Abstract
The recognition of manipulable objects results from the encoding of sensory input in combination with predictive decoding of experience-derived visuomotor information stored in conceptual-semantic representations. This grounded interpretative processing was previously found to subsist even under unaware perception of manipulable object pictures. In this fMRI study, we first aimed to extend this finding by testing whether experientially grounded visuomotor representations are unawarely recruited when manipulable objects are not visually depicted, but only referred to by words presented subliminally through continuous flash suppression. Second, we assessed the generalizability of decoding experience-derived conceptual information to other semantic categories, by extending our investigation to subliminally presented emotion words and testing for unaware recruitment of grounded emotion representations in the limbic system. Univariate analysis of data sampled from 21 human participants (14 females) showed that manipulable object words selectively activated a left-lateralized visuomotor network, both when words were presented below perceptual threshold and when participants subjectively reported lack of stimulus awareness. Emotion words selectively engaged the bilateral limbic network, although univariate analysis did not provide evidence for its recruitment under subliminal perceptual conditions. In turn, multivariate pattern analysis showed that neural codes associated with both manipulable object and emotion words could be decoded even in the absence of perceptual awareness. These findings suggest that the brain automatically engages in conceptual-semantic decoding of experience-derived information not only when circumstances require to interact with manipulable objects and emotions, but also when these referents are dislocated in time and space and only referred to by words.