Traditionally, there are five lexical tones in Taiyuan dialect, but previous studies have shown that some speakers merged the four falling tones T1/T2/T4/T5 in the production. The current study investigates how 25 participants (14 females) with different age groups perceive four falling tones by means of discrimination and identification tasks. Both accuracy rate and reaction time are measured. Results show that: 1) the stabilization sequence in perception is as follows: T2>T1>T4>T5, indicating that T2 (Shang sheng) is the most stable tone while T5 (Yang entering tone) is most susceptible to merge with other tones. 2) the T4 and T5 (Yin and Yang entering tones) exhibit the highest degree of merger, with a tendency to perceive T5 as T4. 3) while participants across three age groups demonstrate similar patterns of tone merger in perception, the extent of merger varies. In the discrimination task, the middle-aged and elderly groups exhibit a greater degree of T4-T5 merger compared to the young-aged group; in the identification task, middle-aged participants demonstrate the highest identification rate for non-entering tones but the lowest rate for the identification of the Yang entering tone, indicating a pronounced confusion between the Yin and Yang entering tones in perception among this group.