sound_soup_plan.md (13751B)
1 # Sound Soup — Proposal 2 3 > **Status:** proposal / not yet built. High-level only — deliberately free of 4 > implementation detail, since the technical approach is expected to change. 5 6 ## 0. Scope of this document 7 8 This is a design and pedagogy proposal for a new KGames title aimed at 4–5 9 year olds learning to read. It states *what the game is, why it works, and what 10 constraints it must respect*. It does not pick an engine, a file layout, an 11 asset pipeline, or an audio strategy. Those decisions come later and should not 12 be inferred from this document. 13 14 --- 15 16 ## 1. Constraint reset 17 18 **The technical constraints that shaped the existing KGames titles no longer 19 apply to this game.** Specifically: 20 21 - **Any engine or framework may be used.** Phaser 3 is no longer the default. 22 If a different engine, or no engine, serves the game better, use it. 23 - **A build step is allowed.** The "double-click `index.html`" development 24 workflow is not a requirement here. 25 - **Real art assets are allowed and preferred** (see §2). 26 - **Real audio assets are allowed and preferred.** This game depends on 27 accurate speech and phoneme audio, which cannot be synthesised convincingly; 28 recorded audio is expected and welcome. 29 - **External files, CDNs, bundles, and asset directories are all fine.** 30 31 What has *not* changed: 32 33 - **The project remains open source, GPL-3.0.** Everything shipped must be 34 license-compatible and properly attributed (see §2). 35 - **No ads, no signup, no backend, no tracking.** Still a static site serving 36 free educational games to young kids. 37 - **It must still be testable**, and tests must still pass before commit. 38 39 ## 2. Asset policy 40 41 **Prefer real, openly licensed art made by human artists over procedurally 42 generated or AI-generated art.** In priority order: 43 44 1. Existing open-license art (CC0, CC-BY, CC-BY-SA, OFL, GPL-compatible) from 45 sources such as OpenGameArt, Kenney, itch.io free/open packs, Wikimedia 46 Commons, Freesound, and similar. 47 2. Commissioned or hand-made original art, contributed under a 48 GPL-compatible license. 49 3. Procedural generation — acceptable for incidental effects (particles, 50 gradients, simple shapes), not for the game's character and world art. 51 4. AI-generated art — avoid. 52 53 Every third-party asset must have its source, author, and license recorded in 54 the repository, and attribution surfaced where the license requires it. 55 License compatibility with GPL-3.0 is a hard requirement, not a preference. 56 57 The same preference applies to audio: real recorded voice and openly licensed 58 sound effects and music beat synthesised equivalents. 59 60 --- 61 62 ## 3. Why this game — the research case 63 64 The evidence base for teaching 4–5 year olds to read converges on a small 65 number of points: 66 67 - **Alphabet knowledge and phonological awareness are the strongest preschool 68 predictors** of later decoding, comprehension and spelling (National Early 69 Literacy Panel). Letter-sound knowledge, phoneme awareness and rapid 70 automatised naming each predict reading growth independently. 71 - **Letter *sounds* matter more than letter *names*.** Names help, but transfer 72 to actual reading runs through sound. 73 - **Blending is the bottleneck.** Meta-analysis shows phonemic awareness 74 instruction has a medium effect on segmentation but only a small effect on 75 blending — yet blending (fusing /k/ /a/ /t/ into "cat") *is* the act of 76 reading. It is the least well-served skill and the highest-value target. 77 - **Systematic synthetic phonics works**: teach a handful of letter-sounds, 78 then start blending real words immediately rather than waiting for the full 79 alphabet. The classic `s a t p i n` opening exists because those six letters 80 alone spell fifty-plus words. 81 - **Sound boxes (Elkonin boxes)** — one box per sound, one tile moved per 82 sound — are the canonical manipulative for this, with sixty years of use 83 behind them. 84 - **Digital versions demonstrably work.** GraphoGame/GraphoLearn, an adaptive 85 letter-sound game played ~10–15 minutes a day, produced measurable 86 letter-sound gains in classroom-randomised trials across several languages. 87 88 **The gap in KGames today:** `letter-find` teaches letter names, 89 `animal-letter` teaches initial sounds, `animal-spell` teaches letter-tile 90 spelling, and `work` builds oral language and vocabulary. Nothing in the 91 catalogue teaches a child to *blend sounds into a word* — nothing teaches 92 decoding. That is the hole this game fills. 93 94 --- 95 96 ## 4. The game 97 98 ### 4.1 Setting 99 100 A **tropical island town** — the same island world as Fire Truck, seen from 101 street level, with a fantasy tilt. Palm-lined streets, painted shutters, 102 market awnings, hand-painted signage, boats in the harbour, warm light. 103 104 The player runs a **soup stall** on a busy market street. The customers are 105 the fantasy element: island spirits, a sea-dragon, parrot-folk, a very 106 polite lizard in a hat, a coconut crab with opinions. Friendly, characterful, 107 never scary. 108 109 This setting does three things at once: it reuses and extends a world the 110 project has already established, it gives the game a cheerful urbanism — 111 a lived-in street with regulars, neighbours and daily rhythm — and it makes 112 a cooking game make sense. 113 114 ### 4.2 Core idea 115 116 **A customer orders. The child cooks the order by sounding it out.** 117 118 Each order is a short, decodable word. The ingredients are *sounds*. The child 119 assembles the sounds in order, stirs the pot, and the pot blends them — the 120 sounds are played with the gaps closing pass by pass until they fuse into the 121 spoken word. The dish is then served, and the customer reacts. 122 123 **That blend — the audible, visible fusing of separate sounds into one word — 124 is the whole product.** It makes the invisible mental act of blending into 125 something a four-year-old can see happen, hear happen, and cause to happen. 126 127 Two modes, the second unlocking after the first is comfortable: 128 129 - **Cook** (decoding): the word is shown; the child sounds it out and blends it. 130 - **Take the order** (encoding): the customer *says* the word; the child 131 breaks it into sounds and picks the ingredients. Segmentation — the skill 132 with the larger measured effect size — as the natural inverse of cooking. 133 134 ### 4.3 Progression and difficulty 135 136 Words are introduced along a deliberate synthetic-phonics sequence, starting 137 with the handful of letter-sounds that unlock the most words, and favouring 138 stretchable continuant sounds early because they are far easier for a young 139 child to hold together while blending. 140 141 Progression is **mastery-based, not level-based**: a new sound is introduced 142 when the child is fluent with the ones they have, and the pace adapts to the 143 individual child. 144 145 ### 4.4 Persistent memory and spaced repetition 146 147 **The game remembers each child's words across sessions** — which sounds are 148 known, which words are solid, which are shaky, and when each was last seen. 149 Old words are re-injected on a spacing schedule so that mastery is maintained 150 rather than assumed. 151 152 This is expressed diegetically, which is the nicest part of the design: 153 **spaced repetition is regular customers coming back.** A creature you served 154 last week walks back up to the stall and orders the same thing again. Review is 155 not a drill; it is someone remembering you. 156 157 Memory is local to the device — no accounts, no backend, no data leaving the 158 machine. A simple way to reset, and to keep more than one child's progress on a 159 shared family device, is required. 160 161 ### 4.5 Fun and gamification 162 163 The game has to be genuinely fun to a four-year-old who does not care that it 164 is educational. Directions worth pursuing: 165 166 - **Characterful customers** with distinct looks, voices and reactions — 167 delight on being served, comic dismay at a wrong dish (never punishment). 168 - **A recipe book** that fills in as words are mastered — a visible, growing 169 collection, which is intrinsically motivating at this age. 170 - **Regulars and relationships**: recurring characters who become familiar, 171 whose return is a small event. 172 - **Stall and street growth**: the stall gets better equipment, decorations, 173 a bigger awning; the street around it slowly livens up as the child 174 progresses. Long-arc visible reward. 175 - **Everyday spectacle**: steam, bubbling, ingredients tumbling in, the pot 176 erupting when a word lands, the dish sliding across the counter. 177 - **Ambient island life** in the background — passers-by, boats, birds, 178 weather — as in Fire Truck. 179 180 ### 4.6 Age-appropriate design rules 181 182 Non-negotiable for the 4–5 age band: 183 184 - **No timers, no lives, no fail state.** A wrong answer is answered with a 185 re-demonstration and another try, never a loss. 186 - **Every instruction is spoken as well as shown.** The audience cannot read; 187 that is the entire premise. 188 - **Short natural stopping points** — a few minutes to a satisfying beat, with 189 a typical session in the ten-to-fifteen minute range that the research 190 supports. 191 - **Meaning is always checked**, not just pronunciation. A word that is 192 correctly sounded out should still resolve to a thing the child recognises, 193 so the game teaches reading rather than barking at print. 194 - **Accurate phoneme audio is a hard requirement.** Sounds must be taught 195 cleanly (a stretched /sss/, not "ess" and not "suh"); a schwa added to every 196 consonant actively breaks blending. This is the single most important 197 quality bar in the game, and the main reason real recorded audio is 198 preferred over synthesis. 199 200 ### 4.7 Input constraint — keyboard only 201 202 **The game must be fully playable with a keyboard alone. Mouse and touch input 203 must never be required.** 204 205 The child should be able to do everything — choose sounds, place them, stir, 206 serve, navigate menus — from the keyboard. Pressing letter keys to choose 207 sounds is both the obvious control scheme and pedagogically useful, since it 208 builds the letter-to-key association alongside the letter-to-sound one. 209 210 Pointer input may be supported as a convenience, but no mechanic may depend on 211 it, and no part of the game may be unreachable without it. On phones and 212 tablets, the existing on-screen keyboard approach used elsewhere in the project 213 satisfies this constraint without the game itself knowing about it. 214 215 *(How this is implemented is out of scope here — only that it is required.)* 216 217 --- 218 219 ## 5. Open questions 220 221 - **Title.** "Sound Soup" is a working name. Alternatives worth considering 222 once the art direction is settled. 223 - **Voice.** Who records the phonemes and words, and in what accent? This is a 224 significant content decision with pedagogical consequences, and it affects 225 which openly licensed audio sources are usable. 226 - **Art sourcing.** Whether a coherent enough open-license art set exists for 227 this specific tropical-fantasy-market look, or whether it needs commissioning. 228 - **Word list and picture coverage.** Which decodable words have a referent a 229 four-year-old reliably recognises in a single image. 230 - **Multi-child handling** on a shared device — how simple can it be while 231 still being useful. 232 233 ## 6. Recommended next steps 234 235 1. Settle the art direction and confirm that assets for it can be sourced under 236 compatible licenses. 237 2. Prove the phoneme audio quality bar — get clean, blendable sounds and 238 validate the blend ramp with a real four-year-old before building around it. 239 3. Draft the word list and teaching sequence, since that is the pedagogical 240 core and is independent of every technical decision. 241 4. Only then choose the engine and build. 242 243 --- 244 245 ## Sources 246 247 - [Developing Early Literacy Skills: A Meta-Analysis of Alphabet Learning and Instruction](https://pmc.ncbi.nlm.nih.gov/articles/PMC2910925/) 248 - [Letter-Sound Knowledge, Phoneme Awareness and RAN, but Not Letter-Sound Integration, Predict Variations in Reading Development](https://www.tandfonline.com/doi/full/10.1080/10888438.2019.1622546) 249 - [Alphabet Knowledge Trajectories and U.S. Children's Later Reading and Spelling](https://www.tandfonline.com/doi/full/10.1080/10888438.2025.2481064) 250 - [Phonemic Awareness: A Meta-Analysis for Planning Effective Instruction](https://www.researchgate.net/publication/361585439_Phonemic_Awareness_A_Meta-Analysis_for_Planning_Effective_Instruction) 251 - [A Meta-Analysis of Phonemic Awareness Instruction Provided to Children Suspected of Having a Reading Disability](https://pubmed.ncbi.nlm.nih.gov/35858272/) 252 - [The effectiveness of a classroom-based phonological awareness program for 4–5-year-olds](https://www.researchgate.net/publication/321196744_The_effectiveness_of_a_classroom-based_phonological_awareness_program_for_4-5-year-olds) 253 - [Combined Language and Code Emergent Literacy Intervention for At-Risk Preschool Children: A Systematic Meta-Analytic Review](https://srcd.onlinelibrary.wiley.com/doi/10.1111/cdev.14252) 254 - [Dynamic assessment of the effectiveness of digital game-based literacy training in beginning readers: a cluster randomised controlled trial](https://pmc.ncbi.nlm.nih.gov/articles/PMC10399564/) 255 - [Assessing the effectiveness of a game-based phonics intervention for first and second grade English language learners in India: A randomized controlled trial](https://onlinelibrary.wiley.com/doi/full/10.1111/jcal.12592) 256 - [Blending and segmenting — Five from Five](https://fivefromfive.com.au/phonics-teaching/essential-principles-of-systematic-and-explicit-phonics-instruction/blending-and-segmenting/) 257 - [Synthetic phonics](https://en.wikipedia.org/wiki/Synthetic_phonics) 258 - [Elkonin Boxes — Reading Rockets](https://www.readingrockets.org/classroom/classroom-strategies/elkonin-boxes) 259 - [Elkonin sound boxes: An evidence-based literacy strategy — Understood](https://www.understood.org/en/articles/evidence-based-literacy-strategy-elkonin-sound-boxes)