Projects

Previous Research Accomplishments

My entire research career has been devoted to understanding the development of language, and how it is related to other cognitive processes. I am particularly interested in the structure of representations, and how they influence one another over acquisition. I have brought the strengths of both behavioral and neuroimaging methods to bear in elucidating underlying mechanisms and honing theory. Much of my work has focused specifically on the interactive specialization theory, which argues that interactivity between brain regions allows for specialization to emerge. Not only does this theoretical perspective provide key insights into developmental mechanisms, but it also holds exceptional translational potential, with implications for both early identification and intervention. For decades, my research has focused on individual differences and neuro-diverse populations, with an eye towards promoting learning in all children.

Two key aspects of language central to my work are the sound structure of words (i.e., phonology) and their meanings (i.e., semantics). Across many experiments, we have used rhyming judgments (e.g., lake and cake) and meaning judgments (e.g., king and queen) to examine specialization of brain function with respect to these core processes. In the course of this work, we discovered that children as young as 5 years of age show specialization for phonological and semantic processing, that this specialization increases over development, and that it is related to behavioral improvements in skill. We have shown that specialization for these processes is related to underlying brain structure and to more complex language, like when processing sentences. We also demonstrated that this specialization process is evident across languages (e.g., Mandarin, Spanish), suggesting that it is universal. We recently broadened our investigation of specialization to include syntax using grammaticality judgments (e.g., she throw) and semantics using plausibility judgments (e.g., bake rocks). Consistent with the interactive specialization approach, language development appears to be characterized by increases in the dissociation between these processes. Not only did we observe developmental changes in specialization, but the evidence further suggests that syntax and semantics bootstrap one another across development.

Over the years, I have been interested in how the development of oral and written language are related. Although decades of research had shown that phonological processing is associated with reading acquisition, my own work further questioned the mechanism underlying this association. Specifically, we asked whether learning to read changes the nature of phonological representations in the brain. We discovered that phonological representations in superior temporal cortex become more multisensory in the process of reading acquisition. We also demonstrated that orthographic representations (i.e., spelling patterns) in the fusiform cortex become increasingly structured along audiovisual lines, and that automatic activation of orthography during oral language processing seems to be a hallmark of reading skill. These discoveries challenged the view that these regions are purely unisensory. We additionally showed that multisensory interaction is supported by structural and functional connectivity through the parietal cortex. In subsequent longitudinal work, we demonstrated that the nature of these phonological and orthographic representations, and their interactivity, are important drivers of reading gains, thereby reflecting truly transactional developmental effects.

Given that our correlational studies strongly suggested that multisensory mapping is crucial for reading acquisition, we wished to directly test this in intervention experiments. Across several studies, we showed that systematically training the mapping between orthography and phonology through the alphabetic principle results in greater generalization. Moreover, electrically stimulating the parietal cortex, where this mapping appears to occur, results in better reading. In related work, we have shown that deficits in these mapping processes result in the use of semantics as a compensatory mechanism. Together, this work highlights the importance of mapping between representations and provides further support for the interactive specialization theory.

Because the Chinese writing system is very different from English, it represents a strong test of the generality of the interactive specialization theory. Chinese characters are organized holistically in two-dimensional squares with embedded subcomponents (e.g., 国). We discovered that this unique orthographic characteristic is related to developmental increases in the engagement of right visual cortex and alterations of this region in dyslexia. In addition, Chinese is less systematic in its mapping from orthography to phonology, with only about 25% of the phonetic radicals giving a reliable cue to pronunciation. For example, gōng jī can be written 攻击 to mean attack and 公鸡 to mean rooster. We discovered that this less systematic mapping results in reduced involvement of phonological regions during reading, and less influence of orthography on the nature of these representations in superior temporal cortex. The structure of the writing system influences the interactivity between these specialized regions.

Much of our work has investigated how language interfaces with other cognitive processes. Although dyslexia and dyscalculia often co-occur, we do not yet fully understand the common or unique mechanisms underlying skill in these domains. To elucidate these relations, we examined the role of phonological processing in the development of arithmetic. We argue that skill in multiplication facts (e.g., 8 x 9 = 72) relies on access to memorized phonological associations between the operands and the answer. In contrast, subtraction problems (e.g., 9 – 3 = 6) require quantity mechanisms representing analogue magnitude. Our imaging work revealed not only developmental increases in the specialization of these mechanisms for arithmetic, but also that children who do not appropriately rely on these specialized mechanisms show less improvement over time. Using a longitudinal approach, we further demonstrated that phonological mechanisms scaffold multiplication. Indeed, reliance on these verbal mechanisms for automatic retrieval seem to be associated with better attitudes towards math. We have also shown that quantity mechanisms scaffold subtraction. Parallel to our previous work on reading and phonological representations, and consistent with the interactive specialization theory, this relation appears to be bidirectional with arithmetic skills also refining quantity mechanisms.

Interestingly, the role of phonological and quantity mechanisms in math is moderated by experience. Chinese culture emphasizes rote verbal learning of arithmetic, and we discovered that this is related to their greater reliance on phonological mechanisms as compared to Americans. In addition to cultural differences, we have found that there are socio-economic differences in the neural basis of math. Better performance in higher socio-economic status (SES) groups is related to phonological reliance, likely due to more extensive language input, but better performance in lower SES groups is related to greater engagement of quantity mechanisms. These results support an adaptation model wherein lower and higher SES children engage different neural mechanisms to achieve similar performance. This adaptation model is also supported by our work in reading and deductive reasoning.

Our work on math and reading suggests that a common mechanism of skilled performance may be phonological processing, but unique mechanisms may be quantitative and orthographic processing, respectively. Neurodevelopmental conditions are multidimensional, so we have argued that the only way to accurately identify groups is to use an ensemble of theoretically motivated tasks. We now have a patent showing that diagnosis in attention deficit hyperactivity disorder (ADHD) can only be achieved with a factorial manipulation of working memory, feedback and reward. Our results suggest that deactivating reward regions and decoupling them from frontal regions when presented with the potential for a large reward is beneficial to cognitive control.

Finally, I am committed to open science through pre-registrations, transparent documentation and data sharing. Indeed, we have shared several large-scale neuroimaging data sets with representative developmental samples.

Research Goals for the Next Decade

Our work has consistently supported the interactive specialization theory by demonstrating developmental and skill-related effects. However, this theory also argues that early connectivity drives later dissociation. One key near-term goal is to use the existing data from our recently completed NIH R01 to directly test this hypothesis with respect to phonological, semantic and syntactic representations. We have some initial evidence that functional connectivity is indeed related to later specialization. We will also use this dataset to test whether interactive specialization is influenced by language and literacy experience in the home, building upon our work on socio-economic status. We have some preliminary results suggesting that more shared reading is associated with greater specialization for language. Our previous work has also demonstrated that Chinese cannot be fully assimilated into the neural network for English, so in our recently completed NSF SBE grant we taught adults an artificial language and manipulated the similarity of the structure of the new words to English. In the near-term, we will investigate whether dissimilar words are less effectively transferred to the neocortex over the course of learning.

In the mid-term, we will complete data collection on our NIH R01 focusing on the reading skills of children who are deaf and hard of hearing, with a partnership of that community. Some of these children struggle with learning to read, so our goal is to determine what mechanisms support better outcomes. Specifically, we are testing the idea that hearing children will rely more on the phonological pathway because they use the statistical regularities between letters and their sounds. We have some initial evidence that higher skill children rely on speech reading mechanisms. In contrast, we expect that children who are deaf and hard of hearing will rely more on the semantic pathway because they use a direct mapping between whole word spellings and their meanings. This project builds upon our previous cross-linguistic and neurodevelopmental work suggesting that semantics plays a greater role when mapping between orthography to phonology is less effective. In the mid-term, we also plan to complete a recently funded NIH RO1 which focuses on late talkers. Some preschool children who are language delayed are simply late bloomers, but others end up being diagnosed with developmental language disorder. Despite decades of research, we do not know why. Building on our prior work on neural specialization, we will test whether phonological specialization drives early language development, whereas semantic specialization drives later language development.

We are in the process of starting three new projects, all of which have been submitted to NIH as R01s. In first project, we aim to extend our work on reading to examine its inter-relation with anxiety. Despite their high co-occurrence, we do not know whether dyslexia leads to anxiety disorders or vice versa. Dyslexia may negatively impact self-concept leading to anxiety, or anxiety may lead to internal distraction which may impede learning. We will test whether these developmental effects are due to differences in bottom-up automaticity or top-down cognitive control in the reading network. In the second project, we aim to determine whether alterations in the language network are similar or different in children with autism compared to developmental language disorder. Based on relative differences between groups in behavioral skills on semantic and phonological tasks, we expect within- and between-network connectivity to differ between groups. This would be consistent with a distinct, rather than a shared, model of language impairments between groups. In the third project, we aim to examine outcomes of children with speech sound disorder with or without co-occurring language impairment. We expect alterations of phonological processing in the dorsal stream to be particularly related to speech and reading outcomes, but that alterations of semantic processing in the ventral stream to be related to language outcomes.

I am co-investigator on two neuroimaging projects studying phonological and semantic processing in Chinese English bilinguals with dyslexia and examining syntax and semantic processing for sentences in Spanish English bilinguals with developmental language disorder. I am also a co-investigator on a project using dense longitudinal neuroimaging to examine the emergence of specialization of numbers and letters in naturalistic scenes using a micro-genetic approach in young children.