Does the reading brain possess top-down connections from morphemes to letters?

No psycholinguist will contradict the importance of morphemes in language processing. But as a cognitive construct, the morpheme and its operations are not fully mapped out yet. Here, we investigated whether morphemes provide top-down support for their constituent letters: a so-called morpheme superiority effect. In two experiments (N = 100 per experiment), French participants identified single letters in nonword strings that either comprised a French affix (e.g., 'o' in 'cojuste') or not ('o' in 'sojuste'). In Experiment 1 participants provided free-recall responses by typing letters seen at postcued locations. Experiment 2 was a Reicher-Wheeler (i.e., two-alternative forced-choice) version of the same task. Neither experiment evidenced better identification of target letters embedded in affixes than nonaffixes. On the other hand, we did observe an impact of the frequencies of embedding N-grams (i.e., letter chunks). Hence, while the brain learns and represents letter combinations, we did not observe evidence for morpheme representations partaking in the brain's feedback loop. Findings are discussed in light of the recent PONG model of word recognition.