The use and nature of grapheme coding during sub-lexical processing and lexical access

This work aimed to investigate grapheme coding during sub-lexical processing and lexical access. Using the letter detection task in Experiment 1, we compared letter pairs that could be considered as a grapheme unit or not depending on context (referred to as weakly cohesive complex, e.g., an in chant vs cane) to real two-letter graphemes (highly cohesive complex, e.g., au in chaud) and single-letter graphemes (simple, e.g., a in place). Three experimental conditions were used, one of which was designed to prevent phonological influences. Data revealed that only highly cohesive complex graphemes were processed as units, not the weakly cohesive ones. The same pattern was found across experimental conditions in favor of an orthographic mechanism. In Experiments 2 and 3, a primed lexical decision task was used with two stimulus onset asynchronies (SOAs) and two different ranges of lexical frequency. We manipulated the number of graphemes removed from partial primes (d**che vs do**he-DOUCHE) and relatedness. In contrast to Experiment 1, no evidence was provided in favor of a role of graphemes during lexical access. We suggest that graphemes can be conceived as sub-lexical orthographic units per se but can only be captured within a sub-lexical route to reading.