Four methods used for the neurohistochemical demonstration of horseradish peroxidase (HRP) were quantitatively compared by counting retrogradely labeled neurons found after each method was used. HRP used as a retrograde marker is an important neuroanatomical tracing method, and maximum sensitivity in its demonstration of retrogradely, labeled neurons is important if these neuroanatomical studies are to completely demonstrate afferent neurons. The four methods compared were a diaminobenzidine (DAB) procedure, a Hanker-Yates procedure using P-phenylenediamine and pyrocatechol, an o-dianisidine procedure, and a tetramethyl benzidine (TMB) procedure. The TMB procedure resulted in a more complete topography of neurons afferent to the HRP application site, and demonstrated many more neurons in all afferent cell groups that either of the three other procedures. Use of the TMB method was especially critical in the cases of small HRP applications, a size useful for neuroanatomical studies, where the other methods demonstrated very few or no retrogradely labeled neurons. Neurons were judged to be retrogradely HRP labeled if they had small granules of the reaction product (the color varying with the chromogen) describing the somal shape, usually extending into the processes, and a clear nucleus. In addition, after the o-dianisidine or the TMB reaction a small number of retrogradely labeled neurons had soma and processes especially well filled with reaction product, giving the appearance of neurons from Golgi preparations. For a sensitive TMB reaction giving good results, exact H2O2 concentration, freshly prepared solutions, minimal postreaction exposure to alcohol, counterstaining, and clean glassware were each found to be important.