Reinforcer-ratio variation and its effects on rate of adaptation.
Wider reinforcer-ratio ranges speed up preference shifts in pigeons — apply the same logic to token boards and praise schedules.
01Research in Context
What this study did
Researchers watched pigeons choose between two keys. The team changed how often food followed each key peck. They made the payoff difference small some days and large other days.
The birds worked in daily sessions. Scientists tracked how fast the pigeons switched their preference when the odds changed.
What they found
Big payoff gaps made the birds change faster. Small gaps made them change slower. The range of reinforcer ratios, not just the size, controlled adaptation speed.
Pigeons showed two layers of control. Local reinforcers nudged each peck. Long-term patterns shaped overall preference shifts.
How this fits with other research
Hawkes et al. (1974) first showed orderly preference curves in pigeons. Weiss et al. (2001) adds that the speed of those curves depends on ratio range.
LeBlanc et al. (2003) extends the idea to stimulus control. Higher rates not only shift preference faster; they also make discrimination more resistant to disruption.
Thrailkill et al. (2018) sounds contradictory at first. They found richer schedules cause bigger relapse after extinction. The studies differ in time frame. J et al. looked at minutes-to-hours adaptation; Thrailkill tracked days-later recovery. Fast adaptation does not guarantee long-term stability.
Why it matters
When you adjust token or praise ratios, think in ranges, not just amounts. A wide ratio jump (5:1 vs 1:1) will produce faster skill shifts. A narrow jump (2:1 vs 1:1) gives you slower, smoother change. Plan the range to match your client's tolerance and your session length.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Try a 4:1 token ratio for a tough skill and a 2:1 ratio for an easy skill; track which one produces faster acquisition.
02At a glance
03Original abstract
Six pigeons were trained in sessions that consisted of six or seven concurrent-schedule components, each of which could have a different reinforcer ratio arranged in it. The components were unsignaled and occurred in a random order separated by 10-s blackouts. The overall reinforcer rate arranged in each component was 2.22 reinforcers per minute. In Experiment 1, the range of reinforcer ratios in the seven components was varied from a condition in which the ratios were always 1:1, to a condition in which the ratios varied between concurrent variable-interval 27 s extinction (EXT) and concurrent extinction variable-interval 27 s (ratios of 1:EXT, 9:1, 3:1, 1:1, 1:3, 1:9, EXT:1). In Experiment 2, the range of reinforcer ratios was always 27:1 to 1:27, and the presence and absence of the intermediate reinforcer ratios used in Experiment 1 (9:1, 3:1, 1:1, 1:3, 1:9) were investigated. Log response-allocation ratios in components changed rapidly with increasing numbers of reinforcers in components, and Experiment 1 showed that sensitivity to reinforcement was usually higher when the range of reinforcer ratios was greater. When the range of reinforcer ratios was kept constant in Experiment 2, the presence or absence of less extreme reinforcer ratios had no clear effect on sensitivity. At a local level, individual reinforcers had predictable quantitative effects on response ratios: Successive same-alternative reinforcers in a component had rapidly diminishing effects in both experiments. Reinforcers obtained on the opposite alternative to one or more prior reinforcers always had large effects on preference, and these changes were greater when the range of reinforcer ratios was greater. The effects of such reinforcers in changing preference were enhanced, and produced clear preference reversals, when intermediate reinforcer ratios were absent in Experiment 2. Two processes, one local to reinforcers and one with a longer time course, may be necessary to account for these results.
Journal of the experimental analysis of behavior, 2001 · doi:10.1901/jeab.2001.75-207