Objective: To compare acoustic and linguistic features between individuals with Mild Cognitive Impairment (MCI) and Healthy Controls (HC) across different language tasks and explore effective discriminative features for MCI identification. Methods: Speech data from 66 eligible older adults (26 MCI, 40 HC) recruited from Study 1 were collected in quiet environments using digital recorders (DJI-MIC, ZOOM-H1). Participants completed three tasks: verbal fluency, picture description, and text reading. Multi-dimensional acoustic and linguistic features were extracted using Python and Praat software, with statistical analyses conducted to examine group and task differences. Results: Across all tasks, the MCI group exhibited significantly weaker performance in fundamental frequency mean, jitter, and energy distribution. Linguistic impairments manifested as poorer lexical selection and syntactic organization. Feature variations showed significant interaction effects across tasks. Compared to HC, the MCI group demonstrated more pronounced declines in acoustic features and linguistic flexibility during verbal fluency and picture description tasks, while exhibiting longer durations, increased instability in articulation, and more pauses during the reading task. Conclusion: Individuals with MCI show significant language deficits across tasks, with acoustic feature differences being consistently discriminative. The verbal fluency task revealed the most prominent group differences in speech characteristics.