Carbachol stimulation of rat parotid gland fragments prelabelled with myo-[3H]-inositol results in a large accumulation after 15 min of [3H]inositol trisphosphate. Only some of this is the D-1,4,5 isomer which would be expected to be derived from the known phosphatidylinositol bisphosphate. The predominant inositol trisphosphate is not susceptible to hydrolysis by human erythrocyte membranes. It yields altritol after periodate treatment followed by reduction and dephosphorylation, and, from partial dephosphorylation experiments, does not have a phosphate in the 2 position; the most likely structure of this inositol trisphosphate is therefore (D/L)-myo-inositol 1,3,4-trisphosphate. The possible origin and significance of this compound are discussed.