Abstract: Repetition blindness (Kanwisher, 1987, 1991) refers to the phenomenon that participants have difficulty detecting repetitions of words (C1 and C2) presented in a rapid serial visual presentation (RSVP) paradigm. Several accounts focused on why and how the recognition deficit for C2 occurs, such as Token Individuation Theory (Kanwisher, 1987, 1991; Kanwisher & Potter, 1989, 1990; Chun, 1997), Refractory Period Hypothesis (Luo & Caramazza, 1995, 1996), Potential Retrieval-Based Models (Fagot, & Pashler, 1995), Construction and Attribution Theory (Masson, 2004; Whittlesea & Masson, 2005; Whittlesea & Hughes, 2005) and Competition Hypothesis (Morris, Still, & Caldwell-Harris Harris, 2009). Among these theories, Token Individuation Theory and Construction and Attribution Theory are the most widely influential. Up to now, it is difficult for them to come to an agreement on the processing level of repetition blindness. For example, Token Individuation Theory argued that repetition blindness was perception deficit, occurring at perception level, while Construction and Attribution Theory argued that repetition blindness was memory deficit, occurring at semantic level. The present research investigated the processing level of repetition blindness with Chinese Polyphones.
We used a mixed design with three variables, one between-subject variable, Time Interval with two levels (100ms and 200ms), and two within-subject variables, Repetition with three levels (completely repeated, homophones repeated and different phonetic repeated), and Position with two levels (position homology and position dissimilarity). Each trial included three words, two experimental words (C1 and C2) and a high frequency word in between. All materials were presented on a computer screen in a rapid serial visual presentation (RSVP) paradigm. Participants were instructed to verbally report all words in a trial, no matter whether a word was repeated or not. The dependent variable was the accuracy rate for reporting both C1 and C2 correctly in a trial. If the repetition blindness occurs at the perception level, the accuracy rate in the homophones repeated condition should not be different from that in the different phonetic repeated condition. Furthermore, if the repetition blindness occurs at the perception level, the accuracy rate for in the position homology condition should be different from that in the position dissimilarity condition. Otherwise, results will support that the repetition blindness occurs at retrieval level of memory.
Results from a three-way mixed design analysis of variance (ANOVA) showed significant main effects of Repetition and Position. The interaction of Interval and Repetition was also significant. In both Time Interval conditions, the accuracy rate in the complete repeated condition was lower than that in the homophones repeated condition and the different phonetic repeated condition. However, only in 100ms condition, the accuracy rate in the homophones repeated condition was lower than that in the different phonetic repeated condition.
The results indicated: 1. For two completely repeated words, repetition blindness showed in both 100ms and 200ms conditions, but for homophones repeated words and different phonetic repeated words, repetition blindness showed in the 100ms condition only; 2. The repetition blindness occurred at the perception level, not at the semantic level.

Key words: repetition blindness, repetition priming, RSVP task, processing level