Hydrogen peroxide is now reported to be a normal aqueous humor component present, in the low μ m concentration range, in the animal species which have been studied. This finding was established with the exclusive use of the dichlorophenol-indophenol method of analysis. In this procedure, aqueous humor is added to a blue, oxidized dichlorophenol-indophenol solution. The 605 nm absorbance of this solution immediately decreases in response to the reducing action of ascorbate present in the sample. The extent of reoxidation of the solution upon the addition of peroxidase, as measured by the increase in its 605 nm absorbance, can be quantitatively related to the concentration of H 2O 2 in the sample. A close examination of this method revealed that reduced dichlorophenol-indophenol spontaneously reoxidizes at a rate of 0·03 nmol min −1 μ m −1, with generation of H 2O 2. H 2O 2 generation was unequivocally established by analysis of the temporal dependency of the absorbance increase produced by peroxidase in the absence of added H 2O 2 and by the sensitivity of this phenomenon to catalase. This spontaneous production of H 2O 2, on the other hand, cannot be attributed to ascorbate auto-oxidation because added ascorbate quantitatively reacts with dichlorophenol-indophenol, provided that an excess of the latter is maintained. This method then has an enormous potential to overestimate H 2O 2 in any sample. On the other hand, the response of the assay system to a given level of H 2O 2 depends on the level of reduction previously produced by ascorbate. This results in an artifactual positive correlation between ascorbate and H 2O 2 levels in samples containing variable amounts of ascorbate. In spite of these serious limitations the method can still be useful to measure H 2O 2 if appropriate precautions are taken. When using it for the analysis of rabbit aqueous humor H 2O 2 without correcting for the H 2O 2 generated during the assay and ignoring differences in the level of ascorbate in the samples, we obtained an average value of 25·3 μ m H 2O 2, which coincides with that reported in the literature for the rabbit, but is obviously incorrect. When analysing aqueous humor there was the additional variable of the aqueous humor itself inhibiting the rate of dichlorophenol-indophenol auto-oxidation and so the final, corrected figure for H 2O 2 concentration in the aqueous humor became uncertain, since the auto-oxidation of the substrate could not be properly subtracted. Nevertheless, by inhibiting dichlorophenol-indophenol oxidation with argon, the apparent concentration of H 2O 2 measured in the aqueous humor became a function of the autooxidation rate of the phenol, H 2O 2 being no longer detectable for extrapolated zero auto-oxidation rates. We conclude that rabbit aqueous humor does not contain H 2O 2 that can be detected with the dichlorophenol-indophenol reagent. H 2O 2 could still be present in this ocular fluid, however, beyond the lower limit of detection of the method, in the n m range or below.