Sequence-level recursion in nonhuman vocal communication A comparative test of five recursion types across 43 species

A recurring question in comparative bioacoustics is whether nonhuman vocal sequences show recursive structure of the kind central to human language (Yip 2006). We tested five recursion types drawn from the phonological literature on recursive prosodic structure — center-embedding, tail-end embedding, left-boundary embedding, external (prefix) phrase expansion, and internal phrase expansion — against call-type sequences from 43 species of bats, birds, primates, a pinniped, and two cetaceans. Center-embedding, tail-end embedding, and left-boundary embedding were tested across 223 files; external and internal phrase expansion were tested across a subset of 82 files spanning nine species. Center-embedding was rare overall (5.8% of files); under Benjamini-Hochberg false discovery rate correction, two species retained significant positive results — yellow-naped amazon (Amazona auropalliata) duets and house wren (Troglodytes aedon) song — while bowhead and humpback whale showed raw-significant but FDR-non-surviving results, indicating weaker statistical support than an uncorrected reading would suggest. Gibbon song (three species, two genera, 30 sessions) showed no center-embedding, but Hylobates agilis showed significant external and internal phrase expansion in the majority of sessions, consistent with the progressive build-up structure documented for gibbon great calls. Bowhead and humpback whale, similar in their center-embedding results, diverged sharply on phrase expansion: humpback showed pervasive repetition across all sequence positions; bowhead did not. Tail-end embedding was the most broadly replicated pattern (49.8% of files, with 97 of 112 raw-significant results surviving FDR correction); a phylogenetically corrected analysis found its association with vocal-learning category weakened but not eliminated relative to a naive estimate (PGLS slope = 1.01, p = 0.121, versus naive slope = 2.10, p = 0.004). The same phylogenetic treatment applied to external and internal phrase expansion was inconclusive, limited by the small number of species (n = 6) with a defined vocal-learning category. These findings are discussed in relation to the phonological recursion literature, in which recursive prosodic structure is documented as a real but non-default, minority-realized phenomenon even in human production.