Parkinson's disease (PD) patients show a good response to levodopa in the morning, and reduced duration or complete failure of response later in the day, but the pathophysiology of this phenomenon remains unclear. We evaluated motor performance hourly over a twelve-hour period in patients treated with levodopa/carbidopa (group A), with bromocriptine (group B), and in "de novo" patients (group C). At 8 am, 12 and 4 pm, group A patients received standard doses of levodopa/carbidopa, whereas patients of group B and C took, respectively, 5 mg bromocriptine and placebo. In "de novo" patients and in patients under bromocriptine we did not observe significant diurnal changes in motor score, whereas in patients under levodopa a progressive daytime worsening, which significantly correlated with progressive increase in 3-O-methyldopa plasma levels, was visible. These data seem to indicate a contributory role of pharmacokinetic or pharmacodynamic factors related to levodopa assumption, rather than to the underlying disease, in the afternoon worsening in PD.