[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83818-en":3,"doc-seo-83818-105":28,"detail-sidebar-cat-0-en-105":89},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":13,"seo_description":14,"update_tm":26,"read_time":27},83818,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","Failures and Successes to Learn a Core Conceptual Distinction from the Statistics of Language","Generic statements like “tigers are striped” and “cars have radios” convey information that is generally true, yet they differ from claims that are only statistically true. People can distinguish principled from statistical properties, and prior work argues this ability is not learnable from experience. This study tests whether the principled-vs-statistical distinction can be learned from language itself. Language models track statistical prevalence but struggle to represent the distinction after controlling for prevalence; only GPT-4 succeeds.","arXiv :2607 .04523v 1 [ cs .CL] 5 Jul 2026  \nFailures and Successes to Learn a Core Conceptual Distinction from the  \nStatistics of Language†  \nZhimin Hu1 , Jeroen van Paridon1 , and Gary Lupyan*1  \n*corresponding author  \n1Department of Psychology, University of Wisconsin-Madison, Madison, USA  \nAbstract  \nGeneric statements like “tigers are striped” and “cars have radios” communicate information that is, in general, true. However, while the first statement is true *in principle*, the second is true only statistically. People are exquisitely sensitive to this principled-vs-statistical distinction. It has been argued that this ability to distinguish between something being true by virtue of it being a category member versus being true because of mere statistical regularity, is a general property of people’s conceptual machinery and cannot itself be learned. We investigate whether the distinction between principled and statistical properties can be learned from language itself. If so, it raises the possibility that language experience can bootstrap core conceptual distinctions and that it is possible to learn sophisticated causal models directly from language. We find that language models are all sensitive to statistical prevalence, but struggle with representing the principled-vs-statistical distinction controlling for prevalence. Until GPT-4, which succeeds.  \nKeywords: distributional semantics; generics; world models  \n1. Introduction  \nPeople interpret generic statements such as airplanes have wings and dogs bark to mean that the named property is, in general, true of the category (Hollander et al., 2009) . Other statements of this form, however, such as airplanes carry passengers and dogs wear collars, while also being judged as generally true, have a decidedly different quality. In a series of papers, Prasada and colleagues (Prasada & Dillingham, 2006; Prasada, 2016; Prasada et al., 2013) drew a distinction between generics that express principled properties and generics that express merely statistical properties. A statement expressing a principled property, such as airplanes have wings, retains its truthfulness when asked whether it is true because of  \n†Published at the 15th International Conference on the Evolution of Language (Evolang XV) .  \n(or by virtue of) being that thing. For example, in the experiments we describe below, on a scale of-3 = completely false to +3 = completely true, people judged the statement airplanes have wings with a mean of 2.9 . This declines only slightly if asked whether it is true that airplanes have wings because they are airplanes (M=2.6) . A statement like airplanes have passengers is judged to also be mostly true (M=1.8), but if asked whether airplanes have passengers because they are airplanes, the truth estimate drops (M=0.6) . Importantly, this key result remains when one controls for confounds such as prevalence and cue-validity, showing that it is not simply an artifact of principled connections being more common or it being harder to come up with counter-examples.  \nResults like these have been used to argue that people’s ability to distinguish between principled and statistical generics requires an a priori sensitivity to a distinction between statistical vs. “in-principle” properties. Because there are no structural differences between generics that could inform this distinction, it is thought that the distinction cannot be learned through associations (see Prasada et al., 2013; Haward, Wagner, Carey, & Prasada, 2018), and perhaps cannot even be represented by an associative mechanism (Prasada, 2021) .  \nHowever, even though generic statements do not encode the principled/statistical distinction in their structure, the distinction might still be captured in the distributional structure of language itself. In this study, we investigated whether the statistical/generic distinction is recoverable from the statistics of language. We did this by predicting human judgments of gener","cbCaifmYjV889iuA","https://ap.wps.com/l/cbCaifmYjV889iuA","pdf",147627,1,"English","en",105,"# Abstract\n# Introduction\n# Human ratings\n## Participants\n## Procedure","[{\"question\":\"What core distinction does the paper investigate in generic statements?\",\"answer\":\"It investigates whether generic statements reflect principled properties (true because of category membership) versus merely statistical properties (true due to prevalence patterns).\"},{\"question\":\"How do the authors test whether the distinction can be learned from language?\",\"answer\":\"They predict human judgments of generic statements using judgments derived from distributional language models, then assess whether the models capture the principled-vs-statistical split after accounting for confounds like prevalence and cue-validity.\"},{\"question\":\"What is the main finding about current language models?\",\"answer\":\"All tested language models are sensitive to item prevalence, but they generally struggle to represent the principled-vs-statistical distinction when controlling for prevalence; GPT-4 succeeds.\"}]",1784190612,20,{"code":4,"msg":29,"data":30},"ok",{"site_id":23,"language":22,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":26},"failures-and-successes-to-learn-a-core-conceptual-distinction-from-the-statistics-of-language","",{"@graph":34,"@context":83},[35,52,66],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/research-report/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/failures-and-successes-to-learn-a-core-conceptual-distinction-from-the-statistics-of-language/83818/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":22,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"What core distinction does the paper investigate in generic statements?","Question",{"text":73,"@type":74},"It investigates whether generic statements reflect principled properties (true because of category membership) versus merely statistical properties (true due to prevalence patterns).","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"How do the authors test whether the distinction can be learned from language?",{"text":78,"@type":74},"They predict human judgments of generic statements using judgments derived from distributional language models, then assess whether the models capture the principled-vs-statistical split after accounting for confounds like prevalence and cue-validity.",{"name":80,"@type":71,"acceptedAnswer":81},"What is the main finding about current language models?",{"text":82,"@type":74},"All tested language models are sensitive to item prevalence, but they generally struggle to represent the principled-vs-statistical distinction when controlling for prevalence; GPT-4 succeeds.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,118,121,125,128,132],{"id":20,"doc_module":4,"doc_module_name":44,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":45,"doc_module":4,"doc_module_name":44,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":44,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":44,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":44,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":44,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":44,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":44,"category_name":123,"show_sort_weight":27,"slug":124},9,"Religion & Spirituality","religion-spirituality",{"id":27,"doc_module":4,"doc_module_name":44,"category_name":126,"show_sort_weight":27,"slug":127},"World Cup","world-cup",{"id":129,"doc_module":4,"doc_module_name":44,"category_name":130,"show_sort_weight":129,"slug":131},10,"Lifestyle","lifestyle",{"id":133,"doc_module":4,"doc_module_name":44,"category_name":134,"show_sort_weight":104,"slug":135},19,"General","general"]