What has empirical research shown regarding the reliability of police lineups?
Empirical research has found that the traditional method of doing a police lineup is accurate more often than not. It has additionally found that traditional police lineups do not typically result in the wrong person being identified as the culprit.
For example, in the New York Times link below, we see that the correct person was identified by witnesses 60% of the time when using the traditional police lineup. Just as importantly, a wrong identification was made only 3% of the time.
Of course, different results can be obtained if the test is done in different ways. However, empirical evidence suggests that the right person is identified most of the time and that the wrong person is rarely identified. (Though, of course, that would be of little consolation to the 3% who were wrongly identified.)