Prompting Hierarchies

How to give a prompt without taking over — the most common move in special education, done well

Why this brief

Prompting is the most frequent thing paraprofessionals do all day. It is also the place where the difference between great practice and well-intentioned mediocre practice is largest. A skilled prompter can produce student independence in months that an unskilled prompter would still be working on years later. The difference is not effort — it's pattern.

This brief covers the main prompting hierarchies in the field, when each is the right choice, how to fade prompts so the student becomes independent, and the small set of mistakes that account for most of the bad outcomes.

| |

| :-: |

| If you read one thingPlan the fade from day one. The single biggest failure mode in 1:1 instruction is delivering high-quality prompting that never tapers — the student gets very good at responding to prompts and no better at responding to the natural cue. The prompt is supposed to disappear. If the data don't show the disappearance, something is wrong. |

1\. What's a prompt?

A prompt is any added stimulus — beyond the natural cue — that helps the student produce the correct response. The teacher asks "What letter is this?" The natural cue is the printed letter B. A prompt might be: pointing at the letter (gesture), saying the first sound (verbal partial), modeling the answer (verbal model), or guiding the student's hand to point at it (physical).

Prompts vary on two main dimensions:

Intrusiveness — how much the prompt does for the student. Verbal cue is less intrusive; full physical guidance is more.

Timing — when in the chain the prompt arrives. Antecedent prompts (before the student responds) prevent errors. Consequence prompts (after a non-response or error) correct errors.

2\. Levels of intrusiveness

From least to most intrusive — usually the order learners encounter them in a least-to-most procedure.

| Level | What it looks like |

| :-: | :-: |

| Natural cue (no prompt) | The teacher asks the question or sets up the task. Student responds (or doesn't) on their own. The goal is for this to be sufficient. |

| Indirect verbal prompt | A general nudge that doesn't reveal the answer. "What do you think?" "Take a look at this part." |

| Direct verbal prompt | Specific words that point toward the answer. "It starts with the /buh/ sound." "What letter does buh start with?" |

| Gestural prompt | Pointing, tapping, looking at the right answer. |

| Visual prompt | A picture, written word, color, or arrow that highlights the answer. (Visual prompts are often built into the materials and can be hard to fade.) |

| Modeling prompt | Demonstrating the response. "Watch — that's a B." "My turn first, then your turn." |

| Partial physical prompt | Touching to guide. "Tap the elbow to start the motion." Light touch on the back of the hand. |

| Full physical prompt (hand-over-hand) | The adult physically moves the student through the response. The most intrusive level. |

| |

| :-: |

| "Verbal is least intrusive" is a generalization, not a lawFor some students, verbal prompts are highly intrusive — a noise to filter, a demand to escape. For non-speaking students, verbal prompts may have minimal effect. For some autistic students, gesture and visual prompts are easier to follow than language. The hierarchy above is a useful default; calibrate to the student in front of you. |

3\. The four main hierarchies

3.1 Least-to-Most (LTM)

Start with the least intrusive level (typically the natural cue alone) and escalate as needed. Wait the predetermined response interval (commonly 3–5 seconds) at each level before moving up.

Procedure

Present the natural cue. Wait the response interval (e.g., 5 seconds).

If correct response — reinforce.

If no response or incorrect response — deliver next-level prompt. Wait the response interval again.

Repeat through the hierarchy until the student responds correctly.

Reinforce the correct response, regardless of which level produced it.

When to use LTM

The student has begun to acquire the skill — they can do it sometimes.

You're working toward independence and want to test that level first.

Errors are not dangerous and not heavily reinforced.

When NOT to use LTM

Brand-new skill, where errors are likely and unhelpful (you'll teach error patterns).

Skills where errors are dangerous (knife handling, street crossing, swimming).

Students for whom errors trigger escalation.

3.2 Most-to-Least (MTL)

Start with the prompt level that guarantees a correct response (often a model or full physical prompt) and systematically fade to less intrusive levels over trials or sessions.

Procedure

Begin with the highest necessary prompt level. Several trials at this level.

Once the student is responding correctly with high accuracy, fade to the next-less-intrusive level.

Continue fading systematically until the student responds to the natural cue alone.

Pace the fade with data — if accuracy drops, return to the prior level briefly.

When to use MTL

Brand-new skill, especially one with a clear correct sequence.

Errorless learning is the goal (no errors during acquisition).

Discrete-trial-style instruction.

Safety-critical skills.

3.3 Time delay

Time delay teaches by inserting a controlled wait between the natural cue and the prompt. Two main variants:

Constant time delay

Trial 1: 0-second delay. Natural cue + immediate prompt.

Subsequent trials: fixed delay (e.g., 4 seconds). Natural cue, wait 4 seconds, then prompt if no response.

Goal: the student begins responding before the prompt arrives.

Progressive time delay

Begin at 0 seconds; increase delay gradually (e.g., 2, 4, 6 seconds across sessions).

Especially useful when the student needs longer processing time as the skill is acquired.

Time delay is one of the best-evidenced fading procedures and is often used as a stand-alone hierarchy. The student learns to anticipate the prompt and produce the response before it's needed.

3.4 Graduated guidance

A physical-prompt fade method specifically. The adult begins with whatever level of physical support is needed (often hand-over-hand) and fades the support in real time as the student takes over the movement.

The fade sequence in practice

Hand-over-hand: full physical support.

Hand-on-wrist or forearm: less direct, still guiding.

Hand-on-elbow or shoulder: gentle redirect.

Shadowing (no contact, hand near): "safety net."

No physical support.

Graduated guidance reads the student moment-to-moment — when the student initiates the next part of the movement, the adult removes contact. The fade is fluid, not stepped, and often happens within a single instruction.

4\. Errorless learning

Errorless learning is a teaching philosophy that minimizes student errors during acquisition by providing whatever prompt level is necessary to ensure correct responses, then fading those prompts. Most-to-least and time delay are the typical hierarchies that implement it.

Why errorless learning

Errors get reinforced (even briefly) and become habitual. Repeated errors during acquisition are surprisingly hard to extinguish later.

Errors lower student confidence, especially for students with anxiety or learned helplessness.

Errors during acquisition slow learning compared to errorless procedures, in the research literature.

Trade-offs

Errorless learning is more adult-intensive. The student gets less practice generating the response from scratch and may show weaker generalization to novel exemplars unless the program is designed to address that. Most contemporary teams use errorless during acquisition and shift to more student-led approaches (LTM with planned errors) once the skill is established.

5\. Error correction

When errors do happen — and they do — the response procedure matters. The standard pattern (sometimes called the "rule of three" or "model-lead-test"):

Stop the trial. Don't reinforce the error. Don't dwell on it.

Model the correct response.

Lead the student through the correct response (prompted).

Test — present the same item again with no prompt or minimal prompt. Reinforce the correct response.

The point is that the student leaves the trial having performed the correct response, with reinforcement, even though they got it wrong the first time. Errors don't end the learning sequence; they restart it.

| |

| :-: |

| Avoid common error-correction mistakesDon't repeat the question the student got wrong as if hoping for a different answer. Don't lecture. Don't go through five different prompts before reaching the right one. Don't show frustration. Three-step model-lead-test, reinforced, move on. |

6\. Prompt fading — making prompts disappear

The whole point of prompting is for prompts to go away. A student who needs the same prompt level forever has not actually learned the skill — they've learned the prompted version. Fading is the engineering problem of moving from prompted responding to independent responding without losing accuracy.

6.1 Plan the fade in advance

Before the first instructional session, the team should know:

What "independent" looks like for this skill (the natural cue alone produces the response).

What hierarchy will be used.

What data will be collected (prompt level used per trial).

What the decision rule is for fading (e.g., "once 80% accurate at this level for two consecutive sessions, drop to the next level").

What the decision rule is for backing up if accuracy drops.

6.2 Track prompt level data

Right/wrong data is not enough. You need to know which prompt level produced the correct response. "He got 8 of 10 right" is much less informative than "He got 4 with no prompt, 3 with verbal, 2 with model, 1 needed full physical, 0 incorrect." The fade lives in the prompt-level data.

6.3 Move the prompt earlier in the chain

Antecedent prompts (delivered before the student responds) often fade faster than consequence prompts (delivered after a non-response). Front-loading the prompt — "Before you start, remember it's the green button" — reduces errors and lets you fade the prompt by gradually moving it earlier (until it's part of the natural cue) and then removing it.

6.4 Generalization-aware fading

Fade across:

People — different adults running the program.

Materials — different exemplars (different printed B's, not just one).

Settings — different rooms, different times of day.

Cues — slight variations in how the natural cue is presented.

If the same para always runs the program with the same cards in the same room, the student learns to respond to that combination. Build variation in deliberately.

7\. Decision tree: which hierarchy when?

| If… | Then likely… |

| :-: | :-: |

| The student is acquiring a brand-new skill and errors are problematic. | Most-to-least (errorless) or time delay. |

| The student has begun to acquire the skill and you're testing for independence. | Least-to-most. |

| The skill has a physical motor component (handwashing, dressing, fine motor). | Graduated guidance. |

| The student needs longer processing time and pre-emptive prompts trigger escape. | Time delay (progressive). |

| The skill is safety-critical and errors are dangerous. | Most-to-least with errorless approach. |

| The student has a strong response history of dependency on prompts. | Time delay or most-to-least with structured fade plan. |

| The student has high anxiety and errors trigger escalation. | Errorless approach (most-to-least, time delay). |

8\. Common pitfalls

Two paras using two different hierarchies for the same skill. The student gets confused and the data are uninterpretable. The supervising teacher picks one and trains everyone.

No fade plan. The team starts strong, the student responds to prompts, and the prompts never come down. This is the most common failure mode.

Reinforcing the wrong response. Praise after a model-prompt is fine; praise after a full physical prompt should be muted (the student barely participated).

Verbal prompt drift. The para keeps adding small verbal prompts that aren't documented ("What's the next step?" "Don't forget the…"). Generalization fails.

Pre-loading the answer. "What letter is this? It's a B." — student never has a chance to respond. No learning happens.

Inconsistent wait times. Wait times need to be the same across staff and trials. Some students need 8 seconds; some need 3. Pick one and use it.

Physical guidance that lasts forever. Hand-over-hand should be a teaching tool, not a daily occurrence on a known skill.

Skipping the data. Without prompt-level data, the team can't see whether the fade is happening.

Treating prompts as helping. Helping the student finish a worksheet ≠ teaching the student. Fading is what teaches.

9\. Resources

AFIRM — Prompting and Time Delay modules — afirm.fpg.unc.edu — Free, self-paced, with implementation checklists and fidelity tools.

VCU Autism Center — Using Prompts to Promote Skill Acquisition — vcuautismcenter.org — Practical fact sheet.

IRIS Center — High-Quality Mathematics Instruction (includes prompting examples) — iris.peabody.vanderbilt.edu — Free modules.

Cooper, Heron & Heward — Applied Behavior Analysis — Pearson — Chapter on prompting and transfer of stimulus control is foundational.

Brief 04.03 — Prompt Fading — this library — Cross-reference.

Brief 04.07 — Promoting Independence — this library — Why fading matters at the system level.

Brief 06.03 — Prompt-Level Data — this library — How to track the fade.

Page of