Science, Grades 3–10 — the spine, and the decisions behind it

What was settled before a single Science artifact was written, and why. This is Steps 1–3 of NEW-LEARNING-AREA.md worked through for the first area after Mathematics.

Step 0 — the gate — was already passed. The curriculum guide is held: docs/research/source/MATATAG-Science-CG-2023-Grades3-10.pdf, 72 pages, August 2023, with a plain-text extraction beside it. Everything quoted below is Tier A — read out of that document, not searched for. The machine-readable extract is data/science-cg.json.

1. The five parameters, settled

NEW-LEARNING-AREA.md §2 says to settle these before writing anything, because discovering one at artifact seven means rewriting artifacts one to six. Settled:

The sixth parameter, which Mathematics did not have

A substantial performance task is required every quarter, and the guide supplies the tasks.

The Science curriculum requires learners to complete at least one substantial performance task for each quarter. — CG p. 17

This is a requirement, not a suggestion. It changes three things:

1. Every Science syllabus gives the last week of each quarter block to the performance task. Mathematics folds its performance tasks into the end-of-term block; Science cannot, because there are four quarters and three terms, so one term would carry two.
2. The performance tasks in the Science syllabi are quoted from the guide, marked Tier A. The Mathematics ones are the syllabus's own suggestions and are marked as such. This is the only place in the repository where that label flips.
3. The assessment pack has to carry a rubric for a task the teacher did not choose, so the rubrics are written against the guide's sample tasks rather than against invented ones.

2. Content domains — and the answer to the most consequential question

NEW-LEARNING-AREA.md §1 calls domain stability across grades "the single most consequential question in the whole build". For Science the answer is no, and in two separate ways — which is why the naive read of that guidance ("an area whose domains change at a key stage boundary is effectively two areas") does not apply here.

Way one — the names change at Grade 7. Three of the four are renamed:

This is a renaming, not a re-partitioning: the same four strands run all eight grades. So it is not two areas. It is one area whose vocabulary matures, and a syllabus that carries "Living Things" into Grade 7 is using an elementary label for a junior-high domain.

Way two, and this is the one that would have gone wrong — the quarter order changes at Grade 7.

Grades 7–10 form a 4 × 4 Latin square. Assuming Quarter 1 is Materials throughout is right for five grades and wrong for three — a 62% hit rate, which is very close to the 30% that the first Mathematics spine achieved by exactly the same reasoning. The difference is that this time the guide was open.

This table is CG p. 28, headed SEQUENCE OF DOMAIN PER QUARTER. It is the single most load-bearing page in the guide for anyone building materials against it, and no amount of searching had surfaced it.

3. The spine, grade by grade

One competency per grade, chosen from the competency list, not the content-standard summary — the rule that changed three Mathematics rows late. Every row is CG Quarter 1, because Quarter 1 is taught in Term 1 and Term 1 is where a first build lands.

Why these eight, and not eight others

The selection rule that mattered most was the fourth in NEW-LEARNING-AREA.md §3: pick a competency a learner can be wrong about in an interesting way. Four artifacts — the misconceptions card, the remediation sheet, the game and the diagnostic item on the worksheet — are thin if there is no characteristic error. Each row above has one, and most of them are famous:

• Grade 3 — a learner states an opinion and calls it an observation. "Maganda ang bulaklak" is not an observation; "May limang talulot ang bulaklak" is. The Key Stage 1 standard lists observing, inferring and predicting as separate skills for exactly this reason.
• Grade 4 — "if it changes, it is a chemical change". A learner who has just met chemical properties applies them to melting, bending and dissolving.
• Grade 5 — sand pours and takes the shape of its container, so a learner calls it a liquid. The CG's competency is stated in terms of shape and volume, which is precisely the pair that separates them, and its own suggested investigation is "Do gases (like air) or liquids (like water) have mass?" — the misconception is named in the guide.
• Grade 6 — "burning is reversible, the ash is still there", and its mirror, "dissolving is irreversible, the sugar disappeared".
• Grade 7 — the canonical particle-model set: the particles themselves expand when heated, there is air in the spaces between particles, particles in a solid do not move. All three are contradicted by the CG's own competency 3.
• Grade 8 — "3 : 1 means three of my four children". A ratio read as a count is the reason Punnett squares get taught and then not believed.
• Grade 9 — "action and reaction cancel out, so nothing should move". They act on different bodies. The CG names the two examples that pull apart hardest: stepping off a boat, where the learner can feel it, and a book on a table, where the weight and the normal force are not an action–reaction pair and almost every learner says they are.
• Grade 10 — "the plates float on liquid magma". The asthenosphere is solid and ductile, and the whole of plate tectonics reads differently depending on which of those a learner believes.

Grade 9 and Grade 10 are the two rows that moved

Both were chosen, written down, and then changed after reading the competency list properly.

• Grade 9 was going to be Newton's first law — inertia, Q1 · 1. The competency list gives inertia one line and gives the third law two, one of which explicitly demands force diagrams for action–reaction pairs. The richer competency is the third law, and it is the one with the misconception.
• Grade 10 was going to be climate change, on the assumption that a Grade 10 Earth and Space quarter would lead with the topical material. It does not. The quarter opens on plate tectonics and reaches climate at competency 8 of 12. The topical assumption is exactly the species of error that put seven wrong rows in the first Mathematics spine.

4. Week placement — and why it is not Week 7 eight times

Nine of the ten Mathematics sets sit at Term 1 Week 7. No Science set does, and the reason is worth stating so the difference does not read as carelessness.

Mathematics Quarter 1 leads with Measurement and Geometry in every grade, so the Number and Algebra competency this repository builds against sits two-thirds of the way through the quarter — hence Week 7, eight times over. Science Quarter 1 has no such consistent lead: each grade's quarter is a single domain, listed in its own order, and the competency with the interesting failure mode falls where it falls. It falls at Week 2, 3 or 4.

Placement follows one rule applied identically to all eight grades:

Quarter 4 gets the ninth week because it falls entirely inside Term 3, and DO 009 s. 2026 is reported to give Term 3 61 instructional days against 54 and 55 for Terms 1 and 2. That is the only asymmetry, and it has a reason rather than a rounding.

⚠️ This mapping is this repository's, not DepEd's. The CG assigns content to quarters and says nothing about weeks. A school's approved Budget of Work replaces every week number above. None has been obtained — see source-catalogue.md. What is DepEd's is the quarter-to-domain assignment, and that is reproduced exactly.

5. Grade 3 sits on a boundary in three ways at once

Worth collecting in one place, because each one caught something during the build:

1. Language. Grade 3 is the last MTB-MLE year, so its artifacts are Filipino and its five teacher documents carry .en.md reading copies. Grades 4–10 are English and carry none. The boundary runs through this learning area rather than beside it.
2. Grading. Grade 3 is Key Stage 1. DO 015 s. 2026's descriptive Key Stage 1 grading is phased — Grade 1 from SY 2026–2027, Grade 2 from SY 2027–2028, Grade 3 from SY 2028–2029. So Grade 3 Science is numerically graded this school year and descriptively graded from the year after next. Its class record is built accordingly and carries the expiry date in writing.
3. Prior learning. The guide states plainly that Kindergarten to Grade 2 science content "form[s] part of other curricula" and is not in this guide, and that Grade 3 use "should be built on and incorporate the content of other curricula". A Grade 3 Science lesson therefore depends on prior learning that this guide does not show, which is called out in the Grade 3 syllabus rather than assumed away.

6. What could not be established, and is therefore absent

Kept short and specific, in the repository's habit of making an omission visibly a decision.

That last one is the practical gap. The CG names a compound microscope at Grade 7, measuring cylinders at Grade 5, an ammeter and voltmeter at Grade 9 and a ripple tank at Grade 6. Every artifact in this set is built to run without any of them, and each grade's offline activities name the improvisation explicitly — that is a design response to an unknown, not a claim that the equipment is unnecessary.

7. What the eight builds settled

Written after all eight shipped, so the next learning area does not re-derive any of it.

The generator was the decision that made eight grades affordable

ARTIFACT-SPEC.md §7 records that two shipped crashes came from copying a game engine between grades and leaving a literal that named the source grade's data. Copying a folder eight times would have offered eight chances to repeat that.

Instead tools/science.mjs emits every structural artifact from data/science/gNN.json — the hub, slides, game, worksheet, remediation sheet, wall chart, parent guide and class record. There is one game engine, one slide runner, one print stylesheet and one class record for the whole learning area, and no grade carries a literal it could fail to edit.

What was not generated: the five markdown documents per grade stay hand-written. Generating prose from data only moves the prose into the data file, loses the diff, and makes every review a review of JSON. The split held for all eight grades and is the recommendation for the next area.

The eight-grade shape, for reference

The week mapping above was applied identically to all eight grades and is this repository's inference, not DepEd's — every syllabus says so on the page. The quarter-to-domain assignment is DepEd's and is reproduced exactly.

What the guide contradicted, grade by grade

Each syllabus carries a What opening the guide changed table. Collected, the eight tables are the strongest argument in this repository for holding the source document before writing:

Three of those eight are the domain rotation, and each one is worth seven weeks of misdirected preparation.

The performance-task count is not one per quarter

Page 17 says at least one substantial performance task for each quarter, and the guide supplies more than one in several places. Counted across the eight grades, the number of tasks per quarter ranges from one to three — Grade 9 Quarter 1 has three, one for each of its unrelated topics.

Every assessment pack therefore carries a rubric for every task the guide gives, names which one the syllabus placed in the performance-task week, and says why. A pack that silently assessed one of three would have looked complete and been wrong.

The cost nobody budgeted: a grade number stopped being an identity

The largest piece of unplanned work in the whole build had nothing to do with Science content. g03-math and g03-science are both "Grade 3", and the grade number was quietly the key in:

• tools/inventory.mjs's cell map — which silently merged the two Grade 3 directories and made a grade's byte total disagree with itself
• tools/build.mjs's short-URL table — /g3 was declared twice in _redirects
• the gallery's INVENTORY lookup and both of its filters
• five assertions in the deploy, gallery and browser tests, written as constants

All of it is now keyed on the directory slug or derived from data/grades.json. The next learning area should not pay this again — but it is worth an hour checking rather than assuming.