Generalization gradient shape and summation in steady-state tests.
Generalization curves follow a simple summing rule that works for birds, people, colors, and tones.
01Research in Context
What this study did
The team worked with pigeons in a small lab chamber.
Birds pecked a disk for food when certain colors of light appeared.
After training, the researchers tested many nearby colors to map how responding spread.
They kept each test going until pecking stayed flat for many minutes.
What they found
The birds made smooth, hill-shaped curves around the trained color.
When two trained colors sat close together, the curves added up in the middle.
The shape matched a simple math rule, like stacking two small hills into one bigger hill.
How this fits with other research
Hoffman et al. (1966) saw the same stacking effect with tones and mild shock.
Their pigeons also showed two peaks that merged, giving early proof the rule works across setups.
Okouchi (2003) moved the test to college students looking at line lengths.
People gave the same lopsided hills, showing the pattern is not just for birds.
HOFFMAN et al. (1963) showed that these hills can last years but fade with repeated tests.
Together the four papers say: gradients form fast, add up predictably, survive a long time, and hold across species.
Why it matters
You can plan stimulus sets knowing that nearby cues will blend.
If you teach two similar colors, responses will grow in between without extra work.
Pick spacing that lets the hills overlap for smooth transfer, or separate targets to keep them distinct.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Place two taught colors 20 nm apart on a color tablet and probe the shade in between; expect extra correct responses without further training.
02At a glance
03Original abstract
Pigeons' pecks at one or two wavelengths were reinforced intermittently. Random series of adjacent wavelengths appeared without reinforcement. Gradients of responding around the reinforced wavelengths were allowed to stabilize over a number of sessions. The single (one reinforced stimulus) and summation (two reinforced stimuli) gradients were consistent with a statistical decision account of the generalization process.
Journal of the experimental analysis of behavior, 1969 · doi:10.1901/jeab.1969.12-91