Corvids and causal reasoning: Further evidence of higher-level cognitive processing in crows

Corvids (birds of the crow family) have long been accepted to be among the most intelligent bird species.

Crows have proven imaginatively playful, adept at problem solving, as well as being keen surveyors of their environment. One study published in the journal Animal Cognition in May, 2012 demonstrated that individuals from a species of carrion crow were able to recognize and differentiate between familiar and unfamiliar human voices. Authors of the book Gifts of the Crow, John N. Marzluff and Tony Angell, communicate the findings of researchers in northern Japan that observed crows stuffing deer dung into the deer’s ears, seemingly in a spirit of playfulness. In another case, researchers revealed New Caldonian crows to be industrious tool manufacturers and users, capable of forming hooked implements from pandanus tree materials for prying food from crevices. And we can now add “capable of causal reasoning” to that list.

The same species of New Caldonian crow just mentioned, prized for its completion of complex cognitive tasks and tool-using strategies, was featured once again in a study released this summer. Published on Aug. 16, 2012 in the journal Proceedings of the National Academy of Sciences, Alex H. Taylor et al. observed, “that the crows can reason about hidden causal agency.” Causal reasoning—the ability to make inferences about unknown cause(s) based on their known effects—was previously only known to occur in the human species.

When from our vehicles we see, for example, a murder (or group) of crows circling and diving an area of the highway off on the distant horizon, though being unable to directly see what is causing them to behave in this way, we might make inferences to make sense of their behaviour. Based on their nature as scavengers, we might surmise that the crows are picking over road kill or maybe assaulting another bird – this would be an example of causal reasoning. Because we cannot readily detect what is causing them to behave in this way, we infer the cause based on a mix of well-established observations and conventional wisdom. That is, until our vehicle reaches the horizon and affords a direct view of the goings-on.

Taylor et al. designed an experiment with two scenarios in order to measure the sample group of eight crows’ capacity to make connections between a series of events, and infer the presence of a hidden causal agent. “[The crow] had to infer what caused an inanimate object to move,” cite the authors. The object: a wooden stick protruding from a hide (or blind).


The crows were first habituated to using a tool to extract food from a box that lay atop a table, facing the hide (and protruding stick) in the crows’ aviary enclosure. In both scenarios, a passive researcher stood in the room with its back turned to the crow, hide, and table.

In the first scenario of the “hidden causal agent (HCA)” experiment, crows perched atop a beam in the aviary witnessed a second actor enter the room, walk behind the hide, and a wooden stick protrude from said hide. The stick was then wiggled back and forth by the hidden actor, followed by the actor reemerging from the hide and leaving the room. The crow then came in to investigate the box for food with a small tool, turning its back to the previously moving, now static stick.

The second “unknown causal agent (UCA)” set of events occurred similarly, although no second actor ever entered or left the hide, and the behavioural outcome observed in the crows was quite different. Researchers operated the movement of the protruding stick from outside of the enclosure. Following the cessation of stick-movement from the hide, the crows came to investigate the food-box once more; however, this time the crows were more cautious then in the previous trial, exhibiting flighty behaviour suggestive of increased nervousness.

“To test between [hypotheses], we measured the degree of caution in the crows’ behaviour,” explained Taylor et al. “[In] the second, unknown condition, if the crows were capable of causal reasoning, they would predict that the stick might move again because they had not observed a potential causal agent leave the hide.”

The authors conclude that the increased “inspection” rates in the unknown causal agent scenario is proof that the crows were capable of something analogous to the kind of causal inferences humans are capable of.

“Animals are predisposed to associational learning,” remarked animal behaviour specialist and professor at the University of Manitoba, James Hare. “[It] is highly adaptive in their environment to associate things that are occurring, the movement of objects, with those things that preceded them.”

In a recent conversation with the Manitoban, Hare discussed the merits of the causal reasoning experiment, where he emphasized the importance of moving objects in nature. These crows, in some sense, are assessing the causal relationship underlying the movement of the protruding stick.

As Hare noted, “Moving objects should never be ignored [ . . . ] Sensory systems are tuned to alert animals to moving objects in their environment at a very basal level in the sensory system, activating all sorts of responses. So where there’s something moving, it’s not surprising they’re cuing in on other elements.”

The experimental design, according to Hare, is not unlike previous expectancy violation experiments that have been performed with mammals and birds. Hare cited an example where rhesus macaques were presented with eggplants. A curtain would drop down, and when it was lifted an eggplant would be present and the macaques would happily eat it. However, when the curtain would drop and lift to mysteriously reveal, say, two or three eggplants, the confused macaques would “fix their gaze on that outcome because it had violated what they had expected, revealing that they have this expectation [of one eggplant] and a cognitive ability to work with numbers.”

What’s interesting about the HCA experiment with the crows is that their increased level of caution, inspection, and nervousness is demonstrably related to the absence of the human actor entering the blind. “They are clearly having their expectancy violated when they can’t have a human associated with [the movement of the protruding stick],” reflected Hare.

When the human is known to be behind the hide, the crows may be attributing the stick movement to the benign actor they’ve witnessed enter and leave the hide. When that actor isn’t present, information pertaining to the nature of the mechanism operating the moving stick is no longer publicly available, or reliably known to be benign, and thus suspicious. This explains why the crows are, in Hare’s words, “engaged in more vigilance to collect more private information” in the UCA condition versus when they witness the human enter and exit the blind.

It appears that Taylor et al. may have made a very important contribution to the growing cannon of crow cognition research, providing tentative first evidence of sophisticated “causal reasoning” processes previously only attributed to humans.