Models of Discourse Representation

This chapter first describes what has been learned about how readers represent the meaning of discourse by integrating the meanings to individual sentences to construct the representations needed to understand larger segments of text. The chapter reviews the key findings related to text processing and how this sparked an ongoing debate about the extent to which the making of inferences during reading is obligatory. The chapter reviews precursor theories and models of discourse representation that attempt to explain how discourse representations are generated via the interaction of language processing and memory. The chapter then reviews a large, representative sample of the models that have been used to simulate and understand aspects of discourse processing. They are reviewed in their order of development to show how the models have evolved to accommodate new empirical findings. The chapter concludes with an explicit comparative analysis of the discourse-processing models and discusses the empirical findings that each model can and cannot explain.

Formal Models

This chapter introduces formal models of cognition and explains how they are similar to verbal theories but use computer programs and mathematics to avoid the many limitations of human reasoning, thereby adding precision and rigor to their explanations. The chapter discusses Marr’s (1982) levels of analyses and how information-processing systems can be understood and described in terms of the task being performed, the representations and algorithms used to perform the task, and how the latter are implemented by physical systems. This then motivates discussion of three common approaches to modeling human cognition and behavior: process models, production-system models, and connectionist models. Each of these approaches is critiqued, with discussion of its merits and limitations. The three modeling approaches are then further illustrated by showing how each might be used to explain the finding that words can be identified more efficiently if they occur in predictable sentence contexts. The chapter closes with a discussion of how cognitive models are evaluated using their simplicity, theoretical scope, compatibility (e.g., with biology), and their capacity to generate novel predictions for guiding research.

Synthesis

This chapter opens with a discussion of the limitations of current models of reading, and moves on to the reasons why more comprehensive models of reading are necessary to advance our understanding of the mental, perceptual, and motoric processes that support reading. The chapter then provides a comparative analysis of the various approaches that have been adopted to model reading, and how the theoretical assumptions of models of word identification, sentence processing, discourse representation, and eye-movement control might be combined to build a more comprehensive model of reading in its entirety. The remainder of the chapter then describes one such model, Über-Reader, and a series of simulations to illustrate how the model explains word identification, sentence processing, the encoding and recall of discourse meaning, and the patterns of eye movements that are observed during reading. The final sections of the chapter then address both the limitations and possible future applications of the model.

Introduction

This chapter provides an introduction to reading research and computer models. The chapter discusses the information-processing metaphors that have been used to study the human mind, drawing parallels between components of the latter and similar distinctions in computers (e.g., short- vs. long-term memory/storage systems). The chapter also introduces the modal model, or most commonly used metaphor for describing the human mind. The chapter then provides examples illustrating how behavioral experiments can be used to make informed inferences about the operation of the mind and its information-processing components. The chapter closes with a discussion of reading, including how it is similar and dissimilar to spoken language, and how the latter differs from other forms of communication (e.g., animal).

Models of Sentence Processing

This chapter first describes what has been learned about how readers process sentences, using information from individual words in combination with linguistic knowledge to generate larger units of meaning corresponding to phrases and sentences. The chapter then reviews what has been learned about sentence processing using various methods, but most notably, the measurement of readers’ eye movements. The chapter then reviews precursor theories and models of sentence processing—models that provide early attempts to explain how readers construct the meanings of phrases and sentences, and that motivate much of the subsequent research to understand the relative contributions of syntactic versus semantic information in sentence processing. The chapter then reviews a large, representative sample of the models that have been used to simulate and understand various facets of sentence processing. These are presented in their order of development to show how the models have evolved to accommodate new empirical findings. The chapter concludes with an explicit comparative analysis of the sentence-processing models and discussion of the empirical findings that each model can and cannot explain.

Models of the Reading Architecture

This chapter describes what has been learned about reading architecture, or how the mental processes that support word identification, sentence processing, and discourse representation during reading are coordinated with the systems that support vision, attention, and eye-movement control. The chapter reviews key findings that shed light on the nature of reading architecture, mainly using the results of eye-movement experiments. The chapter then reviews precursor theories and models of the reading architecture—early attempts to explain and simulate reading in its entirety. The chapter goes on to review a large, representative sample of the models that have been used to simulate and understand natural reading. Models are reviewed in their order of development to show how they have evolved to accommodate new empirical findings. The chapter concludes with an explicit comparative analysis of the models and a discussion of the empirical findings that each model can and cannot explain.

Models of Word Identification

This chapter first describes the tasks that are used to study how readers identify printed words (e.g., the lexical-decision task) and then reviews the key empirical findings related to skilled and impaired word identification (i.e., dyslexia). As explained, these findings have both motivated the development of computer models of word identification and been used to evaluate the explanatory adequacy of those models. The chapter then reviews several precursor theories and models of word identification that provide recurring metaphors (e.g., generating word pronunciations via analogy vs. the application of rules) in the development of later, more formally implemented word-identification models. The chapter reviews a large representative sample of these models in the order of their development, to show how the models have evolved in response to empirical research and the need to accommodate new findings (e.g., how the letters in words are perceived in their correct order). The chapter concludes with an explicit comparative analysis of the word-identification models and discussion of the findings that each model can and cannot explain.