Abstract: Language is very informative. Using language to communicate allows communicators to communicate efficiently. Communicators could form predictive cognition based on speech information. The predictive cognition is not invariant on the semantic processing. For instance, there are also some occasional misunderstandings during communication. Thus, in order to better understand the meaning of language in the process of sentence comprehension, it is necessary to combine more information beyond semantics, which we called external-linguistic information, such as the speaker's body posture, voice, social status, and scene. Consequently, extra-linguistic information about the speaker's attitudes and emotional states might be expected to play an important role in language comprehension. Which means the notion of word meaning or lexical meaning in a sentence can be understood in different ways. It can be argued that meaning is constructed online each and every time a word is used, by combining extra-linguistic information and contextual cues. The combination of language with different external information produces different meanings, which may lead to deviations in the listener's semantic understanding, that is, ambiguity. In fact, there are many situations in every-day life where pragmatic information about the speaker is relevant for language comprehension, and may specifically influence the interpretation of a speaker's intended utterance meaning. This phenomenon has attracted extensive attention from researchers. Studies found that listeners rely on semantics to understand external information, and at the same time combine external information to understand semantics, and the two links are closely integrated. In linguistics, this combination of individual cognition difference and external information is called context predictive processing. Context predictive processing refers to the effective prediction to form predictive cognition before lexical semantic processing. There are two main factors influencing the formation of predictive cognition: prior knowledge and contextual information. The two factors contribute to predictive cognition and promote each other. It has been widely acknowledged that context predictive processing can affect semantic processing. In the specialization of language research into individual disciplines, and under the umbrella of two-stage theories that suggest pragmatic language processing follows semantic and syntactic processing, pragmatic language processing has often been investigated separately and independently from other aspects of language. In recent years, however, the view of pragmatics as a separable aspect of language processing has begun to change. Multiple studies show that context predictive processing affects the neurophysiological correlates of semantic processing. This paper combs the relevant studies on the formation of cognition by context predictive processing and finds that, in the process of semantic understanding, people's cognition formed by context predictive processing is not the invariant cognition formed by the initial state, but the dynamic processing process that is constantly adjusted with the increase of information received. However, most of the previous studies rarely included the influencing factor of the change of cognition in predictive processing. In order to better investigate the influence of the change of cognition on the semantic processing, this paper reviews the theories that can explain the formation of predictive cognition and explores the neural mechanism of the formation of predictive cognition. Firstly, This paper reviews the influence of predictive cognition on semantic processing and how predictive processing forms predictive cognition. Then, This paper makes clear the influencing factors of predictive cognition formation, that is, the role of prior knowledge and contextual information on predictive cognition formation. Next, this paper focuses on the theoretical explanation of predictive processing and verifies each other with the neural mechanism of predictive cognition formation. It is an attempt to illustrate the dynamic nature of predictive cognition. At last, the possible research directions of dynamic cognition in predictive processing are prospected: from the aspects of gender, the time of presentation of contextual information, the arousal of contextual information and the control of independent variables.

Key words: predictive processing, predictive cognition, prior knowledge, contextual information