The ability to abstract a grammar rule that underlies strings of seemingly meaningless sounds is an important aspect of language processing. Rule abstraction allows humans to learn about regularities in their native language and to generalize these regularities to novel input. This ability is a core mechanism underlying  the language faculty. But is it specific to humans and language learning? Up until now, there has been no conclusive evidence of  similar rule abstraction abilities in non-human animals. In this study we test both songbirds (zebra finches) as well as parrots (budgerigars) on their rule learning abilities.
Subjects were trained to discriminate between sound sequences of elements that followed either an XYX or an XXY structure. After this discrimination was acquired, each subject received a number of test sounds (intermixed with the training sounds) that followed the same structural rules, but consisted of new elements or combinations of known elements.
Results show that the budgerigars generalized their discrimination to sequences consisting of new sounds, so had been able to form an abstract representation of the structure of the training sequences. Zebra finches followed a different strategy. Their responses indicate a positional learning mechanism, where the sound sequences with known elements are categorized in accordance with the positional similarity to the training sequences.
This implicates that although the ability to abstract and generalize structural rules is not readily found in the animal kingdom, there are species that share this mechanism with humans.