SCIENCE AT THE SHINE DOME canberra 2 - 4 may 2007

Symposium: Development and evolution of higher cognition in animals

Friday, 4 May 2007

Professor Giorgio Vallortigara
Professor of Psychobiology and Physiological Psychology and Dean of the Faculty, University of Trieste, Italy

Giorgio Vallortigara, Professor and Chair of Behavioural Neuroscience and Animal Cognition at the University of Trieste, Italy, is the author of over 150 scientific papers, most in the area of animal cognition and comparative neuroscience. He discovered the first evidence of functional brain lateralization in the so-called “lower” vertebrate species (fish, amphibians); he also worked on comparative cognition, in particular on visual perception of biological motion and spatial learning and memory. He served in the editorial boards of several cognitive science and neuroscience journals and has been the recipient of several awards.

The cognitive chicken: Higher mental processing in a humble brain

I would like to start with a personal recollection, because I am an experimental psychologist by training, supposed to be a scientist of the mind, and I remember that my father was very surprised by the subject I selected for my research…

(Click on image for a larger version)

...which is not traditionally considered a champion of mental life! To be honest, after a while my father was ready to accept my quite esoteric field of research, but still very puzzled whether there would be somebody interested in giving me a salary for this research. I might say yes, they pay me, but not very much.

(Click on image for a larger version)

Actually, there are good reasons to use the young domestic chick as a model to investigate cognition. One reason is that this little creature is capable of a very amazing form of learning, filial imprinting, which has been widely investigated, most notably by Professor Bateson. But I am interested in filial imprinting as a tool to investigate cognition, with methods which are very similar to the methods which developmental psychologists use to investigate human infants’ cognition.

The second reason is that this is a precocial species, which means that we can have very precise control on the sensory and motor experience of these animals – or lack of it – and this can be combined with quite sophisticated behavioural tests thanks to the early motor development of the chick.

(Click on image for a larger version)

I will start with some old work on a very basic mental ability of humans – that is, the ability to recognise partly occluded objects. Usually, when we see an object which is partly occluded by another, opaque object, we have a strong, compelling impression that the object is continuing in some way behind and beyond the obstacle.

(Click on image for a larger version)

In fact, if the obstacle is removed, this complete figure is what you expect to see.

(Click on image for a larger version)

You will be very much surprised if you see something like a figure with gaps still in it! This is strange, in a sense, because the formerly occluded parts are not readily accessible to the peripheral receptor, to the eye, so they are really a construction of your brain.

Sometimes people are not very surprised by this evidence, because they say, ‘Well, it’s just a matter of past experience, of knowledge, of memory.’ However, there are several nice examples in which you can see that amodal completion, as this phenomenon has been dubbed, could act in contrast, in opposition, to past experience and knowledge.

(Click on image for a larger version)

This is an example of what I mean: quite an elongated car, which is not very familiar in the environment.

(Click on image for a larger version)

This is another example. Even if you look at the true, complex shape of this object, and you learn about its real shape, when the middle portion of the object is covered by a finger you perceive it as a triangle. It doesn’t matter how many times I do this trick, and it doesn’t really matter what sort of shape I am using, because what is at work here is the same very basic, early mechanism that Professor Srinivasan mentioned before, that there is an interpolation mechanism for boundaries which are co-linear. So this is very early visual processing.

(Click on image for a larger version)

What about animals? Some years ago we started to investigate this sort of phenomena using imprinting as a tool. We imprinted chicks on an artificial object – a triangle, for instance.

(Click on image for a larger version)

And then we tested chicks for choice between two different versions of this ‘mother’, a partly occluded version and an amputated version.

(Click on image for a larger version)

We tested chicks for choice in a runway test between the two stimuli. Consider the two stimuli from an anthropomorphic point of view, that is from the point of view of our own perceptual experience. In spite of the fact that the overall red and black areas are exactly the same in the two stimuli – obviously the spatial arrangement is different – the two patterns look very different. The one on the left looks like a complete triangle, which is simply by accident partly occluded by a bar. The other is something different, a small, irregular triangle plus another geometric figure.

(Click on image for a larger version)

It turns out that chicks behave like us in this task – they prefer to approach the partly occluded object – but obviously several controlled experiments were needed before we could say that they really do perceptually ‘complete’ partly occluded objects.

(Click on image for a larger version)

I would like to show a couple of these controls. For instance, the reverse experiment can be done, imprinting chicks on a partly occluded triangle and then removing the occluder. Chicks prefer the complete to the amputated triangle. This is not because they prefer the larger red area, because in the control condition, when they are imprinted on the amputated triangle they do prefer the amputated triangle to the complete triangle.

(Click on image for a larger version)

Some colleagues working in development psychology, Alan Slater and Stephen Lea at Exeter University, UK, were interested in duplicating our results with exactly the same stimuli which are used with human infants. The procedure used with human infants is habituation-dishabituation. Human infants are presented with a rod moving back and forth, partly occluded by a bar, until they are habituated to the stimulus.

(Click on image for a larger version)

Then the occluder is removed and they are shown a complete rod.

(Click on image for a larger version)

Or the two pieces of rod which were really visible during habituation. So the idea is that if human newborns do perceive completion in this case, they will not be very surprised to look at the complete-rod stimulus but they will be surprised to look at the two-pieces stimulus – or vice versa if they do not perceive completion. It turns out that it takes about four months of age for human infants to exhibit amodal completion.

(Click on image for a larger version)

But chicks on their first day of life, with just two hours of imprinting on a moving rod or a partly occluded rod or a two-pieces rod, and tested for choice between a complete and an amputated rod, preferred – as suspected – the complete when imprinted on the complete; the two-pieces when imprinted on the two-pieces. But in the crucial condition, when imprinted on the partly occluded, they preferred the complete, in spite of the fact that physically there was exactly the same red area visible in the partly occluded rod as in the two-pieces rod.

(Click on image for a larger version)

This ability to mentally complete objects is just an example of a more general ability of animals to deal with information in very impoverished stimuli by interpolation and inferential-like procedures. An example is shown here.

It is very difficult, I imagine, to figure out what sort of stimulus this is. But if it is put in motion you can immediately perceive a human figure, a walking man.

(Click on image for a larger version)

This sort of biological motion stimulus has been used recently with chicks in our lab. We imprinted chicks on a point-light moving hen, like the one at the top of this slide, or on a scrambled version of it, shown below that. The local vectors of movements are exactly the same, but the dots are spatially changed in position. We found a pattern of sex differences. Females preferred the familiar and males preferred the unfamiliar stimulus. But chicks were clearly able to tell the two stimuli apart.

Obviously, this does not demonstrate that they perceive the same 3-D structure as we perceive in this condition – they can simply use the kinematics of point-light movements. But we have some further evidence for that.

(Click on image for a larger version)

This is from the inversion effect in biological motion. On the left here you see the point-light display of a hen, and on the right is exactly the same stimulus but simply shown upside down. You can notice how difficult it is to recognise a hen when the stimulus is upside down.

(Click on image for a larger version)

We found that before imprinting, chicks have some preference to approach biological motion stimuli; moreover, chicks also tend to align their body in the same direction as the apparent movement of the hen. For instance, if during the first four minutes of test the hen is moving in one direction, they orient their body in the same direction of movement of the hen. And when starting from minute 5 the direction of motion of the hen is reversed, they also reverse their orientation accordingly, in such a way to remain oriented with their body with the apparent direction of movement of the hen.

(Click on image for a larger version)

But when presented with an upside-down version of the hen, they were completely random in their body orientation. So they do experience the inversion effect, as humans do.

(Click on image for a larger version)

The ability to extract three-dimensional information from two-dimensional displays is shown here in the so-called stereokinetic phenomenon. This is a nice example of convergence evidence from arts and science, because stereokinetic phenomena were discovered independently by the visual artist Marcel Duchamp, in 1930, and by Vittorio Benussi, the founder of the tradition of perceptual psychology in the north-east of Italy (one of his most important followers was Gaetano Kanizsa, in Trieste).

(Click on image for a larger version)

This stereokinetic effect is a Rotorelief developed by Marcel Duchamp.

If you look at these stimuli, particularly with just one eye in such a way as to avoid the use of stereoscopic depth, after a few seconds you can perceive them as 3-D objects tilting in three-dimensional space. For instance, the middle image looks like a sort of glass or a lamp. It usually takes some time for naïve subjects to obtain the stereokinetic transformation.

(Click on image for a larger version)

We used a very simple version of the stimuli which are shown here, the original Benussi stimuli. Here you can see (image ‘b’) a cone after a while, with a point towards the observer or sometimes farther away. Image ‘c’ looks like a cylinder when set in motion, after some inspection.

(Click on image for a larger version)

We exposed chicks to this very strange type of ‘mother’. Soon after hatching, chicks were presented with these moving stimuli for just 90 minutes.

(Click on image for a larger version)

The idea is that if they do perceive something solid, three-dimensional, after looking at this 2-D stimulus, they would probably prefer to associate with a real 3-D cone rather than a 3-D cylinder, or vice versa.

(Click on image for a larger version)

So this was the test.

(Click on image for a larger version)

When imprinted on the cone, chicks preferred the cone to the cylinder.

(Click on image for a larger version)

And when imprinted on the apparent cylinder – the stereokinetic one – they preferred the cylinder to the cone.

Note that no information was available on the 2-D frontal plane display to perform the discrimination.

(Click on image for a larger version)

Similar results were obtained recently, with Professor Lesley Rogers – not using the imprinting procedure, however, but using a conditioning procedure in a primate species. We found that marmosets also experience stereokinetic illusions.

(Click on image for a larger version)

You are probably not terribly surprised by these findings, because these are in some way examples of perceptual rather than cognitive abilities, ie, quite ‘early’ visual perception. So let’s consider something more complicated.

Sometimes objects are not simply partly occluded by other objects; they may be completely hidden by other objects. They might completely disappear from sense organs. However, we have this compelling impression that an object which is no longer available to direct sensory experience is still in existence, so to speak. This object permanence concept has been widely studied by developmental psychologists and it has been shown that it took some time for young infants to be able to exhibit the object permanence concept.

(Click on image for a larger version)

We studied this again with an imprinting procedure in the young chicks. Against the left-hand wall is the ‘mother’, a red ball. The chick is located in this sort of a maze. There is a window so the chick can look at the ‘mother’ but it is impossible for the chick to rejoin the mother. However, it could perform a detour, moving, entering one aperture or another, and then turning right or left. Obviously, we did know that chicks are able to learn the detour task, but our question was: what happens at the very first presentation, when the chick is for the first time faced with the problem?

If out of sight really means out of mind, then turning to the right or to the left after entering one or other of the two apertures along the corridor would be exactly the same. However, if chicks have some sort of representation of the disappeared object, together with an idea of its spatial location, we can expect it to turn in to the D or C compartments rather than to the B or A compartments. And this is exactly what happens.

[A video was shown, with the following commentary.] I would like to show you an example. We did a lot of comparative work with other species of birds, so I will show you the performance of a quail rather than a chick, just to show you that the behaviour is in no way a stereotyped behaviour. To put it in an anthropomorphic way, it seems that the animal is in some way ‘pondering’ the task.

(Click on image for a larger version)

Obviously, one problem with this test is that the amount of time during which the ball, the imprinting object, is no longer visible is very short, and it is out of the control of the experimenter. So I will show you briefly a better experimental condition.

But before that I would like to show some evidence for object permanence in another species of birds. We have European jays in the lab, and we tested them with this very sophisticated experimental apparatus!

We used the traditional Uzgiris-Hunt battery of tests for object permanence which has been developed for human infants.

[A video was shown, with the following commentary.] This is a very simple test. The experimenter is using a mealworm which is covered by a towel. The animal is not trained; this is the first presentation so there is no training. You can see that the bird solves the task without any difficulty. This is an ability which is not available to young infants up to about 18 months of age.

Here is another example. It is taking some time to convince the jay to collaborate, and so the experimenter will attract the attention of the animal in just a few seconds. So the first little box with some tape, then another one, and finally a towel. Remove the first, remove the second, and then go to the small one.

The next task is the invisible displacement task, which is a bit more complicated. The mealworm is placed into a cylinder and then the mealworm is displaced, invisibly, from the cylinder behind a towel. The bird can look at the cylinder, which is empty, and then he searches behind the towel. That is the invisible displacement.

The final task is associated with a serial search, which is at the top of stage 6 of object permanence. There are three screens and there is a mealworm in the hand of the experimenter, and the hand is moved slowly. Actually, the mealworm is in the sleeve now; the hand is empty. So the bird searches in the last one in the series of screens and then, ‘It should be here’ – no. ‘Here?’ – yes. Actually, there was one in the first one as a control for olfactory stimulation.

(Click on image for a larger version)

What sorts of brain mechanisms are available to mammals to perform these sorts of object permanence tasks? We know that the prefrontal cortex is very important.

(Click on image for a larger version)

So, recently, we tried some lesion experiments to the presumed equivalent of the prefontal cortex in the avian brain, which is the NCL, the nidopallium caudolaterale area.

(Click on image for a larger version)

We posed this sort of working memory task for chicks. The chick is confined within a transparent partition, and a ‘mother’, an imprinting object, disappears behind one of two screens – the procedure is repeated a couple of times, in order to get the attention of the animal. Then an opaque screen is placed in front of the cage and there is a delay which is controlled by the experimenter, 10 seconds in this case. Actually the chicks behave quite well with a delay of up to about 60 seconds, which is a very long delay. Then they are allowed to search for the ‘mother’.

In each trial they have to remember behind which screen the ‘mother’ has disappeared, then erase the memory and look for the subsequent trial in which the mother is hidden in the same or in a different screen, so it is a sort of working memory task.

(Click on image for a larger version)

When chicks were tested after NCL lesion, if the delay was zero – that is, if they were allowed to search for the ‘mother’ immediately after its disappearance – there was no effect at all. But a 10-second delay was enough to obtain an animal which was completely random in searching for the ‘mother’.

(Click on image for a larger version)

At this point, I would like to move from object cognition to another cognitive ability which is encoded precociously in the vertebrate brain. I would like to discuss briefly the ability of young chicks to use number. Again we used imprinting as a procedure. We imprinted chicks on three or one imprinting objects and then we tested them for choice between three and one imprinting objects, using both an absolute or a relative discrimination – so, in the latter case, the overall number of objects was exactly the same. Chicks were very good at the task.

(Click on image for a larger version)

However, they go for more. If they were reared with three, they prefer three; and reared with one, they prefer three again. When tested with two versus three, a more difficult task, there was a similar result – reared with three, they prefer three; reared with two, they prefer three again. The choice was clearly related to imprinting, because spontaneous choice without any imprinting did not yield any preference for the larger number of objects.

However, it is quite clear here that they are not using number but probably the overall amount of ‘stuff’ – that is, the volume, the overall area or something like that. This is true also for human infants. But if the objects are different in such a way as to be treated as individuals, things are different.

(Click on image for a larger version)

We used this sort of stimuli this time, in which different shape, colour, area and volume can be used. So there are three objects or two objects.

(Click on image for a larger version)

And then chicks were tested with completely different, novel objects – different in size and colour, and in this case either the volume or the overall surface area was identical. This time, chicks chose the familiar.

(Click on image for a larger version)

If reared with three, they chose three; reared with two, they chose two. So this time they were really using number.

(Click on image for a larger version)

This was confirmed using conditioning experiments.

(Click on image for a larger version)

For small numerosity, we trained chicks to discriminate between one and two dots – obviously, in this case there were a lot of differences in area, overall contour length and so on.

(Click on image for a larger version)

So we checked for that after the initial training, modifying the spatial position, and chicks were still able to discriminate.

(Click on image for a larger version)

Or, when we used an identical surface area though a different colour – that is, amodal completion – and a one versus two discrimination, they were still able to discriminate.

(Click on image for a larger version)

Then we went on with a discrimination of three versus two.

(Click on image for a larger version)

Again there was good discrimination when they were tested with a change in spatial position.

(Click on image for a larger version)

When the overall surface was made identical, they still maintained their discrimination.

(Click on image for a larger version)

And then we checked for the overall contour length, the perimeter – in this case we used square-shaped dots because the computation was easier to do.

(Click on image for a larger version)

Again there was good performance with the spatial position.

(Click on image for a larger version)

And also when the contour length was identical.

(Click on image for a larger version)


(Click on image for a larger version)

Similar results were obtained for amodal completion with identical contour length.

(Click on image for a larger version)

However, when chicks were tested with a four versus five discrimination, no discrimination at all occurred.

(Click on image for a larger version)


(Click on image for a larger version)


(Click on image for a larger version)


(Click on image for a larger version)


(Click on image for a larger version)

Number obviously has not just a cardinal but also an ordinal aspect, so we trained chicks to look for the third in a series of 10 holes, or for the fourth or for the sixth, and then we tested them in a 20 trials test without any food in the holes. They were very good at this task.

(Click on image for a larger version)

Obviously, it could be that they were using the length, the distance between the holes, rather than the number, so we checked for that.

(Click on image for a larger version)

We trained chicks on the fourth hole, and this time the distances between the holes was changed in such a way that if they were using the length they would go to the second; if they were using the number they would go to the fourth. And they were very good in this performance, using number.

(Click on image for a larger version)

We did a series of controlled experiments.

[A video was shown, with the following commentary.] I would like to finish with an example of the performance of chicks in this sort of task. In this task the chick is trained to check for the third container from the right.

Note that reinforcement is available in this case in all the containers; thus, no olfactory cue can be reliably used to perform the requested discrimination.

In conclusion, it seems that the vertebrate brain is equipped from the start with a sort of tool-kit of very basic cognitive abilities to deal with physical (and social) objects, and with their numerical and spatial properties. Unfortunately I don’t have time to discuss spatial properties here. Apparently, cognition of objects, number and space requires very little experience and interaction with the environment, being available from the start in newly-hatched chicks as well as in human newborns.

Discussion

Question 1: In that last experiment I got the impression they were counting from the left, rather than from the right. Is that what they were really doing?

Giorgio Vallortigara: Yes, that’s very interesting. There is a very strong bias in this and several other experiments toward starting from the left and then to the right. It could be that it has to do with the brain lateralisation, that the attention is directed prevalently to the left hemispace, or it could be – that is difficult to check for – that the mental line is, as in humans, from left to right. You would know that in humans there is evidence that the mental line runs from left to right (for reasons which are not very clear) and, apparently, irrespective of cultural aspects such as direction of writing.

Question 1 (cont.): So they are really learning a larger number by counting on the left?

Giorgio Vallortigara: Yes, that would change.

Question 1 (cont.): I don’t know if you would change the total number of cylinders to see what the effect would be. I can’t remember if you did that in the experiment – maybe not.

Giorgio Vallortigara: No, we did not. We stopped at ten, but we are now trying with larger number sets.