The quantal nature of controlling stimulus-response relations as measured in tests of stimulus generalization.
Score generalization probes as all-or-none quantal jumps, not gradual shades.
01Research in Context
What this study did
Crossman et al. (1985) read every lab paper on stimulus generalization they could find.
They asked: do the data look smooth, or do they jump from one level to the next?
They treated each stimulus-response pair as an on-off switch, not a dial.
What they found
Most curves looked like steps, not ramps.
A pigeon either pecked at the new light or it did not; there was little in-between.
The authors say this quantal view fits the numbers better than older curved models.
How this fits with other research
HANSON et al. (1961) drew smooth gradients after giving pigeons pipradrol.
K et al. say those same data show clear steps once you count responses the quantal way.
Rosenthal et al. (1980) talked about generalization in kids’ naming.
K et al. add that each name is either learned or not; there is no half-name.
Sutphin et al. (1998) later used the same review style for habit reversal, showing the quantal idea spread to other ABA topics.
Why it matters
When you probe for generalization, think in yes-no bins.
If a client sorts red and orange blocks together, record “same” or “different,” then move on.
This small shift gives cleaner data and faster decisions.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Record each probe response as 1 (correct) or 0 (error) and graph the step function.
02At a glance
03Original abstract
This paper is a selective review of research that addresses the validity of two interpretations of stimulus generalization. One interpretation, referred to as the descriptive stimulus-control interpretation, proposes that during stimulus generalization a continuous relation exists between stimulus and response dimensions. The other interpretation, referred to as the quantal interpretation, proposes that a stimulus-response relation functions as a unit that may or may not occur. From the latter viewpoint, the continuity typically obtained during generalization tests is deemed to be artifactual and to result from averaging across multiple controlling stimulus-response relations. Studies examining the contribution of these multiple relations to generalization gradients are reviewed. With few exceptions, the quantal interpretation appears to better characterize the results of these studies. Implications for peak shift, selection of analytical level, and identification of the behavioral unit are discussed, as well as factors that may determine the acceptability of the quantal interpretation.
Journal of the experimental analysis of behavior, 1985 · doi:10.1901/jeab.1985.44-245