ABSTRACT
Although there is a large consensus on the idea that readers connect each current processing unit to the immediately preceding ones (i.e., local coherence), theories of comprehension have evoked certain conditions under which distant units in the surface text structure are related and incorporated into the reader’s mental representation. For example, such connections are more likely to be made when participants are engaged in special comprehension strategies (McKoon & Ratcliff, 1992). The experimental situation, including the texts and the nature of the task to perform, may encourage readers to access distant information and build a global representation. A break in local coherence may also increase the likelihood that readers will seek to establish global coherence. These are the main factors in the minimalist view (Dell, McKoon, & Ratcliff, 1983; Haviland & Clark, 1974; Kintsch & Keenan, 1973; McKoon & Ratcliff, 1992). McKoon and Ratcliff (1992) consider the processing carried out by readers to be “minimal” in the sense that they will draw only those inferences necessary for establishing local coherence (i.e., automatic inferences), or based on readily available information. Global coherence would therefore depend on nonautomatic, strategic processing and would be established only when local coherence fails. Contrasting with the Minimalist Theory of the inferential process, the Constructionist Theory embraces a principle of “search
after meaning” that may or may not be successful regarding reading conditions (see Graesser et al., 1994). In this vein, inferences that are drawn during reading (i.e., online) are knowledge-based, because they are crucial to situation-model construction. The principle of “search after meaning” has three critical assumptions that make clear distinctions between the two theories of inference generation (i.e., minimalist and constructionist) and that constrain the types of inferences readers will generate. First, it is assumed that when readers comprehend a text, they are motivated by one or more goals, and therefore construct inferences that address those goals; this is the goal assumption. Second, in contrast to the Minimalist Theory, which assumes that global (or strategic) inferences are constructed only when local coherence cannot be established during clause interpretation, and that these inferences are not encoded as consistently during comprehension, readers attempt to construct a meaningful representation that is coherent at both a local and a global level (i.e., the coherence assumption). As already discussed in chapter 4, the distinction between local and global coherence is an important one. Local coherence refers to structures and processes that organize elements, constituents, and referents of adjacent clauses or short sequences of clauses; global coherence is established when local pieces of information are organized and interrelated into higher order elements. This higher level of coherence requires accounting for relationships between nonadjacent units, and consequently, is not as dependent on surface text cues as adjacent relations are. Also, unlike local inferences, global inferences are more time-consuming (see also McKoon & Ratcliff, 1992) because a reinstatement search has to take place. Thus, in line with this “global” approach to inference generation, text understanding requires readers to construct the most global representation of meaning, based on text information in interaction with background knowledge (for a typology of global inferences, see Graesser et al., 1994).
