The present study investigated how five first-elements of shoochuu types affect rendaku in the second element across the six dialect regions of Kagoshima, Oita, Fukuoka, Yamaguchi, Hiroshima and Shizuoka. The 'decision tree' analysis (405 participants) was conducted to predict a voice-or-voiceless decision by the two variables of shoochuu type and dialect region. Results indicated that the first element of the type of shoochuu was a significant factor for voice-or-voiceless decisions. 'Imo+shoochuu' showed the highest frequency of voicing at 93.83%. 'Kome+shoochuu' (88.89%) and 'Soba+shoochuu' (84.69%) showed the similar percentages. 'Mugi+shoochuu' (72.59%) was significantly lower than imo, kome or soba. 'Kokutoo+shoochuu' (56.44%) was the lowest. The present study also demonstrated that the six dialect regions appeared to have no influence on voice-or-voiceless decisions.