In the early 1900s, the German public was fascinated by a mathematical Mr. Ed named Clever Hans, an Orlov Trotter horse that seemed to be capable of counting, doing basic arithmetic, and even solving elementary word problems—which, lacking the dexterity to grasp a number two pencil, it would answer by stamping its hooves. Eventually, of course, it was proven that Hans was doing nothing of the sort: the horse was perceptive rather than clever, and had been picking up on subtle, subconscious cues from his handler that let him know when to begin stamping and when (having arrived at the correct answer) he should stop.

A century later, academic researchers have shown that even well-trained drug-sniffing dogs are subject to the “Clever Hans Effect,” often alerting to non-existent drugs or explosives in locations where their human handlers have been falsely told they were present. Nor are those findings strictly academic. A recent analysis by reporters at the Chicago Tribune found that field records showed that drug-sniffing dogs produced a disturbingly high level of false positives: in only 44 percent of cases where dogs alerted did a subsequent search turn up contraband. Their success rate was even lower when it came to certain minorities: when dogs alerted on a Hispanic driver, only 27 percent of ensuing searches found any drugs, suggesting that the pooches may be picking up on their handlers’ subconscious bias, effectively legitimizing a form of racial profiling.

All this should make the Supreme Court’s unanimous decision today in Florida v. Harris disappointing to anyone who cares about the Fourth Amendment right to be free of unreasonable searches and seizures. Overturning a ruling by the Florida Supreme Court, the decision holds that a well-trained drug dog’s alert during a traffic stop generally provides probable cause for a warrantless search of the vehicle—even though, as in this case, the dog repeatedly alerted at a car that turned out not to contain any of the chemicals it had been trained to detect. Urging the need for a “flexible” standard, the Court saw no need for police to maintain or provide any record of a dog’s reliability in the field—such as a count of false positives—and even suggested that apparent “false positives” might not be errors at all, since a dog might be picking up “residual odors” from drugs that had previously been in contact with the vehicle. Even if that’s true, however, it’s not clear why it cuts in the government’s direction here: if the dogs are that sensitive, it seems like an additional reason to doubt that an alert provides probable cause to believe contraband is currently present.

The bigger problem, however, is that the controlled conditions in which dogs are evaluated don’t typically match field conditions very well: the handlers there often know where on the test course drugs are located—and even when they don’t, have no incentive to want the dog to alert at any particular location, which removes those subconscious signals from the equation. Bizarrely, the Court nevertheless held that the “better measure of a dog’s reliability… comes away from the field, in controlled testing environments.” Worse, the opinion also provides police deparments with an added perverse incentive to avoid collecting data on the real-world reliability of their sniffers: while a dog’s alert provides prima facie probable cause for a search, the Court held, defendants must be given an opportunity to challenge the reliability of a particular search in court—with field performance as one potential grounds for challenge. But, of course, if that’s the case, keeping records of false positives can only serve to give defendants grounds to invalidate a search that would otherwise be presumed valid. In effect, then, the Court has handed police what may well be a blank check for pretextual searches, while discouraging the collection of data that might prove that’s what they’ve done.