Biswas, G., Leelawong, K., & Davis, J. (2007). Animations of
thought: Interactivity in the teachable agent paradigm. In R. Lowe & W. Schnotz
(Eds.), Learning with animation: Research and implications for design. Cambridge:
Cambridge University Press.
Scruggs, T., & Osguthorpe, R. (1986). Tutoring interventions with special education
settings: A comparison of cross-age and peer tutoring. Psychology in the Schools,
23, 187–193.
Shah, F., Evens, M., Michael, J., & Rovick, A. (2002). Classifying student initiatives
and tutor responses in human keyboard-to-keyboard tutoring sessions. Discourse
Processes, 33(1), 23–52.
Sharpley, A., Irvine, J., & Sharpley, C. (1983). An examination of the effectiveness of
a cross-age tutoring program in mathematics for elementary school children.
American Educational Research Journal, 20(1), 103–111.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Understanding Tutor Learning
Sprinthall, N., & Scott, J. (1989). Promoting psychological development, math
achievement, and success attributions of female students through deliberate psy-
chological education. Journal of Counseling Psychology, 36(4), 440–446.
Topping, K. (1996). The effectiveness of peer tutoring in further and higher education:
A typology and review of the literature. Higher Education, 32, 321–345.
Topping, K. (2005). Trends in peer learning. Educational Psychology, 25(6), 631–645.
Topping, K., & Bryce, A. (2004). Cross-age peer tutoring of reading and thinking:
Influence on thinking skills. Educational Psychology, 24(5), 595–621.
Topping, K., Campbell, J., Douglas, W., & Smith, A. (2003). Cross-age peer tutor-
ing in mathematics with seven and 11-year olds: Influence on mathematical
vocabulary, strategic dialogue and self-concept. Educational Research, 45(3),
Topping, K., & Ehly, S. (2001). Peer assisted learning: A framework for consultation.
Journal of Educational and Psychological Consultation, 12(2), 113–132.
Topping, K., Peter, C., Stephen, P., & Whale, M. (2004). Cross-age peer tutoring of
science in the primary school: Influence on scientific language and thinking.
Educational Psychology, 24(1), 57–75.
Tschannen-Moran, M., & Hoy, A. (2001). Teacher efficacy: Capturing and elusive con-
struct. Teaching and Teacher Education, 17, 783–805.
van der Meij, H. (1994). Student questioning: A componential analysis. Learning and
Individual Differences, 6(2), 137–161.
van der Meij, H. (1998). The great divide between teacher and student questioning. In
S. Karabenick (Ed.), Strategic help-seeking: Implications for learning and teaching
(pp. 195–217). Mahwah, NJ: Erlbaum.
VanLehn, K., Graesser, A., Jackson, G., Jordan, P., Olney, A., & Rose, C. (2007).
When are tutorial dialogues more effective than reading? Cognitive Science, 31,
VanLehn, K., Siler, S., Murray, C., Yamauchi, T., & Baggett, W. (2003). Why do only
some events cause learning during human tutoring? Cognition and Instruction,
21(3), 209–249.
Verba, M. (1998). Tutoring interactions between young children: How symmetry can
modify asymmetrical interactions. International Journal of Behavioral Development,
22(1), 195–216.
Wagner, L. (1982). Peer teaching: Historical perspectives. Westport, CT: Greenwood
Webb, N. M. (1989). Peer instruction, problem-solving, and cognition: Multidisciplinary
perspectives. International Journal of Educational Research, 13, 21–39.
Webb, N., Nemer, K., & Ing, M. (2006). Small-group reflections: Parallels between
teacher discourse and student-behavior in peer-directed groups. Journal of the
Learning Sciences, 15(1), 63–119.
Webb, N., Troper, J., & Fall, R. (1995). Constructive activity and learning in collabo-
rative small groups. Journal of Educational Psychology, 87(3), 406–423.
Wong, B. (1985). Self-questioning instructional research: A review. Review of
Educational Research
, 55, 227–268.
Wood, D., Wood, H., Ainsworth, S., & O’Malley, C. (1995). On becoming a tutor:
Toward an ontogenetic model. Cognition and Instruction, 13(4), 565–581.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
Roscoe & Chi
ROD D. ROSCOE is currently a research associate in the Institute for Software Integrated
Systems and Department of Electrical Engineering and Computer Science at Vanderbilt
University, VU Station B #351679, 2301 Vanderbilt Place, Nashville, TN 37235-1679;
e-mail: He received his PhD in cognitive psychology from
the University of Pittsburgh. His research interests include peer tutoring, learning strate-
gies, and self-regulation of learning.
MICHELENE T. H. CHI is a professor in the Department of Psychology and a senior sci-
entist at the Learning Research and Development Center at the University of Pittsburgh,
3939 O’Hara Street, Pittsburgh, PA 15260; e-mail: Her research focuses
on ways of learning, such as tutoring, collaborative learning, or observing tutorial dia-
logues. She also studies how instructional intervention can overcome science miscon-
ceptions about emergent principles and concepts.
at ARIZONA STATE UNIV on April 12, 2013http://rer.aera.netDownloaded from
... Tutoring can be described as when "people who are not professional teachers are helping and supporting the learning of others in an interactive, purposeful, and systematic way" (Topping, 2000, p. 6). Substantial evidence indicates that human tutoring is effective, where both participants gain an understanding (Bloom, 1984;Topping, 2000;Chi et al., 2001;Topping et al., 2003;Roscoe & Chi, 2007). One-to-one tutoring is the most exclusive form of a teaching-learning situation, comprising one teacher and one learner. ...
... It is a method of cooperative learning where students are organized in pairs, assigned the roles of tutor and tutee, and given a common goal in a pre-planned teaching-learning situation (Duran & Monereo, 2005). Typically, the tutor has more advanced domain knowledge than the tutee, but the knowledge gap can be minimal (Roscoe & Chi, 2007). ...
... Tutoring has other benefits than teaching for the learner and the teacher. Tutors do not necessarily need to be experts in the subject domain; rather, it is considered favorable if the tutor is just slightly more knowledgeable than the tutee (Topping, 2000;Roscoe & Chi, 2007;Alegre et al., 2019). Students can be better tutors than adults since they have more recently learned the material and, therefore, relate more naturally to the problems a tutee may face (Duran & Monereo, 2005). ...
Full-text available
Human peer tutoring is known to be effective for learning, and social robots are currently being explored for robot-assisted peer tutoring. In peer tutoring, not only the tutee but also the tutor benefit from the activity. Exploiting the learning-by-teaching mechanism, robots as tutees can be a promising approach for tutor learning. This study compares robots and humans by examining children’s learning-by-teaching with a social robot and younger children, respectively. The study comprised a small-scale field experiment in a Swedish primary school, following a within-subject design. Ten sixth-grade students (age 12–13) assigned as tutors conducted two 30 min peer tutoring sessions each, one with a robot tutee and one with a third-grade student (age 9–10) as the tutee. The tutoring task consisted of teaching the tutee to play a two-player educational game designed to promote conceptual understanding and mathematical thinking. The tutoring sessions were video recorded, and verbal actions were transcribed and extended with crucial game actions and user gestures, to explore differences in interaction patterns between the two conditions. An extension to the classical initiation–response–feedback framework for classroom interactions, the IRFCE tutoring framework, was modified and used as an analytic lens. Actors, tutoring actions, and teaching interactions were examined and coded as they unfolded in the respective child–robot and child–child interactions during the sessions. Significant differences between the robot tutee and child tutee conditions regarding action frequencies and characteristics were found, concerning tutee initiatives, tutee questions, tutor explanations, tutee involvement, and evaluation feedback. We have identified ample opportunities for the tutor to learn from teaching in both conditions, for different reasons. The child tutee condition provided opportunities to engage in explanations to the tutee, experience smooth collaboration, and gain motivation through social responsibility for the younger child. The robot tutee condition provided opportunities to answer challenging questions from the tutee, receive plenty of feedback, and communicate using mathematical language. Hence, both conditions provide good learning opportunities for a tutor, but in different ways.
... As the benefits are being seen across studies (Brannagan et al., 2013;Christiansen and Bell, 2010;Irvine et al., 2019;McKenna and French, 2011;McKenna and Williams, 2017), then discussion around embedding peer teaching into the curriculum is warranted. Furthermore, Roscoe and Chi (2007) have highlighted the need to support reflective knowledge-building in higher education students. Educators should be fostering activities that promote explaining and questioning rather than the simple transmission of knowledge (Roscoe and Chi, 2007). ...
... Furthermore, Roscoe and Chi (2007) have highlighted the need to support reflective knowledge-building in higher education students. Educators should be fostering activities that promote explaining and questioning rather than the simple transmission of knowledge (Roscoe and Chi, 2007). ...
Full-text available
Background It is now a requirement that all qualified nurses act as practice supervisors and support student nurses' education in practice, hence preparing third-year students for this role is a priority. This study evaluates students' experiences of peer teaching in clinical skills setting from the perspective of these students taking up the supervisory role once they graduate. Method An evaluative survey was utilised to explore and understand student nurse participation in peer teaching. Seventeen students took part in a questionnaire containing closed and open questions. Results This research suggests that students who engaged in peer teaching gained confidence in their own skills, through the revision of their own skills and knowledge. It also triggered reflection upon continuous professional development and inspired students to consider a future career in teaching. Conclusion Peer teaching provides an opportunity to reinforce the students' knowledge, clinical and communication skills. It helps prepare them for the role of practice supervisor upon qualifying by building confidence and enhancing their teaching skills.
... This requires reliable identification of class members with specific profiles of skills and deficits to choose or construct appropriate learning tasks and materials for them (e.g., Park and Datnow 2017). While students with homogeneous skill profiles might be grouped to receive targeted interventions to meet specific task demands (such as additional explanations or worked examples; e.g., Paas and van Gog 2006), students with heterogeneous skill profiles might engage in reciprocal tutoring, thereby helping each other to enhance knowledge structures or fill knowledge gaps (e.g., Dioso-Henson 2012; Roscoe and Chi 2007). 3. Setting priorities for deliberate practice in different classes: Dependent on class-average proficiency levels, a teacher may choose to devote instructional time to extensive, reflective exercises for particular types of tasks that currently pose a major challenge for many students in one class but not in another (Bloom 1968;Fuchs et al. 2010;Lehtinen et al. 2017). ...
Full-text available
A critical limiting factor for adaptive teaching is the availability of diagnostic tools that allow reliable and valid assessments of students’ domain-specific skills in a way that produces detailed information for planning subsequent instructional strategies. The present study demonstrates how Cognitive Diagnosis Models (CDM) can deliver fine-grained diagnostic information on students’ skills in dealing with domain-specific tasks, using introductory accounting as an exemplary field of application. Based on data from a sample of 773 students from secondary business schools in Austria, statistical analyses that incorporated several criteria for evaluating model fit corroborate theoretical assumptions on distinct skills as multiple dimensions of accounting competence. Moreover, they illustrate that CDMs allow not only to quantify the shares of students who have mastered or still lack each accounting skill but also to identify individual skill profiles, which can serve as reliable classification criteria to distinguish homogeneous or heterogeneous ability groups among the learners. We conclude by discussing the practical implications of different diagnostic information obtained from CDM outputs for generic strategies of adaptive teaching, again with an illustrative focus on introductory accounting instruction.
We investigated the learning outcome of teaching an agent via immersive virtual reality (IVR) in two experiments. In Experiment 1, we compared IVR to a less immersive desktop setting and a control condition (writing a summary). Learning outcomes of participants who had explained the topic to an agent via IVR were better. However, this was only the case for participants who scored high on absorption tendency. In Experiment 2, we investigated whether including social cues in the task instructions enhances learning in participants explaining a topic to an agent. Instruction manipulation affected learning as a function of absorption tendency: Low-absorption participants benefitted most from being instructed to imagine they were helping a student peer pass an upcoming test, while high-absorption participants benefitted more when they were to explain the text to a virtual agent. The findings highlight the crucial role of personality traits in learning by teaching in IVR.
This paper considers whether scientists can improve their visual design abilities by participating in critiques. In design education, a critique is a class session where designers present their work-in-progress and receive feedback from faculty, peers, and invited critics. In this study, we show that an intervention consisting of (1) an introduction to visual principles, (2) an explanation of critique methodology, and (3) participation in a group critique led to a significant increase in both the quantity and quality of feedback that scientists provided on a set of figures. These findings indicate that critiques can be a valuable practice for scientists to integrate into their research labs.
Virtual Reality (VR) has been widely adopted for the creation of training experiences, since it appears to allow overcoming, especially for practical training, some of the limitations of real exercises. Many previous works focused on investigating the effectiveness of VR Training Systems (VRTS) in a variety of fields, but the efficacy of these systems is very task-dependent, and the best way to integrate them into existing learning paths has yet to be thoroughly investigated. The goal of the present paper is to explore the latter aspects considering a case study in the context of energy management in industry, and focusing on a measuring procedure that is part, e.g., of energy audit inspections. To this aim, a VRTS was developed and used to conduct two user studies: a first study designed to investigate the effectiveness of the devised system when used as an alternative or in combination with lecture and laboratory-based teaching experiences, and a second study aimed to compare two different approaches (traditional learning, TL, and learning by teaching, LBT) for exploiting the provided VR-based functionalities in a learning path. Experimental results showed that the use of the VRTS alone improved the participants’ performance compared to traditional experiences, and that LBT proved to be more effective that TL for practical learning purposes.
Cette thèse pluridisciplinaire avait pour but de comparer les interactions agent biologique-robot à celles entre deux agents biologiques de la même espèce chez de jeunes individus de deux espèces hautement sociales : l’humain et une espèce d’oiseau, le Diamant Mandarin (Taeniopygia guttata). Trois populations ont été testées : des enfants au développement typique (DT) âgés de 3 à 6 ans, des enfants présentant un Trouble du Spectre Autistique (TSA) léger âgés de 4 à 8 ans et des diamants mandarins âgés de 35 à 100 jours. Chez les enfants, nous avons comparé l’apprentissage de vocabulaire (uniquement pour le groupe DT), la connaissance des parties du corps et des émotions ainsi que la tendance à aider selon que la démonstratrice était une adulte non familière ou le robot humanoïde NAO féminisé (voix humaine) et renommé NAOMIE. Nos résultats montrent que les scores des enfants (DT et TSA) étaient comparables quelle que soit la démonstratrice, suggérant que le robot NAO peut être utilisé en situation d’interaction sociale auprès de jeunes enfants. Pour réaliser nos expériences avec les diamants mandarins, nous avons développé un robot-diamant mandarin interactif : MANDABOT. Nous avons comparé l’apprentissage du chant selon que les jeunes avaient été élevés en présence d’un mâle adulte non familier ou de MANDABOT. Nous avons trouvé que l’imitation des chants produits par les tuteurs était aussi élevée dans les deux groupes suggérant que MANDABOT peut être utilisé comme tuteur de chant. Nos trois études nous ont permis de valider des protocoles en utilisant une approche expérimentale innovante et prometteuse en psychologie du développement, en psychomotricité et en éthologie : la robotique sociale.
This study studied the types of peer scaffolding presented in scientific experimental activities. The study included 14 university students. For the experimental activity of ‘determining temperature changes using the meridian altitude of the sun,' information regarding experimental behaviors, thinking aloud, discourse, and retrospective interview data were collected and analyzed. A model was derived and utilized for experimental activities to analyze peer scaffolding, wherein students internally structured their experiences with the experimental activities. The results indicated that students utilized seven means of peer scaffolding: ‘demonstrating’, ‘assisting’, ‘monitoring’, ‘posing’, ‘questioning’, ‘explaining’, and ‘suggesting’. Three types of peer scaffolding emerged: task completion-, model elaboration-, and learning support-oriented scaffolding. Each type differed in purpose, main mean, and major explanation details. Additionally, this study has observed the level of the model for the experimental activities and the time that had been provided to understand the experimental procedure influenced the three peer scaffolding types. These outcomes indicate that providing ample time to students independently structure the experimental procedure and supporting other students during experimental activities are essential. Moreover, providing assistance to students in focusing on observable phenomena by reducing the cognitive load required to process the experimental procedure is crucial. Keywords: peer scaffolding, scientific experimental activities, students’ interaction, qualitative case study
Full-text available
The evidences and generalizations of the preceding literature have assured that no solitary study has been conducted on theeffect of Peer Tutoring on Performance of Slow Learners in Southern Punjab, Pakistan yet. The proposed study aimed at revealing the marvelous secrets about the “Effect of Peer Tutoring on Slow Learner‟s Performance.” Experimental design was opted to accomplish the study. A sample of 40 students was taken from Govt. Institute for Slow Learners, Layyah. A Self Designed Test was used to gauge the performance of slow learners in mathematics after implementing intervention plan. Statistical analysis entailed pretest and post- test findings and inculcations were analyzed by using Paired Sample T test. The inferences that SPSS exposed were significant. In a nutshell, numerical results evinced the hypotheses of proposed study. The value of p in Paired Sample T test which was (p=.000) claimed that there was notable difference in mathematics performance of experimental group after undergoing in peer tutoring session. To explore the gender disparity on the level of performance an independent t test was applied. The significant value (p=0.001) depicted by independent t test further armored the second hypothesis, „there may be gender difference in performance of mathematics after being exposed to peer tutoring‟. The study and its empirical authentications prompted the educators to used peer tutoring as an effective teaching strategy and assisted psychologists to implement peer tutoring as a therapeutic intervention to augment the learning abilities of the slow learner students. Keywords: Peer Tutoring,, Slow Learners
Full-text available
The evidences and generalizations of the preceding literature have assured that no solitary study has been conducted on theeffect of Peer Tutoring on Performance of Slow Learners in Southern Punjab, Pakistan yet. The proposed study aimed at revealing the marvelous secrets about the "Effect of Peer Tutoring on Slow Learner"s Performance." Experimental design was opted to Aqsa Batool, Gul E Zahra, Muhammad Rehan, Israr Ahmed 6482 accomplish the study. A sample of 40 students was taken from Govt. Institute for Slow Learners, Layyah. A Self Designed Test was used to gauge the performance of slow learners in mathematics after implementing intervention plan. Statistical analysis entailed pretest and post-test findings and inculcations were analyzed by using Paired Sample T test. The inferences that SPSS exposed were significant. In a nutshell, numerical results evinced the hypotheses of proposed study. The value of p in Paired Sample T test which was (p=.000) claimed that there was notable difference in mathematics performance of experimental group after undergoing in peer tutoring session. To explore the gender disparity on the level of performance an independent t test was applied. The significant value (p=0.001) depicted by independent t test further armored the second hypothesis, "there may be gender difference in performance of mathematics after being exposed to peer tutoring". The study and its empirical authentications prompted the educators to used peer tutoring as an effective teaching strategy and assisted psychologists to implement peer tutoring as a therapeutic intervention to augment the learning abilities of the slow learner students.
Full-text available
Human one-to-one tutoring has been shown to be a very effective form of instruction. Three contrasting hypotheses, a tutor-centered one, a student-centered one, and an interactive one could all potentially explain the effectiveness of tutoring. To test these hypotheses, analyses focused not only on the effectiveness of the tutors' moves, but also on the effectiveness of the students' construction on learning, as well as their interaction. The interaction hypothesis is further tested in the second study by manipulating the kind of tutoring tactics tutors were permitted to use. In order to promote a more interactive style of dialogue, rather than a didactic style, tutors were suppressed from giving explanations and feedback. Instead, tutors were encouraged to prompt the students. Surprisingly, students learned just as effectively even when tutors were suppressed from giving explanations and feedback. Their learning in the interactive style of tutoring is attributed to construction from deeper and a greater amount of scaffolding episodes, as well as their greater effort to take control of their own learning by reading more. What they learned from reading was limited, however, by their reading abilities.
To examine co-operative learning between pairs of college students in the field of education, the reciprocal peer tutoring (RPT) procedure was used in two experiments with 97 graduates (Experiment 1) and 100 undergraduates (Experiment 2). Contrary to initial studies using a different population of college students, results in both experiments indicated that RPT failed to improve students' understandings of course material compared to an individualised study task. In addition, RPT neither increased students' feelings of self-efficacy nor did it decrease students' levels of test anxiety relative to the control condition. However, participants overwhelmingly reported RPT to be helpful for studying the course content. Results suggest careful consideration of ecological validity in research and expected gains in practice.
The need for a theoretical analysis of the peer tutoring process has been recognized by reviewers of the research in this area. This paper addresses itself to this need by discussing the process of peer tutoring from the various psychological, educational, and social perspectives. The main areas covered in the paper are: definitions of the peer tutoring process, the psychological and educational processes in peer tutoring as a learning and teaching experience, and the conceptualization of peer tutoring in terms of a cooperative social system and as a group reward structure. This theoretical discussion provides both the researcher and practitioner with a framework for studying and implementing peer tutoring for enhancing specific academic and social goals.
The purpose of this article is to discuss peer tutoring strategies as an effective class of peer-mediated procedures for both classroom behavior management and direct instruction. In this article, we discuss the need for alternative classroom procedures, review relevant research, discuss recent advances in procedures, and identify implications.
Before and after testing on the Wide Range Achievement Test (WRAT) over a four month period assessed gains for student tutors and their pupils in a rural school system. 13 tutored pupils (grades 2 to 5) showed a mean net growth advantage of from 3 to 5 months on WRAT subtests over 14 comparable control untutored pupils. 10 student tutors (grades 8 to 12) showed a mean 9 month edge over controls, a gain of 13 months achieve-over 10 comparable control non-tutors. On the three WRAT subtests all experimental means exceeded control means, but only the reading subtest was significant at the 5% level (tutors held a mean 9 month edge over controls, a gain of 13 months achievement in 4 months). Implications discussed include suggesting a much more institutionalized role reversal between teacher and student, student tutoring, and an educational cooperative with graduated salaries and personal involvement for all participants.