The notion of 'coherence' might be problematic, especially when viewed in relation to cohesion. There are some models which consider cohesion and coherence synonymous, being both derived from the verb 'cohere'. Other models consider the two disparate. However, the approach adopted here is the one that opts for the separation of cohesion and coherence. Moreover, only coherence are discussed here according to two influential models, and the deficiencies in these two models are explored through the approach adopted by Hoey and Jordan (1994). The paper concludes that the two models of coherence presented reflect the way coherence-based interpretation is effectuated. The van Dijk model maps the way comprehension can be approximated in the form of micro- and macro-structures. The two levels of analysis are interrelated, since micro-structures usually operate at the sentence level. de Beaugrande and Dressler's model, on the other hand, is based on the same cognitive background, but it employs intricate networks in order to map how coherence is established. Yet the two strategic points from which de Beaugrande and Dressler's model starts off are too rigid to be realistic. The model is remiss about the global structure of a given text, and this is why the more the text enlarges, the more intricate the network-cum-schema becomes.