Semantic unification is one of the core components of language production and comprehension. In this paper, the most relevant models about semantic unification were described, including multiple phases model (Friederici, 2002), Bilateral Activation, Integration and Selection model (BIAS) (Jung-Beeman, 2005), Memory, Unification and Control model (MUC) (Hagoort, 2005), N400 model (Lau, Phillips, & Poeppel, 2008). The issues at core of debate include the definition of semantic unification and the corresponding locus, as well as whether there is a linguistic specific unification locus. Following the BIAS model and the N400 model, semantic unification is apart from semantic selection and controlled processing. While according to MUC model, semantic unification refers to the process by which we construct complex meaning based on elementary semantic blocks (e.g. words). It relies on selection and inhibition due to the inherent ambiguity nature of language comprehension. However, selection is not always a prerequisite for semantic unification. Unification constructs semantic representation that is not already available in long-term memory. In addition, some information which should be unified is unique and does not engage selection. For instance, Tesink et al. (Tesink, et al., 2009) found left inferior frontal gyrus activation when content violated against character’s gender information as indicated by the acoustics of the voice (for example, a men said he was in pregnant), a situation unlikely relies on selection. In the future, it would be helpful to further investigate whether the activation in left inferior frontal gyrus during semantic unification is linguistic-specified or domain general. Since it was found that activation between semantic unification and general cognitive control such as flanker task induced control processing was overlap (Ye & Zhou, 2009). Moreover, cumulative evidence suggested that functional connected network rather than single region support cognition. Thus the field would be benefit from connectivity analysis for semantic unification. The connectivity network may involve the cortical and subcortical structures, and bilateral hemispheres.