This study investigated whether learners of Japanese with different first language (L1) writing systems use different recognition strategies and whether second language (L2) exposure affects L2 kanji recognition. The study used a computerized lexical judgment task with 3 types of kanji characters to investigate these questions: (a) pseudo-homophones, (b) pseudo-homographs, and (c) real words. Three groups of learners participated in the study: (a) beginning-level learners of Japanese whose L1 was alphabetic, (b) beginning-level learners of Japanese whose L1 was logographic, and (c) intermediate-level learners of Japanese whose L1 was alphabetic. The results showed that both levels of learners whose L1 was alphabetic had poor results on the computerized lexical judgment test, possibly due to poor L2 orthographic awareness. The learners with L1 alphabetic knowledge used a poor visual recognition strategy for L2 kanji decoding, whereas those with L1 logographic knowledge were able to access individual kanji characters due to sufficient knowledge of the characters. Some of the learners also preferred phonological coding to recognize kanji characters. In addition, reaction time for the judgment task differed significantly between beginning- and intermediate-level learners. Results indicated that different reading strategies were used by learners of L1 alphabetic or logographic backgrounds and the beginning and intermediate learners who had had different degrees of exposure to the L2.