In an essay for this year’s Edge Question, I argue that ’associationism’ is not an explanation for how humans and other animals learn, and that the sooner we realise this the sooner the science of learning will take off.
Edge doesn’t allow footnotes or references, so I thought I’d include some here, by way of background.
Popper developed his critique of induction, and his alternative hypothetico-deductive method, in the context of philosophical debates about epistemology and scientific method. In his later work, Popper applied this same line of thinking to the related psychological problem of how organisms -- including humans -- acquire knowledge about the world. Popper argued that because induction was logically invalid, it must also be psychologically invalid; in other words, given that there is "no such thing" as induction in logic, then induction (or ‘association’) is not available as means by which organisms could learn about the world. Popper maintained that "there is no such thing as association", it is "a kind of optical illusion"; "we can, and must, do without [it]" (Popper, 1972).
Aside from Popper, it’s difficult to find any definitive refutations of induction / association. Chomsky’s review ofVerbal Behaviour is often cited in this context, but he never really says that learning by association is impossible in principle, only that it would be extremely difficult in practice. Nevertheless, Chomsky summarises his general attitude to associationism as follows: the framework of “empiricist ideas that has dominated much of modern linguistics, psychology, and philosophy . . . [is] largely mythology, and that its widespread acceptance is not the result of empirical support, persuasive reasoning, or the absence of a plausible alternative.” (Chomsky, 1959)
So how do animals learn? Popper proposed that organisms come into the world equipped with innate knowledge about what to expect (the product of previous iterations of trial-and-error by natural selection); and that, during their lifetimes, individual organisms increase their knowledge by testing their expectations against the world and learning from their mistakes. In later lectures he put it like this: “everything we know is genetically a priori. All that is a posteriori is the selection from what we ourselves have invented a priori." (p46), and "all knowledge is a priori, genetically a priori, in its content. For all knowledge is hypothetical or conjectural: it is our hypothesis. Only the elimination of hypotheses is a posteriori, the clash between hypotheses and reality. [“our senses can serve us . . . only with yes-and-no answers to our own questions" p46-7 (Popper, 1990)] In this alone consists the empirical content of our knowledge. And it is enough to to enable us to learn from experience; enough for us to be empiricists" (p47) (Popper, 1999).
For the alternative information processing account of learning, start with (Dawkins, 1998; Dawkins & Dawkins, 1973) for a primer on uncertainty, information and behaviour, and see Randy Gallistel’s work applying this approach to learning (Gallistel, 1990, 1999; Gallistel, Brown, Carey, Gelman, & Keil, 1991). Gould also has a section on ‘cognitive trial and error’ (Gould, 1986).
Empirical evidence that animals do not ‘associate’, but instead follow naturally-selected decision rules, comes from a wealth of well-documented studies and experiments: (Breland & Breland, 1961; Garcia & Koelling, 1966; Gould, 1986; Gould & Marler, 1987). When discussing how the rat learns to avoid certain foods, Garcia comments: "The hypothesis of the sick rat, as for many of us under similar circumstances, would be, 'It must have been something I ate.'"
When it comes to humans (and some other animals) we have to explain how we are able to learn new things (not just things natural selection prepared us for). But we can apply the same logic to overcome this obstacle. If uncertainty is a prerequisite for the acquisition of information (including learning), then in order to learn novel things humans must be able to generate novel uncertainty. To explain how, I had to combine a number of different elements of the literature that, as far as I am aware, hadn’t been put together before. Spelke & Carey (Marcus, 2004; Spelke, 1994) (Carey, 2011; Gopnik, 1996) talk about innate knowledge, Marcus and Pinker talk about discrete combinatorial (recursive) systems, and Gopnik talks about children (and scientists) learning by trial and error. Put all this together and you get: the acquisition of information (of which learning is one kind) requires uncertainty; humans can generate novel uncertainty by recombining innate knowledge to form new hypotheses, and then putting them to the test. All of this is sufficient for humans to enter the cognitive niche (Boyd, Richerson, & Henrich, 2011; Pinker, 2010).
This essay is an example of itself. In order to explain how humans recombine original elements to form a new hypothesis, I had to recombine original elements to form a new hypothesis. And, the essay itself is a tentative conjecture about tentative conjectures. And in order further the growth of knowledge, critical feedback would be very welcome.
References
Boyd, R., Richerson, P. J., & Henrich, J. (2011). The cultural niche: Why social learning is essential for human adaptation. PNAS, 108, 10918-10925.
Breland, K., & Breland, M. (1961). The misbehaviour of organisms. American Psychologist, 16, 681-684. Retrieved from http://psychclassics.yorku.ca/Breland/misbehavior.htm
Carey, S. (2011). Précis of The Origin of Concepts. Behavioral and Brain Sciences, 34(03), 113-124.
Chomsky, N. (1959). A Review of Skinner's "Verbal Behavior". Language, 35, 26-58. Retrieved from http://www.chomsky.info/articles/1967----.htm
Dawkins, R. (1998). The Information Challenge. The Skeptic, 18(4). Retrieved from http://www.skeptics.com.au/publications/articles/the-information-challenge/
Dawkins, R., & Dawkins, M. (1973). Decisions and the uncertainty of behaviour. Behaviour, 45, 83-103.
Gallistel, C. R. (1990). The organisation of learning. Cambridge, MA: MIT Press.
Gallistel, C. R. (1999). The replacement of general-purpose learning models with adaptively specialized learning modules. In Gazzaniga (Ed.), The Cognitive Neurosciences (2nd ed., pp. 1179-1191). Cambridge, MA: MIT Press.
Gallistel, C. R., Brown, A. L., Carey, S., Gelman, R., & Keil, F. C. (1991). Lessons from Animal Learning for the Study of Cognitive Development. In S. Carey & R. Gelman (Eds.), The Epigenesis of mind: essays on biology and cognition (pp. 3-36). Hillsdale, NJ: Lawrence Erlbaum Associates.
Garcia, J., & Koelling, R. A. (1966). Relation of cue to consequence in avoidance learning. Psychonomic Science, 4, 123-124.
Gopnik, A. (1996). The Scientist as Child. Philosophy of Science, 63(4), 485-514. doi:10.2307/188064
Gould, J. L. (1986). The Biology of Learning. Annual Review of Psychology, 37, 163-192. doi:Doi 10.1146/Annurev.Psych.37.1.163
Gould, J. L., & Marler, P. (1987). Learning by Instinct. Scientific American, 256(1), 74-85. Retrieved from <Go to ISI>://A1987F394500005
Marcus, G. (2004). The Birth of the Mind: How a tiny number of genes creates the complexities of human thought. New York: Basic Books.
Pinker, S. (2010). The cognitive niche: Coevolution of intelligence, sociality, and language. Proceedings of the National Academy of Sciences of the United States of America, 107, 8993-8999. doi:10.1073/pnas.0914630107
Popper, K. R. (1972). Objective Knowledge: An evolutionary approach (Revised ed.). Oxford: Oxford University Press.
Popper, K. R. (1990). A World of Propensities. Bristol: Thoemmes.
Popper, K. R. (1999). All Life is Problem Solving. Bristol: Routledge.
Spelke, E. (1994). Initial Knowledge - 6 Suggestions. Cognition, 50(1-3), 431-445.