1.2.3. Event-Indexing Model

The event-indexing model was discussed in Zwaan et al. (1995). It proposes that comprehenders continually monitor the discourse model to maintain coherence on five dimensions: protagonist (who is involved in the events being described?), time (at what time is the event taking place?), space (what are the spatial relations between the characters, objects, and events being described?), causality (are the events in the discourse causally related to each other?), and intentionality (is the incoming information relevant to the protagonist’s goals and intentions?). Shifts on any one of those dimensions (e.g., if the discourse indicates some temporal delay between one event and the next) are typically associated with some processing cost as comprehenders update their representation. The largest processing costs tend to be associated with discontinuities on the time and protagonist dimensions.

1.2.4. Memory-Based Approach

Myers and O’Brien (1998) described the memory-based approach to discourse processing. Whereas other models might posit the operation of ‘active’ processing mechanisms (in the sense that these mechanisms actively retrieve information when building a representation of the discourse), the memory-based approach is built on passive mechanisms of memory retrieval. The memory-based approach is based on Hintzman’s (1986) MINERVA, and its ‘resonance’ process of memory retrieval. On this view, incoming information resonates both with the existing model of the discourse and with information in long-term memory. Information from memory is used to interpret the incoming sentence to the extent that it resonates with the new information.

1.3. Theoretical and Empirical Aspects of Discourse Processing

1.3.1. Integrating Sentences into a Coherent Discourse

One of the most basic questions in discourse-processing research is how comprehenders connect a series of utterances or sentences into a coherent whole. This issue can be separated into a series of smaller questions. First, how does the comprehender know that a series of utterances or sentences are intended to be interpreted as a connected discourse? There are several linguistic cues that signal the coherence of a discourse. When the series of sentences began with the definite article, comprehenders tended to interpret the series of sentences as part of a connected discourse. On the other hand, when the same sentences were presented with the indefinite article, comprehenders tended to interpret the sentences as being unrelated to each other. Thus, cues as subtle as the articles that are used in subsequent sentences indicate whether those sentences should be interpreted as related or not. Connectives (because, however, meanwhile, etc.) play a similar role. For example, in the sentences ‘‘The mother was preparing food for the party. Meanwhile, the grandparents were loading the minivan,’’ although the focus of the two sentences is different, the connective meanwhile indicates that both sentences should be interpreted as part of the same discourse model.

Once comprehenders have begun to integrate a series of sentences into their mental representation of a discourse, they must continue to process each subsequent sentence. There are several strategies that might be used. Comprehenders might connect each incoming sentence to the most recently encountered sentence. In doing so, the comprehender is attempting to maintain local coherence in their discourse model. On the other hand, comprehenders may attempt to maintain global coherence in their discourse model. That is, comprehenders might connect each incoming sentence by virtue of the overarching themes of the discourse (e.g., the goals of the protagonist).

In general, comprehenders attempt to maintain both local and global coherence. Experiments using the inconsistency detection paradigm discussed above demonstrate the guidance afforded by global coherence. For example, if a narrative begins by presenting a goal for Jane (she needs to get money into her bank account), and several sentences later, the comprehender is told that Jane is going to the beach without meeting her goal, comprehension will typically slow down. The slowed comprehension suggests that the overarching goal – a component of global coherence – affects processing individual sentences. Additionally, it is clear that comprehenders attempt to maintain local coherence from sentence to sentence. If a given discourse has too little local coherence, it will be viewed as altogether incoherent. Comprehenders pay attention to several elements of the discourse in attempting to maintain coherence. Chief among them are space (where are the events taking place?), time (when are the events taking place?), the protagonists being described, causality (is there a causal relationship between the events being described?), and the goals of the characters (Zwaan et al., 1995). An incoming sentence can potentially violate coherence on these dimensions. The more dimensions that are violated by the incoming sentence, the more difficult that sentence will be to understand and to integrate into the existing representation of the discourse.

There is a great amount of evidence in support of the claim that these dimensions are important for maintaining the coherence of a discourse. Space has been shown to be important in several ways. First, when two objects are described as being in different locations, focusing the discourse on one of those objects makes the other less accessible to the comprehender. Second, in studies in which participants study the layout of a building before reading a narrative about events that take place in the building, reading times are sensitive to the relative distances between the rooms in the building. Sentences that describe a character moving between adjacent rooms of the building are read more quickly than those describing a character moving between nonadjacent rooms.

Time is less well studied than space (i.e., the spatial dimension) but has an equally strong effect on language processing. Zwaan (1996) asked participants to read passages which contained a shift in temporal perspective. The time shift was either small (‘‘A moment later. . .’’) or large (‘‘A day later. . .’’). Comprehenders took more time to read sentences with a large temporal shift than to read sentences with a small temporal shift. Subsequent experiments have provided more support for claims about the importance of the temporal dimension. The remaining elements outlined above – protagonist, causality, and intentionality – have received less attention than space or time. The shifts on any of these dimensions can lead to processing difficulty. A more detailed understanding of how each of these dimensions works, and how the dimensions are used together during discourse comprehension, is needed.