Language research has never been richer; spanning formal theory, documentation, neuroscience, psychology, education and AI. Yet it remains partitioned by disciplinary silos, methodological habits, and Western, Educated, Industrialized, Rich, and Democratic (WEIRD) sampling biases. We argue that sustained progress on core problems, i.e., how language is learned, processed, varies, breaks down and can be engineered, requires “retiring linguistics” as an isolated discipline and consolidating expertise within an integrated Language Science. This shift does not compromise rigor; rather, it situates formal modelling alongside quantitative evidence, field-based research, clinical and technological applications. Recent advances that bridge traditional boundaries such as neurosemantic mapping, speech neuroprosthetics, and computational approaches to sociolinguistic variation illustrate the potential of such integration. At the same time structural barriers including departmental incentives, fragmented training pathways, and inconsistent terminology continue to limit coordination across fields. This article combines a critical review of contemporary language research with a concrete proposal for institutional and epistemic reforms: transdisciplinary institutes and appointments, evaluation criteria that reward collaboration, curricula that braid theory, computation and field methods, funding and venues for cross-field work, ethical frameworks centred on partnership, and benefit-sharing with language communities. Unifying around problems rather than on departments can deliver more universally applicable research and greater societal benefit, from equitable language technologies and education, to improved clinical outcomes, by aligning explanations across levels from neurons to social networks.