Examiner Intelligence

Provenance: compiled 2026-09-29, from research/veridian/WMA14-verified-facts.md §5 (the per-topic citations, re-read in full specifically for their examiner-report prose — mark-scheme mechanics, M/A/B notation and isw/cao/awrt conventions are deliberately excluded, since those are already documented in that file's own §4 and re-analysed for technique in forge-exam-technique-guide.md §2) and §6 (series/document accounting, used to build this document's own evidence-base table below). Also read in full before writing, specifically to check for duplication: signal-frequency-and-boundaries.md (topic-frequency counts — a different question from this document's "how do reports talk," not restated here), forge-exam-technique-guide.md (mark-scheme mechanics and a technique checklist, not report rhetoric), all eleven codex-deep-dive-*.md files and codex-formula-and-method-bank.md (technique and formula synthesis, not language patterns), and content/veridian/question-bank/wma14.ts (marked-question items, no report-language commentary). None of those five documents analyses how the reports themselves are written — the recurring vocabulary, the sentence shapes, what gets named versus left implicit — so this document is additive to all of them, not a restatement. No new primary Pearson PDF was fetched this session; every quotation below reuses a fragment the facts bank already independently verified against the source PDF text, re-read here for a different purpose (cross-series rhetorical pattern, not topic-specific error content).

#WMA14 · SIGNAL — Examiner Intelligence

#What this document is, and isn't

This is not another pass through what WMA14 candidates get wrong — the twelve curriculum lessons, the facts bank's own §5, and FORGE's common-error digest already do that, exhaustively, per-topic. This document asks a narrower, different question: how do Pearson's WMA14 examiners actually write these reports, series after series — what phrases recur, what gets emphasised versus left as a bare statistic, and what that vocabulary and structure implies about how a script is actually read. Every pattern below is grounded in at least two independently-cited series, per this course's standing discipline: a phrase or emphasis seen in only one series is reported as a single finding, in its own section, not folded into a "pattern." Nothing here invents a likelihood, a percentage, or a grade boundary — Pearson's examiners themselves never state one in any quotation available to this course (see Pattern 1), and this document does not do what its own source material doesn't.

#1. The evidence base for this document, specifically

The facts bank holds primary documents for eleven series, but only some of those are full examiner's reports — the genre this document is actually about. Mark schemes and question papers are excluded from the table below; they carry no report-writer's prose to analyse.

The honest denominator for this document is 5 — the count of series with any examiner-report prose captured in the facts bank. Four of those five (Jan 2021, Oct 2021, Oct 2022, Jan 2024) are full reports spanning most or all of the paper; the fifth (Oct 2024) contributes only two questions' worth of commentary, not a full-report read. Where a pattern below leans on Oct 2024 for one of its two required series, that is stated explicitly rather than presented as equal-weight evidence to the other four.

#2. The recurring shape of a single report entry

Read across all five series, a report's comment on a given question follows a fairly stable shape, almost always in this order: (1) a headline judgement on how the question went, using a non-numeric severity word (Pattern 1 below); (2) the specific wrong move, often quoted close to verbatim, sometimes as a literal "wrong expression" rather than a description of the error type (Pattern 5); (3) the mark cost of that move, named explicitly rather than left to be inferred (Pattern 2); and, less consistently, (4) an acknowledgement that another valid method existed (Pattern 4) and/or (5) a generalised, forward-looking sentence of advice (Pattern 7). Not every entry carries all five moves, but no entry reviewed skips (1) and (2), and the great majority include (3).

#3. Six recurring patterns (each confirmed in two or more series)

#3.1 A fixed, non-numeric severity vocabulary — and no percentage, ever

Across every one of the five series with report prose, the reports quantify how many candidates did something without ever stating a number. The vocabulary sits on a real, ordered scale, and the same words recur series after series:

• Lower end: "a very small minority" (Jan 2021, Q1), "a small minority" (Jan 2021, Q4), "relatively few students" (Jan 2024, Q2a).
• Middle: "a disappointing number of responses" (Oct 2022, Q1), "many candidates did not attempt this question" (Jan 2021, Q3), "a significant proportion of candidates failed to show any method" (Oct 2024, Q2).
• Upper end: "A great many responses to part (a) did correctly factorise the expression" (Oct 2021, Q10), "The majority successfully set up their initial assumption" (Jan 2024, Q8), "almost all candidates were able to state that the volume" (Jan 2021, Q9), "extremely well done with the majority of candidates scoring at least 5 marks" (Oct 2022, Q2).

What this implies. No quotation available to this course, in any of the five series, gives a number. This is the practical reason this document (and the facts bank, and the SIGNAL frequency document before it) treats "seen in N of M series checked" as the only honest unit — the source genre itself never quantifies, so a lesson or a student reading these words should read them as relative severity signals, not measured proportions. "Extremely well done" and "a disappointing number" are doing real evaluative work even without a figure attached; the ordering of the vocabulary above is itself the only scale that exists.

#3.2 Every named error is paired with its exact mark cost, not left implicit

In every one of the five series, when a report names a specific wrong move, it also states, in the same sentence or the next one, exactly what that move cost:

• "This immediately resulted in 0 marks for (a)(i)" (Oct 2021, Q7).
• "scores of 0 marks were very common" (Oct 2021, Q4).
• "many candidates assumed that ln(x²) was identical to 2 ln x" and "failed to score" (Oct 2021, Q8b).
• "gave no explanation as to why" and "gained no further credit" (Jan 2024, Q8).
• "a significant proportion of candidates failed to show any method for solving the resulting quartic," forfeiting the method mark (Oct 2024, Q2).

What this implies. These reports are not written as general pedagogy about common misconceptions in the abstract — the mark consequence is stated as part of the same observation, every time, across all five series. That is a genuinely useful reading habit to copy: the report is telling a teacher not just what went wrong but exactly how much it cost, in the same breath, which is a stronger, more actionable signal than a misconception described on its own would be.

#3.3 The same split recurs across two unrelated topics: correct calculation, missing conclusion

Read the proof reports and the vectors reports side by side, and the same underlying gap shows up under two different topic labels, confirmed independently in four series:

• Proof by contradiction. "but few made a comment to state that it was odd" (Oct 2021, Q10) — candidates correctly factorised but didn't draw the needed conclusion from it. "The majority successfully set up their initial assumption" but "gave no explanation as to why" a resulting equation was impossible, so "gained no further credit" (Jan 2024, Q8).
• Skew and parallel lines. The majority "failed to state or show that the given lines were not parallel" despite correctly finding the line parameters (Jan 2021, Q8). A "fairly high proportion who achieved a pair of conflicting values" — the correct calculation — "failed to then conclude that this implied the lines did not intersect" (Oct 2022, Q9).

What this implies. Four series (Jan 2021, Oct 2021, Oct 2022, Jan 2024), across two spec points that have nothing mathematically in common, show the same reading habit: examiners are marking for a stated conclusion that follows from a correct calculation, not crediting the calculation as if the conclusion were automatic. A script that computes the right numbers and stops is, on this evidence, a repeatedly-documented way to lose marks that have nothing to do with arithmetic ability.

#3.4 Multiple genuine methods are named and credited neutrally, sometimes with a practical (not evaluative) steer

Three series independently show reports naming more than one valid route to the same question, without ranking them as "better" or "worse" mathematically — only, sometimes, as more or less error-prone in practice:

• "Other methods were seen in part (a) including minimising the length of AX using either differentiation or by completing the square" (Oct 2021, Q7).
• "Once one of the above two methods were clearly seen to be applied, all marks were then available in part (a)" — for either a direct chain-rule route or a full binomial-expansion route (Oct 2022, Q11a).
• "A few used the tabular 'DI' approach" while "the majority opted for the more traditional method" (Jan 2024, Q5a) — named without preference. Separately, the same series states plainly that simultaneous-equation expansion was "more prone to error than attempts via substitution" (Jan 2024, Q2a) — a practical steer, not a claim that one method is mathematically more valid.

What this implies. The reports treat "which correct method did you use" as a genuinely open choice, never a right-versus-wrong axis — but they are also willing to say, plainly, which choice tends to go wrong more often in practice. Read together, that's a specific and useful signal: don't infer from a report naming an "alternative" method that it's a lesser one, but do take seriously a report's own stated observation about which of two valid methods candidates handle less reliably.

#3.5 The exact wrong transformation is quoted, not just categorised

Three series show reports naming a wrong step as a literal mathematical expression rather than a description of an error type — an unusual level of specificity for prose commentary:

• "differentiating 3x²y → 6x dy/dx" (Oct 2021, Q1) — three such literal wrong-transformation fragments appear in this one question's commentary alone.
• "1/(3sin²t) → 3cos²t" (Oct 2022, Q6).
• "obtaining the vector (12,−10,24) rather than the value of 12−10+24" — naming the exact wrong output type against the exact right one (Jan 2021, Q2a).

What this implies. This level of detail means the report-writer is quoting, or very closely paraphrasing, literal working from real scripts — not summarising a category of mistake from memory. For a teacher, this is the single most drillable kind of finding in these reports: it names the precise wrong line a student is likely to write, not just the general shape of the misunderstanding behind it, and it is worth teaching against the exact wrong expression, not only the concept.

#3.6 A topic's multi-year reputation is occasionally invoked above that series' own data

Two series step back from the specific sitting to characterise a topic's standing reputation across years, independent of how that particular cohort actually did:

• "Students find new topic of 'proof' challenging" — a general framing statement, not scoped to the January 2021 sitting alone (Jan 2021, general report).
• "Vectors questions are often unliked by students," followed immediately by "showed some improvement over previous years and there were far fewer blank scripts" (Jan 2024, Q6).

What this implies. In both instances, this framing sits above the series-specific commentary, not inside it — a meta-level judgement about the topic's history that the report-writer evidently felt was worth stating on its own. Read narrowly (two series, two different topics), this is enough to call a real pattern rather than a single aside, but it should not be over-read as something every report does for every topic; it surfaces only where the report-writer apparently judged the topic's reputation itself to be worth remarking on.

#3.7 The thinnest pattern here: an occasional shift into direct, prescriptive address

Two series — the fewest of any pattern in this document, and one of the two is the thin, two-question Oct 2024 extract rather than a full report — show the commentary briefly leaving retrospective description and addressing future candidates directly:

• "It is really important in a proof to include all necessary steps" (Oct 2021, Q10).
• "Candidates need to be aware that it is important to substitute x" into the original expression, not the series expansion, to see what's actually being approximated (Oct 2024, Q1).

What this implies, and the limit of that implication. This is included as a pattern rather than a single finding because it clears this document's own two-series bar — but only just, and readers should weight it accordingly. Where it does appear, the shift is notable: the report stops describing what happened and states, plainly, what a future candidate should do — a stronger and more directly actionable sentence than the surrounding descriptive prose, and worth flagging to students precisely because it is rare enough that when a report does it, the underlying point is evidently one Pearson considers both common and genuinely preventable.

#4. Two single findings — real, but not generalised into patterns

Per this course's own discipline, these are reported as exactly what they are: one real, cited observation each, not evidence of a recurring habit.

An explicit year-on-year improvement claim, found in exactly one series. "The idea of expanding a binomial expansion in x² did not cause the same issues as last year, so clearly candidates have learned from previous series" (Oct 2022, Q4). No other quotation available to this course, in any of the other four series, makes this kind of direct backward comparison to a named prior sitting. It is genuine evidence that Pearson's examiners do, at least once, track and state a longitudinal improvement by name — but one instance is one instance, not a habit of year-on-year commentary.

One vivid, single-instance naming of a fudged "show that" answer's mechanism. "The given answer here persuaded many candidates to 'adjust' their working following obvious mistakes" (Jan 2021, Q10). The broader principle behind this — that a printed target answer is graded on the genuineness of the path to it, not just the final match — is already well-attested across three separate topics and series (this course's FORGE guide names it the "fudged-reverse-fit" meta-trap, citing Jan 2021 Q3, Q4 and Q10 independently). What is attested only once, here, is the specific word "adjust" and the framing of intent behind it — the report naming the mechanism of the fudge, not just its existence. Worth quoting for its vividness; not generalised into a claim that reports usually describe fudging this explicitly.

#5. Flagged gaps

• The denominator is 5, and it is a mixed 4-full/1-partial denominator. Two of the seven patterns above (3.2, 3.7) partly rest on the October 2024 series, whose report coverage in the facts bank is a two-question extract, not a full-report read the way the other four series are. Pattern 3.7 in particular clears this document's two-series bar only because of that thin fifth series; it is the single weakest-attested pattern in this document and is flagged as such in its own section too.
• This document does not extend the facts bank's own evidence base. Every quotation above was already independently verified against the source PDF text by WMA14-verified-facts.md; this pass re-read that verified material for a different purpose — cross-series rhetorical pattern, not topic-specific error content — and fetched no new primary Pearson document. A future pass that reads the remaining unreviewed series' full examiner reports (June 2022's report specifically, January/June 2023, October 2023, June 2024, and a January 2026 report once Pearson publishes one) would meaningfully test whether the six confirmed patterns above hold up against a wider sample, and could either strengthen Pattern 3.7 past its current thin footing or surface a genuine counter-example.
• No grade-boundary, percentage or probability claim appears anywhere in this document, matching Pattern 3.1's own finding that the source genre never states one — consistent with, and reusing, the same no-fabrication discipline already stated in signal-frequency-and-boundaries.md §2.
• This document is silent on January 2026 because no examiner's report exists for it yet (confirmed directly in the facts bank, §8) — nothing here should be read as implying anything about that series' report language, since none exists to analyse.

#Sources used

research/veridian/WMA14-verified-facts.md §5.1–§5.7 (examiner-report prose only, re-read for cross-series rhetorical pattern rather than topic content) and §6 and §8 (series/document accounting, and the confirmation that no January 2026 report exists yet). Checked in full to confirm this document does not duplicate them: signal-frequency-and-boundaries.md, forge-exam-technique-guide.md, all eleven codex-deep-dive-*.md files, codex-formula-and-method-bank.md, and content/veridian/question-bank/wma14.ts. No new primary Pearson document was fetched for this file; every quoted fragment above is reused, under fifteen words, with its own series/question attribution, from material the facts bank had already independently verified against the source PDF text.

Not affiliated with or endorsed by Pearson Edexcel.