Using network science in the language sciences and clinic

Abstract

A number of variables—word frequency, word length—have long been known to influence language processing. This study briefly reviews the effects in speech perception and production of two more recently examined variables: phonotactic probability and neighbourhood density. It then describes a new approach to study language, network science, which is an interdisciplinary field drawing from mathematics, computer science, physics and other disciplines. In this approach, nodes represent individual entities in a system (i.e. phonological word-forms in the lexicon), links between nodes represent relationships between nodes (i.e. phonological neighbours) and various measures enable researchers to assess the micro-level (i.e. the individual word), the macro-level (i.e. characteristics about the whole system) and the meso-level (i.e. how an individual fits into smaller sub-groups in the larger system). Although research on individual lexical characteristics such as word-frequency has increased understanding of language processing, these measures only assess the “micro-level”. Using network science, researchers can examine words at various levels in the system and how each word relates to the many other words stored in the lexicon. Several new findings using the network science approach are summarized to illustrate how this approach can be used to advance basic research as well as clinical practice.

Keywords::

• Adults
• children
• psycholinguistic

Notes

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

This work was supported in part by NIDCD grant T32 – DC00052 Training Researchers in Language Impairments.

Supplementary material available online

Supplementary Appendix.

Notes

1. The words used to make the network came from the Webster’s Seventh Collegiate Dictionary (1967), which also forms the basis of several widely used databases in psycholinguistics (e.g. Storkel & Hoover, 2010a; Vitevitch & Luce, 2004). Although estimates of the size of the vocabulary of the average person vary widely, this sample is believed to be sufficiently representative.

2. Interestingly, such an “oddball” strategy may be advantageous for triggering the acquisition of novel phonological word-forms in typically-developing individuals (Storkel, 2011). That is, novel words that are phonologically less similar to other known words can be more easily identified as a novel word to which resources should be allocated in order to acquire it. Novel words that are phonologically similar to many known words may be erroneously identified as an already known word, thereby delaying the acquisition of that novel word (in typically-developing individuals)