ABSTRACT
Bridging inferences have been assumed to differ from predictive inferences in that they would be encoded as part of a propositional text representation. In other research it has been conjectured that at least for some text materials, bridging inferences may be encoded similarly to predictive inferences, which are represented solely by a situational encoding. In two experiments, we tested these assumptions by determining the strengths of the various components of representations in different tasks and by measuring how fast bridging and predictive inferences are accessed in hemi-field presentations. The first experiment showed that the processing goal of verification made the encoding of bridging inferences more similar to paraphrases. A memorization goal, on the other hand, yielded bridging inferences more similar to predictive inferences. The second experiment furthermore indicated that bridging inferences are accessed faster in the left hemisphere, relative to predictive inferences that are accessed faster in the right hemisphere. A model which assumes that bridging inferences contain a propositional component explains these data well. In addition, it is concluded that processing goals should be taken into consideration in reading comprehension.
