The present analysis interrogates the epistemic foundations of corpus linguistics in Uganda by examining the archival silences that shape its institutional memory. While corpus linguistics is often presented as a neutral, data-driven enterprise, the Ugandan context reveals that the selection, preservation, and accessibility of linguistic data are profoundly conditioned by colonial legacies and postcolonial institutional practices. Drawing on a decolonial hermeneutic framework, the analysis demonstrates that the absence of certain languages, registers, and communicative practices from Ugandan corpora is not an incidental gap but a systematic pattern of epistemic erasure. By reinterpreting the relationship between archival power and linguistic representation, the article argues that a decolonised corpus linguistics in Uganda requires not merely the addition of more data but a fundamental reconsideration of what counts as linguistic evidence and whose voices are authorised to speak. The contribution is a conceptual framework for understanding how institutional memory operates in corpus construction and a set of principles for building corpora that resist rather than reproduce archival silences.