Experimental modal analysis is highly dependent on the quality of the frequency response functions (FRFs) used to extract the mode shapes and other modal parameters. Therefore, the key to success in experimental modal analysis is to obtain FRFs which contain accurate and reliable information;In spite of manufacturers' efforts to minimize the cross-axis sensitivity of their transducers, most transducers used to measure the response of the structure have a primary sensing axis and a perpendicular plane which contains a direction of maximum cross-axis sensitivity. Due to the cross-axis sensitivity, the measured signals are contaminated and this contamination can lead to serious errors in the measured FRFs as well as the resulting modal analysis. It was found that the contamination is not random in nature but is systematically accumulated in the measured FRF, and that the amount of cross-axis measurement error depends on the components of the motion to be measured relative to the primary sensing direction as well as the cross-axis plane;This cross-axis sensitivity error can be compensated for in either the time domain or the frequency domain. The method employed is dependent on the data acquisition and data processing system being used. Accurate calibration for the cross-axis sensitivities should be done using the same compensation process. Both compensated and uncompensated frequency response functions are used to show the effect of cross-axis sensitivity on the modal analysis results obtained for a simple structure.