Methodological problems in the measurement of drug-induced rotational behaviour: continuous recording reveals time-course differences undetected by previous techniques.