Education-line Home Page
Papers from the 28th Annual SCUTREA Conference
Research, Teaching and Learning: making connections in the education of adults

Enquiring into Group Learning On-Line

Celia Graebner, Centre for the Study of Networked Learning, University of Sheffield, UK

On-Line Discursive Environments for Adult Learning

Formalised computer-mediated learning through asynchronous discussion in what we would recognise as an adult learning mode dates at least from the Western Behavioural Sciences Institute’s experimental seminars of 1982 (Feenberg, 1989). Yet many of the issues arising from these early experiences, and reflected on by Feenberg in his contribution to the seminal 'Mindweave' collection, remain live and problematic for facilitators and designers of on-line learning.

With the availability of the Internet as a medium, computer-mediated communication (hereafter, CMC) exercises a growing attraction for all sectors of education; both for the kinds of intrinsic promise recognised by its early proponents, and increasingly for extrinsic reasons, under the influence of national and international policy decisions on delivery and organisation of recurrent education. This paper reflects on some motives for researching CMC discussions; some styles of investigation used, and their purchase on the interactions between individuals and groups in on-line distance learning; and suggests, with some small illustrations, how a pluralistic investigative approach can support co-operative learning and reflective practice in on-line adult learning settings.

The particular model for organising learning on-line which this paper assumes as relevant, on account of its congruence with the andragogical emphasis on group interaction and self-directed learning, is that of a 'virtual learning community'. This is the perspective followed by the MEd in Networked Collaborative Learning at Sheffield, from which the detailed illustrations in later sections of the paper will be drawn. A brief extract from the course description will serve to characterise the responsibilities of members of a learning community:

[to] manage their own learning, and co-operate with others in theirs through processes of negotiation and discussion

[to] participate in developing the learning community perspective, which is based on participants and tutors taking collective responsibility for the design and evaluation of he programme, via constant review and modification of the design, procedures and ways of working

(1998 MEd Course Description: online at

In considering the implications of researching this collaborative style of learning, the paper draws on the more extensive literature of virtual communities in general.

On-Line Discussion as Retrievable Interaction

In some respects, the CMC discussion medium is uniquely fitted for researching practice, since all utterances are captured and archived along with basic contextual information. Moreover, most discussion software includes rudimentary tools for re-ordering the discussion contributions - by sorting, searching or indexing - so that alternative perspectives on the interactions can be derived from, and to some extent within, the discussion environment. (They can, equally, of course, be reported back there.)

For a cogent summary of the implications of participating in the 'written community' of on-line discussion, we can turn again to the Feenberg chapter mentioned above:

A group which exists through an exchange of written texts has the peculiar ability to recall and inspect its entire past. Nothing quite like this is available to a community based on the spoken word. The modification of language through CMC can best be understood as a new variety of ‘social memory’ comparable to such other mediated memories as storytelling, books, and mass communications.

(Feenberg, 1989)

The lack of tacit cues in this written group interaction dictates compensatory practices: the only tacit sign we can transmit is our silence, a message that is both brutal and ambiguous. . . the solution to this dilemma is explicit meta-communication... participants must overcome their inhibitions and demand further information. . . request clarification of emotional tone and intent.

(Feenberg, 1989)

Where these compensatory techniques are well-developed in a group, the record of a CMC discussion can, paradoxically, be more complete than a transcript from an interaction embodying significant tacit elements.

Retrievability can also be a resource for democracy, empowering the individual by allowing space for reflection to make sense of discussions, and formulate views, without the face to face group pressure for instant assimilation and interpretation:

In principle there is no reason why [such] retrieval should not be entirely under individual control since the technologies of retrieval do not require the presence of other human beings.

(Feenberg, 1989)

Thus CMC has an inbuilt potential for supporting reflection and co-operative enquiry. How any of these possibilities are taken up, however, is in practice closely coupled with the social context or climate of a particular CMC implementation.

Interpreting CMC Discussion - Quantitative and Qualitative Orientations

Large bodies of text data with associated contextual detail invite abstractive and replicable styles of analysis. Quantitative measurement of on-line activity derived from system embedded functions was quickly adopted in the pioneering phase of educational CMC to demonstrate the flexibility of access arrangements and the emancipatory effect on learners. System information about levels of activity, spread of participation or volume of messages provided generalised phenomenal descriptions from which broad conclusions could be drawn. (See, for example, Harasim’ s (1989) use of box-plots of on-line activity.)

The level of sophistication possible with more recent software technologies for embedding micro-scale analysis of messages is illustrated by the approach of the Project H research group (Berthold, Rafaeli, Sudweeks, & Coyne , 1997.) Their research developed a method for deriving characteristics of an effective CMC communication by training neural network software to check for co-occurrences of a range of parameters in individual postings. In the corpus of 3000 newsgroup messages investigated, the number of constitutive parameters considered was nearly thirty - giving some sense of the complex multidimensionality of even an unstructured interaction space. It needs to be mentioned that in this investigation, although the inferential process was handed over to software, the messages had been pre-coded by researchers for characteristics such as humour.

Whatever the scale of investigation, these broad quantitative styles of analysis read the discussion material as a single artefact, analogous with a literary text, consequently collapsing the temporal (or more precisely, for an asynchronous discussion, sequential) dimension of its creation. They thus offer scant purchase on claims such as Feenberg’s about the constitutive practices which develop a social reality.

Approaches which most respect this temporal/sequential dimension are those qualitative, broadly grounded styles of interpretation which use the participant observer approach, handle a discussion transcript as conversation, or reconstitute it as narrative (eg , McConnell, 1994, 1998). The multi-dimensionality of the many-to-many discussion environment poses problems for this investigative style, beyond the gross level of information management. Reconstitution of the discussion transcript as narrative or periphrasis, by attending closely to the sequential, recognises individual actors and their (apparent or recovered) intentionalities, but risks reducing the activity to a single convergent space. It may also underplay the performance aspect of participation, the aspect of self-creation on-line identified by Feenberg (1989) or Paccagnella (1997), in favour of an essentialising 'authenticity' of individual experience.

Contextualising Investigation of On-Line Interactions

In a recent discussion based on empirical research with an undergraduate population using CMC, Jones and Cawood (1998) caution against adoption of the on-line discussion transcript as an authoritative account of learning interactions, criticising it as both partial and misleading. They point to supplementary activity absent from the transcript - social interactions and negotiations occurring outside the CMC environment; and to the status of the transcript as a publicly engineered record.

Their caution is worth heeding, but their findings are heavily influenced by the specific source used in their study: a single network-delivered module in the context of a place-based undergraduate degree course in communication studies, a module which also included supporting face to face meetings. Their ethnographic observations had indicated that course members actively constructed through off-line collaboration an on-line artefact to be presented to their tutors for assessment. The social context was one of pre-existing face-to-face networks, a competitive learning culture, and a perception of conflicting interests between teachers and taught, crystallising around issues of assessment of the on-line discussion contributions. The students’ ability to bend the technology to their own ends imay be taken as a reminder that:

groups are realities in their own right, with socially specific needs that must be served by CMC technology.

(Feenberg. 1989)

In itself, the difference between collaborative and competitive learning cultures would produce a different practices in discussion groups.

The remote, as opposed to place-based, learning situation changes the status of transcript within the learning group: without a parallel strand of face to face activity, reality is invested in what takes place on-line (and everyday experience may take the role of the complementary inferred dimension - cf Paccagnella’s (1997) observations on the status of off-line identity as a fiction from the perspective of on-line communication). And where a group comes together for the first time without any recognised common context, the place of group convergence needs to be deliberately constructed within the space of on-line discussion - the transcript of such activity may perhaps be interpreted as an archive of a collective performance.

If the role and status of the discussion transcript varies in different learning situations, the researcher may need to consider divergent methods for divergent contexts. The following sections address the question - what interpretative approaches are appropriate for co-operative group learning which takes place substantially or entirely on-line?

The On-Line Community as Context of Investigation

In the longitudinal study made by Baym of a group constituted around a specific external domain of interest, the RATS soap opera Usenet discussion forum (1992-1995), a distinctive range of social conventions and styles of interpersonal support was observed. Baym concluded that what constitutes community in on-line interaction is the emergence of common understandings through a body of practice:

Community is generated through the interplay between pre-existing structures and the participants’ strategic appropriation of the resources and rules those structures offer in ongoing interaction. (Baym, 1995)

Following from Baym’s analysis, the investigation of community requires the identification of methods for studying the development of custom and convention. and of social roles and identities, through the changing use of the on-line environment over an extended time period.

In the case of potential communities primarily focused on learning activities, learning tasks and resources will serve as significant structures around which community can be developed, and in relation to these specific questions need to be addressed, particularly:

is the group engaged in the productions of shared knowledge about its domain of interest ?

as opposed to creating a shared social culture - however valuable that might be as a support for individual learning;

does the group cultivate a sense of its own learning process?

in other words, does its collective memory encompass the learning that has been manifested in individual contributions.

Paccagnella (1997), following from Baym’s analysis, argues that only a naturalistic and longitudinal approach is adequate

to discover and understand the dynamic and the slow evolution of the specific culture and social climate of a particular piece of cyberspace. There is therefore need for future research that, besides being conducted in the field and not in laboratory conditions, is also longitudinal.

and that for naturalistic analysis of on-line material, the software-embedded tools are an essential resource, since they avoid the interventionism inherent in both survey and participant observer approaches. Adopting Weberian terminology, he proposes that the software tools carry the function of explication (Erklaren) leaving the complementary interpretation (Verstehen) to the researcher.

The Verstehen for a collaborative learning group, however, could also be construed as individual and collective reflective activity. The diagram below illustrates how data on interactions can contribute to a cycle of collective experiential learning.

Diagram 1: Investigation of CMC transcript as a contribution to the learning cycle

Diagram 1

The Place of Aggregative Approaches in Investigating Learning Community

Appraising embedded information in CMC transcripts from the perspective of collaborative learning, Guzdial (1997) proposes that aggregative date can provide a kind of ecological profile of a learning forum, which allows the investigator to observe whether some of the pre-conditions of collective learning are being met and which because of the relevant ease of gathering the data can be used as a formative evaluation tool.

An example relevant to the theme of community development would be the 'recency' effect observed in Web-based communication whereby on-line messages are perceived as having a very short effective life, after which they no longer attract responses, almost irrespective of content. If we are concerned with community behaviour as a process of aggregation and accumulation, we might wish to investigate the time-span over which messages are responded to within our learning group , and if a recency effect is seen to be occurring to the detriment of continuity and accumulation of meaning, to consider ways of designing the on-line environment or managing online interaction to counter it in the interests of community development.

Investigating those activities identified earlier as crucial for a learning community - collaborative knowledge generation and the collective monitoring of the learning process - would involve extending the comparative ecological top a variety of levels of investigation, encompassing smaller entities in the learning environment. To investigate, for example, inter-topic and inter-message relationships within a single discussion area, some finer-grained approaches such as network representations or content analysis may be helpful. Although not embedded in the CMC environment, these may be software facilitated.

While it is beyond the scope of this paper to provide a taxonomy of investigation options, the table below gives some examples of investigation at different levels of granularity. Each level of description provides different types of material for an interpretative activity, which may take place within the discussion area, as a contribution to the cultivation of the learning culture. (Example 4 below shows this happening informally.)

Table 1 - Fields of enquiry, types of evidence and issues for co-operative reflection

                               AVAILABLE FROM        COOPERATIVE REFLECTION   
                           DISCUSSION TRANSCRIPT                              

Whole CMC environment     relative patterns of use  How have different        
                          for areas of different    discussion areas          
                          functionality: topics     contributed to            
                          initiated, pace and       learning/?                
                          quantity of interactions                            

Single discussion area    topics instituted  in     How have patterns of      
                          relation to  function of  use, and focus of         
                          the discussion area.      concerns over time        
                          Activity of individual    shifted?                  
                          participants: frequency   What roles have been      
                          and types of response     adopted by different      
                          (substantive, social,     members of the group,     
                          etc).                     and in what contexts?     

Single content-focused    variety of contributors,  What common learning      
discussion strand         sequencing of             concerns are being        
                          contributions,  focus of  shared?                   
                          message subjects;         How are meanings being    
                          occurrence of common      negotiated?               
                          concepts, semantic                                  

Single process-focused    as previous level - plus  What collective insights  
discussion strand         explicit reflections on   have been gained from     
                          other messages and        the discussion? How are   
                          discussion process        these preserved and       

Contextualising Interpretation for the Learning Community Context

Only a few brief examples of these different levels of investigation can be illustrated here from the discussion data.

Example 1: Overview of a differentiated CMC environment

The transcript under consideration consists of over 1400 messages posted to whole group discussion areas by a cohort of twenty-one part-time Master’s students and five course team members, as contributions to topic strands initiated over the fist six months of their course. Just over half of these were in the 'general purpose' area; the remainder spread across a library and resource a technical support area, and an additional social area set up to bring together both cohorts on the course. Much discussion also took place in small closed groups ('learning sets') where the focus was on learning issues, projects and collaborative assessments, but we focus here on the potential community as a whole. (The design of the overall learning environment and its motivation are detailed in Graebner, 1998.)

Table 2: Overall volume of postings to 4 open discussion areas over 6 months

General Discussion   Technical Support        Library &          Social Area      
       Forum                                  Resources                           

        733                  308                 113                 268          

Where there are plural and differentiated community spaces an aggregative view can begin to characterise the ecology of different spaces, and can raise questions about shifts over time. The example below compares the rate at which new themes were introduced into the different discussion areas, and also illustrates the number of dimensions in the data which may need to be investigated; in this case an overall trend to longer time-spans and more inclusive strands, as well as changes of balance between whole group and the subgroup areas of activity seasonal underlie the general decrease in new topics.

Table 3: New topics in discussion areas by month

General Discussion  Technical Support   Library & Resources      Social Area     

F. 13                                15          9                            14 
M.  6                                 9          3                             1 
A. 10                                 4          5                             3 
M.  5                                 1          3                             1 
J.  2                                 0          1                             1 
J.  4                                 3          0                             0 

Example 2: Evolution of a single discussion area - the Technical Support forum

This was the area set up to address immediate concerns about use of working environment and report problems. As the space where success with setting up the communications software was to be reported, it was ensured broad initial participation. Half of the course cohort originated topics in this area over the six month period.

The area had a resident expert who was also the system administrator for the conferencing software. Of the 15 new discussion threads in the 1st month of use, 5 were initiated by the resident expert. However, all subsequent topics were initiated by students - or occasionally tutors. Most of these (17 out of 27) were requests for clarification, or worries about malfunctions. In other ways, the discussion went beyond a simple technical 'question and answer ' pattern:

although new topics in the first six weeks were largely reactive, subsequently a new type of strand began to appear - raising 'what if' issues about the software environment, including enhanced or alternative uses (7 topics);

there were also some (self-consciously ) stereotypical roles being played out from very early on: the naive user, the Luddite, the Macintosh (or PC) partisan.

Overall, this practically focused area seems to have provided a strong initial base for co-operative interaction.

Example 3: a single content-focused discussion strand

The network below represents a single discussion strand which begins four weeks into the on-line period and runs for eleven days, spanning thirteen messages. Within the pattern of interaction discussed for the whole area, a group member is seeking an explanation of a feature of the Notes software. Like many discussion environments, Notes helps users to manage the flow of information by providing visual marking of unread messages; but in some cases this seems to the user to be applied inconsistently.

Q is the initial questioner, Pn other course group members, RE the resident expert; course tutors are represented as T.

Diagram 2: Content-focused discussion thread

The right hand branch of this thread shows a linear interaction of a 'Q and A' pattern between enquirer and expert, typical for technical support interactions on-line: RE’s first explanation - Q’s observations on its goodness of fit - RE’s acknowledgement of the observations. But in the interim, peers are offering a variety of responses in different roles - supportive ( 'me too'), explanatory hypotheses, and general banter. In his first posting, the expert acknowledges the contribution this sharing of ideas has made to his conception of the problem; although he takes responsibility for a final (though partial)solution, the co-operative input has been very significant.

Example 4: a single process-focused discussion strand

This last example illustrates the use of a general discussion area as a space for reflection on community process. As often in a process-oriented strand, the initiating message takes the form of a request for opinions, referring back to the author’s summary of a previous more general discussion thread:

Diagram 3: Reflective discussion thread

The posting attracts immediate responses from three course members; although the breadth of response is not great, the discussion continues over eight or nine weeks. The most vigorous chain of responses actually represents the spinning of a collective fantasy about virtual versions of a real-time experience, a direction tangential but by no means irrelevant to the main theme; and it would seem to be a measure of the maturing of this community that perspectives from this playful 'digression' are taken up in the general reflections on the separation between on-line and off-line realms of experience which follow them, thus contributing to the collective self-knowledge of the group.

Issues Raised by the Illustrations

The examples of interactions depicted here have been taken up in retrospect.

Since the cycle of learning from on-line experience cannot be considered complete until interpretations of on-line patterns of activity are taken to the group for reflection, a more productive approach to learning from past interactions could be for a community to adopt a more integrated and timely interpretation process for itself. However, the practical and ethical demands of such an approach would be very considerable, and there is arguably a continuing role for the reseracher as initial and tentative chronicler of on-line social realities.


Baym, N.K, (1995):'The Emergence of Community in Computer-Mediated Communication' in Jones, S.G. ed, CyberSociety: Computer-Mediated Communication and Community, London, 1995

Berthold M.R., Sudweeks, F. Newton, S. and Coyne, R.D.,(1997) 'Clustering on the Net: Applying an auto-associative nerual network to computer-mediated discussions'. Journal of Computer-Mediated Communication, 2,4 (March,1997) On-line at: http://www.usc.annenberg/jcmc/index.html Accessed February 1998

Feenberg , A., (1989): 'The Written World: on the Theory and Practice of Computer Conferencing' in Mindweave, eds Kaye, T., and Mason, R., Oxford, 1989.

Graebner, C.F., (1998): Networked Lifelong Learning: Innovative Approaches to Education and Training through the Internet, eds Banks, S., Graebner, C., and McConnell. D.. Sheffield, 1998.

Guzdial, M., (1997): An Information Ecology Perspective of CSCL. Keynote address, CSCL 1997, Toronto. On-line at: guzdial/papers/ Accessed February 1998

Harasim, L. (1989): 'new Domain' in Mindweave, eds Kaye, T., and Mason, R., Oxford, 1989.

Harasim, L., (1995) 'Networlds: Networks as Social Space' in Harasim, L ed, Global Networks: Computers and International Communication, MIT, Cambridge,1995

Jones, C., and Cawood, J, (1998) 'The Unreliable Transcript, Contingent Technology and Informal Practice in Asynchronous Learning Networks' in Networked Lifelong Learning: Innovative Approaches to Education and Training through the Internet, eds Banks, S., Graebner, C., and McConnell. D.. Sheffield, 1998.

McConnell, D., (1994) Implementing Computer-Supported Co-operative Learning, London, 1994.

McConnell, D., 1998 'Developing Self-Assessment in Networked Lifelong Learning Environments' in Networked Lifelong Learning: Innovative Approaches to Education and Training through the Internet, eds Banks, S., Graebner, C., and McConnell. D.. Sheffield, 1998.

Paccagnella, L., (1997): 'Getting the Seats of Your Pants Dirty: Strategies for Ethnographic Research on Virtual Communities' Journal of Computer-Mediated Communication, 3,1 (June 1997) On-line at http://www.usc.annenberg/jcmc/index.html Accessed February 1998