Bilingual mental lexicon (ML) is a dynamic system under constant evolution by developments in lexical and conceptual knowledge being accessed through associative and semantic strategies. L1 ML is conceptually oriented; however, the nature of L2 ML has remained greatly controversial. Literature suggests that limitations in methodology, response categorization, and stimulus onset asynchronies (SOA) have led to inconsistent findings. On the other hand, the field lacks a comparison of the Persian and English MLs. In the present research, we sought to examine the most accessible link to the ML content in both Persian and English through comparing associative, semantic, and associative + semantic relations. We also explored the influence of L2 proficiency on this organization. To this aim, 79 Iranian EFL learners (19 to 55 years old) from two proficiency levels, lower-intermediate and advanced, were tested through two (Persian and English) masked-priming picture naming tasks (PNT). The PNT overcomes the weaknesses of word association tasks. In addition, the strength of lexical relations was contrasted in short and long SOAs to resolve the resulting ambiguities due to the processing costs of each relation. The reaction time data was analyzed using a linear mixed-effects regression model. The multilevel analysis revealed that associative and associative + semantic relations are similarly the strongest connections in the Persian and English MLs. Interestingly, no significant effect of English language proficiency was observed in the kind of lexical access strategies used. However, it significantly enhanced performance both in L1 and L2. These findings imply that L2 ML is established based on L1 word knowledge and world experiences attained through L1; and knowledge in both languages interactively promotes cognitive development. Accordingly, the L2 textbooks and teaching/learning strategies are recommended to be designed based on links between language forms and experiences i