Mass bias occurring during analysis of the light stable isotopes of oxygen, carbon, and sulfur in geological materials by secondary ionization mass spectrometry has been investigated. The effects of instrumental parameters (primary ion beam, secondary ion energy, and polarity) were evaluated by measuring sulfur isotope ratios in conductive sulfide minerals. The role of analyte chemical composition (matrix effect) on mass bias was investigated in sulfides (sulfur), silicates and oxides (oxygen), and carbonates (oxygen and carbon). For oxygen and carbon, various correlations between mass bias and matrix parameters have been identified. The application of several empirical models for prediction of oxygen isotopic mass bias indicates that for silicates, depending on mineral composition, bias can be predicted with an accuracy that is typically within two times that of the precision. However, extension of these models to other matrices has proved problematic, indicating that additional factors are important. (Int J Mass Spectrom 178 (1998) 81-112) (C) 1998 Elsevier Science B.V.