Important words in the lexicon: The influence of closeness centrality on lexical processing

Network science is an interdisciplinary field drawing on computational and mathematical tools from mathematics, computer science, and physics. Network Science utilizes networks to examine real world complex systems. Within the network models nodes represent individual entities and links represent relationships between entities. A key finding of network science is that the underlying structure of a system will influence how that system functions. A network model of the phonological lexicon was created by Vitevitch (2008) using nodes to represent words and links to represent phonological similarity. The present work explores the influence of closeness centrality (a network measure of the average distance between a node and all other nodes in a network) on lexical processing. A word with a high closeness centrality value, such as CAN, will be centrally located and close to many other words in the lexicon. A word with a low closeness centrality value, such as CURE, will be located in a remote, sparse area of the lexicon and will be far from many other words in the lexicon. Three experiments were performed. Experiment 1 used a lexical search task in which participants were to turn one word into another by changing one sound at a time in the word. Participants were more successful at completing the task when it began at a word with low closeness centrality than at a word with high closeness centrality. Experiment 2 used an auditory lexical decision task and results show participants responded more quickly to words with high closeness centrality than to words with low closeness centrality. In Experiment 2, confounding variables were controlled during the initial selection of stimuli. However, in Experiment 3 an auditory lexical decision task was used again, but confounding variables were controlled via statistical analysis. In addition, a number of individual differences in participants were measured (e.g., vocabulary size, working memory span, processing speed, and inhibition processing). Experiment 3 results suggest an interaction between closeness centrality and frequency of occurrence on reaction times, but no impact of individual differences was observed on the closeness centrality effect. Results are explained in terms of a partial activation framework and implications of the work are discussed.