AT A SMALL research institute in Cambridge, Mass., an enlightening little fight is playing out that could significantly affect the way Americans see and think about the economy.

The debate involves the National Bureau of Economic Research, a private group that is the accepted judge of recessions and expansions in this country.

NBER officials were the ones who declared last fall that the U.S. economy had been in recession since March 2001. Eventually, the bureau will pick an official ending date for the contraction, but that call is months away, and most analysts think the economy has begun to grow again.

What has gone unreported is an internal debate at the bureau over its business-cycle dating methods and whether what the economy went through last year was really a recession.

The discussion, which had been percolating for months, came to a head a few weeks ago in Washington.

On April 4 at the Brookings Institution, NBER research associate William D. Nordhaus presented a biting paper in which he questioned whether last year's slowdown was a true recession, called the bureau's approach "archaic" and suggested that "the traditional NBER business-cycle methodology has outlived its usefulness."

Taking issue with the bureau's semiofficial status as recession arbiter, he called for new players to identify downturns and suggested that "like other monopolists, the NBER has resisted innovation in the definition and measurement of business cycles."

Nordhaus, who teaches at Yale, advised President Jimmy Carter and co-wrote a widely used college economics text, thinks the NBER worries too much about industrial trends that reflect only small bits of the economy and doesn't pay enough attention to nuanced, modern statistics from the Commerce Department.

He also faults the NBER for its either-or approach to recessions and expansions. His main recommendation: Bring more subtlety to the process by grading slowdowns the way meteorologists measure hurricanes.

Don't just tell the public "yes" or "no" on the recession question, Nordhaus says. If it's a 1932-style blockbuster, give it a Category 5, Hurricane Camille rating. If it's a mild pause in overall growth, as happened last year, maybe call it a stiff gale.

Nordhaus is a research associate at the NBER, not a member of the bureau's Business Cycle Dating Committee, which makes the recession calls. But behind a public facade of consensus, the panel has been subject to considerable doubt and argument over the nature and timing of modern recessions.

I questioned five of the panel's six members, who insisted there was a recession last year but acknowledged that pinpointing its beginning was difficult.

"It was a very intensive discussion" before the committee settled on March 2001 as the peak of the last expansion, said Victor Zarnowitz of the Conference Board. "It was not easy."

Princeton's Ben S. Bernanke, one member told me, was perhaps the panel member most skeptical of the idea that March inaugurated a downturn, although he accepts that a recession did occur. Bernanke awaits congressional confirmation for an appointment to the Federal Reserve Board of Governors and is not commenting publicly.

The problem facing the NBER, or any economist, is that recessions don't behave as they once did. In old-time slumps, the job base and economic output shrank more or less in unison, prompting a rule of thumb that two consecutive quarters of declining output - gross domestic product - make a recession.

Modern downturns have been milder and harder to nail down. Factory automation and office computers sometimes allow output to grow while employment falls. This has happened twice in recent years, flummoxing the NBER, and the bureau has treated the incidents differently.

GDP grew while the job base shrank from March 1991 through the end of that year, but the NBER, focused on output, declared March the start of a new expansion, that the 1990-1991 recession was over. This led to accounts of a "jobless recovery."

The same thing happened in spring 2001 - rising output, falling employment - but this time the NBER focused on jobs and decided the rough patch was part of a recession, not an expansion.

Part of the switch can be ascribed to Zarnowitz, who places great stock in the employment numbers and says he initially wanted to extend the official length of the earlier recession beyond March 1991 but bowed to his colleagues.

Nordhaus has removed the more inflammatory phrases from his Brookings critique, now published on the NBER Web site. But he still suggests that last year's downturn, with only one quarter of falling GDP, might not have been a recession. He tentatively rates it in the same category as brief, nonrecessionary slumps in 1963 and 1967.

Robert Hall of Stanford, chairman of the NBER business-cycle committee, rejects this and the idea that his group should rate recessions like earthquakes.

But Nordhaus' grading plan could bring new finesse to public economic discussion, and if there wasn't really a recession last year, the view out the rear window looks a lot different.