Estimative language is the standardized vocabulary through which intelligence analysts convey degrees of probability and analytic confidence. Sherman Kent first proposed in 1964 that words like “probable,” “likely,” and “possible” should correspond to defined numerical probability ranges, arguing that imprecise language allowed consumers to read whatever certainty they wished into analytic judgments — and analysts to evade accountability for the precision of their assessments.
Kent’s original spectrum assigned rough ranges: “almost certain” (93% ± 6%), “probable” (75% ± 12%), “chances about even” (50% ± 10%), “probably not” (30% ± 10%), “almost certainly not” (7% ± 5%). The project was resisted by analysts who viewed intelligence as an art rather than a science and by consumers who preferred narrative judgment to numerical hedging. Kent himself acknowledged the difficulty: the same word means different things to different readers, and the act of assigning a number can create a false sense of precision about fundamentally uncertain judgments.
The post-9/11 intelligence reforms revived Kent’s project. Intelligence Community Directive 203 (2007, revised 2015) established a standardized framework distinguishing between two dimensions: the probability of an event (conveyed through phrases like “unlikely,” “even chance,” “likely,” “very likely”) and the analyst’s confidence in the underlying judgment (rated “low,” “moderate,” or “high” based on source reliability, information quality, and analytic corroboration). The distinction matters: an analyst can assess with high confidence that something is unlikely, or with low confidence that something is likely — and the policymaker needs both dimensions to make decisions.
The recurring difficulty is that estimative language operates at the boundary between analytic rigor and rhetorical persuasion. A National Intelligence Estimate that says Iran is “probably” pursuing nuclear weapons has different policy consequences than one that says Iran is “possibly” pursuing them — and the distance between “probably” and “possibly” may rest on a single contested source or a disputed interpretation of ambiguous SIGINT. The history of intelligence failures includes cases where estimative language was either too imprecise to communicate genuine uncertainty (the 2002 Iraq WMD estimate’s “high confidence” claim) or too hedged to prompt necessary action.
Related terms
- Estimative intelligence — the analytic tradition that produced estimative language
- National Intelligence Estimate — the product in which estimative language is most formally employed
- Analysis of competing hypotheses — a structured technique for arriving at the probability judgments estimative language conveys