Table 5. .
Results of a model fitted with both surprisal and SU.SGM and aimed at predicting the amplitude of the N400 component. The SU was estimated from an SG model trained on predicting the lexical form of the fillers, and not their semantic feature representations.
| β | t | p | |
|---|---|---|---|
| N400base | −0.41 | −21.14 | <0.001 |
| SU.SGM | 0.27 | 7.99 | <0.001 |
| surprisal.LM | 0.20 | 3.60 | <0.001 |