Historical problem solving: A study of the cognitive processes used in the evaluation of documentary and pictorial evidence.

A small expert-novice comparison - practising historians versus high-achieving high school students (single-digit numbers per group)

Academic historians and US high school seniors, reading a document set on the Battle of Lexington

The founding descriptive study of the whole cluster, and descriptive is the operative word: think-aloud protocols from historians and from able high school students reading the same primary and textbook sources. There is no intervention, no control group, no outcome measure and no causal claim - it is a cognitive task analysis. It is recorded here at grade D because everything downstream (Reading Like a Historian, sourcing heuristics, Civic Online Reasoning, the historical-thinking arm of the Fendt et al. 2025 meta-analysis) is built on the constructs it named, and a reader needs to see that the constructs come from an n-of-a-few protocol study rather than from a psychometric programme. Wineburg is both the author of this study and the architect of the curriculum built on it, so independence is developer-led by construction. Read at abstract/secondary depth here.

Historians and students read the same documents in categorically different ways. Historians used sourcing (looking at who wrote a document, when, and why, BEFORE reading the body), corroboration (checking claims across documents), and contextualization (placing a text in its historical moment); the students, though academically successful, treated the textbook as the most trustworthy source precisely because it was anonymous and authorless, and read documents for information rather than for the argument behind them. Two consequences run through everything after it. First, sourcing is a real and describable expert-novice difference, not a pedagogical invention - the phenomenon is solid. Second, the study establishes that experts DO this, not that teaching novices to do it produces any downstream benefit; the causal question was left entirely open, and the trials in this cluster answer it far more modestly than the descriptive contrast suggests.

High if the expert-novice gap is read as evidence about instruction. Historians are a hyper-selected population on verbal ability and domain knowledge; the difference between a professional historian and a seventeen-year-old confounds decades of training, selection into a PhD, and general cognitive ability. Nothing in the design separates them, and the study does not claim to.

The descriptive expert-novice contrast has been reproduced many times across countries and document sets, and its later analogue in the digital domain (fact-checkers read laterally, students and academics read vertically) is the empirical basis of the Civic Online Reasoning line. What has never been established by replication is the inference the field draws from it - that installing the experts' heuristics in novices produces expert-like outcomes on anything beyond a test of those heuristics.