I am here on Zoom to answer your questions about the poster (during the Thursday poster session 12 - 2)! https://mit.zoom.us/j/590389235 Meeting ID: 590 389 235 US : +1 646 558 8656 or +1 669 900 6833 International Numbers: https://mit.zoom.us/u/ac4AjF5H9U Join by SIP 590389235@zoomcrc.com Join by Skype for Business https://mit.zoom.us/skype/590389235 Generic statements (e.g., "Birds fly") convey generalizations over kinds, though these generalizations vary in how generally they apply (e.g., penguins don’t fly). Tessler and Goodman (2019) propose a model where the strength of the generalization implied by a generic depends upon background knowledge of the property being generalized. This background knowledge can have extensive structure that affects the interpretation of the generic, which we investigate using generics of conjunctive predicates (“Ks F and G”). Conjunctive predicates are of particular interest because knowledge of one predicate can influence knowledge of the other via mutual exclusivity (ME). For example, “Elephants live in Africa” implies that most do, whereas “Elephants live in Africa and Asia” implies some live in Africa and others live in Asia, due to the ME of the predicates. In other words, the full conjunctive generic implies different values for the prevalence of elephants living in Africa after the first predicate ("Elephants live in Africa") vs. after the second ("...and Asia"). A simple extension to the generics model of Tessler and Goodman (2019) predicts a listener’s interpreted prevalence will decrease over the course of a conjunctive generic sentence, which we test in Expt. 1. Such a belief change raises the question of when throughout the course of a sentence this change occurs. Human listeners may wait until content words such as “Asia” appear in order to update their beliefs (weak incremental processing); alternatively, listeners could use their syntactic expectations to predict interpretations of generics they haven't yet heard, in which case, they may begin updating their beliefs even at non-content function words such as “and” (strong incremental processing), which we formalize by integrating a syntactic prediction mechanism into the generic interpretation model. We test the strong vs. weak incremental predictions in Expt. 2. We collected human judgments about the implied prevalence of novel generics described in a story book. Each chapter contained background information designed to facilitate the ME inference and included page-breaks that occurred in the middle of sentences. In Expt. 1, we used a gating paradigm in which an implied prevalence question (“How many elephants live in Africa?”, “...in Asia?”) appeared on critical trials after a page break occurring either in the middle of a conjunctive generic before the word “and” or at the end of the conjunctive generic. The generic either described ME (“...live in Africa and Asia”) or non-ME properties (“...live in Africa and eat bugs”). The data in Expt.~1 produced the predicted pattern of prevalence judgments; for ME predicates, implied prevalence of the first predicate decreases after reading the second conjunct; non-ME predicates served as a control condition: implied prevalence was not affected by observing the second conjunct. In Expt. 2, we modified the gating paradigm to test the incrementality of the listener’s belief change. The last sentence of critical trials was always ME (“...live in Africa and Asia”) and was interrupted at three possible points (randomized within-participant): after the first predicate, after “and”, and after the second predicate. Consistent with the strong incremental view, prevalence estimates decreased after “and” as well as after ``Asia’’. This and the previous experiment were reported in Tessler, Gu, & Levy (2019, cogsci). When participants rated implied prevalence in Expts. 1 & 2, they rated prevalence of both the first property and an alternative (e.g., what \% of elephants live in Asia). The explicit query may have brought to mind this alternative and caused the decrease in their prevalence estimates. To control for this, we replicated both experiments while only asking participants about the first property (% live in Africa). Both effects were observed in the replications. Together, these results suggest that listeners’ syntactic expectations integrate in a fine-grained manner with their pragmatic language understanding mechanisms.