The cognitive linguistic framework of Thinking for Speaking (TFS) (Slobin, 1987) classifies patterns for expressing motion events. Depending on how languages codify the manner and path of these events, languages may be classified as satellite-framed (e.g., English), verb-framed (e.g., Spanish), or equipollently-framed (e.g., Chinese). But while several studies have used this cognitive linguistic framework to explore the cross-language transfer patterns of learners' L1 when expressing motion events in a typologically different L2, this phenomenon has not been observed in the acquisition of a third language. This study examines the effect of lexicalization patterns in L1 Mandarin Chinese and L2 English on the description of motion events of L3 Spanish for intermediate and advanced students of Spanish living in the U.S. and China. Narrations were elicited from 16 adult Mandarin native speakers whose L2 is English. The data were subjected to statistical and qualitative analyses, with results indicating that the main factor conditioning the encoding of movement in L3 Spanish is not the learners' proficiency in Spanish, but rather their language of habitual use. Those who have spent more time in the U.S. and who use English as their habitual language display lexicalization patterns typical of English, codifying the manner of the main verb and using a higher number of prepositions to indicate direction. In contrast, participants whose habitual language was Chinese showed lexicalization patterns aligned with Chinese, encoding manner and direction in different verbs in compound sentences and by using deictic verbs. These results are contextualized within previous literature on both L2 TFS patterns as well as L3 acquisition and production more generally.