Two experiments investigated the influence of the harmonic content of background music on taste perception.
The participants evaluated samples of mixed fruit juice while listening to soundtracks that had either been harmonized with consonant or dissonant musical intervals. Each sample of juice was rated on three computer-based scales: One scale was anchored with the words sour and sweet, while the other two scales involved hedonic ratings of the music and of the juice.
Participants reliably associated the consonant soundtracks with sweetness and the dissonant soundtracks with sourness, rating the juices as tasting significantly sweeter in the consonant than in the dissonant music condition, irrespective of the melody or instrumentation involved. These results provide empirical support for the claim that the crossmodal correspondence between basic taste and a higher-level musical attribute (harmony in this case) can be used to modify the evaluation of the taste of a drink.
This according to “Striking a sour note: Assessing the influence of consonant and dissonant music on taste perception” by Charles Spence (above) and Qian Janice Wang (Multisensory research XXIX/1–3  pp. 195–208).
Another post about Professor Spence’s research is here. Below, some consonance and dissonance imaging.