Dog whistles, or messages that are communicated to only a subset of a speaker’s audience, pose a problem for semantics because they are “simultaneously conventional and socially dependent” (Henderson & McCready, 2017, 2018; cf Saul 2018). We argue that this problem can be solved through the framework of ethnographically-informed frame semantics (Fillmore 1982), that commits to a context-rich and socially mediated semantics. We argue that dog whistles are not semantically anomalous because: (1) all meanings are conventionalized locally through engagement with particular communities of practice and their related ideologies (Eckert & McConnell-Ginet 1992); (2) meaning is interpreted through recognition of many contextual cues that together instantiate a type of scene one has previously experienced (Fillmore 1982). According to Fillmore, frame semantics is fundamentally a semantics of “understanding” rather than “truth,” where meaning depends on the perceiver’s prior knowledge about contextual use of language, which is both a social and conventional phenomenon. Different people have different levels of expertise with particular areas of cultural knowledge that underlie lexical concepts; thus, a word will not evoke or index the same social discourse for those who have not experienced its use in particular contexts (Silverstein 2006). We argue that the attribution of “dog whistle” depends on the recognition that a phrase can ambiguously evoke multiple semantic frames associated with different areas of cultural knowledge. We also discuss why some phrases which have been previously analyzed as dog whistles are now arguably overtly connected to an insidious message (“welfare”), while some remain ambiguous (“coastal elites”). To do so, we employ two scales: level of conventionality and size of social domain (Agha 2005). We aim to show how dog whistles are not semantically exceptional, and that their insidious rhetorical effect can be best understood through an interdisciplinary cognitive semantic and anthropological lens.