Misused linguistic evidence in a study of media bias

Abstract

This paper examines the methodological challenges of using the grammatical contrast between active and passive clauses as a proxy for ideological bias in computational discourse analysis. Paying particular attention to a recent analysis of New York Times coverage of the Israeli-Palestinian conflict, we note the conflation of syntactic categories like “passive” with semantic properties like “agentless,” and we argue that conclusions that equate the two require a warrant that is often not provided. We outline construct-valid alternatives for investigating agency and responsibility in news discourse, emphasizing the independence of grammatical considerations from the semantic matter of agent identification. Our analysis contributes to methodological debates about the integration of computational tools in critical discourse analysis.