Grading the Inference

The Admiralty Code, the two-axis grading of source reliability (A to F) and information credibility (1 to 6) carried in NATO intelligence doctrine, grades the inputs to analysis. It has no address for the analyst's own inferences, in particular the identification and linkage judgements that decide which reports concern the same person, object, or event, and its top credibility grade presumes an independence between confirming sources that the evaluator has no field to check. This paper argues that both gaps are structural, and that a closed historical archive exposes them more cleanly than live intelligence work, because an archive cannot be re-tasked. Using a worked example of invented but plausible fifteenth-century English borough-court, ale-assize, and probate entries, it proposes four repairs. First, a third axis for link hypotheses, built as a ladder of work rather than a scale of feeling: seven rungs from 0 (admissible) through 1 (hunch) to 6 (closed by record), each defined by an inquiry step completed, so that every grade also names the next task. Second, a production-lineage field that reserves both the top credibility grade and the convergent rung of the ladder for confirmation across distinct lineages, naming the failure it prevents genlyd: corroboration that is the same voice returned. Third, aggregation over possible readings with an explicit sensitivity sweep, so that ordinal grades never silently become probabilities, every conclusion carries a robustness verdict, and the links can be ranked into a work plan by how much each moves the conclusions. Fourth, the paper turns the ladder back on the code itself, arguing that the existing grades should travel with a work rung beside them rather than be replaced, and that divergence between grade and work is a diagnostic signal in its own right. The proposal is offered both to intelligence doctrine and to digital prosopography, and six falsifiable predictions are stated for analyst training.