Anouk Dieuleveut, Emmanuel Chemla & Benjamin Spector (to appear). “Distinctions between primary and secondary scalar implicatures”. Journal of Memory and Language.

Main information:
Full paper in pdf
BibTeX reference
Experiment 1 Experiment 2 Experiment 3 (dual task)
R script for the main analyses R script for the main analyses R script for the main analyses
Result file Result file Result file (multiple files in a zip folder)

R package to decode experimental results from Ibex

Abstract: An utterance of Some of the cards are hearts usually triggers the inference that the speaker does not have the belief that all cards are hearts (primary scalar implicature). It often licenses a stronger inference: that the speaker believes that not all cards are hearts (secondary SI). Previous experimental studies were not designed to detect three distinct readings (literal reading, primary implicature, secondary implicature), but only the existence of a strengthened reading on top of the literal reading. We present an experimental paradigm which, contrary to previous studies, is able to detect these three distinct readings as such. Furthermore, we show that the secondary (strong) reading can be accessed even in contexts where the speaker is not presented as being well-informed, which goes against the classical neo-Gricean pragmatic approaches to Scalar Implicature. This finding is compatible with both the 'grammatical' approach to Scalar Implicatures and some more recent game-theoretic pragmatic models in which speakers and hearers engage in sophisticated higher-order reasoning about each other. We use our paradigm to test standard scalar items such as some and expressions whose interpretation has been argued to involve SIs, but for which this conclusion is controversial: almost, numerals and plural morphology. We find that a) speakers do access three distinct readings for standard scalar items and for almost, and b) participants seem to only access the literal reading and the secondary implicature reading (and no primary implicature reading) in the case of numerals and plural morphology, which suggests that the pragmatic and semantic mechanisms at play are different for both types of items.