Figure 1.
Schematization of the current proposal, illustrating the relationship between comprehenders’ beliefs, the effect that observations in a new linguistic environment have on those beliefs, and the manner in which behavioral measures such as reading times provide a window onto these changes.
Figure 2.
Relationship between reading times and linguistic probabilities.
Illustration of the log-linear relationship between reading times and linguistic probabilities, following Smith and Levy (2013). We indicate the changes in probabilities (horizontal lines) for relative clauses (RCs; blue) and main verbs (MVs; red) for Experiment 1 and indicate the predicted changes in reading times for both (vertical lines).
Figure 3.
Predicted changes in surprisal for Experiment 1.
Qualitative illustration of the changes in surprisal for MVs and RCs predicted by our account throughout Experiment 1. Black arrows indicate the predicted changes in surprisal at the end of Experiment 1: large changes for RCs (blue), but tiny changes for MVs (red).
Figure 4.
By-region reading times for Experiment 1.
Mean length-corrected reading times at each sentence region (indicated in (2)) for all conditions in Experiment 1. Error bars represent 95% confidence intervals on the mean.
Table 1.
Summary of model results at each sentence region in Experiment 1.
Figure 5.
Adaptation effect for Experiment 1.
Change in length-corrected RTs plotted against item order, for ambiguous (dark lines) and unambiguous (light lines) MVs (red) and RCs (blue). Gray shaded regions correspond to 95% confidence intervals on the slopes.
Table 2.
Summary of the design and materials used in Experiment 2.
Figure 6.
Predicted changes in surprisal for Experiment 2.
Predicted changes in surprisal for the RC-First (left panel) and Filler-First (right panel) groups for both RCs (blue) and MVs (red). Dashed lines indicate predicted changes in surprisal of structures not present in a given block (e.g., the surprisal of MVs is predicted to change throughout Blocks 1 and 2 as a result of the occurrence of RCs, but no MVs actually occur in those blocks). Black arrows indicate the average predicted surprisal of MVs relative to RCs during Block 3.
Figure 7.
Mean length-corrected reading times in each block for the RC-First and Filler-First experimental groups. Error bars represent 95% confidence intervals on the means. Brackets with asterisks (*) indicate significant comparisons. For details, see main text.
Table 3.
Summary of model results at each sentence region in Experiment 2.