Meadow
Deliverable Review
Incorrect. Try again.
AbleNet · M1 Deliverable · Vocabulary

Vocabulary & Routines

Every word, every scene, every tier — how vocabulary grows with the child.
400+ words · 22 core at T1, 47 at T2, 74 at T3, 260+ scene
Emoji symbols (prototype) · AI-generated symbols in production
Fitzgerald Key color system · May 2026 · v9
Updated — v9 — May 28, 2026

Custom vocabulary added to V1 — “My Words”

In response to AbleNet review feedback on M1-002, the closed vocabulary decision (D7) has been revised. V1 now includes a personalizable vocabulary layer: ~27 preset slots for people, things, and places (child’s name, family members, pets, comfort objects, favorites) plus ~15 wildcard slots configurable by parents and SLPs. Custom words integrate into existing scenes, edges, and the sentence engine — they are not siloed. Additionally, the child’s communication voice has moved from ElevenLabs to Apple’s premium text-to-speech, ensuring custom words sound identical to bundled vocabulary. ElevenLabs is retained for the companion voice only. Sections marked UPDATED v9 below reflect these changes.

For AbleNet reviewer

Use this page to see the product scope decision: what words Meadow includes, how routines are organized, and how V1 stays intentionally focused instead of trying to cover the entire AAC market.

For SLP reviewer

Use this page to validate vocabulary selection, routine organization, omitted categories, and whether Meadow’s word system supports communication at the three tiers without becoming clinically brittle.

What this page decides

The core vocabulary footprint, routine structure, and boundaries of V1 scope.

01 At a glance
22 · 47 · 74
T1 · T2 · T3 core words
400+
Total vocabulary
7 · 4+
Routines · Scenes, expanding
≤3
Taps to any word

Meadow is built on one idea: children learn language through pictures, colors, and context — not text. Every word in the app has an emoji symbol, a color that teaches its grammatical role, and a place it lives in the child's world. Below is every word the child will have, shown exactly as they'll appear in the app.

Vocabulary density adapts to the child’s language milestone level. At the earliest milestones, children aren’t selecting vocabulary themselves — the app models words to them. As symbolic understanding develops and expressive vocabulary grows, the interface opens up. Communication is always the primary path.

ðŸŧ
Tier 1 · First Words

Exploring & Modeling

Guided Engagement

Pre-symbolic to emerging symbolic. The companion models vocabulary through caregiver co-use — spotlight, speak, pause, celebrate. No direct selection expected.

💎
Tier 2 · Word Combinations

Word Combinations

Routines + Grids + QuickChat

The child selects words — this is communication. Visual scenes with hotspots, grid displays, persistent core bar. Carrier phrases and QuickChat conversation loop emerge.

âœĻ
Tier 3 · Sentences

Sentences

Full Vocabulary

Expandable grids, sentence engine predictions, morphological modification. Full QuickChat with rich follow-ups. World map vocabulary expansion. The app gets out of the way.

📊 Read the language milestones & interaction design →
Yellow = People & Pronouns
Green = Verbs & Actions
Blue = Descriptors
Orange = Nouns
Purple = Social & Questions
Gray = Function Words
02 Core Vocabulary · Always available, every screen

The words that carry communication

Core words are the most-used words in any language. They live in the persistent compass frame, category drawers, and avatar — always available within three taps, always in the same place. Not all words are visible at once: the compass shows the right number for each developmental tier. Color teaches grammar through repetition.

15 tiles per row (page layout at 1440px × 72px min-width + 8px gap)
T1 = 21 compass words · T2 = 31 · T3 = 49 + categories
Total vocabulary below: 155 items across all tiers, categories, and avatar

Research basis & tier key

Each tile shows when it first becomes available. T1 = compass edge at 12–18mo (16 words). T2 = added at 18–24mo (36 cumulative). T3 = added at 24–48mo (72 total with categories). All = available at every tier (via avatar or global). Words behind category tiles (colors, manners, questions) are accessible within 2 taps but not directly on the compass edge.

Sources: Project Core 36 (UNC, Erickson & Koppenhaver), Banajee et al. (2003) toddler core, PrAACtical AAC First 12, AssistiveWare Crescendo, ASHA language milestones for body parts & early nouns.

Teal dot = Meadow addition beyond Project Core 36. “I love you” is a permanent, globally accessible phrase (Decision D11).

People & Pronouns · 15 words
T1ðŸ‘ĪI
T1👆you
T1ðŸ‘Đmom
T1ðŸ‘Ļdad
T1ðŸŦĩme
T2✋my
T2ðŸ‘Ķhe
T2👧she
T3✋mineEXTEND
T3ðŸ‘Ŧwe
T3ðŸ‘Ĩthey
T3👉it
T3👈thatEXTEND
T1🧒[child]MY WORDS
T1👧[sibling]MY WORDS
T1🐕[pet]MY WORDS
T1ðŸ‘ĩ[grandma]MY WORDS
T1ðŸŦ‚[friend]MY WORDS

My Words (v9): Dashed gold tiles are personalizable — parent types the real name (e.g., “Maya,” “Nana,” “Bongo”), and the label changes while the symbol stays or is replaced by a photo from camera roll. Mom, Dad, Grandma, Grandpa, and Teacher override their bundled counterparts when personalized. Up to 18 people slots, 5 things, 4 places, plus 15 wildcards. See the My Words system section below for the full catalog.

Verbs & Actions · 26 words
T1ðŸĪēwant
T1ðŸšķgo
T1ðŸīeat
T1ðŸĨĪdrink
T1🆘help
T1🛑stop
T2ðŸŦąhave
T2ðŸŦīneed
T2👀look
T1ðŸŽŪplay
T2👍like
T3ðŸŦģget
T3🏃do
T3👁ïļsee
T3💊can
T3ðŸ”Ļmake
T3ðŸŦģcome
T3ðŸĪgive
T3âĪïļlove
T3💭feel
T3ðŸ“Ĩput
T3📂open
T3🔄turn
T2👀look
T2🙅don’t
T3🎭pretend
Descriptors · 16 words
T1👐more
T1😋yummy
T1ðŸ”Ĩhot
T2🐘big
T2🐜little
T2⭐good
T2👎bad
T1ðŸĨķcold
T2😄funny
T2❌not
T2⏰now
T3ðŸĪĒyucky
T3🔀different
T3🟰same
T3🌐all
T3ðŸĪsome
Feelings · 13 words via avatar · immediate speech
All😊happy
AllðŸ˜Ēsad
All😠mad
AllðŸ˜Ļscared
AllðŸ˜ītired
AllðŸĪĪhungry
All💧thirsty
AllðŸĪ’sick
AllðŸĪ•hurt
AllðŸĪĐexcited
AllðŸ˜Ūsurprised
AllðŸ˜Īfrustrated
AllðŸĪŠsilly
Colors · 10 words via category tile at T2+
T2ðŸ”īred
T2ðŸ”ĩblue
T2ðŸŸĒgreen
T2ðŸŸĄyellow
T2🟠orange
T2ðŸŸĢpurple
T2ðŸĐ·pink
T2âšŦblack
T2⚩white
T2ðŸŸĪbrown
Body Parts · 8 words via avatar (My Body tab) at T2+ · companion models at T1
T1ðŸ—Ģïļhead
T1👀eyes
T1👃nose
T1👄mouth
T2👂ears
T2ðŸĪšhands
T2ðŸĶķfeet
T2ðŸŦƒtummy
Social & Questions · 14 words
T1✅yes
T1ðŸšŦno
T1âœĻall done
T1ðŸ˜Ŋuh oh
T1👋hi
T1ðŸĪšbye
T2🙏please
T2ðŸĪ—thank you
T3❓what
T3🗚ïļwhere
T3ðŸĪ·who
T3ðŸĪ”why
T3🕐when
All💛I love you• global
Function Words · 16 words
T2ðŸ“Ķin
T2📍on
T2⮆ïļup
T2👇here
T2🚊out
T2ðŸ’Ąoff
T2➡ïļto
T2ðŸĪwith
T2🎁for
T2➕and
T3⚡is
T3📎of
T3👉there
T3⮇ïļdown
T31ïļâƒĢa
T3👆the
Numbers · 10 words NUMBERS category tile · top edge at T2+ · companion models counting at T1
T21ïļâƒĢone
T22ïļâƒĢtwo
T23ïļâƒĢthree
T24ïļâƒĢfour
T25ïļâƒĢfive
T26ïļâƒĢsix
T27ïļâƒĢseven
T28ïļâƒĢeight
T29ïļâƒĢnine
T2🔟ten
Alphabet · 26 letters SLP-toggleable · keyboard overlay at T3 (exceeds 10-item fan-out max)
T3A
T3B
T3C
T3D
T3E
T3F
T3G
T3H
T3I
T3J
T3K
T3L
T3M
T3N
T3O
T3P
T3Q
T3R
T3S
T3T
T3U
T3V
T3W
T3X
T3Y
T3Z
03 Routines & Scenes · Where words live in the child’s day

Words are organized by what the child is doing — daily routines like mealtimes, bath time, and bedtime. Each routine opens a scene — an illustrated context where vocabulary lives. Tap a routine or scene below to see every word inside it. See Interaction Design for how routines and scenes connect to the compass frame.

Routines

ðŸĨĢ

Breakfast

Kitchen · morning
ðŸĨĢ cereal ðŸĨž pancakes ðŸĨ› milk +11 more
ðŸĨŠ

Lunch

Kitchen · midday
ðŸĨŠ sandwich 🧀 cheese 🐠 goldfish +14 more
🍝

Dinner

Kitchen · evening
🍗 chicken 🍝 pasta ðŸĨ• carrots +15 more
🛁

Bath Time

~30 words · Bathroom
ðŸŦ§ bubbles ðŸĨ duck ðŸ’Ķ splash +27 more
👕

Getting Dressed

~18 words · Bedroom
👕 shirt 👖 pants 👟 shoes +15 more
🌙

Bedtime

~19 words · Bedroom
ðŸ§ļ teddy 📖 story 🌟 goodnight +16 more
ðŸ§ļ

Playtime

~35 words · Living Room
âš― ball ðŸ§ą blocks ðŸŦ§ bubbles +32 more

Scenes

ðŸģ

Kitchen

~110 words · breakfast + lunch + dinner + snacks + cooking
ðŸĨĢ cereal 🍕 pizza 🍗 chicken 🍊 cookies +106 more
ðŸšŋ

Bathroom

~30 words · all bath, potty & hygiene vocabulary
ðŸš― potty ðŸŠĨ brush 🧞 soap +12 more
🛏ïļ

Bedroom

~35 words · all clothing, sleep & nighttime vocabulary
ðŸ§ļ teddy 📚 book 👕 pajamas +15 more
🛋ïļ

Living Room

~35 words · all toys, play & activity vocabulary
âš― ball 📚 TV ðŸ§Đ puzzle +17 more

Vocabulary categories

ðŸ‘Ļ‍ðŸ‘Đ‍👧

People

~15 words
ðŸ‘ĩ grandma 🧒 friend ðŸ‘ķ baby +12 more
ðŸū

Animals

~18 words
🐕 dog ðŸą cat ðŸĶ lion +15 more
💛

Feelings

~13 words · via avatar
😊 happy ðŸ˜Ē sad ðŸ˜Ļ scared +10 more
ðŸ―ïļ

Meals · Kitchen

~110 words organized by meal — food is the most frequent communication context for toddlers. Each meal surfaces age-appropriate vocabulary a child actually encounters.

ðŸĨĢ Breakfast · morning
ðŸĨĢcereal
ðŸĨ›milk
ðŸĨžpancakes
🧇waffles
ðŸĨ–toast
ðŸĨšeggs
ðŸŦ—oatmeal
ðŸŦ™yogurt
🍌banana
🍓strawberries
ðŸŦblueberries
ðŸđjuice
🧁muffin
ðŸĨŊbagel
ðŸĨŠ Lunch · midday
ðŸĨŠsandwich
🧀cheese
ðŸŦ”grilled cheese
ðŸĨœpeanut butter
🟊jelly
🍕pizza
ðŸŦ•mac & cheese
🍖nuggets
🌭hot dog
🍎apple
🍇grapes
🍘crackers
🐠goldfish
🧃juice box
🍜noodles
ðŸŒŊburrito
ðŸĨŦcanned soup
🍝 Dinner · evening
🍗chicken
🍝pasta
🍚rice
🍜soup
🍔hamburger
ðŸŒŪtaco
🍠fries
ðŸĄfish sticks
🍞bread
ðŸĨ•carrots
ðŸŦ›peas
ðŸĨĶbroccoli
ðŸŒ―corn
ðŸĨ”potato
💧water
🍝spaghetti
ðŸŦ˜beans
ðŸĨĄleftovers
🍇 Fruits & Produce
🍑peach
🍐pear
ðŸŦplum
🍍pineapple
ðŸĨ­mango
ðŸĨkiwi
🍈melon
🍅tomato
ðŸĨ’cucumber
ðŸĨ—salad
🍊 Snacks & Treats
🍊cookies
🍟chips
ðŸĨĻpretzels
ðŸŋpopcorn
🍉watermelon
🍊orange
🍒cherries
ðŸĶice cream
🎂cake
ðŸŊjam
ðŸŸĨketchup
🍋lemonade
ðŸĨĪsmoothie
ðŸŦgranola bar
ðŸŪpudding
ðŸĨ§pie
🍘rice cakes
ðŸģ Cooking & Appliances
ðŸģpan
ðŸŦ•pot
â™Ļïļoven
ðŸ“ŧmicrowave
🍞toaster
ðŸ‘Đ‍ðŸģcook
🧁bake
ðŸĨ„stir
ðŸ―ïļ Tableware & Mealtime Words
ðŸ―ïļplate
ðŸĨ˜bowl
🏆cup
🍞sippy cup
ðŸīfork
ðŸĨ„spoon
🧋straw
ðŸ§ŧnapkin
🊑high chair
🍞bib
🧈butter
ðŸŸĄmustard
🧊ice
ðŸīeat
ðŸĨĪdrink
ðŸŦ—pour
ðŸ§đclean
🧞wash
😋yummy
ðŸĪĒyucky
ðŸĪĪhungry
ðŸ”Ĩhot
ðŸĨķcold
ðŸŒĄïļwarm
ðŸŦ messy
✅all done
ðŸ§ļ

Playtime · Living Room

~35 words — play is communication. Toys, creative activities, and movement vocabulary for social interaction and turn-taking.

Toys
âš―ball
ðŸ§ąblocks
🚗car
🚂train
🎎doll
ðŸ§ļteddy
ðŸĶ•dinosaur
ðŸ§Đpuzzle
ðŸŦ§bubbles
🎈balloon
📚book
📚TV
ðŸŽĩmusic
Creative
🖍ïļcrayons
🖌ïļpaint
📄paper
✂ïļscissors
⭐stickers
ðŸŦķplay-doh
ðŸŽĻdraw
🖍ïļcolor
ðŸ§ąbuild
Movement & Social
💃dance
ðŸŽĩsing
🏃run
ðŸĶ˜jump
🙈hide
ðŸĪshare
🔄my turn
ðŸŦīcatch
ðŸĪūthrow
ðŸĶķkick
ðŸŦļpush
🎉fun
🛏ïļ

Getting Dressed + Bedtime · Bedroom

~35 words — two routines share this room. Getting Dressed covers clothing and dressing actions. Bedtime covers sleep, comfort, and critical safety words (scared, dark).

👕 Getting Dressed · Clothing
👕shirt
👖pants
👟shoes
ðŸ§Ķsocks
ðŸĐēunderwear
👗dress
ðŸ§Ĩcoat
ðŸ§Ēhat
👕pajamas
ðŸĪzipper
🚊closet
🗄ïļdrawer
👕 Getting Dressed · Actions
👆put on
👇take off
👈pick
🆘help
ðŸĪzip
ðŸŦēsnap
🌙 Bedtime
🛏ïļbed
🛌pillow
ðŸ§Ģblanket
ðŸ§ļteddy bear
📚book
ðŸ’Ąlamp
🌙nightlight
ðŸ˜īsleep
⏰wake up
ðŸĪ—hug
😘kiss
ðŸŽĩsing
📖story
💭dream
🌑dark
ðŸ˜Ļscared
ðŸĨ°cozy
ðŸ˜īsleepy
🌟goodnight
🛁

Bath Time · Bathroom

~30 words — potty training vocabulary is highest-priority fringe for 18-36 months. Bath routine covers hygiene, body awareness, and sensory language.

Potty
ðŸš―potty
ðŸšŧtoilet
ðŸ§ŧtoilet paper
ðŸ§ŧwipes
ðŸ‘ķdiaper
🌊flush
ðŸĶĻstinky
Bath & Shower
🛁bathtub
ðŸšŋshower
🧞soap
ðŸ§īshampoo
ðŸŦ§bubbles
🧖towel
ðŸ§īlotion
ðŸĨrubber duck
ðŸ’Ķsplash
ðŸ§―wash
💧wet
âœĻclean
ðŸŒĄïļwarm
ðŸĨķcold
Sink & Teeth
🚰sink
ðŸŠĨtoothbrush
ðŸĶ·toothpaste
🊞mirror
😁brush teeth
ðŸŦ§wash hands
ðŸšŋrinse
ðŸ§đdry off
ðŸ‘Ļ‍ðŸ‘Đ‍👧

People

~15 bundled words + up to 18 personalized via My Words — family, friends, and caregivers with real names and photos

Family
ðŸ‘Đmom
ðŸ‘Ļdad
ðŸ‘Ķbrother
👧sister
ðŸ‘ķbaby
ðŸ‘ĩgrandma
ðŸ‘īgrandpa
ðŸ‘ąâ€â™€ïļaunt
ðŸ‘ąuncle
ðŸĪļcousin
Others
ðŸŦ‚friend
ðŸ‘Đ‍ðŸŦteacher
ðŸ‘Ļ‍⚕ïļdoctor
🏠neighbor
🧑babysitter
ðŸū

Animals

~18 words — pets, farm, and wild animals

Pets
🐕dog
ðŸącat
🐟fish
ðŸĶbird
ðŸđhamster
🐰bunny
ðŸĒturtle
Farm
🐄cow
ðŸīhorse
🐷pig
🐔chicken
🐑sheep
ðŸĶ†duck
🐐goat
Wild / Zoo
ðŸĶlion
🐘elephant
🐒monkey
ðŸŧbear
💛

Feelings

~13 words — accessed via the avatar’s feelings overlay. Tap a feeling and it speaks immediately with its ASL sign. Expressing a feeling is a communicative act, not a request for a conversation.

😊happy
ðŸ˜Ēsad
😠mad
ðŸ˜Ļscared
ðŸ˜ītired
ðŸĪĪhungry
💧thirsty
ðŸĪ’sick
ðŸĪ•hurt
ðŸĪĐexcited
ðŸ˜Ūsurprised
ðŸ˜Īfrustrated
ðŸĪŠsilly
04 Two modes · What happens when the child taps a word

One vocabulary, three interaction tiers

The developmental level page defines the three canonical tiers: First Words, Word Combinations, and Sentences. Tier 1 stays in direct speech. Tiers 2 and 3 add increasing sentence scaffolding. Tier controls what happens after a tap, not what vocabulary is available.

1
First Words
Direct speech
2
Word Combinations
transition zone
SLP decides
3
Sentences
Sentence scaffolding
Tier 1

Tap → Word

Tier 1 · First Words

Every tap speaks one word and shows its sign. No sentence scaffolding, no follow-ups, no choices to make. The child learns the fundamental insight: touching the screen makes words happen.

Child taps 🍕 pizza:
1 TTS speaks: “Pizza!”
2 Signing bubble shows the ASL sign for “pizza”
3 Done — ready for next tap
Tiers 2-3

Tap → Phrase / sentence options

Tiers 2–3 · Word Combinations through Sentences

Tapping a word opens sentence suggestions at multiple MLU levels. The child picks one, it speaks with signing, and three follow-up options appear. Communication becomes conversation.

Child taps 🍕 pizza:
1 Four sentence levels appear (see Interaction Design)
2 Child taps “I want pizza” → speaks + signs
3 Three QuickChat follow-ups appear
4 Loop continues — speak, choose, speak

Guided engagement adapts to the tier

Tapping the companion (bottom-right corner) starts a guided engagement cycle in any routine or scene. What happens during the cycle depends on the tier.

ðŸŧ Tier 1

Word modeling

Spotlight a scene object → speak the word → show the sign → longer expectant pause → celebrate any response. The app models language to the child. Bigger celebrations, more patience. This is the primary interface at Tier 1.

ðŸŧ Tiers 2-3

Sentence scaffolding

Spotlight a scene object → present a fill-in sentence → child completes the gap → sentence speaks + signs. Shorter pauses, sentence-level modeling. Speak With Me becomes a practice partner for building phrases.

Tier is not a gate

All three tiers have access to the full vocabulary. A Tier 1 child who taps “want” then “pizza” will speak both words separately — they won’t be prevented from tapping multiple words. Tier controls what the app presents (sentence templates, follow-ups), not what the child can do.

05 Design principles · Why it works this way

Built for small hands and growing minds

Every design decision serves a clinical purpose. Colors teach grammar. Fixed positions build motor memory. Conversations happen through rhythm, not complexity.

🖞ïļ

Every word gets a picture

Pre-literate children navigate by symbol. Text labels are secondary, for caregivers. No text-only tiles, ever.

ðŸŽĻ

Color is grammar instruction

Fitzgerald Key colors teach sentence roles through repetition. Yellow = who, green = doing, blue = describing, orange = what.

📌

Fixed positions, always

Words never move. Motor planning depends on muscle memory — "want" is always in the same spot, every session.

🔓

No prerequisites

All vocabulary accessible from day one. No “readiness” gates, no “must demonstrate X before accessing Y.” ASHA’s position statement is unambiguous: “AAC should be introduced as early as possible” and “candidacy models that require prerequisite cognitive or linguistic skills are not supported by research” (ASHA, 2004; reaffirmed 2016). Beukelman & Light (2020) confirm: denying AAC access based on readiness criteria delays communication development without clinical benefit. The companion models vocabulary the child isn’t yet using independently — exposure precedes production, as it does in natural language development.

💎

Conversation, not statements

QuickChat's speak-choose-speak loop teaches turn-taking. Communication doesn't end after one sentence.

ðŸ“ī

Offline first

No internet required for any communication feature. The child's voice never depends on a WiFi connection.

ðŸŽŊ

60pt minimum touch targets

Toddler fingers are imprecise. More whitespace than feels right. Missing a button is missing a word.

💛

Feelings are immediate

Tapping "sad" speaks it instantly. Expressing a feeling is not a request for a conversation — it's a complete communicative act.

ðŸĪŸ

Three modalities at once

Every tap fires three things simultaneously: the picture highlights, the word speaks, and an ASL sign animates. Research shows pre-verbal children learn vocabulary fastest when visual, auditory, and gestural channels activate together. This is the gold standard used by Ms. Rachel and validated by SLP research.

👂

Expectant pausing

After speech fires, the app waits. A gentle listening animation holds for 3–45 seconds (configurable) before the UI resets. Toddlers need up to 45 seconds to process and attempt production. Most apps rush to the next input — Meadow holds space for the child's response.

🎉

Warm celebration

After every word tap: a brief “You said ___! Great job!” in warm TTS, paired with a gentle animation. Not slot-machine rewards — the feeling of a delighted adult responding to a child's communication attempt. Immediate positive reinforcement builds the confidence to try again.

🌅

Routine-anchored scenes

Vocabulary is organized by the child’s daily routines — mealtime, bath time, getting dressed, bedtime, playtime — not abstract categories. Words have immediate functional value when they appear in the context where the child actually needs them. This is how Ms. Rachel teaches vocabulary and how toddlers naturally build word-to-world connections.

ðŸ‘‚âžĄïļðŸ—Ģïļ

Receptive before expressive

Children understand words before they produce them. Meadow mirrors this: the companion models vocabulary receptively (the child hears and sees the word in context) before that vocabulary appears on compass edges for independent expressive use at the next tier. At T1, the companion models colors, body parts, and counting — the child absorbs these through exposure. At T2, those words graduate to compass placement for expressive production. This receptive-then-expressive arc is how natural language develops and how SLPs scaffold AAC intervention.

Research basis: Principles marked with triple-modality, expectant pausing, warm celebration, and routine-anchored scenes are drawn from Ms. Rachel (Songs for Littles) — six SLP-validated techniques with 17M subscriber validation. No AAC app currently incorporates these techniques. Sources: ASHA Leader, AAP Guidance, Global Speech Therapy, Aulad journal semiotic study. Full research: The Ms. Rachel Playbook (2026-05-17).

06 Tier progression · How vocabulary grows with the child

The frameworks behind vocabulary placement

Every word in Meadow has a home — a compass edge position, a category fan-out, an avatar panel, or a scene. These frameworks govern where words go, how they get there across tiers, and what rules prevent the system from breaking as vocabulary scales. This section is for SLPs and clinical reviewers who want to see the logic, not just the word list.

Companion as vocabulary multiplier

At Tier 1, the child is not expected to independently navigate to most vocabulary. The companion — activated by the parent or SLP — models words to the child from the entire vocabulary database, not just compass edge words. This means vocabulary like colors, body parts, and counting can be modeled at 12 months without consuming compass edge space. When the child reaches T2 and is ready to use those words independently, they graduate to compass placement — and the child reaches for words they already recognize.

Tier 1 · Companion leads
scene ðŸŧ 24 tiles

Parent taps companion → companion spotlights scene objects, speaks words, shows signs. Colors, body parts, counting — all modeled by the app, not navigated by the child.

Tier 2 · Shared control
scene 49 tiles

Words the companion modeled at T1 become independently accessible. COLORS and NUMBERS category tiles appear on compass edges. The child reaches for vocabulary they already recognize.

Tier 3 · Child leads
scene 66 tiles

Full vocabulary on compass edges and categories. The child navigates world map scenes, builds sentences independently. Companion becomes a practice partner.

Design implication: Vocabulary the child can’t independently use yet (colors at 12mo, counting at 14mo) doesn’t need compass edge space at T1. It needs companion modeling access. When the child reaches T2, companion-modeled items graduate to compass placement.

Why the companion models, not just the parent: Aided language stimulation research (Binger & Light, 2007; Solomon-Rice & Soto, 2014) shows children learn AAC fastest when a communication partner models it in context. The companion does this directly — but it also teaches the parent by demonstration. Parents watch the companion spotlight a scene object, speak the word, and show the sign. They absorb the technique naturally, the same way parents learn vocabulary strategies from watching Ms. Rachel with their child. The companion is a modeling partner for the child and a live tutorial for the parent — one interaction serves both. This matters because parents of AAC users consistently report feeling overwhelmed by the learning demands of AAC systems (PMC, 2024). The companion lowers that barrier by showing, not telling.

Same word, same position, different behavior

Words that appear at T1 persist through T2 and T3 — same compass position, same color, same icon. What changes is what the app does after the tap. Motor memory accumulates across tiers. The child never has to re-learn where a word lives.

Tier 1 · Direct speech
🙏 WANT “want!”

One tap, immediate speech. Done.

Tier 2 · Word combinations
🙏 WANT ðŸĨ› milk ➕ more ðŸĨž pancakes

Tap → speech + bubble offers scene-aware combos.

Tier 3 · Sentence engine
🙏 WANT I want ___ want more don't want

Tap → sentence engine predicts multi-word paths.

Applies to all tier-spanning words: more, help, stop, yes, no, go, eat, drink, I, you, mom, dad. The word is stable. The scaffolding around it grows.

Edge assignment rules

Each compass edge has a grammatical identity governed by the Fitzgerald Key. Words assigned to an edge stay there permanently. New words at higher tiers fill open positions — nothing moves.

DESCRIPTORS + FUNCTION more · big · hot · not · to · with · in · on VERBS + SOCIAL want · go · eat · help · yes · no · all done ROUTINES PEOPLE SCENE contextual nouns Grid World Me Guide
Top edge · Descriptors + Function

Blue descriptors (more, big, hot, cold…) and gray function words (to, with, in, on, up…). Category tiles: COLORS, NUMBERS, PLACES. Row 1 = descriptors. Row 2 = function words + quantifiers.

Bottom edge · Verbs + Social

Green verbs (want, go, eat, help…) and purple social words (yes, no, all done). Category tiles: MANNERS, QUEST, MORE VERBS. Row 1 = T2+ additions. Row 2 = T1 core verbs (never move).

Left edge · Routines

Orange routine tiles — navigational, scene-switching. The child’s daily life: eating, bath, dress, bed, play, outside, car, store. Each opens a scene with contextual nouns.

Right edge · People + Pronouns

Yellow people/pronouns — I, you, mom, dad at T1. T2 adds me, my, he, she. T3 replaces individuals with FAMILY category (9 people behind 1 tile). Category tiles: FAMILY, EXTEND.

Immovability rule: Once a word is assigned a position, it never moves — not across tiers, not across scenes. Category tiles follow the same rule. Non-negotiable for motor planning: the child builds muscle memory that accumulates across months and years of use.

Category fan-out rules

Category tiles multiply vocabulary reach without consuming edge space. One tile gives access to many words — but categories have constraints to prevent cognitive overload and maintain the ≤3 tap guarantee.

ðŸŽĻ COLORS 1 tap fan-out · max 10 items · 1 level deep ðŸ”ī red ðŸ”ĩ blue ðŸŸĒ green ðŸŸĄ yellow 🟠 orange ðŸŸĢ purple 💗 pink âšŦ black ⚩ white ðŸŸĪ brown 2 taps total
10
Maximum 10 items per fan-out. COLORS has 10, FAMILY has 9. Fan-outs exceeding 10 items use a different pattern (keyboard overlay, scrollable panel).
1
One level only — no nested categories. Fan-out items are always leaf words. Category → word = 2 taps. Never category → category → word. This prevents depth-based confusion.
FK
Categories cluster with their Fitzgerald Key group. COLORS (blue) on the top edge. FAMILY (yellow) on the right edge. MANNERS (purple) on the bottom edge. No cross-edge placement.
60
Fan-out items keep 60pt minimum touch targets. No shrinking items to fit more. If the category exceeds 10, it graduates to an overlay pattern.

Category contents — what’s behind each tile

COLORS · top edge · T2+

red, blue, green, yellow, orange, purple, pink, black, white, brown (10)

NUMBERS · top edge · T2+

one, two, three, four, five, six, seven, eight, nine, ten (10)

MANNERS · bottom edge · T2+

hi, bye, please, thank you (4)

FAMILY · right edge · T3

mom, dad, brother, sister, grandma, grandpa, teacher, friend, baby (9)

QUEST · bottom edge · T3

what, where, who, why, when (5)

MORE VERBS · bottom edge · T3

make, come, give, love, feel, put, open, turn (8)

EXTEND · right edge · T3

mine, that (2 + SLP-configurable)

PLACES · top edge · T3

under, behind, next to, between, above, below (6 · spatial prepositions)

Exception — Alphabet: 26 letters exceed the 10-item fan-out maximum. The alphabet uses a keyboard overlay pattern — a full-screen letter grid triggered by an SLP-enabled category tile at T3. This is a literacy tool, not core communication vocabulary, and is SLP-toggleable per child.

Vocabulary housing map

Every word needs a home. This decision tree governs placement — applied below to all vocabulary items that section 02 catalogues but the compass frame mockup doesn’t yet place.

New word Always available? yes ðŸ‘Ī Avatar no Primary communication word? yes 📍 Edge · 1 tap no Group of related items? yes 📂 Category · 2 taps no Bound to a scene? yes 🖞ïļ Scene noun · 2 taps no SLP-specific tool? yes 🔧 SLP overlay ðŸŧ Companion repertoire feelings, I love you want, go, help, more colors, numbers, family milk, spoon, rubber duck alphabet, keyboard colors at T1, body parts

Housing decisions

Body parts (8) T1-T2 Avatar panel (My Body tab alongside How I Feel). 2 taps. Companion models at T1 during bath routine; child accesses independently at T2+. Body parts are self-referential — they belong with the avatar (“me”).
Numbers 1–10 (10) T2+ NUMBERS category tile, top edge at T2+. 2 taps. Companion models counting at T1 during play/meal scenes. Numbers are quantity modifiers — sit with descriptors.
in/on/up/here/out/off (6) T2+ Direct on top edge row 2 at T2+. 1 tap. Critical sentence-building prepositions — “put IN,” “get OUT,” “go UP.” Too important for a category.
Alphabet A–Z (26) T3 · SLP Keyboard overlay, SLP-toggleable. Exceeds 10-item fan-out max. Accessed via LETTERS tile on top edge at T3 when enabled. Literacy tool, not communication vocabulary.
My Words — people (up to 18) T1+ Right edge positions 5+ at T1, scrollable via swipe. Child, siblings, pets, grandparents, teacher, therapist, friends, babysitter. Configured via the My Words tab in the parent area (not during onboarding). Photo from camera roll replaces default emoji. Custom names override bundled counterparts (e.g., “Mama” replaces “mom”).
mine, that (2) T3 Behind EXTEND category on right edge at T3. 2 taps. Extended pronouns for sentence construction.

Scene nouns vs core words

Vocabulary splits into two systems with different rules. Understanding this boundary is essential for vocabulary planning.

more · big · hot want · eat · help ðŸģ Kitchen milk · juice · cereal · spoon cup · water · cookie
same
edges
different
nouns
more · big · hot want · eat · help 🛁 Bathroom towel · soap · bubbles water · duck · brush
Core words · Compass edges
  • Fixed position — never move across tiers or scenes
  • Present on every screen regardless of scene
  • Grammatical vocabulary: verbs, pronouns, descriptors, function words, social
  • ~55–65 direct words + categories at T3
Scene nouns · Center stage
  • Change when the child navigates to a different scene/routine
  • Contextually bound — cereal in kitchen, rubber duck in bath
  • Concrete nouns: food, objects, animals, clothing
  • ~15–25 per scene × 15+ scenes at scale = 260+ words

Cross-scene nouns: Some nouns appear in multiple scenes — “water” in kitchen and bath, “cup” at mealtime and bedtime. Same word, same icon, same data. The sentence engine recognizes them regardless of which scene surfaced them.

The synergy: Core words (edges) + scene nouns (center) = complete communication. “Want” (edge) + “milk” (kitchen scene) = “want milk.” The scene provides contextual awareness that makes edge words smarter — the sentence engine knows where the child is and offers relevant continuations. This is the differentiator. See Compass Frame.

Compound words and multi-word phrases

Some vocabulary items span two or more words but function as a single concept in the child’s communication. These are represented as single tiles with a unified TTS clip, a single symbol, and one Fitzgerald Key color assignment. The child taps once to communicate the whole phrase.

✓ Correct — single tile
ðŸĶ ice cream 1 tap · 1 concept
vs.
✗ Wrong — split tiles
🧊 ice + ðŸĨ› cream 2 taps · 2 concepts · wrong
The three-part rule — when a phrase becomes a tile
1. Conceptual unity: The phrase names a single concept that cannot be decomposed without losing meaning. “Ice cream” is one thing — it is not “ice” + “cream.”
2. Single referent: The phrase points to one object, action, or state. “Belly button” is one body part. “All done” is one communicative act (cessation).
3. Toddler frequency: The phrase appears in early vocabulary inventories (CDI, LENA, MCDI) as a unit. Children learn “high five” as a chunk, not as an adjective + a number.

All three must be true. “Want milk” fails — it is a sentence, not a concept, and the child should compose it via the sentence engine.

Already single tiles

all done · thank you · I love you · uh oh · night night · high five

Scene nouns as single tiles

ice cream · peanut butter · belly button · teddy bear · french fries · apple juice

TTS: One clip for the full phrase. “Ice cream” speaks as a natural unit, not two spliced words. Symbol: One image representing the concept. The ice cream cone, not separate images for ice and cream. FK color: Assigned to the whole phrase. “All done” is purple (social). “Ice cream” is orange (noun). Sentence engine: Compound tiles participate like any word. “I want ice cream” composes naturally — the engine treats “ice cream” as one noun token.

Morphological marking — inflections without tiles

Children at Brown’s Stages II–IV (24–48 months) begin producing grammatical inflections: present progressive -ing, regular plural -s, past tense -ed, possessive ’s. These emerge in a predictable sequence. Meadow handles them through the sentence engine, not separate vocabulary tiles.

ðŸ―ïļ eat child taps ⚙️ sentence engine eating -ing progressive ate -ed past tense eats -s third person
-ing progressive Child taps “eat” → sentence engine offers “eating” as a variant when context is ongoing action. “I eating” → TTS speaks “I’m eating.”
-s plural Child taps a number then a noun → TTS automatically pluralizes. “Two cookie” speaks as “two cookies.”
-ed past tense Sentence engine offers past forms as predictions. After “I” → predictions include “ate,” “went,” “played” alongside base verbs.
’s possessive After a person word → engine offers possessive form. “Mommy” → prediction includes “mommy’s.” TTS speaks the possessive naturally.

Why not separate tiles? Adding -ing, -s, -ed, and ’s as tap targets would require the child to understand morphological decomposition — breaking “eating” into “eat” + “-ing” — which is a metalinguistic skill that develops after morphological production. Toddlers say “eating” as a whole word; they don’t consciously add “-ing.” The sentence engine mirrors this by handling inflections at the output layer. This is consistent with how Proloquo2Go and LAMP handle morphology for pre-literate users.

Morphological marking is Tier 3 only. At Tier 1 and Tier 2, words speak in their base form. The full morphology system is documented in the Sentence Engine spec.

Sentence engine replaces SCS templates

Traditional AAC apps scaffold sentence construction with SCS (sentence construction suggestions) — static, pre-authored 2–4 word templates displayed in a fixed UI area. Meadow replaces these with a grammar-aware prediction engine inside the speech bubble. This is a significant departure from standard practice. Here is why.

STATIC SCS TEMPLATES I want ___ I see ___ I like ___ same in every scene ✗
GRAMMAR-AWARE PREDICTIONS ðŸģ Kitchen scene want milk · juice · cereal more pancakes · water all done eating
Static SCS templates
  • Pre-authored: “I want ___” “I see ___”
  • Fixed options regardless of context
  • Same templates in every scene
  • Compete for center stage space
  • Author-limited — can only compose phrases someone wrote
Grammar-aware predictions
  • Dynamic: predictions adapt to what the child has already said
  • Scene-aware: kitchen scene offers food nouns, not toys
  • Usage-ranked: the child’s most-used words surface first
  • Lives in the speech bubble — center stage stays scene-only
  • Open-ended — the child composes novel sentences

Clinical rationale: The sentence engine serves the same clinical goals as SCS — reducing cognitive load, scaffolding multi-word utterances, and increasing MLU — but through dynamic rather than static means. The transition rules map directly to Fitzgerald Key categories that SLPs already use. Predictions follow Brown’s Stages II–IV developmental expectations (Agent+Action, Action+Object, S+V+O). The child can always override predictions with manual word selection. This aligns with aided language stimulation (ALgS) principles: model language in context, respond to the child’s communicative intent, and provide accessible next steps without restricting choice.

The grammar transition table is documented in the Sentence Engine spec and requires SLP validation against Brown’s Stages before implementation.

07 Voice & sound design

How Meadow sounds

Sound is the product. Every tap produces the child’s voice — warm, age-appropriate, and consistent across every word, including custom names. The companion has its own distinct voice so the child always knows which voice is theirs.

ðŸ—Ģïļ

Child’s communication voice

Apple’s premium text-to-speech (AVSpeechSynthesizer with .premium quality). Parent or SLP selects a voice during profile setup. One voice engine handles every word — bundled vocabulary, custom names, sentence engine output — so the child never hears a voice split mid-sentence. Works offline, zero network dependency, zero audio asset pipeline.

ðŸ§ļ

Companion voice (My Buddy)

Warm, playful voice generated via ElevenLabs Voice Design. Used for Speak With Me prompts, celebrations, and encouragement. Pre-recorded for scripted companion interactions. Vocally distinct from the child’s communication voice — the contrast is a feature.

Emotional variation

The companion voice carries three emotional tones via pre-recorded ElevenLabs clips. The child’s communication voice uses speech rate and pitch modulation for contextual emphasis — not pre-recorded tonal variants, but real-time adjustment based on interaction context.

Neutral

“Juice.” — Calm, clear. The default for ~70% of taps. Requests, labels, statements.

Expressive

“Juice!” — Excited, delighted. Used in celebrations, Speak With Me discovery, and reward moments.

Urgent

“Juice.” — Insistent, needs-based. For feelings (“hurt,” “scared”) and escalation when the child really needs something.

Why Apple TTS, not ElevenLabs, for the child’s voice

Custom vocabulary means arbitrary text the parent types. If bundled words used pre-recorded ElevenLabs clips and custom words used Apple TTS, the child would hear two different voices — potentially in the same sentence (“Mama wants…” with a voice split between “Mama” and “wants”). For children with ASD who rely on predictability, this inconsistency undermines the “this is MY voice” feeling. A single voice engine for everything means: no pre-recorded audio clips (~25–30 MB saved from bundle), inherently extensible to any word the parent or SLP adds, and zero voice consistency concerns. Apple’s premium iOS 17+ voices are natural enough for an SGD — warm, clear, and recognizably child-like.

Why this matters: Voice quality directly affects AAC device adoption. Research shows families abandon devices when the voice feels robotic or age-inappropriate. Meadow’s approach ensures that every word the child speaks — whether it comes from the bundled vocabulary or is a custom name typed by the parent — sounds like the same person. The companion’s distinct ElevenLabs voice provides warmth and personality where vocal contrast is a feature, not a bug.

08 For SLP review

Key questions for your review

Core vocabulary: The compass frame holds 16 core words at T1 (growing to 49+ at T3), plus 13 feelings via avatar, 10 colors, 8 body parts, numbers 1–10, alphabet A–Z (SLP-toggleable), and “I love you” as a permanent global phrase. Drawn from Project Core 36, Banajee et al. (2003), PrAACtical AAC, and ASHA milestones. Are the tier assignments and omissions clinically justified?
Tier progression logic: Section 06 documents how vocabulary placement changes across tiers — companion-mediated modeling at T1, category graduation at T2, full edge access at T3. Does this guided-to-independent progression align with how SLPs typically scaffold AAC access for 12–48mo children?
Category fan-outs: Section 06 specifies contents for all 8 category tiles (COLORS, NUMBERS, MANNERS, FAMILY, QUEST, MORE VERBS, EXTEND, PLACES). Are these groupings clinically defensible? Are any categories missing or misassigned?
Color assignments: Are the Fitzgerald Key category assignments clinically defensible, especially where Meadow departs from the most literal mapping?
Routine organization: Is routine-based fringe vocabulary the right organizing model for this age range? Any missing high-frequency home routines?
QuickChat and templates: Is speak-choose-speak appropriate for Tiers 2–3, and does the MLU-to-level mapping reflect the developmental sequence without over-scaffolding?
Closed vocabulary: ANSWERED (v9) — AbleNet feedback confirmed that fully closed vocabulary is not acceptable. V1 now includes the My Words personalizable vocabulary layer: ~27 preset slots (child’s name, family, pets, comfort objects, favorites) + 15 wildcard slots for SLP/parent customization. Custom words use the same Apple TTS voice, integrate into existing scenes and edges, and participate in the sentence engine. See Section 07 voice changes and the full catalog in the spec.
Compound words: Section 06 documents how multi-word phrases (“ice cream,” “all done,” “belly button”) are handled as single tiles with unified TTS clips. Is the three-part rule (conceptual unity, single referent, toddler frequency) the right test for which phrases become tiles vs. which remain composable via the sentence engine?
Morphological marking: Section 06 describes how the sentence engine handles inflections (-ing, -s, -ed, possessive ’s) as post-tap modifications rather than separate vocabulary tiles. Does this approach align with how you would scaffold morphological development at Brown’s Stages II–IV (24–48 months)?
Sentence engine replacing SCS: The sentence engine replaces static speak-choose-speak templates with dynamic grammar-aware predictions. Section 06 provides the clinical rationale. Does this approach raise concerns about removing explicit SCS templates, or does the dynamic scaffolding adequately serve the same clinical goals?
Receptive vs. expressive: Section 05 distinguishes between companion-mediated receptive exposure and compass-edge expressive use. Does Meadow’s approach — companion models vocabulary receptively before the child accesses it expressively at the next tier — align with how you scaffold the receptive-to-expressive progression in AAC intervention?