Metacognition of Concepts

Project Conference

Mental Representation – Naturalistic Approaches

Two day conference – 20th & 21st May 2016

Venue : Senate House, South Block, Malet Street, London WC1E 7HU


Friday 20th May

10-11.30 Kate Jeffery (University College London) – In Search of the Cognitive Mappa Mundi

12-1.30 Michael Rescorla (University of California, Santa Barbara) – Maps in the Head?

2.30-4 Nikolaus Kriegeskorte (MRC Cognition and Brain Sciences Unit, Cambridge) – CEPT-intentionality: Rethinking representation for computational neuroscience

4.30-6 Manolo Martinez (Universitat Autònoma de Barcelona – Logos Research Group in Analytic Philosophy) – Reference Magnets in Sender-Receiver Models


Saturday 21st May

10-11.30 Rosa Cao (Indiana University) – Indeterminacy and the plurality of neural codes

12-1.30 Jesse Prinz (City University of New York, Graduate Center) – Semantic Naturalism and the Road to Anti-Realism

2.30-4 Marc Artiga (Universitat de Barcelona  and LOGOS) – Teleosemantic Modelling of Cognitive Representations

4.30-6 Peter Godfrey-Smith (The Graduate Center, City University of New York and HPS University of Sydney) – Mental Representation 2016



Kate Jeffery: In Search of the Cognitive Mappa Mundi

Almost 70 years ago, Edward Tolman proposed that the brain creates an internal representation of the environment which he called a “cognitive map”, which could be used for navigation. This proposal sparked much debate about the nature of a map and whether/how the brain could create one. Twenty years later, John O’Keefe began the seminal studies of rat hippocampus that led to his (ultimately Nobel prizewinning) discovery of place cells. The spatially localized activity of these neurons led O’Keefe and Nadel to propose that they form the neural substrate of Tolman’s cognitive map. Once again this claim proved contentious, provoking arguments about whether place cells represent space or represent some other property of the world of which space is merely an example. In the decades since, additional types of neurons have been found that seem indisputably spatial, including head direction cells which respond to head direction independently of location, and grid cells which encode distance.  These neurons, however, seem only to represent local space, i.e. the place where the animal currently is, whereas a true map should encode some kind of distant and relational information as well. This talk will suggest that the problem now is to understand how the local representations implemented by the spatial neurons are linked by their spatial relations so that the animal can navigate across complex space, e.g. out of its burrow to a distant field, or (for humans) across a city. Is there a “master map” of large-scale space, and if so, where is it? What would it look like? How would we know if we found it?


Michael Rescorla: Maps in the Head?

Any creature that moves through space needs some ability to navigate. Edward Tolman proposed that rats navigate using cognitive maps. Numerous researchers have subsequently revisited the cognitive map hypothesis as applied to diverse animal species. What could it possibly mean to say that an animal has a map inside of its head? In what respects do cognitive maps resemble ordinary concrete maps? I will address these questions, with particular emphasis on the distinctive representational format displayed by concrete maps. A concrete map has representationally significant geometric structure. I will explore whether cognitive maps likewise have representationally significant geometric structure.


Nikolaus Kriegeskorte: CEPT-intentionality: Rethinking representation for computational neuroscience

Computational neuroscience views the brain as a computer whose computational principles are incompletely understood and whose architecture differs from those explored by computer science. It’s a computer that controls the animal’s interaction with the world with the goal to achieve survival and reproduction. Like any other computer, the brain is a dynamical physical system and, as such, can in principle be understood without recourse to representational explanations. However, in order to serve its biological purpose, the brain must process detailed information about its environment. As for the computers engineered by humans, the concept of representation helps us abstract from the details of physical implementation and achieve a functional understanding of what the system does. Although the concept of representation (intentionality) is cognitive and computational, it pervades neurophysiology. Whenever a neural activity pattern reflects a property of an external object, the neurophysiologist is tempted to view the activity as a representation. This amounts to imposing a functional interpretation on the experimental finding: that the activity, in the context of the brain’s overall function, serves the purpose to convey information about the object property. Such intentional interpretations are both useful and dangerous. They have great explanatory potential, but are also problematic when engaged in prematurely – before the effects of the activity on other brain regions and on behaviour is fully understood. I will use examples from human and primate brain-activity measurements and deep neural network computational models to discuss representational interpretations with a focus on (1) the representation of the external world in patterns of brain activity, (2) the format in which the world is represented, and (3) the probabilistic representation in the brain of multiple possible states of the organism and its environment. I will attempt the argument that a neuron’s activity can be interpreted as representing the fact that the brain is “currently exploring the possibility that” (CEPT) the organism or its environment was, is, or will be in a particular state.


Manolo Martinez: Reference Magnets in Sender-Receiver Models 

I present an extension of Lewis-Skyrms signaling models in which sender strategies, and payoffs, are sensitive to properties, instantiations of which are informationally connected in various ways. I then argue for a candidate for the propositional content of signals in this model, based on the Lewisian idea that has come to be known as ‘reference magnetism’


Rosa Cao: Indeterminacy and the plurality of neural codes

There are multiple ways of interpreting predictive coding models of perception, ways that appear to be in conflict.  (Does the ascending signal have content at all?  Are predictions about how the world is likely to be, or rather about the proximal consequences of incoming stimuli?)  These disagreements could be resolved if we we had a semantic theory that provided a principled way of mapping signals to contents. However, the debate over indeterminacy in naturalistic theories of content gives us reason to think that no such mapping could be sufficiently determinate to decide the issue, even in principle. This may sound like bad news, but I suggest that it is both what we should expect, given the nature of the systems in question, and appropriate, given our explanatory purposes with respect to them. Depending on what we are trying to explain (e.g. perception vs. learning) it makes sense to highlight different contents for the same signal, while acknowledging that in some sense many contents are present.


Jesse Prinz: Semantic Naturalism and the Road to Anti-Realism

Philosophers of a materialist bent have been attracted to naturalistic psychosematics.  At the same time, they have also tended to favor strong forms of realism.  Putting these together, the hope is that some relations, specifiable in non-semantic vocabulary, put minds into reference relations with real kinds.  This ambition faces insuperable obstacles.  Naturalistic theories are inescapably indeterminate, fundamentally untestable, and posit relations that are only dubiously worthy of the label “real” in the own right.  Some of these theories depend on implausible views about ontology, and even then fail to deliver as promised.  Materialists are better off abandoning strong forms of realism, and recognizing that naturalistic semantic theories can pave the road towards anti-realist ontologies.  This may sound like a big price to pay, but there are also independent reasons for severing ties between naturalism and realism.  Faith in strong realism is not a consequence of naturalism, but a dogma that may run counter to naturalistic scruples.


Marc Artiga: Teleosemantic Modelling of Cognitive Representations

Naturalistic theories of representation seek to specify the conditions that must be met for an entity to represent another entity. Although these approaches have been relatively successful in certain areas, such as communication theory or genetics, many doubt that they can be employed to naturalize complex cognitive representations. In this essay I identify some of the difficulties for developing a teleosemantic theory of cognitive representations and provide a strategy for accommodating them: to look into models of signaling in evolutionary game theory. I show how these models can be used to formulate teleosemantics and expand it in new directions.


Peter Godfrey-Smith: Mental Representation 2016

I’ll put together developments seen over recent decades and try to say where things currently stand.


The conference is part of the project ‘Meaning for the Brain and Meaning for the Person’ funded by the Arts and Humanities Research Council. The support of the Arts and Humanities Research Council, and of the Institute of Philosophy, is gratefully acknowledged.