Every linguistic community … possesses at least some terms whose associated ‘criteria’ are known only to a subset of the speakers who acquire the terms, and whose use by the other speakers depends upon a structured cooperation between them and the speakers in the relevant subset (Putnam, 1975, p.228).
So claimed philosopher Hilary Putnam over twenty years ago in his famous paper ‘The Meaning of ‘Meaning’’. This quotation outlines Putnam’s ‘division of linguistic labour’. The idea is that language represents a social phenomenon. The meanings for all the words we use are not stored in each of our heads, rather, language is part of the global consciousness: when we use certain terms, we tap into the knowledge that other people have – and we lack.
Riding on the back of global linguistic consciousness has increased more and more, according to Putnam, especially since the advent and growth of science and technology. There are certain scientific terms which all of us laypeople use quite correctly, on a day-to-day basis, such as ‘water’ and ‘gold’. Yet most of us are not aware of the expert or technically correct definitions of such terms. The people who are aware of such definitions are expert scientists.
Although the division of linguistic labour does not play a very large role in the everyday use of terms like ‘gold’ and ‘water’, still it really comes into its own when we get into fights. Suppose I take a look at your diamond engagement ring and comment that it doesn’t look much like a diamond to me; I reckon it’s a papier-mâché copy. You naturally feel pretty upset and so take your ring along to the jeweller, who examines the stone closely with his eyepiece and expert knowledge. Much to your relief, he reassures you that it really is a diamond. So you were using the term ‘diamond’ correctly in this case. The division of linguistic labour acts as a back up in this kind of situation – we laypeople make a ‘quick and dirty’ assessment on the basis of our senses and past experience, and the requisite expert either confirms of disproves our assessment when the need arises.
Putnam’s hypothesis sounds pretty convincing. And it is a maxim according to which we actually behave. We may think that we have chicken pox, but we go to the doctor to check exactly what we are suffering from. If we actually have a case of the measles, then our use of the term ‘chicken pox’ is incorrect in this case. Psychological evidence also backs up Putnam’s hypothesis. For instance, psychologist Barbara C. Malt reports an experiment in which subjects were told that a plant was halfway between a marigold and a dandelion in appearance. They concluded that, with respect to the plant, it made more sense to say, ‘We’d have to ask an expert to tell us which it is’ than, ‘I guess you can call it whichever you want’ (see Malt, 1994, p.42, also Malt, 1990). When push comes to shove, we do not feel that we have the expertise to make our own judgements in these kinds of situation, so we defer to the experts.
Why do we defer to the experts? The consensus of opinion seems to be that they have the requisite knowledge, and that knowledge is not only greater than, but different from, that of the layperson. Expert knowledge is taken to be different from lay knowledge in two ways. Firstly, experts concern themselves with special types of property. According to Putnam, two individuals can only be instances of the same kind of thing if they possess the same ‘important physical properties’. Since scientists (and not laypeople) deal in important physical – which generally translates as microstructural – properties, it is their job to determine whether any number of individuals share the same relevant properties and so whether, in fact, they are the same kind of thing. Putnam’s classic example is that of water – he suggests that the important physical property of water is comprising H2O molecules; no matter how superficially similar two samples appear, they cannot both be water unless they both comprise H2O molecules. To summarise: laypeople deal in surface macro properties accessible to the five senses, whereas scientists deal in inner micro properties. Secondly, thinkers like Putnam imply that experts can conclusively and unanimously answer the question: ‘What kind of thing is that?’ This is because they consider that experts concentrate on those inner properties which reveal the true nature of things – and they interpret ‘true nature’ to mean the unique, unchanging, inherent classification of the natural world which it is the job of science to uncover. Scientific experts should reveal those classifications which cut nature at its joints.1
Psychology and the Explanation-Based View
Sorting individuals into kinds does not only fall within the domain of philosophy. There is an important area of psychology, known as the psychology of categorisation, which examines the way in which we sort and categorise the objects which we experience, the aim being to clarify how we mentally represent concepts. The psychology of categorisation, however, examines the way laypeople – not experts – categorise objects. Over the past twenty years or more, psychological theories of categorisation have changed radically. Yet one implicit assumption appears to remain constant – expert classification is different from lay categorisation and, in times of doubt, laypeople defer to experts.
One of the more recent theories in the psychology of categorisation – the explanation-based view – allows for a highly flexible conception of lay categorisation. It allows for objects being similar to or different from one another in more than one way; it emphasises that classification is based not only on attribute matching, but also on more holistic inferential reasoning; it accepts that our concepts are sensitive to context and do not remain stable across all situations; and it acknowledges that conceptual representation is not necessarily uniform, but varies from person to person. Yet still the commitment to the difference of expert categorisation remains. Psychologists Douglas Medin and Andrew Ortony have described a specific explanation-based model, which they call ‘psychological essentialism’. According to this model, we make categorisation decisions on the basis of surface properties. However, at the very deepest level, our concepts comprise an ‘essence placeholder’ which may contain (among other things) the belief that there are experts who really know what makes an object an instance of a particular concept, even though the layperson cannot specify what that all-important something might be.
These beliefs about the special nature of expert classification may be pervasive, but do they justify the further (frequently made) assumption that there is a fundamental difference between the way experts and laypeople sort objects into kinds? I think not. The traditional distinction between lay and expert categorisation is misconceived, I contend. I want to suggest that expert classification is much more similar to lay classification than researchers have previously assumed. More specifically, I aim to show that the basic principles of the explanation-based view apply to expert as well as to lay categorisation.
Although the explanation-based view is an umbrella term for several different psychological theories of categorisation, still we can isolate a number of core features running through all explanation-based models:
Theories of categorisation predating the explanation-based view take quite a different stand on similarity. Not only do we group objects together because we judge them to be similar, they insist, but that similarity is objective and unchanging, just waiting out there in nature to be noticed by us. The entailed implication (as discussed in the previous section in relation to expert categorisation) is that things are similar in certain fundamental respects only and, once we have based a category around those fundamental similarities, then we have struck gold – we have recognised the category as nature intended. Putnam sums up the spirit well when he says: ‘Once we have discovered that water (in the actual world) is H2O, then nothing counts as a possible world in which water isn’t H2O’ (Putnam, 1975, p.233). H2O comprises the fundamental nature of water. Our work is done.
The explanation-based view, by contrast, recognises that, since different things exhibit numerous similarity relations to one another, similarity alone may be too flexible to account for the categorisation decisions we make. Psychologists Gregory Murphy and Douglas Medin provide a lively illustration: any two objects can be arbitrarily similar or dissimilar, they argue, giving the example of a lawnmower and a plum which are similar in that they both weigh less that 10,000 kg, both did not exist 10,000,000 years ago, both cannot hear well, both can be dropped and so on. Psychologist Lloyd Komatsu gives a second example – a Bedlington terrier seems to share as many similarities with a lamb as with a Great Dane, yet we class it along with the Great Dane and apart from the lamb.
So the question (as highlighted by the explanation-based model) becomes: why do we pick out the particular similarity attributes that we do as the important ones? Similarity alone cannot account for our decisions – we need to ask why. And this question leads into the second broad characteristic of the explanation-based view:
We have already seen that more similarity relations exist than we make use of when categorising objects. This means that we must employ some kind of formula when choosing similarity features which are to count for matters of categorisation – and so categorisation must involve us in more than attribute matching. The question becomes: why do we choose the features that we do and what holds that list of chosen features together – what makes them a coherent and meaningful whole? Take the example of a bird – we might say that a typical bird would have wings, feathers, a beak, would fly, sing and build nests. But a bird is much more than a collection of those features. There is something underlying all those features, something which unites them to form a meaningful whole (a kind of unifying explanation, which gives the explanation-based account its name). What might underlie in the bird example? Well, most of us might plump for some kind of genetic code or structure which all birds share with one another but not with other creatures and which results in the surface features by which we typically identify birds. And this, of course, is where our reliance on the expert scientist comes in. We assume it is a genetic code that is important here, but most of us do not know that for sure. To confirm our suppositions, we need to turn to the experts – geneticists or molecular biologists in this case. We expect that they will be able to tell us what makes a bird a bird, what comprises the essential nature of all birds.
This leads to the third and final defining characteristic:
The explanation-based view is a relational account in two ways. Firstly, as we saw in point two above, it assumes that the features we use to make particular categorisation decisions are related by some kind of underlying unifying structure (explanation) which explains why we treat the individuals which display those features as equivalent and class them together. The features do not stand alone. Secondly, our conceptual knowledge also involves the notion that entities (and categories) relate to one another – they are not isolated individuals. The world is a relational place and things become relevant in relation to other things. For example, our concept of a dog does not simply involve the idea that all dogs share a genetic structure resulting in four legs, two ears, a wagging tail and an infuriating bark. Rather, the concept ‘dog’ becomes relevant because of the way in which dogs relate to human beings and to other entities in the world – dogs bite humans, dogs can be dangerous and this has resulted in the British government introducing the Dangerous Dogs Act, dogs chase cats, people who suffer from asthma may be allergic to dogs and so on. Meaning comes not only from static features and underlying structure, but also from dynamic interaction and connection. In the vocabulary of psychology, when we categorise, we use encyclopaedic (relational) as well as definitional knowledge.
So how do these various characteristics of the explanation-based view (as a theory of how laypeople categorise) apply to expert categorisation? I want to illustrate with examples taken from biological taxonomy. In particular, I want to explore the debate over species concepts which currently rages in biology. This debate concerns how biological entities should be classified, or which characteristics biologists should exploit in order to arrive at categorisation decisions. We will see that different groups of biologists opt for different characteristics, and that the use of different characteristics reflects alternative theoretical motivations.
Philosopher of science David Hull tells us in his book Science as a Process that by 1971 it was possible to isolate three quite distinct schools of biological classification – phenetics, evolutionary systematics (better known as the biological species concept) and cladistics (cladism or phylogeny). In recent years, however, the school of phenetics has largely fallen into disfavour, 2 while both the biological species concept and cladistics continue to thrive. I therefore restrict my discussion in this paper to the latter two schools only.
a. The Biological Species Concept
The biological species concept is typically associated with evolutionary biologist Ernst Mayr. The classic statement of this concept defines species as:
Groups of actually or potentially interbreeding populations which are reproductively isolated from other such groups (Mayr quoted in Sokal and Crovello, 1992).
According to the biological species concept, a species exhibits three fundamental characteristics. Firstly, the members of a species together comprise a reproductive community and so respond to one another as potential mates. Secondly, the species also comprises an ecological unit, interacting as a whole with other species occupying the same ecological area. Thirdly, the species forms a genetic unit which comprises a sizeable interacting gene pool, in contrast to an individual member of the species which ‘is merely a temporary vessel holding a small portion of the contents of the gene pool for a short period of time’ (Mayr, 1984, p.533).
Mayr emphasises that a species is a protected gene pool which has so-called ‘isolating mechanisms’ which protect it from the potentially harmful flow of genes from other pools. Genes from the same pool combine harmoniously because they have become adapted to one another as a result of natural selection, whilst the mixing of genes from different pools will result in disharmonious combinations, hence mechanisms which prevent such mixing are favoured by natural selection.
It is clear, then, that the biological species concept rests upon the presupposition of an inextricable link between taxonomy and evolution (natural selection), and Mayr himself states that: ‘An understanding of the nature of species, then, is an indispensable prerequisite for the understanding of the evolutionary process’ (Mayr, 1984, p.531).
b. Cladism or Phylogeny
The central tenet underlying cladism is that classifications should reflect genealogy or evolutionary branching patterns. In order for a group of organisms to form a species, they must share some kind of common ancestry. Cladism therefore posits a direct link between taxonomy and the history of evolution:
Phylogenetic definitions are thus firmly rooted in the concept of evolution, that is, of common descent...What is both necessary and sufficient is being descended from a particular ancestor (De Queiroz, 1992, p.300).
Perhaps the best known version of cladism is that which was pioneered by zoologist Willi Hennig. Hennig was of the opinion that traditional hierarchical classifications were not complex enough to reflect all the details of phylogenetic development. He therefore concentrated on one particular element in phylogeny – the sister-group relationship. Two taxa (A and B) represent a sister group if they are more closely related to one another than they are to any other taxon (C), the proximity of relationship being based on characters which A and B share with one another but do not share with C. The sister-group relationship is collateral (not ancestor-descendant), hence A and B must share a more recent common ancestor with one another than either one of them does with C. None of the taxa within the statement of a sister-group relationship are said to be ancestral to any other. Despite the fact that Hennig recognised two types of phylogenetic relationship (the sister-group and the ancestor-descendant relationships), he still insisted that a truly phylogenetic classification concerns itself with sister-group relationships alone.
The sister-group relationship can be represented in a phylogenetic diagram (or cladogram), in which speciation events (creation of species) are represented as the splitting of a single line (the stem species) into two (the daughter species which form a sister-group). Hennig argued that when the ancestral species splits, it must be considered extinct and only when this splitting occurs should new species be recognised. Thus an ancestral species cannot exist alongside its descendants and so, as we saw above, two concurrently existing species can only be connected by the sister-group relation, never by the ancestor-descendant relation.
Biological Classification and the Explanation-Based View
As we have seen in the previous section, two quite different accounts of the delineation of species exist in contemporary biology. Each account employs a different underlying principle for guiding classification decisions and each account makes use of a different set of characteristics for grouping individuals into species.
Recall the three core features, which, I argued, characterised the explanation-based account of categorisation:
All three of these defining features can be seen to apply to both the biological species concept and to cladism:
The biological species concept concentrates on facts about breeding for classification purposes, the primary question being whether certain individuals do or do not (can or cannot) interbreed, whether they do or do not (can or cannot) exchange genes. Those that do (or can) interbreed and so do (or can) exchange genes are said to form a coherent species, those that do not (or cannot) are said to fall into separate species. The salient similarity for adherents to the biological species concept, then, is mating preference.
By contrast, the cladist concentrates on facts about common descent for classification purposes. He examines genealogy or evolutionary branching patterns and, he believes, those individuals boasting a common ancestor form a biological species. The salient similarity for adherents to cladism is shared heredity.
Why, then, does the biological species concept count facts about interbreeding and gene exchange as salient? It is because the biological species concept takes the evolutionary process and, more specifically, certain results of natural selection (the adaptation of genes to form a protected gene pool) as the formula according to which the feature(s) that are to count as relevant for species membership should be isolated. This formula specifically highlights interbreeding and gene exchange as the features which make or break a species. And the formula becomes the equivalent of the explanation in the explanation-based account – it underlies and unites species as delineated by the biological species concept.
Cladists, by contrast, operate with a quite different formula. They believe that the history of evolution, resulting in common descent, should be the deciding factor when choosing which properties are to count for the grouping of entities into kinds. This formula leads them to isolate shared common ancestor as the salient characteristic for classification purposes. Once again, this formula represents the explanation which underlies and unifies cladistic classifications.
The biological species concept relates individuals through actual and potential interbreeding, while distinguishing separate species on the basis of reproductive isolation. It also acknowledges that species form ecological units which interact in typical ways with other species in the same ecological area. Species, according to the biological species concept, form part of the wider ecological picture.
Cladism relates and separates individuals through shared ancestry – species can only be isolated by acknowledging that genealogies have merged, interacted and split in the past and will continue to do so in the future. It also insists that, once an ancestral species has split, that species no longer exists – the emergent species can, at the present time, bear no taxonomically meaningful relations to their forebear. The very existence of species, according to cladism, depends on the complex evolutionary environment.
We can therefore see that expert classification (at least as it occurs in biology) bears a strong resemblance to the explanation-based account of lay classification. Expert biologists consider entities to be similar in different ways and which similarities individual biologists choose to focus on for classification purposes will reflect their particular theoretical bias. It seems, then, that the common assumption shared by non-experts, philosophers such as Putnam and psychologists of categorisation that the expert’s schema for classification is somehow radically different from the non-expert’s is misconceived. Certainly, biologists have a deeper knowledge of (amongst other things) the evolutionary process and evolutionary history than do laypeople. They also concentrate on different features when grouping entities into species – perhaps not quite Putnam’s microstructural properties, but characteristics concerning interbreeding, gene exchange and evolutionary branching patterns, characteristics which mean little to the layperson. Yet, more importantly, both schemas share the central notion of an underlying formula or explanation which drives a choice between several characteristics.
Of course, none of this detracts from the credibility of experts, nor should it change our behaviour towards them. Biologists will, by definition, possess greater biological knowledge than the average person-on-the-street and we will still rely on expert knowledge for settling disputes or for confirming our personal opinions – rightly so. The common notion that expert knowledge is greater than lay knowledge is correct (increased knowledge is what individuates an expert, after all). However, the thought which often accompanies this notion – that expert classification must somehow differ fundamentally in kind from lay classification – is mistaken.
Not only does our exploration of biological taxonomy display a structural similarity between expert classification and lay classification (as defined by the explanation-based account), but it also puts paid to the common assumption, articulated by thinkers like Putnam, that categories are simply ‘out there’ in nature waiting to be discovered by virtue of their fundamental characteristics – and that it is the job of scientists to uncover them. Rather, we have seen that biological entities can be grouped into species on the basis of competing formulae and this is possible because of the complexity and richness of those entities and of the similarities and relations between them. Neither should we assume that one of these competing accounts must be right and the other wrong. Of course each has its own peculiar disadvantages. Due to the extremely slow pace at which natural selection operates, for example, it can at times be very difficult to determine what is and what is not a species according to the biological species concept. 3 And, by its very nature, the biological species concept cannot be used to classify asexual entities. On the other hand, the phylogenetic species concept recognises many more species level taxa than the biological, which arguably leads to a decrease in clarity and simplicity. These represent minor problems, however, and in no way serve to demote either account. Both concepts produce workable classifications which reflect important discontinuities in the natural world and both concepts have large numbers of adherents. The fact remains that, in the field of biology, alternative underlying formulae or explanations lead to classifications which exploit different shared properties or similarities between entities. 4
In this paper I have shown that expert classification is much more similar to lay classification than many people believe. More specifically, I have suggested that the explanation-based model of categorisation might help us understand the structure of expert classification as well as that of lay classification. Thus, while expert biologists have a much deeper knowledge of the similarities and relations between biological entities than do lay people, the way in which they make use of that knowledge in classifying entities mirrors the way in which lay people classify. Putnam’s division of linguistic labour remains correct. The definitions of scientific terms do reside in the global consciousness and we do rely on experts and their knowledge in times of uncertainty or disagreement – but the form which expert classification takes is different from what has previously been supposed.
1. It should be noted that Putnam has since abandoned his strong realist views, as expressed in his early works such as 'The Meaning of 'Meaning''.
2. The basic notion behind phenetics is that biological organisms should be classified according to overall morphological similarity. Yet, as the explanation-based account of categorisation makes clear, some kind of independent criterion is needed for assessing which morphological features are relevant for classification purposes. This recognition has led to a waning of interest and confidence in the school of phenetics.
3. So, for instance, it is possible that in some areas two populations are reproductively isolated, whilst in others, they interbreed. This arises from the fact that isolating mechanisms are established at different rates in different populations.
4. See Bryant (Forthcoming) for further, more detailed examples of how expert classification fits the explanation-based model.
Bryant, R. Discovery and Decision: Exploring the Metaphysics and Epistemology of Scientific Classification. Forthcoming from Fairleigh Dickinson University Press.
De Queiroz, K. 1992: Phylogenetic Definitions and Taxonomic Philosophy. Biology and Philosophy, 7, 295-313.
Hull, D.L. 1988: Science as a Process. Chicago: University of Chicago Press.
Komatsu, L.K. 1992: Recent Views of Conceptual Structure. Psychological Bulletin, 112, 500-526.
Malt, B.C. 1990: Features and Beliefs in the Mental Representation of Categories. Journal of Memory and Language, 29, 289-315.
Malt, B.C. 1994: Water is not H2O. Cognitive Psychology, 27, 41-70.
Mayr, E. 1984: Species Concepts and their Application. In E. Sober (ed.), Conceptual Issues in Evolutionary Biology. Cambridge, MA: MIT Press.
Medin, D. and Ortony, A. 1989: Psychological Essentialism. In S. Vosniadou and A. Ortony (eds.), Similarity and Analogical Reasoning. Cambridge: Cambridge University Press.
Murphy, G. and Medin, D. 1985: The Role of Theories in Conceptual Structure. Psychological Review, 92, 289-314.
Putnam, H. 1975: The Meaning of ‘Meaning’. In H. Putnam, Mind, Language and Reality. Cambridge: Cambridge University Press.
Sokal, R.R. and Crovello, T.J. 1992: The Biological Species Concept: A Critical Evaluation. In M. Ereshefsky (ed.), The Units of Evolution. Cambridge, MA: MIT Press.
Mail to: Dr. Rebecca Bryant