Semantic Prosody and Judgment David J. Hauser University of Michigan Norbert Schwarz University of Southern California Some words tend to co-occur exclusively with a positive or negative context in natural language use, even though such valence patterns are not dictated by definitions or are part of the words’ core meaning. These words contain semantic prosody, a subtle valenced meaning derived from co-occurrence in language. As language and thought are heavily intertwined, we hypothesized that semantic prosody can affect evaluative inferences about related ambiguous concepts. Participants inferred that an ambiguous medical outcome was more negative when it was caused, a verb with negative semantic prosody, than when it was produced, a synonymous verb with no semantic prosody (Studies 1a, 1b). Participants completed sentence fragments in a manner consistent with semantic prosody (Study 2), and semantic prosody affected various other judgments in line with evaluative inferences (estimates of an event’s likelihood in Study 3). Finally, semantic prosody elicited both positive and negative evaluations of outcomes across a large set of semantically prosodic verbs (Study 4). Thus, semantic prosody can exert a strong influence on evaluative judgment. Keywords: semantic prosody, corpus linguistics, social cognition, evaluation, language Why does “work” seem worse when someone causes work for us rather than produces work for us? Some might say that produce and cause mean different things, but the words themselves are largely synonymous— both take outcomes that are brought about to exist (like “work”) as their objects, and both are cross-listed in popular thesauri as being strong synonyms. Yet each word seems to prompt different interpretations of “work,” with caused work seeming additional and burdensome, and produced work seeming like a provided opportunity. Why do these synonymous words color “work” with such different valences? Analyses of the co-occurrence of words in text and natural language have shown that some words occur predominantly in contexts with strong negative or positive valence (Louw, 1993; Partington, 2004; Sinclair, 1991; Stubbs, 1995). Frequent co- occurrence, in turn, can give rise to the expectation that the context is likely to reflect the usually associated valence whenever the word is encountered. These expectations are not inherent in the word’s ascribed definition (Partington, 2004; Stubbs, 1995) and are not drawn upon when native speakers are asked to consider a word’s meaning in isolation (see review in Xiao & McEnery, 2006). Linguists refer to this phenomenon as semantic prosody, 1 which denotes the covert valenced connotation of a word derived from frequent co-occurrence in natural language. Language and thought are heavily intertwined, such that minor variations in wording can exact profound effects on judgments and memory. Asking people how they feel about themselves leads them to more negativity than asking how they think about them- selves (Holtgraves, 2015); accidents in which cars were said to smash into one another are recalled as more violent than accidents in which cars hit one another (Loftus & Palmer, 1974); and saying Daniel helps X elicits fewer dispositional attributions of Daniel’s helpfulness than saying Daniel is helpful (Semin & Fiedler, 1991). Nearly synonymous ways to express the same information can lead the reader to very different inferences. We therefore predict that the valence of a word’s typical co-occurrences (i.e., semantic prosody) can also influence judgment, affecting evaluative infer- ences and creating disparate valence implications for similar sen- tences as illustrated in our opening example. Semantic Prosody Lexical priming theories of language suggest that context is key to concept representation (Hoey, 2005). Words do not occur in isolation but appear in context with critical links to other elements of a sentence (Casasanto & Lupyan, 2015; Elman, 2011). The typical context in which a given word appears allows readers to infer attributes of the word that go beyond its lexical definition. 1 Additional terms for the phenomenon include discourse prosody, eval- uative prosody, and semantic preference. Note that the term prosody here is used metaphorically. Semantic prosody does not directly involve speech patterns of stress or intonation; rather, it references them. Just as the speech intonation of vowels can depend upon neighboring letters, the semantic profile of words can depend upon neighboring words; hence, the use of the use of the term prosody. This article was published Online First May 30, 2016. David J. Hauser, Department of Psychology, University of Michigan; Norbert Schwarz, Mind & Society Center, Department of Psychology and Marshall School of Business, University of Southern California. We thank Matthew Brook O’Donnell, Nick Ellis, and the members of the UMich OLab for their valuable insight, and Anni Subar and Meghan Brown for their assistance with Study 5. Correspondence concerning this article should be addressed to David J. Hauser, Department of Psychology, 3233 East Hall, 530 Church Street, Ann Arbor, MI 48109-1043. E-mail: djhauser@umich.edu This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Journal of Experimental Psychology: General © 2016 American Psychological Association 2016, Vol. 145, No. 7, 882– 896 0096-3445/16/$12.00 http://dx.doi.org/10.1037/xge0000178 882