Entailment and implicature in the language of thought

Human language processing involves constant computation of entailments and implicatures, yetthe extent to which these inferences depend on linguistic communication remains unclear. Thisstudy investigates whether logical structures akin to those in natural language can be derived frompurely visual input. Specifically, we examine the polarity sensitivity of scalar inferences (SIs) innon-linguistic contexts. Prior research has established that SIs are preferentially computed inUpward Entailing (UE) rather than Downward Entailing (DE) environments, and we extend thosestudies by examining whether this pattern extends to visual conceptualization. We conducted threeexperiments using an oddball paradigm with minimal instructions to our participants, where theywitenessed animated ball collisions. Across studies, participants consistently detected thesechanges to the quantified structure of events (changes from some to all) more readily in UE thanin DE conditions, mirroring SI patterns observed in language. These findings suggest that theconceptualization of visual scenes parallels linguistic representations, implying that entailmentrelations are not exclusive to communication but emerge naturally from cognitive structures. Ourresults support grammaticalized theories of implicatures, challenging traditional Gricean viewsthat implicatures stem solely from communicative intent. This work contributes to ongoing debateson the relationship between language and thought, demonstrating that logical inferences can ariseindependently of linguistic input.