Recursion Without Reference: A Cross-Disciplinary Framework for Detecting Structural Embedding in Nonhuman Acoustic Signals

Recursion has occupied a central and often contested place in debates over the evolutionary origins of language, largely because comparative research has treated it as inseparable from syntax and, in turn, from semantics. This coupling has made recursion difficult to test directly in nonhuman acoustic signals: existing evidence comes almost entirely from operant grammar-learning paradigms that assess whether an animal can discriminate a recursive grammar under training, rather than from analysis of whether recursive structure is present in naturally produced vocalizations at all. This paper proposes a structural, cross-disciplinary definition of recursion — hierarchical embedding of self-similar units — that is deliberately decoupled from any requirement that those units carry symbolic or referential meaning. Building on this definition, I distinguish five acoustically testable embedding topologies — center-embedded, tail-end embedded, left-boundary embedded, external (prefix) phrase expansion, and internal phrase expansion — propose indexical semantics (rather than symbolic reference) as the appropriate default explanation for any correlation between recursive structure and context, and lay out four operational commitments — segmentation criteria, a self-similarity metric, an embedding-topology test, and a null model — needed to make recursion a measurable property of a naturalistic acoustic signal. I then describe how existing and in-development analytical methods, including recursive contour analysis, wavelet decomposition, and the MOMEL/INSTINT contour-modeling pipeline, can be triangulated to implement this framework and derive a set of falsifiable predictions that follow from treating recursion as a domain-general property of hierarchically organized motor sequences rather than a language-specific syntactic device.