ABSTRACT
Expert readers have a wide tolerance for distortions of the letters that make up a word. Nevertheless, the limits of this invariance are still under debate. To scrutinise this issue, we focused on a single parameter, letter rotation, as it serves to disentangle the predictions from neurally-inspired models of word recognition. Whereas the Local-Combination-Detector (LCD) model predicts invariance up to 45°, the SERIOL model predicts a linear cost until 60°. To test these predictions, Experiments 1 and 2 employed four rotation angles (0°, 22.5°, 45°, 67.5°) in lexical decision and semantic categorisation. The cost was minimal at 22.5°, sizeable at 45°, and considerably large at 67.5°. In Experiment 3, we focused on four moderate rotation angles (<45°). We found a gradual reading cost that increased at 45°. Thus, while there is a resilience limit around 45° favouring LCD, less steep angles also produce a reading cost, backing the SERIOL model.
Acknowledgments
This study was supported by the Spanish Ministry of Science and Innovation (grants PRE2018-083922, PSI2017-86210-P, and PID2020-116740GB-I00 [MCIN/ AEI/10.13039/501100011033]). We would like to thank Andrew Aschenbrenner and Melvin Yap for sharing their word and pseudoword stimuli with us.
Disclosure statement
No potential conflict of interest was reported by the author(s).