In the practical application of multivariate calibration,instrument noise is one of the key factors influencing the prediction ability of partial least squares(PLS) modeling.Plenty of research works have been reported to mitigate the negative influences of instrument noise by various de noised methods.However,effects of instrument noise in PLS modeling have not received adequate attention.The main point of this work is to illustrate how instrument noise could be involved in PLS model and demonstrate its propagation effect on calculations of weight vectors and loading vectors through theoretical analysis.Moreover,the prediction error of PLS model was divided into two parts,ie.the prediction error of noise free model and the error caused by noise propagation and accumulation.Results showed that instrument noise could not only decrease the prediction ability of PLS models,but also influence the model selection.