The prenuclear glide in Southern Min has been analyzed as grouping with the rime (Chung, 1996; Bao, 2001) or as a secondary articulation of the onset (Duanmu, 1990). Four word-blending tasks were conducted to test whether the prenuclear glide in Southern Min syllables tend to cluster with the rime or with the onset. Participants had to combine two syllables into one. For instance, [ham11] and [tjong11] may be blended as [hjong11], implying the prenulcear glide goes with the rime, or as [hong11], implying the glide goes with the onset. The experimental results showed that the prenuclear glide preferred to go with the rime in Southern Min, but the cohesion between the prenuclear glide and the rime was affected by phonotactic probability. When the probability between the prenuclear glide and rime was high, the prenuclear glide tended to cluster with the rime. But the prenuclear glide was less preferred to cluster with the rime when the phonotactic probability between the prenuclear glide and rime was low. Besides, when the prenuclear glide was in the first input syllable, participants preferred to cluster the glide with the onset. But they preferred to group the prenuclear glide with the rime when the prenuclear glide was in the second syllable. This tendency shows that the participants followed the Maximality principle (Prince, 1985) when combining two syllables into one, thus preferring to produce the maximal possible syllable as output.