Parrots do probability


Study finds first evidence of statistical inference beyond great apes.


New Zealand’s smartest bird continues to prove its skills. 

Amalia Bastos

By Paul Biegler

A New Zealand-based PhD student has teamed up with a parrot called Blofeld and, together, they have raised the “spectre” that the birds are much smarter than we thought.

The parrots can, it appears, make complex mathematical judgments about probability, according to a study published in the journal Nature Communications.

Until now, only two members of the great apes clan – humans and chimpanzees – were thought to be able to detect the relative frequency of an object.

To understand that bit of maths minutiae, take an example from the confectionery domain.

Say you like black gummy bears and someone puts twenty in each of two glass jars. Then they add 100 orange bears – which you hate – to the first jar, and 4 orange bears to the second jar. And give both jars a good shake.

If you get a blind lucky dip which jar would you plunge your hand into?

Each jar has the same number of black bears, but odds on you’ll go for the second jar. It has the highest proportion or “relative frequency” of black bears, which ups the chances you’ll get your preferred sugar hit.

Doctoral candidate Amalia Bastos, from the School of Psychology at the University of Auckland, had a hunch that the quick-witted New Zealand Kea parrot (Nestor notabilis), known for its ability to make snowballs and raid wheelie bins, might have similar smarts.

To find out, she and supervisor Alex Taylor trained Blofeld and five other Keas to learn that black tokens meant a food reward and orange tokens meant no food.

Then they filled two jars with black and orange tokens in the same proportions as the gummy bear lucky dip. The birds watched as a token was plucked from each jar by an experimenter, who presented it to the bird in a closed fist.

The birds consistently pecked their preference for the fist that came from the jar with the highest proportion of black, despite each containing the same number of black tokens.

The Kea had, officially, risen above the status of rogue bin raider and claimed a prized spot with humans as a detector of relative frequency.

Bastos and Taylor weren’t finished, however, and wondered if the parrot could mirror other human traits.

They presented the birds with two jars but, this time, each had exactly the same proportion of black tokens. But there was a twist in the form of a barrier halfway up the jar, which meant only the top half was accessible to the experimenter.

In one of the jars black tokens were in a higher proportion above the barrier.

Kudos to the Kea, the birds chose more tokens from that jar. The parrots were integrating knowledge of physical structures to enhance their use of relative frequency.

This, say the researchers, is evidence of something called “domain-general intelligence”, a mix and match of different streams of intelligence that some scientists suspected was unique to humans.

Finally, the team went looking to see if the parrots had social smarts, something useful in their native environment.

“Kea have a complex social structure where many individuals can live in a group and they come and go as they please,” says Bastos.

“That means they need to remember the identities of multiple individuals.”

Bastos and Taylor told one of the samplers to pick out black tokens on purpose, even when it was in a minority black jar. The canny parrots were on to this quickly, developing a decided preference for the closed fist of the black afficionado.

This, the authors say, is evidence the birds use social information to override the dictates of relative probability, another example of domain-general intelligence.

Humans and birds last shared a common ancestor 312 million years ago which suggests, say the authors, that statistical inference evolved separately in the birds.

Looking ahead, the results could even influence artificial intelligence.

The brains of parrots are not just smaller than human brains but have a very different structure. The fact that bird brains can support human level reasoning raises questions about the degree to which AI should take cues from the physical structure of mammalian brains.

Indeed, AI researchers may want to keep an eye on a certain species of parrot.

“I can’t wait to see what Kea do next, because they keep surprising us,” says Bastos.

Contribs paulbiegler 2.jpg?ixlib=rails 2.1
Paul Biegler is a Eureka-Prize winning journalist, bioethicist and former physician writing on all things health and science.
  1. https://www.nature.com/articles/s41467-020-14695-1
  2. https://www.youtube.com/watch?v=n9iifaLV15Q
  3. https://www.youtube.com/watch?v=bxoCuRuHlt8
  4. https://pdfs.semanticscholar.org/c805/1bb66b1e2ddbc728e2287bb3caa00725b06a.pdf
Latest Stories
MoreMore Articles