Considerable data exist to demonstrate that Grey parrots use elements of English speech referentially (Pepperberg, 1999). The question has been raised, however, as to whether these birds can, like young children (e.g., de Boysson-Bardies, 1999; Greenfield, 1991; Marschark, Everhart, Martin, & West, 1987; Peperkamp, 2003; Tomasello, 2003; note Gillen, 2007), intentionally demonstrate creative or innovate use of such speech. Such examples may involve using or recombining existing utterances in novel ways to describe novel objects, engaging in sound play to create novel utterances that could relate to items in the environment, transferring of concepts across domains, and segmentationâthe ability to divide and recombine parts of existing labels to create novel utterances for novel situations. Such examples may involve novel ways of replicating models, but primarily entail demonstrating some level of newness. Although the data are not extensive, several examples exist to demonstrate that at least one Grey parrot was capable of several of these forms of creativity and innovation.
Sound Play: Combinations, Recombinations, Simple Extensions in the Presence of Trainers
The bird in question, Alex, had been trained via a modeling procedure (M/R training, Pepperberg, 1981; Todt, 1975) to identify several objects, colors, shapes, and common foods. The amount of practice and creativity involved in acquisition of his initial labels is unknown, as his solitary sound play was not taped until a decade later (Pepperberg, Brese, & Harris, 1991). Early on, however, he produced novel utterancesâutterances he had not heard from humansâin his trainersâ presence.
Some of these were quite simple combinations of existing labels, and were triggered by the presence of only slightly novel objects (Pepperberg, 1981). Thus, after being trained to use the labels âkey,â âwood,â and âhideâ to identify, respectively, a silvery key, uncolored tongue depressors and pieces of rawhide, and to combine these with the label âgreenâ to refer to their colored counterparts, he was shown a green clothes pin. His label for clothes pins was âpeg wood,â and he immediately stated âgreen wood, peg wood.â Similarly, after being trained on the label ârose hideâ for red hide, he transferred to ârose paperâ and then other red items without training. Green corks were also identified immediately, without training, as were blue hide and blue clothes pins after âblueâ was acquired. The same was true after other color labels were learned. After training to label triangular and square pieces of maple as â3-corner woodâ and â4-corner wood,â respectively, he promptly labeled their paper and rawhide counterparts, although he needed to chew the items first, to identify their material. He later similarly transferred other shape labels. After learning the label ârockâ by repeatedly querying his trainers about a lava-stone beak conditioner, he began to call the dried corn kernels in his food ârock corn,â presumably to differentiate it from the fresh, soft variety he was also fed. He used ârock nutâ to refer to a Brazil nut that he could not crack on his own, but his lack of sustained interest in the item led to the disappearance of the utterance. He labeled an unshelled almond âcork,â possibly because the shell looked like one, but accepted our term âcork nutâ to distinguish between the two objects (Pepperberg, 1990). Other examples of labels that were initially applied to objects of little interest and did not remain in his repertoire are discussed in Pepperberg (1990).
In a similar vein, other parrots in my laboratory creatively and spontaneously extended the use of labels trained with respect to specific exemplars to additional items. Thus my birds, somewhat like children (e.g., Brown, 1973), tested meanings of newly acquired labels; possibly they saw humans as a means of providing additional referential information (Pepperberg, 2002). For example, a younger subject, Arthur, in ways similar to Alex, uttered one of his newly acquired labels, used previously in a very specific context, in a novel situationâhere, âwool,â trained as a label for a woolen pompon, subsequently uttered while pulling at a trainerâs sweater. He seemed to be testing the situation, and our responsesââYes, WOOL!â given with high positive affectâstimulated him further, revealing the potential power of an utterance and encouraging his early categorization attempts. Notably, such behavior was not trained in any direct manner.
Creativity with other labels seemed to involve more complicated processes. At one point, we tried to train Alex to identify an apple; at the time, he knew âbanana,â âgrape,â and âcherryâ for the other fruits he was fed. After undergoing considerable training via our modeling procedure, Alex ignored the targeted label and said âBanerryâŚI want banerry.â He not only repeatedly used this term, but also slowed production and sharpened his elocution (âban-err-eeeâ), much as trainers do when introducing a new label (Pepperberg, 1990). No one had ever used the term in his presence, but it did make some semantic sense, apples tasting a bit like a banana and looking like a very large cherry.
Alex also often produced new vocalizations in what appeared to be vocal play by recombining other existing label parts in their corresponding orders (Pepperberg, 1990). His spontaneous novel phonemic combinations would occur outside of testing or training, appearing in contexts and forms reminiscent of childrenâs play (NB: other juvenile Greys behave similarly; reviewed in Pepperberg, 2004a; see also Pepperberg & Shive, 2001). These utterances were rarely if ever used by trainers, but sometimes resembled both existing labels and separate human vocalizations. Notably, these recombinations suggest that Alex, somewhat like children (see Weir, 1962; note discussion in Treiman, 1985), abstracted rules for utterancesâ beginnings and endings. In over 22,000 vocalizations, he never made backward combinations (e.g., never said âpercupâ instead of âcupper/copperâ; Pepperberg et al., 1991). When we referentially mapped these spontaneous utterancesâproviding relevant objects to which they could referâAlex rapidly integrated these labels into his repertoire. He acquired âgrey,â for example, after seeing himself in a mirror and querying âWhat color?â; he then produced sound variants (e.g., âgrape,â âgrate,â âgrain,â âchain,â âcaneâ) that we mapped to appropriate referents (respectively, fruit, a nutmeg grater, seeds, a paper-clip ring, sugar cane; see Pepperberg, 1990, 1999). With the exception of âgrape,â which trainers had consistently labeled, these labels had never been produced in his presence. He promptly began to use them consistently to refer to the items. In contrast, he abandoned sounds whose combinations we could not map (e.g., âshane,â âcheenutâ), or for which mapped referents were not of interest (e.g., the dried banana chips we used for âbanackerâ; Pepperberg, 1990). Thus, Alexâs spontaneous utterances that may initially have lacked communicative, symbolic value could, as they do for children, acquire this value if caretakers interpreted them as such (Pepperberg, 1990): that is, although his initial combinations could have simply been âbabble-luckâ (Thorndike, 1943), he behaved as if our interactions âconventionalizedâ both the sound patterns and sound-meaning connections in the direction of standard communication. Likely, after the first such instance, the excitement of the trainers and the introduction of a novel item may have impelled Alex to attempt more innovative vocal play.
Alex created very few action-related phrases, but the few he did form bear mentioning. To fashion most action-related variants, Alex excerpted part of a standard phrase, which he then used as a request; just as in referential mapping for labels, trainers responded as though he had made an intentional demand (Pepperberg, 1990). Thus, during a long training session, Alex uttered âgo chairââa portion of the (by then) commonplace phrase âwanna go chairâ (see the section âNovel Combinations in the Service of Vocal Learningâ). The trainer sat, temporarily discontinuing the session. Alex subsequently used this utterance to call a halt to whatever was happening. He also excerpted âgo away,â which he used to break contact with trainers. This phrase was closely related to two situations in which it was first routinely used in his presence. The first situation occurred when he refused to comply with our requests for object labeling: he routinely said âno,â turned or walked away, and began to preen. We responded with a brief time-out, accompanied by the phrase âIâm gonna go away!â He subsequently asked (âgo awayâ) that the trainer depart. The second situation involved scheduled departures of trainers (e.g., at lunch or the end of the day). At those times, we added âIâm gonna go eat lunch/dinner,â and placed Alex inside his cage. Interestingly, during long or difficult sessions, Alex began to request students to âgo away,â or occasionally stated âIâm gonna go away,â after which he climbed off his training chair. Generally, attempts to continue training after such a declaration were fruitless.