Are speakers and listeners 'only moderately Gricean'? An empirical response to Engelhardt et al. (2006)

If hearers are sensitive to Gricean maxims of Quantity (Grice, 1975/1989), they should disfavour expressions which give too little or too much information for the unique identification of an intended referent. Accordingly, cooperative speakers are expected to provide all and only as much information as is necessary for their interlocutor to uniquely identify a referent. Engelhardt et al. (2006) report that speakers and hearers are sensitive to under-informativeness but not to over-informativeness. Based on this finding, the authors re-interpret the literature which claims to document pragmatic effects in language comprehension and instead attribute previous findings to structural-lexical biases. We argue that the reason why speakers and hearers seemed insensitive to over-informativeness in Engelhardt et al.'s studies was because certain aspects of their experiments favoured the use of redundant information. Our experiments 1 and 2 manipulate these factors, revealing that hearers are in fact sensitive to violations of over- as well as under-informativeness. A further production experiment shows that speakers do not under- or over-specify when the factors that favoured over-informativeness in Engelhardt et al.'s study are removed. The findings provide evidence that speakers and hearers are sensitive to both Quantity maxims, and suggest that the effects obtained in previous literature should indeed be attributed to pragmatic factors.