06.10.2026 Tudor Popescu

Music and language are both culturally transmitted, and where they meet in sung verbal art, each constrains the other. This talk presents three lines of work on how such constraints shape structure over historical time.

I begin with the pilot study at the core of my ISLE project, asking whether German Lieder texts (1800 to 1900) retain archaic vocabulary longer than period-matched poetry. Many were chosen for music rather than written for it, so the filter is selection, reinforced by repeated performance. Sung repertoires may therefore act as reservoirs, with slower lexical turnover than unsung literary language. I present the design and available preliminary evidence.

The second line runs the same diachronic logic from the other side. Using Billboard Hot 100 songs from the last decades, we asked whether long-term trends previously found in lyrics (rising stress language, falling sentiment and complexity) are replicated in features derived from the musical audio, or whether the two channels are decoupled, with implications for biases that drive selection on the production and consumption sides.

Both studies take musical structure as given. In closing, I turn to an ERC Consolidator proposal in preparation, on the evolutionary origins of multi-part (as opposed to unison) singing. Converging evidence is sought for the hypothesis that tonal structure evolved under social bonding pressures, as a technology to manage the trade-off between voices getting blended vs kept individually-traceable. Neither pole serves on its own: fully fused voices signal unity but become uncountable, while fully separate ones forfeit the signal of acting as one. Tonal systems, on this account, let singers move between the two.

Across different timescales, the underlying question is: how much of the structure we find in music and verbal art is explained by the conditions under which it is created, performed, and passed on?