Abstract
In this chapter I propose a novel thesis about the nature of representation in biological systems. I argue that what makes something a representation is distinct from what determines representational content. As such, it is useful to conceptualize what it is to be a representation in terms of fundamental concepts from biology, particularly the concept of a biological function (or teleofunction). By contrast, representational content is best understood as a structured relation involving two parts, and the explanation of how states of biological systems have content involves the preservation of internal structural relations and causal history.
I review recent literature on the neurophysiologic mechanisms underlying a sensory discrimination task, in which neurons use a variety of mechanisms for encoding, storing, and comparing information about vibrotactile stimuli. These mechanisms include a one-to-one burst code, a temporal code in which periodicity is the operative mechanism, and a variety of rate codes, some with opposite slopes, and some reflecting neither the base nor comparison stimuli, but rather their quantitative difference. In motor cortex, a binary behavioral outcome is reflected in a sigmoidal shape of firing patterns. A theory of biological representation, if it is to be empirically useful, ought to be able to unify these various encoding mechanisms under an overarching conceptual framework that explains what biological representation is and how representational content is determined, from a general standpoint, and I suggest that the theory on offer takes significant steps toward this aim.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The classic “hierarchical processing” view of visual representation adumbrated above is of course complicated by the fact that feedback modulation occurs at every hierarchical level, even prior to primary visual cortex (V1) in the lateral geniculate nucleus of the thalamus. But that does not alter the basic conceptualization of the representational capacity of early sensory neurons as being grounded in a specific causal etiology.
- 2.
This is not to say that edge detectors are, or are not, representational; rather, it is to say that if they are, it is not solely in virtue of their affinity for firing in response to certain types of energy impinging on the periphery.
- 3.
The argument I’m building here is that the fundamental semantically evaluable units are themselves truth-evaluable; hence, those units bear logical structure in the sense I’m using the term here. A different possibility is that the basic semantically evaluable units are not themselves truth-evaluable, but are instead something like subsentential units that concatenate to form larger sentence-like, truth-evaluable complexes. These fundamental units are like words in a language of thought, admitting of syntactic rearrangement which generates the productivity and systematicity of the language of thought, itself responsible for the productivity and systematicity of natural languages (Fodor 1975, 2008). This is the (or at least one of the) standard view(s) in classical cognitive science. However, the key step is the concatenation of numerically distinct, neurologically instantiated symbols: How does it work? How and why do those two neurologically instantiated symbols “come together” in that particular thought, and not some others? In virtue of what is this complex well-formed in its neurological syntax? In virtue of what are these symbols “joined together”? The appeal to concatenating neurologically instantiated symbols at the lowest level introduces a new binding problem: How and why do those particular symbols join together, excluding others, and in what does this joining consist? Just like the more familiar binding problem of explaining how different aspects of an experience (e.g., bluishness and squareness) join together in the brain to form a coherent, unified percept (e.g., as of a blue square), the syntactic binding problem demands an explanation for how distinct symbols join together to form a unified meaningful mental representation. If, however, the fundamental semantic units are, as I suggest, themselves logically structured and hence truth-evaluable, then the syntactic binding problem is avoided for those units. Furthermore, many suppose that even the lowest-level sensory states can accurately or inaccurately reflect peripheral energy states. If that is the case, it follows that the sensory states must have logical structure because neither accuracy nor inaccuracy is possible without it, as argued in the text. There is of course a great deal more to be said on this issue, but I will leave further discussion for a different venue.
- 4.
We’ll need to be careful here: If I “refer” to my dog Mac as “that cat,” it might seem that I’ve mis-referred, but I haven’t. Rather, the ostensive act referred to an individual, and I predicated the property catness of it. The reference relation obtained, whereas I misapplied a predicate of that to which I referred. On the other hand, there are tricky issues regarding reference to nonexistents; can I refer to Sherlock Holmes or unicorns? These are larger issues in the philosophy of language which will not be addressed here; better to understand the simpler kinds of representation first. If you like, consider my claim that there is no mis-reference as both axiomatic and using the word “reference” to mean something like, only the most fundamental kind of reference. The argument for accepting any axiom is, of course, dependent on how well the theory constructed from that axiom works.
- 5.
Akins (1996), for example, argues that the “traditional naturalist” project of Dretske (1981, 1988), Fodor (1987, 1990), Millikan (1984, 2004), and others rests on a mistaken view of the senses, which is that they must be “veridical.” Akins argues instead that sensory systems are not veridical but are what she calls “narcissistic.” That is, they do not “dispassionately” report what is going on out in the world, but instead are highly dependent on local context (as in, “what does this mean for me, the receptor?”). This objection is somewhat strange in that what constitutes veridical representation is precisely the question. Thus, in order to say that sensory systems are not veridical, one must first be committed to some theory of representational content. Her claims that thermoreceptive systems are not veridical, therefore, cannot be used as an objection to the very project of understanding veridicality itself. Akins, apparently, considers thermoreceptors and the neural machinery attached to them to be narcissistic and non-veridical because they do not have linear response profiles, but instead have very complicated response profiles depending on local context. This doesn’t show that they are not veridical, just that they behave according to complicated nonlinear correlations to the environment, and can change in different contexts. These complicated response profiles nonetheless describe mapping functions between relational systems composed of neural activity and relational systems composed of energy states, and bearing these response profiles may very well be what these thermoreceptors and other neural machinery are supposed to do; that is, have the teleofunction of doing.
- 6.
The primary somatosensory cortex is composed of four areas: 1, 2, 3a, and 3b. Each area has a complete topographic map of the body’s surface composed of the receptive fields of the respective neurons. Further, the specialization of peripheral fibers seems to continue in S1; neurons are classified in S1 as rapidly adapting, slowly adapting, or Pacinian, because their firing activities are similar to their respective primary afferents (Romo and Salinas 2001, 109). The areas associated with the rapidly adapting circuit here under consideration are areas 1 and 3b. Within those areas, there are subpopulations, one of which appears to encode stimulus information using a temporal, periodicity-based code (described in the text), and the other using an aperiodic firing rate code (also described in the text). The terms ‘subpopulation-1’ and ‘subpopulation-2’ should not be confused with areas 1, 2, 3a, and 3b. The subpopulations here under consideration are defined by their behavior in this task and are subpopulations of anatomical areas 1 and 3b.
- 7.
A function is bijective if it is injective and surjective. A function is injective (or one-one) if each member of the range is mapped to by only one element of the domain. A function is surjective (or onto) if every member of the range is mapped to by some element of the domain.
- 8.
More specifically,
and 
are isomorphic if there exists a bijective function f: A → B such that for every a and b in A,
aRb iff f(a)Sf(b).If f is surjective but not injective, then
and 
are homomorphic. A variety of other kinds of structure-preserving mappings can also be defined, by selectively loosening certain criteria. See (Swoyer 1991) for some examples. - 9.
Furthermore, note that r 3 only describes the specific relationship discovered among neurons in subpopulation-1 of S1 with vibration frequency. Presumably, the populations of neurons in S2, PFC, VPC, and MPC, which also show positively sloped response profiles, admit of different specific relationships with stimulus frequency (i.e., different baselines and different slopes). They have not however been published (to my knowledge). Note that these different equations don’t change the overall philosophical analysis of biological representation presented here; the theory easily accommodates differing correspondence relations between neural states and represented states, due to the versatility of the concept of structural preservation.
- 10.
Proving isomorphism is not trivial, and furthermore, measurement theory is concerned with one empirical and one numerical relational system, not two empirical relational systems as I’ve described here. But the technical details are outside the scope of this chapter, so I’ve made simplifying assumptions. Namely, I’ll assume that
and 
both have uncountable domains with countable order dense subsets, and their respective relations generate a total order on the domains. This suffices for isomorphism between two empirical relational systems 
and 
(Collins 2010, 406). Whether these assumptions are justified depends on whether making idealizing assumptions in general are justified. - 11.
There are a variety of intermediate events between the stimulator’s vibrating and a particular pattern of neural firing that it caused, say, in S2. For example, ion channels have opened and closed, neurotransmitters have been released, a variety of firing patterns have occurred in upstream areas in the spinal cord, brainstem, thalamus, internal capsule, S1, and so on. Determining which of these causal antecedents is the one to which the representation refers is known as the causal chain problem, which is a problem for any theory of representation that appeals to causation. While I won’t attempt detailed discussion here, a reasonable solution (at least in this instance) is to appeal to teleofunction. The correlation of neural activity in S2 with upstream neural activity is not what confers survival advantage. Rather, by covarying with energy states at the periphery of the organism, in well-defined ways, distinct neural mechanisms can use that activity to perform transformations and computations which ultimately result in behavior that is appropriate to the environment. Hence, it is not arbitrary to claim that the neural activity refers to the stimulator and not some other link in the causal chain.
- 12.
Notice I write that the content is something like … (rather than that the content is …). It is unjustified to assume that the representational content of the lowest-level biological representations instantiated in the firings of individual neurons can be translated straightforwardly into a natural language. Rather, we should be satisfied with describing the content using natural languages, though should not expect a straightforward translation. Furthermore, note that it is equally justified to describe the content as “that thing is vibrating at…” as compared with “the stimulator is vibrating at….” The neural activity under question does not predicate the property of being a stimulator, only the property of vibrating at a certain frequency. Again, for the purpose of describing the content, rather than expressing or translating it, either rendering is acceptable because both expressions refer to the stimulator in this context.
- 13.
As mentioned in the text above, the equation published in Salinas et al. (2000) includes a noise term, so should be written as: r(s) = 22 + 0.7s + σ Œ, where Œ is noise with zero mean and unit variance and σ is the standard deviation of the mean firing rate. Since noise is by definition not a signal, I’ve deleted the final noise term. Nonetheless, noise in neural systems is a significant conceptual and practical issue to be addressed by a theory of representation; any plausible view must be able to account for it because there is no such thing as a noiseless signal in the brain. Many biochemical mechanisms such as ion channel opening, vesicle release, and ion diffusion are stochastic processes, so there will always be “random” electrical activity which is not a result of stimulus representation or neural computation. Although I don’t have space for an in-depth discussion of this here, the theory on offer does have the resources to account for noise in neural systems. The general idea is to distinguish those alterations in the content-bearing properties of a vehicle of representation (e.g., firing rate) which are due to alterations at the source (e.g., vibrotactile frequency) from those alterations which are not due to alterations at the source; these latter alterations constitute noise. A firing rate that is within the range of noise, given its particular (empirically discoverable) noise range, representation function, and the value of its represented parameter, is a noisy-but-true signal, whereas one that is outside the noise range is a noisy-and-false signal.
For more detail see Collins (2010, 359–363).
- 14.
A function preserves a relation R only if aRb → f(a)Sf(b). A function counter-preserves R only if f(a)Sf(b) → aRb, and thus, a function respects R only if it preserves and counter-preserves R; for isomorphism between relational systems, the mapping function needs to respect the relation R. As I mentioned earlier, there are good reasons to relax the strict requirements on isomorphism when using this tool to construct a theory of representation while keeping the basic idea of the preservation of internal relational structure across systems. The type of structural preservation appealed to in the text is a ∆ / Ψ-morphism (Swoyer 1991), which preserves a subset of relations in one system while counter-preserving a subset of relations in the other (in this case, identity is preserved, while greater-firing-rate is counter-preserved; see Collins 2010, 329–330 for the details).
and 
are isomorphic if there exists a bijective function f: A → B such that for every a and b in A,
aRb iff f(a)Sf(b).
and 
are homomorphic. A variety of other kinds of structure-preserving mappings can also be defined, by selectively loosening certain criteria. See (Swoyer 
and 
both have uncountable domains with countable order dense subsets, and their respective relations generate a total order on the domains. This suffices for isomorphism between two empirical relational systems 
and 
(Collins References
Akins, K. (1996). Of sensory systems and the “aboutness” of mental states. The Journal of Philosophy, 93(7), 337–372.
Bargmann, C. I., & Horvitz, H. R. (1991). Chemosensory neurons with overlapping functions direct chemotaxis to multiple chemicals in C. elegans. Neuron, 7(5), 729–742.
Bechtel, W. (2001). Representation: From neural systems to cognitive systems. In W. Bechtel, P. Mandik, J. Mundale, & R. S. Stufflebeam (Eds.), Philosophy and the neurosciences: A reader. Malden, MA: Blackwell Publishers.
Collins, M. (2010). The nature and implementation of representation in biological systems. PhD dissertation, Department of Philosophy, CUNY Graduate Center, New York.
Devitt, M., & Sterelny, K. (1999). Language and reality: An introduction to the philosophy of language (2nd ed.). Cambridge, MA: MIT Press.
Dretske, F. I. (1981). Knowledge and the flow of information (1st MIT Press ed.). Cambridge, MA: MIT Press.
Dretske, F. I. (1988). Explaining behavior: Reasons in a world of causes. Cambridge, MA: MIT Press.
Ferree, T. C., & Lockery, S. R. (1999). Computational rules for chemotaxis in the nematode C. elegans. Journal of Computational Neuroscience, 6(3), 263–277.
Fodor, J. A. (1975). The language of thought. New York: Crowell.
Fodor, J. A. (1987). Psychosemantics: The problem of meaning in the philosophy of mind. Cambridge, MA: MIT Press.
Fodor, J. A. (1990). A theory of content and other essays. Cambridge, MA: MIT Press.
Fodor, J. A. (2008). LOT 2: The language of thought revisited. Oxford/New York: Clarendon Press/Oxford University Press.
Gardner, E. P., & Kandel, E. R. (2000). Touch. In E. R. Kandel, J. H. Schwartz, & T. M. Jessell (Eds.), Principles of neural science. New York: McGraw-Hill.
Gardner, E. P., Martin, J. H., & Jessell, T. M. (2000). The bodily senses. In E. R. Kandel, J. H. Schwartz, & T. M. Jessell (Eds.), Principles of neural science. New York: McGraw-Hill.
Hernandez, A., Zainos, A., & Romo, R. (2000). Neuronal correlates of sensory discrimination in the somatosensory cortex. Proceedings of the National Academy of Sciences USA, 97(11), 6191–6196.
Hernandez, A., Zainos, A., & Romo, R. (2002). Temporal evolution of a decision-making process in the medial premotor cortex. Neuron, 33(6), 959–972.
LaMotte, R. H., & Mountcastle, V. B. (1975). The capacities of humans and monkeys to discriminate between vibratory stimuli of different frequency and amplitude: A correlation between neural events and psychological measurements. Journal of Neurophysiology, 38, 539–559.
Mandik, P., Collins, M., & Vereschagin, A. (2007). Evolving artificial minds and brains. In A. C. Schalley & D. Khlentzos (Eds.), Mental states, Vol. 1: Nature, function, evolution. Philadelphia: John Benjamins Publishing Company.
Millikan, R. G. (1984). Language, thought, and other biological categories: New foundations for realism. Cambridge, MA: MIT Press.
Millikan, R. G. (1989). Biosemantics. The Journal of Philosophy, 86(6), 281–297.
Millikan, R. G. (2004). Varieties of meaning, the Jean Nicod lectures. Cambridge, MA: MIT Press.
Mountcastle, V. B., Talbot, W. H., Sakata, H., & Hyvarinen, J. (1969). Cortical neuronal mechanisms in flutter-vibration studied in unanesthetized monkeys: Neuronal periodicity and frequency discrimination. Journal of Neurophysiology, 32, 452–484.
Mountcastle, V. B., Steinmetz, M. A., & Romo, R. (1990). Frequency discrimination in the sense of flutter: Psychophysical measurements correlated with postcentral events in behaving monkeys. The Journal of Neuroscience, 10, 3032–3044.
Nicolelis, M., & Ribeiro, S. (2006). Seeking the neural code. Scientific American, 295(6), 70–77.
Romo, R., & Salinas, E. (2001). Touch and go: Decision-making mechanisms in somatosensation. Annual Review of Neuroscience, 24, 107–137.
Romo, R., Brody, C. D., Hernandez, A., & Lemus, L. (1999). Neuronal correlates of parametric working memory in the prefrontal cortex. Nature, 399, 470–473.
Romo, R., Hernandez, A., Zainos, A., Lemus, L., & Brody, C. D. (2002). Neuronal correlates of decision-making in secondary somatosensory cortex. Nature Neuroscience, 5(11), 1217–1225.
Romo, R., DeLafuente, V., & Hernandez, A. (2004a). Somatosensory discrimination: Neural coding and decision-making mechanisms. In M. Gazzaniga (Ed.), The cognitive neurosciences. Cambridge, MA: A Bradford Book. MIT Press.
Romo, R., Hernandez, A., & Zainos, A. (2004b). Neuronal correlates of a perceptual decision in ventral premotor cortex. Neuron, 41(1), 165–173.
Salinas, E., & Romo, R. (1998). Conversion of sensory signals into motor commands in primary motor cortex. The Journal of Neuroscience, 18(1), 499–511.
Salinas, E., Hernandez, A., Zainos, A., Lemus, L., & Romo, R. (1998). Cortical recording of sensory stimuli during somatosensory discrimination. Society for Neuroscience Abstracts, 24, 1126.
Salinas, E., Hernandez, A., Zainos, A., & Romo, R. (2000). Periodicity and firing rate as candidate neural codes for the frequency of vibrotactile stimuli. The Journal of Neuroscience, 20(14), 5503–5515.
Searle, J. R. (1992). The rediscovery of the mind. Cambridge, MA: MIT Press.
Swan, L. S., & Goldberg, L. J. (2010). How is meaning grounded in the organism? Biosemiotics, 3(2), 131–146.
Swoyer, C. (1991). Structural representation and surrogative reasoning. Synthese, 87(3), 449–508.
Vallbo, A. B. (1995). Single-afferent neurons and somatic sensation in humans. In M. Gazzaniga (Ed.), The cognitive neurosciences. Cambridge, MA: A Bradford Book. MIT Press.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Nair-Collins, M. (2013). Representation in Biological Systems: Teleofunction, Etiology, and Structural Preservation. In: Swan, L. (eds) Origins of Mind. Biosemiotics, vol 8. Springer, Dordrecht. https://doi.org/10.1007/978-94-007-5419-5_8
Download citation
DOI: https://doi.org/10.1007/978-94-007-5419-5_8
Published:
Publisher Name: Springer, Dordrecht
Print ISBN: 978-94-007-5418-8
Online ISBN: 978-94-007-5419-5
eBook Packages: Biomedical and Life SciencesBiomedical and Life Sciences (R0)