The idea that the brain represents objects, processes and events in the body and the outside world is ubiquitous. It may well be the pre-theoretical default for anyone who believes that the brain, rather than some incorporeal soul, is the seat of perception, intelligence and voluntary behavior.
But you don’t have to spend too long in neuroscience circles these days to come across anti-representationalists. They claim that the brain does not represent anything at all, and that there are other ways to explain how the brain contributes to perception, thought and behavior. The first exponent of anti-representationalism (or antirep for short) that I encountered was a philosopher, so for some years I assumed that discomfort with representation was a result of confusing a pragmatically-defined neuroscientific concept with a metaphysical bugbear: the idea that the mind has only indirect access to the world. I thought, perhaps naively, that the mechanistic question of what the brain does with the energetic patterns impinging on it could be divorced from the philosophical question of whether the obvious fact of unconscious processing (and the possibility of illusions and hallucinations) has any implications for whether conscious perception is “indirect” or not. 1
In the past year or two I discovered that the philosophers’ quibbles about representation have begun to infect neuroscience. In general I am in favor of metaphysical circumspection in scientists: I like the fact that our notions of causation, agency, and emergence remain hotly contested. But questioning the status of representation seems to generate more heat than light. I used to think the debate was a matter of semantic disagreement about terms that could be replaced without any loss of coherence. This may well be true, but as it turns out people cling quite strongly to their preferred terminology, and I am no exception. Just to be a good sport, I have tried to excise the words “represent” and “representation” from casual conversation about the brain, but I found that this was a recipe for extremely roundabout phrasing. No word does quite the job that ‘represent’ does when one wants to describe a neural mechanism underlying some behavior.
Anti-representationalism has set itself up as a radical alternative to mainstream cognitive science: it can sometimes have quasi-political undertones that are catnip to latent iconoclasts and revolutionaries. If you first encountered the concept of representation in antirep literature, you would be excused for supposing that it is simultaneously a vague metaphysical commitment and a sign of egregious moral turpitude. I agree that ‘representation’ is sometimes vague, but no less so than ‘information‘, ‘affordance’, ‘agency’, ‘will’, ‘mechanism’, ‘causality’, and of course ‘emergence‘. In specific contexts all these concepts have pragmatic definitions, but when philosophers and pop writers (and twitterers?) drag them out of these contexts (a perfectly natural thing to do of course), they take on the whiff of the imponderable.
In any case, vagueness is unavoidable for most concepts related to brain and mind, since we still don’t really know how they work. We rely on the connotative webs surrounding each of these concepts when we use them in descriptions of not-yet-understood phenomena. They function as placeholders, which are sometimes called ‘filler words‘. Unless you are engaging in formal mathematics, it is unlikely that more than a handful of the concepts you use have formal, technical definitions. 2 Consider the word “information”. The existence of information theory has misled many people into thinking that we already have a complete mathematically-grounded theory of information. But outside of contexts involving technical aspects of telecommunication, the word “information” is a placeholder, since we do not know how the ‘things’ that we efficiently communicate using Shannon’s framework take on meaning and/or guide behavior. Similar complaints can be raised about the concept of ‘affordances’ in Gibsonian psychology — the word is evocative and resonant, but is not nearly as straightforward to use as, say, ‘length’ or ‘mass’. 3 All scientific concepts that we think of as solid and well-defined started life as vague allusions and analogies. Formal definitions come after pragmatic understanding, not before. This is especially true of the physics-derived concepts that furnish the ontologies of many if not most scientists: just look at the history of the concepts of “mass”, “energy” and “force”.
How long have we been talking about neural representations?
The antirep literature often creates the impression that representations are a recent invention of something called “cognitivism”: an “ism” that seems to be used only in accusative mode, to criticize certain ideas that bubbled up during the ‘cognitive revolution’ in psychology. This revolution began in the 1950s, and took a couple of decades to dethrone behaviorism, which ruled psychology departments in North America since the early decades of the 20th century. Where behaviorism expressed quietism (bordering on denialism) about internal mental states — attempting to explain everything in terms of stimulus-response conditioning — the cognitive revolutionaries embraced the dialectical antithesis. The challengers to behaviorism pointed to the recently-invented digital computers, as well as to Noam Chomsky’s revolutionary linguistics, as the ideal way to talk about those long-exiled mental/neural states.
It’s worth recalling that the cognitive revolution originated in psychology, so it did not necessarily engender a great change in how neuroscientists 4 had been thinking about their subject matter since the dawn of curiosity about the brain. Clearly anyone who poked and prodded at brains during the reign of the behaviorists was happy to ignore the radical elements of their manifesto, despite adopting many of their methods. This attitude towards behaviorism was almost guaranteed, given that most neural phenomena are (1) unconscious from a subjective perspective, and (2) unobservable if one only looks at outwardly manifest behavior. Some — but by no means all — neuroscientists jumped on the cognitive bandwagon, describing their findings in the fashionable terms of the day: information-processing, computation, and sometimes even ‘manipulation of symbols’. ‘Representation’ was often found in the cognitivist word-cloud, frequently as part of the conjunction ‘symbolic representation’.
These terms still pervade neuroscience, but we should keep in mind that the overwhelming majority of neuroscientists have always been experimentalists rather than theoreticians or modelers. Only very rarely do experimentalists propose integrated theoretical frameworks to describe perception and cognition. Given the complexity of experiments, they usually just don’t have the time for deep theory. When an experimentalist claims that faces are ‘represented’ in the Fusiform Face Area, they are saying little more than that activity in this brain area correlates with the perception, recognition and/or imagination of faces. They implicitly assume that the job of figuring out how exactly the face-selective activity pattern arises in the first place, and what the rest of the brain does with this pattern, will be taken up by dedicated theoreticians and modelers. 5
The debate about representation has already accumulated plenty of historical complexity in the decades since the cognitive revolution. But as it turns out, the concept of representation in the brain precedes ‘mainstream’ cognitive science by at least three centuries. A wider historical lens provides a much clearer sense of the lineage of neural representation in all its vague glory. I performed an unsystematic hunt recently to see how far back the concept could be traced, searching for phrases related to neural representation. Below are some results in no particular order.
Here is a paper from 1878 called ‘On the Localisation of the Functions of the Brain’, WJ Dodds:
This strikes me as a sober and clear-minded analysis of motor control, using anatomical logic in a way that cannot be faulted even now, a century-and-a-half later. The description below of a debate about neural function sounds disappointingly familiar, a century-and-a-half of ‘progress’ later:
The excerpt below is from A Dictionary of Psychological Medicine (1892), edited by D Hack Tuke of the University of London. It sets up an analogy between motor expression and abstract thought — a type of analogy that people might profitably make even today.
What is nice about this speculative excerpt is that the connection between motor control and thought (neatly skipping the fraught issue of perception and the direct/indirect business) gives us a mechanistic picture of what neural representation means to the author, and many people who think like him.
We can go back further. The following excerpt is from a book called Pharmacopia, by JA Paris, from 1831.
This reasoning about phantom limbs would easily be recognized by VS Ramachandran as on the right track. Once again, the context of the use of the term representation says quite a bit about what it might mean: the connection with amputation is crucial, since in this case there is no possibility of “picking up” or “coupling with” information from the world. 6
Here’s an entry from way back in 1684: Lexicon medicum græco-latinum. A Physical Dictionary; in which all the terms relating either to anatomy, chirurgery, pharmacy, or chymistry, are very accurately explain’d by Stephen Blancard.
It is important to stress that I am not arguing that the mere antiquity of an idea is evidence for its usefulness or respectability. The example above involves ‘animal spirits’, a concept that few neuroscientists or philosophers would take seriously. 7 The point is that the antirep crowd is very likely to be misfiring if they assume that a neuroscientists’ use of the concept of representation derives from something in the work of Herbert Simon & Alan Newell, Noam Chomsky, George Miller, or Jerry Fodor. The lineage of “representation” in neuroscience is far older than the apparent revolution of the 1950s, which really just restored to psychology a notion that people tended to arrive at as soon as they became acquainted with the idea of the brain as basis for experience. So “representation” has much to do with the scientific revolution as a whole. In fact, the tendency to look to philosophy for the genealogy of ideas about the mind-brain relation may induce us to neglect the history of medicine. Neuroscience might have been influenced by philosophy (and the upstart that was cognitive science), but its real descent is from medicine.
I have not yet made a clear case for the meaningfulness of ‘representation’, beyond the general point about the acceptability of placeholders in science. In the next post, I will try to show that we have plenty of informal data (including introspective ‘data’) to ground our ideas about representation, meaning that ‘representing’ becomes a mere label for an emergent and visible aspect of behavior, and not a hypothetical mechanism or ontologically separate entity/process. In other words, we call a certain sort of perceptible behavior ‘representing’, and that tells us what representing means, which in turn allows us to attribute representation to inferred processes in the brain. More on this soon!
Notes
- I think that ‘directly’ perceiving things ‘as they really are’ is a meaningless concept, but if we accept this idea, then any philosopher worth their salt can always redefine ‘direct’ perception to be something that has a chance of being meaningful. As far as I am concerned, only abstract relations can be known ‘directly’. Perhaps more on this later.
- And even in mathematics a pure context-free and internally consistent formalism has proved a mirage. A quick exploration of the concept of ‘naturalness’ is a good idea here.
- We must politely humor the petulant Gibsonians who seem to think that affordances are just lying around like pebbles, waiting to be “picked up”.
- Here ‘neuroscientist’ is understood narrowly, and somewhat anachronistically, as anyone who studies the brain itself, rather than the mind.
- I imagine some of them think that they themselves will sort out the theoretical issues at some later time, perhaps when a revelatory finding causes insight to crystallize out of the fluid matrix that is their mind.
- Phrases of this sort, alluding to already-available resources, are often found in antirep explanations.
- But having said this, it’s not a stretch to see that electrical signaling and/or the abstract flow of information serve the same role today that ‘animal spirits’ once did.
