Given what we know about the lower level …

Scientists often give arguments of the form, “Given what we
know about some lower high thing, some higher level thing is not possible.”  Perhaps the most famous case of this for
cognitive science is the 100-step rule that connectionists have alluded
to.  (This goes, very roughly: given what we
know about the neuronal structures underlying cognitive processes and the
reaction times for tests, we know that the mind cannot in general be a serial
computer.)  Stephen Levinson has argued
against innate lexical concepts saying that “There is no biological mechanism
that could be responsible for providing us with all the meanings of all
possible words in all possible languages – there are only 30,000 genes after
all” (Levinson, “Language and Mind: Let’s get the issues straight!” in Gentner
& Goldin-Meadow, Language in Mind,
p. 26).  Any other examples?  Comments on this or other cases?  Cases in cognitive science,
biology, or whatever, welcome.


  1. Some scientists have shown that contrary to the idea you mentioned quoting Levinson, in fact is true that the genome matters: Farah M. and Rabinowitz C. (2003), “Genetic and enviromental influences on the organisation of semantic memory in the brain: is livig things an innate category?”. A. martin and A. Caramazza (eds),A special issue of the Journal Cognitive Neuropsychology, argued that some lesion in specifics regions of the brain during development prevent the individual to acquire semantic knowledge of some object categories.

    Is also known the relevancy of the gene FOX P2 (Dr Simon Fisher).

    Perhaps to me the clue in order to resolve what and how we means with words is in gesture (biomechanical nature of our bodies) and the underlying motor mechanisms (motor cortex in general, mirror neurons…)in addition to perceptual mechanisms.

    Nobody teach another to gesture but gestures are universal in its semantics, and there are highly related to language, and for some scientists the basis for spoken language (M. C.Corballis, M. Arbib, G. Rizzolatti, F. Aboitiz…)

    Many movements are stereotyped and then genetically encoded(¿innate concepts?) but on the other hand can be added to a rich of learneable catalog of new movements (¿experiential concepts?). One problems is with abstract concepts such as “democracy” and the likes.

    The path from gesture to language, very complicated, but not unlikely. There are currently many open doors to formulate rigorous theories arguing that meaning is grounded in actions and perceptual mechanisms providing answers to universals in linguistics and also the mechanisms to learn novel words (contributing to weight evindence toward nativist or interactionists poles because they are in the middle).

  2. My general comment on the degrees of knowledge about levels of organization in the psychological sciences or other sciences is that our lack of knowledge about what happens at the “micro” is not excuse to reject it because the “micro” must play some role in the overall explanation.

  3. kenneth aizawa


    Thanks for both your comments.  The earlier one is more detailed, so will take longer for me to think through.

    The cases I have in mind have a force in the opposite direction from which you suggest.  That is, the point of the arguments I am interested in is to provide evidence against some higher-level theory/postulate on the basis of some lower level evidence.

  4. Hi,

    I don´t know what you think about the form of scientific arguments. Are they correct? Another example may be: “since the atoms of oxygen have these properties (chemistry), the water molecules have these and these behaviors”.

  5. kenneth aizawa


    Thanks for your question and comment.  I tend to think that these kinds of scientific arguments are overrated, but I am wondering if my scepticism is warranted.  Maybe I simply have too few cases in mind.

    As for your comment, I do not mean to challenge or inquire about the use of lower level properties to explain higher level properties.  I am curious about the track record of cases in which lower level hypotheses are used to disconfirm higher level hypotheses.  So, this is perhaps a more refined statement of my concerns.

  6. kenneth aizawa


    Thanks for your comments.

    Just to be clear, I don’t think that Levinson’s view is that the genome is irrelevant to cognition or cognitive processes.  That is certainly not my view.  The issue that concerns me in this example is what a lower level shows (in terms of confirmation) about a higher level. 

    Although I do not know the specifics of Levinson’s view of language acquisition, the little I do know about his views suggests that he might be amenable to the gestural account you suggest. Perhaps gestural abilities do provide a kind of platform for the evolution of language.  But, honestly, I don’t know much about Levinson or the gestural theory of language emergence.

  7. Well, Eric Kandel “cellular alphabet”: common molecular biological pathways within and between species that realizes cognitive functions, used as evidence by John Bickle (Philosophy and Neuroscience 2003) to discredit old functionalism and the argument from multple realization. I think it is an example of lower level data to argue against higher level data.

  8. kenneth aizawa


    This seems to me like a different kind of case.  I’m envisioning something like, given what we know about would have to be the supervenience base for X, X is impossible.  The PotS argument seems to me not to have the level structure to it.  Does that seem right to you?


  9. kenneth aizawa


    This is an interesting case.  Just to plug my own work, Synthese will one day publish a paper that discusses, among other things, Bickle’s book.  I’ve mentioned the basic idea of the paper in an earlier post regarding whether or not all psychological properties are multiply realizable.

    Returning to the case, there are many ways one might try to flesh this example out.  But, what lower level facts are supposed to rule out what higher level hypotheses?  Your example seems not to have this form.

    Let me add one example I think fits the model.  That way folks can potentially turn the tables on me.

    In the 19th Century, one finds something like the following.  Given what we know about the process of inheritance, Darwinian evolution by natural selection cannot give rise to significant evolutionary change. 

    Why?  The idea is based on the (mistaken) idea that inheritance is a blending process.   A white adult and a black adult will produce a grey adult.  (A 19th century example.)  So, any beneficial mutations will be blended away.  This is supposed to show that the Darwinian idea of natural selection (roughly, differential reproduction of heritable variation) cannot work to accumulate beneficial traits, because the heritability part cannot work as Darwin needed it to work.

    So, I have three examples on the table:
    1. neurons against serial computation (100-step rule)
    2. genes against nativism (Levinson)
    3. inheritance against natural selection.

    Comments?  More examples?

  10. gualtiero piccinini

    The following principle seems to me behind this kind of argument, and I can’t find anything wrong with it:

    [A] psychological theory that attributes to an organism a state or process that the organism has no physiological mechanisms capable of realizing is ipso facto incorrect (Jerry Fodor, Psychological Explanation, 1968, p. 110).

  11. 4. ¿abiogenesis (spontaneous generation) against “modern cell theory”?
    -recapitulates much of the misunderstanding about evolution during the early days that you have mentioned, and the ancient view, but generally held idea, of life coming from nothing (or God) rather than other animals (roughly other cells).

    5.¿The neuron doctrine against reticular theory?
    -it was thought that the nervous system was composed by a meswork instead of discrete units (neurons), so given what scientists knew at that time about synapses, neurotrasmitters chemistry… was impossible for a single unit to produce the incredible computations of the nervous system; whereas for a network wil be an easy task.

  12. kenneth aizawa


    I think this principle is correct.  (Carl Gillet has a paper that I like that cites this passage from Fodor.)

    So, what went wrong in the 19th Century “proof” that given what we know about inheritance ( a low level mechanism), natural selection (a high level mechanism) cannot be a major force in evolution?

    I think these arguments rely on certain auxiliary hypotheses that are untested.  These untested auxiliaries are the “Achilles’s heel” of these arguments.

  13. kenneth aizawa


    The neuron doctrine-reticular theory case seems to me plausible, although I do not know the details.  Here is how it might play out.  Golgi might have maintained that, given what we know about the physiology of nerves, there is no way that there could be signaling across a synaptic gap.  Maybe his Nobel Prize lecture would be a good place to look.

    This is a slightly different take, perhaps, on the example, than the one you are giving.  The neuron doctrine versus the reticular theory looks like two theories of the same level structure.  But, insofar as considerations of lower level physiology come into play, one has a promising case.  I’ll look into it.

    I don’t, alas, see how you are viewing case 4. regarding abiogenesis.  Perhaps I don’t know enough of those details.


  14. Tony Dardis

    Hi, Ken,

    There’s an argument against ESP that has this form, I think–no physical mechanism would permit A to know that P (eg some distant happening) by ESP, so no ESP. The same sort of argument works for astrology. Descartes’ Discourse on Method argument for dualism (no physical mechanism could speak or reason, therefore …) might be a variation. It has the same characteristic (about auxilliary hypotheses) that you call attention to: it only works given some unargued assumption about what a physical mechanism can do.


  15. anna-mari

    Dear doctors and others,

    I also think that Fodor`s principle is correct, and I may be in a complete confusion (again), but… is he really speaking about same thing…?

    It seems to me that Fodor is saying that (i) you first have a psychological theory and then (ii) if the theory attributes to an organism a state or process that the organism has no physiological mechanisms capable of realizing is ipso facto incorrect.

    However, it seems to me that it is one thing to say that “ok, this theory cannot be right, because it makes such and such attributions about physiological mechanisms, and there is no such physiological mechanisms in system S. And it is another thing to say that because (i) we know that there is not such and such physiological mechanisms, and then we predict that there cannot be such and such higher level properties (or whatever).

    Are those two things equivalent (epistemologically speaking)? I`d say.. nope. I may be wrong.

  16. kenneth aizawa

    Incidently, notice how these examples seem bear on Patricia Churchland’s comment,

    “After all, one might say, how could the empirical facts about the nervous system fail to be relevant to studies in the philosophy of mind.”  (Neurophilosophy, p. 4)

    These examples seem to give an answer.  Moreover, on her list of possible answers is not “failure of auxiliary hypotheses.” 

  17. kenneth aizawa


    I don’t think I’m getting part of your post.  Here it looks sort of like you are proposing that there is a difference between saying P&Q, on the one hand, and saying Q&P, on the other.

    However, it seems to me that it is one thing to say that “ok, this
    theory cannot be right, because it makes such and such attributions
    about physiological mechanisms, and there is no such physiological
    mechanisms in system S. And it is another thing to say that because (i)
    we know that there is not such and such physiological mechanisms, and
    then we predict that there cannot be such and such higher level
    properties (or whatever).

  18. Abiogenesis, is the higher level data.
    Given that the ancients down to the early modern ages lack the specific knowledge about either biological principles or chemistry, they supposed that life, according to their myths, were an issue of divine comand or spontaneous generations by some intervention of it. Suddenly came a giant of thought, Darwin, and provided a theory of how animal populations appear. The rationale behind the theory of evolution by natural selection was that every animal came from another animal inheriting traits evolved to solve the problems of life.
    But all of this worried reverend Paley and others (Wilberforce) laughing at Huxley wether it was through his grandfather or grandmother that he descent from an ape.

    Given what this people knew about some lower thing, some higher level thing is not possible:

    They lack knowledge about cells and their mechanisms, then for them it was impossible that one animal derived from another so they recurred to “abiogenesis” in the sense mentioned above (spontaneous generations comanded by an intervening God or their prowess).That is, every cell comes from another cell (modern cell theory)and not from nothing or God.

  19. Anna-Mari

    Ken (Can I call you Ken?),

    Yes, something like that were running through my head last night. But this “(i) we know that there is not such and such physiological mechanisms” should be “(ii) we know… (It should be so convenient if one could “preview” these posts and correct all the mistakes before publishing them…)

    I was just figuring out that could the case (i) be more like “top-down”- point of view and the case (ii) bottom-up…

    Could the halting problem, btw, be counted as an example?

  20. kenneth aizawa


    (By all means call me Ken.)

    Here is an interested complementary question.  Are there cases in which scientists argue, “Given what we know about the higher level, we know that this cannot be implemented in such and such a way.”  I would bet there are.  Examples here, too, would be helpful.

    It doesn’t seem to me that the halting problem is an example.  As I have seen the problem developed, the claim is that there does not exist a program P such that for all programs and data P correctly reports whether the other programs halt (don’t go into an infinite loop) on the given datum.  The proof of this that I have seen works by a reductio.  You don’t even have to look at any mechanisms.  The argument is basically, as I understand it, that if you could solve the halting problem (had a program that solves the halting problem), then you could use the solution to solve other problems known to be unsolvable.  Hence the reductio.  So, as I see it, there is no lower level story involved at all in the halting problem case.


  21. Ken

    A summary score here.

    Known (?) failures of the appeal to the lower level:
    1. Physics of matter and a materialist view of mind
    2. Biology (Golgi’s casese (?)) and the neuron doctrine
    3. The physics of life and abiogenesis
    4. The mechanism of inheritance and natural selection

    Successful appeals to the lower level:
    1. Physics and ESP

    Open cases:
    1. Connectionist 100-step rule
    2. Levinson on the human genome and innate lexical concepts

    We are getting kind of a lopsided score against using lower level facts against higher level hypotheses. I figured the best way to solicit support for those who favor these kinds of arguments is to draw attention to this. Any help?

  22. gualtiero piccinini

    A nice example of successful appeal to lower level facts is basic chemistry against homeophathic medicine (the kind that creates purported medications by diluting substances until the probability that any of the original molecules is left is close to zero).

Comments are closed.