CONSUMER GUIDE TO POLLS Suggestions on how to understand the avalanche of election poll numbers. Determine which polls are from reliable firms with a track record and try to follow the results of those polling firms from one survey to the next. Figure out which election polls are of all adults, which ones are of registered voters and which ones are of likely voters. A candidate's support from all Americans is not as meaningful right before the election as how likely voters view that candidate. Do not automatically view changes in results from one company's poll to another company's poll as significant. Those changes may or may not be significant, depending on the way the two polls were conducted. Find out when the poll was conducted, and look for polls that are the most recent. Public opinion can change quickly because of events or gradually as people start paying closer attention to the campaigns. Watch the results of as many reliable polls as available and try to determine if there is a consensus. "If you see five polls that all say a candidate's ahead, the odds are he's ahead," veteran pollster Evans Witt said. Understand that polls from an early primary state like New Hampshire or an early caucus state like Iowa have far more significance than national polls at this stage of the campaign. People in New Hampshire are actually paying attention to the candidates, while people nationally may be less concerned at this point. Know that what happens in Iowa and New Hampshire inevitably has a considerable effect on polls nationally and in other states. Watch the overall strength of each candidate from poll to poll rather than focusing too closely on a close point-margin between two candidates. Understand that small changes in a candidate's poll numbers that are within the poll's error margin are not significant and may or may not indicate a change.

WASHINGTON -- One New Hampshire poll might show Al Gore ahead, a second might have Bill Bradley in the lead and a third might indicate a Democratic presidential race too close to call.

The same thing has been happening with Republicans George W. Bush and John McCain.

What's going on?

The two races in the state with the nation's first presidential primary are very close, for one thing.

Also, polls have an error margin, which means most of those polls generally are accurate within a specified range. Pollsters use different techniques of deciding who to include in their poll. Also, people still are learning about the candidates at this point in a campaign and sometimes are quick to change their mind.

"Bradley and McCain are not that well known," said Evans Witt, president of Princeton Survey Research Associates. "You're asking opinions about someone they don't know well. The opinions may be driven by last night's debate, a story they read in the newspaper, a grimacing photo on a magazine cover."

Sometimes the polls may vary because the pollsters are measuring different things -- something not clear from just seeing a poll result. Two New Hampshire polls this week offered an excellent example.

The American Research Group has a tracking poll, a rolling average of three one-night polls, that keeps an eye on how the candidates are doing leading up to the Feb. 1 primary. In that poll, both party's races are very close, separated by less than the statistical error margin.

A poll released Thursday by the Quinnipiac Polling Institute showed Bradley leading Gore by 10 points, 47 percent to 37 percent, and McCain leading Bush by 9 points, 37 percent to 28 percent. Both leads were larger than the poll's error margin of plus or minus 5 percentage points for Democrats and 4 percentage points for Republicans.

ARG pollster Dick Bennett said his tracking poll tries to reflect the opinions of those almost certain to vote, a process that screens out some independents. Quinnipiac pollster Maurice Carroll said his poll asks people whether they are going to vote, but does not question them as closely about who is certain to vote until much closer to the primary.

The approaches are different; neither is right or wrong.

Carroll acknowledges New Hampshire can be a tough place to poll.

"New Hampshire is dreadfully complicated," said Carroll, noting that more than a third of the registered voters are independent and they can vote in either primary. A third of those who voted in the 1996 Republican primary identified themselves as independent, according to exit polls.

Both Bradley and McCain fare well among independent voters, helping explain why they were stronger in the Quinnipiac poll. Carroll said when he narrows his own poll to just Democrats and Republicans, both races are very close. Bennett said McCain and Bradley fare better in his poll if he does not exclude as many questionable voters.

The job for McCain and Bradley is to make sure those people turn out and vote, Bennett said.

In Iowa, the task is even tougher than in New Hampshire because it is extremely difficult to predict who will attend the Jan. 24 caucuses.

The polls in Iowa do not measure a candidate's organization, which can be crucial in turning out voters for the lengthy caucuses compared with a simpler primary.

The greatest problem in polling in a caucus state is reaching those likely to participate, who are such a small percentage of the party members. On the Republican side, polls have tended to overstate the strength of the mainstream candidates and understated Christian and social conservatives.

"The reason that everyone doesn't poll Iowa is the difficulty of carrying out a valid poll in a caucus state," said Hugh Winebrenner, a professor of public administration at Drake University in Des Moines, Iowa.

Polling is much more manageable for primaries and general elections, where years of experience have led to accurate ways of predicting who will vote.