A decade ago, Tracy Kidder's Pulitzer Prize-winning book, "The Soul of a New Machine," glorified Data General Corp. as a highly charged company that could turn out hot new computers faster than anybody else. This was one American company that was fast on its feet.

These days, the story at the Boston-area company is decidedly different. After stumbling in 1985, Data General has been unable even to regain its balance. Now, headed for its fifth straight year of losses, the firm seems poised to make another wrenching attempt at financial recovery by laying off a sizable chunk of its work force.

The company, while not specifically confirming a Wall Street Journal report yesterday that the cutback could affect as many as 2,000 people, or 17 percent of the work force, said layoffs could not be ruled out as part of its ongoing cost-cutting campaign. A reduction of that magnitude would shrink the computer maker's staff by 45 percent from its 1985 peak of 17,700.

Once a highflier whose sales grew for 20 consecutive years until 1989, Data General needs badly to return to profitability and convince buyers it will be around for the next 20 years. In the most recent nine months, the firm spilled $50.5 million in red ink on revenue of $907.9 million. Its stock closed yesterday at $5.62 1/2, compared with its historical peak of $76 in 1985.

"Data General has very little 'mindshare,' " said Marc Schulman, an analyst with UBS Securities in New York. "They could introduce the greatest product since sliced bread and a goodly percentage of the marketplace wouldn't even notice... . The quickest way to build 'mindshare' is to become profitable."

The problems at Data General are further evidence of the continuing slump in Boston's high-technology corridor, a region that a decade ago was adding jobs at double-digit rates and emerged as the centerpiece of the famed economic "Massachusetts Miracle."

Companies that could do no wrong -- firms like Wang Laboratories Inc., Prime Computer Inc., Data General and even powerhouse Digital Equipment Corp. -- all have stumbled badly in recent years. In the past two years, high-tech firms in the state have laid off 60,000 people, though some of the workers have been absorbed again by smaller firms that continue to grow.

Data General made the crucial mistake of failing to realize quickly enough the importance of personal computers and workstations. Like other makers of minicomputers -- machines that today typically serve up to 100 users in a department and sell for less than $1 million -- Data General found itself squeezed from below by a new generation of more powerful desktop computers from such firms as Sun Microsystems Inc. and from above by more efficient mainframes produced by companies like International Business Machines Corp. and Amdahl Corp.

Buyers also streamed to computers that were built with standard parts, shunning machines like those from Data General and Digital that were difficult to mix and match with equipment from other vendors.

Missing these critical technological sea changes was particularly jarring to Data General, which was founded by Digital defector Edson de Castro and prided itself on its entrepreneurial flair. Nothing typified the company more than the portrait painted by author Kidder -- a small team of gung-ho engineers sequestered for 18 months in a windowless basement, then emerging victorious with a new computer.

Ingrained in the company's culture was the drive to produce what industry insiders call a "hot box" -- a computer that provides more power per dollar than anybody else's.

After drifting in the mid-1980s, Data General tried to recapture that trophy last year with a line of so-called "open systems" computers built around certain standard software and computer chips, making them more easily linked to other machines. Data General's annual report, noting that the new machines offer for the same price 170 times the speed as the company's first computer in unveiled in 1969, boasts that it still beats much of the industry in per-dollar power.

While this new product line is meeting sales goals, the company is still trying to figure out how best to sell them, according to analysts. And Data General continues to be disappointed by sales of its more traditional minicomputers, which previously had sold well in specialized markets like hospitals and banks.