The long-neglected issue of vocabulary acquisition is currently receiving attention in second language pedagogy and research reflecting the importance always accorded it by learners. But it is still far from clear how learners acquire vocabulary or how it can best be taught. As direct vocabulary instruction clearly fails to account for the substantial vocabulary knowledge of intermediate and advanced learners, there has been a broad acceptance that learners could acquire a large proportion of vocabulary incidentally from written contexts. Incidental vocabulary learning has been regarded as common for adults’ life-long learning of a foreign language, which is therefore encouraged to L2 learners for enhancing vocabulary ability. Nonetheless, it is not clear if, or to what extent, different L2 learning tasks have different impact on incidental vocabulary learning. This study aims to examine whether retention of vocabulary acquired incidentally is contingent upon the amount of task-induced involvement load, a construct proposed by Laufer and Hulstijn (2001) in the Involvement Load Hypothesis. Short-and long-term retention of ten unfamiliar words was investigated in three learning tasks(reading comprehension, comprehension plus filling in target words, and composition-writing with targets words) with varying "involvement load"—various combinations of need, search, and evaluation. To be brief, the present study aimed to investigate the following research questions: 1. Are tasks with a higher involvement load more effective for vocabularyretention than tasks with a lower involvement load?2. Do the factors such as participants’ NMET performance, present vocabulary size exert impact on vocabulary retention in an incidental-learning context? Is there any relationship between the result of the immediate posttest and that of the delayed posttest?91 college first-year English Majors from Jinhua college of Education in Zhejiang Province participated in the experiment and were randomly assigned to perform one of the three learning tasks varying in involvement loads: reading comprehension, reading comprehension plus filling in target words, andcomposition-writing with target words. Short and long-term retention of ten unfamiliar words was investigated on two (immediate and delayed) posttests at an interval of one week. The results of the current study are summarized as follows: 1 .Tasks with a higher involvement load, as predicted by the Involvement Load Hypothesis, were more effective for vocabulary retention than tasks with a lower involvement load.2. There is no significant correlation between participants’ NMET performance and vocabulary retention in the incidental-learning context; the participants’ present vocabulary size do not affect the retention score of the immediate posttest but there is a significant correlation between vocabulary size and vocabulary retention in the delayed posttest; there is significant correlation between the performance of the immediate posttest and the delayed posttest. These results are discussed in light of the Involvement Load Hypothesis and implications for further research and L2 pedagogy considered.