The objective of this study has been to investigate the origin of the water vapor effect on the chain length (CL) of peroxy radical chemical amplifiers (PERCA). Results of the investigation of the water interference in the determination of peroxy radicals by using the PERCA technique are presented. The experimental conditions have been analyzed and modeled. A nonlinear dependence of the CL on the relative humidity (RH) has been accurately determined. The combined analysis of experimental and simulated results rules out wall loses as a single explanation of the CL variation observed and indicates three reactions, which possibly account for this water effect: (1) OH + CO nH(2)O--- nonradical products, (2) CO + HO2 nH(2)O--- CO2 + OH, and (3) NO + HO2 nH(2)O--- nonradical products (e.g., HNO3). Assuming a mechanism involving the formation of a HO2-nH(2)O complex, the corresponding rate coefficients and their water dependence have been estimated. The quadratic dependence of these rate coefficients upon the RH implies the participation of two H2O molecules in the proposed reactions. The study has shown that at a RH of 80% an effective second-order rate coefficient of 10(-15) cm(3) molecule(-1) s(-1) for the reaction of CO with HO2, or 10(-13) cm(3) molecule(-1) s(-1) for the reaction of HO2 with NO, explains the observed behavior. Both these complex reactions have potential significance for the chemistry of the marine boundary layer (MBL) and their atmospheric implications are discussed.