Semantic representations – what do we really know?
Talking about the INCREASE project, the coordinator Dr David Vinson explains that “any speaker of a language is likely to have a very deep understanding of the meanings of most words they use, whether they have learned them formally in an educational setting, or more naturally and organically in seeing and hearing how other people use them.” The challenge with this is that there is very little information on how word meaning is represented in speakers’ brains and how the similarity in meaning between words is established. The need for understanding Researchers who have studied this field have either looked at the meaning of words as being made up of a collection of properties or considered the meanings of words as represented by the contexts in which words are used. While both approaches work well for distinguishing the most related words from totally unrelated ones, it is not clear how important these differences are at predicting people’s responses to meaning or their brain responses when reading those words. The INCREASE project, with the support of the Marie Curie programme, set out to address this. “We aimed to investigate the nature of semantic representations – our knowledge of words’ meanings – that could be inferred by similarity dimensions between words, the extent to which two words have similar meaning,” explains Dr Maria Montefinese, a Marie Skłodowska-Curie fellow under the supervision of Dr Vinson. Key discoveries The project found that different similarity measures between words underline semantic representation of abstract and concrete words, and that their involvement depends on the task demands and the depth of encoding. “More importantly, we clarified the role of some brain areas of the left-lateralised semantic network in coding similarity information,” says Dr Montefinese. Dr Vinson adds: “The main results provide important ways forward in understanding how different aspects of information combine to make up the meanings of words, and how readers use meaning differently depending on how deeply they are processing the language they see.” Additionally, they show that people’s intuitions about word meaning do not entirely correspond to how they process meaning when using language. What’s next for INCREASE? Besides the project’s published conference proceeding of the Cognitive Science Society and four empirical papers in the journals Biological Psychology, Cognition and Emotion, Journal of Neurophysiology and Frontiers in Psychology, six more papers will arise from the project. INCREASE is also considering several ideas about its future direction. “We are currently investigating, with electroencephalography, the temporal dynamics of how similarity measures are used in processing abstract and concrete concepts and how their involvement is modulated by the task requirements,” reports Dr Montefinese. Other research questions will also be investigated in partnership with international and national collaborations that were started during the project’s lifetime. Dr Vinson says, “these questions are about the details of meaning representation that can be addressed now that we have obtained a more substantial understanding of the different contributions to word meaning.” Another important step for the project is to investigate the contribution of the context of the word meaning. So far, this has not been possible for practical reasons, whereby the main focus is on single words, read one at a time.
Keywords
INCREASE, semantic representation, word meaning, language, meanings of words, similarity measures, psychology, speakers’ brains
