If we wanted to create an all new alphabet composed of as much letters as possible, with each letter corresponding to one distinctive sound. What's the maximum amount of letters we could have?
Oh and please don't answer "an infinite amount".
The notion of 'distinctive' sounds indicates that the discussion must be limited to phoneme inventories found in a single language. To do this we can consider the largest known inventories of contrastive (i.e. which I'm taking 'distinctive' to mean for the purposes of this answer) consonants, vowels and tonal features.
According to the World Atlas of Language Structures online, chapter 1 'Consonant inventories', the language with the largest known set of consonants is !Xóõ (Southern Khoisan), which has 122 consonants. My understanding is that it is not entirely clear that all the !Xóõ clicks contrast across enough environments that we can be fully satisfied with this number, so we could take instead the largest such inventory that does not include clicks; this is usually said to be Ubykh (Abkhazo-Adyghean) with around 80 consonants (not counting the consonants that only occur in onomatopoeia and loanwords).
WALS takes a simple approach to counting vowel inventories, only counting the vowel qualities rather than all distinctive vocalic segments. Thus they only consider the three features of height, backness and lip rounding, and ignore length contrasts, nasalisation, etc. With this approach they find the largest inventories to be around 14 vowels (German, Indo-European). If we include diphthongs and triphthongs then we have, e.g., Fering/North Frisian (Indo-European) with 22 vowels. Of course, if we move even further from the WALS approach and include secondary articulations such as nasalization, phonation, tongue root retraction, pharyngealization, etc then we could probably find much larger inventories.
There are a few languages thought to have up to 9 tones, e.g. the Kam language (Tai-Kadai)
Largest attested segmental inventories
Combining the WALS maps we can identify a small group of languages that have large inventories of both C and V. !Xóõ is the largest with Ubykh following, but several others have total C+V inventories approaching 60 or so segments:
Largest theoretically possible inventory
If we take the more conservative numbers of 80 C, 14 V and ignore tone (potentially 1,120 CV syllables) then we require the orthography to represent only(!) 94 phonemes.
If however we combine the maximum consonant, vowel and tonal inventories then we could theoretically have a language with up to 122 C, 22 V and 9 T. This would require an orthography with 144 segmental symbols and some kind of tone-marking as well. If tone marking used segmental symbols (as is common in the Tai-Kadai and Hmong-Mien families) then a further 9 symbols could be added bringing the total to 153. It's worth noting that such a language would theoretically have over 24,000 possible CV syllables!
I do not know for an average person, but I can describe my own situation.
Let's count how many phonemes I can recognize and produce. My native language is Russian, so my phonetics is based on that language with some additions. For sounds existing in Russian I will use the transcription system taught in Russian scools, based on Cyrillic, because IPA is not quite suitable for the case.
Most consonants (with exceptions I will mention) come in 4 variants for me: the voiced hard, voiced soft, voiceless hard, voiceless soft. Some of these variants do not exist in Russian language, yet they are readily distinguishable to me and I can produce them. Some quadruples form pairs between each other, corresponding to plosive and fricative variants so to produce octets.
The vowels come in 2 variants for me: the variant coming after a soft consonants and the variant coming after a hard consonant. I would call them "soft" and "hard" vowels, although the Russian phonology does not use this distinction. I think, international terminology is "front" and "back" vowels. For me the distinction is like between German u and ü and o and ö. For me soft vowel comes always after a "soft" consonant and a "hard" vowel always comes after a hard consonant (that's why they are not considered distinct phonemes in Russian phonology), yet for me either "soft" or "hard" vowel can be at the beginning of a word, a distinction not existing in Russian, but existing in German.
[б], [п], [б'], [п'] - four variants
[в], [ф], [в'], [ф'] - four variants, forms plosive/fricative octet with the prevuous
[д], [т], [д'], [т'] - four variants
[ð], [ѳ], [ð'], [ѳ'] - [ð] represents English "th" in "this", [ѳ] represents
the "th" in "thick". The soft-hard distinction is not
phonemic in any language known to me, yet I can easily
produce and recognize it.
Forms a plosive/fricative octet with the prevuous.
[з], [c], [з'], [c'] - four variants
[ӡ], [ц], [ӡ'], [ц'] - Russian has only hard voiceless variant and it is difficult
to me to recognize the soft variants because they mix for
me with [д'] and [т'].
Forms affricate/non-affricate octet with the prevuous.
[ж], [ш], [ж'], [ш'] - Russian does not have the soft variant [ж']
yet I can easily recognize and produce it.
[ӂ], [ч], [ӂ'], [ч'] - Russian has only soft voiceless variant.
Forms affricate/non-affricate octet with the prevuous.
[г], [к], [г'], [к'] - four variants
[ғ], [х], [й'], [х'] - for me the hard variant of й is equal to voiced velar
fricative, which does not exist in Russian.
Forms plosive/fricative octet with the prevuous.
[р], [р'] - this has no voiceless variant for me
[л], [л'] - this has no voiceless variant for me,
groupped with the prevuous as glides.
[м], [м'] - this has no voiceless variant for me
[н], [н'] - this has no voiceless variant for me,
groupped with the prevuous as nasals.
[ŋ], [ŋ'] - velar nasal not existing in Russian
So this constitutes 50 consonants.
I also recognize the following vowels (five basic vowels in back and front variants to form 10 total):
[а], ['а] - in Russian only the first variant can start a word
[о], ['о] - in Russian only the first variant can start a word
[у], ['у] - in Russian only the first variant can start a word
[э], ['э] - in Russian only the second variant can start a word
[и], ['и] - in Russian only the second variant can start a word,
except some foreign placenames starting with Ы letter.
This gives 10 vowels, yet the "softness" of a vowel is totally determined by the preceding consonant, except at the word's beginning.
Also each vowel can be stressed or not, but no more than one vowel per word.
Of course there are other phonemes in different languages, but I have just described what I myself can recognize and produce.
The question asks about "possible", not "probable" -- "What's the maximum amount of letters we could have". Greg Lee's answer is the closest to being correct. SPE does not actually give a final and definitive set of features (they consider it an open question), and even pre-geometric research indicated that their count of features was low (for example, the lack of [labial], [constricted glottis], [spread glottis], myriad tone features). If you up the feature count to 20, the answer would be 1,048,576. These calculations include various "unattestable" combinations, such as the class of [+hi,+lo] segments, which are physically unrealizable but computationally well-defined. If your standard for "can" is not just grammatical distinctness but also physical distinctness, the number would be smaller by some fraction, and nobody was worked out that function (there are numerous little unrealizable distinctions such as pharyngeal nasal stops). The notion of "possible" is meaningful only in the context of some theory that allows a given situation.
We might suppose that the largest number of actually observed segments is related in some fashion to "possible" segments, but how does one relate "actually observed" to "possible"? You might think that a language with 300 segments is "not possible" since this is quite a bit above the largest known inventory, which would be on the order of 140 segments. But that doesn't mean that a language with 300 phonemes is impossible, it simply means that it isn't very likely to exist. Possible, but unlikely. You could, theoretically, make a stronger claim about that probability, if you had some idea of the statistical distribution of phoneme inventory sizes.
So there is a fundamental problem with the question, that the notion of "could be" intended here is quite unclear.
131,072 is the number of distinguishable sound segments in a system of binary distinctive features with 17 different features. My informal count of the distinctive features in the Sound Pattern of English feature theory is 17.
If you want to derive a writing system that can distinguish 131,072 distinctive sounds, I'd suggest the use of diacritics (lots of them).
This depends on the age of the speakers. At birth, humans have the ability to acquire any sound that they are physically able of producing. If the language they are surrounded by has a certain sound, they will most likely acquire it, and will later be able to produce it. As this acquisition process happens, infants gradually lose the ability to acquire new sounds that are not present in the language they are acquiring. The time this takes is often referred to as the critical period.
As for your question of how many sounds a new language can have so that the average adult can speak the language, this crucially depends on what language(s) they already speak. If they know only one language (say, language A), and that language has 28 different sounds, your constructed language can only have those 28 sounds (if you want speakers of language A to be able to acquire the sound system of your new language perfectly). Of course, there are speakers of other languages, say language B with 18 sounds. If your constructed language is tailored to the needs of speakers of language A, speakers of language B cannot acquire it perfectly.
Another complicating factor is the difference between phonemes and phones. Basically, both can be called sounds, but phonemes are a way of grouping different sounds together to a single meaningful unit. But if two sounds A and B belong to a single phoneme p in a language, another language might have sounds A and B as separate phonemes j and k.
I have spent a long time pondering upon this question and according to a former U.S. Air force pilot that I had at the University of Dayton as a professor, a human being can only make 40 distinguishable sounds which are sufficiently different from each other to produce a language that can be understood by the person trying to send the information, and the individual receiving the message. This is very important and why the military uses words like alpha, bravo, and charlie rather than a, b, or c when communicating over the radio. Given the fact that a growl, or a whistle can be considered communicating phonemic sounds (just watch the movies: "Enemy Mine", and "Guardians of the Galaxy" for samples of growling and whistling) when you think about it, we really can produce thousands of tones for a single sound, and thousands of different sounds in scale just like a soprano singer can. But how many of those soprano sounds can be distinguished by an untrained ear?
Therefore the different consonants as well as the different vowels allow us to separate in our brain processing areas, different "breaks" in the flow of words. That is why a K has to be sufficiently different from an L, so our brains can distinguish the words. Imagine a language made only of K's, Q's, and C's. How can you distinguish words in such a language over a radio transmission? How difficult would it be for us to understand each other if everything sounded like cancan, King Kong, or quacking! The human brain does need to identify the differences in order to process the information. And that is precisely the problem; that there is a limit for our neural pathways (such as the ones in Broca's and Wernicke's areas of the brain) to process information.
These aforementioned breaks in spoken communication are called by linguists features of a language and by scientists acoustic signatures. Specifically plosive consonants like p, t, k, b, d and fricative consonants like s, z, and v provide the brain with different enough variations to allow for precise communication. If you are really interested in this follow the work of Dr. Edward F. Chang in how the brain recognizes speech sounds, and an odd language created in the 1950's known as Unifon, consisting of individual phonemic sounds such as the ones in a dictionary.
The real trick behind all this is finding out exactly what those 40 essential sounds are. There is a letter in Hebrew that sounds like you are clearing your throat. Chet. This sound was lost from the English language but it is still used by many Middle Eastern cultures. Did you know that the Egyptians had a hieroglyph drawing of a little chicken which sounded like a W or a diphthong of a Oo (Spanish U) with another vowel! So is a W a single phonemic sound or two? Is it Wal-Mart, or Ooal-Mart?
We have to ask ourselves then, if the original sounds that allowed humanity to communicate in the beginning of our history were enough to be differentiated not only as sounds, but also as symbols? Can you imagine this two symbols: ///////\, O-////// one for water and the other meaning snake, being confused!!! Many archaeologist are pointing to the languages known as Proto-Sinaitic / Proto-Canaanite because several of the written symbols that we use today such as the "A" can be traced to the pictoglyphs Canaanites used; in this case of an ox. As a matter of fact even the name of the letter was maintained. Alp meant ox to the Canaanites. Greeks still call it Alpha. Hebrews call it Aleph. We call it an A. So not only how many sounds can we distinguish is important, but also where do they come from, and what did they mean for the original cultures that used them! Can we simply remove them like we removed Chet? Can we simply substitute the Greek letter Phi for an F? They sound very similar in Phylum and Fielding, don't they? Are the C and K the same acoustic signatures in Car and Karma? If so, why do we have two symbols for the same acoustic signature? How about the V in Volkswagen and the F in Folklore? Have we reached as a species the point where we need to take a more scientific approach to languages rather than a traditional or cultural approach? Can we substitute symbols and sounds for binary codes and tri-dimensional colored shapes? Will this finally solve the problem of communication for blind people, or primitive cultures in Africa, and India without written symbols. Or are we creating in the attempt, an even bigger communication nightmare?
How do we incorporate the different Oriental and Hindu dialects into just 40 sounds and symbols, when sometimes a single symbol represents an entire word, and they have thousands of such symbols? Did you know that the word Lung in some Chinese dialects, means Dragon? One symbol can be an entire word, and may represent or offend an entire culture!!! Let's take for example the symbols of the cross, and the Swastika. What happens when we tell the Jewish people that they have to use this symbols in a New World Language, because the cross and the Swastika were used by the most primitive cultures in the world for millennia, and we are trying to incorporate as many cultures as possible in this new "common" language... Will they accept this? In the words of the immortal Star Trek character Mr. Spock: "Fascinating..."
Did you notice that by the way? The expression "fascinating" is only one word but, the "idiomatic phrase" is: "In the words...", plural. So that's another question, how do we incorporate idiomatic phrases, regionalisms, and other technicalities from one language into another where the same words do not exist? Eskimos have hundreds of words to describe the different types of snow in Alaska. How many words for snow do you think they have in Puerto Rico!!!
"It will be very difficult for humanity to change the many different symbols and sounds that have defined us throughout the centuries, but at some point in our history we will have to do so in order to progress..."
Prof. Ramon R. Malave-Marty, [email protected] 09/02/2015
From several different sites, sources not verified: The number of separate BASIC sounds in a language is often known. I don't know of any languages that have much more than 50 basic sounds or so. All English sounds are made up of just 44-46 sounds, Japanese also has around 44 or so basic sounds, depending on how you define "basic".
The average human can make over 500 distinct sounds of vowels and consonants. If you include variations on pitch and volume the number is infinite.
So, non-tonal would be on the order of 500 letters. With tonal (consider raising half a step or a full step being considered differently) there really could be an infinite number of letters.