Methodological Issues In Evaluating The Occurrence of Inferences

Two central questions in the study of inferences are (1) which types of inferences do readers/listeners routinely draw and (2) when do they draw them—on-line during comprehension or only later during retrieval when they are being queried about the text. Two techniques have been used for assessing inferences on-line. One is to intersperse throughout the text various questions that tap the reader's developing representation of the text to see if it contains information about unspecified information, such as why events have happened and what is going to happen next. The other on-line technique for assessing the occurrence of inferences is less invasive and, consequently, more popular among researchers. Activation is assessed on-line either during or immediately following an inference version of a text versus a control version. Researchers avoid using recognition, even with a deadline procedure, because there is no way to eliminate the possibility of this with recognition. Although the on-line assessment of activation levels of inference concepts avoids some of the criticisms of other measures of inferencing—such as cued recall and on-line question answering—it too has methodological problems. This chapter examines these problems and discusses their theoretical significance and their tentative solutions.