How do (or don’t) we learn letters-sound associations? Neurocognitive processes underlying the learning of Grapheme-Phoneme-Correspondences

• Applicant: Privatdozentin Dr. Xenia Schmalz
• Subject Area: Developmental and Educational Psychology ; General, Cognitive and Mathematical Psychology ; Human Cognitive and Systems Neuroscience
• Term: from 2018 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 405007295

• Final Report Year: 2024

Final Report Abstract

The first step of learning to read is learning relationships between letters and corresponding sounds. We investigated neurocognitive mechanisms and predictors of letter-sound learning. Learning such relationships may rely partly on extracting statistical regularities that children encounter while reading (statistical learning). (1) We developed an artificial orthography learning (AOL) paradigm: Adult participants learned made-up words (pseudowords) written with unfamiliar characters. We tested the same participants on two AOL tasks and found high correlations, suggesting that it reliably captures symbol-sound learning efficiency. (2) we used EEG to investigate how the brain responds to newly learned characters versus Latin letters that participants process with high proficiency. AOL was unsuitable for investigating the neurocognitive underpinnings: we found no significant differences between the baseline and novel symbol condition or the novel symbol condition and Latin letter condition, though the baseline and Latin letter conditions differed from each other, replicating previous work. (3) We manipulated properties of the symbol-sound correspondences. Some symbols had two possible pronunciations (D - /a/ or /o/). The correct pronunciation was predictable from the context (/o/ when followed by /k/) or unpredictable (/o/ or /a/ with a 50% probability in each context). Participants tended to use different strategies depending on the condition: When all correspondences were predictable, they tended to store letter-sound knowledge as all-or-none rules, and when there was unpredictability their representations tended to be probabilistic. (4) We investigated the correlations between the ability to learn context-dependent rules in the AOL and other cognitive characteristics. We finished data collection in spring 2024 and are analysing the data. (5) We examined if children who have learned the statistical relationships relating to common letter clusters have an advantage in a word learning task. Children showed a slight preference for pseudowords with common letter clusters in a choice task, but this did not affect their spelling ability of these pseudowords and did not correlate with their overall reading or spelling ability. (6) In a theoretical article, we conclude that statistical learning, as a domain-general mechanism, is unlikely to have a strong effect on reading acquisition or dyslexia. We propose the Noisy Chain Hypothesis: we need to consider causal chains when interpreting correlations between reading ability and cognitive skills. The more distal a cause (i.e., when there are many intervening steps), the weaker the correlation. Thus, future research should focuss on investigating proximal causes and reading-related processes to understand which cognitive skills we can capitalise on to facilitate reading acquisition.

Publications

• Developmental Dyslexia, Reading Acquisition, and Statistical Learning: A Sceptic’s Guide. Brain Sciences, 11(9), 1143.
  Schmalz, Xenia; Treccani, Barbara & Mulatti, Claudio
  
• What Do Artificial Orthography Learning Tasks Actually Measure? Correlations Within and Across Tasks. Journal of Cognition, 4(1), 7.
  Schmalz, Xenia; Schulte-Körne, Gerd; De Simone, Elisabetta & Moll, Kristina
  
• Effects of complexity and unpredictability on the learning of an artificial orthography. Cortex, 152, 1-20.
  Schmalz, Xenia; Mulatti, Claudio; Schulte-Körne, Gerd & Moll, Kristina
  
• How do we falsify theories of dyslexia? A simulation study. Talk presented at the International Workshop on Reading and Developmental Dyslexia, San Sebastian, Spain, June, 2023.
  Schmalz, X., Leung, A.Y. & Mulatti, C.
  
• Paaz or Pahz: Orthographic learning and ist relationship to statistical learning, reading, and spelling. Talk presented at the biannual conference of the European Society for Cognitive Psychology, Porto, Portugal, July 2023.
  Schmalz, X., Mehlhase, H., Moll, K., Schulte-Körne, G. & Wang, H.C.
  
• Rules and statistics: What if it’s both? A basic computational model of statistical learning in reading acquisition. Center for Open Science.
  Schmalz, Xenia
  
• The role of syllables and morphemes in silent reading: An eye-tracking study. Quarterly Journal of Experimental Psychology, 76(11), 2493-2513.
  De Simone, Elisabetta; Moll, Kristina; Feldmann, Lisa; Schmalz, Xenia & Beyersmann, Elisabeth
  
• What is a Bayes factor?. Psychological Methods, 28(3), 705-718.
  Schmalz, Xenia; Biurrun, Manresa José & Zhang, Lei
  
• Do Faster Learners Know More? How the Learning of Orthographic Regularities Affects Reading in German Primary School Children. Scientific Studies of Reading, 29(5), 500-520.
  Schmalz, Xenia; Mehlhase, Heike; Schulte-Körne, Gerd; Moll, Kristina & Wang, Hua-Chen