How many different vowels are there?

Question

There are generally accepted estimates on how many shades of grey (far less than 50!) or how many colours the human eye can distinguish.

How many different vowels can the human ear distinguish? To qualify as a vowel, the sound must be producible by the human vocal tract, synthesised sounds, sounds of animals, and sounds of musical instruments don't count here. To make the distinction, the ear of a trained phonetician is the right instrument of measurement, lay people will often overhear finer distinctions that are noticed by an expert.

Under these conditions, how many vowel sounds can a human ear distinguish?

Answer 1

Nobody knows, and it's an experiment waiting to happen. The first problem that has to be overcome is creating the stimuli (which is the main non-practical impediment to me doing this). For consistency, the stimuli should be synthesized, thus I disagree with the requirement that the stimuli must be produced by a human. You can synthesize things that can't be produced by the vocal tract, so there needs to be articulatory filtering, and needless to say the product has to be subjectively natural-sounding. (There is a separate question, which Ladefoged wrote his dissertation on, namely how much variation there is in expert phonetician production of the cardinal vowels: a follow-up study would be interesting. There is a lot of variation).

I also disagree with the requirement that trained phoneticians be the only subjects. First, the valid subject pool is quite small, in that trained phoneticians don't receive the level of ear training that they used to a half century ago. Second, limiting the subject pool is invalid given the question, though you could re-phrase the question so that it's not about the human ear (or mind) and make it a question about trained phoneticians. The task is relatively simple: present A, B and answer the question "are these the same?". Since the question is not about phonemic classification, the training has to include A, A (i.e. actual token repetition) as well as A, B (with some difference). Anecdotally, (English-speaking somewhat naive linguistics undergrad) subjects can distinguish absolutely different stimuli.

Anyhow, the other problem is that you have to decide what parameters will remain fixed. Usually when people count vowels, they focus on formant patterns and not F0 differences, source spectrum, amplitude, or duration. As you can imagine, adding all of that on top of formant frequencies would make the project immense. But obviously, these are all (except amplitude) known to be potentially phoneme-defining properties, so not irrelevant to what it means to be "a different vowel".

That is to say, "what do you mean by 'distinguish'?".

Answer 2

There are at least twelve, as there are twelve different vowels in Portuguese.

But then English has a similar number of vowels, which in great part do not overlap with Portuguese vowels. So, just considering those two languages (admittedly two cases of large vocalic repertoire), we would be close to twenty - and neither include the umlaut sounds that are common to French and German.

Now: some English vowels are close enough to Portuguese vowels that if they are used instead of the correct vowels, speech is still intelligible. This means that Portuguese speakers would consider those vowels merely different vocalisations of the same vowel: differente phones but not different phonemes, in other words.

Vowels make up a continuum: if they are close enough, they become indistinguishable. But that "close enough" clause is cultural: there are languages that have only three vowels, /a/, /i/, /u/. To speakers of those languages, /o/ or /e/ are just /u/ and /i/ pronounced with a foreign accent. An Italian or Spanish speaker cannot distinguish the two "e"s of Portuguese; a Portuguese speaker cannot produce the Italian /e/ (wich is an intermediate between the two Portuguese sounds). A language with all those three sounds would probably be difficult, because they would be too close to each others.

So, the answer to your question is probably the following: the number of different vocalic sounds that the human ear can distinguish (in normal conditions, ie, we are not talking about professional musicians or linguists, or poliglots versed in several languages) is variable and determined by the language the human in question speaks.

There is another problem: vowels usually differ in timbre (that is the difference between /a/ and /e/, or /i/ and /u/, for instance). But in some languages duration is phonemic, meaning that /i:/ is a different sound than /i/ (that's how Portuguese speakers will figure the difference between "sheep" and "ship"). So, are long and short "i"s different vowels? Still other languages have phonemic pitch, meaning that a high-pitch "a" contrasts with a low-pitch "a" (or more commonly an ascending "a" contrasts with a descending "a"): are those the same sound, or different ones?

(I have read that Danish has 32 different vowels - including differences in duration -; if this is true, then it must be close to the upper limit of what a language can feature as different vocalic sounds.)