2018 Sat Poster 6710

Saturday, November 3, 2018 | Poster Session II, Metcalf Small | 3:15pm

Statistical learning in reading development and reading impairment
Z. Qi, A. Nguyen, O. Ozernov-Palchik, S. Beach, S. May, J. Arciuli, J. Gabrieli

The implicit ability to detect and extract regularities from inputs, known as statistical learning (SL), plays a key role in spoken language development (Saffran et al., 1996; Evans et al., 2009). Fluent reading, however, relies on efficient cross-modal mapping of phonemes to graphemes. It has been proposed that dyslexia stems from fundamental deficits in extracting regularities (Ahissar, 2006). The relative importance of visual and auditory SL in reading development and reading impairment has not been evaluated. In the current study, we examined the associations between visual and auditory SL and reading performance in healthy adults and children (Expt. 1), as well as in adults with dyslexia (Expt. 2).

Participants in both experiments completed visual and auditory SL tasks (Fig. 1). During the 5-min implicit learning phase, participants were exposed to four triplets of alien cartoon images or tone sounds embedded in a continuous stream of stimuli, while performing a target detection cover task. During the test phase, participants responded to 32 two-alternative forced choice trials (embedded vs. foil triplet). In Expt. 1, 36 healthy adults (mean age: 24;0) and 36 typically developing children (mean age: 12;2) participated in the study (Table 1). Subjects’ reading skills were assessed using the sentence reading fluency test from Woodcock-Johnson III Test of Achievement. Both adults and children showed significant learning on both SL tasks (p’s < 0.001, Fig. 1). Accuracy did not differ between adults and children (p = 0.20). Individuals’ reading fluency standard scores were more strongly correlated (z = 2.05, p = 0.02) with their ASL accuracy (rho = 0.50, p < 0.001) than with their VSL accuracy (rho = 0.22, p < 0.1; Fig. 2). In Expt. 2, 13 adults with dyslexia and 21 healthy controls, matched on age, gender ratio, and non-verbal IQ, participated in the study. Individuals with dyslexia have standard scores below 90 in at least two out of four reading tests (Table 1). The ASL accuracy of the Dyslexia group (Mean: 0.53) was not different from chance (p = 0.13) and was significantly lower than that of the Typical group (Mean: 0.67, p < 0.001, W = 39.5, p = 0.001). In contrast, both groups performed significantly above chance in the VSL task (Dyslexia Mean: 0.69; Typical Mean: 0.67; p’s < 0.003) and there was no significant group difference (W = 124.5, p = 0.85; Fig. 3). Individuals’ word-level reading efficiency, measured by TOWRE total score, was more strongly associated with ASL (rho = 0.62, one-tailed p = 0.03) than with VSL (rho = -0.04, one-tailed p = 0.46) within the Dyslexia group (z = 1.67, p = 0.048). However, TOWRE scores were not significantly associated with either VSL or ASL accuracy within the Typical group (p’s > 0.35, Fig. 4).

The findings from two experiments provide converging evidence suggesting a key role of auditory statistical learning in reading development. Poor capacity to detect and extract statistical patterns in the auditory domain may impair phonological development and subsequently result in the reading deficits that characterize dyslexia.