How phonological neighborhood affects lexical retrieval can shed important light on lexical organization and processing. Yet these effects are unclear, particularly in Mandarin Chinese. This is likely because the working definition of phonological neighbors (i.e., the one-phoneme edit rule) used in Indo-European languages inadequately characterizes the phonological similarity among Mandarin words, which have simpler syllable structures and lexical tones. The current study proposes a graded Mandarin phonological neighborhood and investigates the impacts of near-to-distant Mandarin phonological neighbors on lexical retrieval. In Study 1, we investigated how Mandarin phonological similarity is influenced by the editing of lexical tone, constituent (onset/rime, initial/final) and phoneme. Native Mandarin speakers rated the similarity between the edited monosyllabic words. We found that constituent-edit neighbors were rated as the most dissimilar, followed by phoneme-edit neighbors, while tone-edit neighbors were the most similar. In Study 2, we calculated the constituent-, phoneme- and tone-edit phonological neighborhood densities and frequencies for 4,706 monosyllabic Mandarin words. We then utilized extant datasets to examine how the density and frequency of neighbors at varied distances, as well as of homophonic neighbors, impact response latencies in word naming, visual lexical decision, and picture naming tasks. The results showed that graded phonological neighbors had differential impacts on lexical retrieval efficiency: distant (constituent-edit) neighbors facilitated word retrieval, while near (phoneme-, tone-edit and homophonic) neighbors had inhibitory effects. We discuss these findings within an interactive activation and competition framework and suggest future directions to study the representation and processing of the Mandarin phonological lexicon.