This month I’m writing about a study that was done with Alex decades ago, but that has received some new interest because of recent studies with very young children. It involves how well individuals understand the concept of same-different. The topic is one that would seem, on the surface, pretty straightforward: If you are given two objects, it is pretty simple to state whether they look the same or if they look different from one another. The answer, however, turns out to be pretty complex, because many levels of understanding exist: At the highest level, the response actually involves recognizing degrees of similarity and difference…and not all nonhumans or very young children demonstrate that they understand the most complex levels.
The most basic level of same-different involves identity versus non-identity. You see that the two objects look absolutely the same, or you see that they aren’t absolutely the same—but you don’t have to recognize exactly what is different. Almost every nonhuman tested succeeds on some version of this test, known as “match-to-sample” (or MTS) and “nonmatch-to-sample” (or NMTS). In one common version, the subject is shown a sample item, A, and then A and B together.
Match-to-Sample vs. Nonmatch-to-Sample
If the task is MTS, and the subject picks A, it gets a reward. Even though the task sounds simple, it can take subjects a few hundred trials to figure out what they are supposed to do. When a subject reaches a certain level of correctness, such as 80 of 100 trials in a given day, it is then given C as the sample, and C and D as the choices. If that subject takes fewer trials to learn to pick C than it took to learn to pick A, it is said to have learned MTS.
Then the subject gets what is known as a reversal trial. It goes back to the A and B choice, but now only gets rewarded when it hits B…NMTS. When the subject figures out that task, it is given C, with a choice of C and D, and is expected to choose D. The problem with these tasks, discovered by some of my colleagues (e.g., Premack, 1983; Zentall & Hogan, 1974), is that subjects may not actually be learning the concept of same-different, but rather learning only ‘same’ in MTS and then learning to avoid ‘same’ in NMTS, or learning to respond on the basis of familiarity—that is, they’ve seen A before, keep it in memory and then just choose what is familiar in MTS and avoid it in NMTS. Recent studies on young children showed they also seemed to have problems with difference, but not sameness (Hochmann et al., 2018), which researchers found a bit surprising.
Premack (1983), after analyzing many such studies with nonhumans involving variations on the task described above, came up with a novel analysis. He argued that full comprehension of same/different actually involves the ability to represent very specific relationships of sameness and difference among sets of objects: the ability to recognize not only that two independent objects—e.g., A1/A2—share or differ in just one attribute, e.g., the color blue, but also to recognize that only this attribute—in this case, the category color—is what is targeted (i.e., important) in a particular trial, and that the targeted attribute depends upon whether the task involves sameness or difference, because stimuli can simultaneously share some attributes and differ in others.
Subjects must also recognize that this sameness/difference concept can be immediately extrapolated two other novel items—Bl/B2—having no commonality with the original As (e.g., the targeted attribute in that trial is shape). So the subject is not simply marking identity or non-identity, but knows exactly which attributes are identical or non-identical. Interestingly, Premack also proposed that such abilities were limited to primates and that only subjects that understand what he termed “symbolic representation”—where a non-iconic symbol stands for an object, attribute, action, etc.—could demonstrate such comprehension by using symbols for “same” and “different’ to label their presence or absence, or to use labels to indicate the appropriate attributes that were same or different (e.g., “color”, “shape”).
Spot the Difference
To put it another way, to be correct, a subject had to (a) attend to multiple aspects of two different objects; (b) determine, from the specific question posed, whether the response was to be based on sameness or difference (“What’s same?” vs. “What’s different?”); (c) determine, by looking at the two objects, what was same or different (i.e., what were their colors/shapes/materials?); and then (d) produce (either vocally or by manipulating some physical symbol), the label for the appropriate category (“color”, “shape”, “material”), not just the instance of the category (e.g., “blue”). Thus, the task required a feature analysis of the two objects, recognition that objects could simultaneously share similarities and differences, and the ability to understand which attributes were being targeted.
As it turns out, Alex was probably the only nonhuman (possibly with the exception of Premack’s symbol-trained Sarah) who could succeed (Pepperberg, 1987). If, for example, Alex was given a blue wood triangle and a blue wood square and asked, “What’s different?”, he replied “shape.” If asked “What’s same?”, he would say, “color” or “mah-mah” (his label for matter, or material). He even responded with equivalent accuracy for pairs of objects he had never before seen. He was therefore able to target, and label, the specific attributes that the two items shared or in which they differed.
In sum, full understanding of same-different is much more complex than it may first seem. Many species perceive identity; many can be trained to perceive non-identity or degrees of non-identity. Tasks used in those studies, however, do not enable subjects to demonstrate that they fully understand same-different at the level that Premack specified.
Hochmann, J-R, Carey, S., & Mehler, J. (2018). Infants learn a rule predicated on the relation same but fail to simultaneously learn a rule predicated on the relation different. Cognition, 177, 49-57.
Pepperberg, I.M. (1987). Acquisition of the same/different concept by an African Grey parrot (Psittacus erithacus): Learning with respect to categories of color, shape, and material. Animal Learning & Behavior, 15, 423‑432.
Premack, D. (1983). The codes of man and beasts. Behavavioral Brain Sciences, 6, 125-167.
Zentall, T.R., & Hogan, D.E. (1974)/ Abstract concept learning in the pigeon. Journal of Experimental Psychology, 102, 393-398.
You can help Dr. Pepperberg continue the groundbreaking parrot research she began more than 30 years with Alex, the African grey parrot that won admirers from around the world with his cognitive abilities. If you shop online through sites such as Amazon.com, you can designate the Alex Foundation to receive a percentage of your final sales, or register with the Alex Foundation at iGive.com and a percentage of sales from companies associated with iGive will go to the foundation. The Alex Foundation also has a “Donate” button linked to PayPal. Visit http://alexfoundation.org and click on the “Support Us” link for more information.