I had dinner recently with an eminent academic, a person whose book has been getting a lot of popular attention. We discussed his readings and radio appearances, chatted about the fact that the paperback was in its fourth printing. Finally I asked who represented him.

"What?"

"Who's your agent?" He looked at me like I'd asked if he paid for sex.

"Why give away 10 percent of my earnings to an agent?"

Oh.

First I had to point out that, in fact, the going rate for most literary agents is now 15 percent.

Then I realized that this author, a senior professor at a fancy-pants university, long published by trade presses, well established in his field, had no earthly idea what literary agents do, and why -- in today's publishing environment -- the idea of trying to publish trade books without one is, well, quaint.

If all you want to do is write monographs for your peers, if your authorial intent is to fill a specific scholarly gap, if your publishing ambition is simply to be published, then you do not need an agent. Some university presses still publish monographs -- though not as many as before and not always with financial success. And university-press editors will still read everything submitted to them (unlike trade-book editors, who will not even open an unsolicited -- read "un-agented" -- proposal), even if they may take a year to get back to you; note that even if a university-press editor decides to publish, she is unlikely to actually edit your manuscript, relying on the creaky process of peer review for editorial input.

Not so many years ago writing a trade book would bring accusations of popularizing, an academic sin worse than spending a Sunday night watching the Super Bowl. No more. Now university presses are turning away from cranking out piles of narrow monographs too expensive even for libraries and are actively looking for books that have at least an academic/trade market, books that will cross over to scholars in other disciplines or outside a narrow subfield. At the same time, commercial presses are hungry for serious, well-researched books that will appeal to people who want something more substantial than the next John Grisham. Trade publishers are also willing to pay big advances for the prestige of having heavyweight authors on their list. It isn't hard to think of powerhouse intellectual scholars who have become rock stars of the scholarly firmament. Hey, I'd line up to get Simon Schama's autograph.

How do these "popular" academic books happen? Do their authors instinctively know how to write for a broad audience? No stinking way. For the most part, rock-star academics are made, not born. And the people who make them are literary agents.

Even at trade presses, editors rarely have time for serious editing. They have become buyers, and their Palm Pilot dance card is most likely to be filled with agents who have projects to sell. Savvy agents know the tastes and predilections of individual editors, know when they are starved for projects, and know, too, the trends of the publishing industry. They know about multiple submissions and how holding an auction can artificially inflate the size of the advance. They know about mimetic desire: An editor will more likely want a book if she knows that others want it as well.

Most important, agents know how to help authors write the kinds of proposals that editors will want to buy. The trick with academics, agents say, is to get to them before they actually sit down to write. Agents tend to sell books -- sometimes for staggering amounts of money -- based solely on a proposal. They deal with ideas. Good agents help the author figure out what is interesting, exciting, and relevant about her work. They ask the hard questions like, Who cares?

They know being interesting isn't enough. Agents ask, Why would anyone need to buy this book? The fact is, if you want to write a popular book, you have to give the reader a payoff. Perhaps it's an exciting tale, well told. Perhaps it's an insight into why we are the way we are. Maybe it's learning about a new idea, a new way of thinking. It's not easy to have new ideas, I know, I know. It could be that the same 1,200 people at academic conventions have kicked around an idea for the last 20 years. But if it hasn't gained currency outside of a small scholarly niche, then it is, to the nonacademic world, new.

Agents can help academic authors pull out the implications of their work. Sure, we all sneer at students who ask: Why do I need to know this? But the fact is, you need to make readers, like students, understand why they need to know what you want them to know. An agent will shape, shift, and sort. An agent, a good one anyway, will see through to the core of what makes your work, your argument, your data necessary. That vision will carry through to create a scaffolding for a book. You will still, of course, have to write it, there's no getting around that. But how much easier it is to write with someone there by your electronic side, telling you when you're on course and when you're drifting too deeply into waters all by yourself, where only the few will follow.

After spending a dozen years as an editor, when I started writing I knew I needed help. My agent has sent me back to a blank sheet more times than I would have imagined I ever would be willing or able to do. Drafts of manuscripts come back to me with "THINK HARDER, RACHEL." Pages of what I thought was pellucid prose come back crossed out: "This is exquisite, but it doesn't belong in this book."

Sometimes I argue. Sometimes, if I can convince my agent I'm right, she'll back down. We agree on where I need the most editorial intervention: She never messes with my sentences, but she has thrown out entire chapters. Our relationship is intimate, important, and based on trust, basically on my trust in her abilities to understand where I am trying to go and to help get me there.

The real question, it seems to me, is not why would anyone want an agent, but how could you expect to reach a broad audience without one? Agents hold up a magnifying mirror to your work, helping you mask the flaws while working to bring out the best, most valuable parts. Then, once you get the book written, your agent will mediate your relationship with your publisher. An agent will represent your interests and will tell you, if you're wrong, when to back off. Remember: The publisher is paying your editor; you pay your agent.

Fifteen percent seems like a bargain.

Rachel Toor, a former admissions officer and book editor at Duke University, is the author of Admissions Confidential: An Insider's Account of the Elite College Selection Process (St. Martin's Press, 2001). Her memoir, The Pig and I, will be published by Hudson Street Books in 2005.