Do ravens plan for the future?

The answer to the title question is “apparently so, and at about the level of great apes and four-year-old human children”. That, at least, is the conclusion of Can Kabadayi and Mathias Osvath, the authors of a new paper in Science on raven behavior in the lab (reference and free download below; see also the Perspective summary in Science by Markus Boeckle and Nicky Clayton).

Ravens, like most corvids, are very smart. They’re known to cache food for later consumption; they know when another raven is watching them cache, and then re-hide their food in response; and they can follow the gaze of another raven looking at its cache, and then pilfer that cache later. These, however, aren’t regarded as ravens using new information to plan for the future, though it seems to come close to that!

What Kabadayi and Osvath did was to teach five captive, hand-reared ravens two new skills, not employed in nature, which involved using a tool or token—both of which could be used to obtain food. They then tested whether the ravens could remember how to use those items to get food after a delay of either 15 minutes or 17 hours, testing whether ravens could figure out how to plan for the future. Moreover, the authors asked whether ravens could delay gratification, foregoing a present treat to get a better one in the future.

The first experiment involved teaching ravens (who largely taught themselves) to use a pebble as a tool to open a device containing food. They were also given other “tools” that didn’t work properly. After learning, they were given the treat-containing apparatus on the next day without any tools, presumably so they could ponder its presence and cogitate.

Then the apparatus was removed and the ravens presented with an array of four tools, only one of which worked. They had to choose a tool. The apparatus with the food was then given to them after a delay of 15 minutes.

The experimenters did this again, but with a longer delay: 17 hours.

In both cases, the raven almost invariably selected the right tool and “cached” it for later use to get the food.

In the second experiment, the raven was given a token that it could give to the experimenter in return for food. After they learned to do this, the experimenter “asked” for the token when the birds didn’t have it; this was designed to provide an incentive. Finally, a tray was presented with the token and three useless items that could not be exchanged for food. (Actually, they were given three identical trays so they could choose and cache up to three food-redeeming tokens.)

Again, after either a 15-minute or 17-hour delay the experimenter showed up ready to deal. And again, the ravens selected the correct tokens over 95% of the time, even when they couldn’t be used for many hours.

Finally, the experimenters put a twist in the experiments by putting a treat in among the tools or tokens—a treat that was not as good as the future treat they could obtain by bartering with the token or using the tool to get food. If they selected the first-offered treat, they wouldn’t be able to get a token or tool to obtain the better, future treat. And in both cases (though not as often as when there was no “immediate” treat), the birds chose the tool or token, ignoring the present treat, over 70% of the time—and this was true whether the delay was 15 minutes or 17 hours.  I won’t give the data; you can see it in the paper.

The authors suggest that this says something deeper than does the simple ability of ravens to cache food for the future; and they might be right. Cacheing and watching others cache might just be evolved responses without real learning, while in this case the ravens clearly learned that tools and tokens were useful for getting food–even after nearly a day. Well, we knew they could learn from previous work, but does this study show that they can “plan for the future”?  I’m not sure, as it shows ravens have long memories of associating a stimulus woth a reward. But the authors conclude that even that kind of long-term association is seen only in apes:

This study suggests that ravens make decisions for futures outside their current sensory contexts, and that they are domain-general planners on par with apes. In the tool conditions, including self-control, the ravens were at least as proficient as tool-using apes. In the bartering conditions, the ravens outperformed orangutans, bonobos, and particularly chimpanzees [Detailed comparisons are available in (14).] The first trial performances show that the ravens’ behaviors were not a result of habit formation, and that they perform better than 4-year-old children in a comparable set-up.

What impresses and convinces me more is the fact that ravens can give up a present treat in anticipation of a better one in the future. To me, that suggests that these birds really can anticipate and plan for future actions. So yes, that’s pretty impressive.

This paper has gotten a lot of attention from the press, but since the paper is hard to read (all that data was compressed into two journal pages), people have just accepted the take-home message. But I think the details are fascinating—especially the finding that ravens can turn down a treat if, by so doing, they get a better treat the next day. That is bird smarts!

Here’s a photo from the Perspective piece showing other clever behavior of ravens:

(From Boeckle and Clayton perspctive): Tool-using raven. In a separate experiment, a raven involved in the Kabadayi and Osvath study uses a stick to poke for food in a tube (first photo). She succeeds in pushing the stick into the hole (second photo), but handling the stick is tricky for her. She then invents a new way of solving the problem: She fills the tube with bark pieces (third photo) and thereafter pecks at the bark until the food falls out.

____________

Kabadayi, C. and M. Osvath. 2017. Ravens parallel great apes in flexible planning for tool-use and bartering. Science 357-202-204.