## Info

Values represent the proportion of trials in which the dogs made correct choices (the upper-left and lower-right cells), in which they failed to identify a guilty suspect (false negatives in the lower-left cell), and in which they incorrectly identified an innocent suspect (false positives in the upper-right cell).

Values represent the proportion of trials in which the dogs made correct choices (the upper-left and lower-right cells), in which they failed to identify a guilty suspect (false negatives in the lower-left cell), and in which they incorrectly identified an innocent suspect (false positives in the upper-right cell).

there was only one murderer). These calculations are shown in Table 3.4. The bottom row shows the 20 total suspects divided into two groups, one guilty person and 19 innocent ones. The first column of Table 3.3 remains the same in Table 3.4 because we assume there is only one guilty person. But the values in the second column of Table 3.3 are increased in Table 3.4 because there are 19 innocent suspect s who might be falsely identified by the dog. Specifically, the values in the second column of Table 3.3 are multiplied by 19 to get the second column of Table 3.4. This implies that if the dog selects a suspect (the top row of Table 3.4), the chance that the suspect is guilty is only 27% (0.36/1.21 = 0.27). Even though the probability of a false positive (0.05) is much lower than the probability of correctly picking the guilty person (0.36), there are so many more opportunities for the dog to select an innocent suspect than the guilty person (19 versus 1) that the likelihood that it will pick an innocent person is 73%.

Most criminal cases in which a scent lineup might be used aren't as straightforward as this because the total number of possible suspects isn't known. However, there is often other evidence that links a particular suspect to a crime. At least in theory, this other evidence can be used to estimate the likelihood that the suspect is guilty, independently of whether a dog selects the suspect's scent from a lineup. This is called the prior probability of guilt; that is, the probability of guilt before taking into account the result s of the scent lineup. In the example of the English murder mystery, the prior probability of guilt is 1/20, or 5%, for each suspect because there are exactly 20 suspects and we have no other evidence pointing toward any one of the 20. This prior probability is represented in the last row of Table 3.4. In the scent lineup, the constable's dog matches the scent of one of the suspects to the odor on the handkerchief left by the v ictim's body. Because trained dogs really do have some ability to identify individual humans by smell, this increases the likelihood that the suspect identified by the dog is guilty. After the lineup, the so-

Table 3.4. An application of Schoon's (1998) results on accuracy of scent identification by dogs.
0 0