A test of the on-line status of goal-related inferences

Three experiments were conducted to investigate the on-line status of goal-related inferences. We combined a question-answering procedure and on-line measures of inference generation to test the prediction that superordinate goal inferences have a higher likelihood of being generated on-line during comprehension than do subordinate goal inferences. In Experiment 1, a lexical decision task revealed that superordinate goal inferences were encoded on-line as part of the reader's text representation even though these inferences were not true "bridging" inferences. In Experiment 2, we replicated this outcome in a word naming task. In Experiment 3, we established that the findings were not due to semantic associations between the target words and the lexical items in the text. The pattern of results was consistent with a global-coherence model of inference generation in which the reader generates causal connections that link each episode in a text from beginning to end.