Natural languages begin as conversations. In that sense at least, language is the original interface.
Question is, why is it not the preferred interface for humans and computers?
<<<refactoring slowly -lp>>>
Languages Co Evolved with Human Beings and their brains, growing from crude signals into sophisticated Symbolic Systems under a variety of selective pressures. This is pure speculation. An alternative theory is that the human brain had essentially fully evolved to its modern state prior to extensive human language, and non-linguistic capabilities were subsequently exapted for full language use. This theory probably contends that human culture and language coevolved.
Some of the significant selective pressures exerted on languages may have been:
learnability:
languages more easily learned spread more quickly This is not likely. Language spread is likely to have been largely determined by non-linguistic factors, such as relative power or technological superiority.
usability:
languages that allow more effective communication provide advantages to their hosts
teachability:
languages that enable efficient transfer of knowledge provide advantages for their hosts
[It seems to me that usability and teachability are the same. I cannot clearly find a dividing line between communication and transfer of knowledge. -- David Saff]
The difference between usability and teachability is the difference between telling someone else where the fish are, and instructing someone who's never fished before. A language of high usability can have relatively low teachability {Forth Language would be an example} as far as programming languages are concerned teachability is strongly correlated to maintainability.
At the same time the development of languages placed selective pressure on their hosts, a topic more fully explored in The Symbolic Species.
In terms familiar to Wiki Zens all languages are refactored by each new generation of users. Aspects of a language that are difficult to learn fade out from disuse. Areas of Tolerance For Ambiguity grow to the limits of usefullness. Programming languages are evolving under a similar set of constraints. And can be seen as a subset of natural languages that face the additional selective pressures of correctness, maintainability and marketshare.
[On the contrary, the evidence seems to suggest that native speakers find no aspects of a language "difficult" to learn. Rather, there are "early" and "late" acquisitions. Take irregular verbs, for example. Children usually learn the main irregular forms before they acquire, by generalization, the rules underlying regular forms. They then misapply the derived rules and re-learn the irregular forms by exception. This seems to re-inforce the learning, so irregular forms are resistant to "evolutionary erosion". See Lance Walton's quote from Locke below.]
OK, then why does English have such weird spelling? It is hard to learn, and should have faded. English grammar is odd, too. People say that English is one of the hardest languages to learn as a second language. Perl is like English, with lots of weird things in it, but useful and popular.
[Because English, like Perl, is based on dozens of sources. This gives it its strength, but also the obvious weakness. It has more concepts than most other languages. A fairly moot point, considering most 'fluent' speakers know far less than 10% of the language (possibly less than 1%). A cwm is a fairly complete description of one (uncommon) type of valley. It is, if I recall correctly, from Welsh. Most languages would require many words to describe the same concept. Although English also has two synonyms for cwm (Welsh): corrie (from Gaelic) and cirque (from French) -- Ed Grimm]
[I must point out that English has no more concepts than any language. An English speaker legendarily asked of a Gaelic speaker, what's Gaelic for spaghetti? But I digress. I would suspect that English is very hard to get 'right' as a second language, but very easy to make yourself understood in. The grammar is nonexistent in comparison to, say, Czech (noun cases? aaagh), or even French (noun genders affect several words in each and every sentence). I've heard people similarly describe Chinese as a hard language when in fact it (Mandarin at least) is one of the most straightforward. It's only when you try to write it down - as with English - that the cracks show. -- Brian Ewins]
Wasn't it one of Ronald Reagan's staff who once declared that "The Russians have no word for détente"? As observed by Ed Grimm, English is a mongrel. Funny thing is, when times get tough, it's the mongrels that tend to survive.
I have heard it said quite often that English has more words than other languages, at least in its immediate group ("more than French and German put together" seems to crop up, although how you would test such an assertion in the real world is kind of beyond me). Since Brian is certainly right, I've often wondered where they go. I suspect a lot of them are redundant, words for the same thing but from different lineages (as is a lot of perl's vocabulary, for that matter).
Perl has very high usability in that its possible to express different ideas in different ways. It was intentionally designed to allow mixing different types of code. It's also descended from the Algol Family of languages which have proven high learnability. Perl is noted for its relatively low teachability though.
Eric Raymond recently wrote this about Perl's "user interface":
[problems with Perl's syntax] combined to make large volumes of Perl code seem unreasonably difficult to read and grasp as a whole after only a few days' absence. Also, I found I was spending more and more time wrestling with artifacts of the language rather than my application problems. And, most damning of all, the resulting code was ugly--this matters. Ugly programs are like ugly suspension bridges: they're much more liable to collapse than pretty ones, because the way humans (especially engineer-humans) perceive beauty is intimately related to our ability to process and understand complexity. A language that makes it hard to write elegant code makes it hard to write good code.
The spelling of English words has only become weird since the Great Vowel Shift which I believe happened in about the 15th century. For an excellent discourse on this subject, including theories on why some languages survive in spite of apparently obscure constructs, see Words and Rules: The Ingredients of Language by Steven Pinker ISBN: 0060958405. -- Lance Walton
This may explain some phenomena that have been puzzled over on wiki many times before. My sense is that Smalltalk Language has high teachability and usability but relatively low learnability. That is to say that smalltalk as a language is easier to teach than to learn.
Compare this to C or even Java Language. It is relatively easy for someone to produce functioning programs with C when they learned it out of a book, or picked it up from reading Source Code, whereas smalltalk requires a somewhat higher level of commitment to learn on one's own and relatively few smalltalkers are self-taught, as compared to other languages. --Larry Price
I do not think that Smalltalk is harder to learn than Java. It depends on your background. Suppose the only language you know is COBOL, Lisp, or Pascal. Then Smalltalk is easier to learn than C++, and probably no harder than Java. But suppose you know C. Then C++ and Java seem easier because you are used to the syntax. So, if you know C then you can write a little program quickly in Java, but it takes longer in Smalltalk. On the other hand, even if you know C then it will probably take less time before you are able to write a new widget in Smalltalk than to be able to write one in Java. That is because the better programming environment and the simpler class libraries make it easier.
So smalltalk is more usable than java, at least in producing widgets.
Another advantage in learnability for Java is that there are a lot more books aimed at beginners. -Ralph Johnson
Is that a cause or a consequence.? It seems as though the more learnable a language is, the easier it is to produce an introductory book for it. Also there's a certain amount of positive feedback between popularity and availability. But I was aiming at the intrinsic features of the language.
I find the definitions at the head of the page interesting... Something I miss, though, is adaptability... One language could be very good at explaining the environment for a hunter-gatherer but very poor for a neolithical agricultor... They have very different space-time concepts... As we changed/change from a savage thinking to domesticated thinking (L�vi-Strauss terms, I'm not sure I've used the right English terms), their communication needs changed/change as well... Like a nomadic tribesman might measure his age in months, since the changing phase of the moon is the most obvious long-period phenonemon; a crop farmer might use years for the same purpose, because seasons are so much more important to him. The fun begins when someone's age is recorded in months, but the number is later interpreted as counting years...
It's interesting to note that the more we investigate, the bigger our need to go to old, outfashioned languages to describe our new constructs (Latin, Greek, Sanskrit...). It's like if our thinking way was changing again and we needed to return to concepts we have disposed of long time ago because they are useful again and our languages cannot cope with the high amount of detail that symbol must hold...
I mean, why telephone and not long distance speaker? (I don't believe it's just a matter of laziness... laser is laziness, but "light amplification by stimulated emission of radiation" was described in english and not greek or latin or... Interesting phenomena, the evolution of languages...
Some of this is the domain of those creating words - when it is a technical person, they will use technical resources - thus the telephone and telegraph from men taught in Latin and Greek as the epitome of culture, but ripping for translation from CD audio format to MP3 format, invented by youngsters familiar with music slang. Also consider upload and download, short and descriptive (at the time, even more so).
Anyway, I think it's not that easy. Our primary interface is our hands. As babies, we touch, and hear, then we see, then we listen and finally we tie all together and create a language. There are many cognitive problems related to blind people, with have a haptic space, or with deaf-born people, which won't develop a language as we understand it (speaking language or lingua) Thus, we develop a semiotic understanding of the world and then we develop a linguistic interface to it, but body language is still our primary means of communication (even in the Internet we provide body language with emoticons and context. It would be interesting to develop this further and there probably are people working in this, so if anybody knows of some links, please don't be shy :-) -- David De Lis
This reminded me of one of the many beautiful passages in John Locke's 'An Essay Concerning Human Understanding': ' The senses at first let in particular ideas, and furnish the yet empty cabinet, and the mind by degrees growing familiar with some of them, they are lodged in the memory, and names got to them. Afterwards, the mind proceeding further, abstracts them, and by degrees learns the use of general names. In this manner the mind comes to be furnished with ideas and language, the materials about which to exercise its discursive faculty. ' -- Lance Walton
In Japanese, kanji compounds tend to use the Chinese-derived readings rather than the native Japanese. This seems somehow related to the `telephone' vs `farspeak' example.
ObSF: en.wikipedia.org 
See original on c2.com