Innatism, Concept Formation, Concept Mastery and Formal Education
Innatism, Concept Formation, Concept Mastery and Formal Education.
Introduction.
There has recently been much interest within Philosophy of Education in debates about concept formation and whether the idea of concept acquisition is coherent (eg. Luntley 2009a), and whether we are compelled to acknowledge that concepts are innate, a position maintained by, for example, Fodor (1975, 2008). My aim will be to show that not only is acceptance of the fact of concept acquisition necessary to make sense of our conceptual abilities (in particular the phenomena of: concept change, conceptual depth, and conceptual variation), but that the mode of acquisition of concepts enables us not only to understand concept possession, but also concept mastery and hence the growth of expertise in particular activities, practices and occupations and the role that formal education might play in the growth of concept mastery and the acquisition of expertise.
The issue of innatism is important for understanding learning and education more generally. If concept possession is innate, (the claim that we are born with our conceptual apparatus, we do not acquire it) then there is a sense in which it is also complete. All that experience, learning and maturation can do is activate dormant concepts whose latency is transformed into full potency. But this seems to contradict our understanding of the growth of expertise. We become experts in particular activities and occupations by dint of practice and experience and we are all too well aware of this. Granted, expertise is something more than mastery of the relevant concepts (cf. Bengson and Moffett 2007 for the claim that it is not), but concept mastery is a necessary condition. So what could the claim that concept mastery is innate and yet requires long practice and experience to achieve it amount to? It would appear impossible to distinguish between having a capacity for concept mastery (and mastery could well require long practice and experience to achieve) and having innately already mastered concepts which took long practice and experience to manifest themselves. Since the former is by far the most plausible explanation, the latter should be rejected, as it amounts to saying that a great deal of learning is necessary for concept mastery to manifest itself (which is a form of the non-innatist claim) and thus to show completeness. It also suggests that the innatist is labouring under the confusion between a capacity (an ability to acquire an ability – which in this case could plausibly be said to be innate) and an ability (to make the relevant kinds of judgment which exhibit concept mastery). [1]
A fall-back position for the innatist would be to distinguish between concept possession and concept mastery and to maintain that the former is innate and the latter acquired. But this is inherently implausible and no-one to my knowledge has suggested how it would work. It is implausible because mastery is complete or near-complete possession. Although the distinction between non-expert concept possession and expert concept mastery makes a good deal of sense, it does not follow that there is always a sharp distinction between the two. It may be that in one context someone may be said to possess a concept and in another to have mastered it as in the latter case, that individual is an expert relative to his peers and shows greater conceptual ability than they. But relative to other experts in a different context he may do no more than possess the relevant concepts. If innatism about concept acquisition and possession is unable to account for concept mastery then its persuasive power and explanatory philosophical potential is greatly reduced.
In what follows, first, the ‘learning paradox’ of Fodor is outlined. Next, the reason why it is not a valid argument is stated and an account is given of how concept acquisition can take place, together with empirical evidence which counts in favour of such an account. Fodor’s positive account of conceptual abilities is described and in particular his account of the acquisition of stereotypes and how they are related to concepts properly so-called is critically examined. It is concluded that the account remains obscure and that the role of automatic and rational elements in stereotype acquisition and the ‘locking’ of stereotypes onto concepts is not convincing. By way of contrast, a Wittgensteinian framework, buttressed by empirical research, is offered as an alternative. This involves both an account of the nature of concepts derived from the work of Geach (1957) and of empirical research on how concepts are formed. This alternative not only yields the objectivity required by any satisfactory account of concept acquisition and possession, but has the further virtue of explaining the contrast between concept possession and concept mastery and thus an important element in the transition from novicehood to expertise.
This allows the distinction between preconceptual attention and judgment on the one hand and non-discursive attention and judgment on the other to be clarified in relation to the development of expertise. Finally, the role of formal education in systematising and extending conceptual structures is outlined and an explanation of the growth of practical expertise as an advanced form of know-how is set out.
The Paradox of Learning Revisited.
According to innatists like Fodor and Chomsky, concept learning is a paradoxical idea whose only coherent resolution is to postulate that concepts are innate.
“In fact, try to define a word like ‘table’ or ‘book’ or whatever and you’ll find that it’s extremely difficult. There is, in fact – just to give one example – a recent issue of a linguistics journal that has a long detailed article trying to give the meaning of the word ‘climb’. And it is very complicated. But every child learns it perfectly right away. Now that can only mean one thing. Namely, human nature gives us the concept ‘climb’ for free. That is, the concept ‘climb’ is just part of the way in which we are able to interpret experience available to us before we even have the experience. That is probably true for most concepts that have words for them in a language. This is the way we learn language. We simply learn the label that goes with the pre-existing concept. So in other words, it is as if the child, prior to any experience, has a long list of concepts like ‘climb’, and then the child is looking at the world to figure out which sound goes with the concept. We know that the child figures it out with only a very small number of presentations of the sound.” (Chomsky, 1988, pp.190-191).
Many innatists take a much more cautious approach than this (eg Fodor 2008), and I will not address all the difficulties straightaway. It is worth noting however that innatists have to postulate learning for concepts to be activated. And, unless this learning is itself non-conceptual, it is hard to see how it could get started, since non-activated concepts would need to be pressed into service in order for learning to be possible and this is impossible, since by their nature, they are inert. As we shall presently see, however, there are deep ambiguities in the way in which innatists approach the issue of learning which makes their position on this issue hard to understand.
How then should we approach the innatist argument for concept possession. If we can show what is wrong with this, then one obstacle to the account offered will have been removed. Luntley (2008) provides detailed gloss on the innatist argument, by elaborating the argument for the innateness of concepts. What follows is an expansion of his argument.
1] A hypothesises that a is F.
Explanation: In order to test the hypothesis that a is F, Amust be able to recognise instances of F and so must already possess the concept of an F.
Response: one could implicitly concede the point for hypothesis formation and testing (HFT) and maintain that HFT is a special case and not the only and certainly not the most basic kind of learning. Other kinds of learning do not require the prior possession of concepts in order to be successful. Training, imitation, instruction and memorisation are all effective ways of learning which do not require HFT in order to work and hence do not always require the prior possession of concepts. But as soon as one looks at these forms of learning and teaching one can see that they do require that the learner already possesses concepts.
For example, training focuses on a target behaviour: A will learn to respond to a red light by braking. But in order to respond to a red light A will have to recognise one first and in order to do so will have to recognise it as a red light and hence must possess the prior concept of red light. Similar arguments can be constructed for the cases of instruction and imitation. So the prior possession of concepts is a general requirement for experiential learning, not peculiar to HFT.
As Luntley puts it:
“Any attempt to explain concept acquisition on the basis of experience will still need to appeal to the grammar of some representational system in order for bare experience to be rendered informative in the appropriate way. And, once again, that will require a tacit grasp of the very concept whose acquisition is being modelled so that the account remains at best an account of label acquisition, rather than concept acquisition.” ((Luntley 2008 p.2).
It is true that the explainer will need to appeal to a representational system in order to explain what is going on. But will the agent whose conceptual apparatus is being assessed require one in order for experience to be informative? If not, then the learning paradox cannot get under way. The red light example looks convincing, but it is necessary to distinguish the case where someone like a motorist recognises a red light and responds by braking because he understands its normative import and a case where an animal responds to one out of, say, unfocused interest or because it is an indicator of danger. Consider the following case of someone conjecturing whether something falls under a particular concept:
1] A hypothesises that a is F
2] a is G (a has more than one property)
It does not follow that A hypothesises that a (an instance of F) is G.
Explanation: 3] A hypothesises that φ is ψ.
3] is an example of a referentially opaque context in which the principle of substitution salva veritate does not hold. Both 1] and 2] can be true but 3] need not be (c.f. Linsky, 1971, pp.1-3). This suggests that one can recognise something as a something without at the same time recognising it as a something else (which it also is). This point is important for the discussion about concept mastery.
Furthermore, the general case of recognition, which is a feature of learning generally, seems also to be referentially opaque. From
4] A recognises a
and
5] a is an F
It does not follow that:
6] A recognises a as an F
Although it does follow that:
7] A recognises an F
as a third person report on A’s recognitional abilities.
This remains so even in the unlikely case that a falls under just one concept. It does not follow that because you need a recognition ability to recognise an F, that you need already to possess the concept F. This would only be the case if another premise were assumed:
hidden premise:
8]In order to recognise an F you need to recognise it as an F
This statement is evidently not true, either when something falls under a single or under multiple concepts and also begs the question at issue, namely whether to recognise an F, you need to recognise it under some concept. To put the matter another way, to recognise a, which is F, you do not need to recognise that a is an F. In reply it could be said that there is no other way for an observer to describe someone’s recognition of something except in terms of their recognising the kind of thing that it is. While this is necessarily true for third-person ascriptions of recognition, it hardly follows that the person concerned recognises that thing as falling under a concept (first-person ascription). As we saw, possessing a concept is to possess an ability exercised in judgement (to be able to judge that a is an F) and ex hypothesi, the recogniser is, as yet, unable to do this. It is of course possible that someone can tell that recogniser that they should recognise the thing as a falling under a concept (second-person ascription) and that this should subsequently lead to first person ascription – ‘I recognise that the a is an F’.[2]
A’s recognition can thus be pre-conceptual and dependent only on non-discursive discriminatory abilities. In objection, it might be said that, in order to make sense of bare experience of a and to account for the cognitive process involved in recognition A would need the concept of an F. But the idea of bare experience is debateable. Animals of all kinds possess recognitional abilities based on their needs. They experience the world, not as something that needs interpretation, but as impacting on their interests. Their ability to recognise salient features of the world renders it more than bare experience, it is the experience that they have as members of a species. And because there is no interpretative processing of bare experience, they do not need any cognitive process to do so either.
Thus, in order for someone to exercise a recognitional ability one does not need to recognise the thing recognised as an instance of an F. This is obviously true of animals. Thus my cat Tibbles can recognise a mouse. But it does not follow that he recognises it as a mouse (as something possessing the marks of the concept mouse, thus allowing material inference to properties like mammalian, omnivore, rodent etc.). Therefore he can recognise a mouse without possession of the concept mouse.[3] But the innatist argument needs the fully fledged concept in order to work. The most economical way of explaining what is going on in this example is that Tibbles can recognise prey and different kinds of it (mice, birds etc.) and his discriminatory and agentive properties are sufficient indication of this, without invoking any discursive abilities on his part, although as pointed out, human observers will feel comfortable about describing Tibbles’ recognitional capacities in conceptual terms. Tibbles is not working with bare experience, but with experience with the characteristically salient features which correspond to the needs of cats.
We need, though, to distinguish this case from a distinct one applying to young human beings, which is clearly important for those who do possess concepts:
9] A recognises (conceptually) a as an F
Suppose:
10] a is also G
It does not follow that A recognises a as a G. In order to do that, A would need the concept of a G.
But, in order to work, the innatist argument requires that if someone recognises something then they recognise it as an instance of a concept. Furthermore, if the innatist argument is valid then A will possess all concepts already and, given that concepts are interconnected it would follow that not only does A possess the concept G, but in non-contingent cases where there is a possible material inference from Fa to Ga that A would recognise the F as a G.
To take an example,
11] An oak tree is a source of fuel.
12] A recognises an oak tree
But:
13] A does not recognise the oak tree as a source of fuel.
More generally:
14] A recognises a φ
is a referentially opaque context, where if
15] All φs are ψs
happens to be true, it does not follow that:
16] A recognises a ψ
There are two reasons for why this could be so:
The first is that A’s recognition is pre-conceptual and depends on non-discursive discriminatory abilities, as in the case of the squirrel. The second is that a rudimentary possession of the concept oak tree need not include the mark is a source of fuel.[4] Even if it did, it is not reasonable to expect that a possessor of concept A in a rudimentary form can make all the material inferences that it is possible to make from possession of the concept.
The innatist argument only seems to be valid in cases where:
17] A recognises a as an F (rather than A recognises an F)
And does so in the the first person sense.
and in this case they clearly bring it under the concept F.
This would be a precondition of also recognising a as a G. However, this begs the question because, as we have already seen, it does not follow from the fact that someone can recognise an F that they recognise it as an F, as an object that falls under a concept.[5] But this is precisely what the innatist argument aims to prove. If it only does so by already covertly asserting that recognition involves concept possession then it cannot prove what it seeks to prove, namely that recognition involves concept possession. The argument gains some plausibility from the fact that, for concept possessors, to recognise something is usually to recognise that something as falling under one or more concepts. But it is misleading in this instance to generalise from cases which apply to concept possessors to those for whom concept possession is precisely what is at issue.
It thus looks as if the innatist argument depends on question-begging, namely the contention that in order to recognise an F a creature must recognise it as an F. In order to keep track of what is at issue in this debate we need to acknowledge the possibility that, contrary to Chomsky’s claim , concept acquisition and hence concept possession are not all-or-nothing affairs, but come in degrees. Furthermore, we do not operate with rigid criteria for when someone has acquired and possesses a concept, our inclination to ascribe these to someone very often depends on context and purpose. Of course, the innatist will deny these points, but what we know about how language develops serves to lend plausibility to this claim, as does the alternative explanation of concept acquisition and possession as being part of the larger process of a neonate’s learning to take part in human society (and hence to learn the mothertongue).
Fodor’s Later Argument for Innatism:
Fodor’s later discussion of the innateness of concepts outlined in LOT 2 is cautious and acknowledges that there are significant gaps in the innatist story that remain to be filled. LOT 2 amounts to a reformulation of the innatist position. Fodor is more troubled about the activation of concepts than he appeared to be in 1975 and than Chomsky appeared to be in 1988. The ‘switching on’ of innate concepts is now a more elaborate process than originally set out. It is envisaged as a two-stage process which initially involves the formation of stereotypes. These are non-conceptual and subjective representations of instances of the concept to be activated. They are non-conceptual because they can be developed without the intervention of concepts, because they are built up through associational processes and do not have the capacity to be used in discursive and ratiocinative acts in the way that concepts can be. The process of stereotype formation seems thus to be associational and mechanical, although Fodor describes stereotypes as statistical representations of experience (op.cit. p.162).