Today marks ten years to the day since the world discovered that Madeleine McCann was not a missing person, but a dead one.
On August the 1st 2007, two British sniffer dogs were brought to the resort of Praia del Luz in Portugal to aid in the search for the missing child. These sniffer dogs had been trained to have rather different skills. One had been trained to give an alert upon sniffing out human blood. The other had been trained to give an alert upon sniffing out the peculiar odour released by dead bodies.
The corpse sniffing dog was taken to several locations around Praia del Luz that were relevant to the case. Upon arriving at the McCann’s apartment (from which Maddy’s parents charge she was abducted in her sleep), the corpse sniffing dog began to bark. Two areas were noted by the dog as smelling of death: behind the sofa in the living room and on the veranda in the parents’ bedroom.
Later that afternoon, the blood sniffing dog was taken to the same locations in Praia del Luz. The blood sniffing dog, independently of the corpse sniffing dog, also began to bark inside the McCann’s apartment and, specifically, behind the sofa in the living room and on the veranda in the parents’ bedroom.
The dogs were then, separately, taken to a carpark in which Maddy’s parents hire car was parked among many others. Again, both dogs alerted to the smell of dead bodies and blood only at the McCann’s hire car: specifically, in the boot.
What is the probability that the corpse sniffing dog alone could have been mistaken? Perhaps the probability of error is quite high. Maybe corpse sniffing dogs are routinely unreliable. Let us assume that the corpse sniffing dog had only a 10% chance of giving the right answer to the question “which car once had a dead body in it?” Let’s say that the same goes for the blood sniffing dog. Both dogs are very likely to give the wrong answers to the questions we ask them.
Now, what is the probability that the two dogs would agree to the presence of blood and death in just those very locations that incriminate the McCann’s? In other words, what is the probability that the two dogs would not only get the wrong answer, but get the very same wrong answer? The answer to this question is, I hazard a guess, far lower than 10%. The dogs were used independently and they have been trained to detect different things. It is more likely that the dogs’ agreement has a common cause: the dead and bloody body of Maddy McCann.
We might think of the following analogy, which I borrow from the philosopher of science, Hans Reichenbach. You are reading in the lounge, when suddenly all the lights in the room go out at once. We could, of course, hypothesize that each bulb has blown independently; that the identical timing is nothing more than coincidence. But in practice, we take this to be good reason to suspect that a fuse has blown, or that there has been a scheduled or unscheduled power outage. Of course, it’s not impossible that all the bulbs could blow at the same time, merely improbable to the point of absurdity.
Ten years on, the media persists with the fairy tale that an intruder abducted Maddy McCann and that her parents are innocent victims. Cadaver dogs are unreliable, we are told. Blood sniffing dogs are unreliable, we are told. Of course, perhaps this is true. But when two unreliable, but distinct, processes converge on the same answer to the same question, this is very good evidence that the answer is correct.