The big idea
People always have more than one thing they could be doing. A student can work or chat. A child can ask nicely or scream. The matching law (Herrnstein, 1961) describes how behavior is distributed among alternatives that are available at the same time: the relative rate of responding to an alternative tends to match the relative rate of reinforcement it produces.
The practical lesson is that behavior is choice. Problem behavior is one option among several, and how much of it you see depends on how much it is reinforced compared with the alternatives, not only on how much it is reinforced.
Breaking it down
Concurrent schedules
A concurrent schedule has two or more schedules running at the same time, each for a different response (B.10). Herrnstein’s pigeons faced two keys, each on its own variable-interval schedule. A changeover delay, a short period after switching during which no reinforcer can be earned, kept the birds from being reinforced simply for switching back and forth.
Matching in proportions
The basic matching law says the proportion of responses on one alternative equals the proportion of reinforcers it produces:
B1 / (B1 + B2) = R1 / (R1 + R2)
If appropriate requests earn 30 reinforcers an hour and aggression earns 10, matching predicts that about 75% of responding will be requests (30 of 40). If the reinforcers become 30 and 30, the prediction moves to 50%.
Rate is not the only dimension that matters. Magnitude, quality, immediacy and effort also shape allocation. A response that produces bigger, better, faster or easier reinforcement draws a larger share of behavior.
The generalized matching law
Baum (1974) rewrote the equation in ratio form to describe how real data depart from perfect matching. In its usual log form:
log(B1/B2) = a × log(R1/R2) + log b
The generalized matching law has two fitted parameters:
- Sensitivity (a): how closely response allocation tracks reinforcement allocation. A value of 1 is perfect matching.
- Undermatching (a less than 1): the response split is less extreme than the reinforcement split, closer to indifference. If 80% of reinforcers come from A but only 65% of responses go to A, that is undermatching. It is the most common pattern in data.
- Overmatching (a greater than 1): the response split is more extreme than the reinforcement split. If 80% of reinforcers come from A and 90% of responses go to A, that is overmatching. It can happen when switching between options is costly.
- Bias (b): a consistent preference for one alternative that the reinforcement rates do not explain. Differences in effort, reinforcer quality, location or the person delivering reinforcement can all produce bias.
Herrnstein’s extension
Herrnstein (1970) extended the idea to single responses: the absolute rate of a behavior depends on its reinforcement relative to all reinforcement available in the setting. Adding reinforcement for other things, as in an enriched environment or NCR, can lower a target behavior even when its own reinforcement is unchanged.
Applying it to problem behavior
Problem behavior and appropriate behavior can be treated as concurrent operants. Borrero and Vollmer (2002) found that the allocation of problem and appropriate behavior in descriptive observations closely tracked the relative reinforcement each produced. Practical implications:
- Make the appropriate alternative richer, faster, higher in quality and easier.
- Reduce reinforcement for problem behavior. Full extinction works best, but even partial reductions shift allocation.
- This is part of why DRA can help when extinction cannot be fully carried out (G.2).
Seeing it in practice
| Reinforcer split (A/B) | Response split (A/B) | Description |
|---|---|---|
| 75/25 | 75/25 | Matching |
| 75/25 | 60/40 | Undermatching |
| 75/25 | 90/10 | Overmatching |
| 50/50 | 80/20 | Bias toward A |
In a classroom, Sofia’s teacher responds to her hand-raising about 12 times an hour and to her call-outs about 4 times. Matching predicts that about three of every four of Sofia’s bids for attention will be hand-raises. If the teacher stops responding to call-outs and answers hand-raising 16 times an hour, nearly all of her bids should shift to hand-raising.
Common mix-ups
- Matching does not mean equal responding. It means responding in proportion to reinforcement.
- Undermatching vs. overmatching. Under is closer to 50/50 than the reinforcement split. Over is more extreme than it.
- Bias vs. overmatching. Bias favors the same option no matter which is richer, and it shows up even when reinforcement is equal. Overmatching exaggerates whichever option is richer.
- Matching is about choice. It describes concurrent alternatives. Herrnstein’s extension treats “everything else” as the alternative when only one response is measured.
- Extinction is not the only lever. Changing the relative reinforcement shifts allocation, though extinction plus a rich alternative is the strongest combination.
Exam tip: Convert everything to proportions. Compare the response proportion with the reinforcer proportion. Equal is matching. Closer to 50/50 is undermatching. More extreme is overmatching. A preference that persists when reinforcement is equal is bias.
Quick recap
- The matching law: relative responding matches relative reinforcement across concurrent alternatives.
- The generalized matching law adds sensitivity and bias.
- Undermatching is less extreme than the reinforcement split; overmatching is more extreme.
- Bias is a preference that reinforcement rates do not explain.
- To reduce problem behavior, make the alternatives relatively richer as well as reducing reinforcement for the problem behavior.
Further reading
- Herrnstein, R. J. (1961). Relative and absolute strength of response as a function of frequency of reinforcement. Journal of the Experimental Analysis of Behavior, 4(3), 267-272.
- Baum, W. M. (1974). On two types of deviation from the matching law: Bias and undermatching. Journal of the Experimental Analysis of Behavior, 22(1), 231-242.
- Herrnstein, R. J. (1970). On the law of effect. Journal of the Experimental Analysis of Behavior, 13(2), 243-266.
- Borrero, J. C., & Vollmer, T. R. (2002). An application of the matching law to severe problem behavior. Journal of Applied Behavior Analysis, 35(1), 13-27.
Key terms
- Matching law
- Herrnstein's (1961) finding that, when two or more alternatives are available at the same time, the relative rate of responding to each alternative tends to match the relative rate of reinforcement it produces.
- Generalized matching law
- Baum's (1974) form of the matching law, which describes the ratio of responses on two alternatives as a power function of the ratio of reinforcers. Its two parameters are sensitivity, which captures undermatching and overmatching, and bias, which captures preference that reinforcement rates do not explain.
- Concurrent schedule (conc)
- A compound schedule in which two or more schedules operate at the same time and independently, each for a different response or option. Concurrent schedules arrange choice, and the way responding is divided among the options is described by the matching law.
- Undermatching
- A deviation from matching in which the division of responses across alternatives is less extreme than the division of reinforcement, closer to indifference. In the generalized matching law, sensitivity is less than 1.
- Overmatching
- A deviation from matching in which the division of responses across alternatives is more extreme than the division of reinforcement, favoring the richer option even more than matching predicts. In the generalized matching law, sensitivity is greater than 1.
- Response bias (Bias)
- In the generalized matching law, a consistent preference for one alternative that the relative rates of reinforcement do not explain, often due to differences in effort, reinforcer quality or location.
Check yourself
4 questions on B.23. Choose an answer to see whether you're right and why.
Show the answer
Answer: A. Matching predicts that the proportion of responding to an alternative equals the proportion of reinforcement it produces. Joking earns 45 of 60 reinforcers per hour, or 75%. Choosing 50% assumes equal responding regardless of reinforcement, which is not what matching describes.
Show the answer
Answer: A. The response split (65/35) is less extreme than the reinforcement split (80/20), closer to indifference. That is undermatching, the most common deviation from matching. Overmatching would be a response split more extreme than 80/20, such as 90/10.
Show the answer
Answer: A. With equal reinforcement, matching predicts a 50/50 split. A consistent preference for one option that reinforcement rate does not explain, here caused by lower effort, is bias. Overmatching exaggerates whichever option is richer, and neither option is richer here.
Show the answer
Answer: C. Matching says allocation depends on relative reinforcement. Making the appropriate alternative richer shifts responding toward it even if disruptive comments still sometimes produce attention. The other options make the disruptive alternative relatively richer or the appropriate one relatively poorer.
Adds a step to your study record. Saved in this browser only.