The Concept of Inference in Discourse Comprehension

Abstract In this chapter a short introduction to the subject of inference in discourse comprehension is given. It is a common view in linguistics and philosophy that forming inferences is an important part of communication. The special role of inferences in the various communicative processes, however, can be understood only in the light of experimental and simulation studies. This chapter, therefore, gives an review of inference research from the experimental viewpoint. In the first part of the chapter some conceptual problems of inference are discussed. The central problem discussed here is the distinction of the concepts of inference and of comprehension. Since comprehension is not possible without inferences both notions are sometimes used as interchangeable terms. In order to avoid this we delimit inferences to the semantic part of the comprehension process: An inference is the generation of new semantic information from old semantic information. In contrast, the concept of decoding refers to the generation of semantic information from non - semantic information. Furthermore, the concept of encoding entails the generation of non - semantic information from semantic information. For all three cases examples and borderline cases are presented. The second part of the chapter considers some theoretical aspects of inference in discourse comprehension. First, the representational aspect describes which types of semantic information are relevant for inference making. Several representational levels are distinguished and some interactions of inferences with these levels are discussed. Secondly, the functions of working memory for the processes of drawing inferences are analysed with respect to the procedural aspect. The central question here is the point of time at which the inference process starts, i. e. the question of forward vs. backward inferences. Whereas backward inferences seem to be the usual type of inference, forward inferences may be restricted to special comprehension strategies. Other special processes in discourse comprehension are the strategies for solving comprehension problems. Thirdly, inferences have a contextual component which includes cultural, situational, and personal factors. Inferences will, therefore, depend on all of these factors. The third part of the chapter discusses some classifications of inferences. According to our definition of inference, three aspects are crucial for the classification. The first aspect is the input of the inference process, i. e. the knowledge which exists already before the inference is drawn. The second aspect is the output of the inference process: The resulting mental representation of the text may be categorized according to the levels of semantic representation. The third aspect is the direction of the process between input and output: This may be vertical between different levels of the representation or horizontal at the same level. In the fourth part of the chapter methodological problems of experimental inference research are discussed. Here we argue for a multi - perspective approach in order to get converging evidence about the hypothetical constructs which are necessary for a theory of inference. In the fifth and final part of the chapter some perspectives for future research into inferences in discourse comprehension are presented. On the one hand, it seems necessary in the coming years to take a more functionally oriented view in order to explain the flexibility of the processes of inferring. On the other hand, the interdisciplinary framework of the developing Cognitive Science promises to provide a good theoretical and methodological support for psycholinguistic inference research.