The partial volume effect (PVE) on the arterial input function (AIF) remains a major obstacle to absolute quantification of cerebral blood flow (CBF) using MRI. This study evaluates the validity and performance of a commonly used multiplicative rescaling of the AIF to correct for the PVE. In a group of six patients, perfusion imaging was performed using a T(1)-weighted approach that minimizes confounding susceptibility artifacts. Various degrees of PVE were induced on the AIF and subsequently corrected using four different schemes of multiplicative AIF rescaling. Our results show that a multiplicative rescaling is not always applicable and can introduce a CBF bias. An easily measurable quantity denoted the tissue signal fraction (TSF) is proposed as a measure of the applicability of a multiplicative rescaling. For the present CBF quantification method, a TSF of <0.4 results in a CBF bias <15% after AIF rescaling.