The theory of neurovascular compression has been tested by comparing the neurovascular relationships of the trigeminal nerve in a series of operative observations in patients affected by trigeminal neuralgia with those of a control series of cadavers matched for age, sex and side, in which operative conditions were simulated during simultaneous arterial and venous injection--filling to physiological pressures, as described in Part 1 of this article. A rigorous system of classification of neurovascular relations is defined. In 46 patients with trigeminal neuralgia, 91% had a vessel in contact with the trigeminal nerve adjacent to the brain stem and in all but one a groove was created. Multiple vessels were found in 17% and in two both the root entry zone and lateral portions of the nerve were compressed. However, in 35 randomly selected fresh cadavers, not known to have suffered neurological disease, 14% had neurovascular contact and a further 26% had vessels "near" to the nerve. No vessel was associated with a groove and no multiple vessels, or sites of contact, were encountered. The difference between the control cadavers and the operative findings in patients related to an increase in the number of arteries. Injection-filling of the cadaveric vessels doubled the numbers of vessels in contact with, and near to the nerve. The technique used and system of classification applied showed an association between arterial contact and trigeminal neuralgia. The technique may provide a suitable method for the testing of the neurovascular compression theory in other conditions.