Identify the function of the unwanted behavior

The man who quit smoking five times

He quit cold turkey in January and lasted eleven days. He quit with nicotine patches in March and lasted three weeks. He quit with a support group in June and lasted six days — worse than his first attempt. He quit with an app in September and lasted two weeks. Each time, he experienced the same arc: fierce determination, a brief honeymoon of control, a mounting pressure he could not name, and then a collapse back into the behavior he was trying to eliminate. By December, he had quit smoking five times and was smoking more than when he started.

His sixth attempt was different. Instead of deploying another elimination strategy, he sat down with a therapist who asked a question no one had asked before: "What does smoking do for you?" Not what it costs. Not what it risks. What it provides. He resisted the question at first — smoking does not do anything for me, it is killing me — but the therapist pressed. When do you smoke? After meetings. After arguments. When you get home from work. During what emotional states? Tension. Frustration. The feeling of being trapped. And what happens in the thirty seconds after the first inhale? His shoulders dropped. He exhaled. He said the tension left his body and he could think clearly again.

The function of his smoking was not nicotine delivery. It was nervous system regulation. Every previous quit attempt had removed the mechanism he used to downshift from sympathetic arousal to parasympathetic calm — and had replaced it with nothing. The pressure he could not name was his nervous system screaming for the regulatory function that smoking had provided. Until he understood this, every quit attempt was destined to fail, because he was removing a tool without understanding the job it was doing.

Function before extinction

The previous four lessons in this phase established the mechanics of behavioral extinction: the principle that unwanted behaviors can be systematically eliminated, that extinction requires removing the reinforcing reward, that extinction bursts represent a temporary escalation before a behavior fades, and that true extinction differs fundamentally from mere suppression. All of that is necessary foundation. But there is a critical step that must precede any extinction attempt, and skipping it is the single most common reason people fail to permanently eliminate unwanted behaviors.

That step is functional analysis. Before you try to extinguish a behavior, you must understand what that behavior is doing for you. Not what it costs, not what it looks like from the outside, not what label society assigns to it — but what function it serves in your psychological and physiological economy. Every behavior that persists does so because it is being reinforced, and understanding the specific nature of that reinforcement is the difference between an extinction attempt that succeeds and one that collapses into relapse.

This principle was formalized in applied behavior analysis by Brian Iwata and his colleagues in a landmark 1982 study, later expanded in 1994, that introduced a systematic methodology for identifying the function of problem behaviors. Iwata's work, conducted initially with individuals who exhibited self-injurious behavior, demonstrated that the same topography of behavior — the same observable action — could serve completely different functions in different individuals. One person's self-injury was maintained by attention from caregivers. Another's was maintained by escape from demands. A third's was maintained by the sensory stimulation the behavior itself produced. The behavior looked identical from the outside. The function was entirely different. And the intervention that worked depended entirely on which function was operative.

Edward Carr, writing in 1977, had already laid the conceptual groundwork for this insight in his influential paper on the motivational basis of self-injurious behavior. Carr argued that treating behavior as a function of its consequences — rather than as a symptom of an internal pathology — opened up entirely new intervention strategies. If a behavior persists because it produces a specific consequence, then understanding that consequence tells you exactly what the behavior is "for" and, therefore, what must be addressed if you want it to stop.

Gregory Hanley, Brian Iwata, and Bridget McCord synthesized three decades of functional analysis research in their 2003 review and confirmed what practitioners had observed consistently: interventions based on functional analysis were dramatically more effective than interventions that ignored function. When you know why a behavior occurs, you can design a precise intervention that addresses the underlying need. When you do not know why it occurs, you are guessing — and most guesses are wrong.

The four functions of behavior

Applied behavior analysis has identified four primary functions that maintain human behavior. These are not theoretical categories imposed on reality. They are empirically derived patterns observed across thousands of functional analyses in clinical, educational, and community settings. Every persistent behavior you exhibit — wanted or unwanted — is maintained by one or more of these functions.

The first function is attention or social reinforcement. A behavior maintained by attention persists because it produces a social response from others. The response can be positive — praise, engagement, laughter — or negative — scolding, concern, argument. What matters is not the valence of the attention but its presence. A child who throws tantrums in a grocery store may be doing so because tantrums reliably produce parental attention, even if that attention takes the form of a reprimand. An adult who constantly complains about their workload may be doing so because complaining reliably produces sympathetic responses from colleagues. The behavior is not about the workload. It is about the social connection the complaining generates. Remove the attention, and the behavior should diminish — but only if attention is the actual function. If you remove attention from a behavior that is maintained by escape, you have addressed the wrong variable and the behavior will persist unaffected.

The second function is escape or avoidance. A behavior maintained by escape persists because it allows the person to avoid or terminate an aversive situation. Procrastination is the canonical example. When you delay starting a difficult task, the immediate consequence is relief from the discomfort the task produces — the anxiety of potential failure, the cognitive load of complexity, the unpleasantness of boredom. That relief reinforces the delay. The procrastination is not laziness. It is a successful escape behavior. Similarly, picking a fight with your partner right before a difficult conversation you have been dreading serves an escape function — the argument replaces the conversation, and the dreaded topic is avoided. The behavior looks aggressive, but the function is avoidant.

The third function is access to tangibles or preferred activities. A behavior maintained by access to tangibles persists because it produces something the person wants. A child who cries at the checkout counter is often doing so because crying has historically produced access to candy or toys. An adult who lies on their resume is often doing so because deception has historically produced access to employment opportunities. The behavior serves as an instrument for acquiring something desirable. In the context of self-directed behavior change, this function often manifests as behaviors that provide quick access to pleasure or comfort — opening a delivery app to order food, buying something online during a stressful afternoon, pouring a drink the moment you walk in the door. The tangible is the reinforcer, and the behavior is the retrieval mechanism.

The fourth function is sensory stimulation or automatic reinforcement. This is the most misunderstood function because it does not require any external consequence. A behavior maintained by automatic reinforcement persists because the behavior itself produces a reinforcing sensory experience. Nail biting, hair twirling, skin picking, knuckle cracking, humming, rocking — these behaviors often produce tactile, proprioceptive, or auditory stimulation that the nervous system finds reinforcing. No one needs to be watching. No aversive situation needs to be present. No tangible is obtained. The behavior is self-reinforcing. This is the function that makes certain habits extraordinarily resistant to extinction, because the reinforcer cannot be removed by changing the environment. It is produced by the behavior itself.

The same behavior, different functions

Here is where functional analysis becomes indispensable: the same observable behavior can serve different functions in different people, or in the same person at different times. Consider the behavior of scrolling social media for extended periods. For one person, this behavior is maintained by social reinforcement — they are scrolling to see who liked their posts, who commented, who acknowledged their existence. The reinforcer is attention. For another person, the same behavior is maintained by escape — they scroll to avoid sitting with uncomfortable emotions after a difficult day. The reinforcer is the termination of distress. For a third person, the behavior is maintained by access to tangibles — they scroll to find specific content, deals, or information they want. The reinforcer is acquisition. For a fourth person, the behavior is maintained by automatic reinforcement — the visual stimulation of rapidly changing images and the tactile sensation of swiping produce a sensory experience the nervous system craves. The reinforcer is the stimulation itself.

These four people all look the same from the outside. They are all staring at their phones, scrolling through social media. But the extinction strategy that works for each of them is fundamentally different. Removing social feedback (turning off notifications, hiding like counts) will reduce the behavior for the first person and do nothing for the other three. Providing alternative escape mechanisms (breathing exercises, walks, journaling) will reduce the behavior for the second person and do nothing for the other three. If you apply the wrong intervention because you assumed the wrong function, you will conclude that the behavior is "resistant to change" when in fact you simply misidentified what was maintaining it.

B. F. Skinner understood this at a theoretical level decades before the formal functional analysis methodology was developed. In his functional interpretation of behavior, Skinner argued that the causes of behavior lie not inside the organism — not in personality traits, not in moral character, not in willpower — but in the contingencies of reinforcement that maintain the behavior. To understand why someone does something, you do not look at who they are. You look at what the behavior produces. The function is in the consequence, not the character.

Conducting your own functional analysis

Clinical functional analysis involves systematically manipulating environmental conditions — presenting and withdrawing attention, demands, tangibles, and alone conditions — and measuring behavior across conditions. You are not going to do that to yourself in a laboratory. But you can conduct an informal functional analysis that is rigorous enough to generate accurate hypotheses.

The primary tool is ABC recording: Antecedent, Behavior, Consequence. Every time the unwanted behavior occurs, you record three things. First, the antecedent: what was happening immediately before the behavior started? Where were you? Who was present? What were you doing? What were you feeling? Second, the behavior itself: what exactly did you do, described in observable terms rather than interpretive labels? Not "I procrastinated" but "I opened YouTube and watched three videos instead of starting the report." Third, the consequence: what happened immediately after the behavior? What did you get? What did you avoid? How did you feel in the thirty seconds following the behavior?

The power of ABC recording is in the patterns that emerge across multiple instances. A single observation tells you very little. But after five or ten instances of the same behavior, the antecedent column will start to cluster. You will notice that the behavior reliably occurs in certain situations and not others, following certain emotional states and not others, in the presence of certain people and not others. The consequence column will also cluster. You will notice that the behavior reliably produces the same type of outcome — it reliably generates attention, reliably terminates discomfort, reliably produces access to something pleasurable, or reliably generates a specific sensory experience.

Charles Duhigg, in The Power of Habit, describes an informal version of this process that he calls "craving isolation." Duhigg found himself going to the cafeteria every afternoon for a cookie and wanted to understand why. He ran a series of experiments on himself: Was the craving about hunger? He ate an apple at the same time and checked if the urge persisted. Was it about the break from work? He took a walk around the block instead and checked. Was it about socializing? He went to the cafeteria but skipped the cookie and just chatted with colleagues. Through this process of elimination, he discovered that the function of his cafeteria habit was socialization — the cookie was incidental. The behavior persisted because it reliably produced a few minutes of social interaction in the middle of an isolated workday. Once he understood the function, he could design an intervention that addressed the actual need: he started walking to a colleague's desk for a brief conversation every afternoon, and the cookie habit dissolved without willpower.

This is exactly the method you should apply to your own unwanted behaviors, and you can formalize it further with the "five whys" technique. Start with the behavior and ask why five times, each time going deeper into the functional chain. Why do I check my email compulsively? Because I feel anxious when I have not checked in a while. Why do I feel anxious? Because I worry I am missing something important. Why does missing something feel threatening? Because when I missed an urgent message last year, I was publicly criticized. Why does that criticism still drive behavior? Because my sense of professional competence is tied to responsiveness. Why is responsiveness the axis of my professional identity? Because in my family growing up, being available was how you proved you cared. Five whys, and you have moved from "I check my email too much" to the deep functional architecture of the behavior — a need for security rooted in early relational dynamics, currently expressed as compulsive monitoring. You cannot extinguish the email checking without addressing the security need. If you try, the need will find another outlet. This is the phenomenon clinicians call symptom substitution.

Symptom substitution and the cost of functional ignorance

When you extinguish a behavior without understanding its function, you remove the vehicle but leave the driver. The underlying need is still present, still seeking expression, still requiring satisfaction. And because the need has not been addressed, it will recruit a new behavior to serve the old function. You stop biting your nails, and you start picking at your cuticles. You stop drinking after work, and you start overeating after work. You stop procrastinating on reports by using a website blocker, and you start procrastinating on reports by reorganizing your desk for thirty minutes. The surface behavior changes. The function persists.

This is not a failure of willpower or discipline. It is a predictable consequence of intervening on behavior without understanding function. The psychoanalytic tradition called this symptom substitution and used it as evidence that surface-level behavioral interventions were insufficient. The behavioral tradition initially denied that symptom substitution occurred, arguing that behaviors are independent of each other. The modern synthesis recognizes that both sides were partially right. Behaviors are not connected by mysterious unconscious forces, but they can be connected by shared function. Two behaviors that serve the same function are functionally equivalent — they are interchangeable from the perspective of the reinforcement contingency. Remove one, and the contingency will select another unless you have addressed the function itself.

This is why functional analysis is not optional. It is not a nice-to-have preliminary step that makes extinction a bit easier. It is a structural requirement for extinction that lasts. If you do not know what function a behavior serves, you cannot predict what will happen when you remove it, you cannot prevent substitution, and you cannot design a replacement that serves the same need through a more adaptive channel. Replace rather than just remove will teach you replacement strategies, but those strategies require a functional hypothesis to work. You cannot replace a behavior with something that serves the same function if you do not know what the function is.

Complex and multi-functional behaviors

Not every behavior is neat enough to fit a single function. Many persistent behaviors are maintained by multiple functions simultaneously, and this multi-functionality makes them particularly resistant to change. Consider the behavior of overworking — regularly staying at the office or at your desk for hours beyond what is productive or healthy. This behavior can simultaneously serve an escape function (avoiding a difficult home life or the discomfort of unstructured time), an attention function (being seen as dedicated and receiving praise from managers or colleagues), and an access function (producing more output, which leads to promotions and financial rewards). A behavior maintained by three functions requires an intervention that addresses all three. Address only the escape function, and the attention and access functions continue reinforcing the behavior.

When you conduct your functional analysis, be open to the possibility that your unwanted behavior is doing more than one thing for you. Your ABC log may reveal different antecedents leading to the same behavior with different consequences — same behavior, different function depending on context. Monday's overworking is escape from a fight with your partner. Wednesday's overworking is access to the dopamine of productivity. Friday's overworking is the attention of being the last car in the parking lot. Same topography, three functions, three different interventions needed.

The Third Brain

An AI assistant is an unusually effective functional analysis partner because it has no ego investment in your self-narrative. When you tell a friend, "I procrastinate because I am lazy," the friend often agrees — it matches the cultural script, and challenging it feels confrontational. When you tell an AI the same thing, the AI can ask, "What happens in your body in the moment just before you start procrastinating? What do you feel, and what does the procrastination relieve?" The AI does not accept your label as an explanation. It pushes past the label to the function.

Use your AI assistant to conduct a structured functional interview. Describe the unwanted behavior in detail and then let the AI ask you a sequence of probing questions: When does it occur? Where? With whom? What precedes it emotionally? What follows it immediately? What would happen if you could not perform the behavior — what would you feel? What would go unmet? The AI can also help you map your ABC log data once you have collected several days of observations. Feed the raw data into the conversation and ask it to identify the most likely function or functions. The AI can see patterns across your entries that you may miss because you are too close to the data to notice regularities.

There is a specific prompt structure that works well here. Describe a recent instance of the behavior in as much sensory detail as you can — not just what you did, but what you saw, felt, heard, and thought in the moments before, during, and after. Then ask the AI: "Based on this description, what function does this behavior most likely serve? Is it maintained by attention, escape, access to tangibles, or automatic reinforcement? What evidence in my description supports your hypothesis?" The AI's answer is not definitive — you are the only one who can confirm whether a functional hypothesis feels accurate — but it provides a structured starting point that is far more useful than your own folk explanation.

From function to strategy

You now have a framework for understanding any unwanted behavior not as a character defect or a failure of willpower but as a functional adaptation — a behavior that persists because it reliably produces a consequence your system needs. The smoker at the beginning of this lesson was not weak. He was using the most effective nervous system regulation tool he had. His quit attempts failed not because he lacked motivation but because they removed the tool without addressing the need.

Once you identify the function, the entire landscape of intervention changes. You stop asking "How do I make myself stop?" and start asking "What does this behavior give me, and how else can I get it?" That second question is the bridge to Replace rather than just remove, which teaches you to replace rather than just remove. Replacement works because it satisfies the function through a new, more adaptive channel. But replacement without functional analysis is just guessing at what to replace with. You might replace your evening scrolling with reading, when the function of the scrolling was social connection — and reading, being solitary, fails to serve that function at all. The scrolling returns because the need was never met.

Identify the function first. Always. Before you design an extinction plan, before you set up environmental controls, before you recruit accountability partners, before you try any of the tools this phase will give you. The function is the foundation. Everything else is built on it.