Contents lists available at ScienceDirect
Learning and Instruction
journal homepage: www.elsevier.com/locate/learninstruc
Do those who talk more learn more? The relationship between student
classroom talk and student achievement
Klara Sedova∗
, Martin Sedlacek, Roman Svaricek, Martin Majcik, Jana Navratilova,
Anna Drexlerova, Jakub Kychler, Zuzana Salamounova
Department of Educational Sciences, Faculty of Arts, Masaryk University, Arna Novaka 1, 602 00, Brno, Czech Republic
A B S T R A C T
There have been efforts to investigate the link between classroom talk and student achievement for some time. However, studies considering individual student
participation in classroom talk have thus far been rare. The research reported in this study was carried out on 639 ninth grade students at Czech middle schools.
Observations took place in language arts lessons; talk time and the number of utterances with reasoning were recorded for each student. Achievement was measured
using a standardized reading literacy test.
The results confirmed a strong link between a given student's talk time and number of utterances featuring reasoning and that student's achievement. As for
student talk time, a connection at the classroom level was also identified – students in talkative classrooms had better results. However, there was not a connection
between utterances with reasoning and better results at the classroom level. A positive link between individual participation and achievement was observed in all
students regardless of socio-economic background or gender.
1. Introduction
In recent decades, classroom talk and its contribution to learning
have become a key topic in the educational sciences, as well as in
philosophy (Kennedy, 2014), psychology (Marková, 2003), sociology
(Gurevitch, 2001), and linguistics (Skidmore & Murakami, 2012). There
is relatively wide agreement in the scientific community that there is a
relationship among talking, thinking, and learning (Resnick, Asterhan,
& Clarke, 2017). This assumption has given rise to a range of concepts
collectively referred to as dialogue-intensive or talk-intensive pedagogies
(Snell & Lefstein, 2018; Wilkinson, Murphy, & Binici, 2015). These
concepts share the fundamental premise that good teaching encourages
students to actively participate in classroom discourse.
In talk-intensive pedagogies, students speak often, and their utterances
are long and elaborate and include reasoning words; the utterances
are clearly directed to others; teachers ask open-ended questions
that encourage thinking; students do not repeat memorized facts but
argue and engage in thoughtful activity; teachers are responsive to
students, paying attention to students’ thoughts, building on them, and
developing them; and students listen to each other, asking questions
and having discussions (compare, for example, Alexander, 2006; Lyle,
2008; Lefstein & Snell, 2014).
There have been efforts to collect valid research evidence of any
changes in student achievement relating to the quality of classroom
dialogue (see van der Veen & van Oers, 2017). However, it has been
repeatedly pointed out that more studies of this kind are needed (Howe
& Abedin, 2013; Resnick et al., 2017). The aim of this study is therefore
to examine, in natural classroom environments of Czech middle
schools, the hypothesis that active student participation in classroom
talk may be connected with better student achievement.
1.1. Theoretical background
The interest in how students participate in classroom talk is
grounded in sociocultural theory, especially as it was presented by
Vygotsky. Vygotsky (1978) postulated that each psychological function
appears twice in a child's development – first on the social level (i.e., as
the child interacts with other people) and only later on an individual
level (in the form of internalized psychological processes). He believed
that speech and thought are closely interlinked and that children can
internalize and integrate what they have been able to talk about
(Vygotsky, 1978). It follows from this thesis that as more opportunities
for communication are created, children's internalization of knowledge
could become both faster and of a higher quality. The assumption that
communicating and thinking are closely related is widely accepted by
the scientific community. Sfard (2008) used the term commognition,
coined as a blend of communication and cognition, to emphasize the
indivisibility of these phenomena. Sfard (2008), with reference to Vygotsky,
understood learning as a way of participating in a certain discourse.
The Vygotskian perspective implies that language can be used
https://doi.org/10.1016/j.learninstruc.2019.101217
Received 28 May 2018; Received in revised form 23 May 2019; Accepted 8 June 2019
∗
Corresponding author.
E-mail address: ksedova@phil.muni.cz (K. Sedova).
Learning and Instruction 63 (2019) 101217
0959-4752/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
as a tool in service of intellectual development. This notion necessarily
raises the question of how to handle talk in schooling (Resnick et al.,
2017) – how to implement talk-intensive pedagogies in real classrooms.
It is simultaneously necessary to find research evidence that could
support the theory on which talk-intensive pedagogies are based. The
aim of this study is to provide some such evidence.
1.2. Research on the relationship between classroom talk and student
outcomes
Generally, there are three kinds of research studies designed to
present evidence concerning the relationship between classroom talk
and student outcomes: case studies of a particular teacher in a particular
class, intervention studies, and correlational large-scale studies.
All of them have tremendous potential to extend our knowledge of the
studied phenomena.
A number of case studies have been conducted investigating the
effects of talk-intensive discourse facilitated by a particular teacher in a
particular class (for example Billings & Fitzgerald, 2002; Kutnick &
Colwell, 2010; Mercer & Littleton, 2007; Scott, Ametller, Mortimer, &
Emberton, 2010; Forman, Ramirez-DelToro, Brown, & Passmore, 2017).
These studies have provided valuable insights into the potential of talk
and language to intensify learning. Yet, their results cannot simply be
taken as a reflection of how teaching in accordance with talk-intensive
pedagogies works at school in general, as these studies usually focus on
exceptionally competent and motivated teachers.
Intervention studies have been undertaken to implement a change
towards talk-intensive teaching and monitor its effects on student
learning (see, e.g., Reznitskaya et al., 2001; Andreassen & Bråten, 2011;
Kiemer, Gröschner, Pehmer, & Seidel, 2015; O'Connor et al., 2015;
Furtak et al., 2016; van der Veen, de Mey, van Kruistum, & van Oers,
2017; Alexander, 2018). It is quite difficult to summarize the findings
from these studies as they varied considerably – students were of different
ages, interventions were done in different courses, and different
outcomes were measured.1
These studies have indicated that classroom
discourse characteristics changed as a result of intervention and student
outcomes also subsequently increased (at least partially).
Other studies failed to confirm these links – there were no convincing
changes in classroom talk and/or in student outcomes. Murphy,
Wilkinson, Soter, and Hennessey (2009) carried out a meta-analysis of
42 intervention studies conducted in literacy classrooms and noted that
there was typically an increase in the proportion of student talk and a
decrease in the proportion of teacher talk, which is in accordance with
talk-intensive pedagogies. However, this change in classroom talk was
not always related to improved reading comprehension test results. This
implies that findings from intervention studies have not yet provided
clear answers to the question of whether talk-intensive pedagogies are
connected with better student achievement.
The third type of research conducted to explore the link between
classroom talk and student achievement is represented by correlational
large-scale studies. These studies have extraordinary potential to advance
the understanding of the workings of educational dialogue in
authentic classroom situations, as they are able to capture all kinds of
naturally occurring forms of classroom talk as well as variability in
learning outcomes. We will discuss this type of study in more detail
below.
Nystrand, Gamoran, Kachur, and Prendergast (1997) tested more
than 1100 eighth and ninth graders in language arts lessons. The students’
understanding of literature was tested with a written examination
and essay. In addition, classes were observed to determine the time
devoted to discussion, authentic questions, uptake, and high-level
evaluation. The results of the study were not entirely unambiguous:
they identified a positive effect from the observed characteristics on
student test scores for the eighth graders but not for the ninth graders.
A similar study was later repeated by Applebee, Langer, Nystrand,
and Gamoran (2003). It involved 974 students of different ages in
language arts lessons. The same classroom discourse variables were
observed as in the previous research. Testing took place through productive
written assignments. The analysis indicated a clear positive link
between classroom talk features and student performance.
McElhone (2012) carried out a study on 495 students in fourth and
fifth grade in reading lessons. Achievement was measured through a
reading comprehension test. The study identified high frequency conceptual
press—a pattern involving responses to student contributions
with requests for clarification and elaboration—to be a key feature of
classroom discourse. The study also investigated the opposite pattern of
reducing conceptual press, in which the teacher simplifies and narrows
the original question, prompts the student, answers the question
themselves, or asks for a response from a student other than the one
previously called on. The results demonstrated that the pattern of reduced
conceptual press was negatively related to results on reading
comprehension tests; the results for the high conceptual press pattern
were not statistically significant.
Michener (2014) examined a sample of 236 students in third to fifth
grade. Teaching was observed and students were given reading comprehension
tests. Multilevel modeling indicated a positive impact
stemming from teacher explanations (teacher talk longer than two lines
of transcript) and, in contrast, a negative impact stemming from elaborated
student utterances with explanations (student utterances of
more than two lines of transcript that expressed a coherent idea).
Muhonen, Pakarinen, Poikkeus, Lerkkanen, and Rasku-Puttonen
(2018) conducted research in language arts and physics/chemistry
lessons. The sample consisted of 608 sixth grade students. Researchers
recorded video of 158 lessons which were subsequently coded in terms
of quality of instructional dialogue (i.e., the degree of engagement in
deep and meaningful conversations with clear learning content) and
quality of feedback (i.e., the extent of the teacher's extension and expansion
of student talk). Grades were used as an indicator of student
achievement. The results indicated a positive relationship between the
quality of classroom talk and student achievement: the higher the
quality of classroom talk was, the higher the grades received by the
students in the given class and subject were.
The findings from the research cited above are varied. Some confirm
the assumptions of researchers about the positive relation between talkintensive
discourse and student achievement (Applebee et al., 2003;
Muhonen et al., 2018), while others are less conclusive (McElhone,
2012; Nystrand et al., 1997) or even contradictory (Michener, 2014).
To a certain extent, the variability of the results may be explained by
the variation in observed classroom talk characteristics (for example,
teacher's authentic questions, open discussion time, conceptual press,
student explanation, quality of feedback) as well as by the different
ways used to measure student outcomes (e.g., writing assignments,
comprehension tests, grades).
However, we understand that there may be another important
source of fluctuation in the results of such investigations. The measurement
of classroom discourse characteristics in these studies was
related to the whole class: for example, the total time of student talk,
the total amount of open discussion in a class, or the total number of
authentic questions asked by the teacher. These indicators were then
linked to the aggregate performance of the whole class on tests. This
obscures the distinctions among individual students in the class. There
1
Alexander (2018) measured the outcomes of 4th grade students in English,
mathematics, and science tests; Andreassen and Bråten (2011) evaluated the
reading comprehension performances of 5th grade students; Kiemer et al.
(2015) perceived autonomy, competence, and intrinsic learning motivation in
mathematics results of 9th grade students; Furtak et al. (2016) evaluated
achievement in biology of high school students; O'Connor et al. (2015) charted
the outcomes of 6th grade students in mathematics tests; Reznitskaya et al.
(2001) measured the argumentation of 4th and 5th grade students in persuasive
essays; van der Veen et al. (2017) evaluated communicative competence and
subject matter knowledge in early childhood education.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
2
are almost certainly differences in their participation in classroom talk
and in their test results. Different students, for example, speak for different
amounts of time, differ in the degree to which they are engaged
in open discussion, or answer different types of questions. The failure to
consider the differences among individual students can be viewed as a
limitation of the studies cited above. This study is designed to address
the existing knowledge gap, as we consider individual student characteristics
both in terms of their participation in classroom talk and in
terms of their achievement.
1.3. Differences among students
Empirical research shows that not all students engage in classroom
talk to the same extent. A classic study by Brophy and Good (1970)
demonstrated that both the quantity and quality of verbal interactions
with the teacher was considerably different for individual students in a
class. Similar results have been produced by many more recent studies,
for example Black (2004), Jurik, Gröschner, and Seidel (2013), Clarke
(2015), and Helgevold (2016). Black (2004) examined different patterns
of participation and noted that individual students tended to engage
in habitual and recurring ways. Kovalainen and Kumpulainen
(2007) reported that there are different types of students in classrooms.
Their study describes vocal participants who are very active and take
the initiative in all circumstances, responsive participants who are not
active in themselves but answer questions from the teacher, and silent
participants who rarely take part in communication. The frequency
with which students engage in classroom talk and how long they speak
are identifiable differences among students. According to Black (2004),
students also differ in the level of productivity of their responses. In
non-productive participation, student utterances are short, unsurprising,
and do not require great cognitive effort; the student's primary
goal is to meet the teacher's requirements. In contrast, productive
participation describes student interactions that contain verbal actions
that appear to create and maintain the shared understandings underpinning
the learning process. In such cases, student utterances are
elaborate and contain argumentation and reasoning. Black (2004) argues
that there are differences among students not only in the frequency
of their participation but also in the quality of their contributions.
These findings indicate that there are different participation degrees
and quality in classroom discourse. These differences presumably create
differing learning opportunities and, ultimately, could lead to differences
in student achievement. This assumption was recently verified by
Webb et al. (2014) and Ing et al. (2015). They conducted a study in
mathematics lessons on 111 students. All student talk during wholeclass
and small-group discussions was recorded on videos that were
then coded according to whether the students referred to their own
thoughts or commented on the ideas of others and according to the
degree of conceptual elaboration. Student achievement was measured
using standardized mathematics tests. The study found a positive link
between student participation and achievement.
On the other hand, there were contrasting findings from an experimental
study by O'Connor et al. (2017) of two sixth grade classes
with 44 students total. Both classes had been previously included in an
intervention focused on implementing academically productive talk. A
culture of active engagement had been established in both classrooms.
In both classes, the same subjects were discussed, but the conditions
were manipulated. Each class underwent two teaching units – one unit
taught through talk-intensive instruction and the other unit taught
through traditional direct instruction. Mathematics tests (pre-tests and
post-tests) were used to measure student learning. At the same time, the
total number of words uttered during a lesson was counted for each
student in both classes. Generally, the analysis indicated that students
in the talk-intensive conditions talked many times more and at the same
time scored better on the test. But the results did not indicate an association
between the number of words spoken by a student and that
student's test score in either condition. Hence, the authors claimed that
both silent and vocal students benefited from being in a classroom
where talk-intensive instruction had been established, and it did not
matter whether they verbally participated in the given lesson. This
perspective contradicts the findings of Webb et al. (2014) and Ing et al.
(2015), who stated that individual participation predicts individual
student achievement.
Throughout chapters 1.2. and 1.3., we have referred to studies with
differing results and varied underlying assumptions. Nystrand (1996;
2006) argued that classroom discourse shapes student skills through the
establishment of classroom epistemology. Basically, Nystrand assumed
a homogeneous impact of classroom discourse on all students in the
class. O'Connor et al. (2017) agreed with this assumption and added
that even students who do not speak in class benefit from talk-intensive
instruction. Webb et al. (2014) and Ing et al. (2015) provided a complementary
perspective based on the knowledge that different students
participate in different ways and that therefore the impact of classroom
discourse on different students in a single class may differ. Being aware
of these differences in both assumptions and results, we decided to
carry out our own study. In doing so, we bore in mind both possibilities:
that students may be affected by the discourse established in their
classroom and they may also be affected by the scope and quality of
their own participation in this discourse. This double perspective constitutes
the unique contribution of our study.
2. The present study
The aim of this study is to contribute to the understanding of the
relationship between student participation in classroom talk and their
achievement in the ninth grade. The research was conducted under
natural conditions on a relatively large sample (32 classes, 639 students).
Unlike previous studies of this type, we distinguished two levels.
First, we were interested in determining whether student achievement
might be related to the fact that a given student was in a talk-intensive
classroom characterized by a high degree and quality of student participation.
Second, we examined whether student achievement might
be related to the fact that a particular student participated frequently
and productively in classroom talk. We used student talk time and the
number of student utterances with reasoning made during a lesson as
indicators of student participation. We measured student achievement
by means of the results from standardized reading literacy tests conducted
by the Czech School Inspectorate (CSI).
In line with our research intent, we formulated two main hypotheses.
We expected to identify a statistically significant positive
relation between student participation and student achievement (H1).
At the same time, we hypothesized that a given student's test result
would be significantly accounted for by their individual involvement in
classroom talk and by the overall character of classroom talk (H2).
3. Method
The research design was based on a combination of observational
data obtained by the authors and from standardized reading literacy
tests conducted by the CSI. The CSI is a key central institution in the
evaluation of the education system in the Czech Republic that distributes
and evaluates standardized tests focused on different areas of
student learning to measure student achievement in Czech schools.
Murphy et al. (2009) reported that the use of non-standardized tests
designed by researchers is one of the weaknesses of a number of studies
investigating the impact of classroom discourse on student achievement.
The decision to use tests conducted and evaluated by the CSI in
this study addresses this shortcoming and should help to ensure the
validity and reliability of the assessment.
3.1. Measures
In this study, we measured student participation in classroom talk
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
3
and student achievement. We also included socio-economic background
and gender as contextual variables since these characteristics are believed
to affect student achievement (e.g., Bodovski, Jeon, & Byun,
2017; Pfeffer, 2008; Parker et al., 2018).
3.1.1. Student participation in classroom talk
We operationalized the quantity of student participation in classroom
talk as the amount of time a student talked during a lesson. In
each class, we observed two lessons. We believe that data from these
two lessons represent the nature of classroom talk in the given class and
also participation patterns of individual students.2
Teachers were instructed
to teach as they normally would. They did not receive any
instruction as to the content of the observed lessons or instructional
methods. We therefore assume that both teachers and students behaved
in a way similar to their usual conduct. We only counted student
utterances that were part of whole-class teaching that involved interactions
between the teacher and students and among students. We excluded
such types of talk as reading a text, individual work, and group
work. We also excluded any talk that did not relate to the subject matter
at hand (for example, organizational matters).
We operationalized the quality of student participation in classroom
talk as the amount of utterances with reasoning made by a student
during a lesson. We based our method on the classification of student
utterances proposed by Pimentel and McNeill (2013): (1) no response,
(2) word/phrase, (3) complete thought (resembles a sentence but no
explanation of thinking is included), and (4) thought and reasoning
(resembles a sentence and includes explanation). We counted those
utterances that corresponded to the fourth type of this classification.
An example of an utterance with reasoning:
Teacher: What does the sentence that the universe has no timeless
geography mean? It is quite complex. Could anyone try to explain?
Student: It is not really possible to show what the universe is like
when it is constantly moving so we cannot even know in a timeless way
what shape it will take. When we picture a map, the stars and the galaxies
cannot be placed exactly because they are constantly moving and
no one knows how.
The quantity and quality of student participation were measured
using a tablet application developed by the study authors. The application
is based on a class diagram with places marked where each
student sits. There were always two observers present in the classroom.
The first recorded talk time. The moment a student began to speak, the
observer marked his or her place in the diagram and activated the
measurement. When the student finished, the observer deactivated the
measurement. The second observer recorded the occurrence of utterances
with reasoning. As soon as a student made this type of utterance,
the observer marked his or her place in the diagram and the utterance
was included in the count. While measuring talk time (first observer) is
relatively non-problematic, identifying utterances with reasoning
(second observer) is a difficult task. The reliability of coding could have
been enhanced by the presence of a higher number of coders in the class
or by making video recordings. We did not, however, opt for either of
these options, in order to minimize disruption to the normal course of
teaching. We know from previous field studies (e.g., Sedova et al.,
2014; Sedova et al., 2016; Sedova et al., 2017) that teachers perceive
the presence of a video camera in a lesson very negatively and there is a
significant increase in the likelihood that they will refuse to participate
in the research. We anticipated a similar effect would occur if there
were to be a larger number of observers in the classroom. We did not
want our sample to be significantly affected by negative self-selection
on the part of the participants, so we decided to limit the intervention in
the normal course of action to the presence of two observers in the
classroom. Moreover, any interference with normal conditions raises
the risk that the observed actors will change their behavior. By minimizing
invasive stimuli, we also sought to reduce the possibility that
teachers and students would change their interaction behavior in response
to our presence.
In total, there were four pairs of coders performing the observation.
All coders were thoroughly trained. Four of them counted exclusively
the quantity of student participation and the other four exclusively the
quality – the occurrence of utterances with reasoning. These four coders,
who carried the greatest responsibility during the observation, are
co-authors of this study (the fourth, fifth, sixth, and seventh author).
Prior to the start of the research, each coder practiced individually and
in a group under supervision (by the first and third author of this study)
identification of utterances with reasoning using older video recordings
available to the research team. During the training, coders at first
practiced recognizing utterances with reasoning in transcripts of lessons.
They then watched the video and entered information into the
tablet application in real time regarding which student made an utterance
with reasoning. This was done individually and then checked
for agreement among individual coders. The training was finished when
agreement among all coders reached 90%; this took about 43 h per one
coder responsible for coding utterances with reasoning. The other four
observers were responsible for the talk time measurement. Their
training was much shorter. We presented the tablet application to them
and instructed them how to work with it. Then they practiced real-time
coding of one video recording until we were certain that the differences
among observers in talk time measured for individual students were
minimal. This took about 9 h per one coder responsible for measurement
of talk time.
Observations took place in November and December 2017. Only
those students who attended both lessons were included in the analysis.
This procedure did not result in a significant data loss (up to 4%). For
each student, we considered their average talk time for the two lessons.
We also worked with the average number of arguments that each student
articulated in the two lessons.3
For data at the class level, our
calculations included total student talk time, which was calculated as
the sum of all individual student talk times per one lesson. We also
calculated the total number of student utterances with reasoning per
one lesson.
3.1.2. Student achievement
We used the results of individual students on reading literacy tests
designed and evaluated by the CSI as an indicator of student achievement.
In Czech curriculum, developing reading literacy is mostly concentrated
in the language arts, which is the subject where we carried
out the observation. Testing took place in November 2017. The test
consisted of 16 tasks. An example of one of the tasks is attached in
Appendix. Students were given several longer extracts from texts
written in Czech and the level of their understanding was assessed with
several open- and closed-ended questions. The tasks covered all areas of
reading literacy that were operationalized in accordance with the curriculum
for the grades concerned. In this study, we use the student
success rate on the main section of the test (100% is a perfect score) as
the indication of student achievement.
2
We subsequently gathered additional data intended for close qualitative
analysis from four classrooms. We conducted six video recordings of lessons in
each of these four classrooms. This allowed us to compare values of talk time
and number of utterances with an argument for individual students from
measurement 1 (two lessons) and measurement 2 (six lessons). The aim was to
test whether there were significant differences between the two measurements.
We used the Wilcoxon signed ranks T-test and paired samples test. Based on
these tests, we confirm the null hypothesis (p = 0.11). There was no significant
distinction between the mean of the measure 1 and measure 2.
3
We paid careful attention to the consistency of student seating so as not to
cause confusion in the data due to different seating positions of students in the
two lessons.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
4
3.1.3. Students’ socio-economic background
Students' socio-economic background was measured through the
index of the highest occupational status of parents (HISEI).
Occupational data for the student's father and mother were obtained
through open-ended questions. The responses were coded as four-digit
International Standard Classification of Occupations codes and then
recoded to the International Socio-Economic Index of Occupational
Status (ISEI) (Ganzeboom, De Graaf, & Treiman, 1992). The HISEI for a
given student corresponded to the higher ISEI score of the two parents
or to the only available parent's ISEI score. Higher ISEI scores indicate
higher levels of occupational status.
3.2. Sample
The sample for this study consisted of ninth grade students (ISCED
2A) who were involved in the CSI's selective reading literary testing. In
this national sample survey, CSI included altogether 163 schools from
the total of 4221 Czech lower secondary schools. From this sample, the
researchers chose schools from three Czech districts.4
In total, 23
schools were approached; two of these schools refused to take part in
the research. The study included a total of 21 schools and 32 classes (all
ninth grade classes). The sample consisted of 639 students. There were
approximately 21 students on average in the class; 52% boys and 48%
girls. The age of the students at the time of the measurement ranged
between 14 and 15 years.
In all of these classes, researchers conducted structured classroom
talk observations in two different language arts lessons (2 × 45 min).5
Teachers were instructed to teach the lessons in their usual way. In
total, we recorded data from 64 lessons.
Students completed questionnaires to determine a number of contextual
variables, such as socio-economic background, school grades,
academic self-esteem, and attitudes toward the subject and the teacher.
Table 1 summarizes the descriptive results of all variables relevant to
this text.
Students in our sample had average test scores of approximately
44% (Standard Dev. = 15.36).6
The classroom discourse indicators revealed
that they spoke relatively little in the monitored lessons – on
average, each student spoke for 11 s in one lesson and made 0.5
utterances with reasoning. In our sample, 46 students spoke for 0 s.
Those students had average test scores of approximately 41% (Standard
Dev. = 15.12). These general data indicate that the classroom talk in
the sample has relatively traditional features – it is dominated by teacher
talk, with only a small number of elaborate student utterances
containing arguments. This finding is in line with our previous studies
carried out in Czech schools (Sedova et al., 2014). However, our results
show a great variability in student participation. In almost all classes,
some students spoke for significantly longer amounts of time than
others and some students remained completely silent. The number of
utterances with reasoning for individual students also varied substantially.
Of all the students observed, 91% made some utterance. At
least one utterance with reasoning was produced by only 42% of the
students.
3.3. Research ethics
We first made an arrangement to work with the CSI. They agreed
that we could use the results of their reading literacy tests in our research
if the schools involved and the students’ parents agreed. We then
sought oral consent from the school principals and all the teachers to
allow us to conduct the research in their schools and classrooms. In the
next step, we sought written consent from all of the parents of the
students participating in the observed classes. Participants were assured
of confidentiality and of the ability to withdraw at any time. No one
withdrew during the study. All participants were assigned numbers, and
any personally identifying information was removed from the data
prior to processing.
3.4. Analytical strategy
In this study, our main focus was on investigating the effect that
student participation in classroom discourse might have on student
achievement. We also considered student gender and socio-economic
background in the analysis.
We constructed multilevel models. Multilevel models, also referred
to as random coefficient models, enable testing whether the constants
and slopes of the function symbolizing influence (the regression coefficient)
differ in different groups (see Hox, 2002). The purpose of
multilevel models is to test this proposition using group characteristics,
which in our study were represented by classroom characteristics.7
The
purpose of the analyzed models was to determine, at both the class level
and the individual student level, whether classroom discourse might
affect student achievement. To do so, we worked in the first step with a
zero model. This model enables the variance of the dependent variable
(student academic achievement) to be broken down into a component
caused by differences among individual students and a component
caused by differences among classes. Intra-class correlation (ICC) was
determined as the ratio of variance caused by differences among classes
to total variability (Soukup, 2006). To test the other hypothesis,8
we
used fixed-slope models9
that involve factors at both the student and
classroom levels; in all cases, we looked for models that contained only
factors with an effect on the explained variable at a statistical significance
level of 0.05. The equation of the models is as follows:
Yij = β0 + β1 Xij + β2 Aij + β3 C i + β4 (Xi * Xj) + β5 (Xi * Ci) + β6 (Xi *
Ai) + u0j + eij
where Y represents the academic achievement level of student i in class
j, Xij is the amount of talk time by student i and the total talk time for all
students in class j,10
Aij identifies the socio-economic status of student i
and the group status of class j, and Ci represents the gender of student i.
It follows from the equation that we intend to estimate any impact for
the variables of talk time and student socio-economic status at both
possible levels, i.e. at the individual and whole-class levels. Furthermore,
the model indicates that we also intended to estimate the impact
of three interaction effects (the interaction of the talk time for a student
and for the whole class, the interaction between time and gender, and
the interaction between time and socio-economic status). Parameters u
and e are random errors at the class (u0j) and student level (eij).
4
The Czech Republic is divided into 14 districts. The data were collected in
three districts (South Moravian Region, Olomouc Region, and Vysocina
Region). We addressed all schools in these three districts that were involved in
CSI testing of reading literacy.
5
There were one or two weeks between the two observations.
6
This seems to be a high standard deviation; however, we should note that
similar results occurred with Czech students in an international assessment (see
Blažek and Příhodová, 2017).
7
Another advantage of multilevel models is that the nested structure of the
data can be taken into account.
8
We expected to identify a statistically significant positive effect from student
participation on student achievement. At the same time, we hypothesized that a
given student's test result would be significantly accounted for by both their
individual involvement in classroom talk and the overall character of classroom
talk.
9
The randomness of the first-level variables in all cases proved to be statistically
insignificant.
10
Our second variable conceptualizing classroom talk, i.e. the number of
utterances with an argument, can be substituted for X.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
5
4. Results
We used the following strategy to construct the regression models.
First, we focused on the aforementioned zero model. We then gradually
added individual independent variables. For each predictor,11
we tested
both options as random or fixed components. Due to space constraints,
here we only present the final results. The introduction of each variable
in the form of a random effect was shown to be insignificant for all of
the variables considered at the student level. This phase resulted in the
elimination of any factors not involved in explaining the variability in
students. In the last step, we estimated the effect of some interactions
between significant predictors. We present only significant interactions.
Table 2 summarizes the parameters of the main estimated models.
Because we hypothesized that differences among classes might play a
role in explaining variability in student achievement, we first estimated
the zero model. Using this model, the average success rate on the
reading literacy test when eliminating all factors was estimated at approximately
43%. Intra-class correlation (ICC) equaled 0.18. This result
can be interpreted as meaning that 18% of the differences in student
success rate on the test could be attributed to the class they come from.
In the next step, we introduced the average time a student actively
spoke during a lesson (measured in seconds). The model estimated that
this was a significant predictor (p < 0.01). The calculated model
parameters also indicated that talk time had a positive impact on student
achievement. Its strength can be interpreted as meaning that an
increase in average talk time of approximately 100 s (1.7 min) correlates
with an increase in test scores of 13.7 percentage points. This is
undoubtedly a high figure if we take into account the average score in
the test (43.7%). It seems to us that this confirms our main hypothesis.
We also assessed how to fit the estimated model to the data. We used
the Bayesian information criterion (BIC), which compares the estimated
model with a zero or saturated model. Raftery (1995) suggested the
following key for the evaluation of the credibility of compared models:
a 0–2 point decrease means the evidence was weak, 2–6 positive, 6–10
strong, and more than 10 very strong. A comparison of Model 1 with
the zero model reveals a difference (decrease) of more than 800 points.
At the student level, it seems that talk time was therefore a highly
significant predictor of student achievement.
In Model 2, we added information about the student's socio-economic
background as indicated by the occupational status of the family
(HISEI). This model better matched the obtained data (there was a
decline in the BIC). The impact of talk time at the student level remained
virtually the same. At the same time, the HISEI also seemed to
be a significant predictor. Here we can see a positive link. Since the
HISEI is centralized, the calculated estimate can be interpreted as follows:
an increase in parent status by one unit appeared to correspond to
an average increase in test scores of nearly 2 percentage points.
The next explanatory variable at the first (student) level added to
the model was student gender (Model 3). Although the decrease in BIC
was not great (13 points), this model fits the data better. The prediction
strength of talk time and family status replicated the estimates from
Table 1
Descriptive indicators for the variables tracked.
Variable Mean Stand. deviat. Median Min Max Percentiles 25 Percentiles 75 N
Reading literacy (success in %) 43.86 15.36 43.47 6.52 82.61 32.61 54.35 602
Students
Time (s) 11.45 18.02 5.35 0 180.06 1.74 13.90 534
Boy time (s) 10.29 14.66 4.89 0 98.50 1.68 12.04 274
Girl time (s) 12.44 20.46 5.76 0 180.06 1.74 15.66 257
Reasona
0.57 1.31 0.01 0 15.50 0 0.50 535
Boy reasona
0.51 1.29 0.01 0 15.50 0 0.50 278
Girl reasona
0.63 1.34 0.01 0 13.00 0 0.50 254
HISEI 47.52 18.56 45.00 12.00 88.00 34.00 54.00 582
Class
Total time (s) 189.55 32
Total reason (sum)a
9.67 32
HISEI 47.59 32
a
Amount of utterances with reasoning made by a student during a lesson.
Table 2
Results from multilevel models of the impact of talk time on academic
achievement.
Estimate Std. Error df T p
Model 0 Intercept 43.758 1.258 28.090 34.773 0.000
BIC12
4963.224
ICC13
18%
Model 1 Intercept 42.441 1.332 34.077 31.868 0.000
Time 0.137 0.038 495.135 3.576 0.000
BIC 4161.216
ICC 14%
Model 2 Intercept 42.856 1.276 33.191 33.593 0.000
Time 0.135 0.038 456.203 3.529 0.000
HISEI 1.721 0.744 467.652 2.314 0.021
BIC 3883.940
ICC 13%
Model 3 Intercept 44.334 1.452 56.646 30.523 0.000
Time 0.132 0.039 450.680 3.372 0.001
HISEI 1.822 0.751 465.377 2.427 0.016
Boy −2.722 1.330 452.624 −2.047 0.041
BIC 3870.053
ICC 12%
Model 4 Intercept 39.080 8.940 29.115 4.371 0.000
Time 0.112 0.041 453.371 2.724 0.007
HISEI 1.940 0.778 443.375 2.494 0.013
Boy −2.844 1.332 450.207 −2.135 0.033
Time_Class 0.018 0.011 31.939 1.571 0.106
HISEI_Class 0.044 0.165 29.953 0.266 0.792
BIC 3876.483
ICC 12%
Model 5 Intercept 38.277 2.793 49.667 13.704 0.000
Time 0.323 0.104 416.903 3.103 0.002
HISEI 2.036 0.752 463.968 2.707 0.007
Boy −3.584 1.576 449.838 −2.275 0.023
Time_Class 0.032 0.012 49.860 2.594 0.012
Time*Time_Class −0.002 0.000 377.551 −2.604 0.010
Time*Boy 0.061 0.077 463.434 0.789 0.430
Time*HISEI 0.018 0.034 462.966 0.539 0.590
BIC 3848.742
ICC 11%
11
With the exception of gender, where it was not appropriate given the exhaustive
categories.
12
Bayesian Information Criterion.
13
Intra-class correlation.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
6
previous models, and there were also significant differences in success
rate between boys and girls. Boys' scores on the reading literacy tests
were an average of nearly 3 percentage points lower than girls’ scores.
In the subsequent models, we included class-level variables. In
Model 4, we estimated test success relative to the talk characteristics of
the entire class. This second level is represented by the total amount of
time that students in the class actively participated in talk. We also
introduced the average occupational status of the parents of the entire
class. The estimated model indicated that aggregate talk time was also a
significant predictor of student achievement. The significance of this
variable was 0.1, which is higher than what is common in the social
sciences (0.05). In this case, however, it is necessary to take into account
that multilevel models require a relatively high number of groups
(ideally 100 and more). Our number of classes did not meet this requirement.
Hox (2002) showed that in some situations multilevel
models can be estimated with even fewer groups (even less than 10). He
noted that in such cases the fact that the result underestimates the
standard error needs to be taken into account. Therefore, he recommended
avoiding working with standard statistical significance, i.e.
an alpha level of 0.05. If the number of groups is around 20 to 30, the
critical threshold should be higher (0.09–0.1). Such is the case for
Model 4. Taking these criteria into account, the hypothesis predicting a
significant influence from student talk time at class level was confirmed.
The second aggregate variable—the average occupational status
of the class—turned out to be statistically insignificant. It therefore did
not contribute to explaining the variance in academic achievement.
This is a surprising finding to some extent. Parental occupational status
was a significant predictor at the individual level. Data from the Programme
for International Student Assessment have repeatedly demonstrated
in the Czech context that socio-economic status plays an
important role at not only the individual level but also the school level
(Straková, 2016). Our data at the class level did not confirm this
finding. The relatively small number of classes for the second-level
models could have been an influencing factor. Looking at the descriptive
data, however, it seems to us that the relative similarity of the
classes included in our sample in terms of average occupational status is
a more prominent reason. In our model, this was not a characteristic
that explained student achievement on the literacy test. Overall, it
appears that Model 4 did not provide a more accurate estimate than the
previous model (the BIC slightly increased). We therefore eliminated
the aggregate occupational status in the steps that followed.
The last model presented, Model 5, estimated student achievement
with the contribution of selected interactions between several variables
that had proved in previous models to be significant at the first or
second level. The information criteria indicated that this estimate best
suited the data obtained. New slopes (regression coefficients) were estimated
for the inclusion of all variables. The strength of the influence
of individual talk time increased substantially. The new estimate indicated
that the average test success rate was about 30 percentage
points higher for the individual students who spoke for about 100 s in a
lesson. The significance of the impact of total talk time in a class was
therefore confirmed; the significance of this predictor was now at the
usual level (p < 0.05). The predictors of student gender and occupational
status remained at approximately the level of Model 3.
Interesting conclusions also arose from the tested interactions. The first
tested interaction, focused on the relationship between the talk time of
a particular student and the total time of all students in the class, was
significant. Although the influence is negligible, it indicates a negative
effect of the interaction in question. This means that the positive link
between a student's result in the test and their overall speaking time is
Table 3
Results from multilevel models of the impact of the number of arguments on academic achievement.
Estimate Std. Error df T p
Model 0 Intercept 43.758 1.258 28.090 34.773 0.000
BIC 4963.224
ICC 16%
Model 1 Intercept 43.197 1.300 29.898 33.238 0.000
Reason 1.598 0.574 497.457 2.785 0.006
BIC 4198.613
ICC 15%
Model 2 Intercept 43.527 1.256 29.430 34.665 0.000
Reason 1.645 0.577 462.477 2.853 0.005
HISEI 1.732 0.741 473.500 2.337 0.020
BIC 3927.571
ICC 13%
Model 3 Intercept 45.396 1.415 51.255 32.089 0.000
Reason 1.579 0.574 457.337 2.751 0.006
HISEI 1.820 0.747 470.682 2.436 0.015
Boy −3.475 1.335 458.749 −2.603 0.010
BIC 3911.072
ICC 12%
Model 4 Intercept 50.582 7.759 30.624 6.519 0.000
Reason 1.474 0.607 449.272 2.429 0.016
HISEI 1.989 0.775 445.111 2.567 0.011
Boy −3.443 1.337 456.508 −2.575 0.010
Reason_Class 0.051 0.136 35.897 0.374 0.710
HISEI_Class −0.120 0.157 30.443 −0.762 0.452
BIC 3914.326
ICC 12%
Model 5 Intercept 49.857 8.050 31.455 6.193 0.000
Reason 3.245 1.221 466.511 2.658 0.008
HISEI 1.564 0.854 447.160 1.830 0.068
Boy −3.803 1.520 454.183 −2.502 0.013
Reason_Class 0.128 0.145 38.898 0.881 0.384
Reason*Reason_Class −0.080 0.040 463.808 −2.031 0.043
Reason*Boy 0.986 1.430 449.184 0.689 0.491
Reason*HISEI 0.748 0.712 460.769 1.050 0.294
BIC 3911.327
ICC 12%
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
7
weaker in more talkative classes (i.e., in classes that are above the
sample mean). In other words, a student from a very talkative class
must also be very talkative for the individual-time effect described
above to apply. It is nonetheless important to repeat that despite being
statistically significant, this effect is negligible. A segment of individual
student talking time that amounts to 10 s may be connected to a 3.2%
increase in the student's academic achievement. The same amount of
individual student talking time may be connected to a 3% increase in
more talkative classes. The remaining interactions for the influence of
talk time were insignificant. This indicates that the talk time was neither
strengthened nor weakened in any group of student characteristics.
This is again a remarkable finding, and it appears to confirm our assumptions.
The impact of individual talk time on student achievement
was the same for girls and boys and especially for groups of students
with different socio-economic backgrounds.
We used the same analytical procedure in the next step, in which we
changed the key explanatory variable. Talk time was replaced by the
number of utterances with reasoning in the lesson. We included in the
estimates the number of utterances on both the individual and wholeclass
levels (total utterances with reasoning made by all students in a
lesson). The other predictors were identical. Table 3 presents the results
of these estimates.
A comparison of the results (Tables 2 and 3) indicates that the two
factors of classroom discourse were very similar in terms of their impact
on student academic achievement. In other words, the number of
utterances with reasoning made by a student during a lesson appears to
be a significant explanatory factor for their reading literacy test results.
When a student articulated on average an extra 1 argument during a
lesson, the average test success rate was about 1.5 percentage points
higher (Model 1). When all predictors were included, there was an increase
in the impact by as much as 3 percentage points (Model 5). The
control variables (gender and occupational status) were used in essentially
the same way as in the models based on talk time (see models 1 to
5).
There was a substantial change in the effect from the number of
utterances with reasoning at the whole-class level. It turned out that
this was not a significant influence (Model 4). The interaction between
utterances at the first and second levels was statistically significant
(Model 5) but materially negligible.14
Our data in this respect supported
the conclusion that the quality of talk was a significant predictor
of student achievement in reading literacy; its impact seemed to be
connected to the individual performance of each student. At the wholeclass
level, the students’ talk did not contribute to explaining the variance
in student performance in the dependent variable of reading lit-
eracy.
5. Discussion, limitations, and implications
The aim of this study was to test, under natural conditions in ordinary
classes, the existence of a connection between the frequency and
quality of student talk during a lesson and student achievement on a
reading literacy test. Our results indicated that this relationship existed
and was strong. Students who talked and argued more in language arts
lessons had better performance results on reading literacy tests.
At present, the central role of talk in relation to learning and
knowledge construction in the classroom has come to the forefront of
scholarly attention. It is assumed that students learn through talk and
their learning outcomes can be largely attributed to the quality of
classroom discourse (for example Mercer & Littleton, 2007; Resnick
et al., 2017). These hypotheses are widely shared, but robust empirical
evidence obtained in whole-class interactions is still scarce. This study
can be therefore viewed as a contribution toward supporting the stance
that talk-intensive pedagogies are based on correct premises.
This study contributes to the theory development in the field with
the notion that it is useful to pay attention to general characteristics of
classroom talk and also to individual participation patterns of particular
students. Unlike previous studies of a similar type, we examined the
relationship between participation and achievement in parallel at the
individual and class levels. In other words, we asked whether it is
sufficient to place a student in a talkative classroom or if they have to
actively speak, express their ideas, and reason. The assumption of the
influence of the nature of student talk at the class level is based on the
notion of a learning community in which students work together as a
group to acquire new knowledge (Lyle, 2008), while the capacities of
more advanced students are available to other members of the group
who may observe, try out, and gradually internalize new ways of
communication and thinking (Reznitskaya & Gregory, 2013). It has
been suggested that students benefit from listening to their actively
participating classmates (O'Connor et al., 2017). Our findings support
this idea only to a limited extent. We found that students had better test
results when they were in a class with an overall high amount of student
talk. At the same time, there appeared to be a stronger link between
individual student participation and their individual achievement. As
for utterances with reasoning, it is not possible to say that to be located
in a classroom with a high frequency of such utterances is itself linked
to better student results. It does seem true that there is a link at the
individual level between the number of utterances with reasoning
spoken by a particular student and that student's achievements on a
reading literacy test.
Our findings are thus more in line with the perspective, previously
outlined by the studies of Webb et al. (2014) and Ing et al. (2015), that
classroom discourse affects different students in a class in different
ways, depending on how actively they participate. Our analysis indicated
that those individuals who spoke and argued in class had better
results on a reading test. In theory, this finding can be explained
through Vygotsky's concept of internalization (Vygotsky, 1978). Talk is
essential in this process; the more opportunities for talk there are, the
faster and better the internalization of the newly learned knowledge is.
If we imagine the classroom as a community of practice (Lave &
Wenger, 1991), it is possible to distinguish between students who
participate centrally (talk and argue) and those who participate peripherally
(listen to the talk and argumentation of their peers). Our
findings may indicate that peripheral participation, witnessing the activities
of more skillful and experienced members of the community, is
not connected with learning outcomes comparable to the outcomes for
those who participate centrally.
It is possible that the differences between our finding that vocal
students have better learning outcomes than silent students and the
findings of O'Connor et al. (2017) that silent students have the same
outcomes as vocal students may be, to some degree, explained with
reference to different study designs. O'Connor et al. (2017) collected
data in classes in which students underwent an intervention focused on
the implementation of talk-intensive instruction and had been socialized
for a long time in a culture of active participation, including active
listening. The authors concluded that even the students who did not
speak in the observed lessons were engaged and therefore were able to
benefit from the whole-class discussion. This was probably not the case
for the classrooms in our sample. Our data indicate that teaching in
these classrooms was rather traditional, with a small proportion of
student talk. Consequently, our study does not contradict the study by
O'Connor et al. (2017). It was implemented in different conditions and
shows the relationship between participation in classroom talk and
student achievement in an environment that does not appear to be
dialogue-rich and does not seem to have an established culture of active
participation.
If there is a connection between how students talk in classes and
their learning outcomes, it seems to make sense to give students room
14
This is an inverse relationship; the weakening of the significance of individual
arguments along with an increase in the sum of these arguments in a
class cannot be considered materially relevant with regard to the scale.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
8
for elaborate talk and also to strive to ensure that their talk is of good
quality and contains arguments. This is an important message for
schools and teachers. Teaching in Czech schools is still very traditional
and not much student-oriented (Sedova et al., 2014). The results from
this study may help persuade Czech teachers and their educators that a
change toward talk-intensive pedagogies could be desirable as it might
help improve student learning. Training in the use of dialogic procedures
is not yet a common part of Czech teacher education. This study
provides some evidence for their inclusion in teacher education or
professional development. A number of studies have reported that
students participate unevenly in class (for example Jurik et al., 2013;
Clarke, 2015; Helgevold, 2016). In light of our findings, we can anticipate
that such a situation creates unequal learning opportunities in
terms of the potential link between participation in classroom talk and
student learning gains. We believe that there will always be students
who will talk more, because they know more about the topic (Clarke,
2015) or are more extroverted (Young, 2014) or more motivated (Jurik
et al., 2013). Despite this, we think that our analysis implies a clear
recommendation for teachers – to be aware of the possible link between
classroom talk and achievement and to strive to invite all students in
the classroom to participate in classroom conversations. However,
teachers often perceive this task as very difficult (Snell & Lefstein,
2018). In the future, it may therefore be necessary to devote energy to a
focused examination of the circumstances in which teachers are able to
get students to participate more evenly. In addition to the need to devote
research attention to these issues, it is also vital to shift the focus of
teacher education or professional development in this direction. This
means that teachers need to learn about methods of talk-intensive
teaching and they also need to be guided in creating conditions for a
more proportionate participation of all students, regardless of their
achievement, skills, or other characteristics.
Another important implication emerges from our analyses. The link
between participation and achievement appeared to apply equally for all
students, regardless of socio-economic background and gender. The existence
of educational inequalities according to socio-economic background
has been confirmed repeatedly (for example Bodovski et al., 2017;
Pfeffer, 2008). The effort to remedy unequal conditions and minimize
educational inequalities is one of the most important missions of contemporary
educational sciences. One can find in the literature a suspicion
that instructional strategies that emphasize student responsibility and activity,
including talk-intensive pedagogies, create more of a disadvantage
for already disadvantaged students, especially students from families with
low socio-economic backgrounds (Andersen & Andersen, 2017). However,
our findings indicated that when students, including disadvantaged ones,
are successfully involved and prompted to participate and reason, the
positive link was universal. Given how few widely applicable solutions
capable of optimizing learning opportunities for all students are available,
we consider our conclusions to be extremely important.
5.1. Limitations of the study
In this study, we strove to explore the link between student participation
in classroom talk and student achievement on a reading literacy
test. We conducted the study under natural conditions in order to
capture the processes that are typical of the environment and represent
the common educational experience of Czech students. We also strove
to have a sufficiently large sample of classes that would not be burdened
by self-selection. This intention led to some methodological
choices that make up the limitations of this study.
The first limitation is presented by the fact that we conducted a
correlational study. Hence, we do not have control over the direction of
causality. Our argument that participation in classroom talk affects
achievement is derived from our theoretical premises. The identified
connection might also indicate that student achievement predicts involvement
in classroom talk. In other words, high achievers, as measured
by reading literacy tests, speak more and argue more. The causal
direction could be controlled if we conducted an intervention study
including a pre-test and a post-test, as well as intervention and control
groups of students. In conducting such a study, however, we would not
capture common and typical processes and phenomena.
Another limitation of our study is the fact that the quality of student
contributions was coded by a single coder, directly during the lessons. This
kind of real-time coding can be prone to error. This decision was driven by
our attempt to get a sample representing common teachers and classes. We
were concerned that the use of recording technology or a larger number of
observers in the class would lead to the reluctance of schools and teachers
to participate in the research and to the tendency to change the interaction
behavior. We tried to minimize the risks of this choice by thoroughly
training the coders, who co-authored the study.
Further limitations—having to do with the possibility of interpreting
our findings—are related to the data we gathered in this data
collection setting. Above all, we observed a small amount of talk and
reasoning on the part of students in Czech classes. Therefore, our
findings cannot illustrate how the relationship between participation
and achievement works in a dialogue-rich environment. We can only
assert that in a situation involving the limited participation of students
in classroom talk, the degree of their individual involvement appears to
play a highly important role.
Another fact that undermines our interpretation is that the observed
classes were relatively homogeneous in terms of the students’ socio-economic
backgrounds. In other words, there were not considerable differences
among the students in the sample in this respect. Our finding that a
positive relationship between participation and achievement appears to
apply regardless of socio-economic background should therefore be accepted
with caution. It is possible that in the case of a more heterogeneous
sample this conclusion would not be confirmed.
5.2. Next steps
This study has brought important, but in many ways limited,
knowledge. One advantage is that it is part of a wider project in which
we have additional sources of data to analyze.15
Not all of the data that
we collected from this sample of students have been utilized in this
study. In addition, the present quantitative study was followed by a
qualitative study (February to May 2018) in which we included four
classes from the original sample. In these classes, we made a series of
video recordings of the lessons and conducted a number of interviews
with both teachers and students.
This material should allow us to shed more light on the predictors of
student participation in the near future. In other words, it may enable
us to discover what leads some students to be vocal and others to remain
silent. In addition, thanks to qualitative data, we should be able to
identify with greater specificity the reasons for non-participation: are
silent students simply disaffected, or rather do they not feel the need to
speak even though they are fully engaged? Are there any barriers that
discourage them from participation? In addition to the focus on the
characteristics of individual students, the wider context of classroom
culture needs to be explored in greater detail because the conditions of
participation are necessarily socially construed. The question remains:
In what ways do the teachers’ behavior and expectations influence
whether the student engages? And further, what is the role of peer
relationships and interactions in the classroom in this respect?
Funded
This article is an output of the project On the Relationship between
15
Such as questionnaires on student engagement and other additional data about students (Sedova et al., 2016).
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
9
Characteristics of Classroom Discourse and Student Achievement (GA17-
03643S) funded by the Czech Science Foundation.
Appendix. An example of one of the tasks from reading literacy test
Read the text and choose the right answer.
Towards the end of the 19th century, the disease known as beriberi (in Singhalese “I cannot”) with characteristic symptoms ranging from
weakness to paralysis spread dramatically in Dutch East Indies (present-day Indonesia). The Dutch government, concerned that there is hardly
anybody left to slave on the plantations, set up a special committee. Christian Eijkman (1858–1930), a physician of the state prison in Batavia
(present-day Jakarta) was among its members. On one beautiful day, Doctor Eijkman was enjoying a view out of the window of his official apartment
overlooking the prison courtyard. His head full of the cursed beriberi, he was watching the hens pecking around the yard. He was intrigued by their
strange movements and postures; somehow they reminded him of the sick inmates… It turned out that the hens were, through kitchen garbage,
eating basically the same diet as the inmates—mostly rice. Specifically, it was husked rice. The husks were removed because a product that had been
processed in this way looks better and as a result can cost more.
The already suspecting Eijkman only needed to ask the “competitors” a few questions and everything was clear. Jails where the inmates consumed
only husked rice had high numbers of beriberi patients, while in the jails where the managers economized and fed their inmates with cheap
rice that had not been husked, the disease almost did not occur.
In 1897, Eijkman published his discovery. Despite its significance, there was almost no response. At that time, nutrition science was dominated by
caloric assessment of nutrients, while the causes of diseases were decided by young and ambitious bacteriology. The idea that a mere deficiency in
some trace element in diet could cause a serious disease or even death simply seemed ridiculous.
However, in 1911 Eijkman's work was discovered by a young biochemist of Polish origin Kazimierz Funk (1884–1967) who was living in London.
He first tested Eijkman's conclusions on pigeons and then got a kilo of rice husks and used them to laboriously prepare six grams of white powder.
This powder, even in milligram amount, reliably cured beriberi. Funk called it vitamin B; vita means “life” in Latin and amin was for the amin group
that the powder contained. He used letter “B” to avoid confusion with a substance of a similar category that had been shortly before discovered in
milk by Funk's colleague Frederick Hopkins (1861–1947) who called it growth factor A (today known as vitamin A). It is evident that Funk's name
was widely accepted, even though the amin group after which it was named is out of all the thirteen vitamins that are known contained only in “his”
vitamin (it is at present known as vitamin B1). Eijkman and Hopkins were awarded the Nobel Prize in 1929.
(Houdek F., Tůma J.: Objevy a vynálezy tisíciletí [Discoveries and Inventions of the Millennium], Nakladatelství Lidové noviny 2002, p. 233)
Which of the following statements is directly contradicted in the text
above?
[ID477754]
Christian Eijkman died in prison in Batavia.
The beriberi disease can be cured with six grams of a special white powder.
Kazimierz Funk was later awarded the Nobel prize.
The beriberi disease was widespread in all prisons of the Dutch East Indies.
Vitamin B1 can be found in particular in husked rice; rice that has not been husked does not contain it at
all.
References
Alexander, R. (2006). Towards dialogic teaching: Rethinking classroom talk. Cambridge:
Dialogos.
Alexander, R. (2018). Developing dialogue: Genesis, process, trial. Research Papers in
Education, 33, 561–598.
Andersen, I. G., & Andersen, S.,C. (2017). Student-centered instruction and academic
achievement: Linking mechanisms of educational inequality to schools' instructional
strategy. British Journal of Sociology of Education, 38, 533–550.
Andreassen, R., & Bråten, I. (2011). Implementation and effects of explicit reading
comprehension instruction in fifth-grade classrooms. Learning and Instruction, 21,
520–537.
Applebee, A. N., Langer, J., Nystrand, M., & Gamoran, A. (2003). Discussion-based approaches
to developing understanding: Classroom instruction and student performance
in middle and high school English. American Educational Research Journal, 40,
685–730.
Sedova, K., Salamounova, Z., & Svaricek, R. (2014). Troubles with dialogic teaching.
Learning, Culture and Social Interaction, 3(4), 274–285.
Sedova, K., Salamounova, Z., Svaricek, R., & Sedlacek, M. (2017). Teachers’ emotions in
teacher development: Do they matter? Studia Paedagogica, 22(4), 77–110.
Billings, L., & Fitzgerald, J. (2002). Dialogic discussion and the Paideia seminar. American
Educational Research Journal, 4, 907–941.
Black, L. (2004). Differential participation in whole-class discussions and the construction
of marginalised identities. Journal of Educational Enquiry, 5, 34–54.
Blažek, R., & Příhodová, S. (2017). Mezinárodní šetření pisa (PISA)2015. Praha: ČŠI.
Bodovski, K., Jeon, H., & Byun, S.-Y. (2017). Cultural capital and academic achievement
in post-socialist Eastern Europe. British Journal of Sociology of Education, 38, 887–907.
Brophy, J. E., & Good, T. L. (1970). Teachers' communication of differential expectations
for children's classroom performance: Some behavioral data. Journal of Educational
Psychology, 61, 365–374.
Clarke, S. N. (2015). The right to speak. In L. B. Resnick, C. S. C. Asterhan, & S. N. Clarke
(Eds.). Socializing intelligence through academic talk and dialogue (pp. 167–180).
Washington, DC: American Educational Research Association.
Forman, E. A., Ramirez-DelToro, V., Brown, L., & Passmore, C. (2017). Discursive strategies
that foster an epistemic community for argument in a biology classroom.
Learning and Instruction, 55, 32–39.
Furtak, E. M., Kiemer, K., Circi, R. K., Swanson, R., de León, V., Morrison, D., et al.
(2016). Teachers' formative assessment abilities and their relationship to student
learning: Findings from a four-year intervention study. Instructional Science, 44,
267–291.
Ganzeboom, H. B. G., De Graaf, P. M., & Treiman, D. J. (1992). A standard international
socio-economic index of occupational status'. Social Science Research, 21, 1–56.
Gurevitch, Z. (2001). Dialectical dialogue: The struggle for speech, repressive silence, and
the shift to multiplicity. British Journal of Sociology, 52, 87–104.
Helgevold, N. (2016). Teaching as creating space for participation – establishing a
learning community in diverse classrooms. Teachers and teaching, 22, 315–328.
Howe, C., & Abedin, M. (2013). Classroom dialogue: A systematic review across four
decades of research. Cambridge Journal of Education, 43, 325–356.
Hox, J. (2002). Multilevel analysis. Techniques and applications. London: Lawrence Erlbaum
Associates.
Ing, M., Webb, N. M., Franke, M. L., Turrou, A. C., Wong, J., Shin, N., et al. (2015).
Student participation in elementary mathematics classrooms: The missing link between
teacher practices and student achievement? Educational Studies in Mathematics,
90, 341–356.
Jurik, V., Gröschner, A., & Seidel, T. (2013). How student characteristics affect girls' and
boys' verbal engagement in physics instruction. Learning and Instruction, 23, 33–42.
Kennedy, D. (2014). Neoteny, dialogic education and an emergent psychoculture: Notes
on theory and practice. Journal of Philosophy of Education, 48, 100–117.
Kiemer, K., Gröschner, A., Pehmer, A.-K., & Seidel, T. (2015). Effects of a classroom
discourse intervention on teachers' practice and students' motivation to learn
mathematics and science. Learning and Instruction, 35, 94–103.
Kovalainen, M., & Kumpulainen, K. (2007). The social construction of participation in an
elementary classroom community. International Journal of Educational Research, 46,
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
10
141–158.
Kutnick, P., & Colwell, J. (2010). Dialogue enhancement in classrooms. Towards a relations
approach for group working. In K. Littleton, & C. Howe (Eds.). Educational
dialogues: Understanding and promoting productive interaction (pp. 192–215). London,
New York: Routledge.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation.
Cambridge: Cambridge University: Press.
Lefstein, A., & Snell, J. (2014). Better than best practice: Developing teaching and learning
through dialogue. London: Routledge.
Lyle, S. (2008). Dialogic teaching: Discussing theoretical context and reviewing evidence
from classroom practice. Language and Education, 22, 222–240.
Marková, I. (2003). Constitution of the self: Intersubjectivity and dialogicality. Culture &
Psychology, 9, 249–260.
McElhone, D. (2012). Tell us more: Reading comprehension, engagement, and conceptual
press discourse. Reading Psychology, 33, 525–561.
Mercer, N., & Littleton, K. (2007). Dialogue and the development of children's thinking: A
sociocultural approach. London & New York: Routledge.
Michener, C.. Features of dialogic instruction in upper elementary classrooms and their relationships
to student reading comprehension. (2014). Retrieved from http://hdl.handle.
net/2345/bc-ir 103554.
Muhonen, H., Pakarinen, E., Poikkeus, A.-M., Lerkkanen, M.-K., & Rasku-Puttonen, H.
(2018). Quality of educational dialogue and association with students' academic
performance. Learning and Instruction, 55, 67–79.
Murphy, K., Wilkinson, I. A. G., Soter, A. O., & Hennessey, M. N. (2009). Examining the
effects of classroom discussion on students' comprehension of text: A meta-analysis.
Journal of Educational Psychology, 101, 740–764.
Nystrand, M. (2006). Research on the role of classroom discourse as it affects reading
comprehension. Research in the Teaching of English, 40, 392–412.
Nystrand, M., Gamoran, A., Kachur, R., & Prendergast, C. (1997). Opening dialogue.
Understanding the dynamics of language and learning in the English classroom. New York,
London: Teachers College Press.
O'Connor, C., Michaels, S., & Chapin, S. (2015). “Scaling down” to explore the role of talk
in learning: From district intervention to controlled classroom study. In L. B. Resnick,
C. S. C. Asterhan, & S. N. Clarke (Eds.). Socializing intelligence through academic talk
and dialogue (pp. 111–126). Washington: AERA.
O'Connor, C., Michaels, S., Chapin, S., & Harbaugh, A. G. (2017). The silent and the vocal:
Participation and learning in whole-class discussion. Learning and Instruction, 48,
5–13.
Parker, P. D., Van Zanden, B., & Parker, R. B. (2018). Girls get smart, boys get smug:
Historical changes in gender differences in math, literacy, and academic social
comparison and achievement. Learning and Instruction, 54, 125–137.
Pfeffer, F. T. (2008). Persistent inequality in educational attainment and its institutional
context. European Sociological Review, 24, 543–565.
Pimentel, S. D., & McNeill, K. L. (2013). Conducting talk in secondary science classrooms:
Investigating instructional moves and teachers' beliefs. Science Education, 97,
367–394.
Raftery, A. (1995). Bayesian model selection in social research. In P. Marsden (Ed.).
Sociological methodology (pp. 111–163). Oxford: Blackwell Publishers.
Resnick, L. B., Asterhan, C. S. C., & Clarke, S. N. (2017). Student discourse for learning. In
G. E. Hall, D. M. Gollnick, & L. F. Qzinn (Eds.). Handbook of teaching and learning.
Wiley: Wiley-Blackwell.
Reznitskaya, A., Anderson, R. C., McNurlen, B., Nguyen-Jahiel, K., Archodidou, A., & Kim,
S. (2001). Influence of oral discussion on written argument. Discourse Processes, 32,
155–175.
Reznitskaya, A., & Gregory, M. (2013). Student thought and classroom language:
Examining the mechanisms of change in dialogic teaching. Educational Psychologist,
48, 114–133.
Scott, P., Ametller, J., Mortimer, E., & Emberton, J. (2010). Teaching and learning disciplinary
knowledge. In K. Littleton, & C. Howe (Eds.). Educational dialogues.
Understanding and promoting productive interaction (pp. 289–303). London, New York:
Routledge.
Sedova, K., Sedlacek, M., & Svaricek, R. (2016). Teacher professional development as a
means of transforming student classroom talk. Teaching and Teacher Education, 57,
14–25.
Sfard, A. (2008). Thinking as communicating. New York: Cambridge University Press.
Skidmore, D., & Murakami, K. (2012). Claiming our own space: Polyphony in teacher–student
dialogue. Linguistics and Education, 23, 200–210.
Snell, J., & Lefstein, A. (2018). ‘‘Low ability,’’ participation, and identity in dialogic
pedagogy. American Educational Research Journal, 55, 40–78.
Soukup, P. (2006). Proč užívat hierarchické lineární modely? Czech Sociological Review,
42, 987–1012.
Straková, J. (2016). Mezinárodní výzkumy výsledků vzdělávání. Metodologie, přínosy, rizika a
příležitosti. Praha: Univerzita Karlova.
van der Veen, C., de Mey, L., van Kruistum, C., & van Oers, B. (2017). The effect of
productive classroom talk and metacommunication on young children's oral communicative
competence and subject matter knowledge: An intervention study in early
childhood education. Learning and Instruction, 48, 14–22.
van der Veen, C., & van Oers, B. (2017). Advances in research on classroom dialogue:
Learning outcomes and assessments. Learning and Instruction, 48, 1–4.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes.
Cambridge: Harvard University Press.
Webb, N. M., Franke, M. L., Ing, M., Wong, J., Fernandez, C. H., Shin, N., et al. (2014).
Engaging with others' mathematical ideas: Interrelationships among student participation,
teachers instructional practices, and learning. International Journal of
Educational Research, 63, 79–93.
Wilkinson, I. A. G., Murphy, P. K., & Binici, S. (2015). Dialogue-intensive pedagogies for
promoting reading comprehension: What we know, what we need to know. In L. B.
Resnick, C. S. C. Asterhan, & S. N. Clarke (Eds.). Socializing intelligence through academic
talk and dialogue (pp. 37–50). Washington, DC: American Educational Research
Association.
K. Sedova, et al. Learning and Instruction 63 (2019) 101217
11