Two new papers on recognition of Japanese kana

In collaboration with Roderick Gawthrop and Sachiko Kinoshita (Macquarie University) and Rinus Verdonschot (Hiroshima University), I've recently had two papers published that examine the process of abstraction in letter recognition. In English, there are two forms for each letter: uppercase and lowercase (e.g., A and a), which are referred to as allographs. One indication of the abstract link between allographs that share a letter identity is that presenting either allograph primes processing of the other, regardless of whether they look similar (C and c) or dissimilar (A and a). Other orthographies, such as Japanese kana, also have multiple allographs representing a single identity. We asked whether the allograph forms of kana behave similarity to the allograph forms of English letters by looking for evidence of kana priming that is not dependent on visual similarity. For example, does あ prime ア (visually-dissimilar allographs) to the same degree that り primes リ (visually-similar allographs) ? We found that allograph priming in kana is based on abstract representations, just like allograph priming in English. We suggest that this abstraction process is a general solution to the problem of mapping across two visual forms in orthographies with this property.

Read the papers and see the raw data here and here.



Schubert, T., Gawthrop, R., & Kinoshita, S. (2018). Evidence for cross-script abstract identities in learners of Japanese kana. Memory & Cognition, 46(6), 1010-21. doi: 10.3758/s13421-018-0818-4

Kinoshita, S., Schubert, T., & Verdonschot, R. (2018). Allograph priming is based on abstract letter identities: Evidence from Japanese kana. Journal of Experimental Psychology: Learning, Memory, and Cognition. doi: 10.1037/xlm0000563

New paper in press at Cortex: Schubert, Reilhac & McCloskey

With collaborators at (and formerly at) Johns Hopkins University, we've just had a paper accepted at Cortex, entitled "Knowledge about writing influences reading: Dynamic visual information about letter production facilitates letter identification." This is a cognitive neuropsychological study of an individual we call NGN, who had difficulty identifying letters shown to him. He might look at an 'B' and call it 'R', for example. However, we found that tracing out a letter helped him to identify it more accurately. Most importantly, the type of tracing mattered- if you traced the letter in a standard way (for B: writing the vertical, then the top loop, then the bottom loop), he improved. But if you traced the letter in reverse, or the letter shape appeared over time in a random sequence, his performance did not improve. We concluded that in addition to processing static letters (like those on this page), the letter recognition system can also process dynamic input. And most critically, crosstalk between the letter production system (your knowledge of how a letter is written) and the recognition system provide a boost when the dynamic input matches your stored letter production plans.

Read the paper and see videos of sample stimuli here.


Schubert, T., Reilhac, C., & McCloskey, M. (in press). Knowledge about writing influences reading: Dynamic visual information about letter production facilitates letter identification. Cortex. doi: 10.1016/j.cortex.2018.03.020