check
קולוקויום חברתי-קוגניטיבי מפגש 6: If you don't have anything nice (or interesting!) to say, don't say anything at all - Hannah Rhode | המחלקה לפסיכולוגיה

קולוקויום חברתי-קוגניטיבי מפגש 6: If you don't have anything nice (or interesting!) to say, don't say anything at all - Hannah Rhode

תאריך: 
ה', 24/12/202012:30-14:00
מיקום: 
https://huji.zoom.us/j/87991423807?pwd=MzJRckNvOGI0M0w3Q0lUS1lQVEM4QT09
מרצה: 
Hannah Rhode

If you don't have anything nice (or interesting!) to say, don't say anything at all
Hannah Rohde
University of Edinburgh

This talk considers the role of predictability in models of communication, asking how knowledge of the world influences people's predictions about the content of what a speaker will sayCommunication depends in part on the hope that people will tell us things we're interested in hearing. In this sense, language can be understood as a channel by which speakers convey, among other things, newsworthy and informative messages (i.e., content that is otherwise unpredictable to the comprehender).  We therefore might expect comprehenders to show a preference for such messagesHowever, research on language comprehension often points to a preference for the opposite (i.e., processing ease for real-world predictable content)Several decades of research have shown that comprehenders can deploy knowledge about situation plausibility to generate fine-grained context-driven predictions about upcoming wordsThis talk introduces evidence that comprehenders can also deploy this knowledge in favor of newsworthy content---distinguishing between expectations about what a speaker is likely to encounter in the world and what they are likely to *say* about it.

This distinction is relevant in considering two domains:  how a speaker says something and whether they choose to say anything at allOne study targets the use of modifiers in referential descriptions, specifically, the presence of color modifiersAlthough bananas are typically yellow in the real world, speakers don't typically mention this yellowness in their descriptionsIn a study testing anticipation of upcoming words, we find that comprehenders who hear a speaker mention an object's color show a preference for an object that has a lower probability of possessing that color in the real worldAnother study asks about the choice to speakFor example, it's not unusual for people to drink a moderate amount of coffee, and so a speaker would only be expected to spontaneously report the number of cups of coffee that someone drank if that number were beyond the ordinary.  Indeed, participants’ estimates of the content of a speaker's beliefs are found to correspond more closely to estimates of real-world priors, whereas estimates about the content of spoken utterances are found to diverge from these priorsFinally, for testing real-time effects, self-paced reading studies measure whether newsworthy messages can be easier to processSentence-final reading times show that situation-atypical (but pragmatically informative) messages can indeed be easier than situation-typical (but pragmatically uninformative) messages.  

These findings are compatible with models that describe comprehension as a process of reverse engineering a speaker's production process. The findings highlight that not only do comprehenders need to update their mental model of the situation being described, but they also need to resolve why a speaker would have chosen to talk about that situation in the first place.