False memory is a phenomenon in which people report deviations from facts in past events. The current study used Deese-Roediger-McDermott（DRM）paradigm to explore the effect of image encoding on vocabulary-based false memory. The reason why DRM paradigm was chosen for research is that the false memory effect induced by DRM paradigm is stronger. In a typical DRM procedure, Participants study a list of words (e.g., milk, toast, flour, butter, jam, biscuits, cream, oat, milk powder, paste ) that are semantically associated to a critical lure (bread). During a memory test, participants often falsely remember the critical lure(bread) as a studied list item, illustrating a false memory for the list. How to effectively reduce the false memory is a practical question. Imagine items intentionally not only can suppress semantic network activation during the encoding phase, but also can facilitate the extraction of real learning materials. Several researches using DRM paradigm had found that instructing participants to imagine items intentionally can reduce false memory rates, however, some researches claimed that there is no significant difference. By comparing the experiment procedure used in those researches, we found different image encoding time and depth of image encoding were used and mixed up, and the effect of image encoding on false memory is relatively small totally. We conducted two experiments to clarify the effects of image encoding times and the degree of image encoding on false memory. Experiment 1 used a 2(image encoding time: 3 seconds, 5 seconds) × 2 (test type: free recall, recognition) mixed experiment design, with image encoding time the within-subject variable, and test type the between-subject variable. Experiment 2 used a 2(image encoding time: 3 seconds, 5 seconds) ×2(image encoding degree: deep processing, shallow processing) ×2 (test type: free recall, recognition) mixed experiment design, with image encoding time and image encoding degree the between-subject variable, test type the within-subject variable. Dependent variable is the false memory rate. In the learning stage, the participants remembered the presented words. During the testing phase, at the end of each word list learning, the participants recalled the word list freely, the recognition test was performed after the distraction task. The results showed that false memories under 5 seconds of image encoding time was significantly less than that under 3 seconds of image encoding time（p<.05）. Under the condition of 5 seconds of image encoding time, false memory in the deep-processing group was significantly lower than that of the shallow-processing group (p<.01). The Activation-Monitoring Framework can be better explained the result. In the learning stage, participants imagine items intentionally, it can suppress its automatic image and reduce the probability of activation of critical lure, on the other hand, image encoding can process more details. These results suggested that image encoding can reduce false memory rates. More specifically, when given more image encoding time, with deeper processing, participants were able to remember more details, which mediately led to reducing the false memory rate. Further study of the effect of image encoding on false memory can consider the different situation.