Skip to content

Aided language modeling for AAC users

By the IEP Atlas editorial teamLast updated

Short answer

Aided language modeling means the adult talks to an AAC user while touching the matching symbols on the student's system, so the student sees the language used before being asked to use it. Pick 3 to 5 target words per activity, model one word longer than the student's messages, and never demand a copy. Track how often adults model.

A child learning to talk hears spoken language all day, for months, before saying much of their own. A child learning to use AAC often sees almost nobody use their system. Aided language modeling closes that gap. The adult says "you go" and touches "you" and "go" on the device at the same moment, in a real activity, without asking the student to copy.

It sounds simple and it is, but it takes practice for adults to do it often enough to matter. This guide is for the SLP training classroom staff, and for the teacher or para doing the modeling. For goal wording, see the AAC goals examples. The speech therapy IEP goals pillar covers the rest of speech and language.

What you need before you start

  • The student's system, within reach of both of you. If sharing is awkward, use a second device or a printed copy of the same layout. The symbols must be in the same places.
  • A short list of target words for each routine. Mostly core vocabulary: go, stop, more, help, like, not, you, my turn.
  • One staff member who has seen it done. A 5-minute demo beats a handout.
  • A way to count models. A tally on a sticky note is enough.

AAC has no prerequisites, and partners play a large part in whether a student learns to use it.[1]

Step-by-step aided language modeling

Step 1: Choose 3 to 5 target words per activity

Look at the routine and ask what words fit it many times. At snack: more, open, help, all done, yuck. In a book: look, uh oh, turn, funny, again. Write the list on a card taped near the activity. Keep most of them core words, because core words travel across the day.

Step 2: Talk and touch at the same time

Say the word as you touch the symbol. "Uh oh! It fell." Touch "uh oh" and "fell." The spoken word and the symbol should land together. Keep talking at a normal pace for the rest of the sentence. You do not have to touch every word.

Step 3: Model one step above the student

If the student uses single symbols, model 2-symbol phrases ("more bubbles," "you go"). If they combine 2, model 3. This mirrors how adults talk to young speaking children: a little longer than the child, not a full lecture.

Step 4: Model without demanding a response

Do not follow a model with "Now you say it" or "Show me more." That turns a model into a prompt, and a student who feels tested often stops looking. Model, pause and wait. If the student communicates, respond to the meaning. If not, keep playing.

Step 5: Model different reasons to communicate

Requesting is the easiest to model, so it takes over. Spread models across reasons: commenting ("big tower"), protesting ("not that"), asking ("where go?"), greeting ("hi") and telling feelings ("mad"). The requesting and commenting goals page shows why the balance matters.

Step 6: Respond to what the student says, not how

When the student taps "go," treat it as real talk. Go. If they tap something unexpected, respond to it as if it were meant: "Oh, you said 'eat.' Are you hungry?" Then model what you think they meant if it differs. Accidental hits are part of learning where words are.

Step 7: Count models and student messages

Tally adult models per activity and the student's messages, coded spontaneous or prompted. The AAC data collection page sets out the codes. Without a model count, the team cannot tell whether flat data means the student is stuck or the adults stopped modeling.

Worked example: one student over 8 weeks

Jonah is a fictional first grader with a speech-generating device. He uses single symbols, mostly "more" and "go."

Baseline (week 1). The SLP watches three 20-minute blocks of centers. Adults model a median of 2 times per block. Jonah sends a median of 4 messages: 3 prompted, 1 spontaneous, all single symbols.

Plan. The SLP trains the teacher and para in two 15-minute sessions. Each center gets a card with 4 target words. The team sets its own target of at least 15 adult models per 20-minute block, tallied on a sticky note. The IEP lists staff training as a support for school personnel.[2]

Week 4 check. Adults model a median of 12 times per block. Jonah sends 7 messages, 3 of them spontaneous, and 1 combination ("more car").

Week 8 check. Adults model a median of 17 times per block. Jonah sends 11 messages, 5 spontaneous and 3 two-symbol combinations.

WeekAdult models per 20 minJonah's messagesSpontaneous2-symbol combinations
12410
412731
8171153

These numbers are illustrations. The point is the pairing: the team can see Jonah's growth next to the adults' modeling, and can show the IEP team both. A goal for Jonah might now target 2-symbol combinations, with modeling in the condition. The responding to aided language modeling and expanding a partner's model templates fit this stage.

Common mistakes

Weak: Holding the device away from the student "so they don't get distracted." Better: The device stays with the student all day, and you model on it or on a matching copy.

Weak: "Say 'more.' Show me 'more.' Touch 'more.'" Better: "More bubbles!" while touching "more" and "bubbles," then wait.

Weak: Modeling only nouns on a fringe page (cookie, ball, juice). Better: Model core words that work everywhere (more, go, help, not) plus a few activity words.

Weak: Modeling only during speech sessions, twice a week. Better: Model in every routine the student is in, by every adult, every day.

Weak: Modeling full sentences for a student who uses single symbols. Better: Model one step above: 2 symbols for a single-symbol user.

Weak: Stopping when the student does not respond after a week. Better: Keep going and keep counting. Spoken language takes many exposures, and so does AAC.

Training staff to model

Paraprofessionals often spend the most time with the student, so they need the training most. A short training works better than a long one: demo for 5 minutes, have the para model while you watch for 5 minutes, then give one specific piece of feedback. Check back in a week. The IEP can list this training, because assistive technology services include training for professionals and others who work with the student.[3]

The paraprofessional glossary entry and the paraprofessional responsibilities guide cover the para's role more broadly.

Templates and tools

Frequently asked questions

Is aided language modeling the same as prompting?

No. A model shows the language without asking for a response. A prompt asks the student to respond, such as saying show me more. Mixing the two turns modeling into testing, which many students tune out.

How many times should adults model on the device?

There is no single required number. Many teams set their own target, such as a model every minute or two in a key routine, and count it. The useful step is counting at all, so you can see whether modeling is happening.

Can I model on a paper copy instead of the student's device?

Yes, if the paper copy has the symbols in the same places as the device. The student learns the location of each word, so a different layout teaches the wrong path.

Use it in IEP Atlas

Sources

  1. Augmentative and Alternative Communication (AAC), ASHA
  2. 34 CFR §300.320: Definition of individualized education program, Legal Information Institute, Cornell Law School
  3. 34 CFR §300.6: Assistive technology service, Legal Information Institute, Cornell Law School

Written and checked by the IEP Atlas editorial team under our editorial policy. This guide describes general practice and federal rules. It is not legal advice; your state may add requirements, and the IEP team makes the decisions for each student.

IEP Atlas Pro

Turn your goals into progress reports

Pro tracks every goal on your caseload and drafts the report for you. Free for 14 days.

Count trials, prompts, and minutes live in the session; the graph and trend line draw themselves; progress reports are drafted from your data.