Introduction
As the Internet is increasingly growing,
online education continues to grow too
(Johnson & Aragon, 2003), a phenomenon
expected to continue at a significant rate
(Allen & Seaman, 2004). Online
discussion forums, or Computer Mediated
Discussions, are popular with educators
who aim at using IT (Information
Technology) to enhance the quality of
learning. The use of computer-mediated-communication tools can present new
ways to promote knowledge construction
(Schellens & Valcke, 2006). Computer-mediated-communication tools can help
make the construction of knowledge easier
by working as a social medium to support
students’ learning by representing
students’ ideas and understandings in
concrete forms (e.g., notes) so that ideas
can be more developed via social
interactions (e.g., questioning, clarifying)
(Van Drie, Van Boxtel, Jaspers &
Kanselaar, 2005). One example of such
tools is the asynchronous discussion
forum. The technology which is available
in asynchronous online discussions
provides a number of ways to foster the
construction of collaborative knowledge,
while asynchronicity offers learners the
opportunity to interact at any time from
any place (Scardamalia & Bereiter, 1994).
The debate could be described as a
constructive learning environment which
offers multiple approaches and actual
world examples of the topic of discussion,
that encourages reflection, and that
supports collaborative construction of
knowledge via social negotiation
(Jonassen, 1994).
Early analyses of computer-mediated
communication using asynchronous tools
tended to concentrate more on quantitative
analysis of the data, especially on word
counts and number of postings. Yet,
although this method of analysis provides
a survey of the interactions which occur
online, it does not take into consideration
the content of what is posted on the
discussion boards. The analysis of the
content of the discussion boards, thus,
moves towards a more semantic labeling
of propositions (Donnelly & Fitzpatrick,
2010). The assessment of co-construction
of knowledge based on quantitative
analysis of postings underestimates the
complexity of the available issue.
Although a quantitative analysis allows the
researchers to understand some linguistic
online behaviors, it does not allow deep
investigation of the language complexity
in order to pinpoint the collaborative
learning among learners. Thus, linguistic
models for a qualitative analysis of online
discourses have been elaborated by several
researchers; for example, Interaction
Analysis Model by Gunawardena, Lowe
and Anderson (1997).
More recently, some researchers have
examined if group size might influence the
levels of knowledge construction in online
discussion forums. Schellens and Valcke
(2006), for example, found that discussion
in groups of about 10 participants resulted
in larger proportions of advanced levels of
knowledge construction. Hew and Cheung
(2010) examined if there was any
relationship between the frequency of
advanced level knowledge construction
occurrences and group size. The
researchers found a significant positive
correlation between the discussion group
size and the frequency of advanced level
knowledge construction occurrences.
However, no indication was provided by
Hew and Cheung (2010) about the optimal
group size.
In fact, no research study investigating the
impact of language complexity on
knowledge construction in online
conversations has been reported. Language
complexity refers to the lexical variation of
a given text. Consequently, this study
makes an endeavour to provide some
evidence that seems to be urgently needed.
This paper addresses the effect of group
size and language complexity on
knowledge construction in online debates
and tries to ask these two research
questions: Is there a significant
relationship between knowledge
construction and group size in online
debates? And, is there a significant
relationship between language complexity
and knowledge construction in online
debates?
The study
The goal of this study was to build on the
current literature through exploration of
how group size impacts participants’
construction of knowledge within a
primary asynchronous environment. It also
tries to investigate the impact of language
complexity on knowledge construction.
This study is a longitudinal case study
because the data source is bounded by time
and environment (Creswell, 1998).
Variables of the study
Knowledge construction
Knowledge construction refers to phases
of interaction in the online debates. Phases
of interaction were identified using
Gunawardena et al.’s (1997) Interaction
Analysis Model.
Group size
Group size of an online debate refers to the
number of participants who were involved
in the conversations. Two main forms of
participation are identified in an online
discussion environment: writing and
reading (Hewitt & Brent, 2007). In this
research paper, the focus is on the writing
form of participation because writing is
closely linked to discussion, and it is of
greater importance than reading (e.g.,
when the student is answering postings
from an existing discussion thread)
(Guzdial & Turns, 2000). Moreover,
writing is a more observable phenomenon
than reading. In Debate A, group size is
equal to 326 whereas in Debate B, group
size is equal to 118.
Language complexity
Language complexity (LC) variable is
determined by type token ratio (TTR),
which is a measure of vocabulary variation
within a written text or a person’s speech.
The type-token ratio has been shown to be
a helpful measure of lexical variety within
a text. The number of words in a text is
often referred to as the number of tokens.
However, several of these tokens are
usually repeated. As long as there is only
one type of word, the relationship between
the number of types and the number of
tokens is known as the type token ratio
(TTR) (Williamson, 2009).
A high TTR indicates a large amount of
lexical variation and a low TTR indicates
relatively little lexical variation
(Williamson, 2009). The following table
features the different TTR levels:
Informants
Informants of the study are 444 online
debaters participating in two online
debates:
A: 326 debaters participating in the online
debate: “Technology in Education”
retrieved from:
http://www.economist.com/debate/days/vie
w/244,
B: 118 debaters participating in the online
debate “Internet Democracy” retrieved
from:
http://www.economist.com/debate/days/vie
w/662.
Online debates sampling
The first online debate is entitled
“Technology in education” and was
retrieved from the website “The
economist.com” on March 18th, 2011. It
was carried over 11 days from the 15th till
the 26th of October 2010 and comprised
371 comments. It was coded Debate A.
The second online debate is entitled
“Internet Democracy” and was also
retrieved from the website “The
economist.com” on April 13
th
, 2011. It was
carried over 10 days from the 23rd
February 2010 till the 4th February 2010
and comprised 128 comments. It was
coded Debate B.
Interaction Analysis Model
The informants’ online transcripts were
analyzed qualitatively using Gunawardena
et al. (1997) Interaction Analysis Model
(IAM). The analysis is based on the five
phases of knowledge co-construction that
usually occur during online debates.
Gunawardena et al. (1997) stated
that postings coded Level I and II
“represent the lower mental functions”,
while postings coded level III, IV, and V
represent the higher mental functions:
a) Level I – making statement of
observation or opinion, statement
of agreement among participants;
b) Level II - identifying areas of
disagreement, asking, or answering
questions to clarify disagreement;
c) Level III - negotiating the meaning
of terms, ideas/co-construction of
knowledge;
d) Level IV - testing of proposed
synthesis or construction against
existing literature or personal
understandings, experiences; and
e) Level V - summarizing
agreement/statements that show
new knowledge construction,
application of newly constructed
ideas. In this study, we defined
advanced levels of knowledge
construction as levels II, III, IV, or
V of the model.
Procedure
To apply the Interaction Analysis Model, I
read the postings in the original sequence
and assigned them one or more phases
from the IAM. It is possible to code
multiple sentences or a paragraph or two
with a single phase; this is consistent with
the original application of the IAM
(Gunawardena et al., 1997). I calculated
the frequencies of the coded phases for
each posting and for each informant. Two
raters, myself and an English assistant
colleague, coded the online transcripts. In
order to conduct inter-reliability checks, I
used the most advanced phase from each
posting as the basis for inter-rater checks
(Beaudrie, 2000). Inter-rater differences
were addressed following Chi (1997).
Postings were coded using the five phases
of Gunawardena et al. (1997). For
statistical correlation, Phase I was coded 1,
phase II was coded 2, phase III was coded
3, phase IV was coded 4 and phase V was
coded 5. The ‘absence of phase’ was coded
0. A second researcher reviewed the
coding of the total postings in debate A
and B. The inter-rater was selected based
on her field of specialization, applied
linguistics, and her familiarity with
discourse analysis. The inter-rating
training consisted of an independent
review of the Interaction Analysis model.
Her task was to review the coding made by
the investigator. It was easy to achieve an
agreement of 100% because coding
disagreement concerned only 3 postings in
Debate B. Total agreement was achieved
after discussing discrepancies.
“TextMaster” was downloaded from the
Internet. “TextMaster” is a software tool
for rapid analysis and processing of fixed-length files. This software counts the
number of tokens and types. Each posting
was copied and entered in “TextMaster” to
obtain the number of tokens and types.
TTR was then processed for each posting
through dividing the number of types by
the number of tokens. The value obtained
is referred to as language complexity. The
mean language complexity was processed
for the participants who sent two postings
or more. Numerical data of TTR was
turned into categorical data in order to
process statistical analyses. Values
belonging to the very low TTR levels were
coded 1. Values belonging to the low TTR
levels were coded 2. Values belonging to
the average TTR levels were coded 3.
Values belonging to high TTR levels were
coded 4. Values belonging to very high
TTR levels were coded 5.
The study investigated two online debates.
Debate A comprises 326 participants and
Debate B comprises 118 participants.
Group size in Debate A was coded 1 and
group size in Debate B was coded 2. The
statistical data analysis was based on
descriptive and analytical statistics.
Descriptive statistics were used to
calculate means and percentages of the
selected variables of the study which are
language complexity, knowledge
construction, and group size. Correlation
analysis was used to describe the
relationship between the different
variables. Spearman’s Rho correlations
were computed between different variables
- language complexity, group size and
knowledge construction - to detect any
relationship between them. Simple
regression analyses were computed on
dependent and independent variables to
determine the significant predictors of
knowledge construction. Multiple
regressions analyses were computed on
dependent and independent variables to
confirm simple regression results. The data
were computed using the statistical
Package for the Social Sciences (SPSS)
Findings
Table 2 reveals that in Debate A the
relationship between language complexity
and knowledge construction is positive and
highly significant at the 0.01 level of
significance. It also shows that in Debate B
the relationship between language
complexity and knowledge construction is
positive and significant at the 0.05 level of
significance. These results imply that the
higher the language complexity is, the
higher the knowledge construction would
be.
Table 3 shows that the relationship
between knowledge construction and
group size is negative and highly
significant at the 0.01 level implying that
the less important group size is, the more
important knowledge construction would be.
Table 4 shows that language complexity
has given non-significant results in the
regression equation for knowledge
construction in Debate A. However, Table
5 reveals that language complexity is the
most consistent predictor of the variation
observed in knowledge construction in
Debate B. It accounts for 5.1 % of the
observed variation. The regression
equation is significant as shown by the t-value and the F-ratio.
Table 6 reveals that group size gives non-significant results in the regression
equation for knowledge construction.
Consequently, group size is not a
significant predictor of knowledge
construction.
Table 7 shows that when group size is
added to language complexity in the same
regression equation for knowledge
construction, the adjusted R² falls from
4.3% to 2.8%. Since the t-value is not
significant for the two variables, group
size does not help knowledge construction.
Consequently, the best regression fit is the
simple regression of language complexity
for knowledge construction in Debate B.
Discussion
In both debates the results show that
language complexity and knowledge
construction are significantly correlated.
Correlation is positive and highly
significant in Debate A and positive and
significant in Debate B suggesting that an
increase in language complexity generates
an increase in knowledge construction.
This finding implies that using rich and
complex vocabulary results in consistent
conversations which tend to engender
various ideas, opinions and viewpoints.
Consequently, this could promote
negotiation and higher order thinking.
Furthermore, findings show that language
complexity is a significant predictor of
knowledge construction. Thus, generating
a high lexical variation may foster high
levels of knowledge building. Therefore,
educators should mainly focus on
techniques that promote vocabulary
richness.
Besides, students’ participation may vary
according to the mastery of the language
used. Many learners may feel some
difficulties when communicating in their
second or foreign language which implies
that asynchronous online environment may
be an effective tool in evaluating the
students’ language proficiency.
Furthermore, some actions should be
undertaken to help learners enhance their
language level such as undertaking reading
and writing sessions. The stress should be
placed on English, which is an
international language. Participating in
such debates using the second or foreign
language would be an efficient practice.
Online communication environments are
empowering tools for non-native speakers.
In order to promote rich and consistent
online conversations, students’ online
participation should be fostered. Different
roles can be attributed to students. Some of
them can play the role of moderators. They
may be fight-flaming and stop altercation,
though. Others should have the role of
summarizers, summarizing long and
frequent postings in order to facilitate the
interaction. A group of participants may
also find appropriate theories to back up
informants’ statements, thus playing the
role of theoreticians. Giving such
responsibilities to students will not only
facilitate communication but will also
stimulate them to participate actively in the
discussion, promoting, therefore, language
complexity and knowledge construction.
The results also revealed that the
correlation between group size and
knowledge construction is negative and
highly significant. This implies that high
levels of knowledge construction are
achieved by informants participating in
smaller forums. These findings contradict
the ones reported in Hew and Cheung
(2010). In fact, allowing for an ongoing
increase in the discussion size may have
several limitations. First, it may result in
‘reading without writing’ on the part of the
participants. Second, large groups or
conversations require huge cognitive
efforts from the participants to react to
others. This could result in reading
boredom.
Hew and Cheung (2010) suggest a group
size of about 10 participants in order to
form a critical mass to lead the discussion
to advanced levels of knowledge building
(P.431). Students’ group size should be
limited in order to avoid learners’
exhaustion and withdrawal from the
debate. In fact, a big-sized group often
results in a big-sized conversation; and
students would be overwhelmed by the
number of postings. Limiting students’
number would therefore help them go
through the five phases of interaction.
Limitations and future research
The main limitation of this paper is that it
investigated only two online discussions.
To obtain significant results on the effect
of group size on knowledge construction
in online debates, the number of forums
should be increased. One of the main
limitations of this type of research is the
subjectivity of coding. The classification
of messages is open to individual
interpretation. Using Interaction Analysis
Model is based mainly on personal
opinion. The content might be understood
differently by coders resulting in different
phases of coding.
This research study could be undertaken in
other contexts and by including other
variables. For instance, it could be
conducted in another medium of
communication. Other factors that
influence knowledge construction could be
considered, such as the amount of
participation. Further research is also
needed to discover whether the type of
knowledge or the amount of knowledge
are significant predictors of participation
level and knowledge construction that
occur in online debates. It would also be
quite interesting to study knowledge
construction in online conversations from
a sociolinguistic perspective and find out
how social variables such as age, location,
social status, time or Internet accessibility
could be related to level of knowledge
constructed but future data collection and
analysis are required for more rigorous
findings.
