From today's "link salad" at Jay Lake's blog:
Utopian for Beginners — An amateur linguist loses control of the language he invented.
The link (http://www.newyorker.com/reporting/2012 ... ntPage=all) is an article by Joshua Foer concerning Ithkuil, an artificial language that resembles Speedtalk in form and intent. A fascinating story.

That is one of the most interesting things I have read in a long time. I remember hanging out with Loglan (Lojban) fanatics in the erly 80s. Thanks.

Great piece! I find "Gulf" one of the most intriguing things Heinlein ever wrote, both for "Speedtalk" and idea that there is some kind of secret society fixing things up behind the scenes. Like Crowley, I am still waiting to be contacted by the Secret Chiefs. Guys, if you're reading this, please reply...

Preamble: In “Gulf” Heinlein described ‘thought’ as being the same as ‘speech’, and created what he called ‘Speedtalk’ to facilitate the speed of thought. However, on at least two occasions, he contradicts this concept and demonstrates ‘thought’ without any speech. I am not going into that aspect here, but am discussing only the concept that ‘speedtalk’ as described is even possible. The excerpt that I am using here is part of my book on the influences of Korzybski’s General Semantics in the works of Heinlein.
------------------------------------Begin Excerpt --------------------
Before leaving the discussion of “Gulf”, I would like to deviate slightly from purely G[eneral] S[emantics] to comment on some other aspects of the Speedtalk language which however do impact the ‘thought as speech’ usage that Heinlein was describing.
From a Linguistics viewpoint, the compression of ‘thought’ that Heinlein posits is strictly impossible. Leaving aside the question of ‘thinking only in symbols’, Heinlein is unaware or overlooks the fact that the significant phonological units of any language are not the so-called 100-odd phonetic values he mentions, but rather the set of phonemes of a language usually restricted in number to 30 or so. Each language uses a different set. A phoneme is a cluster of similar phonetic sounds called allophones. Allophones differ slightly phonetically from each other and each is used depending on the specific phonetic contexts in which they occur. However, no matter which allophone is used, even in an arbitrary phonetic context, of a given word, it can be used to distinguish that word from others in meaning.[Bloomfield Language, pp. 74-92] Put another way, any allophone could be substituted for any other in the phonemic cluster and it would make no difference in meaning. Such substitution, if noticed at all, would be considered part of a speaker’s regional or personal dialect, his idiolect. But it is these precise types of phonetic differences that Heinlein talks about for use in Speedtalk to differentiate different 'words'.
There are several reasons why such speed compression using this method is impossible, but in essence, the human vocal apparatus and hearing apparatus is limited in how much variation between sounds can be created reliably or correctly heard. In short, it would be impossible for a human, even homo novis, no matter how well trained, to be able to distinguish 800 or so phonemes based on this scheme because virtually all allophonic variation would have to be non-existant and all phonetic redundancy, (about 50% in normal speech) would be lost.
An additional problem would be the fact that since each phoneme would represent a single 'word', putting such words together would probably create many mixtures of vowel and consonants that would be extremely difficult if not impossible to allow correct pronunciation for each as a distinct phoneme, consisting as they would of multiple clusters of consonants or clusters of vowels. Not only are the combinations which would be produced by clusters of consonants, each representing a different phoneme, and hence each cluster representing a different sentence, be almost possible to pronounce, but the transitions between the individual phonemes in a vowel cluster would be totally impossible, the speaker having to literally stop and restart the voicing, (the sound produced by the vocal cords), between each non-contiguous vowel phoneme to prevent intermediate phonemes from intruding themselves into the speech stream.
As to the gain in speed attributed to such a scheme, one other thing needs to be considered. In a ‘Basic’ language consisting of a limited set of ‘words’, it would take ‘more words’ to express some complex thoughts than in a language with a large vocabulary with a ‘single word’ representing the same thought without lengthy circumlocution. An analogy for this comes from computer terminology. Computers work on binary logic, represented as combinations of ‘0’ and ‘1’. The analogy can be easily seen when using octal notation for groupings of 3 binary digits representing a value stored in a computer rather than working directly in binary notation. For example the binary number ‘010 101 111’ can be represented in octal as ‘257’. Both representations equate to the decimal number ‘175’, but it is easier for a computer person to ‘think’ in octal rather than in binary or especially decimal and thus remains closer to the ‘structure’ of the computer.
In other words, it is the power of 'levels of abstraction' that allows us to compress and express verbally our 'thoughts' in shorter form.

Just from a computing perspective your binary example is odd because no one uses 3-bit nibbles. If we write the number in a more standard format, 1010 1111, we can see it is AF hexadecimal, which is a more compact yet equally common representation.

But can it be considered more compactly yet? Because it has the high bit set, it is in the extended ASCII set as the spacing macron over line, hardly an easy concept to communicate verbally. And that's only if you agree to use extended ASCII as the encoding; Unicode would be more logical but a single byte is only valid Unicode if it does not have the high bit set. So reducing it to a single character poses problems of agreement on the encoding.

Why can't programmers tell Halloween from Christmas?

Because 25 OCT = 31 DEC.

Well, it seems that you are trying to disprove Heinlein's hypothesis and I can't say that you have done so. I am somewhat out of my depth here and clearly you have more knowledge of phonetics than I do so I will have to be clear about just which parts of your argument I am taking issue with.

You correctly point out that spoken language requires redundancy through allowing for multiple ways of pronouncing the same word. This permits someone from Boston to understand someone from Atlanta, for instance. But I don't see why Homo Novis shouldn't be capable of understanding very fine degrees of pitch and not require any redundancy in their language (essentially requiring everyone to speak the same dialect and accent).They could have listening skills to put a master chorister to shame: perfect absolute pitch, perfect relative pitch, and so on. They could equally well have the same control over their vocal chords.

Clearly some level of compression is possible. English wasn't designed, it just happened; a language designed with Huffman encoding in mind could easily have higher entropy. Ithkuil appears to go in that direction although it appears not to be designed to be fast but rather unambiguous (as was Loglan, but I guess it failed). Compressing a spoken language goes down a radically different path from compressing a written language, so the whole binary example was probably dangerous to begin with. So the question is, how much compression would be possible for a Homo Novis language?

Speaking of Heinlein (how odd, around here) - there's a profile on Shuttle commander Mark Kelly in the NYT Sunday Review section, where he says flat out he grew up reading Heinlein and Asimov and then moved on to the real thing.

Actually, this week's SR is one of the most interesting, cover to cover, I've read in a year, and more is between the lines than printed. Huh.