[PS-1.13] The N400 component of the ERP: Some explorations with a connectionist attractor model of conceptual processing

Rabovsky, M. 1 & McRae, K. 2

1 Humboldt University, Berlin, Germany
2 University of Western Ontario, London, ON, Canada

The N400 ERP component is widely used in research on language and semantic memory. Although the component?s relation to semantic processing is well-established, the specific computational mechanisms underlying N400 generation are currently not clear. We explored the mechanisms underlying the N400 by examining which parameters in a connectionist model of conceptual processing most closely covary with N400 amplitudes. The model has 30 input units representing word form that map onto 2526 directly interconnected semantic feature units representing word meaning, according to semantic feature production norms. We simulated a number of N400 effects obtained in human empirical research: influences of semantic priming, lexical frequency, number of features (NOF; also possibly a proxy for concreteness), and repetition, as well as influences of frequency and NOF on repetition effects. Cross-entropy error values were consistently in the same direction as N400 amplitudes. Like N400 amplitudes, error values were larger for low frequency words, larger for words with many features, and decreased for semantically related target words as well as repeated words. Furthermore, the repetition-induced decrease was stronger for low frequency words, and for words with many semantic features. In contrast, there was less of a correspondence between total semantic activation and the N400. Like N400 amplitudes, activation was larger for words with many semantic features. However, activation also tended to increase with frequency, repetition and semantic priming which is opposite to well-established N400 results, and may be more in line with increased activation facilitating decision latencies in lexical and semantic tasks. Our results suggest an interesting relation between N400 amplitudes and error values in connectionist models of meaning. In psychological terms, error values in connectionist models have been conceptualized as implicit prediction error, and we discuss the possibility that N400 amplitudes may reflect this implicit prediction error in semantic memory.