Return to Human Knowledge and Human Nature contents  

 

7. THE CASE FOR INNATE CONCEPTS

 

In this chapter I shall consider the arguments for believing in innate concepts, leaving discussion of innate knowledge to chapter 8.  (Recall from chapter 4, that concepts might be innate even though no knowledge is, and conversely that some knowledge might be innate even though no concepts are.)

 

Kinds of Concept-Possession

We can distinguish at least three different senses in which a creature might be said to possess a given concept.  Firstly, this might be said if the organism has an appropriate discriminatory capacity, being able to act differently depending upon whether or not instances of that concept are present.  For example, pigeons can learn to distinguish triangles from other shapes (pecking at triangles, but not squares or circles, to gain a reward).  They are sometimes described in consequence as possessing the concept of a triangle.  Indeed, pigeons are capable of learning perceptual discriminations which have a remarkable degree of sophistication.  They can be taught to peck at a photograph for food only if it contains some sort of representation of a human being, whether in the foreground or background, sitting or standing, profile or full-face.  They can do the same for photographs of water, or trees, or indeed for photographs of a particular person.¹  Such pigeons might be said to possess the concept human-appearance, or the concept water-appearance.

            However, it is doubtful if this notion of a concept takes us very far.  In particular, it is doubtful whether concepts in this sense will figure in any genuine beliefs.  There are two points to be made here.  The first concerns the notion of belief-content.  It is that beliefs have contents which are essentially structured out of recombinable elements.  The belief that grass is green contains the concepts grass and green, which can each of them figure in many other beliefs, such as the belief that grass is inedible or the belief that emeralds are green.  Indeed, these same concepts can also occur in the content of other propositional attitudes such as hopes and desires, as when I hope that the grass will grow, or want a green shirt.  The only candidates for the content of the pigeon's belief, on the other hand, would be 'There is a human-appearance here' and perhaps 'Human-appearances are a source of food'.  There is no reason to think that the concept of a human-appearance is indefinitely recombinable into other contents and other propositional attitudes.

            The second point about beliefs concerns their causal role, which has already been stressed in previous chapters.  It is essential to beliefs that they should be capable of interacting with desires and other beliefs in such a way as to cause behaviour, where that behaviour is then subject to the practical-reasoning-model of explanation.  Yet it appears extravagant to try to explain the pigeon's behaviour by saying that it believes that the photograph it is pecking contains a human-appearance, wants something to eat, and believes that pecking human-appearances produces food.  This is because the pigeon's repertoire of action is so extremely limited.  The capacity to discriminate human-appearances plays no other part in the pigeon's life than in the production of a pecking-response.  There is insufficient flexibility and variety of behaviour here for us to consider seriously applying the practical-reasoning-model of explanation.

            The second possible sense in which an organism might be said to possess a concept arises naturally out of the limitations of the first.  We could say that a creature possesses such-and-such a concept provided that we are prepared to ascribe to it a variety of beliefs or desires which involve that concept, where we then use the states which we have ascribed in explanations of the creature's behaviour which fit the practical-reasoning-model.  In this sense some of the higher mammals, at least, possess concepts, since we take seriously the attribution of beliefs and desires to them.  For example, we may explain the behaviour of a dog by attributing to it the sequence of thoughts 'I want to get the ball.  The ball is on the table.  If I jump on the chair I can reach the table.  So I shall jump on the chair.'  Moreover, the behaviour of a dog certainly exhibits a wide variety of ways in which it can interact with a ball - fetching, chewing, chasing, and catching - suggesting that the concept of a ball (or something like it) does form a component in a number of distinct canine beliefs and desires.

            However, recall from chapter 6 that beliefs and desires can be nonconscious.  I can believe that there is a large obstacle in the road, and consequently turn the handlebars of my bicycle, because I want to reach my destination safely and believe that unless I turn the handlebars I shall crash - all without conscious awareness of what I am doing or why.  This then gives us our third possible sense of concept-possession, where an organism may be said to possess a given concept provided that it entertains conscious thoughts in which that concept figures (or is at least apt to entertain such thoughts in appropriate circumstances).  A sufficient condition for possessing a concept in this sense would be the ability to use correctly the corresponding term of a natural language.  If you have the capacity to use appropriately, and to understand, statements in which the word 'grass' occurs, then this will be sufficient to show that you possess a conscious concept of grass.  We can leave open the question whether relevant language-mastery is also a necessary condition for conscious possession of a concept, hence leaving open the question whether any nonhuman animals possess concepts in our third sense.  The answers to these questions will depend upon the nature of the difference between conscious and nonconscious thought, and on whether it is true that only language-users can entertain conscious thoughts.²

            Armed now with the distinction between these three different notions of concept-possession, we can see immediately that our basic repertoire of discriminatory-capacity concepts must be innate.  If we could not, in the first place, respond differently to stimuli of different colours or temperatures, or to lines and boundaries within our visual field, then we could never learn anything else.  These elemental capacities for discrimination must be built into the very structures of our perceptual apparatae.  Moreover, even where the discriminations in question are somewhat less basic (and certainly acquired over time), such as the ability to tell human faces apart from one another, it may be that these discriminations are not really learned.  Our face-recognising mechanism may contain much innately determined information about the forms and limits of expression of the human face.  So even these may count as innate (but locally triggered) concepts.

            It is also very likely that the basic conceptual components of nonconscious belief are innate.  For as we noted towards the end of chapter 6 in connection with vision in particular, it seems plausible that much of the representational structure of our perceptual faculties is innately fixed.  That we see individual (often moveable) objects against a relatively stable three-dimensional background is probably not something that we learn.  In which case many of the various perceptual beliefs which can govern our actions without becoming conscious will count as employing innate - though perhaps locally triggered - concepts.  (Are we then to allow, in the light of my previous example, that the nonconscious concept handlebars is innate?  This seems implausible.  But there will surely be some more neutral way of characterising the belief that governs my actions as I ride - for example, that turning the rigid bent structure in my hands will alter my direction of motion.  Here the constituent concepts may plausibly be said to be innate.)  The beliefs in question will be those which enable us unreflectively to pick up and turn over objects in our hands, to step over obstacles in our path when walking, to turn our heads in the direction of a sound, and so on.

            Since the case for innate discriminatory abilities is a powerful one, as is the case for innate repertoires of nonconscious perceptual concepts, all interest devolves on to the question whether conscious concepts are innate.  This is the issue on which I shall concentrate for the remainder of this chapter.

 

Are Concepts Taught?

It is clear that many linguistic concepts are universal, being common to all human societies and to all natural languages.  But this in itself cannot provide sufficient reason for believing those concepts to be innate.  Nor do we need to appeal to the hypothesis of a common historical origin to make the point, as we did in the case of universal features of grammar in chapter 6.  For the universality of such concepts is better explained by their utility.  Societies which did not have concepts of time and place, motion and rest, causality and knowledge, would not survive for very long, let alone prosper.  Our universal concepts can be thought of as prerequisites for successful social life.

            However, the very fact that many of our concepts have a degree of utility which makes them unavoidable, does provide some slight reason for believing them to be innate.  For as we argued in chapter 6, in connection with perceptual concepts of space, the fact that certain linguistic concepts can be regarded as amongst the standing-conditions of all human life, means that there would be survival-value to the individual if those concepts were innate.  This would allow valuable cognitive resources in the developing child to be directed elsewhere.  It would also avoid the risk of mistakes occurring, which attends all learning.  But the most that this argument shows, is that it would not be entirely surprising if those linguistic concepts which are universal were innate; it does not in itself provide sufficient reason for believing that they are.

            Some of the points which gave rise to Plato's Problem in connection with the learning of grammatical rules hold in the case of concept-learning as well.  Children acquire a huge number of concepts in a short span of time.  (Chomsky claims that at certain periods of childhood as many as a dozen new concepts are acquired each day.)³  Moreover, they acquire these concepts on the basis of fragmentary and mostly positive data.  For just as in the case of grammar, very little explicit teaching of concepts actually takes place.  In general, adults simply use concept-words in the presence of children, providing them with a sample of applications which is necessarily limited, only occasionally correcting the child's own mistakes.  Not only this, but much of the data is, from the child's point of view, corrupt.  For much adult usage consists of irony, metaphor, jokes and teasing, in none of which are concept-words used literally, and which can only be understood when the literal meanings of those words are already known.

            However, we cannot assess how powerful the analogue of Plato's Problem is in connection with concept-acquisition, until we have a better idea of what concepts themselves are.  For much of the burden of the argument for the innateness of grammar was taken up by the claim that the grammatical rules governing natural languages form a hugely subtle and complex system.  If, in contrast, concepts were to turn out to be relatively simple entities, then the argument for their innateness would be correspondingly weaker.  To this issue I shall shortly return.

 

Fodor on Concepts

Jerry Fodor has recently provided arguments for the innateness of at least many of our conscious concepts.⁴  The first of these arguments arises out of an objection to empiricist theories of concept acquisition, and is similar to the one we used against Locke's account of abstraction in chapter 4.  Fodor claims that the empiricist view must be that concepts are learned, rather than merely triggered, and then points out that the only theory of learning which has ever been seriously developed involves hypothesis-formation and testing.  In order to learn the meaning of the word 'cat', for example, a child would have to do something like the following.  It might first form the hypothesis that 'cat' means 'animal', adjusting this as further data becomes available (such as a parent saying 'Look at that dog' when they pass a dog in the street), until finally it settles on the hypothesis that 'cat' refers to cats.  Then the obvious point is that in order to go through such a procedure, the child must already have available to it the concepts of 'cat' and 'animal'.  So if this were the only story of concept-acquisition which an empiricist could tell, it would follow immediately that a large repertoire of concepts must already be innate before a child could begin to learn the meanings of the words in its native language.

            A natural objection to this argument is that it assumes that possession of a linguistic concept is a genuine item of propositional knowledge - a matter of knowing that a given word means such-and-such.  But it would be equally plausible (if not more so) to maintain that possession of such a concept is a practical capacity - a matter of knowing how to use a given word.  On this view, to possess a concept is to be capable of classifying presented items correctly, as well as to have the ability to employ the appropriate word in a variety of sentence-constructions.  If this is right, then acquiring linguistic concepts will be much more like acquiring a practical skill, such as the ability to ride a bicycle, than it is like acquiring a new item of information.  And no one would want to claim that bicycle-riding is an innate ability (although of course it does make use of our innate capacities, such as those of balance and leg-movement).

            There are a number of replies to this objection which are available to Fodor.  The first is that acquiring linguistic concepts cannot be a mere matter of habituation, or repetitive practice.  (This parallels Chomsky's point about the acquisition of grammatical rules, discussed in chapter 6.)  For our concept-words admit of an indefinite number of possible uses, and a competent speaker will be able to understand combinations of concepts which they have had no previous experience of.  Someone who understands the word 'cat', for example, does not merely have the ability to say 'cat' when a cat is present, but also to ask whether there is a cat present, to assert that there is no cat present, and generally to connect that word together in an indefinite number of ways with the other words of the language which they understand.  So even if we grant that possession of a concept consists in a practical capacity, rather than being a matter of genuine propositional knowledge, still it must be a special sort of capacity, whose categorical basis in the brain somehow reflects the structure of the concept, and its connections with other concepts.  And then acquiring that concept cannot be merely a matter of practicing an activity, like learning to ride a bicycle.

            Fodor would also reply that in any case all cognitive processes take place in what he calls 'a language of thought'.  This would include the process of acquiring the capacities which we are supposing to constitute grasp of linguistic concepts.  Fodor's view is that all cognitive processes are computational, in the sense that they involve operations upon sentences or sentence-like constructions.  He holds that this is so, whether the processes in question are conscious or nonconscious, or whether they involve the organism as a whole or only sub-personal systems such as vision or the details of language-comprehension.  So even if possession of a concept is best thought of as a kind of practical capacity, rather than an item of propositional knowledge, still the process of acquiring that capacity will involve hypothesis formation and testing in the language of thought.  Then since all processes, including those involved in learning, take place in the language of thought, this language itself must be innate.  It then follows that some concepts - namely the concepts of the language of thought - must be innate in order that linguistic concepts may be acquired.

            The idea of a language of thought is controversial.  But even granting it, Fodor's argument is less powerful than it seems.  For the most that it does is to force us to recognise that nonconscious belief-concepts are innate, in order that conscious (linguistic) concepts may be acquired.  But we had already granted that in any case.  And it does not follow that conscious concepts are themselves innate.  Indeed, the appropriate conclusion may be weaker still.  For it would be much more plausible to assimilate the concepts involved in cognitive processing generally, to the discrimatory-capacity concepts discussed earlier.  For it is doubtful whether cognitive models of hypothesis-formation and testing within some sub-personal mental faculty (such as that of language processing) will genuinely fit the practical-reasoning-model of explanation.  So it may only follow (if we accept the idea of the language of thought) that we are committed to the innateness of sub-personal discriminatory capacities.  But again, this was something that we had accepted in any case.  The interesting question is whether conscious concepts (or something like them) are innate.

 

Fodor on Definitions

Fodor does have a different argument for the innateness of many conscious concepts, which turns on his claim that most linguistic concepts are in fact indefinable.  For he argues that the main point of contention between empiricists and nativists about concepts, concerns the range of concepts which are definable in terms of a primitive conceptual basis.  According to Fodor, the empiricist maintains that all other concepts may be defined in terms of a set of innate concepts which are sensory, which arise directly out of experience (the process of construction through definition then providing the empiricist with their theory of learning).  The nativist maintains on the contrary, that most of our concepts are triggered rather than learned.

            There is one way in which Fodor definitely goes wrong here.  This is in his claim that an empiricist must count the basic sensory concepts as innate (triggered by experience, rather than learned from it) because they are not acquired via a process of hypothesis formation and testing.  For this account of learning is too narrow.  Empiricists such as Locke and Hume certainly thought that our basic concepts were learned from experience, by virtue of their being a record of the initial experience laid down in memory.  What Fodor overlooks, is that our common-sense concept of learning divides into at least three distinct kinds.  There is learning by means of an inference to the best explanation, which is what forms Fodor's paradigm.  But there is also learning how to do something, such as ride a bycyle.  And there is memory-based learning, as would occur when one comes to know what someone looks like, by meeting them.  In fact the early empiricists proposed that basic sensory concepts were learned in the third of these modes, on the basis of prior acquaintance.  All the same, Fodor is quite right about the main issue, which is that it will count in favour of the rationalist thesis that most of our concepts are innate, if it turns out that they cannot be defined at all, let alone in purely sensory terms.

            Fodor's main argument for indefinability is based upon our failure, historically speaking, to come up with agreed definitions.  Despite generations of cooperative labour by analytic philosophers, there is hardly a single concept whose analysis is generally accepted.  The best explanation of this phenomenon, Fodor thinks, is simply that there do not exist any definitions to be found.  But in fact an alternative explanation is suggested by Chomksy.⁵  This is that most of our concepts are extremely complex, being related to one another in all sorts of subtle ways.  This seems on the face of it to be equally acceptable.  However, an interesting possibility opens up here.  For it may be that in either case we shall have an argument for nativism.  For Chomsky, in effect, turns this alternative explanation of failures of definition into the analogue, in the case of concepts, of Plato's Problem for grammar-acquisition.  Supposing that the system of concepts of a mature speaker is very subtle and complex, then the problem of how a child manages to acquire this system on such a slender basis really does become pressing.

            But in fact, even supposing that Fodor is right that most concepts are indefinable, this need not commit us to nativism.  For as he himself notes, 'indefinable' does not necessarily imply 'unstructured'.  It may be that complex concepts are built up out of simpler ones, without being definable in terms of them (at least if the definitions are required to take the form of statements of necessary and sufficient conditions of application).  Just such a possibility will be realised if concepts are represented in the mind by prototypes, which is a thesis for which much psychological evidence has recently begun to emerge.

 

Concepts as Prototypes

A prototype is a specification of a set of prototypical properties, taken together with a weighted similarity-measure.  The prototype for the concept dog, for example, would include such features as barks when angry, wags its tail when happy, is a mammal, has dogs as parents, eats bones, and so on.  But there is no suggestion that all dogs must necessarily have each of these features.  Rather, deciding whether something is a dog is a matter of judging whether it is sufficiently similar to the prototypical dog.

            Fodor considers this theory, but thinks he has a swift dismissal of it.  He claims that it cannot provide an account of compositionality - of how we can combine together linguistic concepts to form new ones.⁶  For example, he asks how the meaning of 'brown cow' is supposed to be determined from the prototypes of 'brown' and of 'cow'.  But the reply is surely straightforward.  It is that the meaning of 'brown cow' may be given by 'is sufficiently similar to this (the prototype of brown) and is sufficiently similar to this (the prototype of cow)'.  The point is, the meanings of complex concepts formed by putting together individual words need not themselves be prototypes.  The false assumption in Fodor's argument is that if prototype-theory is correct, then the concepts expressed by lexically-complex items such as 'brown cow' must themselves be prototypes.  Rather, such concepts may be logical constructions out of prototypes - building them up using notions like and and or.

            However, if concepts like brown cow are logical constructions out of prototypes, where do these logical notions come from in their turn?  I think that they could very plausibly be allowed by an empiricist to be innate.  They can be regarded as belonging to the given (innate) logical structure of the mind.  For these notions do not in themselves constrain what the world may be like.  Empiricists should only object to the innateness of concepts which seem to carry information about the world.  Whereas the logical concepts will be equally applicable, no matter what our experience of the world may be like.  So Fodor's argument against the empiricist fails, even given the way in which he sets up the debate.  It may be that most of our complex concepts are either prototypes or logical constructions out of prototypes, where these in turn are constructed out of simple ideas derived from experience (but without being definable in terms of those ideas).

            I shall not now review the psychological evidence supporting the thesis that many of our concepts have prototype-structure.⁷  Rather, supposing that such a thesis is correct, let us ask whether it is sufficient to defend empiricist theories of concept-acquisition.  One point can be made straight away, in the light of what was said in chapter 4.  For it would be just as implausible to maintain that all complex concepts can be constructed out of a basic set of purely-sensory ones, as it would be to claim that they can all be defined in such terms.  (For example, try giving a prototypical set of sensory concepts which would even begin to be adequate to express the concept of causation.)  So if empiricism were committed to reducing all concepts to those which describe our private sensations, then the introduction of prototypes would bring it no particular advantage.

            However, recall from our earlier discussion that we are allowing there to be an innate set of perceptual belief-concepts which, so far from being purely sensory, involve representations of physical objects in three-dimensional space.  This might provide a sufficient basis on which to construct complex prototypes.  For example, a child might first acquire, through experience, a perceptual paradigm for a given concept.  On hearing an object described as a cat for the first time, it lays down a representation of that particular cat in memory.  (It is this stage which presupposes innate perceptual belief-concepts.)  Then through further experience it begins to acquire a similarity-measure, for judging whether a given object is sufficiently close to the paradigm to be a cat.  For example, the child might start by over-extending the concept, judging that a dog is sufficiently close to the paradigm, until it is corrected by its parents.  As it begins to acquire perceptual paradigms and similarity-measures for other concepts, the child also starts to cross-classify, building up the full prototype for 'cat'.  For example, having acquired a paradigm for 'tail', it learns that cats have tails, and having acquired a paradigm for 'mammal' it learns that cats are mammals, and so on.⁸

            If this picture, or something like it, proved to be correct, then the empiricist view of the acquisition of conscious concepts would be largely vindicated.  The initial perceptual paradigms would be given in experience (although requiring an innate perceptual basis), and the later cross-classifications could also be learned through experience.  The similarity-measures, too, might be learned from experience, if it were to turn out that parental teaching is crucially necessary to prevent children from over-extending or under-extending their concepts.  Indeed, it would be a virtue of this sort of account that it could overcome the problem of an historical regress, outlined in chapter 4.  We need only suppose that the similarity-measures were first settled upon causally, rather than arrived at by prior teaching or conscious choice.  This could either have happened by chance, or as a result of some similarities being more salient, or more relevant to human concerns than others.

            No doubt it is a mistake to treat all concepts as if they are alike.  It may be that some are prototypes, and are acquired from experience, and that some are not.  One class of concepts for which the prototype-theory looks particularly implausible are psychological ones such as 'belief', 'desire' and 'intention'.⁹  (In fact, it is in connection with such concepts that Chomsky urges his alternative response to the lack of agreed definitions, opting for an explanation in terms of complexity, and arguing for innateness as the only solution to Plato's Problem.)  These will come into focus in the next chapter, where we consider the arguments for saying that our knowledge of our own psychology is innate, together with its constituent psychological concepts.  For the moment our conclusion must be that the case for the innateness of conscious concepts is largely unproven.

 

The Concept of Best Explanation

There is, however, one cluster of concepts which are very likely to be innate.  These are the concepts involved in the appraisal of non-deductive modes of argument, particularly the concept of the best available explanation of a given phenomenon.  This concept is undoubtedly a conscious (because linguistic) one, but is notoriously difficult to define.  Yet there is a remarkable degree of agreement amongst speakers in particular cases, as to whether one explanation is or is not better than another.  Moreover, while the concept does to some extent display prototype-structure (a good explanation tends to be simple, consistent, cohere well with surrounding beliefs, have broad scope, be fruitful in generating new predictions, and so on), the constituent notions are no easier to define in their turn; nor is it easy to see how they might be derived from experience.  Then since children receive no explicit training in the use of this concept, we cannot explain how we nevertheless manage to acquire it, unless we suppose that the concept is innate - triggered by the course of our experience, rather than learned from it.

            It is worth stressing that while the concept of best explanation does figure prominently in science, it is by no means an exclusively scientific idea.  On the contrary, we each of us constantly make at least tacit use of it in the course of our daily lives.  For example, when I enter a lecture-hall full of students, and consequently form the belief that they have come to listen to me speak, this is because this belief provides, in the circumstances, overwhelmingly the best available explanation for their presence.  Scientists probably make use of the very same concept, differing only in that their search is for the general laws or principles underlying the observable phenomena to be explained.

            One cannot mention non-deductive modes of argument without discussing Hume.  For Hume is famous for his discovery of the problem of justifying induction, and for his naturalistic explanation of our use of it.  His view is that induction has no rational basis, either in reason or experience.  Its reliability cannot be demonstrated a priori.  But then neither can it be shown to be reliable by past experience without circularity.¹⁰  Our reliance upon induction is rather to be explained by appeal to the basic principles of human psychology, particularly the laws of association amongst ideas.  Hume's theory is that having seen a phenomenon repeated, we are habituated to expect it to continue in the future.  But this is not to ascribe to us an innate concept.  For the psychological principles which give rise to induction are supposed to be general ones.  To say that our psychology is such that we do, as a matter of fact, go in for induction, is not to say that our minds contain an innate concept of a good inductive argument.

            Quite apart from the crudity of his psychological theory, there are two things wrong with Hume's account.  The first is that it ignors the normativity involved in our idea of best explanation.  The fact is that we do not (as a mere matter of fact) reason in certain ways, as a result of our nature.  We also apply standards of appraisal to such reasonings, counting some non-deductive arguments as better than others, and counting people as right or wrong in their assessments of the merits of such arguments.  Yet this is one of the distinctive marks of possession of a conscious concept.  For as we noted in chapter 3, concepts are best construed by empiricists as rules of classification.  This seems to be just what we have in the present instance - we have a rule for classifying some non-deductive arguments as better than others.

            Hume's second mistake is that induction is by no means exhaustive of non-deductive modes of argument.  Induction is a matter of generalising or projecting from observed regularities.  Yet we also frequently reason from such regularities to the presence of some underlying phenomenon (sometimes involving unobservables, such as sub-atomic particles) which would explain it.  Indeed, induction itself is best seen as a particular instance of the more general practice of inferring to the best explanation.  The reason why we infer from 'All ravens so far observed have been black' to 'All ravens are black', is that the fact that all ravens are black is the best available explanation for why all observed ravens should have been black.  Yet there is no question of exhibiting inference to the best explanation as flowing from the general principles of association amongst ideas.  Nor is it easy to see how it might be explained in terms of general features of any more sophisticated psychological theory.  Yet all of Hume's reasons for thinking that induction is cognitively basic and indispensable for us, now transfer to the case of inference to the best explanation.  It is therefore plausible to suppose that we have an innate grasp of the concept of best explanation, which we then employ in appraising particular arguments.

            It may be objected that there are wide variations across cultures, concerning what explanations will be counted as better than others.  How, then, can the concept of best explanation be innate to the human species?  In our culture, for example, we explain illnesses in terms of such things as the activity of viruses on the body, whereas in other cultures the same illness may be explained in terms of the malignant action of a witch.  But this is to miss the point that what counts as the best explanation of some phenomenon is always relative to your background beliefs.  It is these beliefs which vary across cultures.  The reason why the hypothesis that an illness was caused by the action of a witch is not a good explanation for us, is because we do not believe in witchcraft.  But, plausibly, all cultures employ the same general standards for selecting between competing explanations, given their background beliefs.

            Note that if inference to the best explanation is a generally reliable method of forming beliefs, then it is easy to explain how our concept of it could come to be innate.  For individuals will be better able to survive if they are able to attain true beliefs about the underlying processes at work in nature, which can then be harnessed and exploited, or if they can acquire knowledge of the unseen causes of observable phenomena.  (Conversely, if our concept of best explanation has already been shown to be innate, then this gives us good reason to believe that such inferences are generally reliable.  For a concept could not have been selected through evolution unless it conferred survival-value on the individuals who possess it.  Yet it is hard to see in what other way inference to the best explanation could have survival-value, unless it is indeed reliable.  We shall return to this issue in chapter 12.)

            I conclude that, while the case for innate concepts is largely unproven, there is at least one concept which is probably innate, namely our concept of best explanation.  This result will prove to be of some importance in chapters 11 and 12, when we come to discuss the problem of scepticism.

 

On to chapter 8