:: Article

Rattling the Mind

By Richard Marshall.


Doing Without Concepts, Edouard Machery, Oxford University Press 2011

Burroughs said that we should write to make things happen. This is one of those books that does just that. Top cognitive philosopher Jesse Prinz captures the sense that something big is happening here when he writes in his blurb:

“Arguing that cognitive scientists should do away with concepts is like arguing that biologists should do away with genes. Machery’s devastating assault has major implications for philosophy and psychology – it rattles forcefully at the foundations of these fields, and dashes the hopes of those who think we’ll ever find a unified theory of thought. But it is much more than a demolition job. Machery offers a masterful, up-to-the-minute, polemical tour or recent work on learning, induction, and categorization. His bountiful insights and arguments pave a clear and promising path for the journey beyond concepts.”

So what’s going on? This is interdisciplinary work, a mash-up where psychology and philosophy regroup. It’s about the concept of concept. The term ‘concept’ is used by both disciplines but they are used to explain different things. Philosophers talk about concepts as if they are the building blocks of thought. No concepts, no thought is that idea. The big cheese in this work is Jerry Fodor, who wrote his equally big book on Concepts: Where Cognitive Science Went Wrong back in 1998, rocking the landscape in a seismic way just as his Language Of Thought book did way back in the 70s. Fodor is genuinely the guy in this domain, although there are now new people on the block seething with new ideas and nothing is settled.

Fodor rocks because he gives us a way of starting to understand how we are able to think. Following from Chomsky‘s devastating attack on Behaviourism at the end of the 1950s, Fodor put forward a theory describing the actual mechanism that gives us thought. Chomsky, you will recall, argued that there must be something innate about us that allowed us to think and use language. It was something as biological as the mechanism that made us have eyes. The mechanism, whatever it was, must be able to produce the Universal Grammar that allowed everyone to think and pick up a natural language by the time we were four or five. Chomsky left it at that. He didn’t say anything about the mechanics of this idea.
Fodor described the mechanism. The mechanism needed building blocks that enabled thoughts to work as they do. How do thoughts work? Well, thoughts have content. They are about things. When I think, I think about something. So the concepts have to have content too. Another feature of thoughts is that they are combinatorial. I can form thoughts by combining them in new ways. I can think thoughts that have never been thought before, I can express thoughts in sentences that have never been said before, and concepts must therefore be things that can be combined like this to function like that.

The unboundedness of thought, the fact that there is no limit to what can be thought (save limits imposed on us from outside of thought, such as mortality, memory limits and so on) is explained by building blocks that can be combined. Philosophers talk of this in terms of the combinatory and productive features of thought, which are together features of the compositionality of thought. Fodor also explains that thoughts are systematic. By this he means that thoughts can lead to new thoughts without new concepts being needed. For example, if I think that Beckett knew Bataille then I can also think that Bataille knew Beckett. And also Beckett knew Beckett. And also Bataille knew Bataille. What this shows is that a finite number of blocks can create new thoughts without limit. For I can also know that Beckett knew Bataille knew Beckett knew Bataille etc etc forever.

Fodor’s theory of concepts audaciously described the building blocks. Fodor calls these building blocks concepts. Fodor’s concepts are labels for words in a Language of Thought (LOT). This is the language that we think in. We translate LOT thoughts into our natural languages like English or Spanish or Japanese etc. Concepts are without structure. The concept STEWART HOME is just a label, STEWART HOME, inscribed in the LOT. When a Stewart Home enters the environment it triggers in a law like way the concept STEWART HOME in LOT and this allows us to have the thought Stewart Home. The label is just that, a primitive, atomistic thought that represents Stewart Home to us. Fodor can imagine a universe where there are languages with, for example, just the concept STEWART HOME ie a universe where that is the only concept. In that universe you could have the thought Stewart Home and also the thought Stewart Home Stewart Home and also the thought Stewart Home Stewart Home Stewart Home and also etc etc.

Fodor’s theory is counter-intuitive under certain considerations. It seems we need more than just the label STEWART HOME to have the thought about Stewart Home. Don’t we need a concept that decomposes into a bunch of meaningful features, such as ‘author of Red London‘, ‘art striker’, ‘prankster’, ‘afro celt’ etc etc? Fodor’s theory denies this because he needs concepts that are compositional, combinatorial and systematic. If they were decomposable then they wouldn’t be able to function in the way that they would have to if they were to be considered as building blocks, the primitives of thought. The counter intuitiveness is less strong when concepts are considered as the atomized cogs in a working, physicalist machine explanation of how the mind works (Fodor uses computerised modularity to explain the structure of the minds’ mechanism, an idea over-generalised by people like Steven Pinker).

Machery is not writing about concepts in the Fodor sense of the term. So although he is arguing to eliminate concepts from theories of the mind, he is not arguing against the kind of concepts Fodor is considering. It is on the face of it quite consistent for Machery to argue that we should get rid of the term concepts in the sense he is using them to mean and preserve Fodor’s. He can make this move because Machery doesn’t think all concept talk is trying to explain what Fodor’s theory sets out to explain. Psychologists, he suggests, use concepts to explain how we categorise stuff and how induction, the capacity to generalise, is possible. So in this psychological idiom concepts are tools explaining these capacities. So what these psychological theories of concepts are explaining is how we can come to acquire these capacities and why we understand words as we actually do? Because this meaning of concept targets a tool for the pragmatic task of classifying and induction, Machery argues that the theories explaining these functions will not be the same as the theories found in philosophy and don’t need to be. He doesn’t think they will be completely independent of each other, indeed he argues that they will constrain each other somehow, but he is pretty confident that elimination of one idea of concept need not lead to the elimination of the other kind.

Nevertheless Machery argues that it isn’t nonsense to suppose that to think about Stewart Home is to think about everything about Stewart Home. Now Fodor says that when we think of Stewart Home or a class of Stewart Homes or a property of Stewart Home then we have a mental representation that is tokened by the occupants of the property being Stewart Home. There is for Fodor some kind of nomological connection (‘nomological’ meaning the expression of some basic natural laws of causation) between the property of being Stewart Home and the property and it is this nomological connection that makes the thought be a thought about Stewart Home. So here is a way theories might constrain each other.

Machery surveys theories of concepts used by psychologists since the 1970s and comes up with three dominant ones. One is the Prototype Theory. So we come to think of Stewart Home by using a prototype of Stewart Home. This is where we have a typical representation of Stewart Home before us. This typicality need not be perceptual and is what is used to sort out whether this is a Stewart Home before us, or not. A second theory is Exemplar Theory. This is where we identify something by comparing it with all the singular representations of Stewart Home we have had in the past and we then use these stored memories, all of them or a sub set of this totality, to think about Stewart Home. A third rival theory is the Theory Theory. This compares concepts to scientific theories. They provide explanations to the best inferences and store them, such as ‘Neoism subserves deconstruction’.

Machery complains that psychologists have thought that there was a need to choose between these competing theories. They approached concepts as if they were homogeneous. In doing so, concepts were being treated as if they were natural kinds. Natural kinds are the kinds of thing that science targets. Machery argues in his book that this is an error. Concepts as used in psychology are not natural kinds and shouldn’t be treated as such. Instead of being considered homogeneous, they should be thought of heterogeneously. If this is the case then there is no need to choose between the theories because Machery believes that each concept of Stewart Home will be different and independent of the others. And as such, they need not agree with each other. If he’s right, then it follows that we shouldn’t be surprised to find that we do have inconsistent concepts about things.

Machery argues that using the single word ‘concept’ hypnotises psychologists to think that there must be just one kind. He therefore suggests that the word shouldn’t be used. If the three theories about concepts weren’t all labeled identically then they wouldn’t be seen as competing theories. Once understood as independent then the need to choose between them becomes less pressing. And once in this position, Machery argues that it is easier to see how there is a tendency for thoughts to weave in and out of different concepts in ways that can seem to lead to contradictory judgments. So we may think of whales as fish and not fish. They are not fish because they are mammals but they are fish because they match the prototype of fish, and maybe also my stored memories of exemplars of fish.

Machery wants to keep the different concept theories separate because this captures the intuition that we do straddle different concepts much of the time. He wants to be able to grasp the idea that the whale is thought of as being both fish and mammal without there having to be a cancellation of one in deference to the other. Of course, for scientific classification purposes, the whale is best thought of as a mammal because, truth be told, it is a mammal. But of course it’s obviously a fish as well if we classify prototypically, because the prototype of a mammal is nothing like a whale, and the prototypical fish is a lot like a whale. So if we learn about whales in one way we’ll draw different conclusions from those learning about them in another way. Machery thinks we do this all the time and so the heterodox approach to concepts captures genuine psychological processes.


Others in the cognitive philosophical field are not as committed to keeping them as independent from each other as is Machery. Jesse Prinz, who I quoted at the start, wonders whether each of the different concepts – prototypes, exemplars and theories – work more like knowledge stores that we draw on in the process of assigning meaning to something rather than independent theories standing alone. Each of the theories will be drawn on if the others mislead. So in thinking about Stewart Home we may use our prototype of Stewart Homes to reach a conclusion that this is a Stewart Home when reading Blood Rites of the Bourgeoisie, for example. When thinking about a case where the prototype may mislead, say when faced with a Karen Eliot intrusion where the prototype points up a differently gendered phenomenon, then we may draw on stored memories and using a combination of these with the prototype reach a conclusion. When faced with more difficulty we may draw on theory theory to grasp Stewart Home eg ‘the identification of plagarised pornography leads to semantic bafflement’ helps us again to come to a conclusion. The idea is that the three knowledge stores all can contribute to a common process. As knowledge stores for concepts they are theorised as being part of a single theory of concepts rather than three independent ones. Of course, taking this approach undermines Machery’s argument for abandoning the idea of a concept. Instead, it makes the concept of concept more sophisticated and agile and more like a natural kind than ever.

Machery just doesn’t agree that any such process can explain the straddling effect of thought that he believes is intuitively appealing. That we don’t hold consistent beliefs about concepts, as indicated in the illustration above about whales and fish, is data that Machery believes his approach handles better than the ‘knowledge stores’ theory. So Machery believes attempts to bind the three theories of concepts into a single theory is just to fall into the temptation of believing that because there is a single term there must be a single phenomenon. The heterogeneity is what Machery wants to insist on.

One issue that arises from this is how these psychological concepts can be aligned with Fodor’s innate LOT theory where concepts are conceived as building blocks. That there is an issue can be shown (and Fodor has argued this time and again because it’s a problem that gets in the way of many popular theories of learning and cognition) if we return to the feature of language and mind that the philosophical theory of cognition has to explain if it is to count as a genuine theory and not just hand-waving. The unlimited systematic compositionality of language and thought, the idea that finite thoughts can be potentially combined in an infinite number of ways, is a feature that the theories Machery is discussing are not equipped to deliver.

Why not? Fodor uses the example of prototypicality to make his point. Prototypes don’t compose because they have structure and are not systematic. Concepts need to be able to be both compositional and systematic, because thoughts are. The issue is easy to grasp: a goldfish is a protoypical pet fish, but it is neither a prototypical pet not a preototypical fish. If the theory of protypes was right then they would have to be. For this reason Fodor insists that concepts are unstructured labels, stored in the mind and triggered by whatever they label in a law like way. There are several odd things about this: one is that these labels are not learned but acquired and another is that it is possible to have a language with just the word ‘dog’ and nothing else. The first point indicates that we are born with our concepts, just as birds are born with wings, elephants with trunks and so on. This is very odd when you consider that according to this theory we had the concept STEWART HOME thousands of years ago. It also suggests that we have all our concepts fully formed by the age of between four and five. As Fodor insists, by then we are all linguistic geniuses and the issue is not how best to learn stuff but rather what stuff is best to learn?

The second point is the counter-intuitive notion that we have a concept of dog just by dint of having a label for dog that successfully represents any dog we care to have thoughts about. Unlike the psychological theorist of concepts who will say that to have the concept of dog we need to also have the concepts of, say, a mammal, four legs, teeth, bow wow and on and on, a list that, for extremists like Wittgenstein, ends up being the claim that to understand anything you have to understand the whole language, Fodor argues that the compositionality and systematicity of language cannot be explained if those holistic, conceptual package-deal theories are true. So long as language and thought is compositional and systematic then something like the atomistic, building block labeling theory of semantic content along the lines given by Fodor seem to be right. And even if LOT turns out to be in actual fact wrong, even if in theory possible, the other approaches must be wrong in principle. Concepts that can be decomposed can’t be used as building blocks and all three theories from psychology decompose. So Machery and Fodor dramatically depart from each other.

As we noted, Machery thinks that what Fodor is after is different from what the psychologist is after. But if the theories don’t align there is an odd gulf being presented that may be important. Fodor is after the mechanism that lets us think at all. Machery’s concepts are about a couple of things we do with our thinking, that is, classifying things into categories, and using inductive reasoning to generalise from particular to general cases. But if concepts of thought, classification and induction are themselves Fodorian concepts, which they are on Fodor’s theory, then one question this all raises is, what real work is being done by the decomposition of any concepts along the lines suggested by the prototype, exemplar and theory theories of concepts? On the face of it, there can’t be any work being done by them, because concepts are never decomposable in the way those theories assume.

And of course it is this problem, and a whole bunch of other things, that philosophers like Machery and Prinz are working to answer. But this is a book well worth the trouble finding and reading. There are some pretty wild things happening in this area of thought at the moment and its important that we don’t get locked into the idea that only the pop philosophy and psychology books are worth reading or accessible. Machery is a fast thinking and sharp guy, a young turk with loads to say. He’s written something that gives us more smart things to think about and is part of an ever-changing landscape that’s trying to grasp what we are and what we’re becoming. Like a great novel, it is a piece of writing that gives us insights to inscapes that are mainly invisible. It’s the result of painstaking, tough thinking of a kind that is typical of the best philosophers working today. Sure there are things here that make your head hurt – but it’s a thing of beauty like a great poem or novel as well, so we ought to work hard if we want big payoffs.

As experiments and discoveries are starting to fill in the details of how minds work cognitive philosophers are driving forward with ever more wild news from these frontiers. Our self image is becoming deeper and stranger because of this work. There’s an argument I like that concludes that science, armchair philosophy, and the arts are not different in kind but just in degree. Exploration of the very fabric of existence, and of ourselves, requires work in all these domains. At the moment there’s hysteria about the need for everyone to be a scientist at the expense of other areas such as literature, history, philosophy and so on. This reflects a very stupid way of thinking about the relationship between all these disciplines. The humanities, social sciences, arts and sciences are all of a piece and human knowledge increases as all these areas thrive.

The novels of Henry James and Stewart Home have significance for the philosophical theories of cognition and the work of neuro-science because they bring new and significant data to the table that couldn’t be accessed in any other way. So when we read of the reactionary, hug-a-banker Tory and Lib Dem Government in the UK closing down departments of philosophy and stopping grants for student teachers in any subject that is not a science, and when we read of a possible future President of the USA Rick Perry in Texas targeting non-science teaching and departments for closure, this isn’t just an issue for higher education, it’s an issue about what kind of self knowledge do we want to grasp, what kind of society and what kind of people we want to be. The rich and powerful like us to be dumb. Machery’s book kicks sand in their eyes.


Richard Marshall is still biding his time.

First published in 3:AM Magazine: Friday, October 21st, 2011.