关键词: 刺激质量, 语义启动, 语义相关词对比率, 交互激活模型

Abstract: Previously, controversies are observed over the research approach for specifically stimulus quality interacted with?semantic priming during lexical decision tasks (LDTs), and generally between the Interactive Activation (IA) model and the Multistage Activation (MA) model. Both models involve three?stages of processing: feature, letter and word stages. According to the IA model, the processing in each stage feeds activation forward to other stages as soon as activation begins in the original stage (processing is cascaded),?accompanied with?feedback between adjacent stages. The cascaded activation?across representational stages can produce statistic interactions between?stimulus quality and semantic priming?even when?the?relatedness proportion (The?proportion of semantically related primes to targets)is low(0.25). Compared with forgoing IA model, each stage in MA model will not be activated, before the activation in the prior stage reaches?a?threshold or the prior processing stops. This model predicts interactions?as a result of a strategic?retrospective prime retrieval that influences early visual processing?only when the?relatedness proportion?is high. Given that semantic access of a visual word in Chinese system involves less phonological mediation, stimulus quality may possess greater impact on semantic processing via the cascade processing in Chinese than that in English. Thus, two experiments were performed to test the interactions between stimulus quality and semantic priming in different relatedness proportion during lexical decision tasks. Namely, they were to test if the interactions between stimulus quality and semantic priming occur regardless of relatedness proportion (i.e., 0.5 vs. 0.25). The experiments employed a 2 (semantic priming: related vs. unrelated) × 2 (stimulus quality: blurredness of level 5 vs. blurredness of level 6) design which comprised 4 treatment conditions. Four stimulus lists were created so that each condition has equal chance when every prime and every target character appeared by rotating the lists across?the student participants. The target characters were blurred with 66% and 51% proportion of image pixels masked for the levels 5 and 6, respectively. In Experiment 1, there were 35 real character trials in each condition, with an equal number of 140 pseudo-character trials. Experiment 2 shares the design solution with Experiment 1 only taking the exception of addition of extra (140?uncorrelated pairs of prime and characters) and (140?pairs of irrelevant characters and pseudo-characters) to the stimulus list with reduction of relatedness proportion?from 0.5 to 0.25. Overall, all experiments showed that the responses to both corresponding targets and slightly blurred targets initiated more quickly and accurately than those irrelevant and highly blurred?targets behaved. More importantly, as expected from the cascaded processing in the IA model, the experiments together indicate the interactions between stimulus quality and semantic priming on response times (RTs) rather than response accuracy regardless of the relatedness proportion. However, in contrast to previous studies, the designed experiments exhibited greater?priming effects on RTs for slightly blurred?stimuli?than?that of highly blurred?stimuli. Nonetheless, these discrepancies would not undermine the conclusion that the interactions, between?stimulus quality and semantic priming on RTs in Chinese reading, should consistently take place as well when the relatedness proportion?was very low. The interactions in Experiment 2 are not met the MA model prediction while the?strategic retrospective prime retrieval was invalidated in?low relatedness proportions. Meanwhile, these results suggest that the 0.5 ratio would not alter the amount of interactions between stimulus quality and semantic priming compared to the 0.25 one, and the relatedness proportion should not affect the interactions between stimulus quality and semantic priming in Chinese words processing. In conclusion, the results could be interpreted in terms of cascaded processing in the IA model but not applicable in the MA model.

Key words: stimulus quality, semantic priming, relatedness proportion, IA model.