Your estimate seems much too high. I just tried singing the alphabet song in English. It took me 8 seconds. Let’s be generous and say 10 seconds. 10 seconds = 26 letters, or 2.6 letters per second.
I’m going to sing four hours a day, this will be my part-time hobby. Until November first, I count 270 days approximately (I’ll give myself about a week off.)
270 (days) × 4 (hours per day) × 3600 (seconds per hour) × 2.6 (letters per second) = 10108800 letters (ten million letters)?!?
According to my counts of the IPA, there are 77 consonants and 27 vowels that have their own symbols. (This is supposed to add up to 107, but it doesn’t, but I couldn’t be bothered to figure out where the mistakes are.)
I have also looked at the diacritics, 27 can be used with consonants, and 8 can be used with vowels. (This doesn’t add up to 52 either, but again too lazy to figure it out.)
I didn’t check to make sure that every diacritic for vowels can be used on every vowel, but I’m going to assume that there are.
So, this means that if you a rough estimate of the number of unique combinations of characters in the IPA, you get something close to 77 × 27 + 27 × 8, which is 2,295.
Now this doesn’t include supersegmentals, which are those things that allow you to tell the difference between insight and incite, and you’re going to need something for dealing with that, that is the ways that syllables interact with each other. There are 20 symbols for this in the IPA, some diacritics, some really not. If it were up to me, in inventing this new alphabet, I would just call them “function letters”, and call it a day. This brings the rough total up to 2,315.
In my opinion, I don’t think that we need this many symbols. The current situation is currently fine. That the Spanish j is x in IPA and in French it’s ʒ is the cause of no particular problems. Within a language, having many letters make the same sound and many letters with several sounds, can. So, there seems no real need to make every word ever printed in every language obsolete. Moreover, while English could use some reform, that’s not going to happen either, although it certainly would have some advantages.
The IPA has 104-107 by various posters estimates. That’s what, 7 bits?
A bunch of vowels are essentially dipthongs. I (“eye”) is the same as ah - ee. For example. So for a lot of vowels, you can trade the specificity of the IPA for having a shorter alphabet with several letters combining to make new sounds.
I think you can just ignore specialized things like clicks.
OTOH I think that using diacritical marks like tildes is cheating and count as an actual separate letter.
This link has a good explanation of how the IPA sounds. For example, it shows “ch” as basically t + sh.
There’s a lot of other consolidation you can do if you try…
One final note: Your alphabet will keep growing. We’re going to discover new languages, and at least some of them will have new phonemes.
We don’t know how many languages there are. We haven’t explored parts of the Amazon, Indonesia, and other regions inhabited by humans, and every group of humans we’ve encountered has at least one language. Moreover, human phonology is extremely variable: English has a somewhat large stock of sounds native speakers consider distinct and there are still many phonemes utterly alien to the language. This means that, some day, we’ll dig up a group of people in Borneo or Brazil or somewhere and their language will have phonemes that you can’t adequately represent with your existing alphabet.
However many letters we decide on here, it’s going to grow with time.
Arnold, you need to decide what your goal really is: Do you want a writing system or a transcription system? (Don’t say “Both.”)The later is as close as you can get already with the IPA, as many have noted above, so why do you want to do it again? Is it simply that you don’t care for some of its symbols?
If you want the former, then why do you care whether there is a one-to-one correspondence to sounds? So that it will be “easier” to learn? Because a truly one-to-one correspondence to sounds isn’t necessary to make a writing systems easier to learn or use for native speakers, and even to the degree that it is helpful, trying to make it serve universally across languages will only defeat the purpose.
Thank you, RadicalPi, for attempting to answer my question. 2315 seems like a high number to me, but it could be close I suppose. I would have estimated much less.
Let’s take the example of English: 5 vowels, 21 consonants (but some of the consonants are redundant, i.e. c is redundant with either s or k as far as pronunciation goes.)
We know that the five vowels don’t represent all vowel sounds, and that the consonants don’t represent all the consonant sounds. So I’ll multiply each count by five as a wild guess.
I would have estimated that around 25 * 5 = 125 different letters would have been sufficient to phonetically write down any word in any language.
Fair enough. I’m interested in how many we can decide on now.
What’s the difference?
Why not?
My “real” question is this (see post 38)
“the IPA was constructed to represent a certain number of sounds that exist in current languages. What is the total number of sounds represented in the IPA?”
Theoretically, the IPA can be used to represent an infinite number of sounds. However, once you start piling up sounds, you will quickly reach some point at which there is no person alive who is able to distinguish by ear or pronunciation every single one of the sounds represented.
Sure, but that’s like saying that with the English alphabet I can construct an infinite number of words. Nonetheless, if someone asks “how many words are in the English language”, it is possible to come up with a rough estimate.
Yes. This is how the IPA does it. There may be a diacritic you can use to indicate that the combination is in fact a diphthong and not two full vowels next to each other, but that would fit into the “function letters” I talked about.
Just for thoroughness, there is an analogue to diphthongs in consonants, the affricate. Again, normally, the symbols are just juxtaposed, but there is a linking diacritic to indicate the affricate in the IPA.
Then what you want is a transcription system, not a writing system per se. The word “alphabet” can mean different things for different purposes.
Take the English (Roman) alphabet. It can’t indicate the different realizations of medial -/t/- in English, but then it doesn’t need to, as far as the needs of a writing system to communicate in English are concerned. Or take Chinese ideographs. One can communicate pretty easily with speakers of very different dialects that are essentially different languages.
As for the total sounds represented in the IPA? Well, isn’t that a pretty fluid thing? People are constantly adjusting or supplementing it for different realizations. Is the dipthong /aɪ/ in English “eye” (as described by jackdavinci) really the “same dipthong” as the first syllable in Spanish “aire”? You’re going to face the same issue with your new-fangled alphabet. It’s not like this is the first time someone has wanted to do this.
No, it’s nothing like it. The IPA is designed to take into account all the known physical manners of articulating sounds as used in human language. Take a look at the system for writing vowel sounds. You are given 18 cardinal vowels, but these symbols can be modified with diacritical marks to indicate an infinite number of possible sounds. Whatever vowel sound it is possible for a human to utter can be recorded in some way using the IPA’s symbols. And to some extent, each speaker uses a unique set of vowels, in that no two people’s speech is absolutely identical. So you have as many vowel systems as you have users of human language, all recordable in IPA.
There is no language that cannot be written in IPA and no known phone that cannot be written in IPA. However, if you try to gather up all known sounds and distinguish them all, you will run into trouble because no single person is capable of distinguishing all known sounds. There’s a limit to human perception such that even though two sounds can be distinguished when written in IPA, those two sounds are not necessarily distinguishable by an individual person.
guizot, acsenray, pretty much you are saying “it is impossible to say exactly how many sounds used in existing languages are represented in the IPA.”
All right.
Could someone come up with any kind of estimate? Or is that impossible too? If one guy says “the answer is two million”, and the next guy says “the answer is three billion”, and a third guy says “no, the answer is much closer to four hundred thousand”, and another guy says “you guys are way off, it’s more like two thousand five hundred”, are you going to say that all those numbers are equally good approximations?
The number is infinite, the same way that the number of possible points on a line is infinite.
Look at this chart of the cardinal vowels: http://www.langsci.ucl.ac.uk/ipa/vowels.html – the chart represents a drawing of the mouth, with the throat on the right and the teeth on the left. Each point on the chart represents tongue position.
The labeled vowels represent only the extremes of the possible vowels. Every conceivable point on this two-dimensional chart is a vowel that can be formed by the human mouth and, therefore, may be used in a human language.
You said “may be used”, I’m asking how many “are used”. Two different things. Unless you mean that on this infinite line, every single one of those sounds is used in a language somewhere.
It depends on what you intend to do with the system.
Each of 5 billion people in the world has a slightly different phonemic inventory. Thus, if you use the system to record the phonemic inventory of every individual, you will get 5 billion different charts.
The question you’re asking is practically unanswerable. There is no way to reduce the number of sounds actually used in human speech to a manageable set of symbols. Any symbolic system you use will require approximation and averaging in order to create anything comprehensible.
Let’s take a real-world example. Merriam-Webster’s online dictionary, in their pronunciation guide for English, has a link to a chart showing their pronunciation symbols. Why do they only list less than 50 symbols, instead of an infinite number of pronunciation symbols?
Those are approximations or averaging or collectivizing the ranges of sounds used in any particular dialect in order to create a comprehensible phonemic system. They do not reflect a cataloging of every last sound actually uttered by speakers of that dialect.
Whatever system you decide to you, you are faced with choices. It’s impractical to create a system that accounts for every possible sound that can be made in human speech. It’s impractical to create a system that accounts for every actual sound that is made by every individual human engaged in human speech. However, if you want to attempt either one, the IPA is flexible enough to do it. As many sounds as you want to account for, there’s a way to use the IPA to do it.
What you are left with is essentially some form of what current writing systems do – establish an approximation of a group of sounds that speakers of a certain language generally understand to be distinguishable in speech. (A minimal pairs analysis may be useful in this regard.)
From an early age, humans are trained to forget, or unlearn, or ignore distinctions between sounds when those distinctions are not necessary for operating within a certain dialect. Once the language curve has reached a certain point, it is difficult, if not impossible, to relearn every single point of possible distinguishment.
You can even try to expand it to more than one language. But very quickly you will come to a point at which you are faced with many pairs of sounds that no single speaker can distinguish them all.
So you are left with the choice that Guizot gave you. You can try to create a transcription system (IPA) that tries to identify every sound specifically and create a one-to-one correspondence between symbols and sounds. But you essentially end up with something that is useful only when the subject at hand is linguistics, because accounting for every possible pronunciation variation becomes a barrier to communication. Or you can try to create a writing system, which does a great deal of grouping and approximation and fudging, but is a lot more useful for communicating ideas.
Sure, but the bottom line is that, with about 50 sounds, you can come up with a pronunciation guide for English that is good enough for a major dictionary. If I’m the man in charge of editing the Merriam-Webster dictionary, and I ask someone to come up with a pronunciation guide, I won’t be interested in the guy rending his garments and wailing “this is impossible! what you ask cannot be done!” I want to hire the guy who rolls up his sleeves and comes up with the 50 symbols to use in my pronunciation guide.
So, for each language in my list, I have a group of can-do people like the people that came up with the Merriam-Webster pronunciation guide. People complaining that it can never be done need not apply.
How many symbols will I need for my pronunciation guide? RadicalPi came up with an estimate of 2315 symbols. Does anyone else care to venture an estimate? Is that even in the ballpark?