JCPSLP Vol 22 No 1 2020

standardised test of word reading (Rashotte, Torgerson, & Wagner, 1999). In an advance on Arciuli and Simpson (2012), they examined whether SL made a unique contribution in explaining variation in reading ability after taking into account a much larger range of variables known to be related to reading. Another advance included a power analysis justifying their sample size ( n = 65). Results showed a modest but significant relationship between children’s capacity for SL and their reading ability. A regression analysis showed that SL was a predictor of reading ability after accounting for a number of additional variables. Qi et al. (2019) developed two new SL tasks, an auditory and a visual version of the classic triplet task. These tasks included a novel measure of SL tied to response times during the cover task of oddball detection as well as accuracy measures derived from a surprise test phase. Results from 36 children and 36 adults showed that auditory SL rather than visual SL was related to sentence reading fluency after a number of variables were taken into account (Woodcock Johnson III; Woodcock, McGrew, & Mather, 2001). This result was observed in response time data rather than accuracy data. Additional analyses of the child data showed that the relationship between SL and nonword reading (WRMT; Woodcock, 1998) was mediated by phonological processing abilities. Another individual differences study used an SL task modelled after the triplet paradigm, but different to Arciuli and Simpson’s task, to examine 27 English-speaking adults and their L2 literacy acquisition in Hebrew (Frost, Armstrong, Siegelman, & Christiansen, 2013). Results revealed a relationship between the capacity for SL and reading proficiency. A neuroimaging study by Yu et al. (2019) used the same SL task reported by Frost, Siegelman, Narkiss, and Afek. (2013) in their investigation of adult readers of Chinese: 12 native readers and 12 non-native readers. Results revealed a link between SL and both L1 and L2 literacy acquisition, albeit in different reading networks of the brain. As an aside, in addition to studies of individual differences that compare performance on independent tests of SL and reading, there is research showing that SL plays a role in learning to read Chinese (He, W., & Tong, 2017; Yin & McBride, 2015; 2018). Some other studies have not found a link between independent tests of SL and reading ability. Nigro, Jiménez- Fernández, Simpson, and Defior (2015) used a novel visual SL task to examine 28 Spanish-speaking children. They did not find a link between SL and reading ability. They hypothesised that there may be a weaker link between the capacity for SL and variability in reading proficiency in a shallow orthography such as Spanish. However, they also emphasised low statistical power in their study. See also Schmalz, Moll, Mulatti, and Schulte-Korne (2019) who examined the link between SL and reading in 84 German- speaking adults using two novel visual SL tasks: a serial reaction time task and an artificial grammar learning task. Results reported no relationship between performance on these SL tasks and reading ability in German. In line with renewed interest in psychometric properties of psychological measures (e.g., Hedge, Powell, & Sumner, 2017), the reliability of SL tests has become a point of consideration (e.g., Siegelman et al., 2018; Kalra, Gabrieli, & Finn, 2019). Some recent studies have reported reliability. For example, internal consistency has been reported for some of the SL tasks mentioned in this narrative review (Torkildsen et al. [2019] reported Cronbach’s alpha of 0.81; Tong et al. [2019] reported Cronbach’s alpha of 0.56; Qi et al. [2019] reported Cronbach’s alpha ranging from 0.71 to 0.89; Kleij

et al. [2019] reported split half reliability ranging from 0.57 to 0.70). A separate study not discussed here has shown a 32 test-trial version of the Arciuli and Simpson task (2011; 2012) when administered to children has test-retest reliability of 0.57 after cover task performance is taken into account (Kidd, Smithson, Christiansen, & Arciuli, 2019). The body of research examining the link between individual differences in SL and individual differences in reading ability in the general population is growing. Researchers are paying increased attention to task reliability. Researchers are also creating new ways to measure SL in studies of reading (e.g., see the recent study by Siegelman et al. [2020] and the recent study by Snell and Theeuwes [2020]). Literacy instruction that draws on implicit learning techniques This body of research has emerged more recently than the others reviewed here. Some studies have demonstrated the efficacy of incidental learning procedures on broader reading and spelling outcomes (e.g., Protopapas et al., 2017; Tamura, Castles, & Nation, 2017). Other studies have begun to explore for the utility of manipulating the written input provided to the learner using an SL-informed approach. Far from replacing explicit instruction these approaches could supplement explicit instruction. Apfelbaum, Hazeltine, and McMurray (2013) drew on the principles of SL to facilitate literacy acquisition in a classroom setting. Their study of 224 English-speaking first graders was designed to determine whether children could learn implicitly about some regularities concerning graphemes to phoneme correspondences via exposure to words in computer tasks (e.g., vowel learning such as a in bat , i in bit , o in bot, ai in bait , ea in beat , and oa in boat ). Children were assigned to one of two groups where consonant frames were either variable or similar in the learning tasks. Children were given multiple attempts and feedback about accuracy but were not given any instruction regarding specific regularities. Results showed that children were capable of learning these regularities implicitly. Interestingly, variable consonant frames led to greater learning. Following Apfelbaum et al. (2013), Adwan- Mansour and Bitan (2017) found that variability facilitates adults’ learning of a novel script ( n = 60). They suggested that perhaps “variability reduces the effectiveness of whole word recognition and instead increases the salience of regularities, thus enhancing the extraction of letter-sound correspondences” (p. 2849). Most of the studies mentioned in this section on instructional techniques have been conducted with typically developing children and adults. It would be valuable to involve children with developmental disabilities who are struggling with literacy acquisition to ascertain whether gains can be made via these SL-informed approaches which parallel innovations in practices that support oral language acquisition (e.g., Alt, Meyers, & Ancharski, 2012; Alt, Meyers, Oglivie, Nicholas, & Arizmendi, 2014; Plante et al., 2014; Torkildsen et al., 2013). Conclusion For some languages it is not possible to explicitly teach all of the regularities between sounds and letters, and the regularities present within written words, that children need to acquire to become proficient readers. This is not to say that explicit instruction is unimportant. In fact, the opposite is true – explicit instruction is vital when children are learning

6

JCPSLP Volume 22, Number 1 2020

Journal of Clinical Practice in Speech-Language Pathology