6 thoughts on “What primes what – An experimental framework to explore alternatives for Scalar Implicatures”

  1. I wanted to add to Nadine’s question and elaborate on Paul’s answer: Although the effectiveness of str-alt primes do not disconfirm the structural theory, that theory says that, if anything, the nw-alt primes should be at least as effective, if not more effective than str-alt.

    But as Paul says in the talk, it maybe that these are equally effective and what drives the different rates has to do with qud. Nevertheless, until we have a solid qud story, the structural account is challenged

    1. Thank you for this elaboration! That is what I thought as well, that this account would not predict the difference between nw-alt and str-alt you find. Also, I was wondering if this difference and the difference between ad hoc/lexical scales could be made sense of in terms of priming retrieval of alternatives through context search versus priming lexical substitution? (maybe this be connected to the qud story…)

  2. I wanted to add to Paul’s reply to Judith’s question. First, I think we can agree that, in principle, both factors could be at play, working together but maybe only for some and numbers? We can look at the ad hocs case. These are trials where lexical priming would not be at play. And here we would need to tell a qud story (especially as the Str alt is more complex than assertion and this would disfavour the SI, in the RSA framework).

    We can also look at the case where we had Alt primes. These are also trials where lexical priming would not be at play. We made some conjectures (to be substantiated) that Str-Alt primes work better than NW-alt primes precisely because they bias more one kind of qud than another.

  3. I also wanted to add to Paul’s reply to Judith’s question. First, I think we can agree that, in principle, both factors could be at play, working together but maybe only for some and numbers? We can look at the ad hocs case. These are trials where lexical priming would not be at play. And here we would need to tell a qud story (especially as the Str alt is more complex than assertion and this would disfavour the SI, in the RSA framework).

    We can also look at the case where we had Alt primes. These are also trials where lexical priming would not be at play. We made some conjectures (to be substantiated) that Str-Alt primes work better than NW-alt primes precisely because they bias more one kind of qud than another.

    1. Following up to [Richard’s reply to [Paul’s reply to [Judith’s question]]], I think this depends on what what forms we think ‘lexical priming’ could take? You might think in the ad-hoc case, for example, strong priming just increases listener expectation that the LF of “There is an X” sentences is something like exh(there is an X).

      More generally, it seems like there are two ways you could go about fleshing out a QUD uncertainty story: approach 1 is that QUDs are associated with different sets of alternatives, and priming modulates production expectation about those alternatives (though I wonder whether there’s a tension between this approach and your mixed findings w/r/t alternative priming effects, which replicates what Judith & I found). Approach 2 is that QUDs are associated with different form->meaning mappings, an approach used by Justine Kao and collaborators to model metaphoric and other non-literal interpretation. But that second approach – I think – could just end up being a notational variant of the lexical uncertainty analysis that Judith and I propose.

      The Kao et al. paper: https://www.pnas.org/content/111/33/12002.full.pdf

      1. I should also mention, in the 2020 paper we do say that our pattern of results is consistent with two separate levers being pulled: modulation vis a vis production expectation of alternative forms, and modulation vis a vis form-meaning mapping expectations.

Leave a Comment or Question Below