Text comprehension involves the construction of a coherent mental representation of the situations cohesively described by the text. We use the term 'coherence' for these representational relationships and 'cohesion' for the textual indications that coherent representations should be built. Both cohesion and coherence can be indicated by lexical information (vocabulary-driven) or by grammatical information (grammar-driven). In either case, five types of coherence can result: REFERENTIAL, LOCATIONAL, CAUSAL, TEMPORAL AND ADDITIVE strands, corresponding to identification of the who, where, why, when and what. This thesis asks how textual cohesion supports representational coherence, by addressing two research questions: 1. To what extent is each of the types of cohesion independent of the others in its effects on the comprehender? 2. Does coherence develop incrementally throughout the comprehension process, does it come about delayed in a final wrapping-up stage of the comprehension process, or both? To explore effects of vocabulary-driven cohesion on coherence, a connectionist model is used. Vocabulary-driven cohesion is represented as a large number of weighted relations between the elements of the situations described by the text. A series of computational experiments measure the effects of vocabulary-driven cohesion on summary and recall data. The results show that vocabulary-driven cohesion relations both interact and make independent contributions to coherence. Since this computational model integrates information at the end of the comprehension process, its success of the model provides evidence for the plausibility of a delayed-coherence hypothesis. The effects of grammar-driven cohesion are explored in two stages. First, a parameterisation of local grammar-driven cohesion and coherence relations is developed, using the categories TYPE (CAUSAL, TEMPORAL, ADDITIVE) and POLARITY (POSITIVE, NEGATIVE). It is assumed that these relations are processed by combinations of parameters. A series of reading time experiments test the role of these parameters in the comprehension process. Again cohesion relations are shown to make both independent and interrelated contributions to the comprehension process. Furthermore, those cohesion relations that direct resources to coherence-based information in the text support the delayed-coherence hypothesis, whereas relations directing resources to cohesion-based information support the incremental-coherence hypothesis.