(FORTUNE Magazine) – This past March 1, Barry Tarasoff, the research director at S.G. Cowen, sent an amazing letter to his firm's clients. Cowen is a midsized Wall Street firm--owned by Societe Generale, the big French bank--that offers investment banking and institutional trading along with securities research. It caters exclusively to the most sophisticated investors on Wall Street--the hedge funds, mutual funds, and pension funds that collectively make up what's known as the "buy side." In return for Cowen's research in sectors such as media and health-care companies, its clients buy and sell some of those stocks through its trading desk, thus generating commissions for the firm. This, of course, is how buy-side institutions have paid for Wall Street research since time immemorial.

For decades Cowen's researchers spent much of their time upgrading and downgrading stocks, writing up quarterly earnings reports, and setting up meetings between their clients and company managements--just like all the other "sell side" analysts on Wall Street. They also devoted considerable time and effort marketing themselves to institutional clients in an effort to generate both trading commissions and votes for the annual Institutional Investor magazine All-America Research Team, which greatly affected an analyst's compensation. But at Cowen, that was all about to change.

Wall Street research has gone awry, Tarasoff began in his letter. The emphasis on the All-America Research Team "was an insidious development for our profession because it focused much of the research community's attention on the wrong goals and led to a number of unfortunate practices that survive to this day." Thus, effective immediately, the firm would abolish several of those unfortunate practices, starting with its rating system. No more buy, hold, or sell ratings on the stocks covered by Cowen. Its analysts would still have investment opinions, Tarasoff hastened to add, but they would express those opinions "the old-fashioned way, using the full richness of the English language."

Cowen was also abolishing a second pillar of modern securities analysis: the quarterly earnings note. "Most of these voluminous postmortems serve no purpose and consume considerable resources," Tarasoff wrote. "This policy will ensure that our analysts write only when they have something important to say." Finally, in place of ratings and earnings notes, S.G. Cowen would stress "good investment ideas" and "proprietary information flow." That is, the firm would publish research only when it had some unique value and could not easily be duplicated by competitors. "This amounts to a fundamental change in the way we think about our job," Tarasoff concluded.

Indeed, it does--and not a moment too soon. Since April 2003, when ten of the largest Wall Street investment banks agreed to pay $1.4 billion to settle charges brought by New York attorney general Eliot Spitzer that its analysts had betrayed investors during the late 1990s boom, Wall Street research has been in a state of turmoil. Analyst pay has been slashed in half--or more. Highly paid stars have been ushered out the door, replaced by junior analysts. Most big research departments have shrunk by more than a third.

But more profound changes are coming, because Wall Street research finds itself under a far graver threat than even Eliot Spitzer: Its business model is under assault. Dozens of new research boutiques have cropped up in recent years claiming to offer truly independent research. Some are aimed at retail investors, others at the fast-growing hedge fund community--but they're all trying to take advantage of the scandal to grab business from the big firms. What's more, many of the buy-side institutions that use Wall Street research are complaining loudly that too much of what passes for securities analysis is pointless and duplicative--and they are no longer willing to pay for it with their commissions dollars.

Clearly, merely downsizing research departments will not be enough to stave off the coming crisis. It is going to require a fundamental rethinking of the entire business. What is the real purpose of securities research? Who are the customers? How radically does it have to change to be genuinely valuable again? Is it important to even have a research department anymore?

Barry Tarasoff's answers to those questions were embodied in his March 1 letter. But as I discovered after talking to analysts, former analysts, buy-side executives, and research directors, virtually every top executive on Wall Street is engaged in a similar exercise--and they've all got their own answers, many of them quite different from Tarasoff's. The one thing they all agree on is that Wall Street research has to change while there's still time--and that change will have implications not just for the analysts and the firms that employ them but for investors both large and small.

And if it doesn't change? As Sallie Krawcheck, the head of Citigroup's Smith Barney brokerage unit, puts it, "If Wall Street doesn't get its act together and come up with high-quality, moneymaking research, it has the potential of going the way of the dodo bird."

As is usually the case with Wall Street, the grand philosophical questions are prompted by hard-core economic questions. The first is, How much is research worth? The second: How is it going to be paid for? Strangely, those are questions that both the buy side and the sell side have been dancing around for years. Because research costs are folded into trading commissions, securities analysis has often been thought of by investors--institutions and individuals alike--as a freebie. But it's not free; a percentage of every commission dollar covers research costs. As for Wall Street itself, since commissions rolled in whether its research was good or bad, there was little incentive to insist on quality. That's what is changing: The buy side is demanding that the costs of research be quantified--and the sell side is quaking at the thought of it.

This is hardly the first time Wall Street research has found itself facing a crisis after having its economic underpinning kicked out from under it. Think back to 1975, when trading commissions were deregulated after literally centuries of operating under a fixed price. As Krawcheck likes to remind people, "As scary as this moment seems, that must have been much scarier." And it was. Back then, long before the era of the integrated financial services giant, most research was done by boutiques that specialized in securities analysis. They put out lengthy, detailed reports on stocks for clients, who read them at their leisure. If the clients were pleased, they would then direct commissions to the research firm as payment. With commissions between 30 and 80 cents a share in 1975 (they were fixed according to a sliding scale), there was plenty of money to go around for everyone.

Once commissions were deregulated, they began dropping dramatically--and old-style boutiques were suddenly in trouble: The new, lower commissions no longer supported their operations. Boutiques that survived the turmoil of that era, such as Sanford Bernstein, were rare. Most went out of business or were absorbed by large firms like Merrill Lynch, which realized that research could be a tool both to bolster their retail operations and to break into the institutional trading business, Wall Street's big growth area at the time.

The fundamentals of research changed too. Reports became shorter and more frequent, largely because portfolio managers were jumping in and out of stocks quickly. Analysts began rating stocks, and issuing upgrades and downgrades during their firm's "morning call." The telephone became the weapon of choice for analysts, who raced to call important buy-side clients with even the tiniest tidbit of news. Reporting quarterly earnings became important because investors put increasing weight on whether a company met its profit expectations. Analysts also started going on one-man road shows to trumpet their investment ideas to clients. This kind of marketing was a necessity if an analyst hoped to do well in the annual Institutional Investor poll.

All the while, commissions continued to drop; as Sanford Bernstein CEO Lisa Shalett is fond of pointing out, "Research has faced 6% annual compounded price deflation for the past 30 years." But for much of the '80s and early '90s, the decline in commissions was somewhat offset by the increase in trading volumes. Research was hardly a big profit center. Still, it was considered as critical to the institutional trading business (and the retail brokerage business, for that matter) as, say, the firm's ability to execute a trade efficiently. "Research is what causes a transaction," says Scott Cleland, CEO of Precursor, one of the new investment boutiques. In interview after interview, that was a refrain I heard often. "Investment ideas are valuable," says Sanford Bernstein's Shalett. "Research is the intellectual capital of our business," says Suzanne Nora Johnson, head of global research at Goldman Sachs. "It is critical to the way we do business." This became the economic rationale behind research departments: The analysts made calls that caused people to buy and sell stocks, which generated commissions, which translated into profits. The analysts' "intellectual capital" kept institutional trading from becoming commoditized and gave firms something unique to peddle to clients. (For more on a few top analysts who still do old-fashioned stock picking, see "The 2004 Fortune All-Star Portfolio.")

With the stock market boom that began in the mid-1990s, research changed again--and once again, the change was driven by the economics. By then, even huge trading volumes couldn't make up for ever-declining commissions, which were under 10 cents a share. Research departments were employing hundreds of analysts and carrying upwards of $200 million a year in costs. Without another revenue source to buttress commissions, research departments were in trouble.

Just in the nick of time, though, research found its new rationale. It found investment banking.

Research, of course, had long played an important role in the investment banking process. But it was not a role necessarily designed to generate revenue. On the contrary, it was a role that could very well deprive a firm of revenue. Before the '90s, when investment bankers were preparing to underwrite a stock the firm's analyst was brought in to vet the proposed deal and make a judgment about its suitability. The analyst was there not to protect the interests of the firm but to protect the interests of its investing clients. If he (or she) thought the deal was not in the clients' best interests, he was supposed to to say so--even if it meant that the firm lost the business. "Prior to the bubble," says one research director, "I never saw an instance of an analyst being so spineless that he supported a transaction he didn't believe in. It was reputational suicide. But during the bubble, that changed."

Did it ever. As we all now know, many analysts abandoned their responsibility to investors and became pawns of the investment bankers, churning out "research" designed to keep companies happy--and banking deals cooking. "They went from having investment banking deals pushed on them in the 1980s to becoming the greedy pigs at the center of it," says James Freeman, a former First Boston research director who now runs a consulting business. At firms such as Credit Suisse First Boston, the cozy relationship between banking and research even became institutionalized, with a group of analysts reporting directly to Frank Quattrone, who headed CSFB's Silicon Valley investment banking practice. (Quattrone left CSFB in 2003 under a cloud and was recently convicted of obstructing justice.)

The essential purpose of research, to give untainted investment advice, became utterly corrupted--but in purely economic terms, it made perfect sense. Investment banking generated tremendous revenues, and if research could take credit for bringing in some of that money, it could justify itself in an era of dwindling commissions. Indeed, as compensation became tied to banking, many analysts suddenly found themselves making millions. One former analyst, who asked not to be named, says that in the space of ten years his compensation rose from $200,000 to $2 million a year--and he wasn't even in a sector that did many banking deals. (When his pay went back to $200,000, he quit.) Research departments expanded dramatically, as firms got into bidding wars over top-rated analysts and hired teams of new analysts to cover hot banking sectors such as Internet companies. At the peak of the boom, according to Sanford Bernstein analyst Brad Hintz, some 42% of research costs were being covered by investment banking departments. Hintz estimates that by 1999, Merrill Lynch alone was spending over $550 million on its research department.

Did individual investors understand that Wall Street research had lost its independence? Surely some of the more sophisticated individuals did, but many didn't, and they were done genuine harm by the likes of Smith Barney's Jack Grubman and Merrill Lynch's Henry Blodget, to name the two most notorious analysts from the era just past. (Both men, of course, have since paid big fines and been banned from the securities industry for their misdeeds.)

But at the big buy-side institutions, fund managers absolutely understood how corrupt research had become. Yet--and this is the key point--they continued to pay for it with their commission dollars. Partly that was because there was a lot of downside in refusing to pay for Jack Grubman's research. If you cut off Grubman's commissions, you wouldn't just lose access to his tainted research, you'd also lose access to the CEOs he was so close to, like Bernie Ebbers at WorldCom or Gary Winnick, the chief executive at Global Crossing. And you'd lose a chance to get allocations of hot IPOs Citigroup was handling in the late 1990s. That was a very steep price to pay.

Buy-siders also looked the other way because, like many during the boom, they got lazy. If you were a manager of a mutual fund, your portfolio was probably going straight up--so what did it matter if you spent a little more than you should have for corrupt research? Besides, the commissions didn't really come out of your hide; those expenses were absorbed by the fund's shareholders, not the management. If 40 analysts put out post-earnings notes after Cisco announced its quarter, you didn't distinguish between the good and the bad reports. Everybody got paid. "There was no market mechanism to punish bad research," says Scott Cleland. Thus the issue of how much the buy side was paying for research--vs. how much value that research truly provided--became extremely muddy.

And into that morass walked Eliot Spitzer.

You can scarcely talk to an analyst these days without hearing about all the new rules that emerged from the Spitzer settlement. Analysts now have to get approval from a "research recommendation committee" before they can initiate coverage of a new stock or even change a rating on a stock. Their public appearances are monitored closely--if they're even allowed. Though analysts are still supposed to vet banking deals, they can no longer attend the meetings where the investment bankers pitch underwriting business. Analysts aren't even allowed to speak to bankers without a chaperon in the room to monitor the conversation. (A number of firms have blocked phone calls and e-mails between analysts and bankers.) Analysts have to add new disclosures when they put out research reports, including a certification that "all the views in this research report accurately reflect our personal views ..." And beginning in late July, the ten firms that settled with Spitzer will start paying millions annually for independent research that will go out alongside their own reports. (Under the terms of the settlement, the ten firms have agreed to purchase--and disseminate--this independent research for the next five years, at a total cost of $432 million.)

The intent of the new rules, of course, is to ensure that analysts are actually serving the interests of investors again, not bankers. But most analysts I spoke to felt the new rules were largely an annoyance--and demeaning. "They have created a climate of terror on the Street, and that is stifling some people," says Judah Kraushaar, a former top-rated analyst at Merrill Lynch who left the business last year. A current analyst, who asked not to be named (no one wants to be quoted on the record talking about the Spitzer settlement), says, "I roll with the punches, even though I swear under my breath." Analysts feel that people are constantly looking over their shoulders, questioning their work. It is a big reason so many of them have quit.

Have the rules succeeded in refocusing research on investor concerns and away from banking? Up to a point. Sell recommendations, virtually unheard of during the late-1990s boom, are now fairly common. I also heard plenty of stories about analysts, freed from pressure from bankers, who vetoed important underwriting deals. Last December, at Smith Barney, REIT analyst Jon Litt was underwhelmed by one deal. When the bankers went through with it anyway, Litt stood up and announced to the firm's traders, "Your money is better spent elsewhere."

But I heard other stories suggesting that at some firms, banking and research were still a little too cozy. One investment banker, who insisted on anonymity, said that most companies looking for underwriters still want to be sure they'll get positive research coverage once their stock is issued. And investment banks still feel they have to accommodate them. "Before the settlement," this banker said, "the analyst would be with us when we pitched the client, and he would articulate the story--how he was planning to sell the company to investors. Even though the analyst is no longer allowed at the pitch, that still seems to be the criterion: How much will the analyst support the stock?" Some firms, he continued, were encouraging their analysts to visit companies two and three times before the bankers showed up for the pitch meeting. Through winks and nods, they signaled to the company's executives that they planned to publish friendly research. The banker concluded, "Are people still using research to attract business? I think the answer is yes."

Indeed, if all Spitzer had done was set up new rules, it seems likely they would quickly become ineffective as Wall Street searched for loopholes. As the old adage goes, You can't legislate morality. But Spitzer did something far more powerful: He also separated banking and research economically. Under the settlement, analysts can no longer be paid for helping to generate banking deals, and research can no longer view itself as an adjunct to the banking department. Thus, research is back to one revenue stream again--trading. And trading commissions are lower than ever--under 5 cents a share for a full-service "bundled" commission. "The bottom line," says Neil Barsky, a former analyst who now runs a hedge fund, "is that research is over as an obscenely paid profession."

Then Spitzer did something else. After he'd finished with Wall Street analysts, he turned his attention to mutual funds, where he uncovered the next big scandal: insiders who market-timed and late-traded mutual funds. This may not seem connected to the current research imbroglio, but it is. Indeed, strange as it may seem, the mutual fund scandal is likely to affect the future of Wall Street research as much as Spitzer's analyst investigation.

Why? Once again, it is going to change the underlying economics of research. When Spitzer exposed the nasty secrets of the fund business, he also raised awareness about funds' hidden costs to shareholders. One of those costs, clearly, is full-service commissions. (It's also sometimes called soft-dollar compensation--the practice of diverting a portion of commissions to cover the cost of research.)

In the wake of the mutual fund investigation, some fund companies--along with the SEC--began asking the tough questions about research and commissions they had ducked all these years. How much were they truly paying for research with their commission dollars? Given that electronic exchanges can now execute trades for a penny a share, was that additional 4 cents too much to be paying full-service trading desks? Some companies, most notably Fidelity Investments, the biggest fund complex of them all (and a company untainted by the mutual fund scandal), even began asking whether it was more appropriate for funds to pay for research out of their own pockets instead of having shareholders pay for it through commissions.

Although the sell side would still prefer to keep ducking these questions, it can't anymore. Fidelity has been in talks with most of Wall Street, essentially demanding that the firms begin quantifying how much of their commissions are going for research. "They tell us not to take it personally, but they are bound and determined to find a way to unbundle our commissions," says one person who has been involved in some of those discussions. If Wall Street resists, Fidelity is even threatening to single out one firm and stop doing business with it--just to drive the point home. But no one is going to resist. Fidelity's business matters too much.

In early March, Fidelity sent a letter to the SEC calling on the agency to "require mutual funds to quantify and separately report soft-dollar research expenses in dollars and as a percent of assets." In other words, it wants the entire fund industry to be required to tell its shareholders precisely how much of their money portfolio managers are spending on research. In the letter, Fidelity also made it clear that it would like to see the SEC eventually go further--perhaps shifting the expense of research from shareholders to managements. But for now, it sees "transparency"--as the Fidelity proposal has come to be called--as a useful first step.

Fidelity is hardly the only company trying to reform the soft-dollar system. There have been numerous ideas put forward, including the most extreme of all: eliminating the system entirely and having the buy side pay cold, hard cash for research. Still, the SEC, which is currently studying the issue, is likely to adopt the Fidelity plan--quite possibly by the end of the year.

And how will transparency affect securities research? Everyone on Wall Street knows the answer: Disclosing how much the buy side is paying for research will cause the buy side to pay less--a lot less. Take Fidelity again. By most accounts, it pays somewhere around $800 million a year in commissions. Even assuming only 20% of that goes for research, that still means the Boston fund giant is paying a staggering $160 million to get some research it really wants--and a great deal that it doesn't. For years the commission system has allowed all that pointless research to exist. It has allowed what Merrill Lynch analyst Guy Moszkowski calls "research overcapacity"--with 20 or 25 or 30 analysts covering the same stocks and peddling the same ideas. Transparency will expose that problem. At which point the buy side will stop paying for research it doesn't want--and research will change more dramatically than at any time since 1975.

IF YOU WANT TO UNDERSTAND JUST HOW RADical a rethinking is taking place among research directors, consider this: Not long ago a rumor was making the rounds that Goldman Sachs was seriously considering getting rid of research entirely. Even a year ago such a rumor would not have been taken seriously, but now, it seems, everything's on the table. Goldman's Suzanne Nora Johnson insists that the rumor was never true--but she does concede that she led a broad review of research to examine, as she puts it, "whether the model works strategically." She also says--talk about everything being on the table!--that during settlement talks with regulators, a number of firms explored the possibility of setting up a single research department, which they would use as a kind of "shared utility." The effort didn't get very far; they mutually concluded that it was unworkable.

Like every research director I spoke to, Johnson has come to the conclusion that research still matters a great deal to her firm. But Johnson also acknowledges that with the coming of transparency, Goldman is under pressure to produce research that is distinctive enough for the buy side to be willing to pay for it.

And that's the game right now: figuring out how to produce research that is unique enough--and valuable enough--to survive the arrival of transparency. Thus, the biggest firms, such as Smith Barney and Credit Suisse First Boston, are stressing their global franchise and their ability to put together research products that connect trends in Asia with trends in Europe or the U.S. Some firms, such as S.G. Cowen, are eliminating ratings; others, including Smith Barney, are stressing them more than ever. ("Quality of research equals good stock picking," insists Sallie Krawcheck.) Virtually everyone is downplaying quarterly earnings reports in favor of "proprietary" research--that is, research unique to their firm. Goldman did a survey of beer drinkers to find out if their drinking habits were being affected by the low-carb craze. Merrill Lynch did a smart report about the state of the aluminum industry in China. At Credit Suisse First Boston, the buildings-materials analyst has a proprietary network information database--which her bosses are highlighting as a model. (At the other end of the scale, some firms, including J.P. Morgan Chase, have outsourced basic research tasks to India.)

Meanwhile, dozens of new, independent boutiques have arisen in the past few years, all of which think their models will be among the winners. Some are angling for a piece of the $432 million settlement pie that is going to independent research; others think that's fool's gold, and they're better off concentrating on the institutional market. All of them are claiming that their research is inherently superior to the research being produced by the big Wall Street firms. And they all think they've found a profitable niche.

Here's one example: The Independent Research Group, founded a year and a half ago, is run by Paul Noglows, a former tech analyst at J.P. Morgan Chase. Its focus is on stocks with valuations between $200 million and $2 billion. Why that niche? Because small-cap stocks are precisely the ones that have been abandoned by the big firms as they've shrunk their research departments--yet they're making investors money right now. Noglows can't afford to hire big-name analysts, but he doesn't care: "We're hiring the junior analyst who's doing all the work," he says. "That's our secret sauce." IRG's only ratings are buy and sell--Noglows even disapproves of a neutral rating. Toward the end of our conversation, Noglows declared, "The old research model is dead."

The old research model is dead. That may be the only certainty right now. Ask a dozen Wall Street executives what research is going to look like a year from now, and you'll get a dozen different answers. But virtually everyone is sure that research is about to become a very different beast--and better than it's been in a long time. It's strange to think that you can pay analysts $200,000 instead of $2 million, replace stars with junior analysts, shrink whole departments, produce much less actual research--and in the process create a better product. But that's where we are right now. What Eliot Spitzer started, the market is about to finish.