The big idea
F.6 covers the standard conditions, how to read a clear functional analysis (FA) graph, and the main alternative formats. That is enough for the exam, but many real analyses are not clear. Behavior shows up everywhere, or nowhere, or in two conditions at once, or it is so dangerous that even a few occurrences are too many. This guide covers what analysts do in those cases.
The logic is the same throughout: when results are unclear, change one thing at a time and keep a control. Vollmer, Marcus, Ringdahl, and Roane (1995) described a four-phase progression that starts with a brief analysis and moves, only while results stay unclear, to longer analyses and designs that control for interaction between conditions. Clear, replicable patterns emerged for 85% of their 20 participants.
Breaking it down
Ruling automatic reinforcement in or out
Responding at similar levels in every condition, including play, may mean automatic reinforcement. It may also mean the person never discriminated the conditions. Two tools help:
- Extended alone or no-interaction series. Run that condition repeatedly. Socially maintained behavior tends to fall across sessions, since nothing follows it; automatically maintained behavior persists.
- Screening first. Querim et al. (2013) ran a series of 5-minute alone or no-interaction sessions before a full FA. Persistence predicted automatic reinforcement in 21 of 22 cases, and decreasing responding predicted a social function in 7 of 8. Use no-interaction, with a therapist present, for any behavior that is unsafe unattended.
Subtypes of automatically maintained self-injury
Roughly a quarter of self-injury is automatically maintained (Iwata, Pace, Dorsey, et al., 1994). Hagopian, Rooker, and Zarcone (2015) proposed subtypes based on the FA pattern:
- Subtype 1: high when alone, clearly lower in play. Alternative reinforcement competes with it.
- Subtype 2: stays high even in play. Alternative reinforcement does little.
- Subtype 3: the person also restrains themselves, for example by wrapping their hands in clothing.
Across published datasets, Subtype 1 was generally treatable with reinforcement alone, while Subtype 2 generally was not (Hagopian et al., 2017). Subtype 2 cases more often needed added components, such as response blocking or protective equipment, with appropriate review. The lesson: the play condition carries diagnostic information too.
Multiple control and pairwise designs
Behavior is multiply controlled when more than one reinforcer maintains it, seen as elevation in two or more test conditions. In a multielement FA, conditions can also interact: escape in one session can carry over into the next, and rapid alternation of five conditions can blur discrimination.
A pairwise functional analysis (Iwata, Duncan, Zarcone, Lerman, & Shore, 1994) alternates one test condition with its control in each phase, then moves to the next pair: attention against play, then demand against play, and so on. Each phase is easier to discriminate, and each function gets a clean test. In that study, the pairwise design clarified results for two of five participants whose multielement analyses were somewhat undifferentiated.
Idiosyncratic functions
Standard conditions miss some variables. Reviews of modified FAs (Schlichenmeyer et al., 2013) describe examples such as interrupted routines, transitions, divided attention (the adult talks with someone else), particular forms of attention and instructions to stop a preferred activity. When the interview or observation points to one, build an individualized test with a matched control and change only the suspected variable. A test that differs from its control in several ways cannot show which one mattered, which is the same concern raised about synthesized analyses in J.1.
Lowering risk: latency and precursor analyses
For behavior that could cause serious injury, the aim is to show function with as few occurrences as possible.
- In a latency-based FA, each session ends at the first response, and latency is the measure. Thomason-Sassi et al. (2011) found an inverse relation between rate and latency, so short latencies stand in for high rates.
- In a precursor FA, consequences follow a milder behavior that reliably comes first (Smith & Churchill, 2002). Fritz et al. (2013) ran separate analyses of precursors and severe behavior and found matching outcomes in 7 of 8 cases. The method assumes a shared function, so check it where you safely can.
Termination criteria and safety protocols
Write the safety plan before the first session, as part of informed consent (Standard 2.11, Obtaining Informed Consent) and risk planning (Standard 2.13, Selecting, Designing, and Implementing Assessments). It usually covers:
- medical review first (Standard 2.12, Considering Medical Needs);
- a session-ending rule tied to injury severity, for example a rating on the Self-Injury Trauma Scale (Iwata, Pace, Kissel, Nau, & Farber, 1990);
- a maximum number of responses per session;
- protective equipment and trained staff where aggression is likely;
- criteria for stopping the whole analysis.
In a records review of 99 inpatients, injuries during FAs were uncommon and usually minor when such precautions were in place (Kahng et al., 2015).
Reading ambiguous results
Visual inspection of FA graphs is less reliable than people assume. In Hagopian et al. (1997), interrater agreement among untrained raters averaged .46.
Their structured criteria work in three steps:
- Draw a line about one standard deviation above the control condition’s mean, and another the same distance below it.
- For each test condition, count the points above the upper line and the points below the lower line.
- Call the condition elevated when clearly more points fall above than below.
After training on these criteria, agreement averaged .81. Roane et al. (2013) adapted them for analyses of different lengths. Saini, Fisher, and Retzlaff (2018) showed that applying them while the FA is running identified function in about 40% fewer sessions.
FA outside the clinic
- Schools: a trial-based functional analysis runs brief test and control trials, each a minute or two long, within the day’s ordinary routines instead of in separate sessions. Trial-based FAs embedded in class routines matched a standard FA in 6 of 10 cases and partly in a 7th (Bloom et al., 2011).
- Outpatient clinics: brief FAs and the interview-informed analyses in J.1.
- Telehealth: parents coached remotely conducted FAs that identified function in 18 of 20 cases (Wacker et al., 2013).
Seeing it in practice
| Graph pattern | Possible explanation | Next step |
|---|---|---|
| High in every condition, including play | Automatic reinforcement, or poor discrimination | Extended alone series; note the play level for subtyping |
| Elevated, variable data in attention and demand | Multiple control or carryover | Pairwise: attention against play, then demand against play |
| Low everywhere, though caregivers report daily outbursts | A missing variable | Interview again, then test that variable against a matched control |
| Severe head hitting in early sessions | Risk too high for repeated responding | Latency-based or precursor FA with written termination rules |
Common mix-ups
- Undifferentiated does not mean automatic. It can mean automatic reinforcement, poor discrimination or a missing variable. Test before concluding.
- Pairwise is not a reversal design. Each phase still alternates a test and its control.
- Subtype 2 is not untreatable. It is less responsive to reinforcement alone and usually needs more components.
- A precursor FA is only as good as the precursor. If the precursor serves a different function, the analysis points the wrong way.
Exam tip: The exam expects you to read clear FA graphs and to know that high responding when alone, or everywhere, suggests automatic reinforcement (F.6). Subtypes, pairwise designs and structured criteria go beyond it.
Quick recap
- An extended alone or no-interaction series separates automatic from social functions.
- Subtype 1 automatic self-injury drops during play; Subtype 2 does not; Subtype 3 involves self-restraint.
- Pairwise designs separate multiple functions; individualized conditions test idiosyncratic ones.
- Latency-based and precursor FAs, written termination rules and medical review limit risk.
- Structured visual-inspection criteria make unclear graphs easier to interpret consistently.
Further reading
- Vollmer, T. R., Marcus, B. A., Ringdahl, J. E., & Roane, H. S. (1995). Progressing from brief assessments to extended experimental analyses in the evaluation of aberrant behavior. Journal of Applied Behavior Analysis, 28(4), 561-576.
- Hagopian, L. P., Rooker, G. W., & Zarcone, J. R. (2015). Delineating subtypes of self-injurious behavior maintained by automatic reinforcement. Journal of Applied Behavior Analysis, 48(3), 523-543.
- Hagopian, L. P., Fisher, W. W., Thompson, R. H., Owen-DeSchryver, J., Iwata, B. A., & Wacker, D. P. (1997). Toward the development of structured criteria for interpretation of functional analysis data. Journal of Applied Behavior Analysis, 30(2), 313-326.
Key terms
- Functional analysis
- An assessment in which the analyst systematically arranges antecedents and consequences, such as attention, escape from demands and access to items, and compares responding with a control condition to demonstrate which variables maintain a behavior.
- Automatic reinforcement
- Reinforcement that occurs without another person delivering it: the behavior itself directly produces the consequence. It can be positive (adding stimulation) or negative (removing discomfort), and the reinforcer can be unconditioned or conditioned.
- Pairwise functional analysis
- A functional analysis in which each phase alternates a single test condition with its control, such as attention against play, before moving on to the next pair (Iwata et al., 1994). Testing fewer conditions at a time makes them easier to discriminate and reduces interaction between conditions.
- Multielement design
- A single-case design in which two or more conditions alternate rapidly, often from session to session, each paired with a distinct stimulus and plotted as its own data path. Differences between the data paths show the effect of the conditions.
- Visual analysis
- The systematic inspection of graphed data to judge whether behavior changed meaningfully and whether the design shows that the independent variable caused the change. It considers level, trend and variability within phases, and immediacy, overlap and consistency across phases.
- Latency-based functional analysis
- A functional analysis in which each session ends after the first occurrence of the behavior, and the measure is the latency to that response. Shorter latencies in a test condition suggest the function, and far fewer responses occur overall.
- Precursor functional analysis
- A functional analysis in which consequences are delivered for a precursor, a less dangerous behavior that reliably comes before the target behavior, so that function can be identified with fewer occurrences of the dangerous behavior. It assumes the precursor and the target share a function.
- Trial-based functional analysis
- A functional analysis made up of short trials embedded in a person's ongoing routines, each with a control segment and a test segment. Function is inferred by comparing the percentage of trials with behavior in each segment (Bloom et al., 2011).
Check yourself
4 questions on J.3. Choose an answer to see whether you're right and why.
Show the answer
Answer: D. Behavior maintained by social reinforcement tends to decrease across repeated sessions in which nothing follows it, while automatically maintained behavior persists. Querim et al. (2013) found that this pattern predicted automatic reinforcement in 21 of 22 cases. Poor discrimination is a fair worry after the first FA, but the extended series is the test that addresses it.
Show the answer
Answer: D. Automatically maintained self-injury that drops when alternative reinforcement is available fits Subtype 1, which has often responded to reinforcement-based treatment alone. Subtype 2, which stays high even during play, more often needs added components. Low rates in play with free attention argue against an attention function, and self-restraint defines Subtype 3.
Show the answer
Answer: B. Alternating many conditions can produce carryover and poor discrimination. A pairwise design compares one test condition with its control at a time, which can show whether one or both functions are real. Adding a condition makes discrimination harder, and ignoring the demand data risks leaving a function untreated.
Show the answer
Answer: B. The criteria draw upper and lower lines around the control condition's mean, then compare how many points in each test condition fall above the upper line with how many fall below the lower one. Comparing means ignores variability and trend. Overlap with the alone condition says nothing about elevation relative to the control.