Research in English for Academic Purposes has often analyzedlearner language discourse to provide evidence on learner language development (Biber & Gray, 2013; Staples, Egbert, Biber, & McClair, 2013). From the lexical perspective of a second language, previous analyses of learner discourse have focused more on vocabulary frequency measures; however, single word based analyses have been found to be insufficient in capturing learner language development (Read & Nation, 2006). Therefore, together with the measure of single word distributions, analysis of lexical bundles is also called for(Read & Nation, 2006). This study explored the use of lexicon and lexical bundles in written performances across three proficiency levels in an English Placement Test (EPT) corpus (N = 1,388) from a large Midwestern university in the US. The results showed that higher proficiency learners used a higher number of types and tokens and word families, a higher percentage of the second most frequent 1,000 words of English (K2 tokens) and Academic Word List (AWL) tokens, and a higher level of type-token ratio and lexical density than lower proficiency learners. Regarding the use of lexical bundles, although lexical bundles were used more often by lower level learners, most of those bundles were prompt dependent. Prepositional phrase-based lexical bundles with a discourse function were common in higher level responses while noun phrase-based and verb phrase-based bundles with referential and stance functions were frequent in lower level responses. This study has implications for teaching practice, second language acquisition research on learner language development, and second language writing assessment.