Conceptual depth
Early prose sustains concrete contingency; 1979 expands rhetorical purpose; current work exposes abstract literary and institutional relations. Topic affordance explains much of what becomes visible.
Cross-period synthesis
Sophistication, without a scorecard
The texts show changing visible operations. The tasks, populations, timing, media, selection, and scoring systems also change—so observation must stop well before a rise-or-decline claim.
Largest defensible change: task demands and what assessments make visible—not a measured rise or decline in intelligence, underlying capacity, character, or diligence.
How to read the periods
1919 → 1979: the closest match is narrow: timed, source-free narrative. Both sets coordinate concrete relations under first-draft conditions, but their prompts, ages, geography, archive media, selection, and scorer constructs remain different.
1979 → 2024–2025: there is no like-for-like current narrative. The least-bad persuasion comparison puts a brief source-free speech beside longer, typed source-based arguments. A jump from no citations to citations is structurally induced.
| Lane | What it makes visible | Central mismatch |
|---|---|---|
| 1919 calibrated core | Chronology, danger, concrete cause and contingency in five timed narratives | One Virginia prompt; selected scale anchors; typeset excerpts |
| 1928 supplements | Practical, evaluative, sensory, and social-strategy breadth | Unknown writing years/prompts; textbook selection; weaker calibration |
| 1979 NAEP | Adaptation to five short rhetorical purposes | Selected score exemplars from different task-specific constructs |
| 2024–2025 anchors | Source-grounded analysis, academic macrostructure, objection and rebuttal | Longer time, typing, supplied texts, three states, anchor selection |
Ten dimensions of visible sophistication
Number and relationship of ideas
All periods contain multi-idea structures. Counts are transparent inventories within either excerpts or full responses, never interval measurements and never pooled into a composite.
Abstraction
Current source packets elicit more explicit abstraction. Historical narratives were not designed to reveal sustained source-based analysis, so absence is missing matched evidence rather than incapacity.
Causal reasoning
Cause, condition, means–end chains, analogy, and interpretive mechanisms occur across the record. Their objects change from rides and chores to public policy, texts, and institutions.
Qualification and counterargument
Counterfactuals, modality, alternatives, and objections appear where tasks allow. Florida makes rebuttal prominent because its grade-level rubric explicitly asks for multiple counterclaims and source-based response.
Evidence use
Historical support is experiential, sensory, recalled, or prompt-supplied. Current support is textual by design. Quotation is visible; a fully explained warrant is less consistent.
Organization
Chronology and problem/response appear early; claim/reason/consequence in 1979; thesis/body/conclusion and source blocks today. Full-response coding, time, typing, and rubric instruction all affect this contrast.
Sentence control
Meaningful errors coexist with traceable structure in every lane. Typesetting may hide historical manuscript features; present robustness anchors are selected for strong conventions.
Lexical precision
Specific language appears throughout. Current technical diction may be source-derived and can sharpen or overreach; it cannot automatically be credited as independent vocabulary command.
Articulation limits
Early causal assumptions often remain unexamined; 1979 support can be brief; present evidence-to-claim links can repeat or leap. The kind of limit follows the task.
Observed, inferred, unsupported
Observed: language in the records; prompt wording; documented time, medium, score meaning, and institutional comments.
Bounded inference: current retained assessments institutionalize source integration and explanation more explicitly than the retained earlier performance tasks.
Unsupported causal claim: that curriculum, technology, school participation, accountability, automated scoring, AI, or culture caused a change in student capacity. Those remain confounders or hypotheses.
No composite is calculated. There is no ranking, trend line, annualized rate, p-value, confidence interval, nationally representative estimate, or claim of statistical significance.
Sensitivity checks
- Remove all nine 1928 supplements: sensory/evaluative breadth disappears; the five 1919 timed causal narratives remain.
- Use Florida only for the present: claim, source use, paragraph structure, counterclaim, and underexplained-warrant observations remain; Massachusetts/Texas genre breadth disappears.
- Keep only timed source-free narrative: comparison ends in 1979. No three-period performance trend survives.
- Remove institutional annotations: textual observations remain, but explicit assessor praise/criticism claims disappear. The teacher-pair result stays zero.