Clinical assessment of rigidity in parkinsonian patients is largely qualitative. The reliability and validity of the assessments are sometimes in doubt. Several "engineering" methods of quantifying rigidity have been described, but none has been adopted into general clinical practice. A possible reason is that these methods differ in crucial aspects from the clinical exam. We therefore tackled the problem by monitoring the clinical exam itself, using small sensors to measure the forces and displacements applied. Limb impedance (Z) was computed using parameter identification methods and compared to raters' verbalized ratings of rigidity based on a 5-point scale: the Unified Parkinson's Disease Rating System. The qualitative and quantitative estimates of impedance covaried over a fourfold range, depending on the forces imposed and the subject's motor set. Raters differed by up to 1 full point in their mean qualitative ratings and sometimes disagreed on whether levodopa reduced rigidity. This was not due to any significant differences in the overall range of rigidity they evoked, but rather to the way they scored this range [the ratio of mean rating to mean impedance (R/Z) varied between raters and subjects]. On the other hand, the R/Z ratio was reproducible over separate sets of ratings and may therefore serve to convert measured impedance into a standardized rating. Our results indicate that the current clinical exam may be too abbreviated to detect the sometimes quite small reductions in rigidity after levodopa. We conclude that a device that conveniently quantifies the clinical assessment of rigidity is now available and will lead to more standardized protocols for rating rigidity in the near future.