-------------------
  Findings

The ethnographic fieldwork was conducted at an interactive science discovery centre. Our substantive research focus was based around the production and consumption of ‘science’ in the context of an educative, entertainment setting. Hence we were concerned with the means by which the setting ‘produced’ interactive exhibits and displays of scientific principles and the strategies of learning and sense making invoked by ‘consumers’ (children, teachers and families). Thus the methodological aspects of the project were explored, tested and critiqued through a substantive research project that mapped onto existing research interests of the project team. Through this approach we were able to contextualize digital ethnography as an innovative and potentially fruitful way of understanding the social world and social processes.

The methodological findings of the project have enabled us to identify a number of issues confronting researchers wishing to produce hypermedia ethnographies:

Coding and hyperlinking

Analysis of visual data

Multimedia integration

Analysis/representation

Ethnography as design

Sequentiality

Data management and archiving

 

Coding and hyperlinking
There has been some disagreement in the published literature as to whether a meaningful distinction can be drawn between coding and hyperlinking (see Coffey et. al., 1996; Kelle, 1997, Lee and Fielding, 1991). Our work suggests that these are related but distinct approaches to qualitative data analysis. On the basis of our experimentation, we propose a distinction between the largely associative activity of hyperlinking and the indexing work of coding. Hyperlinking focuses attention on the nature of the diverse relationships between segments of data, whilst coding focuses attention on categories of data segment - through assigning them to common categories. When a piece of data is hyperlinked to another, the nature of the relationship between them is specified (through naming it – such as ‘this contradicts that’ or ‘this follows from that’). When a code is assigned to a piece of data, by contrast, the relationship is usually ‘this is an instance of that’. The advantage of hyperlinking therefore lies in its ability to help the researcher specify the nature of the links within the data-set and to see it as a network of semantic and thematic relationships. Coding, on the other hand, tends to focus attention on groups of data-segments and their content.

However, we do not conclude that coding is inferior to, or should be replaced by hyperlinking. On the contrary, our work has shown that a period of initial coding of the dataset is essential before hyperlinking can proceed. Without this, hyperlinking will be a blind process at best, with the researcher making links through the data-set without a thorough knowledge of its thematic patterns and main contours. Coding allows the dataset to be thoroughly perused and its main themes noted. Once this initial coding is complete, hyperlinking can produce valuable results. Hyperlinking can consist of both data-to-data links and data-to-analytical memo links. We experimented with both of these and found that data-to-data linking is where the innovatory potential of hypertext analysis lies, since it forces attention on the relationships amongst data. Data-to-memo linking is more akin to traditional coding. Both, we found, complement each other. Back to menu.

Analysis of visual data
We identified four stages in the analysis of video data.
• Logging the time code
• Storing and capturing the footage
• Transcribing video footage
• Analysing video footage
This last is the big question that confronts hypermedia ethnographers: to what extent should video fieldwork footage be intensively analysed in the same way as written (transcribed) material? In reviewing the literature on visual methods, we found an imbalance in the extent to which procedures have been elaborated for visual-mode and written-mode analysis, respectively. Whilst published work on the latter is full and detailed, principles of visual analysis for social science have been relatively neglected (here we are referring to fieldwork images, rather than pre-recorded images as in film or media studies). Some studies propose coding video footage in a similar way to written transcripts (Nastasi, 1999). But there has been concern that coding reduces footage to a flattened plane of analysis through disaggregating it into segments and recoding it into verbal descriptions (Pink 2001). Such critiques relate to the well-established disagreement between realist or ‘factist’ approaches to image analysis (Collier and Collier, 1986) and more constructionist approaches (Pink, 2001; Banks, 2001). The latter focus on interrogating the relationships between viewer and image and the contexts in which the image was made, rather than seeing them as records of natural action. They are rarely directive about specifying precise procedures for analysis of content per se.

We utilised both approaches - finding it necessary to treat our video footage both as realist records (which we coded qualitatively according to broad content themes) and as narratives shaped and generated by the researcher’s interactions with specific fieldwork contexts (in which we produced a number of edited sequences of selected field activities and processes). We found this approach produced useful insights into the different kinds of information yielded by the same video data. Our experience suggests that attempts to devise intricate content coding schemas for video will be more suited to technical fields of inquiry (such as kinesics) rather than to the sociological interpretation of visual records as complex representations of social worlds. The multimodality of video footage means that its meaning is produced on a number of levels. We found that coding of video was most useful on the level of categorising footage very broadly into general themes. For more ethnographically-attuned analysis, editing the video material into narrative-governed relationships and scenes produced deeper insights into the interactions between filmed participants (and with the researcher and camera). Back to menu.

Multimedia integration
Literature on visual methods now abounds, but little work has been done on the integration of different media within one project. We found that using multimedia encourages the fieldworker to think of more and diverse materials as potential data and to explore how these are inter-related. For example, rather than routinely resorting to one method we considered the potential of each medium for recording different kinds of data (still photography or video? Image- or sound-recording?) Our work underlines, however, that multimedia ethnography is not a simple process of collecting material in different media and then adding them together. Our findings emphasise the distinctive ways in which different media communicate, and indicate some support for the analysis of multimodality proposed by linguists (see Kress and Leeuwen, 2001). Multimodality emphasises that bringing different media together (as in a typical web-page) produces a meaning-making environment which is not simply a sum of the various media present. Different media deploy different modes of meaning. Our use of various media in our data-records (video footage, audio recordings, photographs, scanned images, written fieldnotes) shows how different kinds of ethnographic insight are afforded by each.
However, we found that there is no such thing as multi-media analysis per se. Our experiments suggest that diverse media cannot be lumped together and subjected to generic analysis. Instead, analysis has to proceed initially by examining each media-specific data-set in turn, i.e. photographs have to be analysed separately from video footage and written fieldnotes. Different procedures of analysis are applicable to different media and contain different kinds of information – for example, an image of a space presents it as a map, whilst a textual account provides a narrative or itinerary of it (see Hastrup, 1992). Analysis needs to take account of these differences. Back to menu.

Analysis/representation
In an ethnographic hypermedia environment all data-records can potentially be made available to the reader through hyperlinking. The division of labour between the analysis and representational/authoring stages of research is therefore blurred - the activity of ‘writing up’ becomes increasingly interwoven with that of analysis. This presents both opportunities and costs. On the one hand, it allows the writer to keep incorporating insights and arguments right up until the moment in which h s/he decides to end the authoring process. On the other hand, there are two related dangers. One is that the authoring strategy becomes impossibly unwieldy and complex. The other is that the later stages of analysis become overshadowed by (re)presentational considerations, so that the demands of writing for an audience (such as readability, screen-design, navigation, and so forth) come to govern the direction that analysis takes. For instance, in the following steps in the analysis process, the last two clearly impinge on decision-making about representation:


• In initial analysis, ‘rough’ coding is followed by intensive data-to-data and data-to-memo hyperlinking (as above).
• In the later stages of analysis the ethnographer begins to write analytic texts (interpretative nodes) to explicate key aspects of the emerging analysis.
• These interpretative nodes will be hyperlinked to selected data nodes - i.e. to the relevant data-records including, where appropriate, video footage/montages, sound and photographs/graphical images.


Consequently, at the end of the analysis stage, there will be a number of interpretative nodes hyperlinked to data nodes. In addition, where appropriate, the latter will in turn be hyperlinked to other data nodes. In finalising these hyperlinks, the author needs simultaneously to consider how to order and structure them for representational purposes. Because hypermedia analysis and representation both take place in the same medium (the computer screen), there is bound to be a danger that analysis will be influenced by screen-design and presentational concerns: this is an issue that needs further acknowledgement and discussion in the literature. Back to menu.

Ethnography as design
We found the representation of our analysis to be a complex part of the project. The technical parameters of the book-form have long shaped the classical rhetorical conventions of ethnographic writing (Atkinson 1990). In writing for and within what Bolter (1999) calls the topographical writing space of the computer, these cannot simply be replicated. The electronic computer screen does not naturally lend itself to the construction of a single rhetorical sequential narrative. Firstly, screenfuls of dense text become unreadable: written text must be shortened. Secondly, hyperlinking means that the reader’s experience will be one of exploring alternative pathways, rather than a fixed sequence of pages. Thirdly, use of the screen’s multimedia dimensions (graphics, video, sound) need to be carefully planned if each screen (singly and together) is to make sense to the uninitiated reader. Fourthly, because the medium is electronic and evolving, one may decide to allow readers to annotate and comment, contributing directly to the ‘text’. At the same time the ethnographic hypermedia environment can be linked to and with other electronic texts/web-pages, undermining the idea of fixed content boundaries.


This means that the ethnographer’s task inevitably becomes focused on matters of design: how best to (re)present the material to the reader on-screen? Different media have to be co-ordinated on the screen. Hyperlinked interpretative text has to be navigable. The data-set must be easily browsable and disorientation avoided (see below). We found it necessary to use the modal resources of the screen (font size and style, colour, screen positioning and lay-out, animation, clickable links) carefully in order to make it clear to the reader what reading-choices to make and how to tell ‘where’ one is at any one time. This was a difficult and time-consuming task and would have benefited from available reader-testers to try out decisions made. Trial and error is resulting in a series of recommendations, which we hope to test and take forward in a future project. Here, we note that these design-tasks are time-consuming need to be factored into such a the project. Back to menu.

Sequentiality
Our work shows that the losses and gains of multi-linear writing have to be carefully evaluated. It seems clear to us that the successful communication of ethnographic interpretation will usually require a step-by-step exegesis. So writers will not want to lose the vehicle of a clear linear sequence in order to present an unfolding argument (or series of arguments). Hypertext can easily be used to ensure a default ‘next’ path through a sequence of pages. Any of these can contain small circular detours – perhaps into the dataset – which nevertheless deliver the reader back onto the path again. The navigational structure can be made as simple or as complex as required. Our hyperlinked trail leads the reader through a sequence of argumentation and controlled exploration of the data, through three interlinked paths. We have tried to be reasonably adventurous with the structure in order to explore the potential challenges. Again, the major challenge is to keep the reader orientated.
Hypertext is well-suited to the illustration of competing viewpoints, multiple perspectives and alternative voices (vis-a-vis the debates in ‘post-paradigm’ ethnography over multivocality and reflexivity). We have included some exemplars of this potential in the prototype ethnographic hypermedia environment, and our plans for dissemination will ensure further experimentation in this area. Back to menu.

Data management and archiving
Preparing a good digital data archive is essential, we found, for two reasons. Firstly, the hypermedia approach we favour presupposes that readers of the research can explore the entire data-set upon which analysis has been conducted. In this way, original interpretations can be fully interrogated, alternative readings proposed and secondary data analysis carried out. Secondly, our experience shows that hypermedia analysis/representation can only proceed on the basis of a very organised and well-prepared data archive. The time-consuming preparatory work involved in assembling will be rewarded when it comes to constructing the ‘final’ EHE, for it will ensure that readers can navigate and orientate themselves around the dataset while following their interpretive trail (see above). This data-archive should contain hyperlinks allowing the reader to browse and navigate through the contents with ease. Back to menu.

References

Destinations