Translated from the Portuguese original. Read the original (Português)
The Matching Law is a mathematical description of choice. Discovered by Richard Herrnstein in 1961, it states that, when two activities are available at the same time, the proportion of responses devoted to each one tends to track the proportion of reinforcers each one produces. In other words: we don't choose at random, nor just by what is "good" — we choose by what is relatively better among the options available in that context.
The matching law states that, under concurrent schedules of reinforcement, the relative rate of responding on one alternative equals the relative rate of reinforcers obtained from it: B₁ / (B₁ + B₂) = R₁ / (R₁ + R₂), where B is behavior (responses or time) and R is reinforcement. The original formulation is Herrnstein's (1961). The version accepted today is the generalized, or power-function, version (Baum, 1974): log(B₁/B₂) = a · log(R₁/R₂) + log b, where a is the sensitivity to the ratio between reinforcers and b is the bias, a constant preference for one of the alternatives.
The experiment that gave rise to the law
In 1961, psychologist Richard Herrnstein published a study that looked deceptively simple. Pigeons had two keys available, and each key delivered food on its own variable-interval schedule — that is, both paid off, but at different rates. Nothing forced the animal to stick with one or the other: it could split its responding however it wanted. This arrangement has a name: a concurrent schedule — the term the field's manuals, including Cooper, Heron, and Heward's, use for two or more reinforcement schedules available at the same time, each tied to a different response.
What Herrnstein found was a striking regularity. If one key delivered, say, three-quarters of all the food in the session, the pigeon would devote to it roughly three-quarters of its pecks. The distribution of behavior matched the distribution of reinforcement — hence the name.
Relative reinforcement, not absolute
Until then, reinforcement had mostly been studied in absolute terms: the more reinforcement, the more behavior. The matching law showed that what organizes choice is relative reinforcement — the same consequence can sustain a lot or a little behavior, depending on what the other options are offering.
In 1970, Herrnstein took the idea further: even when we measure a single activity, it competes with all the reinforcement available in the context, including sources the researcher isn't measuring. That's why a behavior can dwindle without anyone having touched it directly — it's enough for the surrounding environment to have become more generous.
Reading the matching graph
The matching law is usually represented on a graph that puts the proportion of reinforcers on one axis and the proportion of responses on the other. If matching were perfect, every point would fall exactly on the diagonal.
When matching isn't perfect
In practice, the data almost never fall exactly on the diagonal. In 1974, William Baum proposed the generalized version of the law, which accommodates these deviations with two parameters — and it was this version that research in the following decades confirmed. Tap the cards to see the three possible outcomes:
What the data showReanalyzing 103 data sets from 23 studies of choice, Baum (1979) found that almost all of them fit the generalized equation closely, and that sensitivity measured in number of responses tended to fall below 1 — that is, undermatching. He also proposed a practical criterion: sensitivity values between 0.90 and 1.11 can be considered good approximations of matching. In a comprehensive review, McDowell (2013) concluded that it's the power-function version — not the original strict formulation — that holds up against the evidence, in both animals and humans.
Not just the frequency of reinforcement
The classic version of the law talks about the rate of reinforcement, but real choice responds to more dimensions than that. In a study with six youths with learning and behavior difficulties, Neef, Shade, and Miller (1994) offered two sets of math problems at the same time and examined four dimensions: the frequency, the quality, and the delay of reinforcement, plus the effort required by the response. In each comparison, the two sets were equal on two of these dimensions and competed on the other two — which made it possible to see which one weighed more. The distribution of time was affected by these dimensions of reinforcement and response, in patterns that varied from student to student. Explore each dimension:
How often reinforcement comes
This is the original law's dimension: the more often an alternative pays off, the more behavior it attracts. But frequency isn't the only factor — and when the other dimensions compete with it, it can even lose.
How good the reinforcement is
Two reinforcers delivered at the same frequency don't carry the same weight if one of them is far more valued by the person. That's why preference assessment is an important practical step: it helps identify what actually works as a reinforcer for that person.
How long until reinforcement
An immediate consequence competes at an advantage with a delayed consequence, even when the delayed one is larger. This is one of the reasons short-term behaviors — and many problem behaviors — have the edge: the response "pays off" right away.
How much responding costs
If an alternative requires much more effort, it loses ground even when reinforcement is equal. In an intervention, this is a practical lever: reducing the effort of the desired response — simplifying the task, teaching an easier way to ask — tips the balance in its favor.
Why this matters outside the lab
The matching law was discovered with pigeons, but its usefulness lies in describing choice in far more complicated contexts. Some published examples:
- Problem behavior and appropriate behavior. Borrero and Vollmer (2002) observed four people with developmental disabilities interacting with their caregivers and recorded, in parallel, the possible reinforcers, the problem behavior, and the appropriate behavior. After identifying through functional analysis what maintained each problem behavior, the descriptive data were reanalyzed in light of the matching equation: the proportional rate of problem behavior, relative to appropriate behavior, approximately tracked the proportion of reinforcement the problem behavior obtained — in all four participants.
- Classroom. Mace, Neef, Shade, and Mauro (1994) gave three adolescents with special educational needs a choice between two piles of math problems, with reinforcers distributed in ratios of 2:1, 6:1, and 12:1. Time devoted to each pile bore a linear relation to the reinforcement obtained — but with substantial undermatching and bias, and only after auxiliary procedures (changeover delays, timers, demonstrations) did changing the schedule start to change the choice.
- Sports. Vollmer and Bourret (2000) applied the equation to the men's and women's basketball teams of a large university, comparing two- and three-point shots and adjusting the equation for the larger magnitude of the three-point reinforcer (1.5 times that of a two-pointer). For players with substantial court time, the overall distribution of shots was predicted by the equation — game by game there was variation, but the cumulative proportion across the season was predicted with remarkable precision, player by player, on both teams.
The practical implication is direct and lies at the heart of differential reinforcement: to reduce an unwanted behavior, it's not enough to try to "take away" its reinforcement. The desired alternative needs to be made relatively better — more frequent, more valued, more immediate, and easier than the option you want to decrease.
Three common misunderstandings1. The matching law does not say that the person calculates proportions in their head: it's a description of how behavior distributes itself over time, not a theory about mental arithmetic. 2. It describes the aggregate — how behavior is divided over a period — not which response comes next. 3. It doesn't work as a magic formula outside the lab: in applied studies, measuring reinforcement "in the real world" is hard, and both Borrero and Vollmer (2002) and Mace and colleagues (1994) pointed out methodological limits, with practical and conceptual consequences, for the quantitative application of the law.
Test your understanding
Quick quiz
Three questions to lock in the concept. You'll see the explanation right away.
Further reading
- Herrnstein, R. J. “Relative and absolute strength of response as a function of frequency of reinforcement”. Journal of the Experimental Analysis of Behavior, 4(3), 267–272, 1961.
- Herrnstein, R. J. “On the law of effect”. Journal of the Experimental Analysis of Behavior, 13(2), 243–266, 1970.
- Baum, W. M. “On two types of deviation from the matching law: bias and undermatching”. Journal of the Experimental Analysis of Behavior, 22(1), 231–242, 1974.
- Baum, W. M. “Matching, undermatching, and overmatching in studies of choice”. Journal of the Experimental Analysis of Behavior, 32(2), 269–281, 1979.
- McDowell, J. J. “On the theoretical and empirical status of the matching law and matching theory”. Psychological Bulletin, 139(5), 1000–1028, 2013.
- Borrero, J. C. & Vollmer, T. R. “An application of the matching law to severe problem behavior”. Journal of Applied Behavior Analysis, 35(1), 13–27, 2002.
- Vollmer, T. R. & Bourret, J. “An application of the matching law to evaluate the allocation of two- and three-point shots by college basketball players”. Journal of Applied Behavior Analysis, 33(2), 137–150, 2000.
- Mace, F. C., Neef, N. A., Shade, D. & Mauro, B. C. “Limited matching on concurrent-schedule reinforcement of academic behavior”. Journal of Applied Behavior Analysis, 27(4), 585–596, 1994.
- Neef, N. A., Shade, D. & Miller, M. S. “Assessing influential dimensions of reinforcers on choice in students with serious emotional disturbance”. Journal of Applied Behavior Analysis, 27(4), 575–583, 1994.
- Cooper, J. O., Heron, T. E. & Heward, W. L. Applied Behavior Analysis. 3rd ed. Pearson, 2020.
This entry is educational and does not replace assessment, diagnosis or treatment by a qualified health professional. Every person is unique — for a specific case, seek individual guidance.