Wednesday, June 4th, 2008

Birdbrain looks back at Alex — the gray parrot subject of the Avian Learning Experiment — and what Irene Pepperberg was able to teach him:

As everyone knows, parrots are remarkably good at mimicking human speech, but they tend to repeat randomly picked-up phrases: obscenities, election slogans, “Hey, sailor.” Many parrots kept as pets also imitate familiar sounds, like the family dog barking or an alarm clock beeping. But Pepperberg taught Alex referential speech — labels for objects, and phrases like “Wanna go back.” By the end, he knew about fifty words for objects. Pepperberg was never particularly interested in teaching Alex language for its own sake; rather, she was interested in what language could reveal about the workings of his mind. In learning to speak, Alex showed Pepperberg that he understood categories like same and different, bigger and smaller. He could count and recognize Arabic numerals up to six. He could identify objects by their color, shape (“three-corner,” “four-corner,” and so on, up to “six-corner”), and material: when Pepperberg held up, say, a pompom or a wooden block, he could answer “Wool” or “Wood,” correctly, about eighty per cent of the time. Holding up a yellow key and a green key of the same size, Pepperberg might ask Alex to identify a difference between them, and he’d say, “Color.” When she held up two keys and asked, “Which is bigger?,” he could identify the larger one by naming its color. Looking at a collection of objects that he hadn’t seen before, Alex could reliably answer a two-tiered question like “How many blue blocks?” — a tricky task for toddlers. He even seemed to develop an understanding of absence, something akin to the concept of zero. If asked what the difference was between two identical blue keys, Alex learned to reply, “None.” (He pronounced it “nuh.”)

Pepperberg also reported that, outside training sessions, Alex sometimes played with the sounds he had learned, venturing new words. After he learned “gray,” he came up with “grain” on his own, and after learning “talk” he tried out “chalk.” His trainers then gave him the item that he had inadvertently named, and it eventually entered his vocabulary. (When Alex devised nonsense words — like “cheenut” — Pepperberg and his other trainers did not respond, and he quickly stopped saying them.) In linguistic terms, Alex was recombining phonemes, the building blocks of speech. Stephen Anderson, a Yale linguist who has written about animal communication, considers this behavior “apparent evidence that Alex did actually regard at least some of his words as made up of individual recombinable pieces, though it’s hard to say without more evidence. This is something that seems well beyond any ape-language experiments, or anything we see in nature.”

Pepperberg told me that Alex also made spontaneous remarks that were oddly appropriate. Once, when she rushed in the lab door, obviously harried, Alex said, “Calm down” — a phrase she had sometimes used with him. “What’s your problem?” he sometimes demanded of a flustered trainer. When training sessions dragged on, Alex would say, “Wanna go back” — to his cage. More creatively, he’d sometimes announce, “I’m gonna go away now,” and either turn his back to the person working with him or sidle as far away as he could get on his perch. “Say better,” he chided the younger parrots that Pepperberg began training along with him. “You be good, see you tomorrow, I love you,” he’d say when she left the lab each evening. This was endearing — and the Times’ obituary made much of the fact that these were the bird’s last words — although, as Anderson points out, it was during such moments that Alex was, most likely, merely “parroting.” It helped Alex’s charisma quotient that he made all his remarks in an intonation that was part two-year-old, part Rain Man, part pull-string toy. His voice, at once tinny and sweet, was easy to understand. Pepperberg tended to speak to Alex in the singsong “motherese” that doting parents use with young children, and he replied in a voice that seemed to convey a toddlerish pride.

Irene Pepperberg got an unusual start as an animal-language researcher in the early 1970s, since she was already in a chemistry grad program:

When Irene Pepperberg went to New York for the Clever Hans conference, she was thirty-one, and had owned Alex for three years. She had arrived in the world of animal communication from “out of left field,” as Diana Reiss puts it. Pepperberg has a Ph.D. from Harvard in theoretical chemistry, not psychology or zoology. But in the midst of her thesis work, which involved modelling chemical-reaction rates, it suddenly hit her, she recalls, that “(a) we don’t know enough at this point to do this exactly right and (b) in the future, what it’s taking me seven years to do with a mathematical model is going to take a computer hours, or seconds.” She decided to pursue something different. In any case, the prospects for women in her field hadn’t been encouraging. Speaking of her class at Harvard, she recalled, “My year was the first year that graduate-school draft deferrals for men were cut way back. So they let in a lot of women for a change. But the women were asked in their job interviews things like ‘What kind of birth control are you using?’”
Despite her graduate-school epiphany at Harvard, she continued with her Ph.D. in theoretical chemistry, receiving her degree in 1976; but she also started attending courses in departments relevant to the bird research she now hoped to do. “I was spending forty hours a week learning psychology and biology and forty hours finishing my doctorate,” she recalls.

The key to teaching parrots was eschewing Behaviorist methods — you don’t put a toddler in a Skinner Box, after all — and playing to the gray parrot’s social nature:

Pepperberg needed a method for teaching a parrot that played to its particular strengths. She came across something called the model/rival technique, which a German ethologist named Dietmar Todt had tried in a 1975 study of parrots. Todt had reasoned that, since parrots learn to squawk by watching each other vocalize, they might be able to learn German by observing people talk. So he developed a system in which one person was the trainer and one was the model for the bird — and its rival for the trainer’s attention. Pepperberg tweaked the protocol: in her version, the model/rival and the trainer periodically exchanged roles, so the bird could see that one person wasn’t always in charge. Parrots started the process by learning referential labels for things they wanted, rather than dialogues of the “Hello, how are you? I am fine” variety, which, Pepperberg figured, didn’t mean much to a parrot. There were no extrinsic rewards. If the parrot named an object, he’d get to play with that object, and, if he didn’t want it, he got the right to ask for something else. Pepperberg explained, “Let’s say you’re the model/rival and I’m the trainer. I have this object that the bird wants, and I show it to you and I say” — she adopted a singsong voice — “ ‘What’s this?,’ and you say, ‘Cork.’ I say, ‘That’s right,’ and you say, ‘Cork, cork, cork,’ while you’re holding it and the bird is practically falling off the perch because he wants it. And he hears that this weird noise is what mediated the transfer of this object. So we change roles, and then, instead of saying ‘Cork,’ I go, ‘Raaaawkk,’” — an uncannily accurate screech — “and you go, ‘No, no, you’re wrong,’ so the bird sees that not just any weird noise transfers the cork.”

The system worked. At first, a parrot might make a sound more like “erk” than “cork.” He’d need practice. Certain sounds are nearly impossible to produce without lips — Alex was never able to say “purple,” for instance, even after he nailed all his other colors. Still, as Reiss says, “Irene really found the appropriate method based on what we know about these birds. If you can tap into what these birds do in their own environment — in this case, the way these birds pair-bond — then you can set up a powerful learning paradigm.”

One of the odd things about watching a gray parrot talk, as this video demonstrates, is the contrast between its face — “goggle-eyed and masklike, and much less expressive than a dog’s” — and its ability to make its wishes clear:

Leave a Reply