This research inspects the allocation of involvement load to the evaluation component of the involvement load hypothesis, examining how three typical approaches to evaluation (cloze-exercises, sentence-writing, and composition-writing) promote word learning. The results of this research were partially consistent with the predictions of the hypothesis: the two writing tasks with greater involvement load led to significantly better word learning than cloze-exercises with lower load, while composition-writing was significantly more effective than sentence-writing despite the same involvement load according to the matrix of the original model. Such results are explained from the perspectives of information organization and pre-task planning, based on which evaluation induced by cloze-exercises is suggested to be allocated with "moderate evaluation" as it involves no use of chunking, hierarchical organization or pre-task planning, evaluation induced by sentence-writing with "strong evaluation" as it involves chunking and pre-task planning at the sentence level, and evaluation induced by composition-writing with "very strong evaluation" for it involves chunking, hierarchical organization and pre-task planning at the composition level.