Deciphering the electrophysiological signature of discourse connectives

The discourse connectives but and so trigger specific kinds of inferences, which are characterizable as contrastive and causal, respectively. Inspired by Diane Blakemore’s notion of procedural meaning, we carry out two studies while relying on a paradigm we developed in previous work, in which 1) connectives are part of thematically bare sentences presented as a word-finding game, e.g. There is a B but there is no T , and; 2) these two connectives are compared to the less informative conjunction and . Experiment 1 instantiated our behavioral paradigm while imposing the kind of constraints one finds in an EEG study (e.g., the discourse connective is isolated and presented for a predetermined duration). The results from Experiment 1 replicated our previous behavioral findings, confirming the design's reliability. Experiment 2, an ERP study based on this modified paradigm, yielded three main findings. 1) relative to and , the discourse connectives but and so trigger a more pronounced P200 followed by Positive Slow Waves (PSW) of greater amplitude. These data point to extra inferential processing triggered by but and so relative to and. 2) the higher amplitude of the P3b component linked to negated segments following but indicates that these were anticipated more than those that follow and or so . 3) post-connective affirmative segments were, surprisingly, linked with a more pronounced P600-like component in and -trials relative to the but - and so -trials. Overall, this study allows one to more deeply appreciate the inferential processing linked to discourse connectives. • We report the online processing of “but”, “so” and “and” in minimal sentences. • “But” and “so” prompt steeper P200s than “and”. • Compared to “and”, “but” and “so” are deemed more relevant for future processing. • Negations following “but” show indications that an expectation has been fulfilled.