Response to all other signals is optional. The fact that animals respond to a much wider range of events than those that immediately affect their physical well-being is a reflection of the non-random nature of the natural world. Stimuli of whatever sort may have predictive value, and a creature that responds appropriately can often avoid future unpleasantness and ensure the regular enjoyment of the better things in life, to the immediate benefit of itself and the ultimate survival of its species.
Learning is largely concerned with establishing the predictive value of stimuli that do not of themselves demand responses. We recognize other animals as intelligent when they predict effectively from these stimuli, and as highly intelligent when they begin to predict by analogy, generalizing to the point where they can take appropriate pre-emptive action without prior experience of the particular sequence of events to which they are reacting.
All invertebrate animals learn, many show glimmerings of intelligence, and some would qualify as highly intelligent by the definitions employed above. A major problem, as we shall see below, is that we are often ourselves insufficiently intelligent to devise situations that will fairly test their performance. This problem becomes particularly acute when an invertebrate is making an elaborate response in an apparently complex situation, so it is perhaps most appropriate to begin by considering 'simple' forms of learning in which it does seem to be possible (i) to define the stimuli to which the animals are responding and (ii) to recognize the likely advantage of the behaviour in question.
One such category is habituation. Animals, in general, soon cease to respond to stimuli that prove to have no predictive value. A shadow passes. It could signal the approach of a would-be predator. In the past, individuals that played safe and ducked, or froze, lived to breed. Their offspring are liable to do the same; caution is genetically determined, natural selection having eliminated the unwary over countless generations in the past history of the species. But a shadow is not invariably dangerous. It may signal no more than a passing cloud, or the waving of a frond of seaweed. An animal that reacts to every moving shadow is doomed to a restless and economically hopeless existence, and cannot effectively compete with its less wary neighbours; in the long run the overcautious are eliminated as surely as the foolhardy. Animals must habituate if they are to remain in business, and the best assumption with which genetics can equip them is that the immediate future is likely to resemble the immediate past. If a stimulus recurs regularly, unaccompanied by consequences of importance to the animal, it is well to ignore it. Rates of habituation vary, as we might expect, because the degree of built-in caution will vary from one sort of stimulus to the next, but caution is always present.
In an opposite direction, we find that nearly all animals will sensitize, becoming more rather than less responsive in circumstances where anything that happens is likely to have predictive value. If recent experience has revealed that something of importance is going on, it is well for the animal to remain more than usually alert, so that it responds to stimuli that it might otherwise have ignored. An animal that has just been hurt will flinch at stimuli that have nothing to do with the damage, and an animal that has just engulfed a tasty mouthful is more than usually attentive; any event could be a signal that more of the same is in the offing. It should be noticed that such effects can be quite unspecific: if things are going badly for the animal it will sensitize in the direction of caution, and take no chances over stimuli that would normally evoke a positive reaction; and if events are proving favourable, it becomes particularly responsive to stimuli that may indicate desirable objectives, like food, or sex. The animal, in either case, is responding to the sum of events (good and bad) in its recent past, and adjusting its response levels accordingly.
Taken together, habituation and sensitization will ensure that an animal behaves economically and opportunistically, cashing in when the going is good and lying low when conditions are unfavourable. The two interact: dishabituation, where a sudden change in circumstances re-establishes responses to a repeated stimulus that the animal has come to ignore, is a special case of sensitization. Between them the twin processes of habituation and sensitization will ensure better-than-even odds on an animal's responding appropriately, even under conditions where the creature is quite incapable of determining the precise nature of the stimulus to which it is responding. The system works, provided only that events occur in a non-random sequence — provided, in short, that the future is likely to resemble the recent past. Inside the laboratory it may, for various reasons, be desirable to randomize trial sequences so that an animal cannot predict what is going to happen without precise identification of the stimuli offered to it, but in the wild the animal is never likely to encounter a random sequence of events. Predators hang about, or go away; food comes in patches, scattered in time and space but never quite at random.
It is relevant to an understanding of invertebrate learning and behaviour to realize that cold-blooded animals in general can afford to be opportunists to an extent unthinkable in a homoiotherm. We poor warm-blooded creatures consume fuel so fast that we tend to discuss motivation in terms of imperatives: hunger must be satisfied or we die. We get thirsty because we live above the temperature of our surroundings and evaporate our body water; the need to replenish the water becomes an imperative because we overheat and die if we fail to do so. This personal experience distorts our thinking about invertebrates; we tend to assume that a starved snail or a thirsty cockroach will take risks to satisfy its needs as we would, becoming increasingly desperate as time passes. There is no evidence for this, at least on the sort of time scales that one would expect from an experience of rats and people. A fed octopus is more, rather than less, liable to attack a strange but potentially edible object lowered into its tank; it cashes in while the going is good, and can starve for weeks if conditions suggest that it is unprofitable or even dangerous to respond to the unfamiliar. Sensitization is relatively unimportant in ourselves, because we can rarely afford to accept its dictates for very long; invertebrates can lie low for months, then stuff themselves, mate, and multiply when conditions improve.
One consequence of this is that it may be difficult to investigate invertebrate learning by conventional methods, forcing the animal to become active and solve problems because it is hungry or thirsty.
A further difficulty that confronts any biologist rash enough to examine learning in the lower animals is that most of them depend predominantly upon chemical stimuli, which we find awkward to classify and almost impossible to measure in the laboratory, let alone in the wild. Experimenters tend, in consequence, to set problems based on visual and spatial cues, so that their unfortunate subjects are more often than not tested with learning problems based on our sensory capabilities rather than theirs, a situation that can only lead to persistent underestimation of their learning abilities.
At a somewhat more subtle level, it is easy to forget that the bodily construction of animals may itself preclude certain forms of learning. Many of the things that we do, and believe simple, are based on an awareness of bodily position that is probably lacking in most invertebrates. When we move, we know how the position of our limbs is changing. If I pick up and examine an object, I am continually moving my fingers; each time I change my grip I know about the new positions adopted and, putting this information together with the feel of the contacts made, I can build up a mental picture of the shape of the thing I am holding. An octopus, for example, apparently cannot. It is quite as capable as I am of feeling over an object that it touches, and there is no doubt whatever that the animal can learn by touch, since it can be taught by simple reward and punishment techniques to distinguish between a wide variety of different textures and tastes (chemoreception again!), learning to perform such discriminations with great accuracy after a dozen or so trials. But it cannot manage shapes. A cube and a sphere are alike to it, apparently because the animal has no means of assessing the relative positions of its equivalent of our fingers, the many suckers arrayed along its flexible arms.
On reflection this is perhaps not very surprising. It is the animal's very flexibility that defeats it. We can compute the precise position of our hands because the human body only bends in a few places and is even there restricted to movement in a few directions. Sense organs in and around our joints can tell us about the angle adopted at each. No such easy computation is possible in a soft-bodied animal, where the position of each part depends upon the degree of contraction or relaxation of muscles all over the rest of the body. Muscle stretch receptors, which all have, are useless in this respect since they can only signal tension in the muscle concerned, not the position achieved as a result. The jointless animal perforce lacks any equivalent of our proprioceptive sense of position.
This divides the animal kingdom rather abruptly into two groups. On the one hand, there are the jointed animals, the vertebrates and the arthropods, which seem to have a double (muscle and joint) set of proprioceptors. These creatures are potentially able to learn to manipulate; they can discover by trial and error precisely which movements it pays to repeat; they can monitor the number of steps they have taken and the angle of any turns that they have made. On the other hand, there is the world of the soft-bodied, forever debarred from learning to make skilled movements, and faced with learning to find their way about entirely from exteroceptive cues. Arthropods, like vertebrates, can readily learn to run mazes, while many of the other things that they do or make (such as honeycombs and spiders' webs) necessitate accurate measurement of lengths and angles. A hermit crab, investigating the inside of an empty shell with its claw, plainly learns about the size and the shape of the hole it is examining before risking exposure of its soft abdomen as it quits its old home in favour of the new. In marked contrast to all this, the soft-bodied snails and worms (and octopuses) rarely succeed in mastering any maze more complex than a T or Y, they never seem to create patterned structures, and they never, so far as we can assess the matter, learn to carry out a skilled movement. As a result we tend to assess their learning capabilities and their intelligence generally as exceedingly limited compared with arthropods and our fellow vertebrates, which find easy the same sorts of tasks as we do ourselves. It may well be, indeed it seems very likely, that the individual adaptive capacity of worms and snails is very limited; but it is well to remember that most of our present evidence comes from tests appropriate to ourselves rather than the creatures we have studied.
Granted, then, that we are almost certainly underestimating the capacities of most invertebrate animals, what generalities can be made in comparing higher forms of learning in vertebrates and invertebrates? Can we, for example, detect signs of latent learning ('the association of indifferent stimuli in situations without patent reward') or insight learning ('the production of a new adaptive response as a result of the apprehension of relations') among invertebrates, as we can among the higher vertebrates?
The answer appears to be 'yes', if one searches. Many invertebrates, inevitably, show little sign of these more complex forms of learning. Any animal that lives a highly specialized existence (an aphid on a rose bush or a lugworm in the mud of a saltmarsh) is unlikely to show conspicuous signs of intelligence. It does not need to: the same very limited range of problems has cropped up generation after generation for so long that almost the totality of its behaviour has come to be programmed genetically. Wherever this is possible, because the future has always rather precisely resembled the past, learning, with its inherent capacity for errors that could be dangerous or even fatal, is plainly a luxury that will be eliminated in the course of natural selection.
To discover what invertebrates can really do, one is obliged to examine cases where the animals live in complex environments. In general, also, it is necessary to look at predators, because predators must always be a little brighter than the prey they feed upon. Two such animals will be considered below, one soft-bodied (a cephalopod, of course), the other a jointed animal, an insect.
The octopus can be taught to make a wide variety of visual and tactile discriminations in the laboratory and it learns rapidly in these trial-and-error situations. It is also plain, from its lifestyle in the sea (where it returns to a home in the rocks after foraging expeditions that may carry it quite far afield) that it is capable of learning its way about a complex landscape, despite the restrictions already considered in relation to the consequences of flexibility. Observations of octopuses in the sea suggest that they can return home from any point in their range without retracing their outgoing steps, and this suggests a capacity for latent learning. In the laboratory octopuses will readily make detours to get at prey that they cannot approach directly; in a typical apparatus the animal was obliged to run out of sight of a crab, down an opaque corridor, and turn appropriately to reach its food. Even untrained octopuses manage this quite readily and will reach their goal with considerable reliability even if delayed for a minute or more by shutting them into the corridor. Plainly the animals are in some manner aware of the spatial relations of the crab and the various baffles that prevent them from reaching it — they show insight. (It is surprising how few vertebrates will detour successfully in similar circumstances.)
Performances that are in some ways even more impressive are shown by some of the hunting wasps. Ammophila, for example, hunts, paralyses, and stores caterpillars. Each mated female operates alone. She digs a hole in sand, covers it over, and, after a brief orientation flight in the vicinity of the nest, departs to search for caterpillars. A caterpillar is paralysed by a sting in the central nervous system and is thus preserved as a living food supply for the egg, which is subsequently deposited with it in the hole that the wasp has dug. The wasp's first problem, however, is to get the caterpillar home. Often the prey is too heavy to fly with, and the wasp sets off to drag it, often for many tens of metres, across the ground and round obstacles to the nest site. This performance is in itself remarkable because it implies a considerable knowledge of the geography of the district round the nest, apparently derived from the brief orientation flight carried out before setting out to hunt down a caterpillar — another nice example of latent learning. Extensive tests show that the cues used are entirely visual and more often than not precisely the sort of landmarks, skyline patterns and the like, that we would choose in the same circumstances.
More impressive still, however, is the discovery that Ammophila will run as many as three nests, in different stages of construction, at one and the same time. After the first egg and caterpillar have been deposited, she begins and provisions a new nest. A day or two later she returns to the first nest, checks whether the egg has hatched, and if it has, begins to stuff further caterpillars down the hole to feed her grub. Any spare time is spent digging out nest three, with a visit to nest two after a couple of days to see whether that egg has now begun to develop. Quite clearly the wasp not only remembers the precise position of each separate hole, but also recalls the state of play at each site, often delaying the response that she makes as a result of an inspection visit for two or three days on end.
Performances like those of Ammophila and the octopus leave no reason to doubt that learning by invertebrate animals can be every bit as complex as that found in vertebrates, which have been far more extensively studied. Whether one regards such creatures as 'intelligent' is largely a matter of taste, that is to say the precise definition of that somewhat elusive property one chooses to employ. We have come a long way from the proposition that the lower forms of life are automata, bereft of the ability to adapt and determine their individual destinies.
(Published 1987)
— Martin John Wells
- Bibliography
- Corning, W. C., et al. (eds.) (1973–5). Invertebrate Learning, 3 vols.
- Hinde, R. A. (1970). Animal Behaviour: A Synthesis of Ethology and Comparative Psychology (2nd edn.).
- Wells, M. J. (1978). Octopus: Physiology and Behaviour of an Advanced Invertebrate.
