# Exact text and nonvisual routes for five A4 aids

All five aids are original A4 SVG with selectable-text PDF counterparts. Their visual order is top-to-bottom and the colours are supplemented by headings and position. Read **model prediction** and **invented record** aloud; colour alone never encodes that difference. A tactile copy can use raised box boundaries, large number tiles and a two-column textured key. Do not pretend the generated SVG/PDF itself is Braille or a raised tactile print.

## [Claim poster](claim-poster.pdf) · [SVG](claim-poster.svg)

Largest headline in two parts: “EACH EXTRA LAYER HALVES / THE SENSOR VALUE!” Next, white on dark panel: “160 × 2⁻⁴ = 10 / at four layers, IF exact halving holds.” The qualification follows in its own light panel: “BUT OUR FIRST RECORD SAID 20 / That is 10 units above the model prediction. / We want a controlled retest before calling it dependable. These are invented arbitrary units.” It ends “LOOK AGAIN / Who made this poster, and who is its audience? / Which number is a model prediction? Which is a record? / What does the small-print limit change?” Byline: fictional Paper Screen team, draft poster, 18 September 2026. **Tactile/voice:** put heading, model and record on separate strips; read the qualification at the same volume as the headline.

## [Claim audit](claim-audit.pdf) · [SVG](claim-audit.svg)

Five numbered boxes in order: **1 Maker + purpose**: “Who created this source? For whom? To do what?” **2 Exact claim**: “Write the words or numbers it asks us to accept.” **3 Source detail**: “Quote a value and say prediction OR recorded.” **4 Assumption + limit**: “What must hold? What method detail is missing?” **5 Revised sentence**: “Name the source; keep uncertainty visible.” **Tactile/voice:** five raised numbered cards can be moved as a reasoning sequence.

## [Model / data / gap](model-and-data.pdf) · [SVG](model-and-data.svg)

Instruction: “A conditional exact-half rule beside one invented record. Units: arbitrary.” Four column headings: `n`, `160 × 2⁻ⁿ`, `B: invented record`, `gap`. Five rows: n=0 record160; n=1 record80; n=2 record44; n=3 record26; n=4 record20. Model and gap cells are **blank for learner work**. Footer: “n = 5 is unobserved. A prediction is not a new reading.” **Tactile/voice:** align each n/model/record/gap as four labelled number strips; keep predictions and recorded values on distinct textures.

## [Retest method mat](retest-method.pdf) · [SVG](retest-method.svg)

Five boxes in order: **Question + variables:** “Change: number of layers. Read: sensor value.” **Hold / record:** “Paper batch and size • lamp setting • geometry / Background light • sensor calibration or zero.” **Runs + roles:** “How many? What order? Who records? Who checks?” **Safety + permission:** “Teacher-approved low-voltage equipment only. / No eyes, mains wiring or public installation.” **Observation status:** “If no approved test took place, write: NOT OBSERVED.” The subtitle says a good written plan does not mean it was done. **Tactile/voice:** move a token from planned to observed only after an actual approved action.

## [Two-panel recommendation canvas](recommendation-canvas.pdf) · [SVG](recommendation-canvas.svg)

Subtitle: “For a fictional manager deciding only about a small supervised retest.” Panel 1 has blank ruled space under “WHAT THE SOURCES SHOW / Source + exact numerical contrast + model assumption.” Panel 2 has blank ruled space under “DECISION + NEXT CHECK / Limit + operational retest step + who may act.” Last line: “First-glance check: could either panel be mistaken for approval?” **Tactile/voice:** use two separately labelled page halves or two spoken segments, evidence then decision; if an Arts media artefact is assessed, keep the designed panel sequence in the submitted work.

The full source packet is [here](../SOURCE-PACKET.md). These aids omit the fresh check values and do not alter source evidence; the teacher-targeted key is public, so the checks are not secure exams.