Leo G. M. Noordman

Toward a taxonomy of coherence relations

Understanding a discourse means constructing a coherent representation of that discourse. Inferring coherence relations, such as cause‐consequence and claim‐argument, is a necessary condition for a discourse representation to be coherent. Despite some descriptively fairly adequate proposals in the literature, there is still no theoretically satisfying account of the links that make a discourse coherent. An adequate account of the relations establishing coherence has to be psychologically plausible, because coherence relations are ultimately cognitive relations. We are proposing a taxonomy that classifies coherence relations in terms of four cognitively salient primitives, such as the polarity of the relation and the pragmatic or semantic character of the link between the units. A classification experiment using fragments of written discourse showed that the 12 classes of coherence relations distinguished in the taxonomy appear to be intuitively plausible and applicable. A second experiment investigating t...

Causal inferences during the reading of expository texts

Abstract Five experiments are reported on backwards causal inferences, which are signaled by the conjunction because . These inferences serve to justify the causal relation expressed by the sentence. Although the inferences are required for a good understanding of the causal relation, they are not made during reading, as is indicated by reading and verification times in Experiments 1 and 2, and judgments in Experiment 3. However, Experiments 4 and 5 show that these inferences are made during reading if they are relevant to the purpose of reading.

Memory-based processing in understanding causal information

The reading process depends both on the text and on the reader. When we read a text, propositions in the current input are matched to propositions in the memory representation of the previous disco...

Modeling knowledge-based inferences in story comprehension

Abstract A computational model of inference during story comprehension is presented, in which story situations are represented distributively as points in a high-dimensional “situation-state space.” This state space organizes itself on the basis of a constructed microworld description. From the same description, causal/temporal world knowledge is extracted. The distributed representation of story situations is more flexible than Golden and Rumelhart’s [Discourse Proc 16 (1993) 203] localist representation. A story taking place in the microworld corresponds to a trajectory through situation-state space. During the inference process, world knowledge is applied to the story trajectory. This results in an adjusted trajectory, reflecting the inference of propositions that are likely to be the case. Although inferences do not result from a search for coherence, they do cause story coherence to increase. The results of simulations correspond to empirical data concerning inference, reading time, and depth of processing. An extension of the model for simulating story retention shows how coherence is preserved during retention without controlling the retention process. Simulation results correspond to empirical data concerning story recall and intrusion.

Readers' knowledge and the control of inferences in reading

Abstract There is a consensus in the literature that inferences which contribute to the coherence of the text representation are made during reading. This study demonstrates that this is an over-generalisation and that one has to make a distinction between relations internal to the structure of the representation and relations that involve reference to the world. It is demonstrated that the readers knowledge of the world is an important factor in controlling inferences. A number of experiments are discussed in which the role of the readers knowledge with respect to the information to be inferred is investigated by varying the materials in terms of their familiarity to the reader, and by having readers with high and low knowledge with respect to the content domain of the text.

Prosodic realizations of global and local structure and rhetorical relations in read aloud news reports

The aim of this research is to study effects of global and local structure of texts and of rhetorical relations between sentences on the prosodic realization of sentences in read aloud text. Twenty texts were analyzed using Rhetorical Structure Theory. Based on these analyses, the global structure in terms of hierarchical level, the local structure in terms of the relative importance of text segments and the rhetorical relations between text segments were identified. The texts were read aloud. Pause durations preceding segments, F0-maxima and articulation rates of the segments were measured. It was found that speakers give prosodic indications about hierarchical level by means of variations in pause duration and pitch range: the higher the segments are connected in the text structure, the longer the preceding pauses and the higher the F0-maxima are realized. Also, it was found that speakers articulate important segments more slowly than unimportant segments, and that they read aloud causally related segments with shorter in-between pauses and at faster rate than non-causally related segments. We conclude that variation in pause duration and F0-maximum is a robust means for speakers to express the global structure of texts, although this does not apply to all speakers. Speakers also vary pause duration and articulation rate to indicate importance of sentences and meaning relations between sentences.

Propositional Integration and World-Knowledge Inference: Processes in Understanding Because Sentences

The issue addressed in this study is whether propositional integration and world-knowledge inference can be distinguished as separate processes during the comprehension of Dutch omdat (because) sentences. “Propositional integration” refers to the process by which the reader establishes the type of relation between two clauses or sentences. “World-knowledge inference” refers to the process of deriving the general causal relation and checking it against the readers world knowledge. An eye-tracking experiment showed that the presence of the conjunction speeds up the processing of the words immediately following the conjunction, and slows down the processing of the sentence final words in comparison to the absence of the conjunction. A second, subject-paced reading experiment replicated the reading time findings, and the results of a verification task confirmed that the effect at the end of the sentence was due to inferential processing. The findings evidence integrative processing and inferential processing, respectively.

Coherence-driven resolution of referential ambiguity: a computational model.

We present a computational model that provides a unified account of inference, coherence, and disambiguation. It simulates how the build-up of coherence in text leads to the knowledge-based resolution of referential ambiguity. Possible interpretations of an ambiguity are represented bycenters of gravity in a high-dimensional space. The unresolved ambiguity forms a vector in the same space. This vector is attracted by the centers of gravity, while also being affected by context information and world knowledge. When the vector reaches one of the centers of gravity, the ambiguity is resolved to the corresponding interpretation. The model accounts for reading time and error rate data from experiments on ambiguous pronoun resolution and explains the effects of context informativeness, anaphor type, and processing depth. It shows how implicit causality can have an early effect during reading. A novel prediction is that ambiguities can remain unresolved if there is insufficient disambiguating in formation.

Inferences in Discourse, Psychology of

An inference is defined as the information that is not expressed explicitly by the text but is derived on the basis of the understander’s knowledge and is encoded in the mental representation of the text. Inferencing is considered as a central component in discourse understanding. Experimental methods to detect inferences, established findings, and some developments are reviewed. Attention is paid to the relation between inference processes and the brain. When listeners or readers understand discourse, they understand much more than what is stated explicitly in the sentences. On the basis of their knowledge of the language and the world, they are able to understand what has been left implicit in the text but is intended to be communicated. Consider the following text (a): (a) There were municipal elections yesterday. Because the majority of the lower town voted for the local party, there was a shift toward the left in the city council. But the right-wing party was not completely disappointed. It had expected to lose much more. In text (a), readers can infer that the local party is a leftwing party; that there is a causal relation between many people voting for a particular party and that party getting a stronger representation; that there are height differences in the town and that the town is probably located on a river bank; that the city council used to be more right oriented; that there is a contrast between the shift toward the left and not being completely disappointed, and consequently that the shift is a reason for being disappointed and that there are other reasons (worse expectations) why the right-wing party was not so disappointed; and that ‘it’ refers to the right-wing party and not, for example, to the city council. All these pieces of information are not stated explicitly in the text, but readers do understand them. They are called inferences. An inference is defined as information that is not expressed explicitly by the text but that can be derived from the text on the basis of the comprehender’s knowledge and that is encoded into the representation the comprehender constructs of the text. In this article, the notion of inference is restricted to the derivation of implicit information that occurs in spontaneous language processing. These inferences are distinguished from logical deductions in reasoning and problem solving (JohnsonLaird, 1983; see also Reasoning with Mental Models), such as the logical deduction ‘Some artists are chemists’ from the premises ‘Some beekeepers are artists. All beekeepers are chemists.’ Logic is concerned with the assessment of the validity of arguments in reasoning. Although listeners and readers certainly make logical deductions, and some inferences can be considered as logical inferences, everyday language behavior is not characterized by the evaluation of the validity of the arguments in reasoning. Two kinds of inferences might be distinguished. The first kind is an inference that is the derivation of new information. This is what in ordinary language is called an inference. For example, from the sentence ‘John selected his stranston shoes because there was much mud,’ the reader can infer that apparently stranston is a material or brand that has advantages if there is much mud. Given that the reader is not familiar with stranston, the reader cannot know that the conjunction ‘because’ is correctly used. However, by assuming that the sentence makes sense, the reader can derive the inference as new information. The second kind of inference is an activation of available knowledge. Examples from text (a) are that ‘it’ refers to the right-wing party, and that, given the contrastive relation indicated by ‘but,’ a shift toward an opponent party is in general a reason for disappointment. This kind of inference in general is not called an inference in everyday language, but most of the psycholinguistic research focuses on this kind of inference.