Results

Overall, the answer to the question "How does collaboration on a text work in a text based virtual reality?" can be answered by stating that other than adding an additional layer of metacognition to deal with the questions of "how do we decide what to write" and "how do we use this environment to write" and so on, the process is remarkably similar to that observed in single-writer-self-talk protocol analyses, modulo a drastic increase in the first layer of metacognition, in which the text events coded as textmaking occur.

As can be seen in the diagram above, the process of collaboration on a MU* begins with tool-making planning, that is, determining the ways which the collaboration would take place, or establishing, as Kesey put it, ground rules. After this came the process of planning the writing itself. Text planning consumed far more time than when they wrote individually, according to the subjects in the postmortem, due to the need to make sure they were all writing the same story. Initial writing followed, initiated by one writer based on the emerging consensus. The paragraph was then reviewed by the others, and as the diagram shows, many things began to occur. The paragraph was revised, causing some re-planning of the text, more text was written, the two paragraphs were evaluated together, more re-planning occurred, more review occurred, more planning of new text occurred, and in general the recursive writing, planning, evaluation, and revision pattern common to protocols done on single writers began to emerge.

What also began to occur during the writing phase(s) was toolmaking evaluation - that is, the methods which the writers had agreed on, and by which the collaboration was done began to be evaluated and in some cases revised. When this occurred, it was generally followed by evaluation of the text, which returned the writers to the recursive text making process.

At the end of the experiment too, the writers went from the text-making writing phase to the tool-making evaluation phase and agreed that in light of the time, technical difficulties, and the fact that several paragraphs had been generated, the experiment could end. The terminal state of the experiment, then, is the toolmaking-evaluation phase.

Overall, the resemblance of this protocol to those done on individual writers engaged in self talk during writing is significant, particularly in the recursive interaction of the various textmaking steps. It stands to reason that once the writers emerged from the planning stage to the writing stage and were then interacting with the JED editor that they would discover that their original tool-making planning did not cover some contingencies encountered during the editing session.

Interesting too, is the additional layer of metacognition - not only what shall we write, how shall we say what we want to write, but also how shall we proceed in this environment to the business of writing, and how shall we resolve the technological issues that arise as we write. A number of the writers commented both during the study and at the post mortem how long it took to generate the fairly small amount of text that was actually produced. In part this can be attributed to this extra metacognition that these experienced solo writers did not normally encounter, although the actual number of toolmaking events, which make up the second metacognitave layer, account for only a relatively small number of the total text events, as shown in the diagram below.

Planning (toolmaking) Planning (textmaking) Writing (textmaking) Review (textmaking) Evaluation (toolmaking) Evaluation (textmaking) Replanning (toolmaking) Replanning (textmaking) Revision (textmaking)
50 263 8 161 14 57 22 39 19

The largest number by far of text acts were involved with planning the text, rather than toolmaking planning, in spite of the technical difficulties experienced. When combined with textmaking review, the planning and review of the text constitute the majority of the text acts in this project. By contrast, the relatively small amount of actual textmaking writing events, combined with the content of some of the discussion, suggest that the subjects found the JED software difficult and time consuming enough to use that it restricted the amount actually written. In any case, it is clear that the subjects spent only a little time on coming to grips with their environment before diving into the process of planning, writing, reviewing, evaluating, replanning, and revising the text.