Levels of glutathione were measured for different cell types in roots of intact Arabidopsis seedlings after labelling with monochlorobimane to give fluorescent glutathione S-bimane (GSB) and imaging using confocal laser scanning microscopy with excitation at 442 nm. Labelling increased to a plateau in most cell types after about 15–20 min and the GSB accumulated rapidly in the vacuole. Formation of GSB in the cytoplasm was not affected by treatment with sodium azide; however, vacuolar transport of GSB was substantially inhibited under these conditions. We infer that vacuolar sequestration was mediated by a tonoplast glutathione S-conjugate pump. Quantitative estimates of the cytoplasmic glutathione concentration involved correction for the loss in fluorescence signal with depth into the specimen using an empirically determined model derived in situ from a permeabilized root. Correction for the dilution experienced on transport into the vacuole also required an estimate of the amount of cytoplasm present in each cell type. This was achieved in two stages: first, the levels of protein were mapped after fixation, permeabilization and labelling with fluroescein isothiocyanate. Second, the corresponding cytoplasmic volume was determined as 40% for epidermal cells in the elongation zone by manual segmentation of the cytoplasm in serial optical sections. Values of relative cytoplasmic volume for other cells were extrapolated in proportion to their protein content. Using this approach, cytoplasmic glutathione concentrations were found to be 2–3 mm in most cell types. There was a marked difference between the central cells and the neighbouring, rapidly dividing initials, and between the columella cells and the outermost cells of the root cap. In the latter case, the difference was equalized in the presence of azide. This might indicate that additional cell–cell movement and preferential sequestration of GSB can occur during the detoxification process in an intact system.