Some other people who might kill you, however, are drunk drivers. They are an especially depraved category, many feel. "The most serious crime in America is not...rape or armed robbery," Alan Dershowitz has said, "it is drunk driving." The organization Mothers Against Drunk Driving (MADD) vilifies those who drink and drive and has lobbied for minimum-sentencing rules in drunk-driving cases.

No one would deny that getting behind the wheel of the car while hog-whimperingly drunk is dangerous and something that ought to be deterred by criminal sanctions. But what about those who drive while only slightly intoxicated, barely over the legal limit? Do they deserve jail time, as is mandatory in some states even for first-time offenders? They are engaging in risky behavior, it's true, but all driving is risky: Sober drivers kill about twenty-five thousand Americans every year. Furthermore, some of these motorists could be taking public transportation but are too lazy; they would rather impose extra risk on themselves and fellow motorists. In fact, the difference in risk between driving drunk and driving sober is less than the difference between driving sober and taking mass transit.

Slightly squiffy drivers, those coming home from suburban cocktail parties and the like, may pose a somewhat higher hazard than the average sober motorist--assuming, that is, they do not try to compensate for their condition by driving extra carefully and staying well within the speed limit, as the sensible among them doubtless do. But how great is this incremental risk?

Take the typical driver with a blood alcohol concentration of 0.1 percent, the current threshold in most states. (MADD wants to lower it to 0.08 percent.) Such a motorist is between three and seven times more likely to cause an accident--a risk of the same order of magnitude as that posed by drivers who are distracted, sleepy, chatting on cell phones, or applying makeup, one would guess. In the case of elderly drivers, or nighttime drivers, or drivers of sports utility vehicles, the risk comparability with drunk driving can be demonstrated with hard statistics. Elderly drivers have an accident record ten times higher than forty-year-olds; nighttime driving is four times more likely to kill you than daylight driving; and in a fatal collision between a sports utility vehicle and an ordinary car, the driver of the latter is four times more likely to perish than the driver of the former. (Beware the geezer barreling down the road in a Range Rover at midnight.)

Then there are diabetic drivers with severe hypoglycemia, who are more than four times as likely to get in a crash as drunk drivers. Ban them from the road, you say? Then what about healthy eighteen-year-old males, whose accident rate is forty times worse than middle-aged women?

Clearly, there are many "impaired" drivers on the highways who do not fall under the dread DUI rubric. (Anyone for Mothers Against Distracted/ Discretionary/Dormious/

Diabetic/Dotard Drivers?) Yet they are punished only if they commit a tangible traffic violation--like crossing over the center line, say, or doing ninety under the Pont d'Alma. Drunk drivers are punished for the mere fact of their impairment--and punished severely.

The criminalization of drunk driving is less of a deterrent than one might expect. In those states that have passed laws mandating jail time for first-time DUI offenders, drunk-driving fatalities have declined by all of 2 percent. But the most compelling critique of drunk-driving laws comes from the Desert Theory of Punishment, which holds, plausibly enough, that an offender should be punished in proportion to the seriousness of the offense.

Douglas Husak, a Rutgers philosopher who also teaches in the law school there, made the case with impressive rigor a few years ago in the journal Philosophy and Public Affairs. The seriousness of a crime is a function of the harm it does

and the culpability of the offender. But, as Husak observed, the gravity of the harms inflicted by drunk drivers is the same as those inflicted by sober drivers who cause an accident. Furthermore, singling out drunk driving on the grounds of riskiness doesn't work, since other kinds of conduct that create comparable risks--driving while sleepy or distracted, driving tout court--aren't offenses at all. As for culpability, even sober drivers should be aware of the grave dangers they pose, often for trivial reasons. (In Japan, it used to be considered a valid defense against charges of hazardous driving to argue that you were drunk at the time and therefore not responsible for your actions.)

"What I am really worried about is the overuse of criminal sanctions in this country," Husak told me. "People are always wringing their hands about the tremendous growth in America's prison population. As a philosopher, it seemed to me useful to try and identify where criminal sanctions might not be appropriate."

Husak is currently using the Desert Theory of Punishment to appraise U.S. drug laws. Meanwhile, those who are moved by his logic might consider joining the organization DAMM--Drunks Against Mad Mothers. Yes, it really exists.

JIM HOLT