Grey Parrot Visual Working Memory Manipulation
In a previous blog, I discussed my parrots’ long-term memory. Could they, for example, remember a student who had been gone for several years? However, another type of memory — what scientists call “visual working memory,” or VWM, is also of interest. It is a very important ability that enables living creatures to survive in their world. For example, VWM is what allows a prey animal (including our human ancestors) to remember that just because a predator has ducked behind a bush, it probably hasn’t disappeared!
VWM similarly functions for predators, who know to keep searching for prey that seems to have vanished. What I’ve just described is only one aspect of VWM, and is formally known as “object permanence.” It involves storage of information and the formation of some kind of mental representation of objects that are no longer directly in view. As it turns out, most humans and a large number of nonhuman species can remember the placement of up to four hidden items at any one time; add a fifth, and accuracy drops. That number is what is known as “storage capacity.”
Recently, I’ve collaborated with colleagues to study another aspect of VWM, involving the ability to update these memories based on additional information. This ability is called VWM manipulation, or VWMM, because we are manipulating the memories in our mind. For example, if we know that a path exists, leading from the bush behind which the predator has ducked, we can imagine the predator moving along that path and thus know to avoid all the other bushes alongside it. The problem for scientists has been finding an appropriate way to study VWMM capacity in order to compare it to storage capacity.
Parrots Win the “Shell Game”
One of my colleagues, Hrag Pailian, came up with a brilliant idea based on the classic “shell game”—the one where a carnival person hides a bean under one of four cups, moves the cups around really quickly, and asks you to bet on where you will find the bean. You usually lose because the game is rigged, but even if it isn’t, the task is rather difficult. Not impossible, but difficult because you must track the bean through many different movements.
Now, imagine if instead of one bean, you had to track four differently colored pompons through multiple moves. Only after all the moves are completed are you asked to “find the yellow one.” That task is really difficult, although, as you will see, not impossible.
Using that exact task, we tested adult humans, 6- to 8-year-old children, and our African grey parrot, Griffin (Pailian et al., 2020). We chose the age range of the children for two reasons. First, we knew from other studies that their storage capacity for four items is very close to that of adults. Second, we also knew that on several tasks (Piagetian liquid conservation and probability, inference by exclusion), Griffin performed at the level of 6- to 8-year-olds. Thus, we could compare Griffin to humans on a developmental scale, as well as see how a “bird brain” compared to that of humans overall—something we have been doing for quite some time!
We also specifically wanted to examine how our subjects fared as the task’s difficulty increased. Thus, we started with two pompons that we hid under two cups, and tested how well the adults, children, and Griffin did when there were zero, one, two, three, or four moves. Unfortunately, the children stopped paying attention after three moves, so we had to stop there for all their tasks. We then increased the number of pompons and cups to three and then four, again looking at the different numbers of moves, zero to three for the children and zero to four for Griffin and the adults. We also gave Griffin a set of five pompons and five cups with no moves to see how he compared with the other species on storage capacity—and he was 100% on those trials, besting most humans!
We were not very surprised when we found that all of our subjects were pretty much at 100% for two cups and up to four swaps. All that is necessary is to track one pompon and, as there are only two possible places, even if the subject is tracking, say, the blue one and is asked about the white one, all that subject has to do is point to the cup that it knows doesn’t have the blue one.
The data were a lot more interesting once we got to three pompons. Again, as expected, everyone was correct for zero moves, but then the results changed dramatically, particularly for the children. With just one move, the children dropped to 80% correct, and for two and three moves, plateaued at 70% correct. That’s still above chance, but quite a decrease. The adult humans also dropped in accuracy with more moves, but only by a few percent each time as we added one, two and three moves, and then had a greater drop, though they were still at 80%, for four. The big surprise was Griffin, who was still 100% at four moves. [He did make one mistake at two moves, which we think was just a lapse of attention, because he improved at three and four.]
When we got to four pompons, the results were again extremely interesting. After just one move, children dropped to 70%, and then plateaued at 50% for two and three moves. That level of accuracy was still well above chance, but clearly the children found the task difficult. Adult humans had a fairly steady, slow drop for each additional move and ended a bit above 70%. Griffin’s results were distinctly different: He stayed at 100% for one move (and thus was above adults), matched adult humans at two moves (~75%), fell below the adults at three moves (60%) and then dropped even further at four moves (50%). At first, we thought that Griffin was just bored with the task…as we got toward the end, we had more mistrials in which he stopped attending before we had finished the moves. So we gave him an almost a year-long break, and then retested him on the four pompon/four move task, but this time paying close attention to the actual types of moves that were performed. Just because four moves were made didn’t meant that all four objects always moved—for example, sometimes only two objects moved multiple times and the question could even be about one that did not. As it turned out, Griffin’s accuracy decreased as the number of moves in which the target was involved increased. Thus, it was clear that tracking multiple moves of any one pompon was difficult.
Nevertheless, the overall result was that VWMM is not a uniquely human ability, and one that does seem to improve in humans with age. That nonhumans need this capacity makes sense—think about tracking fledgling birds to ensure that all are fed even though they are in constant motion, or trying to chase down one antelope in a moving herd. Of course, we need to find out about the abilities of other avian species: We now need to test other birds like crows that cache food in multiple places and have really good memories overall, and also parrots that do not have color labels—we could not be sure that Griffin wasn’t mentally rehearsing the order of the moves by using color labels (e.g., thinking “red, green, yellow, blue; red, green, blue, yellow; green, red, blue yellow”…and so forth). In any case, we have shown that, yet again, Grey parrots are at least as intelligent as 6-8 year old children, and on certain parts of the task, can outperform even adult humans!
Pailin, H., Carey, S., Halberta, J, Feigenson, L., & Pepperberg, I.M. (2020). Age and species comparisons of visual mental manipulation ability as evidence for its development and evolution. Special issue of Scientific Reports,10, Article number: 7689, doi:10.1038/s41598-020-64666-1