I was going to start off this review with a joke about how astronomers make the most authoritative Chicken Littles, but after a while it didn't seem so funny. Returning from Central Asia last month with our newly adopted daughter, my wife and I had midnight encounters with customs officials armed with machine guns and wearing surgical masks to guard against SARS. As this was being written, The Washington Post was running a series of articles about germ and chemical weapons for sale, left over from a secret program in South Africa.

Martin Rees, Britain's Astronomer Royal, a professor at Cambridge University, one of the world's most brilliant cosmologists and a longtime arms control advocate, gives civilization as we know it only a 50-50 chance of surviving the 21st century. The proposal for ''Our Final Hour,'' a breezy but deadpan recital of all the possible ways that the sky could fall on us, was so depressing, Rees has said, that his agent had a hard time selling it.

But that was before 9/11.

The choices we make in the next few decades, Rees contends, could decide the fate of life not only on Earth but beyond, either ensuring its survival -- if we can diversify into space -- or dooming it forever. ''It may not be absurd hyperbole -- indeed, it may not even be an overstatement -- to assert that the most crucial location in space and time (apart from the Big Bang itself) could be here and now,'' he writes.

Humanity has progressed to the point where we are now our own worst enemies: adding to the backdrop of natural calamities that have always threatened us, technology, Rees argues, has now so highly leveraged the power of the individual or the small group that a biological ''unabomber'' or a mistake in a laboratory could wreak havoc only dreamed of by the Strangeloves of the last century, who held the forces of nuclear apocalypse at bay by war-gaming scenarios of mutual assured destruction. He says, in fact, that he has bet $1,000 that an instance of bioterror or bioerror will take a million lives before the year 2020.

But there are many things to worry about, some of which will be familiar to all: global warming, asteroid impacts and that old bugaboo nuclear war, which has been transmogrified by the end of the cold war; the collapse of the Soviet Union, Rees points out, has left the world awash in the raw materials, enriched uranium and plutonium, for some 70,000 bombs.

Others are novel. Engineering advances could lead to the creation of intelligent self-reproducing nanoparticles that could eat us and every other living thing on Earth, reducing the biosphere to what Eric Drexler, one of the pioneers of nanotechnology, calls ''gray goo'' -- the subject of a recent thriller, ''Prey,'' by Michael Crichton.

Certain physics experiments might be even more catastrophic, Rees reports. In principle they could disturb space-time itself, causing the laws of physics to twitch into a new form, like water suddenly freezing to ice, destroying our atoms and everything else. Since we lack a ''battle-tested'' theory of what happens at very, very cold temperatures, he says, we would have been right to be worried when a metal bar -- part of an apparatus to detect gravitational waves, ripples of space-time predicted by Einstein's general theory of relativity -- was recently cooled to near absolute zero, making it what Peter Michelson of Stanford University called ''the coldest large object in the universe.''

In a book whose influences range from science fiction writers like H. G. Wells, Kurt Vonnegut and even Tom Clancy to modern scientists, Rees throws every possible calamity into the pot -- some credible, some not. The point is not that any of these particular disasters will befall us, but that something might. And scientists, he suggests, have sometimes been more interested in public relations than in really leveling with the rest of us about the odds we face.

Take the cold war, for example. Rees does not think the arms race between the United States and the Soviet Union -- a contest that he says was pushed along largely by scientists doing their jobs to improve weapons and thus get an edge on the other side -- was worth the risk. He quotes President Kennedy as saying that the odds of nuclear war during the Cuban missile crisis were between one in three and even. And Rees says Robert McNamara, Kennedy's secretary of defense, rated the risk at ''substantially higher than one in six'' during the cold war. In other words, worse than a game of Russian roulette. Had the public known those odds, would we have played along? ''I personally would not have chosen to risk a one-in-six chance of a disaster that would have killed hundreds of millions and shattered the physical fabric of all our cities, even if the alternative was a certainty of a Soviet takeover of Western Europe,'' Rees adds.

Before the first atomic bomb test, scientists took the time to calculate whether the blast would ignite the nitrogen in Earth's atmosphere and incinerate us all. The risk was low and the test went off, but Rees wonders what the odds would have had to be to discourage the bomb makers.

MORE recently, physicists at Brookhaven National Laboratory calculated the odds that a planned experiment, in which atomic nuclei would be accelerated to collide at high speeds, might cause all the matter in the earth to collapse into exotic dense particles called ''strangelets,'' extinguishing life, among other things. The risk came out to about one in 50 million. That sounds good, and the experiments commenced without tragedy, but Rees takes no comfort in that, pointing out that the results of the calculation can also be expressed as saying that 120 people might be expected to die from the experiments. Not even the most ambitious physicist would advocate accepting such a price for scientific knowledge.

The calculations and decisions relating to a possible extinction event get more complicated if the lives of the unborn are included as well.

In such cases, science is too important to be left to the experts who want to do the experiment, Rees says: ''It isn't good enough to make a slapdash estimate of even the tiniest risk of destroying the world.'' He endorses a suggestion made by the physicist Francesco Calogero of the University of Rome that ''red'' and ''blue'' teams of outside experts should debate experimental risks in public. For Rees the shining moment in this respect was the 1975 conference in Asilomar, Calif., where geneticists agreed to hold off on certain recombinant DNA experiments until it could be ascertained that deadly organisms would not result.

Of course, scientists now do not have the last word on what work or experiments they will do; their paymasters in government or industry do. Rees's call for more public accountability is likely to raise hackles with some of his colleagues who worry about political interference with research. Erica Goode recently reported in this newspaper that AIDS researchers had been advised that certain terms, like ''sex worker'' or ''gay,'' should be avoided in grant proposals lest they draw unwanted attention from ideologues in the Bush administration or Congress.

Fortunately -- or unfortunately for his career as a polemicist -- Rees is too good a scientist to be content with telling only one side of the story. He insists on detailing objections to his own arguments. As a result, we don't always know where he finally comes down on some issues. For example, he begins a discussion of global population growth with the news that it will take the resources of three Earths to support the eight billion people who are expected to inhabit the planet by 2050. But then he reports that the population is expected to fall after that, so that a century from now Earth's population could be less than it is today. So is overpopulation a long-term problem or not? I don't know.

Rees's book comes alive in the last few chapters when he reconnects with cosmology, his own specialty, and confronts the issue of what the extinction of humanity would mean for the universe. Cosmology used to be the most heartless of sciences. Whatever grandeur could be found in the immense wheeling of galaxies, whatever elegance there might be in the Einsteinian warping of the inky loneliness in which they swim, it's a pretty cold grandeur, an acquired and comfortless elegance at best, for most of us. Lately, however, cosmologists have concluded that the universe and the life that lives in it are all of a piece. Life as We Know It seems to depend on a miraculous and improbable juggling of the numerical values of a few atomic constants.

Cosmologists admit they don't know what to make of this. Is life a lucky fluke? Are there zillions of different universes to choose from? What is life, anyway? There are imaginative and impressive arguments that humans are alone in the universe and equally impressive counterarguments that life is ubiquitous. So much for arguments.

Humans are a recent addition to the cosmos. If the whole 10-billion-year projected lifetime of the Sun were compressed into one year, Rees writes, all of recorded history would take up less than a minute, and the 20th century would amount to less than a third of a second. The unspoken assumption behind such expositions, of course, is that we are just getting started; bigger, mightier things lie ahead. But it might be, Rees suggests, that the future of humanity is as slender a thread as its past. If we are alone in the universe, whether we survive or not will determine whether there is a point to the rest of cosmic history.

DOING nothing, Rees points out, is not an option. In two billion years or so, the warming Sun and the greenhouse effect will make Earth too hot for anything but microbes; in a few billion years more, the planet will be toasted to a cinder in the Sun's death throes. The answer lies in space. Once humans have established homes or colonies on separate planets, there will be less chance that a single catastrophe -- be it a plague or an asteroid -- can get us all (although there is a greater chance that the species will diversify genetically).

But NASA is doing it all wrong, Rees says (he wrote the book before the crash of the space shuttle Columbia in February). The space program needs new technology and a new ''style.'' Rather than being a quasi-military government program, he believes, it should become the province of wealthy adventurers who are ready to accept high risks in the pursuit of frontiers and of thrills beyond those provided by yacht racing or ballooning. A Moon colony or a Mars trip would be suitable endeavors for people like Bill Gates or Larry Ellison, he surmises.

Nothing is forever, perhaps not even the universe, but such actions could give us a chance. ''Long before the Sun finally licks Earth's face clean, a teeming variety of life or its artifacts could have spread far beyond its original planet; provided that we avoid irreversible catastrophe before this process can even commence,'' Rees concludes. After taking this stroll with him through the shadow of death, I would be grateful for any good news.

Dennis Overbye is a science correspondent for The Times. His most recent book is ''Einstein in Love: A Scientific Romance.''