Increased sensitivity to proactive interference (PI) was widely observed in patients with memory impairment. There were two competing theories proposed to account for the susceptible PI effect in amnesia patients, with one holds that PI occurs at encoding stage, and the other believes that PI occurs at retrieval stage, however, its underlying mechanism was still unclear. Moreover, the results from patient study may be confounded by the differences of age, gender, education level and intelligence between patients and controls. The present study focused on this issue by in combination using neuropsychopharmacological experiment and computational cognitive modeling technique. Based on the recent findings from amnesic mild cognitive impairments (aMCI) that encoding impairment and susceptible PI effect coexisted in aMCI, and susceptible PI effect still remained in the absence of response competition, we hypothesized that the susceptible PI effect in amnesic patients might be primarily due to encoding deficits. 20 healthy adults (11 females) voluntarily participated in a double-blind, between-subject, placebo- controlled experiment, with a 2 (drug: midazolam, saline) × 3(list: list1, list2, list3) ×3(word pairs: control, interference, practice) factorial design. Subjects were required to participate the experiment twice, one week apart, under midazolam (0.03mg/kg) or saline. In each day, subjects went through 3 lists of word-pair associative learning tasks and a final cued-recall test. For each list, subjects were asked to remember 45 word pairs firstly, and then each word pair was tested twice. Three kinds of word pairs were designed, with control pairs studied on only one list, practice pairs practiced on all three lists, and interference pairs involved recombining cue and response terms from one list to the next. An ANOVA statistical analysis was run on behavioral data and SAC (Source of Action Confusion) models were constructed accordingly. It was found that, episodic memory was significantly reduced after midazolam injection, as contrast to saline injection. In list2, the list directly followed the injection, the PI effect was detected both under midazolam and saline, but the PI magnitude under midazolam was significantly higher than that under saline. The similar pattern was also observed in list3, although not to be significant. The output of the SAC model was fitted well with the experimental data. In conclusion, by using drug studies, we replicated and further demonstrated the susceptible PI effect in amnesic subjects. Moreover, with SAC, the present findings suggested that the increased sensitivity of PI under midazolam, as compared with saline, may be due to the encoding impairment under midazolam. The present finding implies that the sensitive PI effect in amnesic patients, such as aMCI and Alzheimer's Disease (AD), may be due to encoding deficit, and thus may contribute to the diagnosis and cognitive training of these patients.