Entropy Happens
Our original work was based on a hunch: namely, that pigeons’ mastering a same-different concept might be promoted by enriching the available information. Why show the birds just 2 icons that could be the same or different when we could show them 16? It seems like a simple move, but we suspected that it might have more profound theoretical importance if it proved successful.
So, we trained pigeons to peck one button when they saw a 4 x 4 array comprising 16 copies of the same computer icon and to peck a second button when they saw a 4 x 4 array comprising 16 different computer icons (see some examples on the left). The pigeons readily learned the task, and later transferred their learning to new sets of items.
Encouraged by these results, we explored the effect of the number of icons in an array as well as the effect of various mixtures of same and different icons in an array. Pigeons tended to respond “same” when the number of different items was small. Also, pigeons’ “different” reports continuously rose as mixture arrays changed from all same to all different. These results indicated that, rather than performing a categorical (same vs. different) discrimination, pigeons were actually discriminating array variability along a continuum.
Specifically, it seemed that pigeons were solving the same-different categorization by using entropy. Entropy is a measure of variability provided by information theory (Shannon & Weaver, 1949); it measures the amount of informational diversity by computing a weighted average of the number of bits of information in a set of items.
Identifying the neural systems that accomplish the computation of entropy and elucidating how experience and language build on entropy during human development represent key avenues for future investigation.
Wasserman, E. A., Young, M. E., & Castro, L. (2021). Mechanisms of same–different conceptualization: entropy happens! Current Opinion in Behavioral Sciences, 37, 19-28. https://doi.org/10.1016/j.cobeha.2020.06.001
Two-Item Same–Different Discrimination
Research on same–different categorization has shown that mastery of tasks of this kind can be strongly affected by the number of items in the training arrays—for both humans and nonhuman animals. Evidence for two-item same–different categorization in pigeons is decidedly mixed: although some investigations have succeeded, others have failed.
To provide a more convincing resolution of this story, we implemented a series of methodological techniques. First, we used trial-unique stimuli. This manipulation minimized the possibility that pigeons could base their responding on memory of earlier items. Second, we required our pigeons to perform a conditional discrimination; each pigeon had to peck either the same-item pair or the different-item pair depending on the color of each trial background (see examples on the left). Conditional discriminations are substantially more demanding because they require pigeons to attend to the features that discriminate both same and different concepts, rather than just the features that are characteristic of only one concept. Finally, we sought to quantify our pigeons’ same–different categorization behavior by using stimuli that varied along a continuous dimension: color.
All of our pigeons were able to learn the two-item conditional same–different discrimination task to very high accuracy levels. Furthermore, the birds were strikingly sensitive to the discriminability of the items in the different-item pair, exhibiting a systematic improvement in choice accuracy as color disparity increased. Trials on which the colors in the different-item pair were very similar to one another resulted in chance-level responding, whereas trials on which the colors in the different-item pair were very dissimilar from one another supported extremely high (90% correct) accuracy.
We suspect that our pigeons’ successful learning was promoted by our deploying a trial-unique procedure that avoids stimulus memorization. In this way, we demonstrated that stimulus-repetition is not necessary for pigeons’ successful learning in a two-item same–different paradigm.
Diaz, F., O'Donoghue, E. M., & Wasserman, E. A. (2021). Two-item conditional same–different categorization in pigeons: Finding differences. Journal of Experimental Psychology: Animal Learning and Cognition, 47(4), 455–463. https://doi.org/10.1037/xan0000297