Sunday, January 5, 2014

Why you can never recall your dreams

One of life's more interesting phenomenons is waking up from a dream in which you have just unravelled one of life's great mysteries, created the most amazing invention, or imagined the poem you always meant to write. You jump out of bed and rush to your writing desk.  Pen in hand, you sit poised, and ... gone!  Where did it go?

I got a possible answer from Stephen Pinker's The Language Instinct. To explain it, he starts by trashing the Sapir-Whorf hypothesis. That hypothesis was based on bogus data from indigenous North American Indian languages, regarding the influence from language on the world view of the speaker of that language. For example, American Indian languages had such an arcane verb tense and aspect system that their concept of time must be much different from ours. Eskimos had so many words for snow that their visualization of that substance could not possibly be the same for a non-Eskimo. Terms for color in different world languages cause speakers of different languages to fail to distinguish blue from green, for example, ignoring the fact that wavelength of light and rods and cones in human eyes are constant, so as humans we see the same independently of language. The language date used to support these hypotheses was incorrect. Whorf for example relied on interpretations of his study of Apache grammar but didn't actually have informants to present him with material he could use to support his claims in much the way that Margaret Mead idealized the society of Samoans whom she visited and subsequently misconstrued to form the basis of her work.

Having shown that the data underpinning the work of Sapir and Whorf was fatally flawed, Pinker starts putting together an image comprised of puzzle pieces from George Orwell's Newspeak, a simple syllogism and a Turing machine. Newspeak was a language devised by Big Brother which would be devoid of words for certain concepts which its speakers would henceforth be unable to think. This notion forms the outer shell of the following reasoning. The syllogism is that Socrates is a human, all humans are mortal, therefore Socrates is mortal. He introduces the Turing machine to show that this machine can be programmed to reach the same conclusion. If language worked like a Turing Machine then it could be programmed to reach conclusions in a predictable manner, and the machine could be made to work in any construct that repeated the same pattern. For example, if your language works right to left, or left to right, or top to bottom, or puts propositions first or last in utterances, the meaning of a syllogism or any utterance can be understood by speakers of your language, as long at the correct pattern is maintained.

However, the human mind does something more than a machine does.  It can infer from context. So when Chomsky tells us that visiting relatives can be fun, we can understand from context who visited whom (Pinker has collected several newspaper headlines whose ambiguities are funny -- e.g. a child's stool is good for the garden -- but which can be unambiguously understood given the context of the news story, or even once we understand that the ambiguous line comes from a newspaper headline ... ah, so that's what it's supposed to mean!). Pinker points out the fact that a given word can have more than one meaning in different contexts is itself evidence that language is something other than coding. Also, there is interesting data from deaf people who have grown up without language, who are highly intelligent, who can function in society, and even mime narratives to one another.

There is a great Radio Lab program on the research Pinker cites, fascinating stuff about the Nicaraguan deaf man Susan Schaller met who had never learned to sign, and a glimpse inside the head of Jill Bolte Taylor after she suffered a stroke that robbed her for some time of language (http://www.radiolab.org/story/91725-words/; and also Charles Fernyhough about the connection between thought, inner speech, and the voice in our heads: http://www.radiolab.org/story/93554-voices-in-your-head/). Pinker also gives examples of great visual thinkers, such as Einstein, for whom ideas were conceived in thought experiments that were only later rendered into mathematical symbols and other forms of spoken and written language. Coming back through to the outer shell, Pinker concludes that Newspeak would never work. The children of its first speakers would creolize it, and Big Brother would find itself with a fifth column on its hands.

What we glean from Pinker and Radio Lab is clear evidence that thought works independently of language, that is is possible to reach conclusions without having to codify them in any symbol system that can be expressed outside the brain in which the conclusions were reached. You can see that yourself the next time you try to write down or explain to someone the narrative revealed to you in dreams. You might be able to capture some of it if you can move the memory quickly enough into some more permanent location in your brain, but personally, I have never been able to work out how I arrived at the point where I can remember myself flying, for example, or what great insights this gave me that evaporated with the light of dawn. Except that after reading Pinker, I realize now that the dream was a glimpse into thought apart from language.



Stephen Downes covers this post in his Daily for Jan 8, 2014, using it to recall something he had written 25 years ago and extrapolating from that recollection through this post to "the basis for both connectionism, as a philosophy of mind, and connectivism, as a philosophy of education".  In his words on that date
(http://www.downes.ca/archive/14/01_08_news_OLDaily.htm):
My career as a published academic began in 1987-1988 with a couple of papers entitled 'Why Equi Fails' and 'Conditional Variability', both of which suggest that meaning is determined from context, and not merely content. That's the lesson drawn, Vance Stevens writes, "when Chomsky tells us that visiting relatives can be fun, we can understand from context who visited whom." What this told me is that thought is subsymbolic. "Thought works independently of language, that is is possible to reach conclusions without having to codify them in any symbol system that can be expressed outside the brain in which the conclusions were reached." The brain is not a computer. It doesn't encode'data' and it doesn't use rules or procedures process that encoded data. That - to me - is the basis for both connectionism, as a philosophy of mind, and connectivism, as a philosophy of education.
Downes, S. Conditional Variability. Calgary Working Papers in Linguistics Number 13 1-13. . November 1, 1988. Authors: Stephen Downes. NRC . A - Publications in Refereed Journals 


No comments: