Prompt Level Data

Prompt-Level Data

Prompt-Level Data

Paraprofessional Best Practice Library

Brief 06.03

Prompt-Level Data

The most important data type for fading and the path to independence

For paraprofessionals teaching skills and tracking growth

Why this brief

Of all the data types paras collect, prompt-level data is the most directly tied to teaching. It tells you whether the student is becoming more independent over time or whether they're still relying on the same level of help. Without prompt-level data, you can't fade prompts honestly; without fading, you don't get independence. Brief 04.02 covers prompting hierarchies; this brief covers how to record what level of help worked, how to interpret the trends, and what to do when fading isn't happening.

Prompt-level data is also where dishonesty creeps in most often. "Independent" gets marked when the para said the answer first. "Verbal direct" gets marked when there was actually hand-over-hand. The reasons are usually well-meaning — wanting the student to look like they're succeeding, wanting the program to look like it's working — but the cost is real: a student who isn't actually independent gets called independent and the help fades on paper while staying the same in reality.

| |

| :-: |

| The principleHonest prompt-level data is one of the most valuable things paras produce. Independence isn't a state you describe; it's a state you build. The data is the map of how that building is going. |

Who this brief is for

Paras teaching discrete skills (academic, communication, self-help, social, vocational)

Paras working in self-contained, ABA-influenced, or DTT-style settings

Paras working under BIPs that include skill-building components

Supervising teachers and BCBAs designing prompt-level data systems

Prompt hierarchies — quick recap

Brief 04.02 covers the prompt hierarchies in depth. The most-to-least continuum used in many programs:

| Level | What it means |

| :-: | :-: |

| Full Physical (FP) | Hand-over-hand; you do the skill with the student |

| Partial Physical (PP) | You guide the elbow, wrist, or part of the action; student finishes |

| Modeling (M) | You demonstrate; student copies |

| Gestural (G) | You point or signal without speaking |

| Verbal Direct (VD) | Specific spoken direction: 'Pick up the pencil' |

| Verbal Indirect (VI) | Open question: 'What do you do next?' |

| Independent (I) | Student does it without any help |

Other systems use slightly different scales — least-to-most, time delay, graduated guidance, errorless learning. Whatever the program uses, the principle of recording the actual level of help is the same.

Recording prompt-level data

Per trial recording

Most prompt-level systems record one prompt level per opportunity ("trial"). For a multi-step skill, you record per step, per trial. The data sheet is essentially a grid:

| Trial \# | Step | Prompt level used |

| :-: | :-: | :-: |

| 1 | Get toothbrush | PP |

| 1 | Apply paste | FP |

| 1 | Brush front teeth | M |

| 1 | Brush molars | FP |

| 1 | Rinse | PP |

| 2 | Get toothbrush | G |

| 2 | Apply paste | PP |

| 2 | Brush front teeth | M |

| 2 | Brush molars | PP |

| 2 | Rinse | VD |

This kind of record lets the team see exactly which steps are being mastered (toothbrush, getting it down to G or I) and which are still requiring physical support (molars, rinsing).

Frequency of recording

Every trial during teaching — the standard approach

Probe data — periodic, less frequent (once a week, monthly) — checks generalization

Both — daily teaching trials with periodic probe data

Where to record

Paper data sheets — most common; reliable, visible, easy

Apps and digital tools — Catalyst, BehaviorSnap, others; can be more efficient at scale

Whiteboards or quick-jot systems for in-the-moment, transferred later

When to record

In real time, immediately after each trial — period

Not from memory at end of session — accuracy collapses

Note context if anything unusual happened (substitute, illness, missed sleep)

Honesty about the dose of help

This is the central integrity issue with prompt-level data. The temptation is to record the level you intended, not the level you actually delivered. The data only works if it captures what really happened.

Specifically watch for

Recording "independent" when you said the answer first

If you said "What's the next step?" and the student answered, that's verbal indirect. If you said "What's the next step? Toothpaste?" — you said the answer; that's verbal direct, not independent.

Recording the planned prompt, not the actual

Plan was modeling. Student didn't get it. You ended up doing partial physical. Record partial physical.

Skipping prompts not planned for

"He needs hand-over-hand for this step but the data sheet only goes down to verbal direct." Then the data sheet is wrong; either record FP separately or modify the sheet. Don't artificially mark a higher independence than was achieved.

Counting cooperation as skill

If a student does the step because they want to please you, but doesn't actually know the skill, that's not independent — it's compliance. The skill needs to generalize beyond the specific person.

Subtle prompts you don't notice

Eye gaze toward the right answer. Slight head nods. Positioning materials so the answer is obvious. These are all prompts. "Independent" requires no help, including unconscious cues from you.

| |

| :-: |

| If in doubt, prompt one level higherWhen you're not sure whether a response was independent or whether you nudged it, record the more-helpful level. Better to under-claim independence than over-claim it. The data will be more conservative but more honest. |

Why honesty matters

Wrong-direction fading: if the data says independent, the next session drops the prompt — but the student still needed it, so they fail

Wrong celebration: family is told the student is independent on the skill, then surprised when they can't do it at home

Wrong IEP decisions: goals get marked met that haven't been met

Wrong programmatic next steps: the team moves to a new skill before the current one is solid

Self-deception: paras start to believe their own data, which insulates from feedback that things aren't going well

Reading prompt-level trends

Healthy fading

Over weeks, prompt levels should descend (less help needed) on most skills being taught. Visual summary tools that paras and teachers can use:

Bar graphs showing percentage of trials at each prompt level over time

Line graphs showing the modal (most common) prompt level by week

Step-by-step grids showing which steps have moved down which levels

Stagnation

Same prompt level for weeks with no descent. This is data signaling that the team needs to revisit the program:

Is the skill at the wrong level (too hard)?

Is the prompt fading approach being used (most-to-least requires deliberate dropping)?

Is reinforcement strong enough?

Are practice opportunities frequent enough?

Is there a sensory or motor barrier no one's identified?

Regression

Prompt levels going up — student needing more help than before — is real and usually not a problem with the student. Possible causes:

Illness, fatigue, family stress

New staff or substitute

Setting changes

New step or material introduced

Holidays or breaks reducing practice

Skill not yet generalized — works in one place, not in another

Mastered → independent

Most programs define "mastered" with specific criteria — typically independent across multiple consecutive trials, multiple sessions, multiple settings. Common criteria:

80% independent across 3 consecutive sessions

90% independent across 5 trials

Generalized across at least 2 staff and 2 settings

Mastery criteria should be in the program description; if not, ask the supervising teacher or BCBA.

Generalization data

A skill that works only in the specific context where it was taught isn't really mastered. Generalization data tracks whether the skill shows up:

With different staff (you, the other para, the gen-ed teacher, family)

In different settings (classroom, lunch, bathroom, home)

With different materials (different pencils, different toothbrushes)

With different verbal cues ("Brush your teeth" vs. "Time for hygiene")

After breaks (does it return after a weekend? a holiday?)

Recording generalization

Probes during the teaching phase — once a week, run a generalization probe in a new context

Maintenance checks after mastery — re-test periodically

Cross-staff and cross-setting fidelity checks

Why generalization matters

If a student can wash hands only with you, only at the same sink, only after the same prompt, the skill isn't useful. Real independence means the skill survives the variability of normal life.

Fading prompts deliberately

Brief 04.02 (prompting hierarchies) and 04.03 (planned, prompt fading) cover the techniques in depth. The data side:

Most-to-least with planned drops

Most common approach. Start at the level that ensures success (often FP), then drop the prompt level deliberately when the student is succeeding at the current level for several trials in a row. Data drives the timing of the drop.

Time delay

Wait progressively longer before delivering the prompt to give the student room to respond independently. Data captures the wait time at which the student responded.

Errorless learning

Prevent errors by prompting at a high level initially, then fading. Data shows the prompt level used and whether errors occurred.

Common fading mistakes captured by data

Fading too fast — student fails, prompt level has to go back up

Fading too slow — student becomes prompt-dependent and the data plateaus

Inconsistent fading across staff — student gets different prompt levels from different paras

Skipping levels — going from FP to G in one session usually doesn't work

Working with the team

Prompt-level data is generally designed by the supervising teacher or BCBA, collected by paras, and analyzed by the team.

Para's role

Collect honestly and consistently

Bring the data to weekly check-ins

Note unusual context (illness, substitute, etc.)

Flag concerns about the data system or what it's showing

Team's role

Design data system that fits the skill and student

Set mastery and fading criteria

Review data regularly — at least weekly during active teaching

Adjust program based on data, not impressions

Calibrate paras when inter-observer agreement drifts

Common conversations

"The data shows we should fade — should we?"

"The data is flat — what should we change?"

"The student is doing well in trial but not generalizing — what's our generalization plan?"

"My data and the other para's data don't match — let's calibrate"

Considerations for ELL students

When teaching skills to multilingual learners, prompt-level data has specific considerations:

Is the verbal prompt in the language the student understands? A verbal direct in English to a student at WIDA Level 1 may not function as a verbal prompt at all

Is the visual or modeled prompt clear enough across language? Visual prompts are often more accessible across languages

Is the skill being taught in the right context for generalization across both school and home languages?

Distinguish between "can't do the skill" and "didn't understand the language of the prompt"

Brief 08.06 covers WIDA proficiency levels; coordinate with the EL coordinator and SLP when designing prompt-level systems for multilingual learners.

Prompt levels in IEPs

Prompt-level data often appears directly in IEP goals as the measure:

"Marcus will brush his teeth with no more than gestural prompts on 4 of 5 trials across 3 consecutive sessions"

"Maria will identify community signs with verbal indirect prompts at 80% accuracy across 3 sessions"

"Aaron will request preferred items using independent picture exchange on 8 of 10 opportunities"

This makes prompt-level data tied directly to legal IEP progress reporting. The honesty stakes are higher — a goal marked met based on inflated independence claims is a real problem at IEP review.

Pitfalls

| Try this | Watch out for |

| :-: | :-: |

| Record the prompt level that actually delivered the response | Record the level you intended or hoped for |

| When in doubt, prompt one level higher (be conservative) | Round up to make the student look more independent |

| Record per trial in real time, not from memory | Reconstruct the data at end of session |

| Notice subtle prompts you might be giving (gaze, positioning) | Assume 'no help' just because you didn't say or touch anything |

| Define mastery and fading criteria in the program | Operate without explicit criteria for moving |

| Probe generalization across staff, settings, materials | Stop at fluency in one context |

| Bring stagnation to the team — something needs to change | Continue forever at the same level hoping it'll click |

| Calibrate inter-observer agreement when paras' data diverges | Each para go their own way and produce inconsistent data |

| Use the data to drive program decisions | Use impressions and 'feel' instead of data |

| Tie prompt-level data to IEP goals when appropriate | Treat IEP and teaching data as separate worlds |

Scenarios

Scenario 1: Discovering you've been over-claiming independence

Reviewing your data, you realize you've been marking 'independent' on a step where you usually point at the materials. You point so quickly and quietly that you didn't think of it as a prompt.

Honest correction. From now on, that's gestural, not independent. Bring it to the supervising teacher: "I want to flag that I've been recording these trials as independent but I think I was actually giving a quick gestural prompt. Going forward I'll record G; we may want to look at the past data with that lens." Backdated correction is hard and usually impractical, but going-forward correction is essential.

Scenario 2: A skill that won't fade

Your student has been at modeling for tooth-brushing for six weeks. The prompt level isn't dropping.

Bring the data to the supervising teacher and BCBA. Possible interventions: try a different fading strategy (time delay), break the skill into smaller steps and target the stuck ones specifically, check whether reinforcement is still potent, check if there's a motor or sensory barrier (the OT may help). Six weeks at the same level is a signal — don't let it go to twelve.

Scenario 3: Inter-observer disagreement

Your data shows the student needing FP for the rinse step. The other para's data shows G is enough. You can't both be right.

Calibrate. Sit with the BCBA or supervising teacher. Watch each other run the trial. Discuss what counts as a gestural prompt. Often the disagreement comes from one of you giving a quick subtle prompt the other doesn't notice. Take a few sessions of double-data — both observe, then compare. Continue until agreement reaches 80%+. This is especially important when the data is feeding IEP goals.

Scenario 4: Student succeeds with you but not the gen-ed teacher

Your student handles the morning routine independently with you. With the gen-ed teacher, he's back to needing partial physical.

This is a generalization issue. The skill works with you but hasn't generalized. Generalization plan: have the gen-ed teacher run trials too; teach explicitly across staff; identify what's different (your relationship, your prompting style, the setting). Brief 04.08 (planned) covers generalization in depth.

Scenario 5: A regression after a break

After winter break, your student needs more prompts than before the break.

Document the regression. It's expected after long breaks. Run shorter teaching sessions, expect to step back up briefly to higher prompt levels, and resume fading as the student re-engages. This is normal and usually resolves quickly. Bring it to the team if it doesn't recover within a couple of weeks.

Scenario 6: Data sheet that doesn't fit the skill

The data sheet has FP, PP, M, G, VD, I as the levels. But your student responds best to a different prompt — visual modeling using a picture sequence — that's not on the sheet.

Bring it up. The data sheet should reflect the prompts actually being used. Either add the prompt to the sheet, modify the program, or use a different data system. Squeezing real teaching into a sheet that doesn't represent it produces misleading data.

Closing thought

Prompt-level data is the engine of independence. It tells you whether the student is moving forward or stuck. It catches your unconscious helping. It feeds the IEP goal reports. It guides instructional decisions. Done well, it makes you a thoughtful teacher — constantly calibrating help to what the student needs. Done sloppily or dishonestly, it can make a stalled program look like it's working until everyone discovers, painfully, that it isn't.

The skill of prompt-level data is partly technical (how to record, how to define) and partly disciplinary (the willingness to be honest about your own help). The technical part is teachable; the disciplinary part is where excellent paras stand out.

| |

| :-: |

| Bottom lineRecord the prompt level that actually delivered the response. Real time, every trial, per step. Be honest about subtle prompts. Be conservative when uncertain. Watch trends — fading is the goal. Stagnation means change something. Probe generalization. Calibrate with other paras. Tie to IEP goals when appropriate. Bring concerns to the team. |

Related briefs

04.02 Prompting Hierarchies

04.03 Prompt Fading (planned)

04.07 Promoting Independence

04.08 Generalization and Maintenance (planned)

06.01 Data Types Overview

12.06 Working with the BCBA

13.01 FERPA and Confidentiality

Page of