"The way we see things is affected by what we know or what we believe." (p. 8)
           --John Berger (1977), Ways of Seeing.

In her article, "Student Writing: Can the Machine Maim the Message?" Marcia Peoples Halio (1990) raises some compelling questions. Can the features of particular computer systems affect developing writers in different ways? If so, how should teachers respond or adapt to these differences? Because these issues beg an even more fundamental question--How does computer-mediated writing differ from conventional, typographic writing?"--they are bound to arouse controversy at a time when technology may be precipitating an identity crisis in humanities education (Lanham, 1989).

There has been plenty of controversy. Halio's article has elicited extensive critical response on electronic conferences and in other forums, where specialists in computers and writing have called Halio's claims unwarranted and her observations disputable. Most of the discussion so far has focused less on the questions Halio poses than on the way she presents her observations as if they were answers to these questions.

A more productive reply must address broad issues involving writing, technology, and the goals of education. But such a reply is impossible within Halio's perspective, which pits "good" writing technologies against "bad" in a dispute between devoted users of two microcomputer systems. Such a dualistic view is only one rather narrow way of seeing. In this article, we reject the specious "Macintosh versus IBM" contention and instead explore larger questions about the way technology may influence education and about the growing importance of graphic elements in writing.

These issues need much more extensive investigation. But in pursuing this research agenda, investigators must carefully consider the relationship between the nature of the questions posed and the methods of inquiry used to address them. How we define the questions may well determine what kinds of answers we find.

Before taking a broader view of the issues, however, we feel it is necessary to clear up misconceptions about Halio's assertions. We are mindful in doing this that some of our colleagues feel criticism of this work might discourage future descriptive research. Descriptive accounts, case studies, and ethnographies have an important place in writing research, but authors must be cautious both about the limits of their observations and about the ways they generalize. Halio has not been cautious in both respects, yet her article has gained surprisingly wide credence. We begin with a detailed critique here because Halio's article has been circulated at some institutions with comments suggesting that its findings prove friendly user interfaces are detrimental to student writing. Her observations cannot support such a broad conclusion.

Although Halio acknowledges that her data are anecdotal and inconclusive, she nevertheless presents findings based on them--under the heading "Suspicions Confirmed"--and, on their strength, speculates that teachers may need to be cautioned against the pitfalls of oversimplified user interfaces. The validity of the observations reported in the article is highly questionable, and those observations fail to demonstrate any causal relationship between aspects of a computer interface and differences in writing. Thus, her contention that graphic interfaces impair student writing does not stand up to scrutiny. Halio's rush to judgment is both misguided and potentially misleading.

Assuming for a moment that the differences she observes among writers are real and significant, the effects may have been caused by nontechnological factors of which Halio takes no account. Among the alternative explanations, we need to consider her students' writing abilities before they began their college courses, the kinds of writing and computing instruction they received, and, even more broadly, the general culture surrounding the technology students encountered. It is a gross simplification to blame weak writing skills on the "cuteness" of the Apple Macintosh or the "immaturity" of its users and to ignore the social contexts of both writing and computer use.

Writing research, especially when it is carried out in the classroom, is always beset by complexities. Any critique of Halio's assertions, then, should be balanced by an understanding of the general difficulties of research in this field. Many quite valuable studies have significant limitations because of difficulties with research design (Hawisher, 1989).

Based on the research to date, our understanding of student writing is incomplete. For example, we know very little about the effects on writers of gender, socio-economic status, prior education, or other cultural factors. Although anecdotal or even disputable observations sometimes provide important partial perspectives, we do not yet know enough about writing and technology to draw sweeping conclusions. We need to concentrate instead on framing better questions and on defining forms of inquiry appropriate to them.

Seymour Papert (1987) has warned against strictly "technocentric" studies, like Halio's, in which the researcher concentrates exclusively on the relationship between an isolated user and an isolated machine. We may find an antidote to technocentrism in studies more fully "informed by the interaction of technology with the culture in which it exists" (Hawisher, 1989, p. 64). Such studies require us to ask questions other than whether using a particular machine can "maim" a student's writing.

In order to understand why Halio's conclusions are untenable, we need to examine in some detail the weaknesses in her observations and inferences. There are, as we see it, at least five major problems with her methods.

1. Establishing Comparable Test Populations. To yield sound data (even in descriptive work), a comparative study must first establish that the test populations have comparable levels of ability before they begin the test conditions. This is especially important when subjects are self-selecting rather than randomly assigned to the treatments (Slavin, 1984). Halio claims that "all students in the computer sections have roughly comparable levels of writing ability" (p. 17), but her methods are inadequate to establish even a "rough" baseline.
   

   Because Halio's test groups are regular classes, their comparability depends on her university's placement system, which assigns students to a "medium writing-ability range" according to verbal SAT scores and a writing sample. Halio does not indicate what span of SAT scores is involved, so we have no way of knowing the magnitude of variations--50, 100, 200 points? For the assessment of writing samples, Halio provides no controls for rater reliability. The SAT scores and writing samples cannot establish comparability because the middle group, neither honors nor remedial students, is designed to include the large majority of the university's first-year students.

   In his guide to research methods in education, Slavin (1984) explains that "so much of the variance [in any study] . . . is explained by student ability or past achievement that treatment effects are almost always small in relation to student-to student differences" (pp. 28-29). With no reliable ways to establish students' abilities at the beginning of the term, Halio can make no valid claims based on differences she sees later.

   In more controlled experiments that compare two educational technologies, either the experimental groups are assembled randomly or a single group is introduced to both technologies. This approach is necessary if the study is to eliminate the influence of social or personal differences--like the contrast between the "childish" attitudes of the Macintosh users and the businesslike orientation of the IBM users which Halio posits. Students in Halio's study chose the machines they wished to work with. Any differences in writing may therefore reflect differences in their attitudes and backgrounds. Halio's reliance on self-selecting groups makes attributing writers' performances primarily to their writing tools virtually impossible.
   

2. Determining Comparable Instruction and Practice. Differences in teaching--from organization of the syllabus to the conduct and content of class discussion--can exercise a crucial influence on the writing students produce. In studies where teachers act as their own controls, comparability is usually confirmed by classroom observations (Slavin, 1984). These accounts, vital to case studies, ethnographies, and other forms of descriptive research, portray teachers' attitudes and biases, constructing a picture of the classroom culture. Halio reports no such information.

   Similar observations are needed to detail students' practices. For example, did students typically compose on paper and then type on a computer or did they tend to compose at the keyboard? Halio claims that because the labs were open the same number of hours, students in both conditions had equal access to machines. But if one type of computer is generally the preferred tool on a particular campus, students may have more trouble getting enough time on that machine. Students working with the computer less in demand might thus have longer and more frequent work sessions. Any of these differences might be confounding factors.
   

3. Determining Differences in Performance. Halio uses a number of objective measures to suggest that writers who use the Macintosh do not write as well as IBM PC users. For a number of reasons, Halio's statistics are incomplete and inadequate foundations for her claims. To begin with, the results she presents have not been subjected to any test of statistical significance, such as the Analysis of Variance (ANOVA), which accounts for the strength of chance in variations.

   In addition, the sample size (ten essays from each test group) may be too small to be representative. Halio provides no details on the size and breakdown of the population she observed, but if there were 100 students using each type of computer and if each wrote five essays during the term (a total production of 500 samples for each group) then the number of essays evaluated (ten) would represent only two percent of the writing. A sample that small might be representative, but only if it were carefully controlled for extraneous variations. Were the essays Halio assessed all selected from the same point in the term? Were they written in response to the same assignment? Were the ten in each group selected proportionally from all the test classes two from each class in each condition--to control for teacher effects? Without controls on these influences, we have no way of reliably attributing weaker writing to the use of a particular computer.

4. Assessing Performances. Halio relies on the Kincaid readability scale incorporated in the WRITER'S WORKBENCH analysis program to evaluate the importance of the variations in writing she finds. The scores suggest that the Macintosh samples were readable at just below an eighth grade level and the IBM samples slightly above the twelfth grade level. Readability tests rely on quantifiable stylistic measures, such as sentence length and clause type. Halio associates longer and more complex sentences with better writing, but her interpretation may be faulty. In an analysis of topical structure in writing, Witte (1983) finds texts with less complexity and shorter sentences easier to read and understand . Distrusting quantitative methods, many composition specialists prefer evaluations by human raters who can take into account content, purpose, audience, and other features of the rhetorical situation.

   Even if we accept these stylistic analyses as meaningful, they do not imply, as Halio seems to assume, that Macintosh users were less mature than IBM users. Readability scores do not indicate the level at which a writer reads but the level at which (in the prose being analyzed) he or she has written. A good writer's choice of level is contingent on subject matter, audience, and purpose. Though it takes some effort, even a professor of English can produce copy readable at an eighth-grade level. Readers might not judge this prose weaker than more complex, erudite writing.
   

5. Judging Subject Matter. Halio complains that Macintosh writers chose to work not on serious subjects like capital punishment, teenage pregnancy, nuclear war, and drunk driving but on frivolous topics like graffiti and junk food. Her value judgments here are questionable: many teachers believe that students can write seriously and originally about popular culture, particularly if they are encouraged to think critically. Even more troubling is Halio's suggestion that these choices of topic were influenced by the computer students used, the "toy-like" Macintosh encouraging students to think of writing as "playtime," thus leading them to choose inferior occasions for argument. This is all a matter of interpretation unsupported by any relevant observations or data. The correlation Halio claims to find between topics and technology might be more related to student attitudes than computer systems. But in any case, to investigate such a correlation Halio would need to conduct a different kind of research altogether, one that measures and analyzes attitudes.
   

Problems in research design like the ones we have been discussing often result from an approach that lacks a clear vision of causal connections. Research of all sorts begins by noticing something in the course of events that seems unusual or unfamiliar. The observer seeks to verify that the events triggering the inquiry actually exist in a significant way (in other words do not simply reflect the observer's attitudes or biases), but a researcher cannot proceed without a coherent account of how the events in question might be explained. That account, whether explicit or implicit, guides the study's methodology and provides the theoretical framework for constructing more controlled research.

Even if we suppose that Halio could demonstrate that students in the IBM sections wrote better essays than students in the Macintosh sections, she has failed to provide a plausible explanatory narrative, one that might serve to inform further research. Suppose then that some well-constructed experiment were to find that writers of equal ability before using a particular computer system became writers of unequal ability after they had worked with different systems for a period of time. To what factors and influences might we attribute these effects and how might such hypotheses be tested?

For example, a researcher might hypothesize that screen size affects writers' work. What elements of writing might be affected? Paragraph length, we might suppose, could be influenced by the amount of text a screen can display because the aspect ratio of computer screens is so different from that of paper. Before testing that proposition, though, the investigator should probably consider whether screen size also interacts with software features; for example, whether the word-processing program allows the user to scroll easily or restricts movement within the document to discrete screens (as in page-up and page-down navigation). A combination of factors, rather than an isolated feature, is likely to affect a writer's work (Carroll & Kellogg, 1989).

To study the effects of screen size alone, the researcher should use an experimental design. Such a study would test two different screens (preferably using the same word-processing software to minimize confounding factors) on the same group of writers, measuring whatever aspect of writers' performance the study was designed to examine. In a study of this type, Haas and Hayes (1986) compared the effects of large bit-mapped displays, small IBM PC displays, and pen and paper on writers' difficulty working with text they had already produced. The study found that the large bit-mapped display and pen-and-paper conditions were essentially equivalent while writers had more difficulty with the small CRT displays. It is not clear from this study whether the bit-mapped display alone (holding screen size constant) would offer an advantage over the standard CRT.

Similarly, what theory would connect a command-driven system with mechanical and grammatical correctness, and how would such a theory be tested? The researcher might hypothesize that a command system demands syntactical accuracy and that there is a transfer effect. In other words, writers would learn that "computer readers" (operating systems or word-processing programs) are inflexible and might come to believe that human readers are similarly exacting. Writers would therefore be more attentive to precise syntax in English. The role of feedback--immediate machine response to error as opposed to delayed human-reader response--might be crucial here, and the experimental design would have to take into account those factors as well as the substantial differences between artificial command structures and natural languages. The experiment would have to test writers working with command-driven editors, not with programs whose interfaces employ menus or function keys.

Perhaps, as Halio implies, easy manipulation of visual features--everything from fonts and typefaces to full illustrations--affects students' writing on the Macintosh or in other graphic computing environments. The controlling hypothesis might posit that, whether writing on a Macintosh or an IBM, students spend roughly the same amount of time preparing their assignments but that those working on a Macintosh devote less time to text production or revision because they spend more time with visual features. To establish this point, a descriptive approach would be appropriate. The study would track students' activities and measure the amounts of time writers give to producing words, formatting text, illustrating, and completing the whole task. Any report of this type would certainly have to indicate whether the students composed at the computer or merely typed in already drafted texts before they began to "play" with graphics.

Finally, theories about perceptions and attitudes require careful attention to the entire cultural context (Kling, 1980). As part of this context, the hands-on computer training and the documentation students receive may have significant effects on their attitudes and performance. A descriptive approach would be an appropriate way to study cultural context if it includes sufficient and relevant detail not only about computing instruction, but also about the more general computing environment at the institution. It is insufficient to note, as Halio does, that students in both test conditions "receive equal amounts of training" (p. 17). This says nothing about what aspects of the operating systems and available applications are addressed in the training. If Macintosh instruction includes help with fonts, typefaces, imported graphics and the like, while IBM instruction does not, we might expect that students using the one machine would be more likely to try out such features than students using the other.

Similarly, a user's perceptions of a machine's seriousness and a user's choices as a writer (selecting one topic or target audience rather than another, for example) might be correlated by qualitative methods, including attitudinal surveys and analyses of the physical and instructional settings within which writers work. It would be essential in this research to determine the origins of important features in the computing environment. Here Halio's work reveals important misconceptions. Halio's assertion that students regard the Macintosh as a "toy" rests largely on her assertion that these students have "nicknamed the printers . . . Happy, Doc, Dopey, Grumpy and Bashful" (p. 18). But this observation is erroneous. The names given to network printers are not supplied by users but by operators of the network. They are not terms of endearment but practical designations that allow the software running the network to differentiate among output devices.

Manager's attitudes toward machines might conceivably influence students' perceptions of their work environment. If supervisors of consultants complain about the "dumbing down" of the Macintosh interface, some inexperienced users might indeed take the machine less seriously. Likewise, if managers denigrate MS-DOS or UNIX as arcane and authoritarian, users new to these systems could be quick to find them frustrating and constricting.

It might be very interesting to study such social phenomena, but because approaches to writing technology differ from campus to campus, a good study would have to address more than one institutional setting. The kind of names network managers give devices, for instance, varies considerably from place to place. At Cornell and Yale, the printers in Macintosh facilities have such prosaic names a "Printer A" and "ImageWriter 6" while at Carnegie Melon University all printers, regardless of which computer system that they are attached to, have colorful monikers like "maple," "birch," and "dangermouse." To reveal general correlation, any investigation of these issues would have to be broadly based.

Some of the tentative narratives we have suggested lend themselves to quantitative and experimental research; others are more appropriately pursued with qualitative and descriptive methods. In the absence of a coherent explanatory narrative, however, it is impossible to construct meaningful studies of either type. Even if we were to construct such studies, the results would not necessarily be helpful in guiding technological choices for a writing curriculum. All the approaches we have offered are based on two unstated premises: that we all agree on what writing is and that our understanding of that activity is not subject to change. As long as these premises remain unchallenged, our perspectives are likely to remain largely technocentric. They may not help us construct the fuller account of writing, technology, and educational goals that we must have if we are to make wise choices.

Halio's article--along with the considerable debate it has engendered--gives clear evidence that the cultural context of writing is in flux. Halio speculates that elements of graphic communication intrude on the traditional, exclusively typographic domain of academic writing. When her students use multiple fonts to draw attention to the visual presence of their words or add pictures to their essays, she worries that this "gilding" of the text constitutes a frivolous distraction. These activities seemingly divert students from the normal discourse of the writing class, which for Halio consists of formalized arguments on a predefined range of serious issues. Halio implies that graphic interfaces threaten the stability of this discourse, and by extension, the ethos of academic high seriousness. Students who regard writing as "play-time" are apt to write about dating and fast food, not abortion and nuclear disarmament.

But many teachers would reject the "typically authoritarian" (p. 17) approach to writing that Halio defends. Kenneth Bruffee (1984) has suggested that pedagogy works best when it involves students in "the conversation of mankind." To initiate this engagement, Bruffee notes, teachers must allow for what the philosopher Richard Rorty calls "abnormal discourse"--experimentation, subjectivity, and free verbal play. It might thus be more desirable to have students write insightfully about hip-hop music than to rehearse predigested arguments about Roe vs. Wade.

Likewise, other rhetorics besides the strictly verbal may be appropriate and powerful even in relation to highly serious subjects--witness the images of fetal distress or coat hangers employed in the abortion debate. To restrict argument to a narrow range of topics or to words alone may be defining composition far too narrowly, especially in a world where information takes complex and sophisticated forms.

Richard Lanham (1989) foresees a "digital revolution" in English studies, a period in which questions about the impact of electronic technology will become increasingly urgent for teachers of literature and writing. In fact, this "revolution" may involve as much recovery as innovation. Electronic composition may restore and expand an expressiveness that typographic technologies have limited . Lanham observes that bit-mapped page construction and desktop publishing cause us once again to look "at" rather than "through" the printed word.

It is reasonable to be concerned about the changes in writing and rhetoric that this technological shift may bring: print has been a very potent mechanism for conserving and disseminating knowledge (Eisenstein, 1979). But if we insist that rhetoric confine itself to the printed word, we cut ourselves off from the enormously persuasive power of the visual, a power modern technology has greatly amplified.

Decorating or "gilding" a text does not necessarily impair meaning. "Gilt" illustrations (literally, drawings embellished with gold) are often of great value, materially as well as aesthetically. From a puritan or ascetic perspective, this kind of decoration may seem abominable--but there are other approaches to writing besides the puritan. If we take gilding in its metaphoric sense, referring to any sort of textual illumination, it suggests that decorations infuse the text with a "light" that is as much discursive as aesthetic. We should remember that "graphic" comes from a Greek root meaning to inscribe or write. As W. J. T. Mitchell (1986) argues, there is no inherent difference between depiction and description. The two have always been productively combined, and the evolution of electronic technology may enable even more valuable combinations.

This potential is evident in learning tools now becoming available for both Apple and IBM systems. These include THE RIGHT TO DIE? THE CASE OF DAX COWART and THE SAFE AFFAIR: AN INTERACTIVE HEBREW LESSON, both running on IBM computers with INFOWlNDOWS, and PROJECT PERSEUS and THE SHAKESPEARE PROJECT, available for the Macintosh. Some of these tools, like IN THE HOLY LAND from ABC News Interactive, allow users to edit, annotate, and recombine material from the multimedia database on the videodisc. Using such features in conjunction with presentation management software, students in rhetoric courses could create multimedia essays that would combine their own written commentary with graphics, sound, and video.

This type of composition would situate language spoken, written, and iconographic--in a much richer context than the typed or word-processed essay can provide. A course in multimedia composition would still concern itself with argumentative discourse and the arts of persuasion, the native province of rhetoric from classical times to the present. Teachers of such courses would still teach writing, with all the mechanical and grammatical conventions that are essential to it. But writing and textual presentation historically have encompassed more than the printed word, and contemporary technology affords an even wider scope for expression.

All of this may seem beside the point. Halio asks if our profession wants "to change the traditional goals of freshman comp, which have always been to help students express themselves in words" (p. 19). She cautions that "care must be taken as we include the technology in our programs to define clearly (or redefine) what we consider to be acceptable expository prose to be [sic]" (p. 19). In Halio's view, expanding rhetoric to include multimedia composition would distort the function of college writing courses.

Traditionally, the English composition curriculum has trained students to write expository prose because business and academia required that skill. There is growing evidence, however, that educated people in the 21st century will require a wider range of skills. A recent economic forecast by Venture Economics, Inc. ("Hyper is Hip," 1989, p. 6), predicts that hypertext on-line documentation will experience a 128% annual revenue growth rate, and that hypertext and hypermedia applications will grow at an annual rate of 113% in the years ahead. These figures suggest a new need for competent writers and readers of such materials. Literacy in the next century may well mean the ability to compose in multiple discursive dimensions and across media.

If we were to act on Halio's cautions against systems capable of fully integrating print with other media, we would put the computer to work in writing courses as little more than a rather expensive typewriter. Because, as Halio notes, all major computer vendors are moving to implement graphic interfaces, the logical extension of her position would have writing programs prefer first-generation personal computers (like the early Apple II or the original IBM PC) over the richer computing environments now available from all vendors.

This preference might well appeal to some administrators and managers of technology resources. After all, writing programs typically serve all an institution's students, making these courses a notably expensive enterprise. If only because of the scale involved, it would no doubt be cheaper to supply the writing program with outdated or low-grade technology.

But this is not necessarily a rational policy even in the short run. Today's entering students will become "knowledge workers" who must contribute to a viable economy in just four years. If these students are notable to develop the skills that economy will require, they will not be ready to play their part. The institution that fails to educate them properly may find itself at a distinct competitive disadvantage. More important, if we English teachers are unwilling to expand our notions of writing, we relegate ourselves to the study of the past and the instruments of the past.

Technological change calls for ways of seeing that are broad, not narrow. As Lanham (1989) points out, technology often presents an opportunity to redefine an enterprise--sometimes with substantial penalties for rejecting that opportunity: "Newspapers had to decide whether they were in the information business or only the newspaper business; most who chose the newspaper business are no longer in it" (p. 270). This logic may apply equally to the knowledge business. Moreover, if we fail to take the opportunity to redefine writing, we abdicate our responsibility to shape emerging technologies of communication.

As applications of computer technology rapidly develop, higher education faces profoundly serious questions. How should we define writing, composition, and rhetoric as college subjects? What role should technology play in the teaching of these subjects? How can we describe and assess the impact of technologies on learning in these fields? These questions demand broadly contextual and well-designed research efforts from which we draw no premature conclusions and in which we carefully separate questions of value from matters of fact.

While constructing our research agenda along these prudent lines, however, we must also acknowledge that all research is historically and culturally situated. No inquiry is or ever can be neutral. The view of writing we have presented in this article is certainly open to challenge. But any such challenge, we maintain, should address not whether a machine can maim a message, but how a technological culture envisions and re-visions the multitude of messages it produces. As John Berger says of seeing in general: "We only see what we look at. To look is an act of choice. . . . We never look at just one thing; we are always looking at the relation between things and ourselves" (p. 9).

Nancy Kaplan is Director of the writing center at Cornell University: Stuart Moulthrop, previously of Yale University, now teaches English at the University of Texas--Austin.

Berger, J. (1977). Ways of seeing. New York: Penguin Books.

Bruffee, K. A. (1984). Collaborative learning and the "conversation of mankind." College English, 46, 635-652.

Carroll, J. M., & Kellogg, W. A. (1989). Artifact as theory-nexus: Hermeneutics meets theory-based design. In K. Bice & C. Lewis (Eds.), CHI'89 Conference Proceedings. Austin, TX: ACM. (pp. 41-60).

Eisenstein, E. (1979). The printing press as an agent of change. Cambridge, England: Cambridge UP.

Haas, C., & Hayes, J. R. (1986). What did I just say? Reading problems in writing with the machine. Research in the Teaching of English, 20, 22-35.

Halio, M. P. (1990, January). Student writing: Can the machine maim the message? Academic Computing, 4, 16-19, 45.

Hawisher, G. E. (1989). Research and recommendations for computers and composition. In G. E. Hawisher & C. L. Selfe (Eds.), Critical perspectives on computers and composition instruction (pp. 44-69). New York: Teachers College Press.

Hyper is hip. (1989, August 14-27). Mass High Tech, 7(21), p. 6.

Kling, R. (1980). Social analyses of computing: Theoretical perspectives in recent empirical research. Computing Surveys, 12(1), 61-110.

Lanham, R. (1989). The electronic word: Literary study and the digital revolution. New Literary History, 20(2), 265-290.

Mitchell, W. J. T. (1986). Iconology: Image, text, ideology. Chicago: The University of Chicago Press.

Papert, S. (1987). Computer criticism vs. technocratic thinking. Educational Researcher, 16, 22-30.

Slavin, R. E. (1984). Research methods in education: A practical guide. Englewood, New Jersey: Prentice-Hall, Inc.

Witte, (1983). Topical structure and revision: An exploratory study. College Composition and Communication, 34, pp. 313-341.