Cancelation Resistant PCIs Janneke Huitink Jennifer Spenader Dept. of Philosophy Center for Language and Cognition University of Nijmegen University of Groningen j.huitink@phil.kun.nl j.k.spenader@let.rug.nl 1. Introduction An implicature i that arises from an utterance U is cancelable if U is consistent with not i. If i is a can- celable implicature, then it is a conversational impli- cature. However, some particularized conversational implicatures (PCIs) cannot straightforwardly be can- celed. Imagine that Mr. X is applying for a philosophy position and his teacher is writing him the following letter of recommendation: (1) Mr. X’s command of English is excellent and his attendance at tutorials has been regular. He is a brilliant philosopher. The first sentence of (1) carries the PCI that Mr. X isn’t a brilliant philosopher, which we call i. The second sentence is a cancelation phrase, i.e. its lit- eral content is not i. Generally, this phrase should be able to cancel the implicature. But the discourse in (1) allows for two different interpretations. Under the first interpretation the implicature of the first sen- tence is indeed canceled. But under the second inter- pretation, the cancelation phrase itself is interpreted non-literally. The speaker is understood as exploiting the maxim of Quality, i.e. the second sentence is read ironically. As such, the second sentence confirms the implicature of the first sentence. We will call PCIs like the PCI of the first sentence of (1) cancelation resis- tant PCIs. In this paper we argue that these cancelation resis- tant PCIs have become semi-conventionalized. That is, in specific contexts, the implicature is so strongly attached to what is literally said, that canceling the im- plicature becomes very difficult. Hearers often refuse to accept an explicit cancelation, instead preferring to interpret the speaker as being ironic. 2. Background 2.1. Source of PCIs Conversational implicatures arise under the as- sumption that speakers adhere to the Cooperative Prin- ciple (CP): “make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged” (Grice, 1975, 45). Particu- larized conversational implicatures are context depen- dent inferences based on the assumption that speak- ers are being cooperative, following Gricean maxims. The same utterance may give rise to different PCIs in different contexts. This sets PCIs apart from gener- alized conversational implicatures (GCIs). 1 Consider the following examples from (Levinson, 2000, 16-17) that illustrate the difference: (2) A: What time is it? B: Some of the guests are already leaving. PCI: It must be late. (3) A: Where is John? B: Some of the guests are already leaving. PCI: Perhaps John has already left. B’s utterance carries different PCIs in (2) and (3), yet the same GCI arises: not all the guests are already leaving. (Grice, 1975) identified three ways in which implicatures can be derived. 2 Implicatures can arise by speakers exploiting or flouting a maxim, by a clash between maxims, where a speaker breaks one maxim in order to avoid violating another maxim, or, in some cases implicatures may arise without breaking any maxims at all. 1 Note that some researchers ((Carston, 2004), (Hirschberg, 1985)) make no distinction between PCIs and GCIs, considering it to be more a difference of degree than type, but see (Levinson, 2000) for a defense of the distinction. 2 For alternative ways of deriving implicatures, see for instance (Green and Carberry, 1992) for an account based on discourse relations, or (van Rooy, 2003) for an analysis based on exhaustification of answers. Since we know of no alternative means to derive the implicatures of all the examples we discuss, we take Grice’s theory as our point of departure, even though Grice is not explicit on many points.