There is a widely held belief that L2 vocabulary should be taught in an L2 context. However, studies in this field do not always support this idea because the effect of contextualized learning is rarely found in ordinary immediate-posttest scores. To explore the possibility of another way of assessing vocabulary cognition, Hasegawa (2010) used an imageability rating scale and found that context presentation enables learners to realize and imagine what the target word refers to. However, a follow-up study was imperative because the original study participants were only 22 university students and the factor of proficiency was not considered. According to past studies, the effect of contextualized learning might be found only among adult or advanced learners. Therefore, the present study carefully re-examined how context presentation affects learners' imageability ratings, as well as ordinary posttest scores, with an additional 118 high school students. In the experiment, the participants learned 21 unfamiliar adjectives in three types of translation-based learning that differed in context presentation: (a) no context, (b) collocational phrase context, and (c) sentential context. Study 1 analyzed the data from the new participants, and Study 2 compared the current data with Hasegawa's (2010) investigation. The results indicate that the context presentation does not enhance retention of vocabulary meanings and that the context effect on word imageability differed according to learner groups. Therefore, the previous conclusion should be modified as follows: When learners are provided with both translations and contexts, they tend to utilize translations as the immediate source of learning new words; however, university students are ready to enrich representations of lexical meaning in context.