Effects of reinforcement rate and reinforcer magnitude on choice behavior of humans.
Humans under-match by about 15-20%, so boost reinforcement on the desired option.
01Research in Context
What this study did
Thomas (1974) asked a simple question. Do humans follow the matching law like pigeons?
Adults picked between two buttons. Each button paid points on its own timer. The team changed how fast the points came and how big the prizes were.
They watched which button the adults chose.
What they found
People did not match. They picked the richer side 15-20% less than the math said they would.
Changing prize size or speed did not fix the gap.
How this fits with other research
Davis et al. (1972) and Killeen (1969) showed pigeons almost perfectly match. The birds’ data made the law famous.
Thomas (1974) used the same timers and counters, but humans broke the rule. This is an apparent contradiction. The birds matched; people did not.
Later work kept the story going. Innis (1978) switched pigeons to random-interval timers and still saw near-matching. The animal rule stayed strong, so the human gap is real.
Why it matters
If you build token boards, choice programs, or concurrent schedules for kids or staff, do not copy pigeon numbers. Expect people to under-match. Start with 15-20% more reinforcement on the target side, then watch and tweak. This keeps your programs aligned with how humans actually choose.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Add 20% more tokens or praise to the side you want picked more often.
02At a glance
03Original abstract
Two experiments with human subjects investigated the effects of rate of reinforcement and reinforcer magnitude upon choice. In Experiment 1, each of five subjects responded on four concurrent variable-interval schedules. In contrast to previous studies using non-human organisms, relative response rate did not closely match relative rate of reinforcement. Discrepancies ranged from 0.03 to 0.43 (mean equal to 0.19). Similar discrepancies were found between relative amount of time spent responding on each schedule and the corresponding relative rates of reinforcement. In Experiment 2, in which reinforcer magnitude was varied for each of five subjects, similar discrepancies ranging from 0.05 to 0.50 (mean equal to 0.21), were found between relative response rate and relative proportion of reinforcers received. In both experiments, changeover rates were lower on the long-interval concurrent schedules than on the short-interval ones. The results suggest that simple application of previous generalizations regarding the effects of reinforcement rate and reinforcer magnitude on choice for variable-interval schedules does not accurately describe human behavior in a simple laboratory situation.
Journal of the experimental analysis of behavior, 1974 · doi:10.1901/jeab.1974.21-409