Abstract: Dehaene, et al. (1998) and Greenwald, et al. (1996) have brought substantial evidence on the existence of subliminal perception. But there are still controversies about which level can subliminal perception reach. While some researcher argued the subliminal stimuli can be processed semantically, other researcher argued it was a shallow processing based on the visual characteristic of stimuli. Both sides have empirical evidences to support their views. Several theories object the subliminal semantic process, such as direct motor specification hypothesis, target stimuli activation theory and action trigger theory. Van Opstal and his colleagues (2005) believed that even not all of the experiments can be explained by action trigger theory, at least under some masking conditions, real semantic prime occurred. In order to explore whether and why semantic process exists in the subliminal process, two experiments were designed in this study. Based on previous researches, the present paper explored the depth and potential moderators of priming effects with picture and Chinese words as stimuli. 26 college students (age ranged from 20 to 26 years old (14 female, average ages was 22.3) ) participated in the experiment1. The masked priming paradigm and category classification task were adopted in the experiment1. Both indirect and direct measurements were tested on participants to make sure that the priming stimuli were operated unconsciously. At the mean time, big categories such as “animal” and “non-animal” were used to prevent the interpretation of action trigger. In the experiment2, 28 college students participated in the exp2. Two participants were excluded due to their slow reaction. The final sample was constituted by 26 participants (16 female, average ages was 19.2). Priming stimuli and target stimuli were presented by different form, such as word characters and pictures, to prevent the potential stimuli—response connection. Moreover, the size of target category was manipulated to examine their effect on subliminal process. The main results are as follows: when using pictures as materials, there is no interaction between priming stimuli and target stimuli. The fact that priming effect could still be seen in a big category classification task provides further evidence of subliminal semantic process. Meanwhile, subliminal semantic process also can be seen by using pictures as priming stimuli and Chinese characters as target stimuli, since there are no interactions between them. In addition, the depth of subliminal perception is decided by which level of process participants’ use on target stimuli. If participants process the target stimuli on a low level, the subliminal stimuli would be processed in the perceptive level. If participants process the target stimuli on a deep level, the subliminal stimuli would also be processed in the semantic level. The present study has important implications in the area of subliminal perception. Not only does this study provide further evidence of the existence of semantic subliminal perception, but also indicates that the task context could determine whether stimuli will be processed semantically or nonsemantically, which in turn can influence priming effect.

Key words: subliminal perception, masked priming paradigm, semantic processing