Abstract Taking tests enhances learning. But what happens when one cannot answer a test question—does an unsuccessful retrieval attempts impede future learning or enhance it? Although the study has demonstrated that unsuccessful retrieval attempts enhance learning, it has two major limitations. First, studies of unsuccessful retrieval attempts have been based almost exclusively on memory for cue-target pairs. The second major limitation of research on unsuccessful retrieval attempts is that it does not inform us about whether testing enhances the transfer of learning. 　Three experiments were conducted to investigate whether the unsuccessful retrieval complex meaning material would enhance retention and transfer learning, and the effect of the feedback timing as well. Participants in 3 experiments read an essay about Respiratory System, which consisted of eight paragraphs, the paragraphs contained either a single fact or a single concept, question was developed for each fact and concept. In Experiments 1 and 2, Participants in the test condition allowed 10 min to answer the questions. At the end of 10 min, participants given the text passage and told to study it for 8 min. Participants in the extended study condition were given 18 min to study the passage. Participants were then immediately administered the Final test, which consisted of 8 same questions from the initial test conditions (Experiments 1)or 8 new inferential questions (Experiments 2)to assess transfer. The first two experiments, the test group with delayed feedback .Therefore, added a condition of immediate feedback in Experiments 3, subjects were presented with the individual facts or concepts immediately after answer each question. 　Any items that participants answered correctly on the initial tests were removed from the following analyses of final test scores .The findings showed that (1) In Experiments 1, significant main effects of learning condition and type of question were found，F(1,55)=6.491，p<.05，ηp2=.106, F(1,55)=13.072，p<.05，ηp2=.192 respectively, with the test condition and the factual questions exhibiting better recall accuracy. An independent samples t test examined the effects of learning conditions by comparing factual and conceptual questions . For the factual questions, test group resulted in better final test performance than did the study group, t(55)=1.825，p<.05, while the conceptual questions attempt did not show such facility. (2) In Experiments 2, there was a main effect of learning condition, F(1,45)=10.19, p<.05，ηp2=.185. For the factual questions, test group resulted in better performance than did the study group, t(45)=3.059，p<.05. For the conceptual questions, there was no difference between the test and study group. (3) In Experiments 3, significant main effects of feedback timing were found, F(2，64)= 6.946，p <.001，ηp2=.178. Post Hoc revealed that the group of delayed feedback produced greater transfer than both the immediate feedback group and study group, ps <.05. For the factual questions, there was a significant effect of feedback timing, F(2，64)=8.282, p<.05, delayed feedback produced greater transfer than both the immediate feedback group and study group, ps<.05. 　Our study revealed that: unsuccessful retrieval of factual questions enhanced retention and transfer learning, and followed by delay feedback produced greater transfer than did immediate feedback and repeated studying; while the conceptual questions attempt did not show such facility. These results demonstrated that followed the unsuccessful retrieval attempt of factual knowledge, given learners delayed feedback can enhance retention and transfer learning; and for the conceptual knowledge which represented higher difficulty, it could be a boundary condition of unsuccessful retrieval attempts enhance learning.