For all my dissertation studies to be done, I am grateful for the use of an existing data set, which was a joint effort of my previous lab members under the direction of Dr. Neelima Wagley, Marisa Lytle for their help with research idea generation, emotional support, and data organization, and thank Dr. Agnieszka Dębska and Richael Rice for helping me practice and review defense presentation.
Finally, and most importantly, I would like to thank my family, roommate Qimeng Wang, friends Qiuhai Yue, Haiping Huang, and Zimu Su, without whom I cannot succeed and live a happy life in graduate school.
Background
Along with behavioral findings, previous neural studies found that the left vOT was activated during auditory phonological awareness tasks and its activation strength was associated with children's reading ability (Raschle, et al., 2012; Wang, Joanisse, & Booth, 2018; Debska et al. ., 2016; Wang et al., 2021; Derosches et al., 2010; Debska et al., 2019). Previous investigations of this model have focused on how the activation of orthographic representations by visual input relates to reading acquisition (e.g., Seidenberg, 1992; Cunningham et al., 2010; Centanni, et al., 2019). Several previous studies have investigated brain activation during phonological awareness tasks and its relationship with children's reading ability.
Only two previous neurological studies have used a longitudinal design to address causality (Mauer et al., 2009; Yu et al., 2020).
Aims of my dissertation
Two sets of studies
The longitudinal relation between reading skill and automatic orthographic activation
- Introduction
- Method
- Results
- Discussion
However, unexpectedly, brain activation in the anterior vOT for processing of large-grain rhyme (rhyme > onset) at T1 significantly predicted reading ability at T2 (p = 0.049, Bonferroni corrected) after accounting for the effects of the non-covariates are important (see table 2). Results of the hierarchical regression analyzes examining whether early brain activation in vOT predicted later reading skills in the younger cohort (scaffolding hypothesis). To address this issue, we separately examined the relationship between brain activation for rhyme at T1 and the standardized residual of reading comprehension at T2 (see Figure 3B), and the relationship between brain activation for onset at T1 and the standardized residual of reading at T1 T2. (See Figure 3C).
The scatter plots of early brain activation in vOT with later reading skills in the younger cohort. The scatterplot for the relationship between brain activation for rhyme at T1 and the standardized residual of reading ability at T2. However, we did not find that reading ability at T1 predicted functional connectivity between STG and the anterior vOT for large-grain rhyme (rhyme > onset) processing at T2, after accounting for the effects of the covariates of no interest ( see Table 3).
Results of the hierarchical regression analyzes examining whether early reading skills predicted later functional connectivity between STG and vOT in the younger cohort (sculpting hypothesis). As shown in the figure, the better children's reading skills were at T1, the more specialized the functional connectivity between STG and the posterior vOT for small-grain phonemic processing at T2. To answer this question, we separately looked at the relationship between reading ability at T1 and the standardized residual of functional connectivity for rhyme at T2 (see Figure 4B), and at the relationship between reading ability at T1 and the standardized residual of functional connectivity for beginning of rhyme. at T2 (see Figure 4C).
Scatterplots of early reading skill with later functional connectivity between STG and vOT in a younger cohort. Scatterplot for the relationship between reading skill at T1 and standardized residual functional connectivity for Start > Rhyme at T2. Scatterplot for the relationship between reading skill at T1 and standardized functional connectivity residual for baseline at T2.
We found that brain activation in the anterior vOT for large grain size affected subsequent reading ability.
The longitudinal relation between reading skill and phonological processing
- Introduction
- Method
- Analysis 1 -using univariate analyses
- Analysis 2-using multivariate analyses
- Results
- Results from analysis 1 using univariate analyses
- Results from analysis 2 using multivariate analyses
- Discussion
To examine the scaffolding hypothesis, a hierarchical regression analysis was conducted by entering decoding coefficients for onset versus unrelated or rhyme versus unrelated at T1 as predictors, with nonverbal IQ and reading ability at T1 as controlled variables. The scatterplots for the relationship (A) between initial processing at T1 and the standardized residual of reading proficiency at T2 and (B) between rhyme processing at T1 and the standardized residual of reading proficiency at T2 in the younger cohort. However, functional connectivity of IFG with STG for initial processing at T1 did not predict reading ability at T2 after controlling for reading ability and nonverbal IQ at T1 (see Table 6).
The scatterplots of (A) the relationship between reading skill at T1 and the standardized residual of brain activation in IFG at T2 and (B) the relationship between functional connectivity of IFG with STG at T1 and the standardized residual of reading skill at T2 in the older cohort. The regression analysis examining the scaffold hypothesis showed that only decoding coefficients of priming versus unrelated processing at L1 significantly predicted reading proficiency at L2 after controlling for reading proficiency and nonverbal IQ at L1 (p = 0.023). Decoding coefficients of rhyming versus unrelated processing at L1 did not predict reading proficiency at L2 (see Table 7).
As shown in the figure, the higher the decoding coefficients for priming versus unrelated treatment, the more children achieved their reading skills over time. The regression analysis examining the sculpture hypothesis showed that reading skills at T1 only predicted decoding coefficients for baseline versus unrelated treatment at T2 after controlling for decoding coefficients and nonverbal IQ at T1 (p = 0.009). Reading skills at T1 did not predict decoding coefficients for rhyming versus unrelated processing at T2 (see Table 8 ).
Scatterplots for (A) the relationship between onset decoding coefficients versus unrelated processing at T1 and the standardized residual of reading skill at T2 and (B) the relationship between reading skill at T1 and the standardized residual of onset versus unrelated processing decoding coefficients at T2. A regression analysis examining the scaffolding hypothesis showed that decoding onset versus unrelated processing coefficients at T1 did not predict reading skills at T2 after controlling for reading skills and nonverbal IQ at T1. Furthermore, rhyme decoding coefficients compared to unrelated processing at T1 did not predict reading skill at T2 (see Table 9).
A regression analysis examining the design hypothesis showed that reading skill at T1 did not predict decoding onset coefficients compared to unrelated processing at T2 after controlling for decoding coefficients and nonverbal IQ at T1.
General discussion
Specifically, phonological representations in STG played an earlier role than phonological access in IFG in the reciprocal relationship between reading ability and phonological processing. Second, reading small grain sizes is possible sooner than larger grain sizes because reading is expected to evolve from mapping small to large grain sizes. Finally, earlier reading skills always formed subsequent phonological representations at small grain sizes, because phonological awareness is expected to develop from large to small grain sizes.
Second, my dissertation provides the first neural evidence on the role of automatic orthographic activation during auditory processing in reading acquisition. Although the connectionist model suggests that a feedback mapping from phonology to orthography is influenced by reading acquisition, and that this feedback mapping facilitates word recognition, previous studies of this model focused only on how the activation of orthographic representations by visual input relates to reading acquisition. (e.g. Seidenberg, 1992; Cunningham et al., 2010; Centanni, et al., 2019). Using a longitudinal design and fMRI, my dissertation studies addressed the literature gap and found that the feedback mapping from phonology to orthography, as indicated by functional connectivity between STG and vOT during auditory processing, was the result of learning to read in young children .
These findings provide direct neural evidence supporting the feedback mapping hypotheses from the connectionist model. The current connectionist model contains only one phonological representational system to explain the process of reading development. Although previous neural research has shown that phonological representations in the left STG and phonological access in the left IFG are differentially related to developmental reading skill (Raschle et al., 2012; Luniewska et al., 2019; Debska et al., 2016; Chyl et al., 2018; Vandermosten et al., 2019; Corina et al., 2001; Heim et al., 2010; Cao et al., 2017), these studies only examined concurrent correlations and therefore cannot infer the directionality of the relationship .
My dissertation studies addressed this issue and found that early in development only phonological representation in the STG was scaffolded and sculpted by reading skills. These results suggest that it is necessary to add a control system to the connectionist model to account for the role of phonological access in order to provide a better and more neurally plausible account of reading development.
Future directions
For example, my previous studies have shown that beginning readers relied on letter representations in the posterior vOT for phonemic processing, while older children relied on their refined phonological representations in STG for phonemic processing. Furthermore, older children relied on rime representations in the anterior vOT for rhyme processing as a function of their reading ability (Wang, Joanisse, & Booth, 2018; Wang, Joanisse, & Booth, 2021). Although our previous cross-sectional studies have observed a developmental difference in the dependence of automatic orthographic activation in vOT and phonological representations in STG during phonological tasks, we do not know how these two representations interact to achieve this developmental change.
For small-grain phonemic processing, because greater utilization of mapping to letter representations will help young children better recognize the acoustically indistinguishable phonemes in spoken words, I expect that in young children, automatic letter representations in the posterior vOT will refine later phonemic representations in the STG. Thus, I expect that in older children, better phonemic representations in the STG would reduce the engagement of letter representations in the posterior vOT. For the processing of coarse-grained rhymes, because rhyme awareness is well developed early in life (Anthony, 2005) and reading develops into coarser-grained orthography-to-phonology mapping (Frith, 1985), I would expect that both young and older children, rhyme representations in the STG would predict larger later rhyme representations in the anterior vOT.
Some behavioral studies have used a longitudinal design and found an early stage effect and a later formative effect between word reading skill and vocabulary in elementary school students (e.g., Verhoeven et al., 2011). The emergence of this interest is due to the fact that in 6-year-old children we found early automatic activation in the anterior VOT for rhymes, which compromised their subsequent reading skill. A previous study showed that the left vOT is functionally connected to language areas in the infant brain (Li et al., 2020).
In the literature, only a few neuroimaging studies have investigated the role of HLE on pre-reading language processing in the brain. Therefore, in the future, by using a longitudinal design and integrating early brain measures, I would like to answer whether HLE influences pre-reading language processing beyond early brain features and how HLE and early brain features interact to influence pre-reading . language proficiency.
