by Matthew Cobb
Imagining that others experience the same feelings as oneself, or being able to see things from another’s perspective, is an essential part of being an adult human – it’s called having a ‘theory of mind’. Young children find it difficult, and either learn it or develop this ability as part of normal growth. Severely autistic individuals can also fail some of the simple tests that are used to measure this ability. This character, or a primitive version of it, must have been present in our primate ancestors, and there is evidence that chimps can attribute ‘intentionality’ to human behaviour, which suggests they can image what we feel/think.
Now, in a paper just published in PNAS, Professor Nicky Clayton’s group from the University of Cambridge have shown that the Eurasian jay – a beautiful bird that very occasionally comes into my (very small) back garden, are apparently able to attribute ‘desire-states’ to other birds. In children, attributing a ‘desire-state’ to an individual (their wants and wishes) occurs prior to the development of a fully-fledged theory of mind. If Clayton’s group is right, and jays do have the ability to attribute ‘desire-states’ to other individuals, the simplest explanation is that the same ability has evolved at least twice – once in our hominid lineage, and once (and perhaps first) in the avian dinosaur lineage.
The authors set up a cooperative feeding experiment, in which males could provide their mate with food. The reason for this is that it is presumed that there is a link between cooperative behaviour and state-attribution (and food-sharing is a key part of courtship behaviour in jays). The females had already eaten as much as they wanted of one of two kinds of food (either waxmoth larvae or mealworm larvae). The hypothesis was that if the males could attribute a desire-state to their partner and had observed her at the eat-all-you-an-eat buffet with one kind of food, he would be less likely to offer her that food. And this is exactly what happened – the males would share significantly less of the food that the female had been eating.
For the doubting Thomases and Thomasinas, although the sample size was small (7 pairs) they did the experiment 20 times on each pair, swapped around which food the females were stuffing themselves on and also included a maintenance diet control. And they included an ‘unseen’ condition in which the male couldn’t see what the female had eaten; in this case there was no significant effect, showing it wasn’t the female shaking her head and saying ‘no thanks, I’ve eaten too much of them already’ when proffered one kind of food. It looks pretty tightly designed to me.
The authors deal with some of the obvious criticisms:
“The results of the unseen condition negate the possibility that the males might have learned a simple rule (such as “do not feed what has just been eaten”). Learning about an action can only occur when that action is reinforced (regardless of the content of what is being learned). Therefore, in our case, for the male to learn when is an appropriate situation in which to feed the female different foods, he must have experience of the acceptance or rejection of certain foods by the female. As discussed earlier, the results of the unseen condition indicate that the female’s immediate behavior when the male is sharing the food during the test phase is not sufficient to elicit the differential sharing pattern by the males: it is only in the seen condition that the male provides the food that the female desires. This difference between the seen and unseen conditions makes it highly unlikely that males would have been able to previously learn any rule for which the female’s acceptance or rejection of his attempts at sharing would have acted as the reinforcement.”
They are also extremely careful about how far their data should be interpreted. They conclude, circumspectly:
“The results of the current study present a crucial first step in demonstrating state-attribution. They fulfill the necessary behavioral criteria, namely ruling out behavior reading at the time of action and providing evidence of self–other differentiation. Our study suggests that the Eurasian jays’ food-sharing behavior represents a useful paradigm within which to investigate whether these birds, and more generally nonhuman animals, might be capable of desire-attribution.”
So the answer to the question in the title of this post is: we have no evidence that jays (those are the modern dinosaurs, of course) have a theory of mind; but it seems that they can do something that is on the way to having such a theory.
In one way, you might say this is no surprise. Corvids (jays, crows, magpies etc) are amazingly smart, as Clayton’s group has shown on a number of occasions, in particular in their tool use and their ability to pretend to hide food if they see another bird watching them (this would indicate desire-state attribution, too, I think). And magpies are the only non-mammals that have been shown to pass the mirror self-recognition test, which is one of the proxies we have for something like consciousness. Furthermore, the English poet Chaucer knew this, and made a caustic comparison between the intelligence of the jay and the stupidity of the Pope in the Prologue to The Canterbury Tales:
And eek ye knowen wel how that a jay
Kan clepen “Watte” as wel a kan the pope
Ljerka Ostojić, Rachael C. Shaw, Lucy G. Cheke, and Nicola S. Clayton (2013) Evidence suggesting that desire-state attribution may govern food sharing in Eurasian jays. PNAS March 5, 2013 vol. 110 no. 10 4123-4128