Michael Dukakis' difficulties derive from the fact that he is saying little and is saying it gratingly. Suddenly Democrats are looking at him squiggle-eyed, wondering what they have hitched their wagon to.

What worries Democrats is how quickly Dukakis' lead was lost, and how easily.Among Southern whites, Bush's lead has gone from 14 points in July to 31 in early September. If that holds, Dukakis will get few Southern electoral votes. As long as it holds, Bush will not need to spend much time or money in the South-time and money being two great and scarce resources of any candidate.

Among Catholics, independents and white-collar and sales people, Dukakis' leads have become Bush leads. In California, Dukakis' lead has shrunk to three points. That means Dukakis is behind, because on election day in California Republicans usually run four to six points better than the last poll.

Republicans there are better at getting out their voters, including absentee voters. New Jersey is another state which, like California, has gone Republican ever since the 1964 anti-Goldwater landslide, and Dukakis is behind there by 10 points.

Rarely if ever has anyone done as much with as little as Bush did in August. He gave a good speech at the convention, but that is the least you should expect from a man in his third decade in public life, and who had weeks to rehearse what was written for him by an accomplished writer.

Then Bush beat Dukakis about the head and shoulders with the Pledge of Allegiance and furloughs for killers. Then Bush pledged devotion to the future purity of Boston harbor, and -poof! - Dukakis' 17-point lead was gone.

The bounce Dukakis got from his convention was remarkably short-lived. Why? In the last half of the primary season Dukakis was, in the eyes of many Democratic voters, less Dukakis than Mr. Not Jesse Jackson. Furthermore, Dukakis' manner is not wearing well.

The staccato cadences of his clipped sentences suggest irritation. He seems impatient to rush to the end of whatever he is saying because he considers the subject too obvious to detain any clever person. His tone is at once annoyed and complacent, that of a self-satisfied scold.

Since the Republican convention the campaign has resembled (in the words of a Democratic consultant) a tennis match in which only one side gets to serve.

Now we will see if Dukakis can break Bush's service. Dukakis cannot do so by harping on Iran-Contra or Noriega. Any votes that will be won on those issues have long since been won. And if the election comes down to a question of foreign policy, Bush will win.

Assuming Dukakis can change the subject to domestic issues, then the race will be close, at least in terms of popular votes. The last two elections between non-incumbents, 1960 and 1968, were close. Elections following landslides (e.g., 1960, 1968, 1976) usually are.

However, the Electoral College probably will function as it should, exaggerating the decisiveness of the national decision. Such constructive distortion is the work of the state-by-state winner-take-all custom.

In 1960, Kennedy beat Nixon in the popular vote 49.72 percent to 49.55, but 303 to 219 in electoral votes. In 1968, Nixon beat Humphrey 43.42 to 42.92, but won the electoral vote 301 to 191 (with George Wallace winning some). In 1976, Carter beat Ford 50.1 to 48, but won the electoral vote 297 to 240.

Dukakis will be buried beneath an electoral-vote landslide unless he puts a populist and nationalist edge of passion on his economic message.

If he does, he will be accused (he already has been, a bit) of borrowing the themes of a former rival, Richard Gephardt, who would have been harder for Bush to handle.

But that accusation is more bearable than defeat, and borrowed passions are better than Dukakis' sole passion so far, his admiration of his own competence.