We present a novel method for tracking the evolution of political dogwhistles—messages which are only understood by a select in-group, while going unnoticed by others (out-group)—in digital environments. Tracking dogwhistles poses a unique empirical challenge due to their reliance on linguistic ambiguity and intentional concealment. To address this, our method combines computational semantics and survey methodology. We model the contextual distribution of dogwhistle terms in online discussion forums, enabling us to infer semantic representations over time, including interpretations not universally recognized by all readers. Diachronic word embeddings are compared with data from a linguistic replacement task that elicits paraphrases reflecting in-group and out-group interpretations. This allows us to track the gradual semantic change of dogwhistles with regard to their in-group and out-group meanings. We demonstrate our method by analyzing the life cycles of four immigration-related dogwhistles across two Swedish online discussion forums over a 23-year period (2000–2022). Our findings reveal different trajectories of semantic change, both across terms and between communities, highlighting the dynamic and context-dependent nature of dogwhistle communication.
Boholm et al. (Wed,) studied this question.