A poll finding that more than one-fifth of black voters -- 22 percent -- cast ballots for Republican House candidates in last month's elections has been questioned by political scientists, polling experts and party officials who doubt that the black Republican vote was that high.

The finding emerged from a nationwide exit poll conducted on Election Day by Voter Research and Surveys (VRS), a consortium sponsored by the television networks.

The figure has been widely cited in newspaper and television stories and by some Republicans -- including Robert Teeter, President Bush's pollster -- who argue that their party is making inroads on one of the nation's most solidly Democratic groups.

The Republican National Committee hailed the 22 percent figure as proof of a trend, along with the election of Gary Franks, whose victory in an overwhelmingly white Connecticut district made him the first black Republican to win a House seat since 1932.

But casting doubt on the nationwide finding about congressional races is the fact that separate exit polls in contests for governor and the Senate found a much lower proportion of blacks voting for Republican candidates.

In addition, the sample of black voters included a much higher proportion of suburbanites than a comparable 1986 study. This could mean the sample inadvertently overrepresented suburban black voters, who tend to be more Republican than black voters elsewhere. But VRS officials said it might reflect a shift in black turnout patterns.

The 22 percent figure was striking because if accurate, it would mark a significant improvement in Republican fortunes among black voters. In 1986, exit polls showed GOP House candidates winning 9 to 13 percent of the black vote.

At stake here is not an arcane matter for number-crunchers but a finding that has begun to play a role in the political debate. "It's very consequential, whether it's true or not," said Geoffrey Garin, a Democratic pollster who noted that Republicans have already begun citing the exit poll finding as a sign that "it didn't mean anything that George Bush vetoed the Civil Rights Bill."

"This is not abstract political science," Garin said. "This is real politics."

Warren J. Mitofsky, president of VRS, said he believed Republicans did make some inroads among blacks in congressional contests, but that too much has been made of the specific 22 percent figure, which was based on interviews with 514 black voters nationwide.

He noted that the figure was subject to a margin of sampling error of plus or minus 6 percentage points.

"I think there's some change, but I think it's marginal," said Mitofky. "It's something I want to work on more, absolutely. I'm not done with this problem."

Mitofsky has a reputation as a careful statistician, and pollsters are well-schooled in the perils of taking single numbers as gospel. But Garin noted that specific polling numbers, if cited often enough, develop a life of their own and come to be regarded as "a matter of fact."

"I'm starting to see that 22 percent number cited without any attribution to the fact that it comes from an exit poll," he said.

The strongest doubts about the figure come from separate exit polls on statewide contests for U.S. Senate and governor's seats, also conducted by VRS. The congressional figure is based on a nationwide poll.

For example, in 16 governors' races and nine Senate contests in states where the black vote was large enough to be measured by exit polls in a statistically significant way, the Republican share of the black vote averaged only 14 percent. And the black Republican vote seemed especially low in large states and southern states, which have the largest proportion of black voters.

These figures were much closer to the norm of past elections. Voting analysts said it would be surprising for blacks to pick Democrats at the top of the ticket, then shift to the Republicans in such large numbers for congressional contests.

"I can't expect the vote for Congress to be any better than the vote in Senate or governor's races," said David Hansen, director of survey research for the political division of the Republican National Committee. "It's definitely something that can't be taken at face value."

Richard Wirthlin, a Republican who was President Ronald Reagan's pollster, said it would be "unusual, indeed, to find a swing of 7 or 8 points from the top of the ticket to Republicans at the bottom of the ticket in the black constituency."

Academic specialists in black voting were also surprised by the 22 percent figure.

"If I were to guess, I would say the number is somewhere between 12.5 and 15 percent," said David Bositis, a senior research associate at the Joint Center for Political Studies, a think tank in Washington that studies issues of concern to blacks. "Our analysis of voting in 1990 shows absolutely no movement of black voters to the Republican Party. It's basically a pipe dream on the part of the Republicans."

Michael Dawson, a political science professor at the University of Michigan and an authority on black voting patterns, took a similar view. "The studies we've done all show stable and high Democratic identification, high support for Democratic candidates at the national level and low support for Republicans," he said.

But Andrew Kohut, who directs the Times-Mirror Co. series of surveys on national attitudes, said his pre-election study this year showed substantial numbers of blacks drifting away from the Democratic Party and toward independent status.

In coming up with its estimate, VRS faced a thicket of methodological and technical problems.

Wirthlin noted that exit polls have large margins of error for "smaller sub-groups such as blacks." Garin noted that in GOP Sen. Jesse Helms's successful 1984 reelection campaign in North Carolina, the three network exit polls came up with estimates of Helms's share of the black vote ranging from the low single digits to about 14 percent.

There also have been different estimates of the black congressional vote in the past. In 1986, for example, ABC News estimated the Republican share of the black congressional vote at 9 percent, while a CBS News-New York Times poll put it at 13 percent.

One difficulty this year, Garin said, is that when the networks merged their exit polling efforts, there was no way to check the findings against comparable standard.

Sampling the black vote presents problems because of patterns of residential segregation. Less-affluent blacks, who are also the most Democratic, tend to be concentrated in a limited number of inner-city neighborhoods.

In picking a random sample of precincts to poll, Withlin said, it would therefore be possible to under-represent inner-city areas. The results would then show a smaller black vote overall and a sample of black voters that was disproportionately suburban -- and more inclined to vote Republican.

In addition, data initially released by VRS showed that the percentage of the black vote cast by suburbanites increased dramatically over the last four years, from 20 to 31 percent. The results compared the 1990 VRS poll to a 1986 CBS News-New York Times Poll that used similar methods. But when The Washington Post questioned the size of the change, VRS officials discovered they had made a programming error and said corrected data will be available this week.

It is clear that black voters sampled in 1990 were more affluent than those sampled in 1986. In this year's exit poll, 24 percent of all blacks questioned lived in households with total annual incomes of $50,000 a year or more. But in the 1986 exit poll, just 11 percent of blacks interviewed had that income level.

Another significant difference between the two surveys is an apparent decline in the black share of the overall vote, from 8 percent in the 1986 survey to 5 percent in the 1990 survey.

All this raised the possibility that the 1990 VRS poll over-represented affluent black suburbanites, under-sampled inner city blacks and underestimated overall black turnout. But Mitofsky offered an alternative explanation: Less-educated, less-affluent and overwhelmingly Democratic black voters may have stayed home Nov. 6, while more-affluent and better-educated blacks -- those more likely to be GOP voters -- went to the polls.

"What you had was a very small black turnout," Mitofsky said. "The ones who voted are the people who are most likely to vote: the better-educated and the more-affluent. What you may have been left with are affluent blacks who were more likely to vote Republican. The two go together."

Strict methodologists might raise other concerns about comparing the 1990 poll with earlier CBS-Times results. The 1990 VRS national exit poll had serious data processing problems on election night and the firm was unable to provide immediate results as promised to its network and newspaper clients, including The Washington Post.

In order to get results to those organizations the next day, VRS had to abandon a sophisticated computer program used in earlier CBS-Times exit polls to correct the data for slight sampling irregularities. VRS staffers wrote a simpler program that did not correct for certain kinds of sampling problems sometimes encountered in exit polls.

Mitofsky said he doubted that computer differences had a significant impact on the findings.

Staff writer Juan Williams contributed to this report.