A battery of clinical colour vision tests was given to a group of 100 observers with abnormal colour vision who were also tested on the Farnsworth lantern and the Holmes-Wright lanterns types A and B. It was found that clinical colour vision tests are imperfect predictors of lantern test performance. However, observers classified as having a 'severe' colour vision defect were found to fail the lantern tests but only one half to two-thirds of those who fail the lantern tests can be identified in this way. It is not possible to identify with certainty any of the people likely to pass the lantern tests: about one-third to two-thirds of observers classified as being mildly affected fail the lantern tests. The Farnsworth D-15 and City University tests were found to be the best predictors of lantern test performance but other tests such as the Nagel anomaloscope, the H-16, L'Anthony's desaturated test can also be used. The lack of a strong correlation between clinical tests and the recognition of the small coloured stimuli presented by the lantern tests suggests that clinical tests do not test the same aspect of colour vision that is important to the recognition of signal lights. For this reason lantern tests should be retained for occupational testing of colour vision.