Purchase this article with an account.
David Rothlein, Brenda Rapp; Behavioral and neural evidence of stored letter shape and abstract letter identity representations. Journal of Vision 2015;15(12):913. doi: 10.1167/15.12.913.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
What are the learned representations that contribute to object identification? Using letter stimuli in behavioral and fMRI paradigms, we provide evidence for the role of stored “standard” letter shapes and abstract letter identity representations, also identifying their neural substrates within left ventral cortex. The behavioral experiments involved presentation of letter pairs in an unusual novel font; stimuli were 33 letter shapes, including letters that differed in font or case. Participants (n= 90) performed similarity judgments (Experiments 1a/b) and same/different physical identity judgments (Experiments 2a/b). To control for stimulus visual similarity and isolate the contribution of learned representations, stimuli were presented both upright and also rotated (90°ccw, reflected across the vertical axis) to reduce identifiability while maintaining pixel-wise letter similarities across upright/rotated conditions. Pairwise similarity judgments, reaction times and accuracies were correlated with representational similarity matrices representing: pixel-overlap, similarity in a standard font, letter identity (across case/font). The assumption is that matrix correlations greater for the upright than rotated stimuli represent factors not reducible to stimulus similarity. Results revealed: pixel overlap correlations were comparable (r=.65/.63) for upright and rotated stimuli, while matrix correlations for standard shape and letter identity were greater for upright (r=.63; .59) than rotated (r=.34;.30) stimuli. The findings provide evidence for the role of stored standard letter shapes and cross font/case identity in letter identification. In the fMRI experiment, participants (n=11) viewed single upright letters in the novel font and underwent retinotopic mapping. MVPA computed the pairwise similarity of the multivoxel activation patterns which were compared to similarity matrices representing: stimulus similarity (from Experiments 1/2), standard font similarity and letter identity. Results indicated neural representation of: stimulus similarity in bilateral retinotopically defined cortex (V1, V2v, V3v, V4), standard font similarity left lateralized in LO, pF, and the mid fusiform and letter identity in the left mid fusiform.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only