Meaning in brains and machines: Internal activation update in large-scale language model partially reflects the N400 brain potential
- Author(s): Lindborg, Alma
- Rabovsky, Milena
- et al.
The N400 brain potential has been used as a neural correlate of meaning-related processing in the brain, but its underlying computational mechanism is still not well understood. Although efforts to model the N400 as an update of a probabilistic representation of meaning have been promising, the limited scope of earlier models has restricted experiments to highly simplified sentences. Here, we expand modelling of the N400 to naturalistic sentences using a large-scale, state-of-the-art deep learning language model. We investigate the correspondence between updates in the internal state of the model and the N400 in one quantitative experiment and four qualitative experiments. Our findings suggest that activation updates in the model correspond to several N400 effects, but cannot account for all of them.