Sensorimotor ambiguity does not affect reaction latencies in visual lexical decision task

Abstract

Ambiguous words vary on a continuum between words with multiple very similar senses (metonymy; e.g. chicken – animal or meat of that animal) and multiple unrelated and dissimilar meanings (homonymy; e.g. bank – institution or beach). The semantic similarity effect of of senses/meanings was shown to be facilitatory in visual lexical decision task (VLDT; Yurchenko et al., 2020), which is best explained by the distributed processing models where a word representation is seen as distributed in the pattern of activation of basic units representing its different features. The more similar word senses are, the more basic units they share, leading to faster activation and recognition in VLDT (Rodd, 2020). Parallel to this is the view of embodied cognition models that see prior sensorimotor (SM) experience with an object as the basis of a word representation (Barsalou, 1999). However, these models do not account for the fact that different word meanings/word senses can have different SM representations (e.g. chicken as an animal can be heard, while meat can not). We will name this phenomenon sensorimotor ambiguity. Different senses/meanings can be more or less similar in sensorimotor aspect. Sensorimotor (SM) similarity of different word senses/meanings is in positive relationship with semantic similarity (Anđelić & Filipović Đurđević, 2023), which suggests that SM information reflects semantic information. Our goal was to examine if SM similarity is an important factor for online word processing by examining its effect on reaction latencies. Analogous to semantic similarity, we expected it to be facilitatory. Participants (N = 103) took part in a VLDT with ambiguous nouns (N = 95). To ensure variability in predictor variable, we included both homonyms and polysemous nouns (metaphorical and metonymical). SM similarity ratings were calculated as average SM similarity between all sense/meaning pairs of one word (Anđelić & Filipović Đurđević, 2023). Pearson correlation between SM similarity and reaction times in VLDT was not significant, opposite to our hypothesis. Although these results suggest that SM information does not play an important role in activation of the word meaning, this is in conflict to embodied cognition models that suggest simulation of previous SM experience occurs during word recognition. On the other hand, it is possible that the SM effect is too small in magnitude to be recorded in this task, since SM information is postulated as only a subset of basic units whose activation patterns carry information about a word's meaning.