Through analyzing response latencies, errors, and self-repairs in Mandarin, this investigation explores how monolingual, bilingual, and trilingual adults process their speech production differently using cognitive control mechanisms. In this study we conducted two experiments involving speech production in Mandarin. In the two experiments, 81 adults participated and were categorized into three groups: Mandarin monolingual, Hakka-Mandarin bilingual, and Hakka-Mandarin-Minnan trilingual. In Experiment 1 (unpreprogrammed task), each subject read 250 targets that were presented on a computer screen, one at a time, using E-prime; in Experiment 2 (preprogrammed task), each subject read 466 targets printed on A4 paper. Experiment 1 showed that bilinguals and trilinguals outperformed monolinguals in every aspect examined in this research, revealing the presence of bilingual and trilingual advantages in inhibitory control. Interestingly, in Experiment 2, we found a more complex pattern of results: a trilingual advantage in attentional control occurred during L2 production in terms of the number of errors and error correction; however, the bilingual group did not maintain this advantage in the task requiring attentional control. These experimental results revealed that the differences between the language groups became clearer when the inhibitory control demands increased and that compared with the trilingual advantage, the bilingual advantage in inhibitory control emerged in more limited contexts.