Reading a Document: Fact, Frame, and Inference
A three-part method for evaluating any document that makes claims about a person or institution — a claim-classification table, a steelman of the strongest opposing account, and a self-critique applied to your own conclusions with the same rigor.
This tool doesn’t tell you whether a document is telling the truth. It gives you a disciplined way to find out — one that applies exactly as rigorously to your own conclusions as to anyone else’s.
Purpose
How do you evaluate a document that makes claims about someone?
A news article, a police report, a school board communication, a legal pleading, a timeline someone has assembled — all of these documents mix three different things together: what actually happened, how the author has chosen to describe what happened, and what the author believes was going on beneath the surface. Most readers absorb all three at once, at the same level of confidence. This tool exists to pull them apart.
The Method
Three parts, applied in sequence
Claim Classification
Break the document into its individual claims. Sort each one into: Fact (directly supported by an independent, verifiable record), Characterization (a real, documented event described in loaded or interpretive language), Inference (a reasonable conclusion drawn from documented evidence, clearly labeled as a conclusion rather than a fact), or Speculation (an assertion about motive, intent, or an unstated fact with no direct evidence behind it).
Steelman the Alternative
For every claim landing in Inference or Speculation, construct the strongest good-faith explanation available in the record that points a different direction — including one that makes the document’s author look mistaken, unfair, or wrong. If you can’t construct a real alternative, that’s useful information too; it may mean the classification actually belongs closer to Fact than first sorted.
Self-Critique, Applied Evenly
Turn the same classification scheme on your own conclusions about the document — not just the document itself. Where have you used characterizing or pathologizing language without equivalent factual support? Where might a neutral third party, reviewing the same record, reasonably land somewhere else? This step is the one most tools like this skip, and it’s the one that actually makes the other two trustworthy.
The central risk this method is built to resist: a classification table looks rigorous and neutral regardless of what’s poured into it. Applying Parts One and Two thoroughly to someone else’s document while skipping Part Three on your own conclusions doesn’t produce objectivity — it produces one-sided argument wearing objectivity’s clothing.
Worked Example
A fictional homeowners’ association dispute
Entirely invented, to demonstrate the method without any resemblance to a real case: a resident, Jordan, files repeated noise complaints against a neighbor. The HOA board eventually issues a formal notice. Jordan later circulates a written timeline alleging the board coordinated with the neighbor to retaliate.
| Party | Claim | Classification | Basis |
|---|---|---|---|
| Jordan | The board “targeted” Jordan after the third complaint. | Characterization | A formal notice was issued after the third complaint (Fact). “Targeted” frames a documented sequence as retaliatory. |
| Jordan | The board and the neighbor “coordinated” the notice in advance. | Speculation | No email or record showing prior coordination has been produced; this asserts intent not yet evidenced. |
| Board | The notice was issued through “standard procedure.” | Inference | The notice followed the HOA’s documented complaint threshold — a reasonable conclusion, though board meeting minutes for that specific week haven’t been produced yet. |
| Board | Jordan’s complaints were “excessive and unfounded.” | Characterization | Three complaints were filed within one month (Fact). “Excessive and unfounded” frames the volume and merit in the board’s preferred terms. |
Steelman
Jordan’s claim: the board coordinated with the neighbor to retaliate.
Alternative: A three-complaint threshold triggering formal review is common HOA procedure regardless of who is complained about. The timing may simply reflect the complaint count crossing a pre-set line, not evidence of a private arrangement.
Self-Critique
On the analysis itself: Labeling the board’s “excessive and unfounded” language as mere characterization assumes the complaints had some merit — that assumption isn’t yet established either, and deserves its own fact-check rather than a free pass.
On completeness: Without the board’s internal meeting minutes, “standard procedure” remains an inference resting on the board’s own account — not yet independently confirmed the way a truly neutral analysis would require.
Applying It Elsewhere
Where this method transfers directly
- Media articles — separating a reporter’s sourced facts from the framing choices built around them
- Police reports — distinguishing an officer’s direct observations from their interpretive characterization of a subject’s behavior or intent
- Timelines assembled by an interested party — the exact genre this method was built to test
- School board and institutional communications — checking a stated “standard procedure” against the actual documented record
- Legal pleadings — separating the facts alleged from the legal characterizations built on top of them
- Investigative summaries and public statements — the same three-part test, regardless of who authored the original document
Limitations
What this method doesn’t fix, and what it can create if misused
Confirmation bias: a reader can classify claims consistent with what they already believe as “Fact” and claims that challenge it as “Speculation,” without ever noticing the asymmetry.
Over-analysis: not every document rewards this level of scrutiny. Applying a heavy three-part method to a routine, low-stakes communication can manufacture suspicion where none is warranted.
False pattern recognition: finding a “technique” in every rhetorical choice risks treating ordinary imprecise language as evidence of deliberate manipulation.
Assuming intent: a document written under time pressure, institutional caution, or genuine uncertainty can look identical to one written with deliberate strategy. The method can classify language; it cannot see inside anyone’s head.
Mistaking rhetoric for deception: loaded language doesn’t prove a claim is false — a true claim can be described in biased terms, and a false claim can be described in careful, neutral-sounding ones.
The method’s signature risk: applied only to an opponent’s document, this tool produces the appearance of rigor in service of a predetermined conclusion. Part Three exists specifically to guard against this, and skipping it quietly defeats the entire point of the other two parts.
Sources & Further Reading
- News Literacy Project — fact/opinion/inference distinction in media literacy education
- This collection’s own: Why Citizens Feel the System Is Unfair (procedural justice and narrative framing)
- This collection’s own: Reading a Worldview: The Seven Questions (companion methodology piece, comparative rather than adversarial application)