Choice, matching, and human behavior: A review of the literature.

Pierce et al. (1983) read every matching-law paper on humans they could find. They pulled together lab studies where people picked between two buttons, levers, or computer keys. Each option paid off on its own timer or counter. The team then asked: do people’s response rates line up with the payoff rates?

No new lab work was run. The paper is a narrative review, so it sums up what others already showed.

Across studies, people behaved like pigeons. When the left key gave twice as many points, users pressed it twice as often. The review says the matching law “is supported” for human choice on concurrent schedules.

Macdonald et al. (1973) showed the same pattern first, but with birds. Their pigeon data is the base the human review stands on.

Davison et al. (1984) and Hinson (1988) later found limits. When reinforcer length keeps changing, or when both rate and length differ, people stop matching. These null results don’t kill the law; they mark where it frays.

Hall (2005) sharpened the rule: if the two jobs pay off at different speeds, add “earning rate” to your equation. The 1983 story stays true for simple VI-VI cases, but later papers give you fixes for messier ones.

If you run concurrent schedules in skill labs or token economies, expect people to drift toward the richer side. Use the simple matching check: count responses and reinforcers. If numbers don’t line up, look at reinforcer size or earning speed, not just rate. Update your equations and keep the choice fair.