Introduction: Quantification of cerebral blood flow (CBF) using magnetic resonance imaging still suffers from many unresolved methodological issues. In this study we report the successful modeling of monkey CBF data, using the two-compartmental model introduced by Parkes et al. [1]. Absolute flow and transit times were derived including uncertainties of the assumed parameters as well as the signal noise. The precision of the models result was investigated and an acquisition paradigm to maximize the information content is proposed. Subjects and Methods: CBF measurements in the anesthetized macaca mulatta monkey were obtained on a 7T Bruker system with a continuous arterial spin labeling (CASL) sequence using a three-coil approach. Single-shot GE-EPI images were acquired using the following parameters: TR/TE=5300ms/14ms, resolution=0.75x0.9x2mm3, label time=3.5s. Measurements were conducted in the visual cortex under both baseline and elevated blood flow conditions. A temporally stable increase in CBF was achieved by i.v. injection of Diamox. Data analysis was undertaken using Matlab and PMOD (a kinetic modeling software, http: www.pmod.com). Results: Region of interest (ROI) analysis was performed, investigating perfusion of gray matter under both flow conditions. The average fit values were f=109.4ml/min/100ml, tA=0.68s for baseline and f=251.7ml/min /100ml, tA=0.51s for Diamox (n=1). Fig.1 depicts typical data. Monte Carlo simulations yielded coefficients of variation (COVs) of 11.3% and 10.1% in flow and 35.6% and 35.5% in tA for baseline and high flow conditions respectively (n=2). For data acquired during the decay phase of the signal only, the COVs proved to be almost independent of the number of data points acquired. Our results show that for flow-quantification, the chief information content of such data lies in the amplitude of the peak of the signal curve. The amount of information per data point can be increased drastically by sampling all phases of the signal curve as shown in Fig.2. Data acquired in t