Behavior Control

Abstract

Behavior control is the deliberate use of techniques to modify or manage the behavior of a person or group, and MeSH files it under behavioral disciplines and activities (tree number F04.047). The same descriptor sits in three branches of the vocabulary at once — a therapeutic method, a behavioral activity, and a form of informal social control — and that triple placement is this article's subject. The engineering base is operant conditioning: behavior is steered by arranging its consequences, whether by another agent (external control, as in a token economy) or by the person themselves (self-control). The article traces the descriptor from Skinner's analysis of reinforcement, through the applied programs that industrialized it, to the modern reframing of self-control, and works a delay-discounting example.

Keywords: behavior control, reinforcement, self-control, delay discounting

What Behavior Control Is

Behavior control is the application of methods to modify or manage the behavior of persons or groups. In MeSH it is a single descriptor (D032763) that the National Library of Medicine files in three places at once: under Therapeutics (E02.085), under Behavioral Disciplines and Activities (F04.047), and under Social Control, Informal (I01.880.630.099). The triple placement is not an accident of filing — it records the three faces of the same activity. Arranging a person's consequences to change what they do is a treatment when a clinician does it to relieve a disorder, a behavioral technique when a researcher studies it, and social control when an institution uses it to keep order. One set of methods; three institutional uses.

The methods themselves are the machinery of operant conditioning. B. F. Skinner's central claim was that behavior is selected by its consequences the way traits are selected by their reproductive payoff: a response followed by a reinforcing consequence grows more frequent, one followed by a punishing consequence grows less frequent (#ref-skinner-1953). Behavior control is that principle turned into a tool. To increase a behavior one makes a valued consequence contingent on it; to decrease one, the arranger withholds that consequence or makes an aversive one contingent instead. Everything from a psychiatric token economy to a smartphone app that pays its user for going to the gym is a special case of arranging contingencies.

The crucial dividing line inside the descriptor is who arranges the contingencies. When another agent — a therapist, a ward, a court, a parent — controls the consequences, the control is external. When the person arranges their own contingencies, forgoing a smaller-sooner reward to secure a larger-later one, the control is self-control. The history of the field is a slow migration of interest from the first to the second: from Skinner's boxes and Azrin's wards, where the environment did the controlling, to the contemporary science of willpower and precommitment, where the question is how a person controls themselves. Both halves are the same descriptor, and both obey the same arithmetic of reinforcement.

Types of Behavior Control

MeSH gives Behavior Control one narrower descriptor, listed in Table 1. A single child is a thin taxonomy, and that thinness is itself informative: the vocabulary singles out physical restraint as the one form of behavior control distinctive enough to warrant its own heading, because it is the case where control is exerted on the body directly rather than through the arrangement of consequences. The other forms discussed in this article — reinforcement schedules, token economies, contingency management, self-control strategies — are not children of this descriptor but are filed elsewhere in the tree (chiefly under operant conditioning and the psychological techniques), and are related to behavior control as methods it employs rather than as kinds of it. Because MeSH is an indexing classification and not a claim about how the phenomenon is really carved up, the single child below reflects the Library of Medicine's filing conventions, not a settled scientific ontology of control.

Table 1
MeSH child descriptor of Behavior Control (F04.047)

Every direct technique of behavior control either adds or removes a stimulus that is pleasant or aversive. Crossing those two axes gives the four operations below. Reinforcement raises a behavior's frequency; punishment lowers it. Select a cell for a worked case.

Note the two axes are independent: “negative” means a stimulus is subtracted, not that the outcome is unpleasant — which is why negative reinforcement increases a behavior. Computed locally, not stored.

The Operant Foundation

Every technique of behavior control is a way of arranging consequences, and the vocabulary for consequences comes from the operant analysis Skinner laid out in Science and Human Behavior (#ref-skinner-1953). Two dimensions define the space. A consequence can be added to the situation or removed from it, and it can be pleasant or aversive. Crossing those dimensions yields the four operations that exhaust direct behavior control: positive reinforcement (add something pleasant), negative reinforcement (remove something aversive), positive punishment (add something aversive), and negative punishment (remove something pleasant). The first two increase a behavior; the last two decrease it. The demo above lets each cell be inspected; the worked example later quantifies one.

The power of the operant framework is that the schedule on which a consequence is delivered matters as much as the consequence itself. Skinner and his successors showed that intermittent reinforcement — rewarding only some responses, on a ratio or interval schedule — produces behavior far more resistant to extinction than continuous reinforcement, which is why a behavior maintained by an unpredictable payoff (a slot machine, a variable social reward) is so hard to stop. Behavior control in practice is therefore never just a choice of reward; it is a choice of contingency — what is delivered, for which response, and how often. A program that gets the reward right but the schedule wrong will fail, and much of the applied craft is schedule design.

Nathan Azrin, working at Anna State Hospital from the late 1950s, turned this laboratory science into an institutional technology. With Teodoro Ayllon he built the first token economy, a ward-wide system in which patients earned tokens for target behaviors — self-care, work, social participation — and exchanged them for goods and privileges (#ref-ayllon-1968). The token is a generalized conditioned reinforcer: like money, it has no intrinsic value but is backed by a menu of things that do, so it can reinforce any behavior at the moment it occurs and be cashed out later. The token economy is the purest expression of external behavior control ever fielded, and its rise and fall is the subject of the next section.

Applied Programs: Token Economies and Contingency Management

Albert Bandura's Principles of Behavior Modification codified the clinical application of operant control and gave the field its name (#ref-bandura-1969). Through the 1960s and 1970s the token economy spread from Anna State Hospital to psychiatric wards, classrooms, and prisons across the United States, and for a time it looked like the general-purpose solution to institutional behavior. Then it receded. Alan Kazdin's 1982 review, The token economy: A decade later, is the field's honest audit of what happened (#ref-kazdin-1982): the programs worked reliably inside the institution, but the gains often failed to generalize to the outside world or to maintain once the tokens stopped, and the labor of running them was heavy. Deinstitutionalization and a cultural turn against overt behavioral control did the rest. The token economy did not fail as a demonstration that contingencies govern behavior; it failed as a claim that arranging contingencies once, in one setting, would transfer.

The idea did not die — it was refined and renamed. Its most successful modern descendant is contingency management in addiction treatment, where patients earn vouchers or prize draws for objectively verified abstinence (a clean urine screen). A substantial review literature establishes that this works across ordinary community clinics, not just research settings: contingency-management protocols reliably increase retention and abstinence in the treatment of substance use disorders compared with usual care (#ref-davis-2016). Contingency management is the token economy stripped to its essentials — an unambiguous target behavior, an immediate and certain reinforcer — and it is among the most reliably effective interventions in the addictions field precisely because it respects the operant principle that the reinforcer must be immediate and contingent, the very features a delayed and uncertain natural consequence (long-term health) lacks.

A recurring objection haunts all of this, and it is a cognitive one. Edward Deci, Richard Koestner, and Richard Ryan's meta-analysis of 128 experiments found that tangible rewards made contingent on a task reliably undermined subsequent intrinsic motivation for that task (#ref-deci-1999). The finding does not overturn operant control — the reward does increase the behavior while it is in force — but it warns that external control can corrode the internal motivation that would sustain the behavior once control is withdrawn, which is a mechanism for exactly the maintenance failure Kazdin documented. The tension between controlling behavior from outside and cultivating its control from within is the pivot from the applied programs to the self-control literature.

A token economy reinforces a target behavior on a schedule. A continuous schedule (a token for every response) builds the behavior fastest; a leaner intermittent schedule builds it more slowly but makes it far more resistant to extinction once tokens stop. Adjust the schedule and program length.

An illustrative model, not measured data: a continuous schedule (FR-1) maximizes acquisition, while leaner schedules trade acquisition speed for persistence — the schedule effect Skinner identified and the reason token-economy gains can vanish or endure depending on how reinforcement is thinned. Computed locally, not stored.

From External Control to Self-Control

Self-control is behavior control in which the person is both controller and controlled: they arrange their own present circumstances so that a later, larger reward wins out over a nearer, smaller one. The founding empirical result is Walter Mischel's delay-of-gratification paradigm, in which a child chooses between one treat now and two treats after a wait, and the strategies children use to bridge the delay — looking away, reconstruing the treat as something abstract — predict outcomes measured years later (#ref-mischel-1989). Mischel's key insight was that successful self-control is not raw willpower but attention control: the child who succeeds is the one who changes what the situation means, transforming a hot, tempting stimulus into a cool, ignorable one. Control, even self-control, works by rearranging the contingencies of attention.

For two decades the dominant model of that willpower was Roy Baumeister's strength model: self-control draws on a single, limited resource that is depleted by use, so that exerting control on one task leaves less available for the next — a state called ego depletion (#ref-baumeister-2007). The model was enormously influential and generated hundreds of experiments. Then it met the replication crisis. A large multi-lab preregistered replication organized by Martin Hagger found the ego-depletion effect to be, at best, very small and possibly indistinguishable from zero (#ref-hagger-2016), and the strength model has been in serious doubt ever since. This is not a footnote; it is one of the clearest cases in psychology of a well-loved theory failing to survive rigorous replication, and it reset the field's understanding of what self-control is.

Two responses have reshaped the field. Angela Duckworth and colleagues advanced a process model of situational self-control: rather than gutting it out through willpower in the moment, effective self-controllers arrange the situation in advance — removing the temptation, adding a precommitment, changing what is in reach — so that less in-the-moment control is needed at all (#ref-duckworth-2016). Michael Inzlicht and colleagues went further, recasting the apparent depletion of willpower as a shift in motivation and attention: effort feels costly and people reallocate it, but they are not running out of a fuel (#ref-inzlicht-2021). Both moves return self-control to its operant roots. The most reliable way to control one's future behavior is not to strain against temptation when it arrives but to arrange, while still calm, the contingencies that the later, hotter self will face — which is precisely what the demo below and the worked example that follows it quantify.

A delayed reward loses present value as its delay grows: V = A / (1 + kD). The curve shows the discounted value of the later reward; the flat line is the immediate reward. Where they cross is the point of indifference. A steeper discount rate k means more impulsive choice.

At the defaults ($50 now vs $100 in 30 days, k = 0.05) the delayed reward is worth $40, below $50, so the impulsive choice wins; the crossover sits at D* = 1/k = 20 days. Lowering k pushes the crossover outward — every self-control strategy is a way to lower the effective k. Computed locally, not stored.

Worked Example

Self-control problems have a formal skeleton: a smaller reward available sooner competes with a larger reward available later, and control means choosing the larger-later. The standard model of how people value a delayed reward is hyperbolic discounting, introduced in George Ainslie's behavioral theory of impulsiveness (#ref-ainslie-1975), in which the present value V of a reward of amount A available after a delay D is

V = A / (1 + kD),

where k is the person's discount rate — how steeply they devalue the future. Consider a choice between $50 now and $100 in 30 days. The immediate reward is worth its face value, V = $50. For a person with a moderate discount rate k = 0.05 per day, the delayed reward is worth

V = 100 / (1 + 0.05 × 30) = 100 / 2.5 = $40.

Because $40 < $50, this person takes the immediate $50 — the impulsive choice — even though the delayed option is objectively twice as large. Their steep discounting has made the smaller-sooner reward win.

Now ask where the two options are exactly balanced — the crossover delay D\ at which the discounted delayed reward equals the immediate $50. Setting 100 / (1 + kD\) = 50 gives 1 + kD\* = 2, so

D\ = 1 / k*.

For k = 0.05 the crossover is D\ = 20 days: the delayed reward is preferred for any wait shorter than 20 days and rejected for any wait longer. Lowering the discount rate to k = 0.02 pushes the crossover out to 50 days — a more patient person will wait far longer — while raising it to k = 0.10 pulls the crossover in to just 10 days. This is the entire self-control problem in one equation: impulsivity is a large k, patience is a small one, and every behavior-control technique for self-control, from precommitment to reframing, is ultimately a way of lowering the effective k — making the future feel closer so the larger-later reward stays above the crossover. The demo above lets k*, the reward sizes, and the delay be varied so the crossover can be read off any configuration.

Discussion

Behavior control is one of the few places in psychology where the science is settled and the ethics are not. That contingencies govern behavior is as well established as anything in the field; Skinner's operant analysis has survived seventy years of scrutiny, and its applied forms — token economies, contingency management — demonstrably work while they are in force. What remains contested is when it is legitimate to use them, and on whom. The MeSH descriptor's placement under Social Control, Informal is a reminder that the same technology that helps a person quit cocaine can be used to keep a psychiatric ward or a prison quiet, and the history of institutional behavior modification includes genuine abuses. The one MeSH child, physical restraint, sits at the sharp end of that concern, where control is exerted on the body and the harm-versus-safety tradeoff is most acute.

The scientific frontier, by contrast, has moved decisively from external control to self-control, and there the story is one of a theory correcting itself in public. The strength model of willpower was elegant, influential, and — on the best current evidence — largely wrong, and the field's willingness to say so is a mark of health rather than failure. What has replaced it is quieter and more operant: self-control is less a muscle to be strained than a matter of arranging one's circumstances so that strain is not needed. That conclusion loops the modern literature back to Skinner. The person who succeeds at self-control is not the one with the strongest will but the one who best engineers their own contingencies — the same insight, now pointed inward, that Azrin pointed at a hospital ward.

Current Directions

The most consequential recent development is the collapse of the ego-depletion literature and the theoretical rebuilding that followed. The Hagger multi-lab replication (#ref-hagger-2016) did not merely fail to find an effect; it forced the field to ask what self-control is if it is not a depletable resource, and the answer now taking shape is a motivational and attentional one. Inzlicht and colleagues' integrative account treats the feeling of mental effort as a signal that guides the allocation of control, not as evidence that a fuel is running low (#ref-inzlicht-2021), and this reframing is reshaping how self-control is measured and trained — away from willpower workouts and toward the design of environments and precommitments.

A second active front is the maturing of contingency management from an efficacy result into a delivery problem. The technique works (#ref-davis-2016), but reaching patients at scale — through smartphone-based verification, remotely delivered reinforcers, and integration with existing care — is where the current research energy sits, alongside a parallel literature on the ethics and effectiveness of seclusion and restraint that continues to question whether the most coercive forms of behavior control have any therapeutic benefit at all (#ref-chieze-2019). The through-line of both fronts is a discipline that has stopped asking whether arranged consequences change behavior — that is not in doubt — and started asking which arrangements are effective, durable, and defensible.

Common Misconceptions

Behavior control means coercion or manipulation.

The term is neutral in MeSH: it names any deliberate arrangement of consequences to change behavior, which includes a person controlling their own behavior through self-imposed precommitment. Coercive external control is one form, not the definition (#ref-skinner-1953).

Negative reinforcement is a kind of punishment.

It is the opposite. Negative reinforcement removes an aversive condition to increase a behavior; punishment decreases a behavior. The word negative refers to subtraction of a stimulus, not to an unpleasant outcome (#ref-skinner-1953).

Rewarding a behavior always builds lasting motivation for it.

Contingent tangible rewards reliably increase a behavior while in force but can undermine intrinsic motivation for it afterward, a mechanism behind the maintenance failures of token-economy programs once the tokens stop (#ref-deci-1999; #ref-kazdin-1982).

Self-control is a limited resource that can run out.

The strength (ego-depletion) model that claimed this failed a large preregistered multi-lab replication, and self-control is now better understood as a matter of motivation, attention, and situational arrangement than of a depletable fuel (#ref-hagger-2016; #ref-inzlicht-2021).

Glossary

Behavior control.

The deliberate application of methods to modify or manage the behavior of a person or group; a MeSH descriptor filed as therapy, behavioral activity, and informal social control at once.

Contingency management.

An addiction treatment in which verified abstinence earns vouchers or prize draws; the token economy's most effective modern descendant.

Crossover delay.

The delay D* at which a delayed reward's discounted value equals a competing immediate reward; for a doubled reward it equals 1/k.

Ego depletion.

The proposed state in which exerting self-control on one task leaves less available for the next; the central prediction of the strength model, largely unsupported by replication.

Generalized conditioned reinforcer.

A reinforcer, such as a token or money, that has no intrinsic value but is backed by a menu of valued goods, letting it reinforce any behavior on the spot.

Hyperbolic discounting.

The devaluation of a reward as its delay grows, modeled as V = A/(1 + kD); a steeper rate k means more impulsive choice.

Negative reinforcement.

Increasing a behavior by removing an aversive condition contingent on it; not a form of punishment.

Operant conditioning.

The process by which behavior is shaped by its consequences; the engineering base of every behavior-control technique.

Positive reinforcement.

Increasing a behavior by adding a valued consequence contingent on it; the workhorse of applied behavior control.

Restraint, physical.

The physical restriction of a person's movement to prevent harm; the sole MeSH child of behavior control and the case where control acts on the body directly.

Schedule of reinforcement.

The rule governing which responses are reinforced and how often; intermittent schedules produce behavior more resistant to extinction than continuous ones.

Self-control.

Behavior control in which the person arranges their own contingencies so a larger-later reward wins over a smaller-sooner one.

Strength model.

Baumeister's account of self-control as a single limited resource depleted by use; influential but undermined by failed replication.

Token economy.

A behavior-control program in which target behaviors earn tokens exchangeable for goods and privileges; the purest fielded form of external control.

Key Researchers

Nathan H. Azrin (1930-2013). Co-developed the token economy and habit reversal, and ran the Anna State Hospital behavior-control research program that turned operant science into an institutional technology. Wikipedia - Wikidata

Albert Bandura (1925-2021). His Principles of Behavior Modification codified the clinical application of behavior control and introduced self-regulation as its cognitive counterweight. Wikipedia - Wikidata

Roy F. Baumeister (living). Proposed the strength (ego-depletion) model of self-control at the University of Queensland, reframing behavior control as a limited internal resource. ORCID - Wikipedia

Angela L. Duckworth (living). University of Pennsylvania psychologist whose process model of situational self-control shifted the emphasis from willpower to arranging the environment in advance. ORCID - Wikipedia

Michael Inzlicht (living). University of Toronto psychologist central to the reappraisal of ego depletion, recasting self-control as motivated choice rather than depleted strength. ORCID - Google Scholar

Alan E. Kazdin (living). Yale authority on applied behavior analysis whose The token economy: A decade later is the field's canonical audit of behavior-control programs and their generalization problem. ORCID - Google Scholar - Wikipedia

B. F. Skinner (1904-1990). Formulated operant conditioning and radical behaviorism; his analysis of reinforcement schedules in Science and Human Behavior is the engineering base of every behavior-control technique. Wikipedia - Wikidata

Frequently Asked Questions

What is behavior control? It is the deliberate use of techniques to modify or manage the behavior of a person or group by arranging the consequences of that behavior. In MeSH it is descriptor D032763, filed at once as a therapy, a behavioral activity, and a form of informal social control.

Is behavior control the same as brainwashing or manipulation? No. The term is neutral and covers any structured arrangement of consequences, including a person controlling their own behavior. Coercive external control is one form; self-control, in which the person arranges their own incentives, is another.

What is the difference between negative reinforcement and punishment? Negative reinforcement removes an aversive condition to increase a behavior; punishment decreases a behavior. The word negative means a stimulus is subtracted, not that the outcome is unpleasant, which is the single most common confusion in the vocabulary.

What is a token economy? A behavior-control program, pioneered by Ayllon and Azrin, in which people earn tokens for target behaviors and exchange them for goods and privileges. It works well inside an institution but its gains often fail to generalize or persist once the tokens stop.

Does contingency management actually work? Yes. In addiction treatment, rewarding verified abstinence with vouchers or prizes reliably increases retention and abstinence, including in ordinary community clinics, because it delivers an immediate and certain reinforcer for an unambiguous target behavior.

Is willpower a limited resource that gets used up? The strength (ego-depletion) model claimed so, but a large preregistered multi-lab replication found little or no effect. Self-control is now better understood as a matter of motivation, attention, and arranging the situation in advance than of a depletable fuel.

What is hyperbolic discounting? A model of how people value delayed rewards: present value is A/(1 + kD), where k is the discount rate and D the delay. A larger k means the future is devalued more steeply, producing impulsive choices for smaller-sooner rewards.

How can a person improve their own self-control? The most reliable strategies are situational, not effortful: remove temptations from reach, commit in advance, and reframe the tempting option. Each of these lowers the effective discount rate so a larger-later reward stays more valuable than a smaller-sooner one.

References