Interferences of the matrix elements of glass ceramics (Al, Mg, Na and Si) on the titanium signal obtained by atomic absorption spectrometry with a nitrous oxide/acetylene flame were studied by means of experimental design. Quadratic polynomials were chosen as the model; full factorial designs with two, three and four variables at three levels were applied. As expected, aluminium increased the titanium signal, while magnesium reduced it. All the investigated elements interfered nonadditively with the titanium signal; the standard addition method therefore does not provide accurate results. Graphic evaluation of the empirical response surfaces was used to establish optimum conditions for titanium; these surfaces were compared with the polynomial surfaces to check the models. The results obtained on interactions in the system are used with some thermodynamic data to estimate the nature of the compounds formed in the flame. The strong interferences on the titanium signal requires fairly close matrix matching between the standard and sample solutions. The proposed method allows the determination of 3-6% Ti in glass ceramics with a relative standard deviation of 1%.