Can computers reverse the decline in students' writing skills?

Some blame television. Others point to a collapse of school discipline. Still others note the disappearance of homework and the replacement of basic English courses by ''soft'' electives.

The problem: a crisis in composition, with student writing skills declining steadily.

Is there a high-tech answer? Can computers teach good writing?

Ask the question, and you can almost hear the hackles rise on the backs of the traditionalists. ''Of course not,'' they assert. They reason that fine writing is a creative act, born in inspiration and nurtured by reading great literature. And they are right - about ''fine'' writing. But what about good journeyman prose - the stock in trade of managers, bureaucrats, and professionals? Can that be taught electronically?

A firm ''Yes'' is the answer out of Fort Collins, Colo. This year more than 3,000 students at Colorado State University (CSU) will use a version of a writing program developed by Bell Laboratories in Piscataway, N.J. It is known as the Writer's Workbench, and has been tested at CSU since 1981, although it came on the market only last June.

Prof. Charles R. Smith of CSU's English Department, who thinks the program is the best thing for composition since the invention of paper, is quick to point out that it can't replace teachers. Instead, he explains, it helps students to ''reconsider'' the mechanics of their papers before they turn them in - so that ''the teacher can concentrate on content, logic, and organization.'' The process begins when a student types a paper into the school's computer and gets back a printout. Among other things, the resulting analysis covers:

* Passive voice. By locating each use of the verb ''to be'' in all its forms , the program identifies potential overuse of passive constructions (''it was decided by the group,'' for example), which are generally weaker than active forms (such as ''the group decided'').

* Diction. By bracketing some 500 frequently misused words and phrases, the program signals verbose, inaccurate, overblown, or sexist language. ''Due to the fact that'' will be marked, with ''because'' suggested in its place. Other terms flagged include ''advent'' (the coming of something momentous) where ''arrival'' would do; ''employ'' as an inflated way of saying ''use''; and ''man-hour,'' for which the computer politely proposes ''work-hour.''

* Vagueness. The printout also flags words that frequently signal imprecision or generality: ''interesting,'' ''fun,'' ''nice,'' and so forth. It even provides a ''vagueness index'' - comparing the percentage of vague words with a ''good'' paper's percentage.

Also scrutinized are sentence lengths, the ratio of simple to complex sentences, and spelling - in all, some 17 features of prose. But the program does not rewrite. It simply provides what Bell's Lawrence T. Frase, one of its developers, calls an ''educational advisory system'' - noting possible errors and proposing that the student examine them.

And do students respond? Indeed they do, says Kathleen E. Kiefer, an associate professor at CSU. Surveys indicate that students look forward to using the program - and sign up in droves for the courses offering computerized instruction. Even after the course has ended, they make fewer mechanical and stylistic errors than do non-computer-trained students. Early indications also show that they improve their writing proficiency more than do their low-tech counterparts.

Not surprisingly, the word is beginning to spread. Both Bell Labs and CSU steadily get requests for information. IBM is now preparing a writing program known as Epistle. The US Navy is reportedly working on a system for writers of technical and training materials. Even MIT (which to outsiders might seem more interested in numbers than words) is interested. Why? MIT's David Lampe, editor of The MIT Report and a writing-program instructor, points to a recent study by MIT researchers. It shows that scientists, engineers, and managers in a typical research and development (R&D) group spend one-third of their time writing and editing - and that over half of them have had no more than one writing course.

But it is not only to improve America's technical writing that such programs show promise. Their broad value (and I speak as one who spent a decade teaching college composition) is twofold. First, they lift the drudgery of making mechanical corrections off the instructor, placing them instead on a self-teaching system with all the appeal of an individually tailored video game. Result: Student and instructor can focus on content - and on the structuring that makes that content irresistibly readable.

Second, these programs are aimed at the vast middle range of students. They don't work well with remedial cases. Nor do they have much to offer the advanced writer. They focus on that group which modern education (with its two-pronged focus on salvaging the poorest and polishing the brightest) has often ignored: the average writer. If America's writing skills are to improve, that is exactly where the educational systems must concentrate.