The present study investigated cross-modal correspondences between Cantonese tones and two dimensions: (1) spatial motion and (2) emotional valence, via a forced-choice mapping task on Hong Kong native speakers. Results show that the two contour tones (rising and falling contours) could be reliably matched with motions (upward and downward) that are congruent with pitch trajectories; and in contrast, the correspondence between contour tones and emotion valence (rising tone is positive and falling tone is negative) was less robust and limited for selective vowels. In summary, our findings indicate that beyond arbitrary form-meaning correspondence, vertical spatial information, both concrete (motion) and abstract (valence), are also encoded in lexical tones of Hong Kong Cantonese, but that the relative strengths of the two types of correspondences were not equivalent.