17
$\begingroup$

A species with limited capacity for making different vowel/consonant sounds, pitches, vocalizations, and so on, has a limited capacity for the creation of different words in a compositional language. Even if making use of repetition or sound length in word syllables to account for that limited sound availability, there's only so far that that can go (especially without getting too long or tedious to the point of being less and less effective). If using compositional real-world languages as a model for the verbal language of intelligent species, how many unique sounds (phonemes) would be a rough "minimum" requirement for the effective evolution/development of a full, compositional language?

(This may potentially work better in some sort of linguistic stack exchange--if anyone knows a good place do let me know!)

$\endgroup$
13
  • 22
    $\begingroup$ I think you probably want to ask your question here conlang.stackexchange.com From a worldbuilding perspective you can have as few unique sounds in your language as your capacity to explain, comprehend, or handwave allows. $\endgroup$
    – sphennings
    Apr 4, 2023 at 5:58
  • 7
    $\begingroup$ IIRC the Hawaiian language has 8 consonants and 5 vowels (if we don't count vowel length/diphthongs), to give an upper bound for the answer. Also let's not forget that tones could be phonemic too. I think your problem with a species of limited vocalisation capabilities might well be the speed with which it can control its "voice muscles" and process audio signals. $\endgroup$
    – biziclop
    Apr 4, 2023 at 10:41
  • 7
    $\begingroup$ None, they could just choose non verbal and use sign, body language and posture instead or supplement any limited vocal range with those things, are they a species of paraplegics paralysed from the neck down with no ability to make facial expressions, blink or waggle their ears? then why wouldn't they use those things? 🤗 $\endgroup$
    – Pelinore
    Apr 4, 2023 at 11:37
  • 5
    $\begingroup$ linguistics.stackexchange.com does exist. $\endgroup$
    – RonJohn
    Apr 4, 2023 at 17:29
  • 5
    $\begingroup$ Morse code has only 2 sounds. You could actually do it with only one "sound" since technically silence isn't a sound (i.e., replace the dashes in morse code with rest notes instead). $\endgroup$ Apr 5, 2023 at 13:31

12 Answers 12

26
$\begingroup$

I am not sure if it is the lowest number of sounds documented in a language, but Silbo gomero has a limited set of sounds.

According to different studies, Silbo Gomero has between 2 and 4 vowels and between 4 and 10 consonants. [...] It is a whistled form of a dialect of Canarian Spanish. [...] Silbo has many fewer phonemes than Spanish. This means that communication can be ambiguous at times. Context and word choice are important for effective communication.

Morse code used only two sounds to convey messages, and trained user could understand a message just by listening to the alternating sound of dot and lines being transmitted.

$\endgroup$
10
  • 3
    $\begingroup$ Of course Morse code is a encoding and not a language but there's no reason why a creative worldbuilder couldn't create a language with any number of sounds. $\endgroup$
    – sphennings
    Apr 4, 2023 at 6:16
  • 3
    $\begingroup$ Morse code is an encoding of a language, just as speech is. It suffers from being translated to text and back to sound, but it's no more or less of a language than English or Swahili. Also, it only has a single sound that is "enunciated" for a longer or shorter period of time. $\endgroup$ Apr 4, 2023 at 17:18
  • $\begingroup$ Silbo gomero is amazingly interesting! Thanks for bringing it up! $\endgroup$ Apr 4, 2023 at 18:58
  • 2
    $\begingroup$ @RobertRapplean "Morse code is an encoding of a language" and "[Morse code is] no more or less of a language than English" are not compatible statements. Morse code is not a language, full stop. $\endgroup$ Apr 5, 2023 at 2:24
  • 2
    $\begingroup$ @LieRyan Except that the same encoding can be used for multiple languages. For instance you can use morse code to transmit French. Each symbol of the encoding has the exact same meaning, however the messages will be linguistically distinct. Running English text through rot13 does not suddenly create a new language. Gung jbhyq or evqvphybhf. $\endgroup$
    – sphennings
    Apr 5, 2023 at 4:46
48
$\begingroup$

Everything on the internet is conveyed using bits that have only 2 possible states.

While I am not aware of any group of people that has ever natively spoken a binary language, people do use morse code to communicate.

Per Wikipedia, Ham radio operators can routinely send morse code at up to 60 words per minute.

According to John D Cook, Morse code uses an average of 4.53 symbols per English letter.

Furthermore this Stack Exchange post states that morse code operators can generate dots on the order of 50~60ms in duration.

While it's not the most efficient, a useable language could be constructed with as little as two distinguishable verbal states. This could either be two sounds, or one sound and its absence (like on-off-keying in radio). So, we can have a verbal version of morse code.

From the above, it certainly seems possible that a person could get used to generating/listening to something on the order of 20 short tones per second.

For example, one could have a language made entirely with just an "N" sound and gaps with no sound. To say a word the speaker generates a series of Ns and gaps with uniform timing (like 1/20 second per tone/gap).

To make things efficient, use the same principal as Huffman encoding. The most common words are shorter sequences, and more uncommon words are represented with longer sequences. If the 1000 most common words are on average represented with about 10 tones, then someone would be able to speak like 2 words per second.

Note that if the persons speaking can only make a limited set of sounds it makes sense to encode whole words rather than letters. This can increase efficiency a lot. Using English/morse code we might have 5 letters per word and 4.53 dots per letter (so 22.7 dots per word, as compared to 10 dots per word in the example above).

$\endgroup$
19
  • 23
    $\begingroup$ Now I'm just imagining a race of people whose language sounds like the dial-up noises on old modems... $\endgroup$ Apr 4, 2023 at 15:26
  • 4
    $\begingroup$ @DarrelHoffman If they were descended from bats that or something quite similar wouldn't be such an unlikely sound for a language, ever heard a recording of bat sonar? > skip ahead to 1:19 in the video. $\endgroup$
    – Pelinore
    Apr 4, 2023 at 15:58
  • 5
    $\begingroup$ This is surely the correct answer, yet I suspect that no evolved organism would stay at two. If communication is important enough for speech level communication to evolve then it is also important enough to evolve the ability to produce more sounds. Even 3 or 4 sounds would massively improve communication. $\endgroup$ Apr 4, 2023 at 16:14
  • 1
    $\begingroup$ "Ham radio operators can routinely send morse code at up to 60 words per minute." Which is still half as fast as the rate at which people speak. $\endgroup$
    – RonJohn
    Apr 4, 2023 at 17:31
  • 1
    $\begingroup$ @DarrelHoffman memory-alpha.fandom.com/wiki/Bynar $\endgroup$
    – Michael
    Apr 4, 2023 at 18:00
6
$\begingroup$

Many years ago, a linguist told me a story about a conference which had been set up to design a universal spoken language.

The attendees debated endlessly, gradually eliminating sounds which were unused by or unavailable to members with different cultural or linguistic backgrounds. They eventually reduced the list of candidate sounds to a mere two:

ee aw

So the answer is undoubtedly two, particularly if you are a donkey.

$\endgroup$
10
  • 5
    $\begingroup$ This has the appearance of a satirical response, and may not be up to the community guidelines. Maybe it should be marked as satire, or somehow framed as a humorous response. If, OTOH, it is a serious response, it could be improved by adding references to the conference, and its relationship to other work, such as the development of Esperanto. $\endgroup$
    – cmm
    Apr 5, 2023 at 15:00
  • $\begingroup$ @cmm I merely report it as it was told to me. $\endgroup$ Apr 5, 2023 at 15:51
  • 2
    $\begingroup$ They might have been pulling your leg. Even the Khoisan languages of southern Africa have more than two vowels and some non-click consonants. $\endgroup$
    – Davislor
    Apr 6, 2023 at 3:39
  • $\begingroup$ @Davislor I was told this in the 1980s by an elderly Oxford-educated linguist who was considered to be devoid of any sense of humour. However even if it were in jest, I believe his point is valid in the current context: two distinguishable symbols (obvious nod to Spencer-Brown here) which can be emitted (spoken, written etc.) by the population under question, and can also be interpreted by them. As this point an information theorist would start talking about a minimum of one bit per symbol, and a practical engineer about using a ohased-locked loop to extract timing thus delineate symbols. $\endgroup$ Apr 6, 2023 at 5:57
  • 1
    $\begingroup$ @Davislor I believe the idea is that this is the intersection of the sets of all the different language's phonemes. There are only two that are used by every human language, and so would be pronounceable by speakers of any language. Hence, candidate phonemes for universal language. It's not that any single language has only these two sounds, but these two sounds are found in every single language. $\endgroup$
    – tylisirn
    Apr 6, 2023 at 17:27
4
$\begingroup$

The only examples of languages we have are human ones.

Whilst there are systems like Morse code that use an extremely small number of distinct segments, they are still fundamentally encoding other languages, that use a larger system.

So, with that in mind we should only consider the size of phoneme inventories of natural human languages.

Central Rotokas has the smallest consonant inventory globally, with just 6 distinct consonant phonemes. It has 5 phonemically distinct vowel qualities which can occur in two distinct lengths (although it's unclear whether long vowels are best viewed as a sequence of two identical vowels, or a separate segment).

There are languages with just two phonemic vowels (and Proto-Chadic is sometimes reconstructed as only having had a single phonemic vowel), but languages with such few vowels typically have very large consonant inventories, so Central Rotokas probably either has the lowest or one of the lowest number of distinct phonemes at 11.

Generally, I'd expect the number of distinct sounds necessary for language to develop to be higher than the lowest number of sounds usable in language once developed (cf the existence of Morse code and the fact that no species with just two distinct sounds has developed language), so I'd expect a species that was able to evolve language to be able to produce more than 11 distinct sounds.

Humans can reliably produce one to two hundred distinct sounds or so (if trained to do so since birth as part of their native language), although of course most languages use far less, and so most individuals are unable to reliably produce and distinguish all those sounds.

So that gives us a range of around 11-200 for the number of distinct sounds needed for language to develop.

$\endgroup$
6
  • $\begingroup$ This is an interesting answer with some background, but it has non-sequiturs. While morse code and similar things simply encode other languages, there is no reason it has to be that way. They prove it could be done with two states. You also do not provide any support for the claim that a species needs to be capable of producing more distinct sounds than it actually uses in a developed language. Why would that necessarily be? $\endgroup$ Apr 6, 2023 at 17:18
  • $\begingroup$ @TimothyAWiseman it's observational. Many species are capable of producing at least two distinct sounds, and none other than humans has produced something that can meaningfully be called language (there are key features of human language that have not been observed in extant forms of animal communication, and stories about animals learning sign languages or using sound boards are reported in an extremely misleading way and no such actual learning of human language has been observed). Likewise, all extant human languages use far fewer distinct sounds than we are able to produce $\endgroup$
    – Tristan
    Apr 8, 2023 at 10:24
  • $\begingroup$ seeing as human languages are the sole examples we have of natural languages, it's a reasonable bet that things that hold for all human languages have a better than chance likelihood of holding for non-human natural languages, unless explicable by factors obviously linked to humans themselves (e.g. the fact that all human languages use the tongue to form consonants is clearly an accident of our biology) $\endgroup$
    – Tristan
    Apr 8, 2023 at 10:26
  • $\begingroup$ I am very skeptical of that reasoning. You are basing it on a sample size of 1. It may be true, but there is essentially no evidence for it. $\endgroup$ Apr 10, 2023 at 15:29
  • $\begingroup$ @TimothyAWiseman but that is the sole evidence we have. It may be wrong, but a conclusion based on that evidence is going to be better than one which ignores it $\endgroup$
    – Tristan
    Apr 10, 2023 at 15:30
3
$\begingroup$

This is a subjective question to some extent. Two states is all that is required for a communication system, as has been mentioned by other answers. However, I think one has to consider efficiency in this mix. Here is my thought process:

  • English has roughly 1 million words, according to a quick search.
  • If we assume a fixed word size, like computer binary, then you'd need 20 bits to represent the number of words in the English language. That is, 2^20 = 1,048,576 unique outcomes.

This solution works, but efficiency is lacking. You could have a variable length word size of 2 states in order to represent various words, but structurally this would be hard to follow without pauses long enough to dictate the boundaries, which would introduce another point of inefficiency if your word sizes were too small (1 bit, 2 bit...etc). So, I think the solution with these assumptions is to find the number of sounds.

If we assume 1 million words, then our best bet for efficiency is to reduce the word size by increasing the number of sounds, but at a level that balances with the word size.

This by trial and error is roughly: 8 sounds with a word size of 7, providing 2,097,152 different states that can be represented.

One could gain more states if we do vary the word size, but the benefit seems minimal from a purely state based view: (8^1) + (8^2) + (8^3) + (8^4) + (8^5) + (8^6) + (8^7) = 8 + 64 + 512 + 4096 + 32768 + 262144 + 2097152 = 2,396,744

You could use this to vary the number of sounds, but I still argue that the need for more boundary pauses to account for a word size as small as 1 hurts the efficiency more than it helps.

There may be a more mathematical way of optimizing this problem, but this is my quick back of the napkin thought process. Overall, it is a question of optimization and physical ability to produce and hear the different sounds. Emotional conveyance can still be had with other sounds, or structural emotional words, or through loudness or other signals. There are many ways to convey such information, so you definitely have to think outside the box.

Note, my approach is from that of an engineer. I'm sure a linguist would be more appropriate, but hopefully my approach gives you some thought.

$\endgroup$
3
$\begingroup$

In theory, every message can be encoded as a string of 0's and 1's. So you could just map them to two vowel-only syllables like /o/ and /i/.

However, as another answer has pointed out, a sapient species probably wouldn't want to speak in binary for the same reason that we don't count in binary: It makes everything long and unwieldly. Like the current year having 11 digits (11111100111). Having more phonemes allows the same information to be encoded in fewer syllables. So, let's try to make that trade off.

Suppose a language has a “core” vocabulary on the order of 10 000 words, that we'd like to express in three or fewer syllables. To make this work, we'd need a minimum of 22 distinct syllables.

If we use a rigid consonant + vowel (CV) syllable structure, then we can make this work with 10 distinct phonemes. For example, 5 consonants and 5 vowels, giving 25 distinct syllables.

If we allow the slightly more complex syllable structure CVC, then 6 phonemes is the minimum. For example, you can have the three consonants /p/, /t/, /k/ and the three vowels /a/, /i/, /u/.

FWIW, the most phonetically minimalist language IRL is Rotokas or Pirahã, with 11 phonemes.

$\endgroup$
1
2
$\begingroup$

The length of sound in morse code or in a similar fashion sound / no sound pair is a terrible idea for organics. You may understand morse code by listening but it is hard to generate it without being ambiguous. An older person could talk slowly, or you may want to convey emotion.

I would argue you will need two vowels at least but to make it far better add in two consonants. For effective communication speed pair each vowel with a consonant and you will have 4 different states. You could throw a few more consonants in if you wish but I think with 4 sounds you could create not so boring words that could be spoken fast enough. You could also make it more alien to have a specific sound to end words allowing even faster communication by eliminating pause between words. alelakielelekeki.

Also I will oppose other answers that instead of encoding often used words to be shorter, you should encode critical words shorter. eli for warning, aki for stop, etc..

$\endgroup$
8
  • 4
    $\begingroup$ Your first paragraph is from the perspective of someone who uses a more complex vocalization who is unwilling to believe that a society could develop around just two sounds and work out a perfectly viable language. Please don't fall into the trap of believing that since humanity developed this way, there must be no other way. $\endgroup$
    – JBH
    Apr 4, 2023 at 14:41
  • 4
    $\begingroup$ A person talking slowly or quickly is NOT in itself a problem. What you are looking for is calibration. So long as the culture as a handful of appropriate greetings, then when you meet someone and greet them, they can pick up the speed at which you modulate your sounds from the greeting. Also, from a story perspective, unusually slow/quick speakers, or speakers who vary the pace as they speak, are just the equivalent of people with funny/non-understandable accents in our own world. It's normal not to perfectly understand a foreigner at first. $\endgroup$ Apr 4, 2023 at 14:57
  • 2
    $\begingroup$ This is exactly why Morse code is self-referential with regard to timing. A dash is the length of three dots. Letter separation is one dot-time of silence. Word separation is one dash-time of silence. The consistency of those timings is important for understanding it. The actual speed is limited only by what the practitioners can keep up with. Morse also uses special encodings for line signals ("prosigns"). AR, AS, CP,CQ, SOS, etc. When sending these signals the usual letter break is elided and the whole sequence is sent at once. This makes it distinctive to avoid confusion. $\endgroup$
    – Perkins
    Apr 5, 2023 at 0:33
  • 1
    $\begingroup$ Two points there. First, something that we are good at is comparing time intervals and signal coincidence hence deriving what electronic engineers would call a clock from a signal stream: without that music wouldn't work. Second, I believe that beginners are far better at sending well-paced Morse code than they are at understanding a message sent at the same speed by somebody else. So I'd suggest that the fundamental requirement is that there's sufficient information in the message stream to be able to quantise the encoding: a Morse code message comprising a single dot or dash is meaningless. $\endgroup$ Apr 5, 2023 at 19:31
  • 1
    $\begingroup$ If someone can guess the sender of a Morse code message from hearing it, I don't see why emotion or sarcasm would be impossible for the medium. Extremely difficult to detect, perhaps, but not impossible. en.wikipedia.org/wiki/Keystroke_dynamics $\endgroup$ Apr 5, 2023 at 20:58
2
$\begingroup$

How do you define a "unique sound"? If you really just care about phoneme count, and not how those phonemes are composed, well... other answers have covered that. Plenty of human languages get by with remarkably few phonemes.

But, human language is also pretty exceptional among animal communication systems in that we can produce a wide range of different types of sounds, with different qualities, to make our phonemes. Can you get enough phonemes to be useful with the more restricted vocal abilities that you describe?

Yes, you can! And how do we know? Because whistle registers exist! Real-world humans can effectively communicate at normal speeds in a medium which has no distinctions in sound quality at all, with everything encoded in conventionalized (that is, phonemicized) patterns of volume and limited pitch modulation. So, if your species can produce one single kind of vocalization, as long as they can vary the pitch a bit and do it louder or quieter, and hold it different amounts of time, that is sufficient to put together enough distinct phonemic combinations to encode a complete human-like language.

$\endgroup$
2
$\begingroup$

Don't worry about it.

In a book I have out, there is a galactic community of several hundred alien species. None of them are able to correctly pronounce the phonology of the natural languages of any other race; there's the vocal range of humanity, another race that sounds like a goose honking and coughing, another that sound like a madman playing the flute, and so on. While machine translation is possible, reliably translating between any two of thousands of languages is an enormous burden, even with automation.

So at some point in the past, some linguists working together developed a conlang based on a few dozen syllables. Each species has its own way of pronouncing each syllable, and everyone who deals with other races wears an earpiece (or whatever serves as an earpiece) which transvocalizes the speaker's pronunciation of each spoken syllable into whatever the wearer uses.

With some species that have a limited sound vocabulary, they naturally take a bit longer getting each syllable out, but effective communication remains possible.

$\endgroup$
1
$\begingroup$

Consider a click-language.

Let's say your language is made up of language elements. Words, terms, particles; let's call them "tokens", like the AI researchers do.

And let's say we want to be able to represent roughly as many as written tokens as Japanese: so, a few thousand. We can, as the Japanese do, form new words either phonetically, by using the tokens associated with sounds, or by forming compounds of more complex tokens.

If we have only two possible types of click: click, and not-click, then for a thousand, we need ten clicks in a row.

But we don't need to use all ten. We could omit all click-sequences with more then two non-clicks, and have a space of three or more non-clicks mean "end of token" This loses us a lot of tokens, but is kind of needed so that people coming into the conversation partway, or losing track of the clicks, can re-synch. And that then means we can length-compress, so the most common tokens are shorter, and rarer tokens are longer. This also means we're no longer tied to ten clicks as a max length: we can have them any length we want.

But if we add a few more states, such as four different volumes or pitches of click, or clicks with different pincers, then you only need five clicks per token. With ten click-types, you only need three per token. And you don't then have to use non-clicks, which means you only need to use one clickless beat for a "space", and again you have re-synchonization if you lose track, and can length-compress.

We can also use some of the click sequences as context markers, switching between "token-sets" (like code pages, or alphabets!). That means that like Japanese writing, we can change to a numeric token-set rather than a speech token-set. Or shift the politeness register, or switch to jargon, or slang, or... whatever token-set.

And once you have these context markers, suddenly the language balloons, and word lengths get shorter, as with these multiple token sets, you can have more tokens with lower-numbers of clicks. But it's a balance between that saving, and the cost of the initial context marker. If it saves you one click per character, and costs you three clicks to say, then it's only a saving if you typically say more than three tokens from each token-set after a context marker.

So in theory all you need is click and not-click, but the language rapidly gets a LOT richer once you get more than a single sound.

For Morse code, receiving at 140 WPM (11.7 beeps/sec) is I believe the world record, which is about the 150 WPM of normal spoken speech. So a species evolved to use clicks should be able to communicate at a rate we'd consider normal, especially using some of the tricks above.

$\endgroup$
0
$\begingroup$

Zero.

Well. Depending on your definition of "language".

They don't need to use unique sounds, they could just use the sounds that are already around them.

The lyrebird is a fascinating animal, I highly recommend looking up some videos of them. Attenborough has a good one. It seems to be able to "record" sounds around it and then reproduce them almost perfectly at will, like an MP3 player.

In mating rituals, it reproduces as many sounds as possible to attract a female. They've been observed reproducing camera shutter sounds, toy ray-gun sounds, human speech, and various construction equipment, including hammers and chainsaws.

Suppose a society of humanoids with this remarkable ability of perfect audio recall and reproduction. They might not need to develop any unique sounds, and could instead simply communicate in terms of reference to existing things.

Big angry bear? Make a bear noise to warn people. Fresh vegetables to forage? Crunchy biting noises. I love you? A beating heart.

Think Darmok and Jalad, but with the world's biggest soundboard instead of folklore.

$\endgroup$
0
$\begingroup$

Theoretically, two, but an actual language using just 7 exists.

As numerous other answers correctly point out, you can theoretically create communications capable of expressing anything that any other language can using just two states and one of those sounds can be a meaningful absences of sound (which depending on how you define things, arguably means you need only one "sound").

However, as other answers point out, that is probably not overly efficient for organics.

A real language has been developed which uses only 7 distinct "items". It was designed so that those items can be encoded in multiple different ways including into 7 notes that can be made by instruments or 7 sounds easily made by humans. It is called solresol.

Notably, Solresol is a constructed language rather than a natural language (and has even less of a following than some other constructed languages such as Esperanto, but it does have a following). But its existence proves that an efficient, practical and rich language suitable for speech can be effectively made with 7 distinguishable sounds in the real world.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .