My point is that with each new case it is an empirical question whether these models, or models from some other theory, or no models from any theory at all will fit.
You should understand that if you’ve looked through one window,
You’ve looked through the general idea of a window,
Although this claim is entirely conditional and doesn’t apply to stained glass or lancets,
So you have to repeat the procedure repeatedly.
In this paper I criticize a view of functional localization in neuroscience, which I call “computational absolutism” (CA). “Absolutism” in general is the view that each part of the brain should be given a single, univocal function ascription. Traditional varieties of absolutism posit that each part of the brain processes a particular type of information and/or performs a specific task. These function attributions are currently beset by physiological evidence which seems to suggest that brain areas are multifunctional—that they process distinct information and perform different tasks depending on context. Many theorists take this contextual variation as inimical to successful localization, and claim that we can avoid it by changing our functional descriptions to computational descriptions. The idea is that we can have highly generalizable and predictive functional theories if we can discover a single computation performed by each area regardless of the specific context in which it operates. I argue, drawing on computational models of perceptual area MT, that this computational version of absolutism fails to come through on its promises. In MT, the modeling field has not produced a univocal computational description, but instead a plurality of models analyzing different aspects of MT function. Moreover, CA cannot appeal to theoretical unification to solve this problem, since highly general models, on their own, neither explain nor predict what MT does in any particular context. I close by offering a perspective on neural modeling inspired by Nancy Cartwright’s and Margaret Morrison’s views of modeling in the physical sciences.
This is a preview of subscription content, access via your institution.
Buy single article
Instant access to the full article PDF.
Tax calculation will be finalised during checkout.
Subscribe to journal
Immediate online access to all issues from 2019. Subscription will auto renew annually.
Tax calculation will be finalised during checkout.
CA theorists differ in whether they describe their chosen operations quantitatively or qualitatively. I will focus on quantitative descriptions here, since in computational neuroscience most functional descriptions are given in quantitative terms (but see Sect. 5). It should also be noted that these views have developed for each individual theorist over time—Friston (2010) gives a more quantitative reading than Price and Friston (2005) do, for instance. Anderson (2014), in more recent work, seems to soften his CA stance somewhat. I am attempting to focus on the idea of CA itself, so will not catalog these nuances here.
While I have mostly discussed absolutism as a description of the function of areas, it can also apply to functional divisions within an area. In the visual system distinct parts of V1 are classically posited to process displacement, orientation, and wavelength information, and this continues in areas V2 and V3 (Livingstone and Hubel 1988). It is compatible with TA to subdivide areas and give absolutist function ascriptions to each, but this consideration is not pertinent in the arguments I will make regarding MT. See note 4 below, and for more thorough discussion (Burnston 2015).
Influences of disparity information on MT responses had in fact been noticed long before, but a variety of reasons were evinced for considering this influence to not be functionally related to depth perception (Maunsell and Van Essen 1983). As a result, the disparity sensitivity of MT was relegated to footnotes and asides. I discuss this development in more detail in (Burnston 2015).
This speaks against the suggestion that feature-specificity, and thus absolutism, might be saved by subdividing MT into motion- and depth-selective parts, as had been done for color and form responses in V4; see note 1.
Investigation of MT and depth has proceeded considerably beyond these early results, further distinguishing between different types of disparity responses and showing that MT responses are relevant to multiple kinds of depth perception. I discuss these results in detail elsewhere (Burnston, forthcoming).
The movies were not entirely naturalistic—they were “motion-enhanced”, where the “enhancement” consisted of the random insertion of textured objects moving across the screen. The enhancement constrained the movies to meet certain statistics for spatial frequency, which I will not discuss in detail here. It should also be noted that the standards for interpreting model success for naturalistic stimuli are generally different—models for these stimuli generally account for less variance in responses than for more controlled stimuli. I will gloss over this detail as well.
Nor would it be helpful to model “totally naturalistic” stimuli, which contained all possible stimulus elements. Understanding cell responses to such stimuli would require knowing the statistics of the relevant aspects of the stimulus, and this is often difficult to discern in fully unconstrained settings. Drawing conclusions in such studies often requires having a model already in hand of what the cell is responding to (as was, in fact, the case for Nishimoto and Gallant’s)—but, as already shown, a previous understanding of what cells do can be overturned in new contexts. Rust and Movshon (2005) discuss these points in an amusingly titled article, “In praise of artifice,” which criticizes overly optimistic use of “naturalistic” stimuli in computational neuroscience.
Both Koechlin et al. (1999, p. 40) and Simoncelli and Heeger (1998, p. 756) admit that this further investigation and explicit addition would be necessary for generalizing their models beyond motion contexts. Tellingly, despite expressly calling their model a generalization, Nishimoto and Gallant do not even mention depth in their discussion of motion-selectivity properties, other than to note that no disparity is present in their stimuli and therefore is absent from the modeled responses.
Krekelberg and Albright note that there are pragmatic problems with this suggestion—namely the “combinatorial explosion” of needing to look at all of the Fourier components and their combination. Inevitably, this will involve some exploratory investigation. In other work (Burnston, forthcoming) I discuss how contextualist function ascriptions can be used to constrain search through very complex sets of possible contexts, and argue that this kind of search is generally systematic and intelligible despite not attempting a “complete” account of an area’s function.
I would like to point out that while pragmatic virtue is central to the CM view of models, the view is not necessarily instrumentalist, or at least not purely so. For instance, even if a canonical computation were agreed upon tomorrow (which I find doubtful, but not impossible), this would not change the point about explanation in specific cases at all. Put briefly, even if it were “true” in some deep sense that all brain areas fundamentally perform (say) summation and normalization, the point about needing to incorporate contextual knowledge to explain what MT specifically does would hold unadulterated. This requires much more discussion than I can give it here, however (cf. Cartwright 1999).
Adelson, E. H., & Bergen, J. R. (1985). Spatiotemporal energy models for the perception of motion. Journal of the Optical Society of America A, 2(2), 284–299.
Anderson, M. L. (2010). Neural reuse: A fundamental organizational principle of the brain. The Behavioral and Brain Sciences, 33(4), 245–266; discussion 266–313. doi:10.1017/S0140525X10000853.
Anderson, M. L. (2014). After phrenology: Neural reuse and the interactive brain. Cambridge, MA: MIT Press.
Bergeron, V. (2007). Anatomical and functional modularity in cognitive science: Shifting the focus. Philosophical Psychology, 20(2), 175–195.
Bradley, D. C., & Goyal, M. S. (2008). Velocity computation in the primate visual system. Nature Reviews Neuroscience, 9(9), 686–695.
Britten, K. H., Newsome, W. T., Shadlen, M. N., Celebrini, S., & Movshon, J. A. (1996). A relationship between behavioral choice and the visual responses of neurons in macaque MT. Visual Neuroscience, 13, 87–100.
Burnston, D. C. (2015). Perceptual context and the nature of neural function. Doctoral dissertation, University of California, San Diego.
Burnston, D. C. (forthcoming). A contextualist approach to functional localization in the brain. Biology & Philosophy. doi:10.1007/s10539-016-9526-2.
Carandini, M., & Heeger, D. J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience, 13(1), 51–62. doi:10.1038/nrn3136.
Cartwright, N. (1983). How the laws of physics lie. Cambridge: Cambridge University Press.
Cartwright, N. (1999). The dappled world: A study of the boundaries of science. Cambridge, MA: Cambridge University Press.
Chirimuuta, M. (2014). Minimal models and canonical neural computations: The distinctness of computational explanation in neuroscience. Synthese, 191(2), 127–153. doi:10.1007/s11229-013-0369-y.
Chirimuuta, M., & Gold, I. (2009). The embedded neuron, the enactive field? In J. Bickle (Ed.), The Oxford handbook of philosophy and neuroscience. New York: Oxford University Press.
Cummins, R. C. (1975). Functional analysis. Journal of Philosophy, 72(20), 741–765.
DeAngelis, G. C., Cumming, B. G., & Newsome, W. T. (1998). Cortical area MT and the perception of stereoscopic depth. Nature, 394(6694), 677–680.
DeAngelis, G. C., & Newsome, W. T. (1999). Organization of disparity-selective neurons in macaque area MT. The Journal of Neuroscience, 19(4), 1398–1415.
Dodd, J. V., Krug, K., Cumming, B. G., & Parker, A. J. (2001). Perceptually bistable three-dimensional figures evoke high choice probabilities in cortical area MT. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience, 21(13), 4809–4821.
Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.
Grunewald, A., Bradley, D. C., & Andersen, R. A. (2002). Neural correlates of structure-from-motion perception in macaque V1 and MT. The Journal of Neuroscience, 22(14), 6195–6207.
Klein, C. (2012). Cognitive ontology and region- versus network-oriented analyses. Philosophy of Science, 79(5), 952–960.
Koechlin, E., Anton, J. L., & Burnod, Y. (1999). Bayesian inference in populations of cortical neurons: A model of motion integration and segmentation in area MT. Biological Cybernetics, 80(1), 25–44. doi:10.1007/s004220050502.
Krekelberg, B., & Albright, T. D. (2005). Motion mechanisms in macaque MT. Journal of Neurophysiology, 93(5), 2908–2921. doi:10.1152/jn.00473.2004.
Levy, W. B., Hocking, A. B., & Wu, X. (2005). Interpreting hippocampal function as recoding and forecasting. Neural Networks, 18(9), 1242–1264.
Livingstone, M., & Hubel, D. (1988). Segregation of form, color, movement, and depth: Anatomy, physiology, and perception. Science, 240(4853), 740–749.
Maunsell, J. H., & Van Essen, D. C. (1983). Functional properties of neurons in middle temporal visual area of the macaque monkey. II. Binocular interactions and sensitivity to binocular disparity. Journal of Neurophysiology, 49(5), 1148–1167.
McIntosh, A. R. (2004). Contexts and catalysts: A resolution of the localization and integration of function in the brain. Neuroinformatics, 2(2), 175–182.
Morrison, M. (2000). Unifying scientific theories: Physical concepts and mathematical structures. Cambridge, MA: Cambridge University Press.
Nishimoto, S., & Gallant, J. L. (2011). A three-dimensional spatiotemporal receptive field model explains responses of area MT neurons to naturalistic movies. The Journal of Neuroscience, 31(41), 14551–14564.
Nowlan, S. J., & Sejnowski, T. J. (1995). A selection model for motion processing in area MT of primates. The Journal of Neuroscience, 15(2), 1195–1214.
Olshausen, B. A., & Field, D. J. (1997). Sparse coding with an overcomplete basis set: A strategy employed by V1? Vision Research, 37(23), 3311–3325.
Palanca, B. J. A., & DeAngelis, G. C. (2003). Macaque middle temporal neurons signal depth in the absence of motion. The Journal of Neuroscience, 23(20), 7647–7658.
Piccinini, G. (2008). Computation without representation. Philosophical Studies, 137(2), 205–241.
Price, C. J., & Friston, K. J. (2005). Functional ontologies for cognition: The systematic definition of structure and function. Cognitive Neuropsychology, 22(3), 262–275. doi:10.1080/02643290442000095.
Rathkopf, C. A. (2013). Localization and intrinsic function. Philosophy of Science, 80(1), 1–21.
Rust, N. C., & Movshon, J. A. (2005). In praise of artifice. Nature Neuroscience, 8(12), 1647–1650. doi:10.1038/nn1606.
Shadlen, M. N., Britten, K. H., Newsome, W. T., & Movshon, J. A. (1996). A computational analysis of the relationship between neuronal and behavioral responses to visual motion. The Journal of Neuroscience, 16(4), 1486–1510.
Shagrir, O. (2001). Content, computation and externalism. Mind, 110(438), 369–400.
Simoncelli, E. P., & Heeger, D. J. (1998). A model of neuronal responses in visual area MT. Vision Research, 38(5), 743–761.
Snowden, R. J., Treue, S., Erickson, R. G., & Andersen, R. A. (1991). The response of area MT and V1 neurons to transparent motion. The Journal of Neuroscience, 11(9), 2768–2785.
Uka, T., & DeAngelis, G. C. (2003). Contribution of middle temporal area to coarse depth discrimination: Comparison of neuronal and psychophysical sensitivity. The Journal of Neuroscience, 23(8), 3515–3530.
Van Essen, D. C., & Gallant, J. L. (1994). Neural mechanisms of form and motion processing in the primate visual system. Neuron, 13(1), 1–10.
Zeki, S. M. (1978). Functional specialisation in the visual cortex of the rhesus monkey. Nature, 274(5670), 423–428.
I would like to thank Thomas Albright, William Bechtel, Jonathan Cohen, Rick Grush, John Serences, and Ben Sheredos for extremely helpful discussion and comments on earlier versions of this paper. Distant cousins of this material were presented as posters at the Methodology in Neuroscience Workshop at Pitt HPS (November 2013) and the 2014 Society for Philosophy and Psychology Conference in Vancouver, and I benefitted from discussion with audiences at each.
About this article
Cite this article
Burnston, D.C. Computational neuroscience and localized neural function. Synthese 193, 3741–3762 (2016). https://doi.org/10.1007/s11229-016-1099-8
- Computational neuroscience
- Functional localization
- Perceptual neuroscience