Using EEG to test if spoken image labels propel subliminal images into consciousness

This preregistered study investigated whether linguistic information can influence low-level visual perception. We argue that many previous studies supporting this view have significant shortcomings. Specifically, these studies conflate measures of objective performance with those of subjective experience, and cannot exclude the possibility that their observed effects might have been mediated by post-sensory processes instead (e.g., semantics, memory). To address these issues, our paradigm combined elements from two published studies by Sekar et al. (2011) and Lupyan and Ward (2013). Participants (N = 19) observed and rated visual targets of fixed intensity corresponding to their 50% identification threshold; visual targets were preceded by nonpredictive speech cues which matched or mismatched the target. Participants carried out an identification task and rated their subjective experience (“didn’t see”, “couldn’t identify”, “unsure”, and “sure”), while event-related potentials were recorded. We found that identification of targets that remained unseen (“didn’t see”) was not facilitated by congruent cues, but was impaired by incongruent cues. In addition, cue congruence did not affect the relative frequency of nonconscious (“didn’t see”) to conscious visual experiences, nor did it affect the amplitude of the Visual Awareness Negativity, an electrophysiological correlate of visual awareness. The results indicate that language cues did not enhance the processing of subliminal visual inputs, nor boost them into conscious experience. This outcome aligns with accounts of low-level perception that exclude the influence of linguistic or conceptual knowledge.