COMPUTERS and COMPOSITION 8(3), August 1991, pages 83-88

As early as 1980, Allan Collins and Dedre Gentner were theorizing about the potentials of "WritingLand," a computer-based reading and writing environment that might assist students by suggesting strategies and structures to negotiate the several levels of cognitive activity in composing. [1]

Now, a team of a dozen interdisciplinary scholars and researchers report on "the progress being made toward the creation of ideal computer writing environments for college students and professional writers" (ix), defining ideal computer writing environments as "those that support all of the cognitive processes fundamental to good writing" (ix). Psychologist Bruce Britton and educational psychologist Shawn Glynn join three other psychologists, two English professors, two computer scientists, a linguist, a technical communications specialist, and a humanities educator to provide snapshots of the quickly changing computer-support systems available for writers.

In doing so, the authors use a substantial base of theory and cognitive research to validate the appropriateness of computer support for writers in electronic environments. The six chapters of Computer Writing Environments describe computer management of writing and augmentation of cognitive processes, a variety of student approaches to electronic writing, the status quo of present writing aids along with desirable future enhancements, and work in progress on a "comprehensibility aid" for technical manuals.

Britton and Glynn, with co-authors Denise R. Oaks and Linda F. Mattocks, lead into the book with their chapter, "Computer Environments for Managing Writers' Thinking Processes." They praise the computer as a more recursive environment for writing, less linear than writing by hand or by typewriter. Their imagined "wish list" for an ideal computer-writing environment includes five goals: 1) the system should facilitate interaction and recursion, structuring not just bottom-up grammar checks but also top-down organizational probes; 2) it should be graphics-based as well as character-based, using icons and spatially represented information; 3) hypertext capabilities for searching, indexing, and document variation should allow for manipulation of large and multiple documents; 4) multi-tasking should allow flexibility and quick switching among texts, data, and graphics; and 5) group and individual writing environments should facilitate team composing and interaction.

Such computer environments will affect teaching and learning strategies in the writing classroom, but they will not threaten job security: "At the present time, and in the foreseeable future, word process[ing packages] cannot replace the instructor who fulfills the important roles of guide, collaborator, editor, and audience" (p. 11). Student writers using computers may revise more, gain fluency (which may be explained by the absence of the end-of-the-page syndrome), become more open to peer review, and enjoy writing more. Teachers will benefit from these improvements in process while remaining critical of programs that isolate skills, such as the grammar subprogram Grampop. Although such support is available for writers, the authors point out that the real promise for computer-assisted writing instruction lies in "[t]he integration of programs for managing top-down and bottom-up processes, [which] represents a quantum leap in word processing" (p. 15). Nevertheless, such advances will require refinements in artificial intelligence that have only been imagined, not achieved.

Chapter 2, by John B. Smith and Marcy Lansman, is titled "A Cognitive Basis for a Computer Writing Environment." Research by cognitive psychologists in reading comprehension and written communication reveals schema for decoding and gisting (creating memory representations) propositions, coherence links, and hierarchical macrostructures. Certain typographical and rhetorical conventions, such as thematic titles, advance organizers, signaled or cued structures, and topic or theme sentences in initial positions, aid readers and suggest effective guidelines for writers.

Writers' cognitive processes are analyzed with the Flower and Hayes model, which uses three types of representation common in cognitive psychology: the box model, the flow chart, and the production system. A matrix of constituent concepts and cognitive modes for writing (p. 36) weaves together the interdependencies of processes, products, goals, and constraints with seven modes or particular ways of thinking involved with writing. The various cognitive modes that are analyzed range from exploration, situational analysis, and organization, to writing and three kinds of editing (for global organization, coherence relations, and expression).

The design of their WRITING ENVIRONMENT (WE) system emphasizes multimodal cognitive support for writers. The network mode is tailored for exploration. Here, computer screens allow for a graphic clustering type of brainstorming for ideas and relationships. The tree mode allows the writer to organize ideas into a structured hierarchy. The editor mode is for drafting, linguistic encoding, and expansion of concepts. The text mode provides editing support. Although the names seem rather arbitrary and perhaps interchangeable, the rich environment offers flexible guidance and support for a writer's complex tasks.

The authors test their system with users in three ways. An automatic tracking function records a transcript of each writing session. Session replays (speeded up or slowed down) prompt retrospective think-aloud protocols from writers. A grammar parser reduces data from the protocol tracker into a parse tree and infers cognitive processes and modes associated with particular items in the text product.

Ronald T. Kellogg offers a rationale for augmenting word processing to enhance human cognition in chapter 3, "Idea Processors: Computer Aids for Planning and Composing Text." The power of computing can address three common problems of writers: attention overload, idea bankruptcy, and affective interference. Three novel computer functions, in addition to "Computation" and "Memory Retrieval" (archival) functions, can help writers overcome these difficulties.

Allocation of attention can be guided by a "Funnel" function in which the writer's attention is channeled electronically. This may be done by hiding distracting information, such as collapsing outlines or paragraphs, or by encouraging on-screen free-writing by using flashes that prompt pauses. Eventually, physical or mental workload may be further reduced with "Listening Typewriter" functions using voice recognition.

Failure to generate usable ideas, memorably labeled "idea bankruptcy," is addressed by "Inventor" functions to aid forming concepts and relationships. Whether a hypertext NOTECARD device, a tagmemic matrix, or visual synectics, such "inventory devices" enrich writers' thinking. As with any invention aid, such programs generate new ideas only by processing the writer's preliminary (and ideally secondary and tertiary) thoughts on a topic.

Affective interference comes about because of the fears, anxieties, and other emotional reactions students bring to writing. The computer-as-therapist reduces frustration, apprehension, lack of confidence, blocking, procrastination, and other neuroses, Kellogg asserts. Covert reinforcement can be embedded in the programming by choice of words and phrases for on-screen response to writers' behaviors. Other programs, like MORTON, are specifically designed for overt cognitive therapy or perhaps even contingency management treatments.

Kellogg uses Bridwell-Bowles, Johnson, and Behe's classification of "Mozartian" and "Beethovian" revisers to conclude that these funneling, inventory, and therapeutic computer aids are especially effective with Mozartian executors who plan extensively before producing a polished draft. Beethovian discoverers, who compose rough, exploratory drafts, have been less satisfied with augmented word processing.

In chapter 4, "Word Processing and Student Writers," Michael M. Williamson and Penny Pence focus on the value of word processing as a window on composing strategies. They study novice student writers using word-processing systems with dribble files, which save keystroke records. Measuring revision episodes, or chain reactions of text changes, they identify three basic styles of revision: linear revisers who save revisions for last, intermittent revisers who stop periodically, and recursive revisers who continually embed revisions at the point of text generation. Such close attention to texts-in-progress can be disabling, or at least slowing, for basic writers, but Williamson and Pence label the linear style as weaker and praise the recursive, highly embedded revision style of expert writers. These professional writers exhibit "metacognitive awareness;" they know what they're doing. Word processing facilitates a reconceptualization or externalization of the composing process necessary for growth as a writer. The ecology of the laboratory/classroom encourages tinkering with texts, even "slash and trash" rewrites, and "facilitates the dump of human memory into electronic memory" (p. 122). These researchers conclude that word processing frees writers from "the linear conception of writing that is an artifact of writing on paper, and it frees them from the messiness of recursive revision strategies on paper" (p. 122).

In chapter 5, "A Computer-Based Writing Aid for Students: Present and Future," Morton Friedman and Earl Rand analyze WANDAH and present their vision for a SUPERWANDAH of the future. As part of the research and development team that produced WANDAH at UCLA, these authors share their wish list for potential enhancements. WANDAH was designed for cognitive problem solving, with attention given to composition theory and research as well as the human-machine interface, using the Flower and Hayes design model and a toolkit of heuristic devices. While keeping computer jargon to a minimum, the program displays good screen and keyboard design, excellent error correction, and extensive on-line help. It automatically formats output and has a split screen for word processing and for a commenting section ("writers need readers" (p. 134)).

Possibilities for a SUPERWANDAH by 1993 include the addition of templates for supporting details, examples, quotes, or particular subject matter: i.e. the psychology experiment or thesis proposal. [2] Further enhancements that Friedman and Rand speculate about include review and revision checklists plus multi-tasking and windowing to provide a richer and more active environment. Next-generation programs, such as intelligent tutoring programs, would still have limited knowledge domains unless quantum jump improvements are discovered.

David E. Kieras concludes with chapter 6, "An Advanced Computerized Aid for the Writing of Comprehensible Technical Documents." Interested in improving the clarity of writing and in providing more comprehensible technical manuals, Kieras contrasts automated reading detection processes within reach of current computer technology with judgmental and stylistic correction processes that remain in the human domain. Computers can perform the sequential stages of comprehension: word identification, syntactic analysis, semantic analysis, and pragmatic and functional analysis. An advanced writing-aid system can feasibly "identify some of the comprehensibility problems that can occur at each stage of the comprehension process" (p. 151). Kieras adds that computers, at best, imitate poor readers, exhibiting "artificial stupidity" (p. 152). Such natural-language processing is difficult to implement given the status quo of artificial intelligence. Yet he describes a demonstration system designed on the principle of given and new topics and referents. Parsers and augmented transition networks can analyze the grammar, but comprehensibility issues include more than grammaticality. Additional issues of user-interface and usability are critical, so "close integration with the text-editing environment and very high execution speed" are recommended components for such a system design.

With only occasional lapses into unexplained jargon, [3] the authors present a complex theoretical and technical overview of the realities and the possibilities for computer-assisted composition. Although their reliance on cognitive research might be expected to limit their concerns to individual mentation, their focus on the computer-writing environment as a location of cognitive support includes social interactions and cultural contexts that are external to the writer's psyche. The electronic medium is a support structure of the brain (not equivalent to or substituting for the brain), located in the socio-cultural milieu of the classroom, laboratory, or workplace. The authors' defense of computer writing as a whole environment for writers becomes an invitation to teachers and writers to make full use of such electronic support for the various stages of composing, from planning and inventing to reviewing, revising, and stylistically improving our written products.

The concept is catching on. Jane Zeni continues the contagion in her book, WritingLands: Composing with Old and New Writing Tools, recently published by NCTE. Her investigation moves from support of individual writers to classroom environments and communities of writers in computer-enhanced schools. Converts and doubters both will want to read Zeni's book as well as Britton and Glynn's to keep current with the profession's reasons for such enthusiasm about electronic writing.

1. Collins, A., & Gentner, D. (1980). A framework for a cognitive theory of writing. In L. W. Gregg & E. R. Steinberg (Eds.), Cognitive process in writing. (pp. 51-71). Hillsdale, NJ: Lawrence Erlbaum Associates.
   
   
2. William Wresch's WRITER'S HELPER STAGE II offers a limited number of such formats.
   
   
3. Williamson and Pence's embracing of methodological pluralism is understandable, but for the lay-reader their "2x2 between-subjects design with a single within-subjects factor change" is not (102).
   Kieras defines ATN (augmented transition network) and PPS (parsimonious production system) but slips in two uses of ACT (evidently some different kind of semantic network representation) without explanation (155, 159).
   

Collins, A., & Gentner, D. (1980). A framework for a cognitive theory of writing. In L. W. Gregg & E. R. Steinberg (Eds.), Cognitive processes in writing. (pp. 51-71). Hillsdale, NJ: Lawrence Erlbaum Associates.

Zeni, J. (1990). WritingLands: Composing with old and new writing tools. Urbana, IL: NCTE.