An assay technique for measuring anticholinergic drugs in human serum based upon their inhibition of the specific binding of [3H]-quinuclidinyl benzilate to rat brain muscarinic receptors is described. The assay was validated by demonstrating a close correlation (r = 0.99) between serum levels of nortriptyline measured by the radioenzymatic assay and a GLC technique. The assay measures free anticholinergics, and under standard assay conditions, approximately 95% of benztropine is bound to serum protein. Marked variation in serum anticholinergic levels in patients receiving the same oral dose was observed, and in individual patients there was a non-linear relationship between increasing oral dose and serum anticholinergic levels. In a cross-sectional study of 109 patients receiving concurrently neuroleptics and anticholinergics, there was no correlation (r = 0.029) between serum neuroleptic levels measured by radioreceptor assay and extrapyramidal side effects (EPS). In the patients whose serum anticholinergic levels were also determined, there was a significant inverse correlation (r = 0.44) between anticholinergic levels of EPS. In this cohort of patients, there was no significant correlation between serum anticholinergic and serum neuroleptic levels (r = 0.16) and the ratio of serum anticholinergic to serum neuroleptic was a poor predictor of EPS (r = 0.26). The results suggest a marked variation in sensitivity of patients to the EPS-inducing of neuroleptics; nevertheless, the incidence of EPS decreases with increasing serum levels of anticholinergics. An optimal serum anticholinergic level of 10 pmole atropine equivalent per ml was associated with a low incidence of EPS and is relevant to drug action at the striatal muscarinic receptor.