The foreign language vocabulary learning research literature often attributes strong mnemonic potency to the cognitive processing of meaning when learning words. Routinely cited as support for this idea are experiments by Craik and Tulving (C&T) demonstrating superior recognition and recall of studied words following semantic tasks ("deep" encoding) compared to structure-related tasks ("shallow" encoding). However, participants in C&T were not language learners but native speakers of English studying known English nouns. These experiments have never been directly replicated using nonnatives to establish the relevance of the findings to nonnatives and learners. The present study replicated C&T Experiment 5, comparing effects of shallow and deep encoding tasks on subsequent recognition of target words by native and nonnative speakers of English with equivalent short-term memory function. The results showed depth effects similar to C&T for all participants, indicating that C&T's results do generalize to less proficient speakers of the target language. It is crucial, however, that nonnative speakers of English benefited less from semantic encoding than native speakers, suggesting an effect of preexisting knowledge representations on mnemonic effects derived from semantic processing, and hence, a limit to the relevance of C&T for learners. Results are discussed in terms of the constructs of the mental lexicon, expert knowledge, distinctiveness and levels of processing in memory research and language learning.