My research concerns the nature of mind and cognition and its physical implementation in the brain. This involves me in an embarrassingly large number of areas, including philosophy of representation, semantics (of natural language), and theoretical neuroscience. It may come as some surprise that these are all intimately inter-related in my work. I discuss each of these and their inter-relations below.
Philosophy of Representation.
There are three issues of interest for me here: the emulation theory of representation, the skill theory of spatial content, and what I will call the Kant-Strawson-Evans line on objective content. I will explain each of these in order.
Emulation theory of representation. My doctoral dissertation (UCSD 1995) developed and defended the idea that the brain represents 'external' things, such as the body and the environment, by constructing, maintaining, and using inner models of these entities. This basic idea has been around a while, but I tried to make it somewhat precise by exploiting constructs from control theory, specifically forward models -- which I call emulators. (And for the record, while it is not very common for philosophers and scientists to appeal to forward models in a variety of ways, in the early 90s when I was writing the dissertation, this was not the case! No philosophers had ever heard of the term, and in the various sciences of the brain, it was primarily only motor control that the ideas were employed.) Forward models are devices that mimic the input-output function of some other entity. For example, I argued, following some researchers in motor physiology, that certain structures in the cerebellum and brain stem act as emulators of the musculoskeletal system. This emulator receives efferent copies of the motor signal sent to the body and produces as output a prediction of the proprioceptive signal that the body will produce when it executes that motor command. I showed how such an emulator not only can aid fast goal-directed movements, but also how, if run off-line, can produce 'motor imagery', and might even account for phantom limb phenomena. Similarly, the brain can construct inner models of the motor-visual loop, in order not only to anticipate, on the basis of what is currently seen and what motor actions one executes, what the next visual stimulus will be (for example, if I am looking at a circle and issue a saccade to the left, what I will see next is a circle slightly farther to the right on my retina). I argued that these mechanisms can account for visual imagery. More generally, I have made the claim, and tried to argue, that these mechanisms are responsible for internal representation generally: not only for inner imagery, but that such models can be used for planning, and even for perception (via their employment in Kalman filters -- this line of my work is rather new). I have argued that this account of representation is superior to more standard philosophical accounts -- which typically center on information or causal covariation -- both philosophically and neurobiologically. For more on the emulation theory of representation, see Grush 1995, especially chapters 2 and 3; 1997; 1998a; and especially my BBS article Grush 2004 [pdf here].)
Skill theory of spatial content. The skill theory of spatial content is a theory I have developed in large part via inspiration from the work of Gareth Evans on spatial content -- my attempt has been to fully develop some very suggestive ideas Evans raised (especially in 'Molyneux's Question') but was unable to fully develop himself. The problem is, in virtue of what do we grasp egocentric spatial contents? What is involved in my not just seeing a cup of coffee, but seeing it right here, or over there? It turns out that none of the more obvious answers to this question are adequate. Evans' idea was that such contents were made available through their connection with behavioral dispositions. Information from the cup is such that it produces in me a host of dispositions concerning how to act with respect to the cup, such as reaching in certain ways in order to grasp it, moving my head and eyes in certain ways so as to look directly at it, etc. I develop this idea into what I call the skill theory of spatial content (the term originates in Grush, 1998). According to this theory, an organism's sensorimotor skills are such as to generate an amodal region, centered on the torso (typically), and within which both sensory and motor information can be coherently coordinated. I exploit as an example sensory substitution devices, such as the Bach-Y-Rita device that allows blind people to 'see' via an array of stimulators on the skin, and the sonic guide, a device that turns echoes from objects into audible sound profiles in such a way as to allow blind subjects to 'see' objects via their auditory system. (For more on the skill theory, see Grush 1998d; and especially my 'Skill Theory v2.0' (2007), [pdf here].)
Objectivity. Our experience and thought has a feature which stands in need of explanation. We don't simply, as the proverbial ostrich does, think that the world stops and ends with our experience. Rather, we conceive of our experience as objective -- that is, as experience of a world whose existence is independent of us and our experience. This feature of our experience was first systematically explored by Kant, and was developed in detail by P.F. Strawson and his student Gareth Evans. The common idea (which represents something of a departure from the bulk of the apparatus Kant brought to bear on the issue) is that it is via our ability to represent space that we can conceive of objects existing independently of us: that there are places where those objects are perceptible, but that those are places where I am not now stationed, to paraphrase Strawson's formulation. Evans expanded on this, showing how the idea of an objective world depends on superimposing one's egocentric space on an objective spatial map. I develop this idea, both via the skill theory of egocentric space (discussed above), but also via an account of what an 'objective spatial map' amounts to. I argue that this objective map is best conceived of as an off-line or imaginary view of some spatial realm from a different vantage point (typically but not always one within which one is actually located, such as a view from a city from high above, within which one's own actual location is specified via something like a 'you are here' arrow). By running an off-line 'emulation' (see above) of egocentric space, one can generate a conception of space that is independent of one's actual behavioral exigencies. I argue that mechanisms such as these can account for the objectivity of our experience, the fact that we conceive of our experience as experience of things that exist independently of us and our experience. (For more on my account of objectivity, see my 2000 paper 'Self, world and space' [pdf here], though I should point out that while I still largely agree with everything I said in that paper, I think there are some mistakes in it.)
Semantics.
I have two main interests in semantics. The first is the semantics of singular reference, especially demonstratives and indexicals. The second more general interest is in cognitive semantics, especially Ronald Langacker's Cognitive Grammar.
Singular reference. My interests here are in what might be called grounding expressions (the term is due to Langacker): expressions whose semantic import is constitutively tied to the speech event and participants. These include not only demonstratives (that, this, those, these), and indexicals (I, here, now), but also verbal elements such as tense and possibly modality. In all these cases, these expressions or elements have as part of their semantic import the placing of objects or events in space and/or time relative to spatiotemporally oriented speech event participants (the thing here, the thing over there, the person speaking, the current time, the event that occurred before now, the event that is occurring now). My line on this is that the semantics of these expressions is parasitic on the semantics of representational mechanisms within the individual language user: in large part the same mechanisms that provide the language user with an ability to conceive of an objective spatiotemporal world. I have not published anything in this area (yet), but my main sources of inspiration are the work of Gareth Evans (for demonstratives in particular), and, more generally, Ronald Langacker's Cognitive Grammar framework. I have recently (as of late 2007) turned to demonstratives, and hope to publish on the topic soon.
Cognitive Grammar. History will record Ronald Langacker (the developer of Cognitive Grammar) as the person who finally put our understanding of natural language on the right track, even if a general appreciation of this fact is occurring more slowly than one would hope. And the 20th century's infatuation with trying to understand natural language via the tools of formal syntax will eventually be seen to be the gargantuanly counter-productive historical curiosity it is. Cognitive Grammar has been the primary topic of three graduate seminars I have organized (in 1995, 1998, and 1999). My own research interest in the area is, aside from singular reference (mentioned above), the development of an account of island constraints within a cognitive grammar framework. This is a long-term project, and aside from some very schematic ideas in Grush (1995, chapter 6), I have published nothing on the topic, yet.
Theoretical Neuroscience.
I have interests in what I call theoretical neuroscience. That is, I do not conduct any experiments myself. Rather, I am interested in the related activity of trying to provide adequate theories about the principles that govern the representational activity of the nervous system. I have two loci of interest: spatial representation and the relations between perception, imagery and motor control.
Spatial representation. My interests are in the neurobiology of both egocentric and 'allocentric' spatial representation. My interest in egocentric spatial representation is in part connected with the Skill Theory of Spatial Content (discussed above). I have tried to show that the Skill theory is compatible with what is known about the details of how egocentric space is neurally represented, and to even use the skill theory to shed light on the way the brain represents space. Specifically, the areas of the brain (mostly the right posterior parietal lobe) are, in fact, intimately involved with skilled sensorimotor activity in egocentric space. I also have developed a theory of the mechanisms of allocentric representation, according to which (as mentioned above) it is off-line egocentric representation. My interest in the neurobiology of allocentric spatial representation (aka cognitive maps) is centered on the attempt to discern whether this view is compatible with the neurobiological facts. (For more on the skill theory and its relation to neurobiology, see my Skill Theory v2.0 [pdf here], and 'Space, time and objects' [pdf here].)
Perception, imagery and motor control. Since my doctoral dissertation, I have been arguing that the brain constructs inner models, called emulators, of the body and environment, and that these emulators are the brain's representations. On this view, these representations are invoked, or 'run', through their processing of the same kinds of motor commands that would normally result in movement of the actual body in the environment. More recently, I have tried to show how normal perception is a matter of invoking the same models, that is sensory mechanisms work not by providing information that the brain uses to build up a representation of the environment, but rather that the brain has a self-contained representation of the body and environment whose exact state is set or influenced by sensory mechanisms. I am trying to spell this out by articulating a Kalman filter model of representational activity. On this view, our normal perception is largely a matter of top-down processes resulting from imagery-like mechanisms being run in parallel to the real body, and that these mechanisms provide information which is combined with sensory deliverances in order to yield the final perceptual product.