Instincts are amazing. Instinct can enable an animal to do effortlessly what we can do only (if at all) through the exercise of reason and intelligence. Once upon a time, before Darwin, the apparent intelligence of instinctive behaviour was seen as evidence for design by God. Animals lacked true intelligence of their own, but had been endowed with instincts by an intelligent designer. Instincts were intelligence by proxy.
Darwin showed how to explain the appearance of intelligent design in the biological world, how to do so without taking on supernatural commitments. Biological traits that appear to be intelligently designed are in fact the natural products of cumulative natural selection. These traits are adaptations. Adaptations are, normally, adaptive, a fact which is accounted for by natural selection. Konrad Lorenz explains the point:
“If a biologist says that the cat has crooked, pointed claws ‘with which to catch mice,’ he is not professing a belief in a mystical teleology, but succinctly stating that catching mice is the function whose selection pressure caused the evolution of that particular form of claws.” (Lorenz 1965: 24)
Darwin considers that instincts are adaptations. He sketches an evolutionary explanation of instincts in Chapter 7 of The Origin of Species. There he offers the following account of an instinctive action, an account, he says, of the way the concept is understood by any informed observer of animal life (Darwin 1859/1968: 234):
“An action, which we ourselves should require experience to enable us to perform, when performed by an animal, more especially by a very young one, without any experience, and when performed by many individuals in the same way, without their knowing for what purpose it is performed, is usually said to be instinctive.”
Darwin here lists several defining marks of instinct. Instincts are: independent of experience; innate; stereotypical in their expression; species-typical; and nonpurposive. To these five marks, Darwin shortly adds two more: that instincts are inherited, and that they serve the animal’s welfare (Darwin 1859/1964: 235, 236). Any action that has all seven of these properties is a paradigm of an instinctive action. This is not to deny that there are many cases in which some but not all of these properties are found together, and that there are cases in which some or all of them are present to different degrees. It is the paradigm cases, the canonical examples of instinctive actions, that are our primary concern in this paper. These are the cases in which all of these properties are apparently exhibited by some pattern of behaviour.
Instincts in Classical Ethology
Classical ethology is an evolutionary science of instinctive behaviour. Not all animal behaviour is instinctive, of course. Lorenz says that some behaviour is ‘learned’ and some is the product of ‘insight’ (Lorenz 1970: 116). But in order to study the effects of evolution on behaviour, it is necessary to begin with the study of innate elements in behaviour. Those innate elements are instincts. Instincts are the results of natural selection acting on behaviour.
Instincts are expressed in what Lorenz calls Fixed Action Patterns. A paradigm example of a fixed action pattern is egg-rolling behaviour in the greylag goose (Lorenz 1970: 316-350). Geese retrieve eggs that have rolled from the nest by using a stereotyped sequence of movements. The goose reaches out and hooks the egg with her bill, and then pulls the egg back and into the nest. On every occasion on which she performs this activity, she executes the same patterns of movement: the form (or ‘shape’) of the behaviour is stereotyped. Every greylag goose retrieves eggs in essentially the same fashion: the form of the behaviour is species-typical. Another paradigm case is the stereotyped and species-typical pattern of courtship behaviour that is triggered in female sticklebacks by the sight of a male stickleback’s red belly (Tinbergen 1951: 37-8). Fixed action patterns like these are triggered by specific stimuli. The triggering stimulus, called a sign stimulus, is also stereotyped in form and species-typical: an egg-like object outside the greylag’s nest or an object that is coloured red below in the stickleback’s environment, in these two examples.
The fact that such action patterns are both stereotyped and species-typical is remarkable enough. But fixed action patterns have another remarkable property: they are normally correctly expressed by empirically naïve animals. That is, these patterns of behaviour are performed correctly by animals which have not had the opportunity to learn the correct patterns of behaviour from experience. It is as if empirically naïve female sticklebacks have a priori knowledge of the fact that having a red belly is a mark of a suitable mate.
Lorenz accepts the traditional explanation for the existence of these adaptive and unlearned patterns of behaviour. The traditional explanation is that instincts are innate. A trait is innate if it is a part of an animal’s natural endowment, biologically inherited from its parents. Innate traits are encoded in the genes and decoded during development. Lorenz thinks of innate traits as traits that develop endogenously. They unfold from within the organism during the course of its individual development. The contrast is with traits that are acquired, in some sense, from without, traits that develop in response to specific encounters with the environment.
Lorenz’s theory of instinct is the foundation of classical ethology. But the Grand Theory of classical ethology collapsed in the middle decades of the twentieth century. Lorenz’s understanding of innateness was especially harshly criticized. Yet Lorenz could not easily give up this concept, which is fundamental to his evolutionary understanding of behaviour. Evolution by natural selection produces adaptations, and adaptations are inherited by individuals as part of their native endowment: which is just to say (surely) that adaptations are innate. So Lorenz is not prepared to give up the concept of the innate.
The empirical facts which the Grand Theory sought to explain also survive the collapse of the theory itself. It remains more or less true that there are fixed action patterns in animal behaviour. Ethologists now rarely use the language of Fixed Action Pattern, Sign Stimulus, and so on, but the animals that Lorenz observed still sometimes behave, for all that, in stereotyped, species-typical ways, apparently without benefit of learning. This is an empirical fact that still stands in need of an explanation.
The Developmentalist Critique
The most important critic of Lorenz’s theory of instinct was Daniel Lehrman (1953). Lehrman agrees with Lorenz that naïve animals can express stereotyped and species-typical patterns of behaviour. But he disputes the interpretation of those behaviour patterns as innate (Lehrman 1953: 341). The patterns of behaviour that Lorenz describes as innate come to exist only because of processes of biological development that occur in individual animals. The patterns of behaviour do not exist in some hidden form at the start of that developmental process: they are not preformed in the zygote, or in the genes. Like every other part of the phenotype, ‘instinctive’ behaviour appears as a consequence of processes of development in the individual organism. Biological development is a process in which new structures and activity patterns appear as a product (not of interactions between heredity and environment or genes and environment) but of interactions among the existing structures and activities of the organism and its environment (345).
In every developmental process that has been closely studied, it has been found that environmental factors, either by their presence or by their absence, can prevent the development of a normal phenotype, says Lehrman. If an ‘innate’ trait is understood as a trait that is present in the organism independently of the action of environmental forces, a trait that unfolds in some sense from within the organism, then no trait is innate.
Nativism flourishes in the midst of dichotomies between innate and acquired, or instinct and learning, or heredity and environment, or maturation and learning; but none of these dichotomies have a place in developmental biology. There are genetic conditions for the development of any trait, and there are nongenetic conditions for the development of any trait. No trait is exclusively the product of environmental factors. No trait is exclusively the product of inherited factors. So says Lehrman.
Lehrman is, officially, neither a nativist nor an environmentalist. But he betrays a fondness for environmental explanations of behaviour that would have provoked Lorenz. Lehrman cites with approval Kuo’s explanation of the ‘instinctive’ pecking behaviour of young chicks as an after-effect of embryonic head bobbing caused by the chick’s own heartbeat while it is still in the egg (Lehrman 1953: 341-2). Kuo’s explanation makes it just a happy coincidence that the young chick’s pecking behaviour is useful. Lorenz thinks, to the contrary, that the chick’s pecking behaviour is an adaptation. The presence of this behaviour in the chick’s initial behavioural repertoire demands an evolutionary explanation. No matter how rich and complex the individual ontogeny of the chick might be, the evolution of the behaviour must also be considered, if the adaptedness of the behaviour is to be explained. That is why the nativist continues to insist on the presence of inherited factors in ontogeny (Lorenz 1965).
Instinctive traits have a phylogeny as well as an ontogeny. To deny this is to deny the evolutionary foundations of ethology (Lorenz 1965: 31). The developmentalist critique may have persuaded many animal psychologists to give up the concept of innateness (for example, Shettleworth 1998: 13-16). But it did not persuade Lorenz. To give up on innateness is, he thinks, to give up the thesis that adaptation has an evolutionary explanation.
The Virtues of Nativism
The concept of innateness is not a very robust theoretical concept. But it is a concept that is hard to avoid, given a few evolutionary assumptions. We have to explain empirical facts that are surprising. It is surprising that there exist stable patterns of behaviour that are both adaptive and unlearned. We are not surprised when learned behaviour is adaptive: any process in which behaviour is adaptively modified by individual experience counts as learning. But behaviour can be adaptively modified by learning only if the animal has suitable learning opportunities. Adult males of many species of songbirds know the correct songs to sing only because as young birds they heard the songs of adult males of their own species. They learned from exemplars what the correct songs sound like. What is surprising is that animals can sometimes behave adaptively even though they have not learned lessons from experience. The brown-headed cowbird, a brood parasite, produces a reasonable approximation to his species-normal song in the absence of experience of any exemplary songs (Shettleworth 1998: 457). Behavioural capacities such as these are traditionally described as innate. The term indicates that individual learning cannot explain the adaptedness of the behaviour. The adaptedness has a source other than lessons learned in individual experience. Darwin identified that source as natural selection, the source of all nonfortuitous adaptation. From an evolutionary perspective, it is no more surprising that animals have some innate behavioural capacities than that polar bears are naturally endowed with thick fur coats.
Lorenz Replies to Lehrman
Lorenz replies to Lehrman’s critique in his 1965 monograph Evolution and Modification of Behavior. Lorenz now concedes that it is never strictly correct to describe the manifest behaviour itself as innate (Lorenz 1965: 34). Nothing in the phenotype is produced entirely from internal, developmental resources, independently of all environmental forces. Therefore, the behavioural phenotype cannot properly be described as innate. Lorenz now accepts what Lehrman says about this (37).
Lorenz insists, however, that the empirical foundation of the classical theory is not damaged by Lerhman’s critique. There is still the need to explain the reliable reproduction, in each new generation, of stereotyped patterns of behaviour, in those cases in which some elements of the specific form of the behaviour cannot be explained by any suitably specific lessons learned from experience. Instinctive behaviour is reliably expressed in all or almost all healthy animals which have undergone development in evolutionarily normal environments. This is a generalization for which there is reasonable empirical evidence. Instinctive behaviour is regularly expressed in the species-specific, stereotypical form by naïve animals which have had no opportunity to learn the proper forms of behaviour from others of their species. This is also a reasonably well-confirmed generalization. Lorenz developed the technique of the isolation experiment in order to show that instinctive patterns of behaviour really do develop independently of opportunities for learning relevant lessons in individual experience. It is very important to note the reference here to relevant lessons. An isolation experiment does not aim to isolate an animal from all experience. That aim is not only impossible but pointless. The point is to isolate the animal from relevant experiences, where relevant experiences are learning opportunities. An experience is relevant only if it could provide the animal with the specific information that it needs to produce the correct forms of behaviour (Lorenz 1965: 85-89). If a bird on reaching maturity can sing its species-typical song, even though it has never heard an example of that song, then the bird’s capacity to sing that specific song must be innate, in some sense of that term which implies that some properties of the song are unlearned. It is not the singing but the specific form of the singing that cannot be explained by learning, and it is this property of the singing that is explained as innate.
The explanation for instinctive behaviour must begin, says Lorenz, with the fact that instinctive behaviour is adaptive. Instincts typically serve animals well, in their struggles to survive and reproduce. An adaptation is a trait that equips an organism to meet a specific set of challenges. Adaptations are adapted to specific properties of specific environments. The close fit between an adapted trait and a specific environment is not an accident or a fortuitious coincidence. The “amazing facts of adaptedness” (Lorenz 1965: 32) stand always in need of explanation.
Adaptation by natural selection is a process in which an organism is ‘molded’ to its environment. The process is akin, says Lorenz, to the formation, within organic structure, of an image of the environment (Lorenz 1965: 7). The fact that adapted traits ‘fit’ their environments is explained, by Lorenz, as the product of one or other of only two possible mechanisms. Behaviour is adapted through either phylogenetic processes, acting over evolutionary time, or through ontogenetic processes, acting within the individual lifespan. Processes of the latter kind involve learning. Processes of the former kind, phylogenetic processes, produce instincts. These two mechanisms are ‘entirely independent’ (1965: 4).
Lorenz conceives of the adaptive modification of behaviour as a process in which information (an ‘image’ of the environment) flows from the environment into the organism. So the question becomes: “Whence does the organism derive the information underlying all adaptedness of behavior?” (Lorenz 1965: 42). Only two channels are available for the flow of information into the organism. These two channels are natural selection and individual learning (18-19).
Learning is a process in which an animal acquires environmental information during its own individual lifespan. The information is ‘environmental’ in two senses: it comes from the environment and it is about the environment. Learning is more than just change of behaviour as an effect of environmental forces acting on the animal. It is a process that normally produces adaptive changes of behaviour. Learning normally modifies behaviour so as to improve the degree of fit between behaviour and environment. Lorenz introduces the concept of information to explain a very particular fact about learned behaviour, namely, to explain the fact that learned behaviour is normally well-adapted to specific, contingent features of the animal’s environment. Learning is a process in which specific information flows from a specific environment into the animal’s nervous system, where it shapes behaviour to fit some of the specific challenges posed to the animal by that type of environment (Lorenz 1965: 12, 37).
The only other channel through which information about the environment can flow into the animal is a phylogenetic or evolutionary channel. Information from the environment flows into a lineage during its evolution. The information is about the specific environment in which natural selection occurs. Natural selection is a process in which a lineage takes up information from its environment, information that is expressed in adaptations to that specific environment. Among an animal’s adaptations is its repertoire of instincts.
Lehrman and many others have been critical of dichotomies between innate and learned, or heredity and environment, or genes and environment. The basic criticism is that these dichotomies falsely assume that all the various factors that enter into development can be usefully classified under an exclusive and exhaustive dichotomy, and that development is a process in which an organism is produced by interaction between the innate and the learned, or between heredity and environment, or between genes and environment. In spite of these criticisms, Lorenz still considers that there is a sound distinction to be drawn between his two sources of adaptive information. No piece of behaviour considered as a whole can be assigned exclusively either to heredity or to learning. But there is a clean distinction to be drawn between two independent sources of adaptive information. Complex behaviour is normally adapted, in some respects, by information from the environmental channel and is simultaneously adapted, in other respects, by information from the phylogenetic channel. Imprinting, for example, depends on both innate attention biases and sensitive periods, and also on perceptual information. But this causal complexity does not blur the distinction between the two information channels.
Adaptation and the Flow of Information
This model of a flow of information from the environment into the organism has considerable intuitive appeal. It is prominent in the contemporary culture of popular science. Witness Matt Ridley:
“That is what genes are: parts of an information system that collects facts about the world in the past and incorporates them into good design for the future through natural selection” (Ridley 2003: 194).
Lorenz invites us to imagine ancestral animals picking up adaptive information from their environments and passing that useful information on to their offspring (Lorenz 1965: 8). The model has special appeal as an explanation of the astonishing specificity of some instinctive behaviour. Consider for example the potency of the sign stimulus. Why is it that these very specific stimuli have such a powerful capacity to trigger instinctive responses? The intuitive appeal of Lorenz’s information-theoretic account can be confirmed by considering the following stickleback story, told by Ernst Mayr. In normal populations, female sticklebacks are disposed to produce courtship responses to stimuli which carry the information ‘red below’. Isolation experiments show that this response pattern develops in females independently of any opportunities they have for learning that sexually mature male sticklebacks have red bellies. Mayr reports that in one stream system on the Olympic Peninsula, there is a population of jet black sticklebacks. The population evolved from red to black bellies about 8000 years ago, in response to local predatory pressures. For the last 8000 years, the Olympic females have been courted by and have mated with black males. Mayr reports the following remarkable fact:
“when given a free choice between a black and a red-bellied male, (the black female) chose the red male five out of six times after 8000 years of ‘habituation’ to black males!” (Mayr 1974: 653).
Why does the female show such a marked preference for a red belly? She has had no opportunity to learn the association between red bellies and desirable mates, an association that does not even hold in her environment. Yet her response is hardly likely to be pure coincidence. The causal explanation of her response must begin from the fact that suitable mates for her kind had red bellies in the distant past. Lorenz interprets the causal relation between ancient and current populations in terms of an information flow from past to present. Information about the desirability of red bellies in mates has been transmitted over a gap of 8000 years to the current population of Olympic sticklebacks.
It is very tempting to say, with Lorenz, that what is innate in these Olympic sticklebacks is information that has been ancestrally acquired. The female stickleback knows that red bellies make good mates. She does not know this from her own experience, but from the experience of her ancestors. It is inherited knowledge. Lorenz himself toys with this epistemic interpretation, though without actually endorsing it:
“In the special case of our baby Apistogramma and of our male stickleback we want to know how the former can possess information on the external characteristics of its mother and how the latter can ‘know’ what its rival looks like.” (Lorenz 1965: 33; see also 85.)
In what manner does the baby fish ‘possess’ information about its mother’s appearance? How like ordinary knowing is the sense in which sticklebacks ‘know’ what their rivals look like? Suppose that by ‘know’ Lorenz actually means knowing, in some real if modest sense.
If we suppose this, then we have arrived at what I will call the flamboyant interpretation of Lorenz’s theory of innate information. On the flamboyant view, what is innate, strictly speaking, is knowledge. Animals have innate knowledge about the environment into which they are born. Information about past environments is passed down through the genes and expressed in young animals as a priori knowledge. The information from the past is innately present in the animal; and it is also information for the animal. In philosophy, of course, the innate is traditionally conceived in this epistemic way. A priori knowledge is present in a person’s mind prior to all experience. It is knowledge that each person is born with, knowledge that is brought to experience. Lorenz himself makes the comparison with the philosophical conception of a priori knowledge: his own definition of the concept of the innate is, he says, a conscious paraphrase of Kant’s definition of the a priori (Lorenz: 44).
In philosophy, however, innate knowledge is most often understood to be just of necessary truths, the first principles of mathematics and logic, perhaps of metaphysics and ethics. This is one of the ways in which Lorenz’s evolutionary view differs so flamboyantly from the traditional view. For Lorenz appears to be saying that the innate knowledge that animals have is knowledge of empirical facts. Empirical facts are contingent on how the world just happens to be. The facts might have been different. Male sticklebacks might have had blue bellies. But as it happens, they have red bellies, and it is this contingent fact, in all its improbability, that the Olympic female knows innately. On the flamboyant interpretation of Lorenz’s theory of instinct, instincts provide animals with innate knowledge of contingent facts about the environment to which they are adapted.
There are a number of passages in Lorenz which encourage, or at least do not discourage, the flamboyant interpretation. In addition to the passage recently cited about the baby Apistogramma and the male stickleback, consider the following sample. Phylogenetically acquired information “tells the organism which of the consequences of its behaviour must be repeatedly attained and which ought to be avoided in the interests of survival” (Lorenz 1965: 16). All learning mechanisms depend on innate information telling the organism what is ‘good’ and what is ‘bad’ for it (18). Many kinds of geese “possess perfectly good phylogenetic information about how a fellow member of the species behaves when inviting copulation”. “The genetic information may be verbalized as follows: copulate with a conspecific who is lying low in the water and is stretched out along its surface” (39). “The gosling ‘knows innately’ that it should copulate with a fellow member of the species stretching out low in the water” (42).
It is true that there is no explicit endorsement of the flamboyant interpretation in these passages (nor explicit denial). Should Lorenz take the step of removing ‘know’ from scare quotes? No, he should not, because to do so would be to commit himself to purposive explanations for instinctive behaviour. It would commit him to the view that animals understand the purposes of their own instinctive actions. But there is evidence that this is normally false.
Purpose and Instinct
An animal acts purposively if it acts in order to achieve a specific goal. Purposive action requires that the animal represents to itself the goal of its action. The action that results is undertaken in order to achieve that goal, a goal that is mentally anticipated. On a purposive interpretation of instrumental learning, for example, a rat presses a lever because he knows that lever pressing causes the delivery of food. He can form a mental representation of food, even when no food stimuli are currently available. He has a goal, eating food, and knows what to do to achieve that goal. On the purposive interpretation of egg rolling behaviour, the goose has the goal of recovering the egg and acts in order to achieve that goal; similarly, the Olympic stickleback has a racial memory of the colour that originally marked out the most splendid mates, and chooses the red fish because she wants to court and be courted by the best possible mate.
The point of my objection is not that animal behaviour is never purposive. There are reasonable grounds for holding that some animal behaviour is purposive. There are empirical marks of purposive behaviour. Behaviour that is controlled by a representation of a goal has the properties of persistence and plasticity (Woodfield 1976). Persistence is expressed in the ability of an animal to recover from disturbances or interruptions to its plans: having been knocked off course, the purposeful animal responds by taking steps to bring itself back on course. Plasticity is expressed in the ability of an animal to achieve a specific goal in a variety of different ways, depending on its initial situation. The animal setting out (counterfactually) from other starting points would go in other directions, but in each case he would be heading towards the same goal. Some of the hunting behaviour of predators like raptors and carnivores displays these properties of persistence and plasticity. In general, the most likely place to find purposefulness in animal behaviour is in strategic activities, like hunting, where animals draw fully on the lessons they have learned and the abilities they have acquired from their own experience.
Instinctive behaviour is not the place to look for purposefulness. The usual criterion for instinctive behaviour, the most reliable empirical criterion, is its inflexibility, and inflexibility is the absence of persistence and plasticity. Darwin specifically says that animals acting on instinct do not understand the purposes of their own behaviour. Natural selection takes care of the purposes, leaving the animal with responsibility just for delivering a specific pattern of movement on demand. The flamboyant interpretation of the doctrine of phylogenetic information would commit Lorenz to an untenably rich account of the proximal causes of instinctive behaviour.
The classical theory of instinct includes a model of proximal mechanisms. This is the infamous hydraulic model of motivation (a model, incidentally, which resists rich cognitive interpretation). But why should a theory of instinct take on any particular commitments with regard to proximal mechanisms, with the exception, perhaps of a commitment to nonpurposefulness? One of the more telling criticisms of the classical theory of instinct is that it binds together too many different properties of behaviour. In particular, in the present instance, there is no good reason to expect, in advance of actual inquiry, that all behaviour that is stereotyped in form, species-typical and correctly expressed by naïve animals, will be caused by the same kind of proximal mechanism, in every case. Perhaps the instinctive behaviour of invertebrates and ‘lower’ vertebrates is normally under the control of relatively simple, peripheral, stimulus-response mechanisms. This need not be true, however, of the instinctive activities of birds and mammals. Lorenz himself was attracted to the idea that some animals engaged in instinctive behaviour for its own sake. Instinctive behaviour was intrinsically satisfying for the animal. An animal that acts instinctively is acting purposefully in a limited sense, though not for the sake of any further purpose beyond the satisfactions to be found in the activity itself. Lorenz cites with approval Wallace Craig’s distinction between appetitive and consummatory acts (Lorenz 1965: 63, 75; 1970: 270-271,312). Animals seek out opportunities to produce consummatory acts: the seeking out is appetitive behaviour (which might be purposeful), while the intrinsically satisfying consummatory acts are instinctive. This conception of the psychology of instinct is comparatively rich, even though it attributes to animals no understanding of the ultimate purposes of their own instinctive actions. My sense of Lorenz’s background view of animal psychology is that he thought that instincts probably were psychologically real at least to this degree in ‘higher’ vertebrates, but perhaps not in ‘lower’ vertebrates and invertebrates.
The flamboyant interpretation supposes that the phylogenetic information that is expressed in instinctive behaviour is information about the goals (or functions) of that behaviour. Information is ‘expressed’ in behaviour in the sense that a representation of that information is effective in the causal production of behaviour. What makes the information effective in this way is the fact that it is known by the animal, and the animal acts on what it knows. Phylogenetic information is represented both in and for the animal: it is ‘personal-level’ information. A more modest form of cognitivism is the view that phylogenetic information is represented in but not for the animal: it is ‘subpersonal-level’ information. Representations of that information drive only subpersonal cognitive processes. The proximal mechanisms of instinctive behaviour, in this case, are cognitive, but not in a manner which underwrites intentional interpretations of the animal itself as an agent. On one view of the knowing-that/knowing-how distinction, this is how know-how is produced: by representations (whether declarative or procedural) that drive subpersonal mechanisms. Perhaps this is what Lorenz should have said: that the only personal-level innate knowledge that animals have is know-how, and that know-how is produced by subpersonal mechanisms in which phylogenetic information is explicitly represented. This is a possible view. It evades the more implausible implications of the flamboyant interpretation that assigns explicit purposes to animals in their instinctive activities. But it is not acceptable for another reason, which also affects the flamboyant view. This reason depends on a distinction between the source and the content of phylogenetic information, a distinction that I explain and defend below.
Cognitive interpretations of the doctrine of phylogenetic information are hazardous for another reason: they encourage a Lamarckian view of the evolution of instincts. If instincts express a priori knowledge (whether the ‘knowledge’ is personal or subpersonal, explicit or implicit), then it is very tempting to think that the initial acquisition of that knowledge will have involved learning. This rather Platonic thought is expressed in the old notion that instinct is ‘lapsed intelligence’ (Boakes 1984: 205). Ancestral sticklebacks learn from experience that red denotes a sexually mature male of the species, and pass that lesson on to their descendants. The objection of course is that offspring do not inherit the lessons that their parents have learned from experience. If it is true at all that instincts express inherited information, then it is information that has not entered the lineage through cognitive channels. And if phylogenetic information has not entered the lineage in representational form, is it likely that the current expression of that inherited information is representational in form?
What is Phylogenetic Information?
We need a more modest interpretation of the idea of phylogenetic information. Here are some key facts about phylogenetic information, as it is conceived by Lorenz.
There is, first of all, the fact that the information has content: it is information about environments of adaptation, ecological information. Lorenz refers to adaptations as containing “an image of the environment” and “information concerning environment” (Lorenz 1965: 7). Chromosomal mechanisms have the evolved function ‘of acquiring and storing information on the environment’ (8). The baby Apistogrammas who selectively responds to a certain black and yellow pattern, and the male stickleback who selectively responds to an object which is red below, both possess information concerning the normal appearance of (respectively) mothers and rivals (33). Phylogenetic information has content: it is about the environment of selection.
Second, this ecological information enters the lineage during its evolution. Through natural selection, “the species gathers information and stores it, coded in the form of chain molecules, in its genome” (Lorenz 1965: 103). Only two channels exist through which information can be ‘fed into’ the organism. One of these is individual learning; the other is the adaptive processes of evolution (18-19).
Third, phylogenetic information is stored in the chromosomes (Lorenz 1965: 8). To describe something as innate is to assert that it is based on genetical information (42).
Fourth, phylogenetic information has a crucial function to perform in individual ontogeny. The “coded information” that is “stored in the genes” is decoded in development (Lorenz 1965: 20). “What rules ontogeny, in bodily as well as in behavioral development, is obviously the hereditary blueprint contained in the genome and not the environmental circumstances indispensable to its realization” (42).
Fifth, phylogenetic information explains the adaptedness of instinctive behaviour. Some properties of instinctive behaviour may be explained by environmental factors. But the “amazing facts of adaptedness” can only be explained by the phylogenetic origin of “the information contained in and indispensable for the molding of the organism in such a way that it fits its environment and is able to cope with it” (Lorenz 1965: 32-3).
To this list, the flamboyant interpetation adds a sixth claim, namely, that information about the environment of adaptation is represented in the cognitive mechanisms that are proximally responsible for instinctive behaviour. But we have seen reason to discard this claim.
However, the remaining five claims still bundle too much together. In particular, the notion that phylogenetic information is ecological information needs to be unbundled from the notion that it is stored in the genes and expressed in ontogeny. Only the former is properly called phylogenetic. I will argue, shortly, that the only information about environments of selection that is ‘present in’ adaptations is information for the ethologist (not for the animal itself). First I need to say something about ontogenetic information. There are well-known objections to the whole idea that genomes contain developmental information, whether as blueprints or as recipes (Griffiths and Gray 1994). I leave these further worries aside.
Genetic information is supposed to explain some phenotypic facts. For example, the carnivore genotype includes genes for carnassial teeth. These are genes for carnassials in the sense that those genes specify that teeth of that sort are to be produced in development. What the genes specify is a structure, carnassial teeth. They do not specify an adaptive function. The function is specified, not by the genes but by the relevant selection history that explains why those genes are now present, or dominant, in the gene pool of that species.
Suppose that there are genes for instincts. These genes also specify particular structures, in this case, structures (or patterns) of behaviour. Consider again the egg-retrieval routine in the greylag goose, and the genes for that pattern of behaviour. The information in the genotype is information about the form of behaviour, information specifying the form (structure) of a fixed action pattern. That is why it is no accident that stereotypical patterns are so often found in instinctive behaviour.
The concept of phylogenetic information is ambiguous. It is one thing for information that is present in an animal (or its genes) to have an origin or source in phylogeny. It is a different thing for information that is present in an animal (or its genes) to be about an environment. The former fact concerns the source of the information, the latter concerns its content. To this mix should be added a third concept, that of ontogenetic information. This is the concept of phenotypic information that is contained in the genome. This information specifies phenotypes, or specifies processes of development which have those phenotypes as their normal outcome. These three concepts, bundled together by Lorenz, need to be unbundled.
Consider two different explanatory projects. One project is to explain the existence of structures in morphology and behaviour. Why does a particular cat have carnassial teeth? Because she has the genes for that specific phenotypic structure. Why does a particular female stickleback respond with a specific pattern of behaviour to the sight of a red belly? Because she has the genes for that phenotypic structure, that pattern of behaviour. Both of these explanations are proximal explanations for phenotypic structures. They use the concept of ontogenetic information to explain structural facts about phenotypes. There is also a more ultimate explanatory project. What is the explanation of the presence of these phenotypic structures in current populations of sticklebacks and cats? Or, equivalently, why are the genetic structures that specify those phenotypes present in contemporary stickleback and cat gene pools? The explanation is evolutionary, or phylogenetic. The contemporary structures exist because of adaptive functions they performed (directly or indirectly) in ancestral environments. Genetic information specifies structures; phylogeny explains the presence of that genetic information in current genotypes.
Physical Information and Semantic Information
We also need to observe the distinction between physical information and semantic information. Consider first physical information. This is a ubiquitous commodity. The world is full of it. It is sometimes called natural information, or natural meaning or indicator meaning (Dretske 1981). Smoke means fire: that is, smoke carries information about fire, smoke indicates the presence of fire. Tree rings and ice cores carry information about past climates. Strata in the rocks carry information about past geological events. Information of this modest kind is present in the physical world wherever lawful regularity is present. Causes carry information about their effects and effects carry information about their causes. Lawfully coinstantiated properties carry information about each other. Wherever there is physical regularity, there is physical information. The lawfulness of the regularity does not need to be cognitively tractable, the kind of lawfulness that we can understand and describe. It might be immensely conditionalized lawfulness, of the sort that genes, say, have with respect to phenotypes. There is no more to physical information than lawful regularity. In particular, in the sense of meaning as content, physical information does not have meaning: it is not about anything.
Semantic information is marked out by the fact that it has content: it is about something. Ecological information must be semantic, because it is information about environments of adaptation. The flamboyant interpretation attributes semantic information to animals: they act on representations of ecological information, information about the environments to which their instincts adapt them. If the flamboyant interpretation is rejected, what happens to the idea of ecological information? What happens to the idea that phylogenetic information is semantic information?
Adaptations carry a large information load, in the physical sense of information. Properties of adapted traits carry information about evolutionary pathways, natural selection processes, the morphology, physiology and behaviour of ancestral animals, and the ecology of ancestral environments. The carnassial teeth of my cat carry information about the diet and habits of her ancestors. The sexual preference of the Olympic stickleback for red bellies contains information about the existence of red bellied males among her ancestors, and information concerning the fact that in some ancestral populations, female preferences for red bellied males were fitness-enhancing. But all of this is just to say that adaptations have causal explanations which refer to properties of ancestral environments and animals.
It sometimes seems that there must be more to physical information than just the existence of objective regularities. It seems like this because of the implicit presence, in the frame, of observers. An observer who has the right kind of general knowledge already to hand is also able to get his hands on semantic information by studying evidence. The geologist gets information about past geological events from strata in the rocks. The ethologist gets information about the environment of selection from instinctive behaviour patterns. But these are no literal acts of information extraction. It is not literally true that the semantic information that is now in the scientist’s mind has arrived there having been mined from the world. Information in the rich, semantic sense never was out there in the rocks or in the courtship responses of the female stickleback.
The ethologist can learn about ancestral populations and environments by studying contemporary populations and environments. He learns about the ancestors of the Olympic sticklebacks by studying the behaviour of current animals in that population. All the information hereabout is either semantic information that is contained in the knowledge that the ethologist possesses or it is physical information that is contained in the structure of the stickleback’s behaviour. The fish herself does not have semantic information about environmental or evolutionary facts; she is only a channel of information for the ethologist. Her behaviour is an objective structure from which the ethologist learns facts about her evolution. Her phenotype is composed of a variety of other physical structures that also provide the ethologist with information about her evolution. But she is not herself a subjective recipient of the same information that he comes to have by studying her phenotypic structure. She does not know that once upon a time, suitable mates had red bellies. My cat does not know that her ancestors used their carnassial teeth to slice meat: we should not be tempted to think otherwise merely by the fact that this is the adapted function of her carnassials. We should avoid the like temptation when reflecting on the functions of instinctive behaviour. The only knowledge hereabouts is knowledge that the scientist gets by observing animal phenotypes.
Adaptations as Evidence
The ethologist who knows the function of a biological trait can infer facts about the environment of selection. The trait, in virtue of its function, carries phylogenetic information for the ethologist. So far, so good. The adapted function of the trait is evidence for past selection events. But there is a worry: a possible function truly is an adapted function only if this is the specific use for which that trait was historically selected. Hypotheses about adapted functions and hypotheses about historical environments of selection stand or fall together. If the ethologist knows the adapted function of a trait, then he also knows something about historical selection events. Having identified the adapted function of the trait, he cannot go on to increase his knowledge by inferring conclusions about phylogeny. Knowledge about functions is inseparable from knowledge about environments of selection. However, the ethologist can increase his knowledge if he begins not from function in the strict sense, but from current adaptive value. Here is how this idea works.
The ethologist observes a pattern of behaviour. He notes the adaptive value of the behaviour, its utility, the contribution it currently makes to fitness. He notes what Tinbergen called its survival value (Tinbergen 1963). Then he makes an adaptationist move: he conjectures that the current utility of the behaviour is also its adapted function. He conjectures that the specific contribution to fitness for which the behaviour was originally selected is the same as its current adaptive value. On the basis of this conjecture, the ethologist draws conclusions about the environment of selection. This whole process of reasoning does not consist just of the rather trivial inference from adapted function to environment of selection. It is the more interesting inference from current utility to environment of selection. The current utility of the trait provides the ethologist with information about the environment of selection. It does so by virtue of a hypothesis to the effect that current utility is evidence for adapted function.
Where does all of this leave the idea of phylogenetic information?
Lorenz started out with the idea that instincts are evolutionary adaptations. This is a good idea. He thought that instinctive behaviour was therefore innate. This idea came under stringent attack by Lehrman. Lorenz retreated from his original position. In his later work, he adopted the view that strictly speaking, what is innate is only information, phylogenetic information. Information about the adaptive significance of red bellies flowed into the stickleback lineage at some time in the past. This information is now expressed in the courtship preferences of current female sticklebacks. The information has historic origins, is expressed now in phenotypes, and is about adaptive, ecological facts.
On the flamboyant interpretation of this doctrine, it is the animal’s knowledge of those ecological facts that explain its behaviour. This is a purposive explanation. A weaker version denies that instinctive behaviour is purposive, but still gives a cognitive, though ‘subpersonal’, explanation for instinctive behaviour. Semantic information about ecological facts is present in representations that are used by the mechanisms that produce instinctive behavior. But this is all too rich.
Purposive explanations are inappropriate for instinctive behaviour. Cognitive explanations, exploiting semantic information, are too rich for much of the instinctive behaviour of ‘lower’ vertebrates and invertebrates (whatever the truth turns out to be for mammals and birds). Animals do not need to know the evolved functions of their instinctive behaviour. Still less do they need to have knowledge of the ancestral environment of selection in which the behavior evolved to have those functions. If there is any knowledge here, any semantic information, it is the information that the ethologist has about the animal and its instinctive behaviour, knowledge about the adaptive function of the behaviour and correlative knowledge of ancestral ecologies. Such functional and ecological knowledge is psychologically realized only in the mind of the ethologist. All the other information that exists hereabouts is physical information. Physical information does not have content (let alone phylogenetic content) but it does have a source, and sometimes that source is in the evolutionary past. Instinctive behaviour carries a lot of physical information about a lot of things, including historical information about the environment of evolutionary adaptation and information about the adaptive functions of the behaviour.
Every physical thing carries some physical information, including historical information. There is nothing distinctive about the fact that instinctive behaviour carries physical information, or even that it carries physical information about the past. Perhaps something could be made of the fact that instinctive behaviour carries information primarily about past environments whereas learned behaviour carries information primarily about present environments. But once it is explicitly noted that this kind of information is just ordinary physical information, we can see that the corresponding notion of phylogenetic information is just a colourful way of referring to the fact that instincts are adaptations. It is just a colourful way of saying that specific behavioural control structures are present in current animals because of selection events in historical environments.
If it is semantic information that is at issue in the thesis of phylogenetic information, then the choice is either to embrace a flamboyant interpretation of animal instincts or to agree that the only information about phylogeny and function that exists hereabouts is information that is in the ethologist’s head and in his notebook. But there is nothing special about instinctive behavior in the latter regard: anything that is known by someone is a bearer of information in the same sense.
The one context where the idea of phylogenetic information does seem to do some useful work is in scientific reasoning from current utility to conclusions about historic selection events, and hence to adapted function. This is semantic information alright, and information about the environment, but it is also information just for the ethologist, and not for the animal itself.
Thanks to Robert Jackson and Bill Rowland for advice on the Olympic sticklebacks.
Boakes, R.: 1984, From Darwin to behaviourism. Psychology and the minds of animals, Cambridge University Press, Cambridge.
Darwin, C.: 1859/1964, The Origin of Species, first edition, ed. J. W. Burrow, Penguin Books, Harmondsworth.
Dretske, F.: 1981, Knowledge and the flow of information, MIT Press/A Bradford Book, Cambridge, MA.
Griffiths, P. E. and Gray, R. D.: 1994, ‘Developmental systems and evolutionary explanation’, Journal of Philosophy 91, 277-304.
Hinde, R. A.: 1970, Animal Behaviour. A synthesis of ethology and comparative psychology, McGraw-Hill: New York.
Lehrman, D.S.: 1953, ‘A critique of Konrad Lorenz’s theory of instinctive behavior’, The Quarterly Review of Biology 28, 337-363.
Lorenz, K.: 1965, Evolution and Modification of Behavior, University of Chicago Press, Chicago and London.
Lorenz, K.: 1970, Studies in Animal and Human Behaviour, Volume 1, translated by Robert Martin, Methuen, London.
Mayr, E.: 1974, ‘Behavior programs and evolutionary strategies’, American Scientist 62, 650-659.
Ridley, M.: 2003, Nature via nurture. Genes, experience, and what makes us human, HarperCollins, New York.
Tinbergen, N.: 1951, The Study of Instinct, Clarendon Press, Oxford.
Tinbergen, N.: 1963, ‘On aims and methods of ethology’, Zeitschrift für Tierpsychologie 20, 410-433.
Woodfield, A.: 1976, Teleology, Cambridge University Press, Cambridge.