Matching-to-sample Performance In Rats: A Case Of Mistaken Identity?
Teach conditional discriminations with moving positions early so the learner attends to the picture, not the spot.
01Research in Context
What this study did
The researcher taught rats to pick the picture that matched a sample picture. The sample and choices always sat in the same spots on the screen.
Next, he moved the pictures to new spots. He wanted to see if the rats still chose the correct picture when the places changed.
What they found
Accuracy crashed as soon as the sample moved. The rats had learned to pick a place, not a picture.
Spatial location had become part of the stimulus that controlled the response.
How this fits with other research
Collier et al. (1986) saw the same crash with pigeons. When line samples moved, accuracy fell. Hue samples stayed solid. Together, the two papers warn that position can steal control from the feature you meant to teach.
Catania et al. (1972) showed the set-up: restrict training to only a few layouts and you bake in position bias. Iversen (1997) proves the bias bites later when you finally shift the layout.
Bailey (1984) offers the fix. Children generalized matching after the experimenter stabilized their sample-coding responses. Combine lessons: vary positions early, or first build strong coding responses, then accuracy survives layout changes.
Why it matters
If you always flash the SD in the same corner of the table, the learner may bond to the corner, not the image. When you later probe generalization, performance can collapse. Rotate cards, tablets, or objects across trials from day one. Mix top, bottom, left, right so the relevant feature stays the picture, word, or object—not its address on the surface.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Shuffle your flashcards or tablet screen to a new corner every trial today.
02At a glance
03Original abstract
Three rats had previously acquired a simultaneous matching‐to‐sample performance with steady and blinking lights. In training, the sample stimulus had always appeared on the middle of three horizontally arranged keys with the comparison stimuli on the side keys. In Experiment 1, the sample stimulus appeared on any of the three keys with the comparison stimuli on the remaining two. The matching‐to‐sample performance broke down with variable sample and comparison locations; the sample stimulus did not control responding to the comparison stimuli when it appeared on a side key, but it retained control when it appeared on the middle key (as in training). In Experiment 2, the rats were trained with the sample always on the left key. When the sample appeared on either of the trained locations (left or middle key), it retained control for both locations. When the sample then appeared on any of the three keys, as in Experiment 1, sample control did not transfer to the untrained location (right key). The experiments demonstrate that training with fixed sample and comparison locations may establish spatial location as an additional controlling aspect of the stimuli displayed on the keys; stimulus location had become part of the definition of the controlling stimuli. The rats' performance seemed best described as specific discriminations involving the visual stimuli and their spatial locations rather than as identity matching.
Journal of the experimental analysis of behavior, 1997 · doi:10.1901/jeab.1997.68-27