The Metaphysical Function of Concepts

Linguists and philosophers have been attempting to answer what seems to be the same question for some time. What are concepts? Linguists engaged in this question have approached this issue from a cognitive science perspective, which treats language and meaning as being representations stored in the mind. Philosophers, on the other hand, have traditionally taken an externalist approach, looking for ways of establishing the content of concepts independently of the mind. The standard philosophical account, which has been proposed by philosophers for years, is dubbed the decompositionalist account.

The decompositionalist account, or 'Classical View', claims that concepts are defined by having a number of conditions, which together are both necessary (you need them all) and sufficient (you don't need any more). Under this account, to understand the word is to understand the necessary and sufficient conditions for that concept. Making category decisions, deciding whether or not an object fits into a given category, should therefore simply be a matter of checking off whether the object has all of the conditions on the list for that concept. The standardly given example is of 'bachelor', which is said to consist of the conditions 'male', 'adult' and 'unmarried'. If someone fulfils all three of these criteria, they should, by definition, by a bachelor.

Criticisms of the decompositionalist account are based on a wide variety of counter-examples, theories, and experimental evidence. One of the most convincing arguments, it seems to me, is simply the problem of discovering what these conditions are – a task which is extremely taxing, if not impossible, for all concepts which aren't 'bachelor' (and even 'bachelor' has problems). Even supposing, as defendants of the decompositionalist account might, that these conditions do exist but are just hard to discover through introspection, there remains a major problem in that we don't behave as if we have access to these conditions, even subconsciously.

If making category decisions simply requires checking conditions off from a list, then we would expect it to be a trivial task for humans to be able to make decisions about the concepts they understand. However, this simply isn't the case, with experiments showing a huge variance in the time taken to make the decisions, and with some decisions difficult to call at all, such as the traditional 'is the tomato a fruit?' question.

In looking at this problem Georges Rey (1983) draws a distinction between how objects actually are and how people understand and recognise objects. The first refers to the 'metaphysical function' of a concept, whilst the second refers to the epistemological understanding of the concept. An example Rey gives concerns gender, where Rey assumes some underlying essence which makes people either male or female, yet we ascertain someone's gender by examining not this essence, but some outlying indicators such as clothing, figure, hair and so on. This shows a separation between the actual concepts and the identification procedure by which we attempt to discover them.

A further example comes from cognitive science in the work of Lettvin, Maturana, McCulloch and Pitts (1968), in which the cognitive capabilities of frogs' eyes are examined. The study showed that frogs have, within their retinas, 'bug detectors', which send a signal to the frogs' brain on sensing a moving black dot entering their field of vision. Philosophers have suggested that within this example, the frogs can be said to hold the concept FLY, but have the identification procedure of sensing moving black dots, which can generate false positives.

However, it is not just the case that we might mis-identify objects as belonging to a particular conceptual category, in many cases we simply cannot identify what the essence of a concept is. An example from Wittgenstein (1953) is the word 'game', which seems to have no definable meaning that anyone can discover. Putnam and Kripke divert this question however, by suggesting that the question of whether or not anyone knows the defining conditions is a separate one from askig whether or not there are defining conditions.

Kripke and Putnam focus on 'natural kind terms', such as those for species or naturally-occurring substances. It is no co-incidence that these are precisely the kinds of terms which have been subject to scientific inquiry over several centuries, as it is these kinds which are most able to be studied and classified.

Two of the examples given by Kripke and Putnam are 'gold' and 'tiger'. In the case of 'gold', scientists tell us that gold is a metal and the chemical element with atomic number 79, which describes the number of protons found in the atom's nucleus. Most of us come to know gold, however, as the shiny yellow metal used in jewellery, medals, and so on. This information forms the epistemological concept that we have in our minds, yet which may not even be true. Again we may falsely identify shiny metals as being gold, or it may turn out that some other metal is actually the metal used in jewellery. Regardless, Kripke and Putnam claim, the concept gold is still the same. Their conclusion is that this shows that competent users of a term need not know its proper definition after all, in contrast with the Classical View. However, their claim is that there are still necessary and sufficient conditions for concepts, at least for natural kinds, but that these are part of the metaphysical function and not the epistemological function.

Under this theory, concepts are defined outside of the mind, and are external. Whilst such a claim might seem extraordinary, Putnam (1975) points towards terms within a community whose terms are only known by a subset of that community. An example might be that of the beech tree, which is distinguishable from similar trees, such as the elm, only by biologists or tree specialists, with only these people knowing what the actual difference is.

It may not even be the case that anyone knows the true definition, however, as Rey emphasises. Before atomic theory became an accepted theory, scientists may not have been able to correctly define gold at all. Perhaps even, atomic theory is wrong and no-one can today. Rey summarises this inability for humans to be able to hold all the definitions in their head as the Hypothesis of External Definitions.

In his summary of the arguments advanced by Putnam and Kripke, Rey comments that he has not advanced a positive argument for the Classical view of the core of the concept, on which there is still work to do, but claims to have given the argument that the Classical view has not been rendered false by the problems put forward by S&M.

If the definition of concepts is external to the epistemological content stored within our heads, a valid question to ask is what role this metaphysical function takes. One answer may be that this metaphysical function underlies the stability of concepts which ensures that concepts are consistent across time and different members of a population. Stability of concepts is considered a vital element in rational thought, such as being able to conclude Q from the premises 'if P then Q' and 'P', which might be expressed as follows:

If it is raining, I will stay at home.
It is raining.
I will stay at home.

Here, the argument is only valid if 'raining' means the same in premise two as in premise one. This stability can perhaps be accounted for if we attribute it to the metaphysical definition of rain always being rain, whatever that definition may be. Rain is perhaps not the best example though, as rain can be thought of as a fuzzy concept, with a continuum between what is clearly rain, through light drizzle, mist, dampness, humidity, and so on until at some point it is clearly not raining. Whilst this is conceivably a perceptual problem, resulting from our epistemological lack of the true definition of rain, the problem of fuzzy concepts is acknowledged by Rey, who also points towards the graduated concept 'bald'.

'Bald' has been tackled by philosophers before, trying to solve the following apparent paradox: a man with no hairs is bald, a man with one hair is bald, a man with two hairs is bald. If a man with x hairs is bald, a man with x + 1 hairs is bald. Therefore, a man with a full head of hairs is bald.

This paradox can perhaps be explained by appeal to the differences between the metaphysical function and the epistemological function. The metaphysical definition of 'bald' might be the literal meaning of having no hair on the head at all. All descriptions of men with a small number of hairs on their head being bald are therefore approximations, which can be explained by pragmatic processes.

So far, it has seemed possible that metaphysical concepts could have necessary and sufficient metaphysical conditions outside of the mind. However, I think it is clear that there would be many concepts for which there can be no necessary and sufficient conditions, such as human artefacts like 'chairs' or 'tools'. This doubt may well spread to natural kinds.

Whilst the 'tiger' concept may seem to metaphysically refer to a discrete species which has at least some biological essence or specifying DNA, there may be all kinds of hypothetical situations in which it is unclear whether something is a tiger or not. How many limbs a tiger can lose and still be a tiger (the legs, the tail, the head?), for example. Or what happens if it becomes possible to create a continuum of cross-breeds between tiger and lion? Whilst these can perhaps be argued to be epistemological issues, if there is a clear case of a continuum, there will clearly be metaphysical problems which may not be solvable.

The question of whether or not there are metaphysical truth conditions which neatly categorise the world is one for philosophers to ponder and scientists to discover. I fear that in many cases, the further we delve, the more problematic cases we will find.

From a semantic point of view, it is perhaps worth asking the question of whether stability of concepts can be explained without recourse to metaphysical necessary and sufficient conditions. It is clear that there must be some relation between the epistemological concepts stored individually in our heads and the world, otherwise we wouldn't be able to communicate with one another. However, this does not require concepts to have metaphysical necessary and sufficient conditions, it may be possible that there are metaphysically present groups based upon resemblance, but with no defined edges.

Smith, Medin and Rips (1984) respond to Rey's comments on their earlier paper by stating that their psychologically-evidenced problems for the Classical theory were only ever meant to be epistemological. They also claim that Rey misunderstood the distinction that they drew between identification procedure and concept core, with the latter supposedly referring to the metaphysical. SMR state that both only referred to epistemological categorization. They demonstrate this taking one of Rey's examples: gender. They suggest, with Rey, that the identification procedures might involve clothing, hair, and so on, but suggest that the core might be having sexual organs of a particular type. Whereas Rey takes S&M's 'core' as relating to metaphysical properties, SMR clarify that they intend this to characterize 'the lay person's theory of the nature of things'.

This idea adds an extra layer to the picture about the psychological representation of concepts. Returning to previous examples, we can see that such a theory may well explain the 'psychologically real' distinction between the identification properties of a concept and the psychologically-perceived core. To give an example, if someone is looking through a box of coins and picks one out which is shiny and yellowish, initially assuming this to be gold, they are relying upon a primary identification procedure. However, if they later find another coin in the same box which is also shiny, yellowish, but a lot heavier, they may conclude that this second object is gold and that they must have been wrong about the first. For this to take place, the person must have both an identification procedure, based upon easily-accessible perceptual information, and a concept core, which is connected with encyclopaedic information. This 'core' must be present in order for the person to be able to establish the difference between the object as they first believed it to be, on initial identification, and their later thoughts as to how it differs from 'real' gold. Note, however, that both are epistemological beliefs, and both may be wrong when it comes to the metaphysics of the object (perhaps neither is gold).

One of the motivations that Rey gives for having metaphysical necessary and sufficient conditions is down to the stability function, which allows us to refer to the same object across time ('intrapersonal') and between persons ('interpersonal'). Without them, Rey suggests, 'any two people who use different procedures, or one person who uses different procedures at different times, would ipso facto have different concepts'. He gives an example of people identifying birds by either visual information (features, colours, and so on) or aural information (the sounds they make and songs they sing). If the identification procedure also played the role of the psychologically-held concept, this would make a sight-identified-blackbird different from a sound-identified-blackbird.

Whilst this would indeed be the case, the claim made by SMR is more subtle in that it is not the epistemological identification procedures which are the content of the psychological concept, but the (still epistemological) psychologically-held 'core'. If this core is separate from external metaphysical conditions, however, this is still room for the core to be different across people and time, and this might become evident if, say, biologists got into an argument over whether a subspecies counted as being blackbirds.

In their response to Rey, SMR admit that psychological concepts cannot account for absolute stability of concepts, as used in formal reasoning, philosophy, and so on. However, they suggest a subtle distinction between stability of concepts as meaning sameness of concepts (required for formal logic) and similarity of concepts, which they dub 'communality' to avoid confusion. This communality is used to describe how psychological conceptual cores can be similar across time and people.

Before going into details about this, SMR first question Rey's arguments on why psychological categorisation could not give rise to communality. Rey suggests that the different of ways in which we come to know about the world, the many different identification procedures we might have, makes it impossible for concepts to be similar enough to give rise to any stability function. Rey gives the examples of modern spectroscopy and carbon-dating as methods of identification which presumably don't alter the stability of the things they identify.

This diversity argument is though, as SMR point out, entirely a priori, based upon the feeling that psychological categorisations are unlikely to be similar enough to allow for the stability condition. SMR suggest that empirical evidence finds considerable communality in identification and core properties. They point to experiments requiring college students to list attributes for classes of objects in which there were some attributes used by the majority of subjects. SMR also point towards research suggesting a considerable consensus among core properties such as 'rudimentary biological beliefs that an offspring must be the same biological type as its parents'.

Whilst such 'empirical evidence' can hardly be the basis for claiming that metaphysics isn't required to explain the stability function, Rey's arguments become irrelevant anyway if we accept the psychological identification procedure/concept core distinction, as then differences in identification procedure (such as carbon-dating, modality, etc) don't affect stability of concepts because it is the concept core that counts.

Even if it can be shown that psychologically-held concept cores can, it may be the case that we only need a rough communality for everyday purposes. This seems to me to be heading in the right direction. Whilst differences in core concepts might cause some objects to be categorized differently by different people, or by one person across a time period, for the vast majority of objects and circumstances there are unlikely to be any disputes, particularly if pragmatics is taken into account. To take the example I used above, for example, where 'raining' must be consistent across the two premises in order for the conclusion to be valid.

In a real life scenario, if there is rain pouring down outside, there isn't going to be any doubt about it. The identification procedure of looking for the rain falling, hearing it, etc are all going to flash positive, no matter if they differ across people slightly. Furthermore, the core concepts aren't going to be too different either, involving something like lots of water droplets falling from the sky. However, should it only be drizzling to a slight degree, there may well be some disagreement over whether or not two people consider it to be raining. Not only might the identification procedures of two different people bring up different results (one person can see the drizzle but the other can't hear it), the slight difference in core concepts might mean that this borderline example returns different results to the different people. For one person, rain might be closely connected with the understanding that it causes things to get wet. For another, it might be more connected with falling in discrete droplets which are big enough to sense. Faced with a situation where rain is falling in such tiny droplets as to not be easily sensible, yet still enough dampness in the air such that you might get your hair wet – there could easily be a disagreement.

To go down such a path as this isn't to suggest that metaphysical properties can have no influence upon the epistemological concepts held within our head. Far from it, it is the facts about the world which make using our epistemological information to create mental categories at all useful. If the world was full of random things which didn't behave in consistent ways, our psychological categories would be meaningless. However, the hard-and-fast rules of physics mean that the physical world behaves in predictable ways, and the gradual evolution of life has ensures that the living world acts in largely predictable ways. Furthermore, for us as humans, it has been evolutionarily advantageous to have the kinds of minds which can store concepts which capture useful generalisations about the world, such as that there are 'tigers' which are a category of animals that always seem to have teeth and eat meat, and so on, and that there is 'rain', which is water falling from the sky in such a way as to get things wet, and make it hard to see, and so on. These concepts are informed by the metaphysics of the world, but form cognitive concepts in our minds, and have cognitive effects when lit up, and so on.

These metaphysically-informed concepts aren't formed on individual epistemological evidence however, that would require each of us to have enough evidence through contact with tigers to be able to individually make the generalisation establishing the concept within our minds. Instead, it is the collective epistemological information gathered by a community of people over time which enables the concept to be passed from person to person, from generation to generation.

The exact way in which these concepts are passed along, becoming more-or-less stable, can be the subject of much inquiry, but a good suggestion would be that firstly it is the creation of a linguistic label ('tiger') which causes people to look for a connected concept, perhaps in the form of a template. This space may then be filled by finding shared confirming or negating examples, such as the toddler who initially over-generalises 'cat' to refer to all four-legged animals. I would suspect also that this would be combined by underlying innate pragmatic rules based upon optimal relevance, which overcome the poverty of the stimulus and allow us to make inferences upon the linguistic information given to us.

Concepts are also subject to change when faced by new epistemologic information. As the sciences discover ways of examining the world in ever more detail, the generalisations that we make may change. Chemists may discover, for instance, that all of what we currently called gold (metal with atomic number 79) actually splits into two completely different subtypes, only one of which reacts with some acid. They decide that only this latter subtype is 'real' gold, with the other being some other substance. This narrowing of the concept, based upon the new epistemologic evidence coming from the metaphysical properties of the object, may or may not transfer to the wider community. Whilst this is only a hypothetical case, examples such as this must have happened throughout history.

For a community concerned with making jewellery, for example, iron pyrites may have initially been used in simple jewellery in an identical way to gold, and have thus been considered the same concept. However, once it was discovered that iron pyrites isn't as malleable or uncreative as gold, and thus doesn't make such good jewellery, the community would soon develop a different concept for it, even if they haven't yet developed identification procedures able to tell them apart easy (this is what 'Fool's Gold' describes). The community would soon become better at identifying them, however (iron pyrite's gold surface can be scratched off), and would thus end up with contrasting identification procedures as well as concept cores.

Metaphysics is also not the only influence upon our psychological concepts. Many of our concepts are social, used for describing beliefs, mental properties, and so forth. These are going to achieve communality through social interaction rather than any reference to metaphysical properties. An example might be the term 'Indian', which has been used as a social label for someone from any part of the Indian sub-continent, but but which, through social understanding, is being narrowed to refer to persons specifically from the country of India.

In summary, I think we can conclude that psychologically-held concepts do not have to have necessary and sufficient conditions (ala the Classical View). Additionally, it seems sensible to suggest the existence of both psychologically-held identification procedures and psychologically-held concept cores. The former lies within the domain of epistemology and sense perception, and the latter lies within the domain of naïve psychology and cognitive representations and forms the psychological concept. This leaves two questions unanswered. The first is whether there exist any metaphysical necessary and sufficient conditions. This is a question for philosophers, although I would suggest that any such claim is open to problems as it is easy to consider hypothetical situations which refute the example. The second question is the precise relationship between the metaphysical environment and the stability of psychologically-held concepts. Even though there does not need to be a one-to-one correlation, an explanation of how metaphysics might contribute to concept communality would be of benefit, and this is partially a question of epistemology.

Finally, it is worth pointing out that for communication, the stability of psychologically-held concepts might not be that important after all, as pragmatics seems to allow a precise fixing of the communicated meaning regardless of whether this may have significantly narrowed or widened from whatever conceptual core may be held. Perhaps concepts are simply NULL.


  • Kripke, S. 1972. Naming and necessity. In D. Davidson & G. Harman (eds.), Semantics of Natural Language.
  • Lettvin, J.Y., Maturana, H.R., McCulloch, W.S. and Pitts, W.H.. 1968. What the frog’s eye tells the frog’s brain. In Corning, W.C. and Balaban, M. (eds.), The Mind: Biological Approaches to its Functions. pp 233-258.
  • Rey, G. 1983. Concepts and stereotypes. In Cognition 15:237-62.
  • Putnam, H. 1999. Is semantics possible? In Margolis, E. & Laurence, S. (eds.), Concepts: Core Readings. Cambridge, MIT Press. 1998, chap 7.
  • Putnam, H. 1975 The meaning of 'meaning'. In Gunderson, K. (ed.), Language, Mind and Knowledge.
  • Smith, E., and Medin, D. (1981) Categories and Concepts. Cambridge, MA, Harvard University Press.
  • Smith, E., Medin, D. and Rips, L. 1984. A psychological approach to concepts: Comments on Rey's "Concepts and stereotypes". In Cognition 17.
  • Wittgenstein, L. 1953. Philosophical Investigations (3rd edition). New York, Macmillan.