All language is encoded in fundamental repeating
units to convey meaning. For a spoken language, those units are sounds. This is the International
Phonetic alphabet, or IPA. It’s is a collection of every sound from every known language on
earth. No language could possibly have all of these sounds, or even most of these sounds,
but there’s no sound the human vocal tract can produce consistently that isn’t featured
somewhere on this chart. Every language will have what’s called a sound inventory,
or sound chart that displays exactly what sounds are present, all of which are written
in the standardized IPA notation. For your language, the very first thing you’ll want
to do is make a sound chart of your own. Have a look through the IPA and see which sounds
you like. Very conveniently, Wikipedia provides audio files for most of these sounds, so you
can hear exactly what they’re like before you include them. If you want your language
to sound like any real-world language in particular, you can also look up the language and see
its own chart. Be aware that the symbols that the IPA uses to represent a sound may not
be the same as the one English uses. For example, even though this sound is represented with
the letter j, it actually stands for the sound /y/, whereas this, this, and this are actually
/j/, /ch/, and /sh/. Some sounds are much rarer than others. For example, as English
speakers we’re familiar with the sounds /th/ as in ‘think’, and /dh/, as in ‘this’,
but it turns out these are both pretty rare cross linguistically. The uvular stop /q/,
found in Arabic, Mayan, and some Inuit languages, is also quite rare, and the lateral affricate
/tl/, found in Nahuatl and Navajo, is even rarer than that. If you’re making a fictional
universe featuring many languages, it’s good to be aware of these and keep the relative
cross-linguistic frequencies of these sounds in mind. If you’re having trouble thinking
of what sounds you’d like to include, try coining some sample words that you like the
sound of and listing the sounds that they contain. But it’s not as simple as grabbing
any random sounds that tickle our fancy; we’re dealing with a naturalistic conlang here,
and all natural languages share some features in common, and in terms of phonology, one
very strong tendency is sound symmetry. In the IPA, the columns indicate where in the
mouth the sound is produced, starting at the very front with both the lips, and moving
back towards the glottis. The rows represent the way the sound is produced, whether it’s
a stop, produced by completely obstructing airflow, a fricative, produced by leaving
a small gap for air to hiss or buzz through, an affricate, produced by initially obstructing
airflow in the same manner as a stop before releasing in the same way as a fricative,
a nasal, produced by letting some air pass through the nose, or a liquid, produced by
slightly changing the contour of the unrestricted airway. Any natural language is likely to
have more than one sound in any given row or column. There may be one or two exceptions,
but for the most part, the sounds present in a language should come in series. For example,
we would never expect to see a natural language that features these sounds.
Whereas something like this is much more likely to occur naturally. Now, notice how all of
these spaces on the chart contain two symbols. The ones on the right are voiced, meaning
they require the use of the vocal cords, while the ones on the left are voiceless. All of
these sounds are collectively called obstruents. All natural languages in the world feature
at least some voiceless obstruents, but not all natural languages have voiced obstruents.
If a language has a voiced version of an obstruent, then it will very likely have its voiceless
equivalent. These sounds are called resonants or sonorants. All languages in the world will
feature at least some voiced resonants, but voiceless resonants are exceedingly rare.
Among consonants, resonants include nasals like [m] and [n] and liquids like [r] and
[l]. Any natural language is very likely to include at least one nasal, in fact I have
yet to come across any natural language that doesn’t have a [n] sound, and one liquid.
Vowels are also counted as resonants. Every language in the world features at least two
vowels. Some like Arabic and Yupic inuit have only three vowels, some Native American languages
have four vowels, but by far the most common layout is the classic five vowel system, which
is common in every language family on every continent. Some languages like Turkish have
as many as eight vowels, and some languages go even further than that; as much as we’d
like to think that English has five vowels, it actually has something more like 20. How
many sounds your language has is entirely up to you. Some languages like Hawaiian, Rotokas,
and Mohawk have less than a dozen consonants in their entire phonology, while others like
Ubykh and Taa have over 100. Most languages will have somewhere between 20-30. It’s
always best to err on the side of too few sounds, as a smaller inventory will be easier
to work with and help give your language a more distinct and recognizable phonetic character.
Plus, there’ll be plenty of opportunity to add or remove sounds as we evolve the language
further. From these, we can compile all the sounds in the language into a single chart.
So for our sample language, we’ll start out with the basic sounds: /p/, /t/, and /k/
are extraordinarily common, as are [m] and [n]. For now, we won’t include any voiced
sounds, but that can be remedied later. As for fricatives, /s/ is extremely common, so
we’re safe including that, and the glottal fricative /h/ is fairly common too. But just
for fun, let’s throw in something a little less common like the lateral fricative /hl/.
This is a pretty rare sound, but can be found in welsh, zulu, and Navajo. For affricates,
let’s throw in a /ts/, which is a fairly common affricate and matches the other alveolar
sounds we’ve got. /ch/ is also extremely common, but I’m choosing to omit it for
simplicity, and I’ll need it to help illustrate something later on. Keeping the idea of sound
symmetry in mind, we also might as well throw in the lateral affricate /tl/ to match the
lateral fricative. Again, we’re not using voiced obstruents at the moment, so we won’t
include /z/ or /dz/, or /zhla/ or /dla/. There are many different variations of the
r-like sounds, from /r/ to /r/ to /r/ or even /r/, but the most common is the simple tapped
/r/, so we’ll use that one. L also has a bunch of variations, /l/, /L/, /ly/, but we’ll
stick with the basic /l/. Finally, lets throw in /w/ and /y/, both very common sounds, and
voila, we have our consonant chart. As for vowels, we’ll just stick with the
basic 5-vowel system, since it’s the most common. I’m also going to throw in vowel
length as a contrastive feature. A long vowel (usually) sounds more or less the same as
its short equivalent, just held about twice as long. Having a distinction between long
and short vowels is extremely common across the world; English kind of has long vowels,
but they only appear in specific phonetic environments. Compare the pronunciation of
back vs. bag. However, languages like Latin, Finnish, or Mongolian pay much closer attention
to whether a vowel is long or short. Long vowels can also be very useful later on, so
we’ll include a long version of every one of our five vowels.
So there we have it, here’s our finalized sound chart. It’s pretty basic, but it’ll
serve us well when we make the proto-language. But as you’ll see, there’s going to be
plenty of opportunity for modification later. Now for a quick note on Romanization: you
may have some IPA symbols that aren’t practical to use when transcribing your language, so
you’ll have to choose easier characters to represent these sounds. It is strongly
advisable to choose symbols to represent these sounds that are as easy as possible to interpret.
For our language, let’s represent the /y/ sound with a y rather than a /j/ most english
speakers would mistake that for a /j/. Other than that, The only sounds in our language
that are a bit unwieldy to write in IPA are the lateral fricative and affricate. Most
readers would have no idea what sound to make for these symbols, and it would be a real
pain to have to look up and copy and paste these symbols all the time. Instead, lets
simply represent them with a /hl/ and /tl/. These are what we call digraphs, where a single
sound is represented with multiple characters. Now, these might be misconstrued as consonant
clusters, so someone might be tempted to read /atla/ as /at-la/ instead of /a-tla/, but
as long as we detail this in our romanization system, and have a way of potentially disambiguating
any similar consonant clusters that occur, we should be fine. And then there’s the
issue of representing long vowels. A frequent way to do this is by simply writing the vowel
character twice to indicate that you hold it for twice the length. That makes enough
sense, but unfortunately, English’s weird spelling system means that more often than
not, an English speaker will interpret “ee” not as “e:”, but as “i:” and “oo”
not as “o:”, but as “u:”. There are a couple of ways around this, but I’m going
to take the simple way out and just use vowels with macrons, just as latin does. But we’re
not out of the woods just yet: a languages sound system is more than just a collection
of sounds. Every language has a series of rules that govern how sounds can be put together.
This is part of the reason langauges have recognizable phonetic qualities; we can often
make a pretty good guess at what language a person is speaking even if we don’t speak
it ourselves. So, we need to set some phonotactic guidelines.
The first and most basic rule to set is syllable shape. Any given syllable consists of up to
three segments: the onset, the consonant that begins the syllable, the nucleus, the part
of the syllable capable of receiving stress, which is almost always a vowel, and the coda,
the consonant that ends the syllable, if there is one. Keep in mind, for a consonant to qualify
as a coda, it needs to precede the onset of another syllable or appear at the end of the
word. If a syllable has no coda, it’s called an open syllable, and if it does have a coda,
it’s a closed syllable. About one eighth of the words languages don’t allow codas
at all: Hawaiian, Swahili, and Guarani, consist of only open syllables. Some languages like
Siglitun and many Asian languages allow only a very small number of consonants to serve
as a coda. Most European languages are considered highly
complex in terms of codas, like English’s notoriously dreadful “strengths” and “angsts”.
And some languages can go even further than that: It is possible, if not likely, that
your syllable shape will change as you evolve the language. To illustrate this, and to keep
it simple, I’m going to say that our proto-language is going to consist of purely open syllables.
Another rule you’ll want to establish early on is how the stress system works. In most
langauges, a certain syllable of any given word will receive the stress. Some languages
have a fixed stress system, that is the stress always falls on the same syllable no matter
what. In other languages, changing the stress can change the meaning of the word, compare
the English present and present, and in yet others, which syllable is stressed depends
on the phonetic environment. In a language like latin, the third-to-last syllable is
stressed UNLESS the second to last syllable contains a long vowel or has a coda, in which
case, it receives the stress. Since our sample proto-langauge is purely CV, we don’t have
to worry about coda consonants. Let’s say that the stress always falls on the third-to-last
syllable unless the last-to-last syllable contains a long vowel. And there we have it.
One proto-phonology. We now have everything we need to start coining words. Join me next
time when we’ll start figuring out how to put words together.
