The Prehistory of Marpa

Author: Ron Savage
Table of Contents
Introduction
Inadvertent Structured Sound
Deliberate Structured Sound
Human Speech
Reading and Writing
Computer Languages
The Rise of Marpa
Where to from Here
Last updated: 2014-09-18
Introduction
Top

I've been thinking a lot about how we humans lex and parse incoming speech and text, or at least how this relates to Marpa. And I feel it's now time to write something about my imagination's dreaming.

As I intend to somehow cover the geological time spans during which animals developed vocalization and the corresponding auditory sensing organs, it would make sense to begin with fossils, but I have a better idea. Here, 'better' of course means in terms of ease of making a point. To that end, the next section will start with the Artic Fox.

Also, I won't always bother saying 'speech and listening', but shall just use 'speech' or 'language', since I assume (deliberate) speech must have co-evolved with hearing. Likewise, if I say 'writing' I may omit 'and reading', since even pictures on cave walls can be read (by sighted people).

An aside: Instead of 'omit' I was tempted to type 'elide' but I see it's definition covers omitting vowels, consonants and syllables, although not explicitly words or phrases.


Inadvertent Structured Sound
Top

You have I hope seen the fascinating film clips of Artic Foxes (and Polar Bears) trotting across a frozen landscape, and then suddenly jumping up and diving into the ice, digging for a prey animal.

I describe that as the predator lexing and parsing 'Inadvertent Structured Sound'.

It may be that our aquatic ancestors first developed vocalization in order to send pressure waves towards a predator, with the development of auditory sensing organs coming later. Actually, I have no idea what happened in this chicken-and-egg-like series of events, and won't pursue that topic further.

The point here is that the prey does not have to be deliberately transmitting a signal in order to endanger its life. Hence 'inadvertent'.

But inadvertent or not, the predator is acting on the message sent, and no matter what its (the predator's) brain is actually doing, it sure looks like lexing and parsing to me. And above all, such creatures' brains are DNA-based.

And so to our 1st generation:

GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA

Note: In these tables, I won't put an 'etc' at the end of every 'Actor' column entry, but you can imagine one there anyway.


Deliberate Structured Sound
Top

It's now a small step to consider the refinement to these DNA-driven mechanisms to where the transmission is both deliberate and structured.

Since, I assume, animals communicated with their own kind, and also sent warnings to rivals and predators long before humans evolved, this step is called 'deliberate', but does not yet cover human speech.


GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA
2Deliberate Structured SoundBirds/whales singingDNA

Human Speech
Top

I regard the development of speech and writing to be massive achievements in the history of the human race, but, simultaneously, introducing humans into this timeline opens the door slightly to racism, so I'd like to comment on that before preoceeding.

Discussions of human speech can lead to people raving about the FoxP2 gene. For instance some people were desperate to believe that Denisovans and Neanderthals couldn't speak because they lacked this gene. It turns out that they have the same allele as we do.

You can read the fascinating Wikipedia article for details, but here I just want to strongly emphasize that I'm not frightened - as those people appear to be - by the possibility that our evolutionary cousins could produce sounds we might call 'human' speech.

I strongly suspect racism to be the basis of that false assumption, in exactly the same way those sort of people were for decades, if not centuries, adamant that there could have never been interbreeding between the various branches of the hominid evolutionary tree. In fact, nowadays, whenever I see fanaticism, I automatically smell a rat (psychopathology) the size of an elephant. But I digress ...

Collectively then, we humans and our cousins are called Hominidae, and DNA is our master.

If you think some other agent, perhaps a deity, is really our master, consulting anything written by Richard Dawkins ought to, but of course won't, cure of that.

In fact he has a whole book, The God Delusion, dedicated to this topic. Note: Correspondence will not be entered into :-).

DNA builds huge, complex, organisms. It may be true that our brain only weighs 2% of the body even though it consumes 20% of our energy demands, but we don't need to study it all. Rather, let's now just consider a very simplified version of talking and listening:

The DNA has to control the building of a body which:

  1. Has a brain ...
  2. And has ears ...

You can see that, apart from the word 'speech', that process describes what animals have been doing for hundreds of millions of years, so such a list could have been introduced anywhere in this article.

And that neatly brings us to the next generation in the development of DNA-driven lexing and parsing:

GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA
2Deliberate Structured SoundBirds/whales singingDNA
3SpeechHominidaeDNA

Reading and Writing
Top

Idiomatic English demands I write 'Reading and Writing', even though the act of reading what was written would normally have to follow the act of writing. Of course we can extend reading to cover reading of animal tracks, seasonal changes in fruit availability, etc. That naturally, but a bit awkwardly, leads to stretching writing to cover 'inadvertent' writing such as tracks in the dirt and the ripening of fruit!

But why do I single out writing as a separate generation in this process? Because clearly - whether or not it involved more evolution of DNA beyond that required for speech - it is a separate, and complex, DNA-driven mechanism within our brains.

GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA
2Deliberate Structured SoundBirds/whales singingDNA
3SpeechHominidaeDNA
4Reading/WritingHominidaeDNA

Computer Languages
Top

This section could become very complex and unwieldy, but I'll be keeping it brief. Brevity is also an opportunity to circumvent exposing my ignorance of many of the details ...

In order to lex and parse (a program written in) a computer language, we must firstly design one which can be parsed, and then write the software to do the work. But that software must itself be lex and parsed, in order to be run (to do that work). Hence we're immediately faced with a risk of an infinite regression.

You can study how this was handled, in the histories Yacc and Bison, if you wish to investigate more deeply.

I must say in all this, that Lex, Yacc, Flex and Bison are actually the names of software packages.

Here, I've split out Lex/Yacc from Flex/Bison, despite their similarities, because they all have their own, important, place in software history.

You can guess that 'lex' and 'flex' are lexers, and 'yacc' (Yet Another Compiler-compiler) and 'bison' (an Americanism, obviously) are parsers.

GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA
2Deliberate Structured SoundBirds/whales singingDNA
3SpeechHominidaeDNA
4Reading/WritingHominidaeDNA
5Lex and YaccHumansSoftware
6Flex and BisonHumansSoftware

To state obvious, of course humans used DNA-driven brains to design and write the software mentioned, and for that matter, used the same process 10,000 years ago to domesticate wild grasses, and turn them into wheat, rye, barley, oats and rice. But this is an article, not an encyclopedia!

And yes, I realize that these crops could accompany human migration westward (or eastward) from their birthplace in the Middle East across Europe because the path of that trek was parallel to the equator. This means the climate is basically constant across that distance. If you travel 2,000 to 3,000 kilometers south down thru Africa or America, climatic variation will make it vastly more difficult to cultivate the same sorts of crops.

IIRC, I gained an understanding of this from Jared Diamond's book Guns, Germs and Steel, although it might have been another one of his books.

Also, for another view on the timeline surrounding this phase, please jump to Jeffrey Kegler's article Parsing: a timeline, which restricts itself to software matters.

BTW: This level of software sophistication was achieved by about 1970.

After much to-ing and fro-ing about the design of lexers and parsers, to get to this point, the study of such software technology stagnated for decades, apparently under the false belief that major breakthroughs would never (again) happen.

Indeed, the very topic seems to have fallen off tertiary course curricula all over the world.


The Rise of Marpa
Top

The advent though of Marpa, in 2010, changed all that. Or, more precisely, it changed the thinking of a small number of people, who are now its fan club. Broad-based adoption - at least in public - races along like a snail on Valium.

GenerationContextActorAgency
1Inadvertent Structured SoundArtic FoxDNA
2Deliberate Structured SoundBirds/whales singingDNA
3SpeechHominidaeDNA
4Reading/WritingHominidaeDNA
5Lex and YaccHumansSoftware
6Flex and BisonHumansSoftware
7MarpaHumansSoftware

Where to from Here
Top

Well, since this is just my view of the prehistory of Marpa, and its history will only be written in many small stages over many decades, it's time for me to wrap up and and wind down. (Don't you just love a language where 'wrap up' and 'wind down' mean the same thing?)

The links below tell you, in the nicest possible way, where to go (next).


Marpa's homepage
The Marpa Papers
My homepage