HAVE WE reached the point where the media's use of super-modern polling techniques -- and the public's reaction to them -- are skewing the outcome of campaigns? Take a look at the recent example of New Hampshire.

For seven days this month, the state of New Hampshire was transformed into a hothouse of polling, punditry and politicking, the elements of which have never before commingled quite so explosively in a presidential campaign.

By night the pollsters queried the voters; by day they reported back the the latest reading on their shifts in sentiment, like so many blips on an EKG machine. In the end, this orgy of observation did not keep pace with the object under study -- perhaps, several analysts now speculate, because the act of measurement itself kept changing the object under study.

The size of Vice President Bush's 38-29 percent victory over Sen. Robert Dole in the Feb. 16 primary wasn't forseen by any of the news organizations that had been conducting nightly tracking polls for a week. Measured against the 20-point gap he had enjoyed in New Hampshire polls throughout 1987 and January of 1988, his winning margin was hardly remarkable. It's how he got there that caught nearly everyone by surprise.

In the days immediately before and after his third place "humiliation" (that's how it was widely reported in the media) in the Feb. 8 Iowa caucuses, Bush lost all of that 20-point New Hampshire advantage, according to nightly tracking polls, and actually fell behind Dole. Then, in the last 72 hours before the New Hampshire vote, he recouped much of his lost support.

Tracking polls -- repeating the same set of questions to a small sample on successive days -- monitor shifts in voter sentiment typically in the last days before an election. They've been used by well-financed campaigns for nearly two decades to guide last-minute tactical decisions. This year many media organizations began using them as well.

The various tracking polls in New Hampshire all caught the last-minute shift in momentum, but not its magnitude. Even now, pollsters are mystified that sentiment could have moved so sharply in one direction then doubled back just as sharply in the other -- without the intervention of a major news event. "I've studied data from thousand of elections," said one pollster, who asked not to be identifed. "This is the most unique finish I've ever seen."

What happened? With hindsight, there's no shortage of explanations: the Bush campaign's far superior get-out-the-vote operation (a factor that, experts say, could account for 3 to 5 points of his margin); a well-conceived late attack ad against Dole; an eleventh hour boost from Sen. Barry Goldwater; and an almost plaintive "warts and all" appeal from the candidate himself.

The Dole campaign, on the other hand, started the week flush with the Iowa victory and determined to "act presidential" so as to bury once and for all the "mean Bob Dole" stereotype. It never wavered from that strategy, even in the face of tough television ads depicting Dole as a "straddler" and a tax-raiser. Dole's now says he regrets his week spent above-the-fray, attributing it to overconfidence spawned in part by tracking polls. "I felt pretty good there for about 24 hours," he said, ruefully, this week. "I was writing my inaugural adress."

Beyond the second-guessing over tactics, howver, there is an emerging line of analsysis which argues that the heavy publicity surrounding the Bush "collapse" in the early part of the week, almost in and of itself, created a counter-trend.

Dole's New Hampshire coordinator, Thomas Rath, argues the media, in their eagerness to bury Bush, may have set the stage for his comeback. "The press was focusing on 'the death of George Bush story' -- you almost think they already had the leads written in their typewriters -- and I think that it dramatically altered the stakes of the race," he said. "It created a 'poor George Bush' phenomenon, and allowed him to come back and say, 'Don't throw this good and decent man out.'"

Rath is correct that the media -- emboldened by its tracking polls showing a massive hemorrhaging of Bush support in New Hampshire -- captured and, in retrospect, overdramatized the peril of his situation. The night after Iowa, NBC had him suffering a "humiliating " defeat and quoted a senior Bush aide saying that unless he won New Hampshire, "the race may be all over." On CBS, the Bush campaign was depicted as "desperate". The Washington Post chronicled the Dole surge all week, as did the Boston Globe, the Manchester Union Leader, and virtually all of the smaller papers that New Hampshirites read.

"Just as voters in new Hampshire started thinking about the race," said Henry Brady, a University of Chicago political scientist, "they may very well have started saying to themselves, 'My God, if I vote for Dole I might knock Bush out of the race.'"

"This gets into an area psychologists would call 'framing' the way you think of your choice," said Samuel Popkin, a political scientist at the University of California at San Diego. Popkin drew a contrast between what happened to Bush in New Hampshire this year and what happened to Walter Mondale in the same state in 1984. In January of 1984, a Washington Post/ABC poll showed Mondale with a lopsided 50-4 percent New Hampshire lead over Gary Hart. Then came the Iowa vote. Hart got only 16 percent. But, by coming in second in a crowded field, he established himself as the clear alternative to Mondale, who was an unpopular frontrunner with many Democrats. The next week, Hart zoomed past Mondale in New Hampshire, 37-28 percent.

Why couldn't Mondale reverse the collapse, as Bush did in 1988? One difference is that in 1984, the electorate was not so heavily bombarded with daily poll evidence of its own abandonment of Mondale -- the Washington Post and ABC were the only news organization's doing tracking polls that year. So in 1984, the counterforce had less of a chance to set in.

Popkin is one of many academicians who believe that primary voters often engage in "strategic voting"; that is, they take into account factors other than the simple question of preference. Until this year, however, the prevailing evidence has been that strategic voting creates bandwagons in the early phase of the nomination season, and impedes them only in the late phases. In the early primaries, voters who are attracted to longshots or protest candidates may not vote for them, because they don't think they can win. "The overwhelming evidence is that voters' expectations of who they think is going to win correlates in a positive way with who they vote for," said Larry Bartels, a political scientist at the Unviersity of Rochester who has written a book about the dynamics of primaries. "The only example of a negative correlation comes late in the process, and it tends to be weaker."

The Rath explanation of the "strategic" vote to save Bush would break that pattern, and Bartels is skeptical of it. "It seem unlikely to me that 10,000 or 15,000 voters came to the spontaneous conclusion that Bush was on the verge of being eliminated, and they all reacted the same way. You are talking about telescoping into a couple of days a process that usually unfolds over months."

But Dole's media adviser, Larry McCarthy, thinks that's exactly what happened. "If it hadn't been for the tracking polls, Dole would have won," he said. A tracking poll question asked by Dole's pollster, Richard Wirthlin, showed that on Sunday, Feb. 14, and Monday, Feb. 15, more Republicans -- by a margin of nearly 20 percent -- thought Dole was going to be the winner than Bush. Wirthlin's post-election survey also showed that voters who said they were undecided on Sunday and Monday wound up breaking 7-3 for Bush on election day (although network exit polls show a more even break among late deciders). Lots of people appear to have decided "in the last 12 hours," Wirthlin says, to vote for a candidate they thought was going to lose.

Bush's pollster, Robert Teeter, said he believed that what ultimately saved Bush was not so much better execution in the last 72 hours as his larger base of support going in to the final week. "Our polls showed us running at about a 40 percent support level for a year -- and, of that, 30 percent was hard core support. The Iowa loss stripped us right down to our hard core. But we knew exactly who the 10 percent were, and we were able to go right back after them in a targeted way."

The Dole camp, meantime, was a step behind the curve all week. On the Wednesday after Iowa, it filmed an "innoculation" ad on taxes, in which Dole uttered the phrase "I pledge to veto any tax increase which . . . " But because of production and delivery glitches, the ad never showed, expect for a few hours on one Boston station.

A rule of thumb in the closing days of a campaign is never to let an attack go unchallenged -- and all week Dole had been accused of being a tax-hiker by his opponents. Many in the campaign now concede their failure to respond contributed to the margin of defeat. But all also say that larger forces -- not the least of them their early-week surge and the media reaction to it -- were at play. It's a suspicion that can neither be proved nor disproved -- but one that is sure to crop up again in this age of non-stop polling. Paul Taylor is a Washington Post staff writer. Staff Writers David Hoffman and Edward Walsh contributed to this report.