Gras, D. 1 & Nicolas, S. 2
1 Laboratoire de linguistique formelle UMR 7110 CNRS & Université Paris Diderot
2 Laboratoire mémoire et cognition Université Paris Descartes
We studied the time course of predictive inferences activation during the reading of French texts (inferences about what will happen next in the text that we are reading) in three experiments. In the two first experiments, we constructed texts in three versions. The predictive version which arouses the inference, the explicit version in which the prediction is explicitely presented in the text and the control version in which the prediction is not aroused. In experiment 1, we used a naming task after the presentation of the texts; the target (a word which represents the inference) is presented after the text at differents SOA (350, 750 and 1000 ms). The experiment 2 is the same than the first one, except that we use a Stroop like task instead of the naming task. The results of this two experiments show a facilitation in the naming task and an inhibition in the Stroop like task for the explicit version compared to the control version for all SOA. For the predictive version, we observe the same effects but just for the SOA of 1000 ms. The experiment 3 is the same that the experiment 2 but the explicit version is replaced with an infirmation version which arouses the prediction first and then contradicts it. In this experiment, the SOA used are 350, 750, 1000 and 2000 ms to know if the predictive inferences are desactivated after 2s. We observe an inhibition after the predictive version for the SOA of 750, 1000 and 2000 ms and a facilitation after the infirmation version for the SOA of 750 and 1000 ms. This results indicate that the predictive inferences are automatically activated or inhibited between 750 and 1000 ms after the reading of the text.