Children can teach themselves new words via independent text reading. Previous studies on self-teaching heavily focused on learning to read in a first language (L1). Limited work was devoted to learning a second language (L2). We investigated the roles of exposure time of target pseudowords (four vs. six), availability of context (cohesive story vs. scrambled text), and phonological structure of target pseudowords (single consonant vs. consonant cluster) in self-teaching among English L2 learners. Forty-fifth-grade Chinese L1 children who are learning English L2 in Mainland China participated in the study. Children were asked to read through eight texts embedded with one regularly spelled pseudoword in each, consisting of four cohesive stories and four scrambled texts. The target pseudowords appeared four or six times in each half of the texts. Twenty-two children learned pseudowords with single initial consonants, and the other eighteen learned pseudowords with initial consonant clusters. An orthographic decision and then a spelling task were administered both immediately after text reading and after 3 days to assess orthographic learning. Results show that, in learning to spell, English L2 children were able to acquire novel English words with four exposures, and two more exposures and context benefited their performance. Targets with initial consonant clusters posed a challenge. In visual word recognition, context facilitated the learning of targets with initial consonant clusters. Our findings suggest that English L2 learners are able to self-teach; exposure time, context, and phonological structure all play important roles in orthographic learning in L2.