Taste-odor interactions are usually studied by asking assessors to estimate the taste intensity of tastants alone or in mixture with odor compounds. We propose another method to study taste-odor interactions based on the selective attention paradigm proposed by Garner (1974). This paradigm is based on the following principle: If two dimensions are separable (i.e., no interaction occurs), the performance at a categorization task based on one dimension is not affected by variations on the other dimension. Two pairs of perceptual dimensions were studied with stimuli varying in: sucrose/vanillin concentrations and citric acid/lemon aroma concentrations. For both pairs of dimensions, results showed that performance at the categorization task on one dimension were consistently lower when the level of the irrelevant dimension varied than when it remained constant. This gave evidence of an interaction for these two pairs of dimensions. This experiment shows the potential of such a paradigm for taste-odor interactions studies.