Semantic information modulates online morphological learning: neuromagnetic investigation

Morphemes are defined as the smallest meaningful units of a language. They can, among other things, be used to form new words (i.e., derivation, e.g., -ness in darkness) or convey grammatical information (i.e., inflection, e.g., plural -s in houses). Learning the meaning of morphemes and recognizing morphemic boundaries is thus an essential part of language learning, however, neural mechanisms underlying the acquisition of morphemes remain largely unexplored. Here, we investigated the online build-up of memory traces for novel bound morphemes (suffixes) in the left perisylvian language network. Half of the novel suffixes were trained semantically through a picture-word association task and half were untrained, i.e., not provided with a semantic referent. Following the semantic training with twenty repetitions per suffix, we used magnetoencephalography (MEG) to record cortical responses to the novel semantically trained, novel untrained, and real suffixes combined with real and pseudoword stems in a 20-minute-long repetitive exposure task. The participants were exposed to spoken stimuli passively, i.e., without any stimulus-related task, while their attention was diverted away from the auditory input. To assess the online learning, we compared the brain’s neuromagnetic responses to the different suffixes in the beginning and the end of this passive exposure. Our results showed that semantically trained suffixes demonstrated a consistent activation pattern throughout the learning task, suggesting that repetitive exposure did not further facilitate learning after the formation of initial episodic memories during the semantic training. By contrast, neural source activation to untrained suffixes combined with real stems increased over the course of the passive exposure, reflecting a build-up of suffix-related memory traces. Online learning related changes were present in all regions of interest in the left hemisphere (inferior frontal gyrus, anterior temporal lobe, posterior superior temporal gyrus, posterior middle temporal gyrus. The distinction in the neural source activity patterns between the trained and untrained novel suffix types indicates that semantic information can give a head start to the suffix learning process. Furthermore, in the context of a real word stem, neural source activation elicited by both novel suffix types in the late phase of the passive exposure reached the early phase activation elicited by real suffixes. Our results suggest immediate learning effects to novel suffixes gained through a brief semantic learning as well as more incremental suffix learning effects obtained through purely perceptual exposure.