Testing a hypothesis of multi-layer network of orthographic neighbors via a novel measurement of orthographic knowledge

Testing a hypothesis of multi-layer network of orthographic neighbors via a novel measurement of orthographic knowledge

First Author: Bowen Wang -- University of California, Berkeley, Graduate School of Education
Additional authors/chairs: 
Anne Cunningham
Keywords: Orthographic Knowledge, Orthographic representations, Spelling, Measurement, Item Response Theory
Abstract / Summary: 

Orthographic neighbors are usually defined as words produced by changing one letter of the target word (Perea & Gomez, 2010). However, neighbors themselves also have neighbors. Thus, a word’s neighborhood could be viewed as a network with multiple layers. The 1st-layer neighbors differ from the word by one letter; the 2nd-layer neighbors differ from the 1st-layer neighbors by one letter, etc. We hypothesize that a learner’s initial orthographic representation of a word might be in an outer layer of the neighborhood and gradually move towards the inner layers.
To test this hypothesis, we designed a measure that examines orthographic knowledge in a more nuanced way. In our recall tasks, learners’ spelling is scored based on its Levenshtein distance from the target word (i.e., the least number of single-letter changes needed to change a word into another via addition, deletion, or substitution) instead of in an all-or-nothing way. Our recognition tasks ask learners to choose the word (e.g., “kindergarten”) from five 1st-layer neighbors (e.g., “kindergarden”), five 2nd-layer neighbors (e.g., “kindegarden”), and five 3rd-layer neighbors (e.g., “kidegarden”), different from extant measures where learners choose from only two options (e.g., Conrad, 2008).
This test has been administered to 59 English learners in China. Its reliability and validity are supported by item response data analysis. We found that across items, learners’ mean locations consistently increased from the lowest step (spell or choose a 3rd-layer neighbor) to the highest step (correct word), which supports our hypothesis. The next stage of research in Spring, 2020 will utilize 2nd-grade native speakers of English.
Conrad, N. J. (2008). From reading to spelling and spelling to reading: Transfer goes both ways. Journal of Educational Psychology, 100(4), 869-878.
Perea, M., & Gomez, P. (2010). Does LGHT prime DARK? Masked associative priming with addition neighbors. Memory & cognition, 38(4), 513-518.