JCPSLP Vol 22 No 1 2020

Expanding possibilities: Foci on reading and interdisciplinary practices

An introduction to research on statistical learning and reading Joanne Arciuli

Statistical learning (SL) is a form of implicit learning. It is thought that SL plays a role in a range of mental activities including reading. There are several different bodies of evidence to consider when thinking about the link between statistical learning and reading. These include corpus analyses, behavioural studies, and computational modelling which have provided converging evidence that the reading of single words and nonwords appears to proceed via principles of SL. In addition, there have been group comparisons investigating whether individuals with and without dyslexia differ in their capacity for SL. There have also been studies of whether individual differences in the capacity for SL relate to variability in reading ability in the general population. Lastly, some researchers have begun to consider principles of SL when designing classroom- based and other activities for beginning readers to see if this facilitates learning. Although there is ongoing discussion and debate regarding the nature of SL and the extent of its contribution to reading, clinicians may find it valuable to know about SL and the range of methodologies being utilised in this field of research. This invited paper summarises key points from the narrative review by Arciuli (2018) with discussion of additional studies. A guide to statistical learning and reading ability Children must learn to read. In an alphabetic language such as English, learning to read often begins with explicit instruction in the home environment and at preschool (e.g., to encourage children to recognise and reproduce the letters comprising their name). Once children start to tackle the reading aloud of a variety of individual words, they will likely be receiving formal instruction regarding regularities. Some of these regularities pertain to how letters map onto sounds (e.g., k maps onto /k/ as in kitten ) and how adjacent letters are combined within written words (e.g., ck

also maps onto /k/ as in kick ). They may also be instructed to attend to some regularities that pertain to non-adjacent letters within written words (e.g., vowels in certain consonant frames change when combined with a silent e: mat vs mate ). These are just some examples of the kinds of regularities that are taught via explicit instruction. Over time, as children are exposed to a greater amount of text, it is thought that other regularities will be learned implicitly. For example, a recent study by Steacy et al. (2019) provided insight into regularities concerning context dependent vowel pronunciation during children’s reading aloud of English monosyllables. Additional studies and reviews of regularities in the context of reading and spelling include Chetail (2017), Pacton, Fayol, and Perruchet (2005); Pacton, Perruchet, Fayol, and Cleeremans (2001), Senechal, Gingras, and L’Heureux (2016), and Treiman (2018), among others. As discussed in more detail below, large-scale corpus analyses have revealed regularities in the way letters are combined within written words and how these combinations relate to patterns of lexical stress in a number of alphabetic languages including English. To illustrate, Arciuli and Cupples (2006) reported an extensive list of these kinds of regularities (e.g., the majority of disyllabic words ending in -an have first syllable stress while the majority of disyllabic words ending in -uct have second syllable stress). It is not possible for educators and clinicians to convey all of these kinds of regularities to children explicitly. Most of us, children and adults alike, do not have conscious awareness of these regularities. Lesser known regularities that are not taught explicitly In languages such as English, lexical stress is critical for meaning. This is the case during comprehension of spoken words but also during reading. Take for example a word like zebra . In English, this word has first syllable stress. Second syllable stress turns it into a nonsense word and renders it meaningless. Some early research indicated that there might be orthographic cues to placement of lexical stress in the final part of English words (e.g., Kelly, Morris, & Verrekia, 1998; Smith, Baker, & Groat, 1982; Zevin & Joanisse, 2000) but there was no large-scale corpus analysis showing the extent of these cues. Arciuli and Cupples (2006) analysed more than 7,000 disyllabic words from the CELEX database (Baayen, Pippenbrock, & Gulikers, 1993). Their results revealed clear evidence that certain letter sequences in words’ endings are probabilistically associated with certain

KEYWORDS READING STATISTICAL LEARNING THIS IS AN INVITED ARTICLE

THIS ARTICLE HAS BEEN PEER- REVIEWED

Joanne Arciuli

3

JCPSLP Volume 22, Number 1 2020

www.speechpathologyaustralia.org.au