1 Reprinted with permission from M. Idesawa (1991), ‘Perception of 3-D illusory surface with binocular viewing’, Japanese Journal of Applied Physics, 30(4B), L751.
2 We will see later that the brain may operate by extrapolating from vast batteries of examples, rather than working with general principles, whether geometric or not. However, this point, while crucially important, does not affect the present argument.
3 Beautiful theoretical work has analysed how this process of finding the best interpretation of the available data might work, and there are many elegant proposals for ‘idealized’ versions of the nervous system (and some of these proposals can be shown to carry out powerful computations). But the details of how the brain solves the problem are by no means resolved (see J. J. Hopfield (1982), ‘Neural networks and physical systems with emergent collective computational abilities’, Proceedings of the National Academy of Sciences of the United States of America, 79(8), 2554–8). Importantly, there are powerful theoretical ideas concerning how such networks learn the constraints that govern the external world from experience (e.g. Y. LeCun, Y. Bengio and G. Hinton (2015), ‘Deep learning’, Nature, 521(7553): 436–44.).
4 Although in a digital computer, cooperative computation across the entire web of constraints is not so straightforward – more sequential methods of searching the web are often used instead.
5 The idea of ‘direct’ perception, which has been much discussed in psychology, is appealing, I think, precisely because we are only ever aware of the output of the cycle of thought: we are oblivious to the calculations involved, and the speed with which the cycle of thought can generate the illusion that our conscious experience must be in immediate contact with reality.
6 H. von Helmholtz, Handbuch der physiologischen Optik, vol. 3 (Leipzig: Voss, 1867). Quotations are from the English translation, Treatise on Physiological Optics (1910) (Washington DC: The Optical Society of America, 1924–5).
7 D. Hume (1738–40), A Treatise of Human Nature: Book I. Of the understanding, Part IV. Of the sceptical and other systems of philosophy, Section VI. Of personal identity.
8 From this point of view, the question of what we are thinking about needs to be kept strictly separate from the issue of consciousness. Two people might both hear an identical snippet of conversation, but in one case, the speakers are talking about a real couple who, by sheer coincidence, are called Cathy and Heathcliff; in another, the speakers are members of a book group, discussing Wuthering Heights. What might be an identical conscious experience of thinking: ‘Poor Cathy!’ is a thought about a real person in the first case (though the hearer has no clue who this person is); in the second case, it is a thought about a fictional character (though the hearer may have no clue which fictional character, or even that she is a fictional character). The nature of consciousness and of meaning are both fascinating and profound puzzles, but they are very distinct puzzles.
9 For example, dual process theories of reasoning, decision-making and social cognition take this viewpoint (see, for example, J. S. B. Evans and K. E. Frankish, In Two Minds: Dual Processes and Beyond (Oxford: Oxford University Press, 2009); S. A. Sloman (1996), ‘The empirical case for two systems of reasoning’, Psychological Bulletin, 119(1): 3–22. The Nobel Prize-winning psychologist Daniel Kahneman is often seen as exemplifying this viewpoint (e.g. D. Kahneman, Thinking, Fast and Slow (London: Penguin, 2011), although his perspective is rather more subtle.
10 For example, P. Dayan, ‘The role of value systems in decision making’, in C. Engel and W. Singer (eds), Better Than Conscious? Decision Making, the Human Mind, and Implications for Institutions (Cambridge, MA: MIT Press, 2008), pp. 51–70.
11 There is a small industry in psychology attempting to demonstrate the existence of ‘unconscious’ influences on our actions (see, for example, the excellent review by B. R. Newell and D. R. Shanks (2014), ‘Unconscious influences on decision making: A critical review’, Behavioral and Brain Sciences, 37(1): 1–19). From the present point of view, this hardly needs demonstrating: we are only ever conscious of the outputs of thought and our speculations about their origins are always mere confabulation. A consequence of this viewpoint is that any demonstrations of the ‘unconscious influences’ on thought do not imply the existence of hidden unconscious pathways to decisions and actions that compete with conscious decision-making processes (although this has been a popular conclusion to draw: see A. Dijksterhuis and L. F. Nordgren (2006), ‘A theory of unconscious thought’, Perspectives on Psychological Science 1: 95–109). On the contrary, such effects are entirely consistent with the cycle-of-thought viewpoint: there is just one engine of thought, the results of which are always conscious, and the origins of which are never conscious.
Must we conclude that each of us is completely oblivious to the processes which generate our thoughts and behaviour? Within a single cycle of thought, I think this is right. But conscious deliberation – pondering different lines of attack on a crossword clue, planning ahead in chess, weighing up advantages and disadvantages of a course of action – involves many cycles of thought. And each cycle will generate conscious awareness of some meaningful organization (a candidate word for our crossword clue, an image of a hypothetical chess move, a snippet of language, a pro or a con). The output of each cycle will feed into the next – if we are to have a stream of coherent thought rather than an aimless daydream.
12 For example, K. A. Ericsson and H. A. Simon (1980), ‘Verbal reports as data’, Psychological Review, 87(3): 215–51.
13 J. S. Mill, The Autobiography (1873).