Room for intelligence in the context of selective perception and specific responses

It is a natural idea to study learning and intelligence in animals in the context of their natural environment and to investigate how species perceives stimuli before studying how they solve problems.

Selective perception. There are many interesting reviews of the selective perception in animals (for example, Hinde, 1970; Kaufman, 1974; Freeman, 1991, 1995; Zentall and Riley, 2000; Dukas, 2004). It is known that many animals live in a “smelly” world, and others live in a world penetrated with ultra-sounds. Probably, rats live in a world dominated by sensations transmitted to their brains through vibrissae.

A huge body of recent investigations have derived from Tinbergen’s (1951) ideas about the basic process of distinguishing between perceived and effective impulses in animals; this process is called stimulus-filtering. Members of different species selectively perceive sounds, smells, as well as tactile and visual objects because they somehow filter incoming stimuli. A frog sitting in a marsh full of the sounds of evening chorus of all sorts of frog species hears only the call of its own species.

Besides, it is well known that some species can catch signals beyond the reach of other species. Bats, moths and dolphins use sonar, and this is only one small group of examples from a great array.

A pet subject for illustrating stimulus filtering in many books and reviews is a description of how a toad perceives prey units. Toads belong to the so-called “sit-and-wait” predators and they remain motionless for long periods of time. When a prey item moves in their receptive field, the predator lunges with great speed (relative to the prey) and snaps it up. For example, when a toad spies a fly walking along a wall, it begins to orient itself in preparation for the tongue flick that will capture the insect, but if the insect stops, the toad shows no further reaction. The fly is still clearly visible, but now it elicits nothing in the toad’s central nervous system that would compel the toad to continue its orientation. Somewhere between the eye of the toad and its muscles, the sight of the immobile fly is filtered out (Wallace, 1979). The toad sees the fly, certainly, but not as an orienting stimulus. The key to the toad’s motion-based prey detector is the receptive field, the fundamental unit of its perception machinery. Each of the thousands of receptive fields in the toad eye consists of the following components: (1) a single ganglion cell that integrates information from the receptive field and relays a response back through the optic nerve, (2) bipolar cells that are all connected to the single ganglion cell on one synapse and connected on the other side to one or more receptor cells, (3) a circular cluster of receptor cells, the receptive field, that consist of (4) central excitatory photoreceptors that are loosely tethered together through bipolar cells, (5) peripheral inhibitory photoreceptors that are connected to a single bipolar cell. This is the smallest neural unit of stimulus filtering found in the visual system.

Different stimulus filtering is found at the level of specific photoreceptors in that individual neurons pay attention to certain signals (e.g., rods vs cones, and if cones, then the kind of photopigment found in the cone). The stimulus filtering found in receptor sensitivity is hard-wired by evolution. However, receptive field has special cellular interactions built into it that results in certain information being ignored and other information being acted upon. The receptive field is also the smallest unit of the toad’s motion detector. The excitatory and inhibitory cells act in unison to either filter or detect objects from higher centers such as the optic tectum. The stimulus filtering in the receptive field is also capable of being modified by the animal’s internal state - food satiated or hungry. If a large object casts an image over the visual field, the light intensity changes on the photoreceptors. Both excitatory and inhibitory cells from many receptive fields are triggered. Because the ganglion receives impulses from both the excitatory and inhibitory cells (through their respective bipolar cells) the effect of the inhibitory cells cancels out the effect of excitatory cells. No impulse is sent from the ganglion cell to the optic tectum. However, if a small image passes over the visual field, the small image tends to trigger fewer receptive fields. The small image will also tend to excite some of these fields because the image hits many of the central excitatory cells, but only a few peripheral inhibitory cells. The ganglion cell receives impulses from the expiatory cells (through their bipolar cells), but with little inhibitory feedback, the action potential is relayed on to the optic tectum for further integration. The optic tectum receives inputs from all ganglion cells. Several clusters of ganglion cells form a higher-order receptive field at the level of the optic tectum that integrates information form the clusters of receptive fields. Consider objects of different shapes that might strike receptive fields. Receptive fields come in a variety of " flavours". Some are used for detecting long thin objects, others large objects, etc. One of the toad’s favourite foods consists of worms - long thin objects. There are receptive fields that are tuned to fire when long thin objects pass across them. When several " long-thin" receptive field detectors have the image of a bar pass over the receptors, their ganglion cells will relay the information to the optic tectum. The grasp reflex then takes hold, and the toad orients with both eyes. Once both eyes are locked on, other motor neurons cause the toad to lean forward, open its mouth and eat the worm.

It is important to note that if the coding (filtering) is done at the level of the receptor, it is called peripheral coding; if it takes place in the central nervous system, it is called central coding. An advantage of peripheral coding is that useless or irrelevant information is discarded early so that no neural activity is wasted on its transmission. On the other hand, peripheral coding is disadvantageous in that there are fewer neurons operating at the peripheral level than in the central nervous system. Therefore, any peripheral coding, or filtering, is likely to be less specific than central coding. The complexity and coordinating of any resulting behaviour will therefore be reduced. Central coding might be expected in animals living in rich or varied environments with high levels of potential stimuli that could be sorted and generalised in the brain. Animals, such as the frog, that rely more on genetically based and rather inflexible patterns tend toward dependence on peripheral coding. Mammals such as the cat and the monkey, on the other hand, show comparatively little peripheral coding. There are however notable exceptions, and the real picture is more various and complex (Muntz, 1971).

It is worth to note that, studying stimulus filtering in animals, researchers are not always able to remove their professional filters which prevent them from understanding of how an “outsider” species react to its environment. An amusing example comes from Vallortigara’s (2000) paper about animal’s “left and right perceptual World”. The matter concerns visual lateralisation in many species. The first evidence of visual lateralisation in the intact animal was obtained in the domestic chick by temporary occlusion of either the left or right eye (Rogers and Anson, 1979; Rogers, 2002). This procedure revealed that the right eye is better at discriminating visual stimuli, such as grains from pebbles, and that the left is more reactive to emotionally charged stimuli. These results met with initial incredulity; a comment by an eminent animal lateralisation researcher, involved in studies of split-brain monkeys, clearly expressed this concern: “These results lead to the plausible but revolutionary inference that a bird more effectively searches for food with its right eye while it watches for danger with its left! ” (Hamilton, 1988). However, what at first seemed unbelievable turned out to be absolutely correct and was confirmed in a variety of other species of birds (Vallortigara, 2000).

We know now from a huge body of data that organisms perceive and react selectively in their species- specific ways. Studies of this kind pave the way for establishing a link between learning abilities and ecological traits of species.

Fixed Action Patterns: small room for learning. Members of many species display similar sequences of similar behaviours over and over again (Sisyphus is not a bad example here). As it was noted in Part I, Lorenz (1937, 1950) and Tinbergen (1942, 1951) elaborated the paradigm of classical ethology basing on some important ideas and definitions. One of the main ideas in their theory is that every species has a repertoire of stereotyped behaviours called Fixed Action Pattern (FAP). Lorenz suggested that the FAP are (1) innate (and used the German word “Erbkoordination” which translates literally as 'inherited co-ordination' to describe them); (2) common to all members of the species (species-typical), and therefore they are as characteristic of the species as shared structural features; (3) once triggered by sign stimuli (see below), fixed action patterns proceed in the absence of the triggering stimulus. FAP is a motor program provided by a specific neural instruction, a component of a subject’s nervous system, called the Innate Releasing Mechanism (IRM). The IRM is the sensory mechanism that detects the signal, and FAP is a fixed sequence of stereotyped acts.

Natural World is full of displays of FAPs, and among them courtship is the most specialised. Characteristic sequences of behaviours in courting ceremonies are highly stereotypic in many species. Long before the concept of FAP had been accepted, Julian Huxley, Lorenz’s teacher, in his study of the behaviour of great crested grebes, found that, in course of evolution, certain behavioural patterns lose their original functions (such as specific motions that are used for feeding or fighting), and become purely symbolic, useful only in communication. He called this process ritualization (Huxley, 1942). In great crested grebes, for instance, a couple, facing each other, adopt an upright posture and shake their heads from side to side. Fixed action patterns in general, and particularly in courtship, can be used to follow the evolution of behaviour because they vary between related species of animals. Such analysis was one of the main interests of early ethologists. A stereotyped pattern of courtship behaviour in male ducks is one of the classic examples dating from O. Heinroth’s (1911) observations.

In modern studies of evolutionary ethology courting rituals are now known to be one of the major causes of speciation in animals. There is an interesting example of ritualisation among the insects. Several of thousand species of Empid flies are known as " balloon makers." That is, the male flies capture an insect and enclose it in a frothy bag. The male carries this package around with him as a lure to entice a female to mate. Some species dispense with the froth. Empis barbatoides males capture little insects, usually weaker flies like bibionids (March flies), and dangle them in front of females as a preface to courtship. Members of other species present ”symbolic” gifts such as silken balloons (sometimes containing prey and sometimes empty). Other species show signs of sex reversal where females compete for males, and females have secondary sexual traits such as swollen abdomens aimed at fooling males into thinking they are full of ripe eggs. In other species, males attract females by displaying fake dead pray which is, in reality, merely a growth on male legs. The evolution of nuptial gifts is thought to be a classic case of ritualisation in courtship behaviour (Lorenz, 1965). Recently LeBas et al. (2004) have examined the DNA of several species, recording their behavioural data and investigating which factors different mating techniques are associated with. They suggest that the central problem of evolution of nuptial gifts in flies is connected with sexual conflict between males and females and variations in efforts which males and females put into courtship behaviour. This, in turn, is closely connected with cheating and counter-strategies in courtship.

As it was already mentioned, fixed behavioural patterns can be mixed with flexible behaviours. For instance, hunting in some birds involves instinctive behaviour. As a hunting falcon begins searching for prey, its behaviour is highly variable. At such times, the bird is hungry and might be equally pleased at the sight of a flying bird or a scampering mouse, so the falcon can catch its victim in the air or suddenly trop towards the ground. So the first stage of falcon’s hunt may be flexible, focusing only later when prey is spotted. When the falcon spots its prey, there is a point at which its behaviour can no longer be altered. It is committed to a specific pattern. Its actions now are performed in very much the same way as they have been at the same stages in previous hunts. This is the brief moment at which the falcon’s feet are tightly clenched into a fist as it swoops at very high speed. A victim will be knocked from the sky and the falcon’s hunt will be successful. So at this stage a falcon shows increasingly stereotyped behaviour as it ends each instinctive sequence with a Fixed Action Pattern. A very similar pattern is characteristic for big dragonflies of the genus Aechna. In view of a victim, a dragonfly clenches its six legs into a “basket” and grasps a flying insect very effectively.

In general, a final hit of any specialised hunter, such as a bird of prey, a cat or a dragonfly, being a Fixed Action Pattern, calls a series of muscles into play that contract for the same duration and in the same sequences under all conditions. Lorenz (1952) vividly describes a specific hunting hurt in a pussy cat that is very much like that in a lion that hunts a zebra but now is aimed at the owner’s foot.

Humans also exhibit Fixed Action Patterns. Eibl-Eibesfeldt and Hass created a Film Archive of Human Ethology (see: Eibl-Eibesfeldt, 1989) of un-staged and minimally disturbed social behaviour. They filmed people across a wide range of cultures with a right-angle reflex lens camera i.e. the subjects did not realise that they were being filmed because the camera lens did not appear to be pointing at them. Eibl-Eibesfeldt had identified and recorded on a film several human Fixed Action Patterns or human “universals” e.g. smiling and the " eyebrow-flash ". Humans show a rapid brow rising which coincides with raising eyelids. Because all the cultures he examined showed this behaviour, Eibl-Eibesfeldt concluded that it was a human “universal” or Fixed Action Pattern.

Nowadays the term Fixed Action Pattern has been dropped from ethology and substituted by the phrases “ behavioural patterns ” or “ behavioural acts ” because behaviour is not as fixed as implied by the term Fixed Action Pattern. There are subtle variations between and within animals in, for example, the duration of individual components. And, what is more important, Fixed Action Patterns are not simply innate; they can be subtly modified by experience.

For example, long term monitoring of satin bowerbird bowers shows that this highly specialised courtship behaviour includes elements of learning. Recent studies suggest that the attractiveness of a male display comes from the intensity of display and from male ability to modulate display intensity in relation to female signals of comfort. Thus, in terms of sexual selection, in bowerbirds, experience and learning ability may ensure quality of male’s signals (Borgia, 1995; Borgia and Coleman, 2000).

Although some fragments of behavioural patterns can be “polished” by individual experience, it is very important to note that when animals perform a complex of behavioural patterns they lack the ability to recombine various segments of behaviours in their repertoire in new ways so as to achieve new goals.

Selecting parts of the environment: Sign stimuli and Releasers. As a rule, Fixed Action Patterns in animals are triggered by specific sign stimuli (also called key stimuli) in the environment. After the stimulation, it does not require any more stimuli for the continuation of the event. This was demonstrated by Lorenz's work with the Greylag goose. When a Greylag's egg is removed from its nest and placed in sight of the goose, it will extend its neck and roll it back in with its beak. If then the egg is removed out of sight when the neck extension and rolling movement has already started, the goose will still continue as if the egg were still there.

Tinbergen (1942, 1951) described a variety of models that would release displays of behavioural patterns in animals. A classic and widely cited example is reactions of sticklebacks on various models imitating specific features of males and females. In spring male sticklebacks change colour, establish a territory and build a nest. They attack male sticklebacks that enter their territory, but court females and entice them to enter the nest to lay eggs. Tinbergen used several models to investigate which features of male and female sticklebacks elicit the attack and courtship behaviour from male sticklebacks. A realistic model of a male stickleback but with a white (instead of the natural red) belly elicits little response from males in reproductive condition. Crude models having the red bellies were much more effective. Tinbergen's findings demonstrate that a model with a red belly is attacked, while a model with a swollen belly is courted by male sticklebacks. These two simple features: the red belly and the swollen belly serve as releasers that trigger two different behaviour patterns.

The terms sign stimulus and releaser are sometimes used interchangeably. However the term “releaser” stands for stimuli that have evolved to facilitate communication between animals of the same species. “Sign stimuli ” are features of an animal's environment to which it reacts in a particular way. For example, a fly orchid looks like an insect and thus is a releaser which attracts pollinators.

Tinbergen (1951, 1976) conducted many experiments studying how sign stimuli and releasers work. In one of them, he investigated the stimulus responsible for releasing the gaping response in young thrushes. He discovered that the birds would gape at a protuberance (head) on the side of a body model. One might think that the birds would gape at the bigger head, but in fact Tinbergen found that the relative size of the head was more important than its absolute size. Presumably the birds preferred the “'head” that was the proper size for the model’s “body”.

It is important to understand that each instinctive act is not necessarily summoned by a single releaser. Instead, a particular response may be elicited by any one out of a number of releasers. For example, it was found that fighting behaviour in the male cichlid fish is elicited by five stimuli: (1) silvery blueness, (2) dark margin; (3) highness and broadness; (4) parallel orientation to opponent, and (5) tail beating (Seitz, 1940). It was found that any one of these stimuli would elicit hostile behaviour, and that any two would elicit about twice the reaction as one (although the relative strengths of reactions is very hard to measure precisely). Therefore, sometimes responses to releasers appear to be additive, so that the whole is equal to the sum of its parts. This additive effect is called the law of heterogeneous summation (Tinbergen, 1951).

Many examples illustrate how this rule works in different species. For example, in our experiments we applied Tinbergen’s methods of presenting multiple models for animals in order to investigate whether red wood ants recognise images of their competitors and potential prey and, if yes, what features are key for them (Reznikova and Dorosheva, 2004). First, ants appear to react on models of enemies and potential prey on the basis of the key signs which include darkness, bilateral symmetry and the presence of protuberances (legs and antennae). Additional features valuable for the ants were the size of a model and the speed of its movement. Moreover ants’ reactions for models of carabid beetles satisfied the Tinbergen’s principle of heterogeneous summation.

The whole picture of interaction between a subject and stimuli in its natural environment is clarified when we consider a phenomenon of supernormal stimuli. This is the phrase used to describe hyperbolic (usually artificial) stimuli that are more effective than the real thing in eliciting a behavioural response. It seems that the releasing value of any sign stimulus is not fixed. If blue is good, bluer is better (Wallace, 1979). Since the releasing value of each sign stimulus is modifiable, then a question arises regarding the extent to which a releaser may increase its effect.

In experiments of Tinbergen and Perdeck (1950) herring gull chicks peck at a red spot on their parent’s bill to induce their parents to regurgitate food. Chicks will also peck at a model consisting of a red spot against a yellow background. However it is possible to construct a model that is even more effective than a real head by using a red pencil with three white bars at the end. This is an example of a supernormal stimulus. In experimental situations animals look insane when they aim their parental care to supernormal stimuli and ignore their own offspring. For instance, birds such as oystercatchers, herrings and gees showed a preference for giant eggs as products of experimenters’ mockery rather than normal eggs that were bigger than their own eggs.

However, hyperbolic (and thus supernormal) stimuli may work in natural World, not only in experiments. Certain parasitic birds, such as some cuckoos and cowbirds, have young that are larger and more “babyish” than the host’s. This may be the reason why the host attends to the young parasite in preference to its own young. Small passerine birds respond fanatically to a giant red gape of a cuckoo’s chick. Undoubtedly this is a supernormal stimulus for parents. A possible evolutionary mechanism of some of such reactions has become more clear recently basing on investigations of relations between parents and chicks in barn swallows (Sacchi et al., 2002; Saino et al., 2003). Barn swallow nestlings beg vigorously for food from their parents by producing loud calls and displaying their bright orange gapes. The gape display is a reliable signal of health status since a challenge of the immune system with sheep red blood cells reduces the level of colouration, while artificial provisioning with lutein (the carotenoid causing the coloration) increases the level of coloration. Parent barn swallows respond to experimental manipulation of gape colour by changing their allocation of food. Gape coloration reflects the viability of offspring. Nestling begging calls also reliably reflect health status. Both current hunger status and long-term condition of nestlings affect their begging rate and the response of parents to the displays. This is a case when “if orange is good, the most orange is the best”, or, like authors give this, “better red than dead”. New series of questions is raised about a handicap between more and more “super-normality” of stimuli and balanced inter-relations between parents and offspring in some situations, and hosts and parasites in other situations.

We can see, then, that a knowledge of the perceptive capacities of an animal will not tell us what part of its environment is actually triggering a behaviour, since, as we know now, a certain behaviour may be stimulated by cues from only a small part of the environment (Wallace, 1979). In the effort to “prepare” behaviour with the help of knowledge about key stimuli, we should note that some stimuli may work differently in dependence of behavioural and environmental contexts.

In some cases, a set of releasers is necessary to elicit an entire adaptive response in animals. Honey bees may initially be attracted to a coloured paper flower, but they will not land unless the flower scent is also there, so in this case, the visual releaser alone will not initiate the entire behavioural pattern. This example also rises the question of what senses are brought into play in specific behavioural contexts.

We should also take into account that whereas some types of patterns such as capturing behaviour can be elicited at any time, other types of behaviour can only be released at special time. In other words, animal respond to certain environmental cues in dependence of their physiological status at this moment. For example, in many species, none of courtship signals has any effect out of breeding seasons. Should a misguided male perform his mating patterns (such as zigzag dances in sticklebacks) in a wrong time, a female may see him, but since the seasonal day length has not caused her to produce reproductive hormones, she will remain totally unimpressed by his efforts. In many animals, in particular, in some species of fishes, lizards, social insects, birds, rodents, ungulates, and monkeys, increasing levels of density of populations may switch patterns of aggressive territorial behaviour that are otherwise kept in “doze mode”. In these cases concentration of smell and (or) number of encounters with conspecifics serve as sign stimuli.

It is important to note that a simple key stimulus may in some cases trigger an amusingly complex behavioural response. For example, it turned out that beavers react with dam building for a so simple acoustic stimulus as the sound of flowing water. It has been revealed recently by Jung (2003) who noted that beavers living in ponds and lakes never build dams. He therefore obtained several pairs of beavers (all with proven dam building track records), released them in different environments and then watched what they did. Those released in ponds and large rivers burrowed into the bank, set up beaver housekeeping and then showed no more desire to construct anything beyond their holes. Those released along streams, however, found likely looking pools and then proceeded to deepen them by constructing dams at the narrow, shallow, downstream end. The investigator then proceeded to a riffle (the shallow, high gradient part of the stream) and set up a tape recorder to tape the sound of the water rushing over the gravel and stones. He then set up speakers around known beaver haunts and at dusk turned the tape on. When he returned the next morning he found the speakers buried under several feet of sticks, gravel and mud - thus effectively silencing the sound. The result was the same whether done along a beaver dammed stream, a large (and quiet) river or a lake or pond. The beavers always covered the speakers until they couldn't hear the sound of rushing water. Based on experiments with both free living and captive beavers the researcher found that the sound of rushing water was as annoying to a beaver as the sound of fingernails on a blackboard is to humans. And that beavers will pile up sticks and mud in any spot they hear that sound until they can no longer hear it.

It would appear, then, that the beavers are governed by very simple rules in their business. But they are not. Beaver engineering also include two other activities: lodge making and canal digging. Canals can extend hundreds of meters into the forest, and beavers can float branches from trees they have cut and thus move branches to safer feeding locations. The so called beaver’s lodge is a unique structure, that is, an oven-shaped house of plant material, woven together and plastered with mud, increasing gradually in size with year after year of repair and elaboration (France, 1997; Collen and Gibson, 2001). To build and maintain these constructions, beavers behave not only on the basis of their innate patterns but also include a great deal of individual experience and social interactions.

In sum, however strongly may animals be equipped with innate behavioural patterns triggered by key stimuli in their environment, we can not conclude that most species are equipped by behavioural formulae in all living instances. Neither can we believe that animals have time and abilities enough to learn what decisions are most effective in many vital situations. Instead, many species may be equipped with innate predisposition for conditioning specific behaviours in their lives. Let us explore this idea further.

22. GUIDED LEARNING AND COGNITIVE SPECIALISATION

There are now increasing evidences that animals may learn selectively only about a subset of events they face. Studies of learning in ecologically relevant contexts have shown that animals more readily learn about complex functionally critical stimuli such as the appearance of a predator, features of their parents, colouration of venomous insects, or characteristics of their natal songs. The term “guided learning” introduced by Gould and Marler (1987) acknowledges the adaptive nature of such phenomena. They argue that work done in the past few decades has shown that there is no sharp distinction between instinct and learning. The process of learning is often innately guided, that is, guided by information inherent in the genetic makeup of the animal. In other words, the process of learning itself is often controlled by instinct. It now seems that many, if not most, animals are " preprogrammed" to learn particular things in particular ways. In evolutionary terms innately guided learning makes sense: very often it is easy to specify in advance the general characteristics of the things an animal should be able to learn, even when the details cannot be specified. For example, bees should be inherently suited to learning the shapes of various flowers, but it would be impossible to equip each bee at birth with a field guide to all the flowers it might visit.

It is important to note that predisposition of species to build up one set of associations more readily than another may create both a channel for very limited learning and prerequisites for outstanding cognitive development within a specific domain. In this chapter we will analyse different aspects of this phenomenon.