In the interpretation of involuntary intrusive anxiety, psychologists tended to emphasize the importance of unconscious stimulation. Unconscious processing of threat is a type of automatic processing that is especially relevant to the core symptoms of pathological anxiety and may contribute to the experience of what is perceived as free-floating anxiety in generalised anxiety disorder (Mathews & Mackintosh, 1998). But social phobia, panic and obsessive-compulsive disorder are related to consciousness processing of threat, which are called intrusive thoughts (Li & Zinbarg, 2007). The association between conscious and unconscious threat processing remains unclear. Using subliminal affective priming paradigm, we studied the threat information subliminal priming effect and the moderating role of trait anxiety on it. 38 participants took part in the experiment. In the emotional Stroop task and awareness check, Chinese word sets (including dangerous words and neutral words ) were presented subliminally and supraliminally while the baseline did not present words so corresponding stems could be used to assess baseline completion. Subliminal, supraliminal, and baseline trials were presented in a random order. Each trial began with a word and/ or a mask using the same timing, word lists, colours, and font parameters as in the emotional Stroop task. Following the question (‘‘What’s the word?’’), participants either attempted to report the word or said ‘‘no’’. If a participant ever correctly reported a word in the subliminal condition (as happened for seven subjects), data from all trials were excluded. In the words completion participants were instructed to complete the stem to the first word to come to mind using the Chinese character presented while excluding any word just seen. Given the possibility that the stem led to conscious awareness of the word, the instruction to exclude the word when producing a completion provided a further safeguard against contamination by conscious perception of a word. We hypothesized subliminal word perception could elicit unconscious processing and result in an increased likelihood of using that word in the stem-completion task and there was interaction between trait anxiety and word valence on priming. The results were: (1) The hit rate of the target word in the subliminal trials was significantly higher than the baseline condition; (2) For the high TA group, the priming effect of dangerous words was significantly higher than that of neutral words, while for the low TA group, the priming effect of neutral words was significantly higher than that of dangerous words. We concluded that subliminal priming effect was significant and the unconscious processing of threat information of high TA group enhanced the conscious processing of the threat information leading to the involuntary intrusive anxiety. The findings support the notion that unconscious processing of threat among highly anxious individuals tends to break through into consciousness. A scenario whereby unconscious threat processing among highly anxious individuals leads directly to intrusive anxious thoughts may constitute an important mechanism responsible for the development and maintenance of anxiety.