Some very funny things have been said about car design. I think it was K.T. Keller, head of Chrysler in the early 1950s who said something like: "Americans want a car they can wear their hats in, not piss over." He then brought out a line of cars that were shorter, higher and narrower while everyone else was going longer, lower, wider. That disaster made Chrysler turn to fins.

Then there was George Walker, styling director of Ford when they still called them stylists: He said that "No statue was ever erected to the man or woman who thought it best to leave well enough alone."

And Bill Mitchell, who led the General Motors styling team in their greatest days, told me" "It's hard to do a small car. It's like tailoring for a dwarf."

But my favorite was attributed to Phil Caldwell, the brilliant chairman of Ford during the dark days of the early 1980s. He was betting the company - $3 billion - on a radical design called Taurus when bad news came in. The first focus group in Kansas City had been a disappointment. "Get another focus group," Mr. Caldwell reportedly said. "We're doing the car." The rest is history.

Design theory today at GM seems to go in the opposite direction. GM says it wants to learn all it possibly can about its potential customers and then give them what GM thinks they want. Contrast that with what Bob Lutz of Chrysler said a few months ago in Traverse City: "We've put so much faith in analysis and quantification," he said, "that we've often missed the forest for all the well-examined trees."

He talked about creative spark and inspiration and individual genius. Just listening to the customer is like looking into a rearview mirror, he said. "He can tell you what he likes among the choices that are already out there. But when it comes to the future, why, I ask, should we expect the customer to be the expert in clairvoyance or in creativity.?

"Customer-driven is certainly a good thing. But if you're so customer driven that you're merely following yesterday's trends, then ultimately, customers won't be driving your supposedly customer-driven products." Mr. Lutz says you can get carried away with anything, but we get his point. And his team is gaining market share, while the big guy still is slipping.

This isn't an attack on conservative design. We reporters continually knock Honda's Accord and Toyota's Camry as dull. But they are solid, good-looking cars and the gold standard for quality. Marketing, service, reputation, price are all a part of the package that sells a car. It isn't all good looks. Saturn doesn't win any styling awards, but it's the most wanted car in the GM lineup. We all know why.

And Chrysler's stylish cars actually don't sell that well - it's the Ram and the Jeeps and minivans that are setting records, possibly because the cars don't have the highest reputation for quality (or possibly because the dealers are writing so many orders for Rams, Jeeps and minivans that they aren't pushing cars that hard).

Still, a ranking GM official recently told me that they know their cars need more styling reach, but they feel GM must establish its reputation for quality and reliability first. Maybe. Let's hope they have enough customers left when that day is reached.

There's another reason to stretch the envelope. Designs last longer, six years or so. If a car design is going to last that long it must be advanced at its introduction if it's going to be up to date four years later. The new Taurus has been criticized from here to eternity, but it is quite possible that it will gain strength as it ages. This happened in Europe with a midsize Ford called the Sierra, which was called too radical at first, but gained strength through the years. It lasted almost a decade.

The best example of styling excitement I've seen recently comes from Ford's European Ka, a 120-in. (304-cm) long, 60-hp citicar.

It,s actually a chopped European Fiesta, but while the Fiesta is a box, the Ka is a ball. Same platform, same essential systems, but a new look. It's the best example today of building different cars off a single platform. Total cost: only $280 million, Ford says.

"Chrysler couldn't do that," brags Ford's head of Ford Auto Operations, Jacques Nasser. Of course, building different models off a single platform is an old trick that the industry is relearning. Remember how Lee Iacocca turned the plain Ford Falcon into a Mustang? And one of the strengths of the old GM was its talent for creating different cars from the same body. Somehow that got forgotten in the last 20 years.

We are entering an era of styling experimentation. You see it in main-line cars like the Taurus and the big Chryslers of 1998 and 1999. You see it in the Ka, in the coming Mercedes A Class (a citicar) and M Class 4-wheeler and the Mercedes Swatch Smart (I saw one that was on exhibit in a shopping mall under the Louvre).

Interiors are showing faint signs of glitz again, thank goodness. What I call Teutonic minimalism - the Taurus was a great example - is fading. I remember the old days when red cars had red seats and trim, and blue cars had blue seats and trim. Now they all seem to be mouse grey or mud brown.

Experimentation will probably mean mistakes. The classic styling mistake was the Chrysler Airflow of the 1930s. Then there was the Edsel. "Someone hopped on that front end and called it a toilet seat, and it was dead from that minute," said Ernie Breech, the Ford boss. More recently, there were GM's Dustbuster minivans.

It would be a poor design strategy to try to avoid all mistakes by playing it absolutely safe. The trick is to correct mistakes quickly. After all, it shouldn't take six years to give a van a nose job.