Mixing Memory

Crows are smart. Really smart. But just how smart are they? Studying non-human primates, particularly gorillas, orangutans, and chimpanzees, researchers have shown that they’re capable of what’s called meta-tool use, or using one tool with another tool (I’ve mostly seen it defined as using one tool to modify or improve another tool, but more on definitions in a bit), but it’s not always something these primates do readily. Monkeys (macaques, e.g.) are much less likely to display meta-tool use. Meta-tool use is difficult because it requires behaving in a way that isn’t directly linked to a reward, something that non-human animals (and even human children, as the paper discussed below notes) have trouble doing. Do crows display meta-tool use when the situation requires it?

Crows definitely use tools, and they’ll even modify tools if the initial material isn’t sufficient for their purposes. Here is a great video of a juvenile New Caledonian crow (Corvus moneduloides breaking off a twig and fashioning a hook on the end so that it can extract a piece of meat from a hole. That’s impressive, but not as imppressive as meta-tool use. However, a paper to be published in the September 4 issue of the journal Current Biology, Taylor et al.(1) purports to show that crows are capable of meta-tool use as well.

Taylor et al.’s first experiment went like this:

Food (meat) was placed in a 15 cm deep horizontal hole 1.75 m away from two identical “toolboxes”. The front of each toolbox consisted of vertical bars that allowed a crow to insert its bill but not its head. We placed an 18 cm long stick tool 4 cm inside one toolbox. This tool was long enough to extract the meat but out of reach of a crow’s bill. In the other tool box, we placed a stone in a similar position… In front of the toolboxes, we placed a 5 cm long tool. This tool was too short to extract the meat but could be used to extract the long tool from the tool box. Successful completion of the task required a crow to use the short stick to extract the long stick from the box and then transport the long stick to the hole and extract the food. (p. 1)

Here’s video of one of the crows, named “Gypsy,” performing the task, via Henry:

This was Gypsy’s first go at the task. Three out of the other six crows tested also performed the task successfully on their first trial. The other four crows took longer (5, 19, and 23 trials), but were eventually able to get the meat. Taylor et al. note that this performance is similar to that of nonhuman primates, most but not all of which of which perform analogous tasks successfully on their first trials.

How are the crows able to complete this task? Taylor et al. rule out a couple possible explanations, including trial-and-error learning (because the crows didn’t try out the long-stick first, for example), and suggest that the crows must have knowledge of the causal relationships involved in the task. To test this hypothesis, they ran a second experiment in which the short stick was in the box and the long stick outside of it. That means the crows could solve the task by simply using the long stick to get the meat, and thus without meta-tool use. The reasoning is that if they are aware of the causal relationships involved, they will see without trial and error that they can simply use the long stick.

In this second experiment, all of the crows put the long stick into the short stick’s box, but only one extracted the short stick, and “no crow took the short stick to the hole” with the food in it (p. 3). Taylor et al. argue that this indicates causal knowledge. This further implies, they argue, that the crows were using a form of analogical reasoning to solve the first task. That is, they were comparing the problem of getting the long stick out of the box with the short stick to previous examples of extracting food with a stick.

I have to admit that I’ve been going back and forth on how convinced I am by these two experiments since I read the paper yesterday. First, the behavior doesn’t fit the definition of meta-tool use that I’m familiar with: modifying/improving one tool with another. In this case, the crows are simply using a tool in its normal function, but with another tool. Over at Henry’s blog (linked above), Alex (who I believe is Alex Taylor, the paper’s first author) comments that this behavior is in fact meta-tool use, and since I’m not an animal researcher, I’ll defer to his definition. Whatever they’re doing, it does appear from the second experiment that they’re using something like causal knowledge of the situation to solve the task, though I wonder if that requires analogical reasoning. Recognizing that the long stick is sufficient, but the short one isn’t, when looking at the placement of the food, and then understanding that the short stick is sufficient to retrieve the long stick is pretty impressive. I’m not sure it requires analogical reasoning — association might be sufficient — but I suspect that future studies will sort all of that out. However they’re doing it, it’s clear that crows are pretty damn smart.


Taylor, A.h., Hunt, G.R., Holzhaider, J.C., Gray, R.D. (2007). Spontaneous metatool use by New Caledonian crows. Current Biology, 17, 1-4.

Comments

  1. #1 Simon G.
    August 17, 2007

    Yes, Alex would be Alex Taylor.

    –Simon

  2. #2 Ed Yong
    August 18, 2007

    I loved this paper, and I think that while animal intelligence research is always the subject of debate over interpretation and semantics, it’s clear that crows are indeed very clever. Just check out this clip from the Life of Birds.

  3. #3 andy
    August 18, 2007

    What are the implications of this for the evolution of technological civilisations? It seems to be the case that tool-using intelligence is not a fantastically unlikely process (given the existence of multiple examples of species with this capability on this planet at the same time), but what about getting to agriculture and cities and computers (or what Jack Cohen calls “extelligence”)?

  4. #4 Daryl McCullough
    August 29, 2007

    I’m not sure whether to call the crow’s behavior “meta-tool use”, but one thing is clear: The crow is performing a several step process, where the goal of the first step is something that has no value in itself, but is just a stepping-stone towards the ultimate goal. Of course, there are lots of examples of such behavior in animals, such as nest-building, but it seems to me that most such examples are of innate behavior, rather than behavior made up “on the spot” to solve a unique problem.