I am no historian. But I've always been fascinated by the sense of purpose, the intent, of writers of histories that have engaged me -- from Herodotus to Stephen Ambrose, whose final book, To America, I wrote about just last week. I read books of history, and their first cousin, historical biographies, with great interest. If I could read at five times my natural speed, I would surely read more histories.

Never before, though, have I come across a book that so illuminated the craft of the historian than The Landscape of History: How Historians Map the Past, by John Lewis Gaddis (Oxford, 224 pages, $23).

Gaddis knows how the stuff is done. He is a professor of military and naval history at Yale, and has written six previous books, all related to the cold war, to U.S.-Russian relations or to interpretations of both. He researched much of this book two years ago at Oxford University, in connection with a year's visiting professorship, his second there in 10 years. The project was up to him, but he devoted much of the time to examining and lecturing on the purpose and process of historians.

While the work is clearly designed primarily for consideration by scholars interested in writing history themselves or in teaching it, I found it a startlingly provocative guide to reading history. It should engage and inform any reader, I think, who is generally enthusiastic about history books.

Gaddis has a delightful command of language -- and a delight in it. He draws on Gertrude Stein, Mark Twain, contemporary movies, Thucydides, Tom Stoppard, Woody Allen and lots more. In seeking to describe what developing and writing history is all about, he makes "analogies to painting, cartography, and even tailoring as well as to mathematics, astronomy, geology, paleontology, and evolutionary biology."

But he is modest in his expectations for his craft in the future. "It's an unsettling exercise to try to guess what historians two or three hundred years hence will select as significant about our age," he writes. "One depressing possibility might be the defunct website we leave lying around in cyberspace. ... All we can say for sure is that we'll only in part be remembered for what we consider significant about ourselves, or from what we chose to leave behind in the documents and the artifacts that will survive us."

So how do we define history? Gaddis categorically rejects and ridicules the core contention of postmodernism that there is no such thing as certainty, that it is impossible to know surely about the past. Rather, at its simplest, this is the job: "Events recorded in the strict order of their occurrence almost immediately get rearranged into a story with a discrete beginning, middle, and end. These then become histories."

Ultimately, Gaddis argues, the purpose of historians' work is "to achieve the optimal balance, first within ourselves but then within society, between the polarities of oppression and liberation."

In that, he is wary of the perils of abusing history, of using it for political demagogy and worse, citing Hitler, of course, and Marxists in general. "The search for a past with which to attempt to control the future is inseparable from human nature; it's what we mean when we say we learn from experience," he writes. "What's frightening about this process is when it targets victims, when excuses for marginalization lead to discrimination and then to the next logical step, which is authoritarianism."

Gaddis contends that first-rate historians work in a manner that is keenly parallel to that of the stricter -- non-social -- sciences. Physicists seem a favorite simile, but others as well. In this, he suggests that in both categories, what might be called the most important breakthroughs of knowledge, discoveries, come as something like tectonic insights -- even visions.

In the case of historians, their main ambition is to analyze, understand and then communicate the causes and effects of major events. That of scientists is to arrive at insights that provide foundations for replicating and thus controlling future processes. The historian is not in that business, for to do so would make his job one of predicting the future. But Gaddis gives no quarter here -- while making no claims for such predictive responsibilities, he puts sound historians in a far more competent position for comprehending likely future trends and events than, say, economists or political scientists.

He is very tough on the social sciences. With this book, Gaddis will make no friends in faculties of economics, sociology, political science, Freudian psychology or their related fields. "The social sciences" he writes, "[have] operated from the conviction that consciousness and the behavior that results from it are subject, at least in general terms, to the working of rules -- if not laws -- whose existence we can detect and whose effects we can describe."

In perhaps his least harsh dismissal, he argues that social sciences and social scientists "tend to be static, neglecting the possibility that human behavior, individually or collectively, might change over time."

But his ultimate criticism is more contemptuous: "Social scientists seek to build universally applicable generalizations about necessarily simple matters; but if these matters were any more complicated, their theories wouldn't be universally applicable. Hence, when social scientists are right, they too often confirm the obvious. When they don't confirm the obvious, they're too often wrong."

I am not much of an enthusiast for academic writing by academicians. But Gaddis is clearly an exception: He is a distinguished scholar who writes with a clarity and a charm and a lack of pedantry that is quite marvelous. Equally impressive, he's not afraid of a rip-roaring fight with his fellow academics.