Hong Kong Cantonese has been described as having developed a dissimilatory merger in which O becomes A in pre-labial contexts (henceforth OM/P vs. AM/P). This change is also one that has been described as completed by the end of the 20th century. This paper presents what may be the first acoustic study addressing the Cantonese AM/P~OM/P merger. It also addresses the extent to which heritage speakers in Toronto, Canada also participate in this change. Analysis involved midpoint F1, F2, and F3 measurements from a total of 38 sociolinguistic interviews from the Heritage Language Variation and Change (HLVC) in Toronto Corpus (Nagy 2011). This amounted to a grand total of 889 tokens of AM/P and 816 tokens of OM/P. The Year of Birth of participants ranged from 1922 to 1998. Mixed effects modeling showed that OM/P is significantly raised (lower F1, p < 0.001), significantly retracted (lower F2, p < 0.001), and significantly more rounded (lower F3, p < 0.01) than AM/P. Pillai Scores for each individual speaker were also calculated. A Pearson Correlation test showed a significant inverse correlation between Year of Birth and Pillai Score (r(36) = -0.361, p< 0.05). While no significant difference was found in F1, F2, or F3 variation based on City or based on Generational Group, the overall results from this study show that a merger that was previously described as complete is, in fact, still ongoing in both homeland (Hong Kong) and heritage (Toronto) varieties of Cantonese.