On 14 July 2017, the U.S. National Archives and Records Administration (NARA) announced that it would shorten the time period for holding eleven kinds of noncitizen detainee records and invited public comment on these changes. NARA stated that the decision to recategorize many of these records as “temporary” was because they held “little or no research value.” The files included records of abuse, assault, and deaths of people in immigration detention. Curious how NARA valued research, we designed a project asking what could be learned—about abuse in detention and NARA—from these documents. This article describes our methodological approach and our findings and discusses the implications of both for future research on immigration, government archiving practices, and accountability. We submitted Freedom of Information Act requests for the eleven document types slated for earlier disposal and analyzed Immigration and Customs Enforcement’s documentation of neglect and abuse in immigration detention. To do so, we traced the documents’ intertextuality, showing how each document relied on—and further produced—other documents. Challenging NARA’s calculative logics, we show that disposing of these documents would widen gaps, holes, and silences in an already partial, state-centric archive, limiting which future histories of U.S. immigration policies might be told.