Psychological latent structure of 2100 words The relevance of perceptual component in the semantic description of language

Abstract

In the past decade, with the increasing popularity of the embodiment theories (Barsalou, 1999) researchers started to collect measures that are related to sensory-motor experience with an object represented by the word. The need for sensory-motor measures came from embodiment explanations of mental representations of concepts, where representations are simulations of previous sensory-motor experience with a concept. In other words, the neural pathways and brain zones that are active during the perception of an object, are stimulated when we are only thinking about it. Lynott and Connell (2009; 2013) were the first authors who published sensory norms for English words. They defined perceptual strength of word as the extent to which is possible to experience it with the particular sense (to what extent we can experience something by vision, touch, smell, taste and hearing). Recently, Serbian (Filipović Đurđević, et al., 2016), Russian (Miklashevsky, 2018) and Dutch (Speed & Majid, 2017) norms were published and the cognitive relevance of those measures was recorded. For example, it was found that the strength of perceptual experience of words predicts lexical decision latencies over and above concreteness and imageability (Connell & Lynott, 2012; Filipović Đurđević et al, 2016). In the present study, more than 500 participants rated 2100 Serbian nouns on twelve dimensions: perceptual strength for five sensory modalities (vision, hearing, smell, taste, and touch), concreteness, imageability, context availability, familiarity, emotional valence, arousal and age of acquisition. We defined perceptual strength as the extent of being able to perceptually experience an object by relying on individual sensory modality. Since we want to explore whether the experience with an object is crucial for conceptual representation, we also collected per-modality ratings for actual perceptual experiences as introduced in Filipović Đurđević et al. (2016). All dimensions were rated on a seven-point scale, except for the age of acquisition (AoA) for which participants evaluated at what age they had learned the given word. We analysed the data with the principal component analysis (PCA) as we wanted to explore whether is possible to extract the perceptual component as a separate attribute of the semantic word space. The PCA analysis revealed the change in the latent structure depending on which perceptual strength ratings were included in the analysis. When the perceptual ratings were based on 32 possible experience (table 1), three components were extracted, explaining 72% of the variance. The first component contained visual, tactile, olfactory and gustatory modality, as well as concreteness and imageability. The second component contained semantical features: AoA, Familiarity and Context availability. The third factor was loaded with emotional valence, arousal and auditory modality. In the PCA analysis with the perceptual strength ratings based on real experience (table 2), four components were extracted, explaining 77,6% of the variance. The major difference was due to additional extraction of the olfactory, gustatory modality and auditory modality as the separate component. Our results showed that perceptual information is the relevant latent attribute of the semantic word space. It remains to conduct behavioural experiments to explore the cognitive relevance of these measures.