The conductivity of water having parts per billion concentrations of oxygen, hydrogen, and bicarbonate was measured while the water was irradiated by a low-pressure mercury vapor lamp, which was turned on and off periodically. A cell normally used for measurement of dissolved oxidizable carbon was modified for use in these measurements. When the lamp is turned on, the conductivity increases (sometimes decreases) with a time constant of about 50 ms; when the lamp is turned off, the conductivity changes in the opposite direction with a time constant of about 275 ms, but does not return tb its value before the lamp is turned on. The lamp step (difference between conductivity with lamp on and conductivity with lamp off) depends on the intensity of radiation and on the concentrations of oxygen, hydrogen, and bicarbonate. It is negative when [O-2] is less than approximate to 10(-10) M and positive for higher [O-2]. increasing to a maximum at [O-2] approximate to 10(-7) M. The presence of dissolved H-2 increases the lamp step. The lamp step increases in magnitude when the lamp intensity increases, without being proportional to intensity. Experiments were performed that show that the reactions responsible for the changes in conductivity occur in bulk solution and not at the cell electrodes. A theoretical model to explain the changes in conductivity was developed. It assumes that the absorption of a photon of ultraviolet radiation converts one molecule of water to a hydrogen and a hydroxyl radical (H . and . OH), and that these react with H+, OH-, and other dissolved species. Some thirty bimolecular reactions are considered, with rate constants taken from the literature. The differential equations giving the changes in the concentrations of twelve species are solved numerically. The rate of generation of H . and . OH is varied with time to represent the turning on and off of the ultraviolet lamp. From the species concentrations, the conductivity is calculated as a function of time, yielding calculated lamp steps in general agreement with our experimental results. The species responsible for the lamp steps can then be identified, and the important reactions elucidated. The conductivity is always dominated by the contribution of H+. It is shown that a substantial negative lamp step, found for very low oxygen concentrations, cannot occur in completely pure water. Dissolved carbon that has been oxidized to bicarbonate must be present. Hydroxyl radicals produced by irradiation react with HCO3- to give the carbonate radical anion, C .O-3(-). Because the pK of the parent acid HC .O-3 is substantially larger than that of H2CO3, formation of C .O-3(-) leads to a decrease in [H+] and hence a decrease in conductivity. If dissolved oxygen is present, it may be converted by H . to perhydroxyl radical H .O-2, which dissociates to H+ and superoxide anion .O-2(-). raising the conductivity. Furthermore, superoxide can reduce HC .O-3 back to HCO3-, countering the conductivity-lowering effect of bicarbonate. Because superoxide is destroyed mainly by reaction with perhydroxyl radical, and the concentration of perhydroxyl is much smaller than that of superoxide, superoxide is a long-lived species. Thus the conductivity after the lamp is turned on and then off is larger than the conductivity before the sequence. If hydrogen is present in addition to oxygen, it reacts with . OH to generate .H, which leads to the formation of more H .O-2. In addition, the reaction of . OH with H .O-2, which would convert the latter back to O-2, is prevented. For both reasons, hydrogen makes the conductivity step larger, as observed. The concentration of superoxide is limited because high [O-2(-)] leads to high [H .O-2], so the reaction of O-2(-) With H .O-2, which destroys O-2(-), becomes important. The experimental observation that the conductivity step goes through a maximum as a function of O-2 concentration is not explained by our model, but is believed to be associated with absorption of ultraviolet radiation by superoxide, H2O2, or other species formed from O-2.