Second thread about (relatively) recent work from our lab- now up: Samara,Singh & Wonnacott (2019) https://doi.org/10.1016/j.cognition.2018.09.005 with @Anna_D_Samara @dnl_singh 1/9

Backgrd: Acquiring literacy is a recent human development, but studies of spelling errors+ nonword spelling suggest a role for implicit statistical learning process akin to those for spoken lang. acquisition, (Saffran et al. 1996 etc.) 2/9

We demonstrated this learning process directly (see also Samara and Caravolas 2014) in an experiment with 7yr-olds in two linguistic contexts (Eng., N=78; Turkish N=37) 3/9

Under incidental learning conditions, children viewed written novel word stimuli embedding “graphotactics” (e.g. “t” never follows “e”); At test, their judgements of unattested stimuli revealed they had picked up these constraints 4/9

We also contrasted word-initial v word-final units (given theories predicting greater learning in the latter, e.g. Goswami & Bryant,1990): There was subst. evidence for learning in both conditions but no evidence of difference. 5/9

Since no evidence of difference doesn’t mean evidence of no difference, we also ran BF analysis (following method from Dienes 2016)- unfortunately this found the evidence was ambiguous (B=.8) 6/9 ....

....analyses assuming the error reduces in proportion to √SE suggest that it could require > 200 (!) participants to establish a null. This suggests our paradigm isn’t sufficiently sensitive to detect between condition differences. 7/9

TAKE HOME: 7yr-olds show implicit statistical learning over written materials, learning graphotatic constraints in both word-initial & word-final positions 8/9

No-paywall version of paper: https://psyarxiv.com/yz4tr/ ; open access data: https://psyarxiv.com/yz4tr/  and full analysis script @Anna_D_Samara https://osf.io/vxqub/  9/9