Functional Analysis: How ABA Finds Why Behavior Happens

Functional analysis tests why problem behavior happens. Learn how FAs work, which formats to pick, and where ABA experts disagree about them.

Key takeaway

A functional analysis is a test that shows why a problem behavior happens. It changes one part of a person's world at a time and watches the behavior.

Confessions of a New Behavior Analyst in Functional Analysis
Watch the full CEU recording

Confessions of a New Behavior Analyst in Functional Analysis

Matt Harrington · 2.5 CEU · 142 min
Watch on openceu.com →

Free learning path on this topic

Child Development for Behavior Analysts

Typical development for behavior analysts. Kristen Byra's overview, then Kelly Brzak's deep dives on ages 2 to 5, 6 to 8, and 9 to 11, with milestones, common problem behaviors, and programming tips for each band.

Start the path →4 courses · 4 CEUs · Watch at your own pace

Experts who cover this

13 recordings on openceu cover functional analysis.

A functional analysis is a test that shows why a problem behavior happens. It changes one part of a person's world at a time and watches the behavior. If behavior climbs in one condition and stays low in another, you have found its function. A function is the job the behavior does for the person, like getting attention or escaping a task.

This matters because treatment works best when it matches that job. A plan built for attention will likely miss if the behavior is really about escape. The experts below agree on that goal. They do not always agree on the best way to reach it. They also disagree about how far the results can be trusted outside the test room.

What a functional analysis tests#

Dr. Joshua Jessel compares the functional analysis to an allergy test. He had one done himself not long ago, so the picture was fresh for him. In an allergy test, a doctor places small amounts of different allergens on the skin. Then the doctor looks for a reaction, compared to plain saline. The saline is the control, a condition where nothing is expected to happen. The reaction tells the doctor which treatment to use.

A functional analysis works the same way. Instead of allergens, the clinician arranges different parts of the environment. Then the clinician watches which arrangement makes problem behavior show up. Jessel says this is what sets the method apart in health care.

The functional analysis is really the comparable assessment to other health professionals because it's the only empirical demonstration of functional control.

From the talk — Dr. Joshua Jessel

Dr. Tom Szabo uses the same allergy picture in a talk about acceptance and commitment training, or ACT. He walks through what a "rash" looks like in behavior terms. In one condition, problem behavior gets attention, and behavior does not go up. In another, problem behavior gets the person out of a task, and behavior climbs.

We expose you to conditions under which your problem behavior gets you out of having to do tasks and then your problem behavior bumps up.

From the talk — Dr. Tom Szabo

Negative reinforcement means a behavior grows because it takes something unpleasant away. In this example, the unpleasant thing is the task.

Why test conditions make behavior so clear#

Matt Harrington explains why test conditions produce such clear results. He uses an idea he calls contingency strength. A contingency is an if-then link between a behavior and what follows it. He scores its strength on a scale from negative one to one. A score of one means the behavior always pays off. Zero means the payoff is random.

Think about the demand condition of an FA. At first, there is no reinforcement around. Reinforcement is anything that makes a behavior more likely to happen again. Then the behavior occurs, and reinforcement arrives every single time. Harrington says this is why a test condition sits at a contingency strength of one.

He adds that this setup creates a strong rate of behavior. It also makes the behavior hard to change while that setting stays in place. That is exactly what you want in a test. You want the answer to show up fast and clearly.

Extinction sits at the opposite end of his scale. Extinction means the behavior still happens but no longer gets reinforced. With perfect follow-through, the learner is more likely to be reinforced for not doing the behavior. That is the negative one value.

How it differs from other assessments#

Jessel places the functional analysis inside a bigger family called functional assessment. That umbrella holds three kinds of tools. Indirect assessments are surveys or questionnaires about the behavior. You can use them without ever seeing the behavior happen. Descriptive assessments mean watching the behavior yourself as it unfolds in daily life.

Jessel says a descriptive assessment gives you, at most, correlations. A correlation means two things tend to happen together. It does not prove that one causes the other. That leaves the clinician with some uncertainty. In his view, the functional analysis is the best option for reducing that uncertainty.

He also points to a gap between the law and ethics. If a child shows problem behavior, the rules require a functional assessment. But almost any functional assessment will count. A quick caregiver questionnaire can meet the legal bar. So can five minutes of watching the child. Jessel argues that ethics asks for more than that. Some functional assessment methods simply are not as effective as others. Choosing a weaker tool when a stronger one is possible can lead to a plan that misses the real cause.

Matt Harrington pushes back on a narrow view of the method. Many people picture one fixed set of steps from the early Iwata papers. Harrington says that view is far too rigid.

Functional analysis is absolutely not a set of formal procedures set forth by Iwata et al. 1984-92 and meant to be taken in stone at that moment. It's not the Ten Commandments of functional assessment.

From the talk — Matt Harrington

He prefers a definition he credits to Schlinger and Norman in 2013. It calls the FA a pretreatment clinical assessment. It changes possible antecedents and possible reinforcers for problem behavior. An antecedent is whatever happens right before a behavior. Harrington notes the method also works for verbal behavior, not only challenging behavior.

In plain terms, he calls it a simple search for an if-then relationship. Under that view, many formats qualify. A latency-based FA, which measures how long it takes a behavior to start once a condition begins, counts. A traditional FA counts. An ISCA, a test that combines several likely reasons into one condition, counts too.

It's a search for an if-then relationship with at least one or more test and control conditions. These are all functional analyses.

From the talk — Matt Harrington

So the goal shifts. Harrington says we are not hunting for the perfect ISCA or the perfect traditional FA. We are hunting for the method that best serves the client in front of us.

Changing consequences or changing motivation#

Szabo describes two broad ways to build a functional analysis. The first style changes the consequences. You set up conditions where behavior earns attention, or earns escape, and compare the results. That is the classic analog approach. Analog means a planned session that copies parts of real life under tighter control.

The second style changes motivation instead. Szabo talks about establishing operations and abolishing operations. An establishing operation makes a reinforcer more valuable right now. Being hungry, for example, makes food worth more. An abolishing operation does the opposite and makes a reinforcer worth less.

Szabo says this second idea was suggested back in 2001 but not studied at the time. Later, a book on functional behavior assessment laid out a strategy for doing it. More recently, he says, Greg Hanley has described changing these operations as the basic strategy in his practical approach to FA.

Szabo admits that running an ACT version of an FA takes a while to learn well. Still, the core logic does not change. You alter one part of the situation and see what the behavior does. Whether you change what comes after the behavior or what comes before it, you are looking for a clear, repeatable shift.

Where the functional analysis came from#

Stephanie Peterson uses the FA as her go-to example of research turning into practice. She describes a loop. Basic researchers study principles like reinforcement and extinction. Applied researchers build on those findings with real clients. Over time, that work can become a tool clinicians use every day.

She says the functional analysis grew out of Brian Iwata's work, which came from basic research on reinforcement. After that, researchers kept studying how to run FAs more effectively and efficiently. She names the ISCA as another tool still being developed from that work.

So if we think about assessment and treatment of problem behavior, that development of functional analysis has changed the way we do practice.

From the talk — Stephanie Peterson

Peterson says the shift happened while she was in graduate school. It reshaped how the field assessed problem behavior. That, in turn, reshaped how the field treated it.

Harrington adds a note about the size of the evidence. He points out that the traditional FA has been around since 1984. It has hundreds and hundreds of replications in the literature. A replication is a study that repeats earlier work to see if the results hold up. That long track record is part of why the traditional FA still sets the standard others are measured against.

The problem of time#

Jessel's main concern is efficiency. His team reviewed every functional analysis published over a ten-year span. They multiplied the number of sessions by session length to get a total. That number left out setup, prep, and staff training.

What we found is that the standard functional analysis takes an average of around three hours to conduct.

From the talk — Dr. Joshua Jessel

Jessel adds that this average assumes a clear result on the first try. It does not count unclear results or repeat analyses. He wants an FA built for practitioners, not for researchers in an ivory tower. He wants something quick, practical, and low in cost. He calls himself a "practitioner scientist" to show which role comes first for him.

He then looks at the ISCA. It usually needs about five sessions: three tests and two controls. Across the 127 ISCAs in the literature at that time, the average was 30 minutes. He calls that around an 80% gain in efficiency.

the functional analysis is the only time in the entire assessment treatment process where the problem behavior is intended to be evoked and reinforced.

From the talk — Dr. Joshua Jessel

Jessel gives three reasons speed matters. The FA is the one step where problem behavior is meant to happen and pay off. A shorter FA means less exposure to that. And a faster FA moves the learner to treatment sooner. He says the FA is a necessary step, but not one he enjoys doing.

Picking the right format#

Harrington lists many formats a clinician can pick from. They include the ISCA, the practical functional assessment (PFA), and the traditional FA from Iwata in 1984. They also include latency-based, trial-based, and brief FAs. A trial-based FA places short test and control conditions inside natural settings. One recent study even used it with pet dogs, and the results matched a standard FA for every dog.

He adds that indirect and direct assessments, like ABC data and caregiver surveys, also exist. ABC data means writing down what happened before, during, and after a behavior. These sit lower on what he calls the continuum of scientific rigor. He wants an assessment that is both quick and accurate.

I highly recommend things like the latency-based functional analysis, the single-session ISCA, the brief trial-based functional analysis.

From the talk — Matt Harrington

He says newer research is making FAs very fast. A single-session ISCA can take about 10 minutes. A brief FA might take five. Then you confirm a function-based treatment and move on to teaching.

In another talk, Harrington explains which client details should drive the choice. A language skills score matters a lot for a complex communication plan. But that score likely matters little for choosing an FA. The setting matters much more.

Are we going to do a latency based FA? Are we going to do a trial based FA? Are we going to do a full FA? Are we going to do a PFA?

From the talk — Matthew Harrington

Isolated versus synthesized conditions#

An isolated contingency analysis tests one possible reason at a time. Attention gets its own condition, and escape gets its own condition. A synthesized analysis, like the ISCA, combines reasons that an interview suggests go together. Harrington reviewed a 2015 study that compared the two for one client named Dan.

Dan's ISCA showed zero challenging behavior in the control condition. His test condition combined escape to tangibles plus attention. Tangibles are physical items, like toys or a tablet. The result was clear. The team then tested attention alone, because the interview hinted that attention mattered.

Now let's go to isolated contingency analysis... you can see here that there is no differentiation throughout this process indicating that this was not an effective way to identify a function.

From the talk — Matt Harrington

Differentiation means behavior looks clearly different between test and control. Without it, you cannot name a function. Harrington also points to a study that gathered 30 replications of the ISCA. He frames it as Hanley's team working to catch up to the traditional FA's large evidence base.

Harrington shared one of his own cases as well. A boy reacted strongly to transitions, almost like a fight-or-flight response. He started the case as he started most cases.

I ran a functional analysis. I ended up doing an ISCA and we saw a really clear synthesized reinforcement contingency for, um, access to preferred activities, as well as escape from an aversive transition.

From the talk — Matt

From there, the team taught a safe way to ask and built tolerance for hearing "no." The plan matched what the FA had found.

Choosing by client, not by camp#

Harrington sees camps forming in the field. Some people favor synthesized analyses. Others want to stay with isolated ones. He calls that fight a bit of a red herring, a distraction from the real issue.

We should pick our functional analysis methodology based on what assessment is going to best serve the individual in front of us.

From the talk — Matt Harrington

He lists the barriers to weigh before choosing. These include environmental barriers, implementer barriers, and competency barriers. Ethical barriers count too. He tells a real story about stopping his own assessment midway and finishing the next day. He says he still blushes when he thinks about it. The lesson is that the format has to fit the real situation, not a favorite researcher.

He raises a related debate from feeding work. There, an FA or ISCA often shows escape and attention as dual functions. Dual functions means the behavior does two jobs at once.

There's a lot of debate over if attention is a relevant function or if that's an artifact of the functional analysis.

From the talk — Matt Harrington

An artifact is a result created by the test itself rather than by real life. This question shows why format choice is not a small detail.

Setting, staff, and resources#

Harrington warns against copying research without checking its context. He describes an inpatient unit with full meal service, protective equipment, and plenty of materials and time. He says great research comes from settings like Kennedy Krieger Institute. But a learner may later move to a group home with five clients for each staff member.

Was there a functional analysis under the constant supervision of nurses who were constantly intervening and changing the environment to maintain safety? And we don't have nurses to do that.

From the talk — Matt Harrington

In that case, he suggests a different style of FA may fit better. He also checks who ran the sessions in a study. Sometimes the first author ran every session. That is very different from a newer, younger staff member.

Schools raise their own questions. When an escape extinction plan is failing and aggression is rising, Harrington asks first if the function is even right.

what are the results of the functional analysis? Was there an experimental functional analysis? Can we do an experimental functional analysis within the school setting?

From the talk — Matthew Harrington

His search tip is simple. Put "functional analysis" in quotes and add school, not home, clinic, or hospital.

Dr. Jessica Osos adds a billing angle. In her talk on training RBTs (Registered Behavior Technicians, the front-line staff who run many ABA sessions) to support assessments, she says FA work typically falls under the older T codes. It is not part of the technician-supported assessment an RBT would bill for.

What happens when the FA is skipped#

Osos shows the cost of skipping this step. In one case, an analyst treats aggression with strategies that worked for other clients before. No structured assessment ever happens.

The problem is that they skipped a structured assessment, so they don't do an FA.

From the talk — Dr. Jessica Osos

Because of that, the function was never found. The plan did not match what kept the aggression going. Osos calls this an over-reliance on past experience. The research says to match the treatment to the function, and that step got lost.

Her second case is a three-year-old who bites. Here the team starts with a structured assessment. The FA shows the biting is maintained by escape from demands. That points straight to the research on escape-maintained behavior. Functional communication training has a strong evidence base there, so it becomes a fitting choice.

Harrington makes the same point from the other side. An assessment should always serve the treatment, never the other way around.

the only reason why we would run a functional analysis to identify the function of a behavior would be if we have an intervention that needs that information in order to work better

From the talk — Matt Harrington

He adds that running an FA on a random behavior would be a waste of time. That is especially true if the behavior has nothing to do with the current teaching.

Where experts disagree#

The experts in these talks agree the FA has value. They split on how far its results reach outside the test room. Jessel calls it the only empirical demonstration of functional control. Osos treats it as the step that must come before choosing a treatment.

Louis Sandler sees a real problem with the analog setup. He says the process has value, but it is decontextualized. That means it is cut off from the person's real life. He says we pretend someone has a certain escape need and build situations that are not real.

the behavior plans which may work in the clinical context then become basically artificial, too. They work. They're internally valid. They work in the clinical sense. But they don't transfer very well. They don't generalize.

From the talk — Louis Sandler PhD BCBA-D

Internally valid means the results hold up inside the test itself. Generalize means the skill or result carries over to new places and people. Sandler also argues that a true clinical FA is often not even possible in schools.

you can't really do a clinical FA in a school. you don't have the room you don't have the trained people you don't have the materials

From the talk — Louis Sandler PhD BCBA-D

He would rather watch behavior in real time. If writing sets a child off, he wants to see writing presented during language arts. He wants to see how the task is delivered, not copy it in a side room. Jessel, by contrast, warns that watching alone shows only correlations and leaves uncertainty. Harrington takes a middle path, searching for FA formats that do fit schools. Sandler's full case is in Intersection of Cultural Ontogeny, Ecological FBA and the Behavior Instructional Plan.

"Too clean" for the natural world#

Dr. Shane Spiker raises a related worry in a talk on crisis management. He lists mistakes clinicians make with antecedents. First, he says ABC analyses are not enough. Then he says FAs are typically too clean.

the natural world is not arranged in the way that FAs are arranged.

From the talk — Dr. Shane Spiker

By "too clean," Spiker means FA results can be so specific that they fit only one context. He admits this opens a can of worms. He points listeners to Hanley's work on the same idea.

His bigger concern is what clinicians do next with the results. If we bubble wrap learners and remove every trigger, we lose chances to teach. We lose motivating moments to work with. We also lose practice handling things when they go wrong. Spiker is clear that he is not saying to set off problem behavior on purpose. He is saying a clean test should not lead to a world with no hard moments at all. He also urges clinicians to watch for precursors, the small early behaviors that come before bigger ones.

When language changes the test#

Stu Law brings a different angle. He separates two meanings of functional analysis. One is a general way of thinking about why behavior happens. The other is a specific tool, the formal FA built around four functions.

there is a distinction that is sometimes made and not made regarding functional analysis as just a general way of going about things and functional analysis as a technology

From the talk — Stu Law

Law says the broad approach works well. Many therapists who help veterans, or people with drug problems, see their job as functional analysis in that broad sense. But he is skeptical about the Iwata-style four-function FA. He thinks its reach is limited to people with very little language. He also notes his wife's early intervention program never ran a single FA.

He points to research on mutual entailment. That is the skill of reversing a relation. If an uncle is cooler than an aunt, the aunt is less cool than the uncle. Law says once a learner can do this, the FA often stops working.

the chances of an FA working on you are like fall to zero. They fall to like near zero, because you can kind of understand what I'm trying to do.

From the talk — Stu Law

In his view, the learner figures out the game and can play along. He adds that this skill shows up earlier than many people expect.

What the research says#

Recent papers add detail to these talks. Brown, Helvey, Kranak, and Lavin (2025) wrote a guide for clinicians new to running FAs. It covers when to include each test condition and how to pick a method for a case. It also discusses how to keep FAs safe and ethical in any setting. The guide notes that brief FAs need high procedural integrity, meaning steps are done exactly right. Brief FAs have shown lower agreement with longer methods. It also warns that a fixed order of conditions can create sequence effects, where an earlier condition skews the results of a later one.

Brown, K. R., Helvey, C. I., Kranak, M. P., & Lavin, A. (2025). Functional Analysis Decision-Making Considerations. Behavior Analysis in Practice, 18(4), 1237-1254. https://doi.org/10.1007/s40617-025-01057-w

Hoffmann and Boyle (2025) surveyed 15 recognized FA experts about 11 FA variations. They call FA the most rigorous form of functional behavior assessment. Yet they note it is used less in practice than its track record suggests. Weak training may be one reason. The experts urged clinicians to consider several FA methods.

Hoffmann, A. N. & Boyle, M. A. (2025). A Survey of Expert Decision‐Making in Functional Analysis Methodology. Behavioral Interventions, 40(4). https://doi.org/10.1002/bin.70045

Nesselrode, Falcomata, Hills, and Erhard (2022) reviewed FAs run in public schools. They found that many failed trials could confound results. Broad labels like "problem behavior" were another possible source of confusion.

Nesselrode, R., Falcomata, T. S., Hills, L., & Erhard, P. (2022). Functional Analysis in Public School Settings: A Systematic Review of the Literature. Behavior Analysis in Practice, 15(3), 958-970. https://doi.org/10.1007/s40617-022-00679-8

Jessel, Hanley, and Ghaemmaghami (2020) reviewed FA studies from 1965 to 2016. They tracked which parts of the Iwata model became more or less common over time.

Jessel, J., Hanley, G. P., & Ghaemmaghami, M. (2020). On the Standardization of the Functional Analysis. Behavior Analysis in Practice, 13(1), 205-216. https://doi.org/10.1007/s40617-019-00366-1

FAQ#

What is the difference between a functional analysis and a functional behavior assessment?

Functional assessment is the umbrella term. It includes surveys, direct observation, and the functional analysis. Jessel explains that surveys and observation can show patterns, but not proof. A functional analysis actually tests the environment to show what controls the behavior. That makes it the strongest tool under the umbrella.

How long does a functional analysis take?

It depends on the format. Jessel's review found the standard FA takes about three hours of session time on average. The ISCA averaged about 30 minutes across 127 published cases. Harrington says some single-session ISCAs take about 10 minutes, and some brief FAs take about five.

Can you do a functional analysis in a school?

Experts disagree. Sandler says a true clinical FA is not really possible in most schools. He points to missing space, trained staff, and materials. Harrington searches for school-based FA research and weighs formats like trial-based or latency-based FAs. A 2022 review of school FAs found they can work, but failed trials and broad behavior labels can muddy results.

Turn this topic into a CEU

You just studied this. Now get credit for it.

Watch Confessions of a New Behavior Analyst in Functional Analysis with Matt Harrington and earn 2.5 free BCBA CEUs. Detailed certificate, delivered the moment you finish.

Watch and earn the CEU →Free account · No card · detailed cert

Keep going with a free learning path

Child Development for Behavior Analysts

Typical development for behavior analysts. Kristen Byra's overview, then Kelly Brzak's deep dives on ages 2 to 5, 6 to 8, and 9 to 11, with milestones, common problem behaviors, and programming tips for each band.

Start the path →4 courses · 4 CEUs · Watch at your own pace

Want the primary literature? Read the Functional Analysis research roundup on our sister site, Behaviorist Book Club.