We used several approaches to assess the reliability and sensitivity of computer-assisted densitometry to detect regional changes in tissue antigen content as a function of immunohistochemical staining density. We designed a model system to mimic variations in antigen concentration in postfixed, slide-mounted rat brain sections by varying the ratios of conjugated (biotinylated) to unconjugated secondary antibody. Antigen concentration was also varied in tissue discs made from mixing rat brain homogenate with increasing amounts of tissue embedding compound. The monoclonal antibody bd-17 to the beta2/3 subunit of the GABAA receptor was used as the primary antibody. Immunostaining density was visualized with diaminobenzidine (DAB). There was a significant, positive linear relationship (r = 0.97-0.99) between immunostaining intensity and antigen concentration. With this approach, changes in antigen content of less than 10%, as reflected in immunostaining intensity, were detectable in brain sections. The low degree of variability in measures of regional variation in immunostaining in sections from naive rats (n = 7) suggested that the method was suitable for quantitative analysis and indicated the reliability of the method. This systematic study of the utility of computer-assisted image analysis for semiquantitative immunohistochemical analysis found the method to be both reliable and sensitive.