Production effect refers to the typical finding that words learned by reading aloud are better remembered than those learned silently. Currently, there are two theoretical explanations to account for the production effect, namely, the distinctiveness account and the strength account. According to the distinctiveness account, words being read aloud are particularly distinctive compared to words being read silently, pronunciation improves memory. However, strength account claims that the production effect is due to the difference of coding strength, coding strength is simply greater for words being read aloud. Previous studies of production effect mainly focused on a specific group, lack exploration of effect changing from the perspective of development. In the present study, we aimed to test the two theoretical explanations and further explored the development characteristics of the production effect of Chinese written word. One hundred and ninety-two participants with normal reading ability participated in the experiment, of which 40 are in grade three, 39 in grade five, 37 in grade seven, 36 in grade eight, and 40 in college. We investigated the influence of reading mode and learning frequency on the production effect using the learning-testing (corresponding to encoding and recognition, respectively) paradigm. Therefore, the present study was a 2 (reading mode: loud reading, silent reading) ×2 (learning frequency: one time, three times) ×5 (grade: grade three, grade five, grade seven, grade eight, college students) mixed design, and the dependent variable was the memory performance. According to the distinctiveness account, learning frequency should not influence the production effect, whereas the strength account predicts that the production effect should increase when learning frequency increase. The results showed that: (a) Main effect of grade occurred, with the memory performance of college students was significantly better than that of students in grade three, grade five, and grade seven; the memory performance of grade five, grade seven, and grade eight was significantly better than that of grade three. (b) The main effect of reading mode was significant, with the memory performance of reading aloud being significantly better than that of silent reading. (c) The main effect of learning frequency was also significant, with the memory performance of learning three times being significantly better than that of learning only once. (d) The interaction between grade and reading mode was significant. Further analysis showed that the memory performance of college students was significantly better than that of grade three in the loud reading memory, but not better than that of grade five, which indicated that the memory of loud reading tends to be mature in grade five; the memory performance of college students was significantly better than that of grade three, grade five and grade seven in the silent reading memory, but not significantly better than that of grade eight, which indicated that the memory of silent reading tends to be mature in grade eight. The difference in the production effect between grade seven and college students was not significant. The interaction between grade, reading mode and learning frequency was not significant, which supported the distinctiveness account.