What do we do with what we learn? Statistical learning of orthographic regularities impacts written word processing

Abstract

Individuals rapidly become sensitive to recurrent patterns present in the environment and this occurs in many situations. However, evidence of a role for statistical learning of orthographic regularities in reading is mixed, and its role has peripheral status in current theories of visual word recognition. Additionally, exactly which regularities readers learn to be sensitive to is still unclear. To address these two issues, three experiments were conducted with artificial scripts. In Experiments 1a and 1b, participants were exposed to a flow of artificial words (five characters) for a few minutes, with either two or four bigrams occurring very frequently. In Experiment 2, exposure took place over several days while participants had to learn the orthographic and phonological forms of new words entailing or not frequent bigrams. Sensitivity to these regularities was then tested in a wordlikeness task. Finally, participants performed a letter detection task, with letters being either of high frequency or not in the exposure phase. The results of the wordlikeness task showed that after only a few minutes, readers become sensitive to the positional frequency of letter clusters and to bigram frequency beyond single letter frequency. Moreover, this new knowledge influenced the performance in the letter detection task, with high-frequency letters being detected more rapidly than low-frequency ones. We discuss the implications of such results for models of orthographic encoding and reading.

Keywords: Artificial script; Letter identification; Orthographic regularities; Reading; Statistical learning.