To learn words, infants must be sensitive to native phonological contrast. While lexical tone predominates as a source of phonemic contrast in human languages, there has been little investigation of the influences of lexical tone on word learning. The present study investigates infants' sensitivity to tone mispronunciations in two groups of infants. For one group (Chinese learners), tone is phonemic in their native language, and for the second group (English learners), tone is non-phonemic and constituted suprasegmental variation. In Experiment 1, English learners were trained on novel word-object pairings and tested on their recognition of correct pronunciations, tone and vowel mispronunciations of these words at 18 and 24 months. In Experiment 2a, bilingual English-Chinese learners were tested on a similar task translated into Chinese at the same age intervals. Results demonstrate that non-tonal learners treated tonal and vowel substitutions alike as mispronunciations at 18 months but only treated vowel substitutions as mispronunciations at 24 months. Tonal learners treated both tonal and vowel substitutions as mispronunciations at both ages. In Experiment 2b, bilingual non-tone language learners were tested on the same set of tasks replicating a similar set of results as monolingual non-tone language learners (Experiment 1). Findings point to an early predisposition to treat tone as a defining characteristic of words regardless of its lexical relevance at 18 months. Between 18 and 24 months, learners appear to ascribe lexical relevance to tone in a language-specific manner. The current study identifies the influences of tone variation on memories for newly learned words and the time period during which lexical tone - a highly frequent constituent of human languages - actually becomes lexical for early learners. Findings are contextualized with prevailing models of the developing lexicon.