Abstract The perceptual interference effect of memory refers to the phenomenon that masked presentations of English words during encoding counter-intuitively results in improved later memory. However, this effect is basically found with English words of high frequency. Further, controversies remain unsolved regarding how this effect is explained. More recently, a reverse effect has been found with Chinese characters of high frequency: characters that are followed by a mask are difficult to recognize in later test. This counter-effect of perceptual interference provides evidence against current theories of the interference effect. This counter effect has been suggested to reflect the reliance of Chinese characters on orthographic processing for semantic (deeper) processing during encoding, compared with English words and Chinese characters. Using a “study-test” paradigm, two experiments were run to examine this later suggestion with Chinese characters and double character compound words of low frequency. 384 Chinese characters (Experiment 1) and 384 compound words (Experiment 2) were presented to 55 native speakers of Mandarin（29 in experiment1, 26 in experiment 2） as participants in one of following 3 interference conditions in study phase. Specifically, a 500 ms prompt “+” was first presented, and then a character (or compound word) was either presented continuously for 2500 ms, or briefly (for 110 ms or 260 ms) and then followed by a mask of random stokes for the remaining time. The participants were then required to read the stimulus presented on the screen aloud. All the characters and compound words in the study portions were presented randomly. After a distractor task, the participants were given a yes-no recognition test. Two experiments were both run on DELL computer using E-prime software. As a result, the accuracy rate of character-recognition test was significantly higher and the RT was shorter in the no-mask condition than in other mask, especially stronger mask conditions (Exp. 1), whereas, no differences in both accuracy rate and RT of word-recognition test between any two conditions are significant (Exp. 2). In conclusion, counter interference effects occur for Chinese characters regardless of frequency, but not for Chinese compound words. It is again suggested that deeper processing for memory in Chinese characters more relies on orthographic processing than on phonological processing.