Operationalizing shared phonetic space in bilingual speech: A quantitative proof of concept for the Revised Speech Learning Model

The Revised Speech Learning Model (SLM-r) is highly influential in research on bilingual speech, but its central notion of a shared phonetic space is not directly observable. This article proposes Quantitative SLM-r (QSLM-r), a theory-driven framework that operationalizes that construct through four diagnostics: discriminant language axes, equivalence-band occupancy, bilingual--monolingual gap contrasts, and latent category structure. The framework is illustrated with an exploratory corpus of Brazilian Portuguese (BP) monolinguals, American English (AE) monolinguals, and BP--AE bilinguals producing speech in both language modes. Across five acoustic domains, Voice Quality was the only domain in which all diagnostics converged: the BP--AE axis was valid, bilingual productions clustered toward an acoustically ambiguous region, the bilingual gap was markedly smaller than the monolingual gap, and latent structure was associated with speaker condition. The remaining domains produced weak or non-convergent outcomes. The contribution of the study is therefore operational rather than population-general: it shows how a theoretically important but latent construct can be rendered explicit, testable, and partially falsifiable in multidimensional acoustic space.