To recognize or report the second occurrence of a target in RSVP paradigm is challenging, a phenomenon known as Repetition Blindness (RB; Kanwisher, 1987). Some scholars argue that RB is due to an encoding failure of a repeated stimulus (e.g., Kanwisher, 1987, 1991; Kanwisher & Potter, 1989, 1990), while others believe it is caused by a retrieval failure (e.g., Armstrong & Mewhort, 1995; Fagot & Pashler, 1995; Masson, 2004; Whittlesea & Masson, 2005). Morris, Still and Caldwell-Harris (2009) proposed the Competition Hypothesis to further explain RB with inter-item competition. However, under the time press, people may allocate attention resources according to the task to get the best recall performance. In the RSVP paradigm, to serve the purpose of detecting more distinctive stimuli, repeated stimuli receive less attention resource than the non-repeated stimuli during the perceptual phase. Consequently, repeated items become difficult to report during the retrieval phase. Recently, Leng, Tan, Zeng, Cheng, and Lu (2012) illustrated these views in their Optimization Allocation of Attention Resources Hypothesis. The present study aimed to further examine the mechanism of RB through a new method. The typical RSVP paradigm includes one pair of repeated item (single repetition, C1 and C1’).In our research, two pairs of repeated items (double repetition, C1 and C1’, and C2 and C2’) were created. A within-subject design with one variable was used in all three experiments, repetition of words with three levels (non-repeated, one pair of word repeated, and two pairs of words repeated). The dependent variable was the accuracy rate for reporting targets. If RB occurs in a single repetition RSVP and it occurs at the perceptual phase, then the effect of RB from the second repeated stimulus (C2’) should not occur in a double repetition RSVP because the first repeated stimulus (C2) can not be recognized in the RSVP streams. If RB occurs at the retrieval phase in a single repetition RSVP, then the effect of RB from the second repeated stimulus should occur in a double repetition RSVP because the first repeated stimulus can be recognized in the RSVP displays. The second pairs repeated stimuli C2 and C2’ as well as an item between them construct a single repetition RSVP, in which the RB from C2’ should occur. In Experiment 1, both targets and non-targets were neutral words; while in Experiments 2a and 2b, targets were emotional words and non-targets were neutral words. Experiment 1 found that the accuracy rate for targets in the singly repeated but not in the doubly repeated condition differed from accuracy rate in the non-repeated condition. However, in Experiments 2a and 2b, the accuracy rate for targets in the doubly repeated condition but not in the singly repeated condition differed from the accuracy rate in the non-repeated condition. The results indicated: (1) The methodological merits of the repetitiveness of doubly repeated stimuli in a RSVP paradigm depended on the value intensity of targets and non-targets. When the intensity between them was equivalent, the second repeated targets showed advantage. When the intensity from the targets was stronger relative to the non-targets, the second repeated targets showed disadvantage. (2) People can allocate more attention resources to the stimui with higher intensity. RB occurs at the retrieval phase, not at the perception phase. The Optimization Allocation of Attention Resources Hypothesis can reasonably explain the effect of RB.