One of the unresolved questions in multisensory research is that of automaticity of consistent associations between sensory features from different modalities (e.g. high visual locations associated with high sound pitch). We addressed this issue by examining a possible role of selective attention in the audiovisual correspondence effect. We orthogonally manipulated loudness and pitch, directing participants’ attention to the auditory modality only and using pitch and loudness identification tasks. Visual stimuli in high, low or central spatial locations appeared simultaneously with the sounds. If the correspondence effect is automatic, it should not be affected by task changes. The results, however, demonstrated a cross-modal pitch-verticality correspondence effect only when participants’ attention was directed to pitch, but not to loudness identification task; moreover, the effect was present only in the upper location. The findings underscore the involvement of selective attention in cross-modal associations and support a top-down account of audiovisual correspondence effects.