(FORTUNE Magazine) – ON A TUESDAY AFTERNOON IN LATE OCTOBER, Phil Purcell is in his office standing behind a four-foot-tall desk--an odd-looking piece of furniture installed to relieve his chronically aching back. The chairman and CEO of Morgan Stanley is a gigantic man, 6-foot-5, with the broad shoulders of a former high school football star. Beside his desk, near photos of his wife and seven sons, a Bloomberg terminal is flashing Morgan Stanley's stock price--up a few pennies on the day but down 21% since March. An article on the news service says that the firm is struggling and hints that Purcell will have to sell out to a big commercial bank if he wants Morgan Stanley to remain competitive. The topic strikes a nerve with Purcell, who publicly maintains that he doesn't need a partner but who is rumored to have explored a deal with American Express, Deutsche Bank, Bank One, and J.P. Morgan. He gets the merger question all the time and ordinarily just dismisses it. But today he appears annoyed. His eyes narrow, and he suddenly looks like Gary Cooper staring down a gunfighter in High Noon.

Purcell turns to the ticker and punches in another code. A table flickers up on the screen. It shows the latest rankings of the top Wall Street firms. "No. 1 in global equity underwriting," he says. "That's some failure." He bangs in another code. Up comes a second list. "No. 1 in IPOs. Is that a failure too?" he asks rhetorically. "Morgan Stanley is winning. The business model works."

If Purcell sounds a tad defensive, he has good reason to be. A chorus of critics--analysts and institutional investors--don't think the model works. They see Morgan Stanley as an awkward hybrid--a top-notch investment bank and trading operation saddled with a struggling retail-broker network (the former Dean Witter) that has nearly as many advisors as Merrill Lynch but earns just one-fifth the income; an asset-management business (the Morgan Stanley and Van Kampen mutual funds) that has historically underperformed its peers; and a credit card operation (Discover) that has a mere 5% of the market.

Moreover, this line of thinking goes, Morgan Stan-ley is too small to compete with giants like Citigroup that use their lending power to woo investment-banking clients. Yet it's too big to be as profitable as nimbler, more focused securities shops like Goldman Sachs. So Purcell must either merge Morgan Stanley with a big commercial bank or sell off slow-growing, noncore parts of the company. Morgan is not alone in this; Merrill faces a similar quandary, but one difference is that Merrill is having the best year in its history. Morgan Stanley is not. Profits are 30% below the peak reached in 2000. "They're in a strategic pickle," says analyst Dave Trone at Fox- Pitt Kelton. "Morgan has become Goldman with baggage."

What does Purcell have to say about all this? Well, that's part of the problem. An admitted loner, Purcell is not a popular figure on the Street and is usually not accessible. He spends hours each day in his office contemplating strategy but declines to give anyone except his closest allies a peek inside. Until he agreed to sit down with FORTUNE--participating in more than ten hours of interviews--Purcell had rarely met with analysts, investors, or the media to address questions about the firm's future.

Purcell's failure to articulate a clear vision for Morgan Stanley has become an irritation for some on the Street. "Investors understand that the problems Morgan Stanley has are manageable," says Brad Hintz, a Sanford Bernstein analyst who likes the stock. "They know Discover has issues. They know Dean Witter isn't the best brokerage. Rather, it is a communication issue.Purcell and the management team won't talk about the issues, and that has put an uncertainty discount on the stock." Hintz himself has felt stonewalled. "I've been a ranked analyst for three years," he says. "For ten years I was a partner at Morgan Stanley Group. I have an outperform on the stock, and I have never been allowed to meet with the chairman."

IT'S A PERFECT AUTUMN DAY AND Phil Purcell, 61, is perched on the edge of a chair outside the grill room at the Winged Foot Golf Club. He begins reading news clippings and e-mails his assistant has printed for him, then glares at me. He's just come across an e-mail I sent asking whether the firm's retail brokerage unit, which Purcell ran for years, is still an important player on Wall Street. Reaching into a folder, he pulls out a recent Barron's that ranks the 100 best high-end stockbrokers in the country. "Take a look at this," he says sharply. "Thirteen of our people are on that list. This is proof that you're wrong."

A little later we're standing on the picturesque tenth tee. Purcell has a beer in one hand and a fairway wood in the other. "Don't knock over my beer--that's important," Purcell jokes, placing the plastic cup on the grass. After a practice swing, he launches the ball 183 yards to within 15 feet of the cup. Barely pausing to acknowledge the impressive shot, he turns to answer a pending question about whether Morgan Stanley can stay competitive without a merger. "You're asking if our business makes sense," he says. "I'll tell you this: The model stood up pretty well through a very tough cycle, and I wouldn't want to be CEO of any other franchise."

OF COURSE, PHIL Purcell has defied the conventional wisdom before. In 1996, Morgan Stanley CEO Dick Fisher and John Mack, his heir apparent, began talking seriously with Purcell about a merger with Dean Witter, where he was CEO. Purcell made extensive demands, including that he be CEO--not co-CEO as was becoming customary in mega-mergers. Surprisingly, Fisher and Mack, who would be president of the merged company, agreed. Their acquiescence is often attributed to simple eagerness to get a deal done. But there is another theory. "These guys figured Phil was some hick from Chicago who they'd blow away in six months," says John McCoy, former CEO of Bank One and a Purcell friend.

Purcell was happy to let Mack and other Morgan Stanley hands, like institutional securities head Peter Karches, run the firm day to day. He focused his attention on the directors. After Fisher and other Morgan loyalists retired from the board, Purcell replaced them with outsiders and old associates from the Midwest such as Charles Knight of Emerson Electric. And when Mack began angling for the CEO job, he was rebuffed time and again by the board. He finally quit in early 2001. Says McCoy: "Phil outsmarted the best dealmakers on Wall Street. He figured out how to play the game, and he won."

FOR A WHILE, PURCELL SEEMED TO BE managing the company as well as he was managing his career. When he merged Dean Witter into Morgan Stanley in 1997, Purcell envisioned a new kind of firm that would have all the upside of a traditional Wall Street house without the risk that comes from relying solely on volatile trading and investment-banking income. While remaining focused on the securities business, Morgan Stanley would maintain four distinct revenue streams: trading and investment banking, retail brokerage, asset management, and credit cards. He described the business model as a sort of relay team. If one business unit faltered because of a change in the macro economy, another would always be there to pick up the baton.

The plan operated perfectly for three years. Morgan Stanley was in the sweet spot of the late 1990s. Its market share in global M&A rose from 23% in 1997 to 35% in 2000. It handled one of every five IPOs, up from one in 20 at the time of the deal. Return on equity surged, from 15% premerger to more than 30%. Earnings more than doubled, from $2.6 billion in 1997 to $5.5 billion in 2000. The stock gained nearly 400%.

But when the crash came, it turned out that Morgan Stanley wasn't so insulated against market risk after all. Its equity- underwriting deal flow fell from $58 billion a year at the peak to $40 billion in 2003. Its percentage of M&A was halved by 2002. As a result, net income for the institutional securities business plunged from $3.5 billion in 2000 to $1.7 billion two years later.

The crash hit Morgan Stanley's other market-related divisions just as hard. The brokerage saw net income plunge from $715 million in 2000 to $23 million the following year. From 2000 to 2002, asset management's profit fell by a third, to $418 million. Discover was the only division left to pick up the baton. But its $760 million profit (in 2002) wasn't enough. Morgan Stanley's earnings plunged to $3 billion in 2002. By contrast, the much more diversified Citigroup saw its earnings jump from $13.5 billion in 2000 to $15 billion in 2002.

And now new threats were arising. In response to the crash, the Federal Reserve began lowering interest rates, making it more attractive for companies to borrow. This, in turn, amplified the competitive advantage of big banks with commercial lending operations. The banks escalated efforts to win underwriting business--which has fat profit margins--by offering generous loans to potential clients. Suddenly names like UBS, J.P. Morgan, and Bank of America had joined Citigroup on the league tables. J.P. Morgan, for example, currently ranks second in global M&A for the first nine months of 2004, supplanting Morgan Stanley. (Although the $261 billion in deals handled by the other half of the old House of Morgan includes $58 billion from its purchase of Bank One--a fact Purcell gladly points out.) Others, such as Bank of America, with an 8% market share, and UBS, with a 12% share, have also been creeping up in the rankings.

Purcell says that some Morgan executives have urged him to do a deal with a bank to bolster Morgan Stanley's lending power. But his belief has always been that Morgan Stanley could compete with the banks without becoming a bank. Recent experience, he says, has confirmed that view. He cites disasters like Enron and WorldCom where banks loaned heavily and lost billions of dollars. Morgan Stanley didn't do business with any of those firms, in part because it couldn't offer the same generous loans. "Maybe we got lucky in some cases," says Purcell. "I also think that we were more disciplined about the way we used our balance sheet." But the larger point, Purcell says, is that Morgan Stanley doesn't need to make loans to win business. "Our advantages are the ability to attract better talent, the ability to innovate, the ability to move much quicker," he says.

While Purcell insists that the business model is right, he now admits that some parts of the model weren't working. For one thing, he believes Morgan Stanley's bankers were too focused on cooking up the next deal rather than serving as long-term strategic advisors to their clients. "We had to return to our roots to focus on clients rather than the next transaction," says Vikram Pandit, a 20-year Morgan Stanley veteran and head of institutional securities.

In the wake of the crash, Purcell, who began his career at McKinsey & Co., came up with a new structure for banking that turned the dealmakers into quasi--management consultants. After putting the bankers through an informal training process that involved listening to presentations about customer surveys, he told them he wanted them to stop focusing on deals and concentrate on building relationships that would ultimately make even more money for the firm. He took star bankers who had previously worked on big mergers and IPOs and assigned them to specific corporate clients.

Has the new approach paid off? Purcell points to the firm's No. 1 ranking in global equity underwriting as proof that it has. And he's been practicing what he preaches. His presence has helped land several deals, including the IPOs of Google and Deutsche Post Bank and an expected lead role in the IPO of China Construction Bank. "Meetings with Phil are insightful," says IBM CEO Sam Palmisano, who has hired Morgan Stanley on several big deals in the past two years. "He consciously tries to understand our strategy in detail, vs. coming in and just presenting transactions."

When it comes to the divisions he once headed, Discover and the old Dean Witter, Purcell has been less eager to make changes. His argument is that each is best in class. But both businesses were built on models that now seem out of date.

Discover, for instance, has been treading water for years. Purcell conceived the card when he was the head of strategic planning at Sears. The idea was to compete with Visa and MasterCard by offering rebates to cardholders. But Discover never claimed more than 6.3% of the card market. A recent federal court ruling found that Visa and MasterCard broke antitrust laws by preventing Discover (and American Express) from marketing its card through banks. Analysts say a settlement could drop billions of dollars into Morgan Stanley's coffers and open the way for pacts with the banks. But the card market is already overloaded with competitors spending billions on marketing trying to steal customers from one another; it is far from clear that Discover can ever be anything more than a niche player.

The brokerage has an even fuzzier outlook. Dean Witter once thrived by selling stocks and in-house mutual funds to "mass affluent" investors--upper-middle-class families. But those folks seem to have lost their taste for playing the market. Studies show that most investors are putting their money into mutual funds rather than stocks, with the American Funds, Vanguard, and Fidelity grabbing most of the fresh cash. Merrill Lynch, which has aimed for higher-net-worth clients, has been less vulnerable to the trend. Last year the average Merrill broker produced $750,000 in commissions and fees, vs. $450,000 for the typical Morgan Stanley broker.

Meanwhile, regulators have come to take a dim view of brokers who push in-house funds--the very backbone of the old Dean Witter's business. While Merrill Lynch and Smith Barney have adapted well to the change, Morgan Stanley seems wedded to the old ways. In a survey by Registered Rep. magazine, Morgan Stanley was shown to be more likely than any others to pressure its brokers to sell in-house goods. In fact, regulators nabbed the firm last year for doling out more than $1 million in incentives, including Britney Spears concert tickets, to brokers who sold in-house funds, and for misleading fund investors about fees. Last year Morgan Stanley had to pay more than $50 million in regulatory settlements related to mutual fund sales practices. That was on top of the $125 million it paid in connection with New York State attorney general Eliot Spitzer's investigation into research practices. Although profits have recovered from the post-crash low, the brokerage still makes only a small contribution to Morgan Stanley's earnings. Analysts are increasingly eager for Purcell to ditch the unit all together. "The old Dean Witter is increasingly irrelevant," says Trone from Fox-Pitt Kelton. "They know it, but they don't want to admit it."

If Purcell has a grand strategy for fixing Morgan Stanley's woes, he continues to keep it a mystery. He talks about doing minor acquisitions to bolster core businesses but says he doesn't need a transforming deal. He claims he isn't looking to merge the company but hastens to add that anything is possible. He says that he plans to retire in four years and then declines to identify any potential successors.

Speculation is that Purcell wants to go out on top of yet another big deal that will leave a final mark on Wall Street. But Purcell isn't even hinting at his plans as he finishes his round at Winged Foot. He's tired of answering questions. His focus is on making one last putt to win a match. It is now late in the day, and the sun is setting. His eyes become steely again. He strokes the ball into the hole and collects $16 for the victory. Turning his back to the question of his future, he walks off toward the clubhouse. "Morgan Stanley," he says, "doesn't have to do anything."