The starting point for the epistemology i will present here -- which applies to all animals, including humans -- is the concept of Umwelt. According to Jacob von Uexküll [48,49], each animal has its own Umwelt (surrounding world) with itself at the centre. This `world' consists of those factors in the surroundings which the animal is able to recognise and react to in a differentiated manner ( Merkwelt), and those factors towards which it directs its activities ( Wirkungswelt). It can only be conceived of as part of a whole which includes the animal itself. For our purposes it is important to note that the factors entering into the Umwelt -- what is relevant to the animal -- can be very different for different animals. It is also important to be aware that the Umwelt is not something merely `given' to the animal from outside -- it is shaped by the life form of the animal; or to be more precise: Umwelt and life form ( Lebensform) have been formed in a mutual process.
An important feature of this Umwelt is that the phenomena are in no way purely passive qualities (registered by the animal) -- they will from the outset be imbued with (practical) significance. They come to belong to the Umwelt for precisely this reason -- the Umwelt arises through an evolutionary process where those animals survive who are best able to distinguish between and react adequately to those features of the environment that are relevant to their lives. This significance is most direct in the simplest life forms, where almost every feature of the Umwelt corresponds to a unique action. As we move towards more flexible forms of life -- which are able to function in varying conditions -- this direct connection between sensation and action becomes weaker, and perception more internalised, while the nature of the significance changes from immediate towards potential. But it will never disappear completely.
I will now identify the Umwelt, which can be constructed (more or less) objectively from studies of the physiology and behaviour of the animal, with the animals lifeworld ( Lebenswelt), which is the world as the animal itself experiences and lives in it. The life world is a web of expectations, significant patterns, and objects the animal relates to. It is quite true, as Merleau-Ponty stresses, that it is impossible to arrive at the lifeworld by studying the excitations of the sensory nerve cells -- but then this does not tell us much about the Umwelt either. In both cases what we must look at is the overall relation to the environment.
Thus i believe it is reasonable to claim that lifeworld and Umwelt are essentially the same thing, as seen from the subjective and objective points of view respectively. This differentiation and subsequent identification requires the web which is the lifeworld to be sufficiently complex, and the animals relation to its environment to be internalised to a high degree. Uexküll himself [49] goes quite far in the direction of seeing the Umwelt as a subjectively experienced world.
If we depart from Uexkülls definitions, and instead define Umwelt as everything that influences or has importance in the life of the animal, lifeworld and Umwelt will again be different. This concept of Umwelt is epistemologically more interesting, since it enables us to talk about more or less adequate cognition of the Umwelt - while retaining the focus on what is relevant to the animal.
From this account it emerges that mental states are predominantly intentional -- they are primarily directed towards the subjects Umwelt. The starting point for all mental activity is, after all, how we perform in our environment -- and the aim of this activity is adequate action in a world which is adequately perceived. It is obvious that this only makes sense if the mental states are experienced as relating directly to the outside world without any ifs or buts -- ie., we do not experience mental states as such, but we relate to the world through them. States that are primarily non-intentional also function within an intentional framework. They often appear as dispositions for action: pain, for instance, is a signal that i should withdraw or take it easy, while hunger is a signal that i should find something to eat. As i said above, there are no such things as neutral, isolated impressions or qualia -- they are all woven into a web of significance.
How, then, is it possible to have adequate cognition of the Umwelt or environment? First, allow me to point out the obvious: an ability to adequately perceive and understand the relevant features of the environment has a strong positive selection value (it greatly increases the animals fitness), while inadequate understanding will be correspondingly negative. Because of this, evolution will tend to bring forth processes which reliably generate correct cognition of the Umwelt of the animal (subject).
It is clear that all that counts in the context of natural selection is the reliability of the process and the correctness of the resulting beliefs or attitudes. Reliability here includes both the process reliably producing correct cognition when it occurs, and that it occurs with a sufficient regularity to be relied upon. The weight put on these two criteria will vary -- it is unclear what confers greater fitness of a process that can hardly go wrong, but only occurs once in a blue moon, and another which is far more uncertain, but will be invoked in almost all relevant situations. If we decide to call these processes epistemic -- and i see little reason otherwise2.1 -- it shows that cognition not only may occur among other animals, but it would be quite unresonable to assume that it does not occur. The advantage of symbolic thought, which is assumed to be the unique domain of humankind, is that it allows us to use the same processes in many different situations. However, in order to guarantee that the resulting beliefs are correct, input from reliable sources is required.
If we look at this more closely, we find (as i have indicated) that the proper object of knowledge must be the Umwelt, not the surroundings. Knowing all kinds of irrelevant details of the surroundings confers no advantage; perhaps it is even a disadvantage, since it distracts attention from more important things. And the relevant features of the surroundings is precisely what constitutes the Umwelt.
But then, you may object, it cannot be real knowledge after all! What characterises real knowledge is objectivity, that it is knowledge of the same, objective world, independent of the subject; you can only have real knowledge when you have abstracted from your own, contingent position. To this, one could ask rhetorically, is human knowledge really that objective? -- or do we just perceive and underestand the human Umwelt?
Let us not dwell on this, but concentrate on the main defect of the objection, which is that it (like physicalism and behaviourism) falsely identifies the real with the objective. The basic requirement of knowledge must be that it is knowledge of something real, not that it is completely objective. The problem with egocentrism is not its subjectivity per se, but that i believe the world appears to everyone the way it appears to me -- or that only my perspective reveals the real world. The aspect of the world that i see is real, but it is not the only one. Even though the Umwelt varies from one animal to the next, it does not become less real.
But when i say this, do i not end up in a kind of relativism, which makes it impossible to talk about truth and reality, since there is no common reality? No; even though there is a different Umwelt for each animal, the same things will enter into all of them. For instance, an animal is often part of the Umwelt of several other animals, where it can be a threat, protection, food, landscape, thing etc. But it is the same animal that has all these rôles. The features that enter into different Umwelten are all equally real aspects of the thing or how the world is. Furthermore, from a biological point of view it is clear that there is no room for either idealism or relativism of the kind `one thing is as good as the other': the Umwelt for each individual animal is a hard fact, which it just has to cope with. But since different animals have different ecological niches and are unequally equipped, these hard facts will also vary.
My main point is that we must take seriously what acts as reality for the animal, what is relevant to it,and not fall into ther trap of what Arne Næss [34, 16] calls labyrinth epistemology -- that we decide what kinds of challenges and facts the animal should repond to (like rats in a labyrinth), and judge their understanding or behaviour as adequate or inadequate according to our Umwelt (as if we were some kind of omniscient, all-encompassing subjects), not according to what is real for the animal. What is irrelevant, can be true or untrue; it is still irrelevant. Out Umwelt (which we have a tendency to consider `objective') includes phenomena which do not enter into the world of the bat, and its world includes phenomena which are not in ours. Sonar properties are just as real as colours, but to not exist for us.
But could we not still talk about objective knowledge above and apart from all the specific lifeworlds, as a higher level, which is something that is at least approached by human, scientific knowledge? And do i not presuppose an objective reality when i talk about `the surroundings'? Not necessarily. The surroundings are best understood as the totality of all possible perspectives of Umwelten, encompassing an infinity of aspects. Our Umwelten will encompass more or less of the surroundings, but the surroundings themselves cannot be captured by any possible experience. On the other hand, we can think of the surroundings as a necessary limiting concept, or -- in kantian terms - as a regulative idea.
If we go to the opposite extreme, the scientific knowledge that will abstract from any point of view, we find that this concept of objectivity is of no help in biologically oriented investigations, and cannot be anything but a limiting concept either. Pure physicalism, which is what this implies, has no room for the organism as a natural unit, or for any other natural units. The epistemic subject dissolves together with all of the known world. Arthur Eddington [10, introduction] describes it in the following way:
`In the world of physics we watch a shadowgraph performance of familiar life. The shadow of my elbow rests on the shadow table as the shadow ink flows over the shadow paper ... The frank realisation that physical science is concerned with a world of shadows is one of the most important of recent advances.'
The Umwelt of the simplest life forms consists of a small number of features, with a one-to-one correspondence between stimulus (feature of the sensory world) and response. Each phenomenon in the Umwelt has its specific significance, in the shape of one reaction, which in its turn defines the phenomenon. Two phenomena with the same significance in one respect, are the same phenomenon. Or to put it another way: the phenomena are uniquely determined by one and only one `property'. Admittedly, there may also be cases of completely separate chains of reflexes, where the same object is treated according to two (or more) distinct stimuli, which in turn lead to two distinct reactions to the same object. But in this case the behaviour and cognition of the animal are clearly not coordinated.
Internalisation of cognition breaks the direct link between stimulus and response. Information from sensory organs is now processed and organised according to schemata representing important types of states of affairs and patterns of reaction. Each schema may consolidate information from several sensory sources, resulting in a coordinated movement. Here we may already find something which can evolve into categorisation, or ordering of the Umwelt into stable, general, meaningful patterns.
But there is still a long way to go. A tentative account of categorisation may shed some light on this. Firstly, it implies that you realise that two different phenomena or states of affairs belong to the same category -- in other words, you must not only assimilate and consolidate information according to one schema, but consolidate it in different ways, so that both differences and similarities emerge. A fish will probably see a large moving object (or a corresponding electromagnetic field) as a threat, and react by fleeing or hiding, while smaller objects are potential food. But can it differentiate between large objects and see them as anything apart from only a threat? Will it, while maintaining awareness of the threat, be able to take other feature of the object or situation into account -- features which might be of some importance -- or will all available information be `swallowed' by the threat schema? This would require a form of object-consciousness, which in a way entails two schematisations: one which indicates what counts as objects and other things we are aware of, while the other categorises these objects, situations etc., placing them into significant patterns of life. This is not a two-stage process, but a parallel process, since factors that we are in no way consciously aware of may be relevant to the categorising.
Secondly, categorisation in a true sense is an activity where the situation or the object receives a meaning which, although it is an extension of the immediate action-significance, is also partly separate from it.2.2 The subject has a certain `control' over its categories -- they are formed without any (more or less) stereotypical action automatically being attached, and they may have several different `uses', not only the primary one(s). We can see this in laboratory experiments, where categories are formed and used even though the natural context is almost completely absent. An extension of this is our ability to form categories that have hardly any place or use in natural life.
One reason why i emphasise this is that it shows how categories (in this sense) and our concepts are related. We may think of concepts as verbalised categories. It also turns out that categorisation, thus understood, is a fairly advanced epistemic process, which does not appear to occur in fish and invertebrates. However, many of them can discriminate -- between certain objects, properties and situations. This is a sign that the Umwelt is experienced `theoreticaly', as consisting of objects and situations, not only through primitive schemata like food, threats, etc.
Categorisation implies a qualitative leap compared to discrimination, in that we are able to assimilate situations, properties, relations and objects we have not yet come across, to a category we are already in posession of. Thus we are in general able to master the distinction between what does and what does not belong to this category. Discrimination only means being able to distinguish between each single concrete case -- and the ability to simultaneously remember a large number of such distinctions is limited [46]. One example which illustrates this quite well, is a set of experiments [19] (see also [40,46]) where pigeons were shown slides of trees, people, water etc. After they had been shown some slides and learnt that (for example) they should peck on the slides showing trees, they would peck on all pictures of trees, even those they have never seen before. Considering this, as well as the large number (several hundreds) of pictures used, and the great variation in the motives (eg, the tree-slides showed many kinds of trees, and whole trees as well as parts of trees), it seems unreasonable to claim that the pigeons learned to distinguish between picture A and picture B, A and C, C and D, etc. -- or that they used particular properties (stimuli) of the pictures to distinguish between them. Nor was the ability to recognise a tree as a tree founded in any lack of ability to distinguish between different trees (or pictures of trees) -- they could learn different kinds of discrimination and categorise in different ways.
It can be useful to distinguish between natural and acquired categories. Natural categories are those that have an immediate significance or are important in the Umwelt of the animal -- categories which it is important for the animal to master in order to survive or function properly, which it will therefore be genetically disposed to learn or master. In addition to these, the animals may acquire categories that may be useful in various situations, enabling them to adapt to different environments. In the aformentioned experiment, trees and water are examples of natural categories. The pigeons learned to react correctly to these very quickly, and had an extensive and precise mastery of them: drops of water as well as oceans were correctly categorised as water, while they distinguished between celery sticks and trees. Given more time, they could identify an individual human being, or distinguish between the letter A in different fonts and the number 2 -- all examples of acquired categories.
A reasonable assumption is that the natural categories emerge from `connections' of the nervous system which are either innate or developed during the animals formative period. They may be seen as `ready-made' transformations of `data'. In addition, possibilities exist for learning new ways of organising experience, by developing new connections. Categorisation will thus be an implicit skill, which cannot be formulated as explicit rules for symbolic manipulation, nor can the categories be `defined' by a set of properties. The constraints on this kind of learning, such as the number of `dimensions' according to which experience can be organised, whill be of a fairly general kind.
The categories can be of different types. In the pigeon experiment we have seen classification categories, in the shape of general classes of objects (trees, water) or individuals (one person). Natural categories for particulars often arise in one of two ways. Firstly, the advantages of good geographical and topographical knowledge leads to the ability to recognise individual things like rivers, trees, small hills etc. Secondly, having to keep track of social structures and rôles means that you will recognise individuals, especially within your own flock. More complicated classification categories, relating to properties or situations, may also be found. Another important type of categories are relations, such as `greater/less than' or `similarity/difference'. These have so far only been clearly identified in primates [46].
Is it possible to construct a hierarchy of cognitive skills? Is it possible to rank the animal species, and say that one animal is more intelligent than the other? Is it possible to find a clear definition, clear and unambiguous criteria or an unambiguous measure of intelligence, which makes such a ranking possible, and does not just serve to confirm predetermined conclusions?
Some people will answer these questions in the negative. They will point out that every animal has its unique ecological niche and Umwelt, where it alone is completely at home. No other animal can function any better in this ecological niche. Uexküll says, accordingly [48, p. 4], `Ein Pferd [würde] de Rolle eines Regenwurms nur sehr unvollkommen ausfüllen.' If we define biological intelligence according to how well the animal functions in its Umwelt or ecological niche, the natural conclusion is that every animal is the smartest.
This conclusion gains further currency if we look at the character of the evolutionary process which has produced all the various species. Natural selection works equally on all phylogenetic strands towards producing more well-adapted varieties, and where a branching occurs, this is because both varieties are well-adapted, each in its own way. All present species may thus be considered the pinnacle of evolution. There is no evolutionary ladder.
However, common sense supports a different conclusion. Some animals are smarter than others; there are dumb animals. The earthworm is dumber than the fish, which is more stupid than the dog, which again is less smart than the chimpanzee, which is not quite as mentally able as the human. Is this belief merely an effect of anthropocentrism, of us uncritically applying our own standards to other animals? Can we really be so wrong in our own assessments?
Both arguments touch on something significant. I do not believe it is possible to make a single, unambiguous ranking -- the world is simply not that simple and one-dimensional. It is even less possible to give a numerical measure -- it should be sufficient to consider the difficulties we have in agreeing on an adequate measure of human intelligence to be convinced of this. Experience with humans also shows that bias in the definitions is a very real possibility. IQ tests have had a striking tendency to favour white, middle-class males -- and who have made the tests? We must accept that there is a multitude of cognitive abilities, and an even greater variety of affairs to be mastered.
Although there are great qualitative variations in the abilities of the species, it does not follow that it is impossible to compare them in a general and approximate manner, as long as we do not try to draw too many conclusions from these comparisons. Species may be generally smarter or dumber, even though each has its own speciality. In particular, it does not follow that an animal has any greater intrinsic value because it is smarter. It is also problematic to link intelligence as closely to fitness as the concept of biological intelligence does, since fitness or the ability to function adequately in the environment is often purely genetically determined -- whereas cognition and intelligence are usually associated with conscious thought.
It is, however, wrong to claim that genetic factors cannot contribute to cognition and intelligence. For instance, a well-developed and reliable sensory apparatus relative to the demands of the environment contributes to an increase in capacity for knowledge. The organisation of experience is also in part genetically determined, in the form of guidelines for formation of categories, among other things. But for us to meaningfully speak of cognition, the unconscious and subconscious must be expressed in or contribute to the conscious level. Completely unconscious cognition or understanding is at least not what i am primarily interested in here.
One starting point for a general ranking is that a richer and more varied cognition or understanding is by and large a better understanding. I am of course not only talking about the amount of data that can be absorbed, but also the possibility of having many kinds of experiences and organising experience in different ways. There is also a biological point to this: through a variety of cognitive skills, the animal is able to function in a wider range of circumstances.2.3 But this variety may still be realised in a wide range of different ways in different animals.
Harry J. Jerison [21,23] has a theory of intelligence wich can be seen as a concretisation of this. In fact, he operates with not just a ranking, but a measure of intelligence -- the encephalisation quotient EQ, which is the ratio of the size of that part of the brain not used for controlling basic bodily functions, to body size. This measure is, clearly, closely related to information processing capacity, and it also provides us with a phenotypic expression of intelligence.2.4 We must of course consider this measure only as an approximate indicator. What is interesting is that it implies that there may be or even must be many different forms of intelligence -- information from the environment can be treated in many different ways. Or to put it another way: different species may experience and structure their Umwelt in widely differing ways; they may create radically different models of the Umwelt -- or, they may constuct completly different or even incommensurate lifeworlds.2.5
To illustrate this, we may compare humans and dolphins, which have approximately the same EQ. The human lifeworld is to a large extent lingusitic -- not only do we communicate through language, but to a large extent we also structure the world through it. The dolphin may well structure the world in completely different ways. But whether or not they have any lingustic abilities, they have as good a knowledge and understanding as us, although we may never be able to understand them.2.6
In traditional, experimental psychology, intelligence is often by definition linked to learning ability, which is taken to include problem solving skills. (This is of course because learning is something which is easy to control experimentally.) In this way, the idea that intelligence and more well-developed cognitive skills are related to increased flexibility and ability to adapt to different circumstances, is maintained. However, the link between intelligence and knowledge is partly broken, and intelligence is no longer necessarily linkted to consciousness. The simplest learning mechanisms require no consciousness whatsoever.
Learning theory may still shed light on the problem of classifying the cognitive abilities of animals. In particular, this is the case when the study of learning uncovers different ways of organising experience. It is then possible to construct a hierarchy of cognitive abilities, from the simplest schematisation, linked to motor reflexes, to problem solving by insight and reflection. The categorisation of the pigeons is one level of this hierarchy, as is the learning by insight of Köhlers apes. It is also possible to gain some insight into the biological contexts and limitations within wich learning occurs, and what cannot be hierarchically ordered. I will not discuss this any further here.
It is possibly to extend learning theory in such a way that evolution itself also appears as a form of learning, and biological intelligence is once more considered `real' knowledge. This is motivated by the clear structural similarities between evolution and learning (and cultural development), noted already by Spencer and James. H. G. Plotkin [36,37] has developed such a model, where learning or evolution occur on four levels: genetic, ontogenetic, individual-cognitive and sociocultural. The first level is the ususal darwinian evolution, while almost all traditional learning theory has the third level as its subject matter. Through the processes of these four levels the organisms acquire knowledge of their Umwelt or environment, knowledge which can be used to tackle the problems of life. Jerisons model may also be incorporated here -- primarily as an interplay between the first and third levels.
The higher levels have the advantage of enabling the animals better to keep up with rapid changes in the environment, while the lower levels -- especially the genetic -- are far more reliable and require less energy. The lower levels also form the basis for the higher ones -- in other words, it is the genetically determined Umwelt which is the most fundamental.
A qualitatively new level in the organisation of the lifeworld is fake activity, such as play and deception. This implies that the animal has not only categorised the Umwelt, enabling it to treat it `at a distance'; it can also consider its own activity as an object of categorisation and reflection. Not only is the direct, immediate action-significance of the external world weakened, but so too is the direct significance or meaning of its own actions. It is now possible to be perform an act without it carrying the usual meaning, ie., to perform it without really performing it. This kind of fake action is an important precondition for many forms of complex social interaction, as well as for symbolic action and thought, internalisation of thinking, and awareness of mental states. In the latter context it is useful to distinguish between different levels of intentionality:
Play is a kind of activity that is performed without a conclusion; it is a kind of imitation of an ordinary life activity or a part thereof, but, contrary to this activity, it is not meant to produce anything -- it is not (directly) linked to the other activities of life. The point of play is that it is pointless -- the moment it gets a point or rationale beyond itself, a conscious function or a purpose in life, it becomes serious and ceases to be play. Play is supposed to be an area where the hard facts of life do not intrude -- it is supposed to be `isolated', its whole point being the performance of the play. This does not mean that play is meaningless -- as an imitation of ordinary activity, it carries with it the meaning of that activity. In other words, the activity is played out and understood, but with the understanding that this is not serious. Play can be carried out in solitude or together with others. A more finely meshed analysis can be found in [30].
Play can be found in the repertoire of a number of mammal and bird species, especially among young individuals. Most of us know the playfulness of cats and kittens; here we clearly find imitations of fighting and hunting, while there is no doubt that we are not looking at real fighting or hunting. Among monkeys we can find quite complicated patterns of play.
Play could hardly have evolved if it did not have some function in life, although this function cannot be the purpose of play, if it is to be real play (however, see [42]). One aspect is the function play has as a `refuge' -- when we play, we feel free to do things without worrying about the consequences. This can be important when `real life' really feels really serious. Another, related aspect, which has often been pointed out, is the function it has as preparation for the real activity. In the sheltered world of the young mammal it is possible to acquire considerable skills through play, without running the risks and paying the costs attached to learning them `the hard way'.
Play also has an important rôle in establishing social relations and social intelligence. Through social play we get to know each other, learn to anticipate the reactions of ourselves and others, and become accustomed to the social lifeworld -- Piaget puts great emphasis on the rôle play performs in creating a decentred self. Play is also a prime example of rule-based activity or praxis (rule, in the Wittgensteinian sense). We limit ourselves in play -- it is the rules of the game, not brute realities, that limit what we do. Play serves to acquaint us with such activity, as well as introducing us to social rules.
Symbolic action differs from play in having its purpose kept intact. But, rather than concluding the activity, a `picture' of it is performed; it is not allowed to become completely serious. When establishing group hierarchies, for instance, it is better to have a pretend fight. Wolves engage in fights which are interrupted at the moment one of the individuals shows signs of giving in. Then, the other will not (ab)use its position to inflict further harm on its opponent. In general, it is the case that a considerable amount of aggression can take place without social bonds suffering.
We can make a distinction between on the one hand mimetic activity, where the `picture' clearly resembles the original activity, and on the other hand completely symbolic action, where it is replaced with conventional symbols. There may be a sliding scale between the two.
What happens here is that the original activity is partly played out, with its original meaning intact, but with a somewhat limited purpose -- placing bounds on the activity. Instead of the original activity in its entirety, a representation of it is performed -- a replacement activity. You symbolise what you would have done if the situation were real. The representation can be more or less complete, from a near faithful copy to mere indicators. The latter approaches symbolic activity proper.
However, does this not require a considerably more complex intentional system than what we may assume non-human animals possess? A reconstruction of a playing or symbolising activity D toward b could be that a performs D toward b without taking it seriously, while knowing b to understand that a does not mean it to be serious. This is third order intentionality, and essentially the same structure as appears in a serious conversation between humans.
I do not know how well this analysis really fits, although i will not go to great lengths to refute it. One problem is that we might then question whether small children are really playing -- and if anyone plays, it is they.... The crux of the issue is whether it is necessary that a knows that b is aware that a is not serious. When it comes to symbolic activity proper, it makes sense to assume that it must involve such a mutual recognition, but in this case we are approaching full-scale communication. But otherwise, is it not sufficient that a performs the activity as an imitation, and that the situation guarantees that b does the same, so a can safely assume that b is also playing? The existence of solitary play appears to support this conjecture. In addition, the clear position that play and symbolic activity have within a complex social framework indicate that they are identified largely from this position. (It is precisely because play is freed from the usual constraints of life that it must belong within a social framework.) Features of the social situation serve to identify and guarantee the play situation, while the physical situation may be identical with that in which the ordinary activity takes place. An identification beyond the situation is really required only when the possibility exists of deception and exploitation. This fits well with my assumption (which i will develop in the next section) that deception is the starting point for higher order intentionality.
It should be noted that many situations which may look like examples of symbolic activity, may be better explained by physiological mechanisms developed by direct selection. This is especially the case for play to attract females (such as practiced eg. by the black grouse), but also territorial behaviour. These activities have more in common with mating dances, which are also performed by insects, than with real symbolic behaviour. So we are not obliged to postulate complex intentional systems in all these cases.
Deception or simulation is an activity which is performed to induce false beliefs in a counterpart, thereby misleading it. The agent is in other words aware of the action, its meaning and rôle, as well as what it means to the counterpart,2.7 but performs it with a quite different purpose -- or rather, merely as a means to a different end or purpose, viz, to make the opponent behave in a certain way. This behaviour thereby differs clearly from play and symbolic action, which are performed without ulterior motives.
An often cited example is the plover, which simulates a broken wing when birds of prey approach the (ground) nest [39,40], until they are at a safe distance. At this point the plover takes off and flies away. In this case, however, we need not assume that the plover has any knowledge of the bird of prey as an intentional agent -- the most natural assumption is that it knows that such a behaviour (which it may or may not recognise as feigning injury) leads the bird of prey away from the nest. Deception proper, on the other hand, is in evidence among some primates, eg., female baboons who take great care to ensure the dominant male does not suspect anything when they mate with younger males [26]. Laboratory experiments [38] may yield firmer evidence of the level of intentionality -- although they will be plagued by the oversimplification which is the typical weakness of all laboratory experiments. Alternatively, the `Sherlock Holmes' method proposed by Dennett [9] for collecting `anecdotal evidence' may also be employed.
There are obvious advantages in deceptive behaviour for survival and reproduction -- someone who can manipulate rivals and enemies into acting inadequately wins a clear advantage for him or herself. What is particularly interesting is that it leads to a kind of arms race, where on both sides there is selection for (respectively) uncovering deception and hiding deceptive intentions. There are grounds for believing that this mechanism is a primary driving force in the evolution of higher order intentionality and complex social relations -- which again demand high levels of intelligence to master. This, combined with issues related to altruism, has probably played an important rôle in the evolution of human intelligence [47,3]. The level of intentionality which can be mastered will vary with the situation -- as we know from ourselves, there is no clear and unambiguous limit. The ability to conceal our intentions, combined with the advantages of `trying out' alternative actions in our heads rather than in concrete reality, will be essential to internalisation of thought, and thereby for the establishment of a self.
If this account is correct, it is a strong argument for the impossibility of solipsism. A solipsist self-awareness would be a form of second order intentionality, but second order intentionality is evolved in order better to predict and manipulate the actions of other animals. Because `mentalism' is so obviously adaptive, behaviourist scepticism is removed long ago, and solipsist scepticism can never arise. Only when we have learnt to recognise the beliefs and intentions of others can we look at our own with a critical eye; on its own this is of little adaptive value.