Selecting and Implementing Interventions

How interventions are chosen and run, one task at a time: goals, procedures, side effects, relapse and the data behind each decision.

Concepts
8
Sections
5
Diagrams
6
In this chapter
  1. 01 From assessment to goals
  2. 02 Choosing what to do
  3. 03 Side effects and relapse
  4. 04 Deciding with data
  5. 05 Working with others

From assessment to goals

An intervention begins on paper. Before anyone teaches or reduces anything, the plan has to say what the person will do, and how everyone will know when it has happened.

Figure 1

From an assessment finding to a measurable goal

Assessment finding

At lunch, Mia waits for an adult to open every container and often runs out of time to eat.

Not yet a goal

“Mia will be more independent at lunch.”

Measurable goal

  1. Condition

    When, where, with what help

    When handed her lunchbox at the cafeteria table

  2. Behavior

    The learner’s own action

    Mia opens all the lunchbox containers without help

  3. Criterion

    How well, how often, for how long

    within 2 minutes of receiving the lunchbox, on 4 of 5 school days, for 2 consecutive weeks

Check: would two people watching the same lunch agree on whether the goal was met?
The assessment finding says why a goal is needed; the goal says exactly what will count as success. A goal is complete when it names the condition, the learner’s own action and a criterion two observers could score the same way.
H.1

Writing observable, measurable intervention goals

A goal turns an assessment finding into a target that anyone can check. Observable means another person could see or hear the action; measurable means it can be counted or timed along a dimension such as frequency, latency, duration or accuracy. A complete goal has three parts: the condition (when, where, with which materials or level of help), the behavior (an active verb for something the learner does) and the criterion (the level that counts as success, usually held across several sessions or settings). The goals worth writing are the ones the assessment showed matter most for this person’s daily life.

Example

Assessment shows that Mia, age six, waits for an adult to open every container at lunch and often runs out of time to eat. “Mia will be more independent at lunch” is a wish. “When handed her lunchbox at the cafeteria table, Mia opens all its containers without help within 2 minutes of receiving it, on 4 of 5 school days for 2 consecutive weeks” is a goal: a lunch aide and a parent watching the same day would score it the same way.

Common confusion

Use observable actions rather than an unmeasured label such as feeling calm or understanding sharing. A reduction goal may validly specify fewer or no target responses during defined observation periods, but it should be paired with a goal for a useful active replacement. The dead-man heuristic helps identify descriptions of absence; it does not invalidate measurement of behavior reduction. Separate what the learner will do from implementation goals for staff. Match the measurement dimension to the need: difficulty starting work calls for a latency measure, not accuracy alone.

Key points (4)
  • A goal names something the learner will do that others can see or hear, not an internal state, trait, attitude or label.
  • A measurable goal states the condition, the behavior and the criterion, so two observers scoring the same session would agree on whether it was met.
  • A goal describes the learner's own action, not only the absence of a behavior and not what staff or caregivers will do.
  • The criterion uses the dimension and the settings that match the problem (for example latency, duration, accuracy, or performance in the places where the skill is needed).
Examples and a practice question: H.1 Writing observable, measurable intervention goals Practice this: H.1 Writing observable, measurable intervention goals

Choosing what to do

Intervention selection integrates assessment, evidence, client values and practical fit. A reduction plan also identifies a useful alternative response and how the person will obtain appropriate reinforcement.

Figure 2

Choosing an intervention: five questions to integrate

Assessment and goals

Which assessed needs and behavioral variables does the plan address?

Options matched to relevant skills, context and supported functions

Evidence

What has research support for similar people, behaviors and settings?

Supported procedures, not popular ones

Preferences and values

What does the client assent to, and what does the family accept?

Options the people involved can live with

Contextual fit

Can these people, in this setting, run it as designed?

Options that match skills, time and materials

Least restrictive

Which remaining option is least restrictive and still likely to work?

The first recommendation

Recommended plan, with how integrity and progress will be measured
Read these questions together and revisit them as information changes. Assessment includes skill needs and, when relevant, maintaining consequences. Client values, assent, safety and feasibility matter throughout; the numbered order is a visual aid, not a required decision sequence.

Figure 3

Can the replacement compete?

Problem behavior

Banging on the counter

Replacement

Handing over a coffee picture card

Outcome it produces At least as good
Coffee gets made Coffee gets made
Effort At least as good
One quick movement One quick movement
Speed and reliability of payoff At least as good
Within a minute, almost every time Within a minute, every time
Acceptable in daily life Better
Startles others, disrupts the room Understood at home and at work
This example favors the replacement on the listed dimensions; actual acquisition, use and outcomes still need measurement.
Compare the outcome, response effort, delay, reliability and everyday acceptability. A readily available response that obtains the maintaining reinforcer is a strong candidate. These dimensions can trade off; an advantage on every row is not required and does not guarantee success. Teach the response and measure whether it competes in daily settings.
H.2

Choosing interventions from assessment, evidence, preference and context

Begin with the assessed needs, the evidence about relevant behavioral and contextual variables, and the goals the client values. For behavior reduction, evaluate maintaining consequences and select procedures that address the supported function; for skill acquisition, also assess prerequisites and useful learning opportunities. Integrate scientific evidence, client preferences and assent, family priorities, cultural variables and practical resources throughout selection. Research with similar learners and contexts informs a recommendation but does not guarantee this outcome. Check whether implementers have the skills, time and materials required. Prefer the least intrusive effective option, document the rationale for any greater restriction, and specify how outcomes, acceptability, safety and integrity will be monitored.

Example

Two children both sweep materials off the table during seated work. A functional analysis shows that for one of them, sweeping ends the task; for the other, it reliably brings an adult over to chat. Same movement, two plans: the first child learns to ask for a short break while task demands are adjusted, and the second receives frequent attention while working and learns an acceptable way to call an adult.

Common confusion

Matching a procedure to the behavior’s form (“throwing calls for a point-loss system”) skips the function entirely. A procedure that succeeded in a well-staffed research study may still fail in a home with one tired parent and three other children; that is a problem of fit, not a reason to ignore evidence. And the fastest or most forceful option is not the default: restrictiveness has to be justified, never assumed.

Key points (5)
  • The function identified by assessment guides the choice: the intervention changes the reinforcement contingency that maintains the behavior rather than reacting to its form.
  • Scientific evidence means research support for similar clients, behaviors and conditions; popularity, novelty, testimonials and anecdotes are not evidence.
  • Among effective options, the client's preferences, assent and values (and those of caregivers) help decide which one to use.
  • Contextual fit asks whether the people and setting can carry out the intervention as designed: implementer expertise and training, cultural variables, time, materials and other resources.
  • The least restrictive and least intrusive procedure likely to be effective is recommended first; more restrictive procedures need a documented reason and risk review.
H.3

Selecting socially valid replacement behaviors

When a plan reduces a behavior, it also names what the person will do instead. The replacement should be functionally equivalent: it produces the same reinforcer the problem behavior has been producing, so the person keeps access to something that matters to them. Prefer a response the person can acquire and access easily, with prompt and reliable reinforcement; effort, delay, quality and amount of reinforcement can trade off, so test actual response allocation rather than requiring an advantage on every dimension. It should also be socially valid: acceptable to the person and to the people around them, and likely to be noticed and reinforced in ordinary settings, not only during sessions.

Example

An adult who uses few spoken words bangs on the kitchen counter when he wants coffee, and whoever is nearby usually makes some. A picture card he can hand to anyone in the kitchen, which gets coffee made just as quickly, gives him the same outcome with no extra effort. A long sentence he struggles to produce would lose to banging, which is faster.

Common confusion

An absence such as not hitting does not specify what to do to obtain a reinforcer. Sitting quietly or keeping hands down may be appropriate targets in a particular contingency, but do not assume they provide a functionally equivalent way to meet the need. Define the response, check that it is accessible and acceptable to the person, and verify what reinforcement it actually obtains. DRA can use other effective reinforcers, but functional equivalence is an important selection priority when replacing problem behavior.

Key points (4)
  • When a behavior is to be decreased, the alternative behavior should produce the same reinforcer that maintains it (functional equivalence).
  • A socially valid alternative is acceptable to the client and stakeholders and is likely to be reinforced by natural contingencies in the client's everyday settings.
  • To compete with the target behavior, the alternative should require no more response effort and pay off at least as quickly and reliably.
  • Quiet compliance or the mere absence of the target behavior is not a replacement; the alternative adds a useful skill to the client's repertoire.

Side effects and relapse

Every procedure that changes behavior can also change things nobody targeted, and a behavior that went down can come back. Both are predictable, so both belong in the plan before treatment starts.

Figure 4

Four possible relapse routes: check what changed

Reduced in treatment What changed
  1. Spontaneous recovery

    A break from extinction; the target contingency is not restored

  2. Resurgence

    Reinforcement conditions for the alternative worsen

  3. Renewal

    A context outside treatment: the original setting or another one

  4. Reinstatement

    A previously effective reinforcer is delivered independently of the target

Illustrative sketches, not recorded data. Assume prior reinforcement followed by reduction during extinction-based treatment. The dashed line marks a change preceding the return. Labels describe possible routes, not diagnoses from curve shape; multiple changes may occur together. Return size and duration vary.
H.4

Anticipating unwanted effects of reinforcement, extinction and punishment

When reinforcement for a behavior is withheld, responding may briefly rise, vary or intensify, and emotional behavior or aggression can appear. Punishment can be accompanied by escape or avoidance of associated people or places, emotional responding, aggression, or suppression restricted to conditions in which the punishing stimulus or event is expected. Reinforcement has its own risks: satiation, dependence on a single contrived reward, or more of a behavior than anyone wanted. A sound plan names these effects in advance and says what to do about them: prepare the people carrying it out, define a consistent response and a safety plan, reinforce an alternative behavior, and measure more than the target so that collateral effects show up in the data.

Example

A school stops letting a student leave class when he rips his worksheets. The plan warns staff that ripping may increase for a few days, gives them a short script and a safety procedure, teaches the student to hand over a break card, and adds two extra counts: leaving the room without permission and reports of outbursts at home, in case the problem shifts somewhere else.

Common confusion

An early increase during extinction is not, by itself, proof that the plan failed; it was foreseeable, and the plan should say how it will be judged. It is not something to shrug off either: if it involves harm, the safety plan applies. Tracking only the target behavior can hide side effects. And a reduction procedure used on its own, with no reinforcement for an alternative, leaves the person without a better way to get what they need.

Key points (4)
  • Extinction can produce a temporary increase (extinction burst), emotional behavior or aggression; the plan warns implementers, sets a consistent response and a safety plan, and reinforces an alternative.
  • Punishment can produce escape and avoidance, emotional responding, aggression and suppression limited to the punisher or setting; it is paired with reinforcement for an alternative and monitored.
  • Reinforcement procedures also carry risks, such as satiation, dependence on one contrived reinforcer and excessive or unwanted responding, which the plan anticipates.
  • Mitigation includes measuring behaviors beyond the target and responding to collateral effects as soon as the data show them.
H.5

Planning against relapse

After a previously reinforced response has decreased during extinction-based treatment, it may return without its reinforcement being restored. Spontaneous recovery follows a break from extinction. Resurgence follows worsening reinforcement conditions for a more recently reinforced alternative. Renewal follows a change from the treatment context, including a return to the original context or entry into another one. Reinstatement can follow response-independent exposure to a previously effective reinforcer. These are possible effects, and several changes can occur together. Plan gradual, data-based changes in alternative reinforcement, practice across relevant contexts, reliable access to useful communication, follow-up checks and a response to early deterioration.

Example

During homework with the same parent in the same room, crying no longer produces help, while an accessible help request does. Crying decreases. Later, the parent continues not to help following crying but often leaves appropriate requests unanswered, and crying returns. With those conditions documented, the pattern is consistent with resurgence. Training another available caregiver and arranging a workable request schedule could reduce the disruption, but the results must be checked.

Common confusion

Check the treatment history and what changed before naming a relapse route. Return outside the treatment context is consistent with renewal; return after alternative reinforcement worsens is consistent with resurgence. Return after the old reinforcer is delivered independently of the target is consistent with reinstatement. If the target itself again produces reinforcement, also consider reacquisition. Several routes may overlap, and the return pattern alone cannot identify which variable caused it.

Key points (4)
  • A reduced behavior can return through spontaneous recovery, resurgence (reinforcement for the alternative decreases), renewal (the context changes) or reinstatement (the former reinforcer is delivered independently of the behavior).
  • Reinforcement for the alternative behavior is thinned gradually and kept available, because losing it abruptly invites resurgence.
  • Treatment that includes the client's everyday settings and several implementers lowers the risk of renewal when the context changes.
  • Meeting the goal does not end relapse risk: caregivers learn to recognize and respond to an early return, and follow-up or booster checks are planned.

Deciding with data

Review implementation and client outcomes together. Check integrity when interpreting progress, and respond immediately to safety concerns rather than waiting for the integrity review to finish.

Figure 5

Measuring procedural integrity, step by step

Token system · one observed session

  1. 1 Materials ready before the student sits down
  2. 2 Instruction given once, as scripted
  3. 3 Waits 3 seconds before any prompt
  4. 4 Tokens given only after correct responses Commission: token also given after an error
  5. 5 Token delivered within 2 seconds
  6. 6 Praise names the correct response
  7. 7 Response recorded after each trial Omission: planned step left out
  8. 8 Exchange offered when the card is full
75%

Steps carried out as written: 6/8

An illustrative checklist from one observed session. Integrity is the share of steps carried out as written, and it counts both kinds of error: a step left out, and something added that the plan does not include.

Figure 6

Reading progress against an aim line

↑ Fields completed independently

Baseline Intervention Aim line
05101520 Goal: 20 Aim line

Sessions →

Check safety and client priorities now; review integrity when interpreting progress.
Acquisition follows the aim line and remains acceptable
Continue and monitor
Acquisition is flat below the goal or slower than planned
Review the plan, measures and goal
Harm or meaningful deterioration appears
Act promptly and review
An appropriate criterion is met
Plan support changes and check generalization and maintenance
Illustrative values for increasing independent completion, not a real case. The aim line joins the starting level to a planned goal date; it does not demonstrate a treatment effect. Here progress falls below the acquisition target, so the team reviews the intervention after checking integrity, safety, measurement and whether the goal remains appropriate.
H.6

Data-based decisions about procedural integrity

Procedural integrity, also called treatment integrity or fidelity, is how closely an intervention is carried out as written. Measure specified components with representative direct observations and, where valid, products or records that verify implementation; self-report can supplement these measures but should not be the sole evidence, and client progress does not substitute for integrity data. Inspect critical components separately because the same overall percentage can conceal very different errors. Errors come in two kinds: omission, a planned step left out, and commission, something added that the plan does not include, such as an extra prompt or a token given after an error. Integrity is measured repeatedly, for every implementer and every setting. When progress is poor, integrity is checked first, because a plan run at half its steps has not really been tested. Low integrity is fixed where the data place it: training and feedback for the specific steps and people involved, or a revised step if the original cannot be done in that setting.

Example

Weekly checks show a token system run at 92% of its steps in the morning group but at 55% in the afternoon, where tokens are often late and sometimes follow a wrong answer. The response is targeted: practice and feedback for the afternoon staff on token timing and on which responses earn a token, rather than a new intervention for every student.

Common confusion

Lack of progress does not show that an intervention is ineffective if it was not carried out as written. Integrity is not inferred from good client outcomes, and high integrity does not guarantee that an intervention works; it only means the plan was actually tested. Do not assume every integrity problem is a training problem: sometimes a step does not fit the setting, and the step is what needs changing.

Key points (4)
  • Procedural integrity is the degree to which an intervention is carried out as designed. Prioritize direct observation against defined steps; validated permanent products can document relevant steps. Self-report is an indirect supplement, and client outcomes alone do not measure integrity.
  • Before concluding that an intervention does not work, integrity is checked; poor outcomes with low integrity call first for better implementation (training, support or simplifying an impractical procedure), not for concluding that the intervention failed.
  • Low integrity is addressed where the data locate it: the specific steps and implementers involved, with targeted training and feedback, or by fixing a step that does not fit the setting.
  • Integrity is measured repeatedly, for every implementer and setting, and counts errors of commission (adding something not in the plan) as well as omission.
H.7

Data-based decisions about effectiveness and modification

Decisions to keep going, change course or move on come from graphed data read by visual analysis: level (where the data sit), trend (the direction they move) and variability (how much they bounce), compared with baseline and with an aim line that runs from the starting level to the goal by its target date. With adequate integrity, data tracking the aim line support continuing. Flat data, or growth too slow to reach the goal in time, support modifying the intervention. A worsening trend, above all one involving harm, calls for a prompt review rather than more waiting. High variability is a cue to look for what is driving it. When the criterion is met, the plan moves to its next stage: fading support, programming generalization and checking maintenance.

Example

A teenager’s goal is to complete 20 fields of a job application on his own by the end of the month. Two weeks in, his count has risen from 4 to 7, while the aim line called for about 12 by now. Integrity checks are high, so the team adds a brief model of each new field before practice instead of waiting until the month is over.

Common confusion

One session does not establish a trend, and an aim line is a planning aid rather than proof that treatment caused the change. Use repeated outcomes, relevant comparisons and integrity data. Even an improving intervention may need revision if it causes harm, imposes excessive burden or no longer fits client preferences. Stable performance at an acceptable goal can support maintenance; a flat acquisition curve below an appropriate goal calls for review. Meeting a criterion prompts a planned transition, with continued checks of generalization, maintenance and needed support.

Key points (5)
  • Decisions rest on visual analysis of level, trend and variability across sessions compared with baseline and the goal or aim line, not on a single data point.
  • With adequate integrity, steady progress toward the goal supports continuing; no progress, or progress too slow to reach the goal, supports modifying the intervention.
  • High variability is a signal to look for the variables behind it before changing the intervention.
  • When the criterion is met, the plan moves to the next step (fading, generalization, maintenance) instead of continuing unchanged or dropping the skill.
  • A worsening trend, especially one involving harm, calls for prompt review and modification rather than more waiting.

Working with others

A behavior plan rarely runs alone. Families, teachers and other professionals touch the same person every day, and the plan works best when they pull in the same direction.

H.8

Collaborating with others to support client services

Most clients are served by several people at once: caregivers, teachers, speech-language and occupational therapy providers, physicians and others. Behavior analysts coordinate with them, with consent to share information, so that plans are consistent and the client’s interests come first. Caregivers and other stakeholders help choose goals and are trained to carry out the plan across settings. Disagreements are worked through with shared data, real attention to what others have observed and a joint plan, without dismissing other disciplines and without simply giving way when the client’s data point elsewhere. Questions outside behavior analysis, such as medical, nutritional or medication issues, are referred to, or coordinated with, the professional who holds that expertise.

Example

A girl’s bedtime plan stops working, and her parents mention that she recently started an allergy medicine. Rather than changing the routine on a hunch, the behavior analyst asks the parents for consent, contacts the prescribing physician, shares the sleep data and agrees on what to keep recording while the physician reviews the medicine.

Common confusion

Collaboration is not deference: if the data show that a recommended approach is harming the client, the behavior analyst raises it, with the data. It is not overreach either: helping a family prepare questions for a physician is appropriate, while deciding about the medicine is not. And a shared professional purpose is not a substitute for consent to share client information.

Key points (4)
  • Behavior analysts coordinate with other professionals serving the client, with consent to share information, so that plans are consistent and the client's interest comes first.
  • Caregivers and other stakeholders take part in selecting goals and are trained so the intervention is implemented consistently across settings.
  • Disagreements are worked through by sharing data, listening to others' observations and seeking a joint plan, without dismissing others or deferring against the client's data.
  • Questions outside behavior analysis (medical, nutritional, medication) are referred to or coordinated with the professional who has that expertise.

Keep going

Practice the whole chapter in the course

Original questions on every task of area H, with explanations and a record of which tasks you have mastered.

Open the course

Try a sample free, without an account