Arguments by analogy, although frequently persuasive, are among the least defensible forms of scientific reasoning. As elegantly stated in a 1912 logic textbook (1):

Applied to qualitative relations, Analogy belongs more to Rhetoric than to Logic. It gives to a statement emphasis, impressiveness, picturesqueness, and often intelligibility, that it would not otherwise have, and that can be given in no other way. Hence it is especially frequent and valuable in poetry, for it adds to mere assertion a powerful reinforcement.

The fact that two phenomena have similar characteristics permits no logical inference that one causes the other, that they have a common cause, or that they are both part of a larger construct (2). One of the major problems with analogies is that they are invariably selective. Consciously or unconsciously, there is always a temptation to underscore semblances that support one's conclusions and to ignore or gloss over those that could support alternative or competing conclusions (3).

Drug policy in the United States is typically framed in terms of competing analogies. Advocates for a public health perspective on the drug problem frequently argue that drug abuse or dependence is a "disease" akin to chronic relapsing medical conditions, such as asthma, hypertension, or diabetes. If drug abuse behaves like a chronic medical condition, so the argument goes, then drug abusers ought to be treated similarly to medical patients. Proponents of a public safety perspective take the contrary position that drug abuse or dependence is like various forms of reckless misconduct. If drug abusers cause as much foreseeable damage to themselves and to society as, say, drag racers, then they ought to be treated like criminals.

The public health perspective recently gained scholarly support from a synthesis of the research literature conducted by McLellan and associates (4), which compared genetic heritability estimates, relapse rates, and treatment noncompliance rates for drug dependence with those for asthma, hypertension, and diabetes. It was concluded from that review that studies of monozygotic and dizygotic twins reveal remarkably consistent heritability estimates for all these disorders, typically hovering around .50. Heritability estimates generally range from .25 to .50 for hypertension, from .30 to .55 for type 1 diabetes mellitus, and from .36 to .70 for adult-onset asthma. They similarly range from .34 for heroin dependence to .55 for alcohol dependence to .61 for nicotine dependence.

Treatment noncompliance rates and relapse rates for these disorders similarly hover around 50 percent. Approximately 40 to 60 percent of adult patients with type 1 diabetes, hypertension, or asthma do not adhere to their medication regimens, and about 70 percent do not adhere to recommended dietary changes and behavioral changes for managing their disorder. Similarly, an average of 40 to 60 percent of drug abuse clients drop out of drug abuse treatment before receiving a minimally adequate dosage of services. Finally, approximately 30 to 50 percent of adults with type 1 diabetes, 50 to 70 percent of adults with hypertension or asthma, and 40 to 60 percent of adults with drug dependence experience a yearly recurrence of symptoms that requires professional intervention.

McLellan and colleagues (4) readily acknowledged the limitations of arguments by analogy and restricted the policy implications they drew from these similarities to recommendations for insurance parity legislation. Nevertheless, the implications for criminal justice policy are unmistakable. If drug abuse or dependence is a "disease," then, arguably, there can be little basis for criminalizing the conduct or for expecting incarceration to serve as a cure. Indeed, this line of reasoning is behind several statewide policy initiatives that have effectively decriminalized drug possession and drug intoxication. For instance, the preamble to Arizona's Proposition 200, a law that diverts most nonviolent drug offenders into treatment in lieu of prosecution or incarceration, states that "we need to medicalize Arizona's drug control policy recognizing that drug abuse is a public health problem and treating drug abuse as a disease" (5).

The problem with this analogistic line of reasoning is that the similarities apply with equal force to criminal and antisocial conduct. Virtually identical genetic heritability estimates and relapse rates are reported in the literature for criminal and aggressive behavior, conduct disorder, and antisocial personality disorder. The issue of treatment noncompliance makes lesser sense in this context, because few adult clients receive treatment targeted specifically at criminal or antisocial behavior. Heritability estimates for criminal and aggressive behavior generally range from .30 to .70 (6,7,8,9), with a mean of .50 derived from a meta-analysis of more than 100 studies (10). Similarly, heritability estimates for conduct disorder symptoms range from .40 to .70, with most studies reporting estimates of around .50 (11,12,13,14). Heritability estimates for antisocial personality disorder symptoms range from .30 to .70 (11,12,15). Finally, relapse (or recidivism) rates for criminal and aggressive behavior generally hover around 50 percent (16,17), which is virtually identical to the relapse rates reported for type 1 diabetes, asthma, hypertension, and drug dependence.

Importantly, we are not advocating that drug abuse or dependence be viewed as akin to criminal or antisocial conduct. On the contrary: it is our position that such analogies are distracting and have little probative value. The fact is that neither the disease analogy nor the antisocial conduct analogy has contributed to effective policy initiatives for drug abusers or drug offenders (18). Without intensive criminal justice supervision and immediate and consistent consequences for noncompliance, drug abusers ordinarily drop out of traditional treatment programs long before one could reasonably expect to see minimally beneficial effects. Moreover, traditional correctional interventions, such as imprisonment, probation, boot camps, and house arrest, have had virtually no influence on drug use or criminal activity (19).

The Latin term explanandum has been used to describe would-be explanations that appear on the surface to convey meaningful information but that actually mask tautological reasoning or merely restate the problem. If by "chronic relapsing condition" one means that the condition has no reliable cure, rarely resolves on its own, and follows an irregular course, then asthma, hypertension, diabetes, drug dependence, and antisociality are all chronic relapsing conditions. Science does not progress by semantics or by selective comparisons. It progresses by rigorously testing falsifiable hypotheses. Once science uncovers demonstrably effective interventions for these conditions, the competition of analogies will become moot.

Dr. Marlowe is director and Dr. DeMatteo is a scientist in the section on law and ethics research at the Treatment Research Institute at the University of Pennsylvania, 600 Public Ledger Building, 150 South Independence Mall West, Philadelphia, Pennsylvania 19106-3475 (e-mail, marlowe@tresearch.org). Sally L. Satel, M.D., is editor of this column.