The purpose of this study was to construct and evaluate a statistical model for the quantitative analysis of computed tomographic brain images. Data were derived from standard sections in 34 normal studies. A model representing the intracranial pure tissue and partial volume areas, with allowance for beam hardening, was developed. The average percentage error in estimation of areas, derived from phantom tests using the model, was 28.47%. We conclude that our model is not sufficiently accurate to be of clinical use, even though allowance was made for partial volume and beam hardening effects.