Sound changes
The notation of a rule, and how to read a change that did nothing.
A rule carries every word that fits it from one form to the next. To add one, point at a row of the matrix, press +, choose Sound change, type the rule and press Enter. The field refuses a rule it can't read and says why, and the IPA keyboard beside it types ∅.
The form of a rule
source > target / left_right // exceptionThe environment after / and the exception after // are optional, and _ stands for the source in both.
| Rule | Word | Result |
|---|---|---|
t > d | tata | dada |
t > d / V_V | pata | pada |
p > b / #_ | pap | bap |
p > b / _# | pap | pab |
t > ∅ / _# | pat | pa |
∅ > ə / C_C | akta | akəta |
s > h / V_V // _i | asa | aha |
s > h / V_V // _i | asi | asi |
a > e / _ // _i | pat | pet |
# is the edge of a word and ∅ is nothing, so ∅ deletes as a target and inserts as a source. Give an insertion an environment: ∅ > ə alone turns ta into ətəaə. An exception needs an environment before it: / _ means anywhere.
Each rule reads the word as it stood before that rule, so a > b / a_ turns aaaa into abbb. Rules and changes run from top to bottom.
Groups of sounds
Braces hold options, separated by commas. Options on both sides pair up in order, so the two sides need the same number of them, or one target for them all.
| Rule | Word | Result |
|---|---|---|
{p, t, k} > {b, d, ɡ} | pataka | badaɡa |
{ɛ, ɔ, e, o} > ɐ | tome | tɐmɐ |
{p, t} > b | pat | bab |
a > e / _{p, t} | at | et |
ɛɔeo > ɐ | hɛɔeom | hɐm |
ɛɔeo > ɐ | hɛm | hɛm |
A space separates nothing: p t k is the three sounds in a row, the same as ptk, and {pt, k} is pt or k. The field refuses a source written with spaces, such as ɛ ɔ e o > ɐ, and gives both readings:
Spaces don't separate sounds. Write {ɛ, ɔ, e, o} for any of them, or ɛɔeo for all in a row.
Classes
C is any consonant and V any vowel. @ names the other classes.
| Rule | Word | Result |
|---|---|---|
V > a / _# | pu | pa |
C > ∅ / _# | pat | pa |
@nasal > m / _p | npa | mpa |
The classes are @vowel, @consonant, @stop, @plosive, @nasal, @trill, @tap, @fricative, @lateral-fricative, @approximant, @lateral, @affricate, @click, @implosive, @voiced, @voiceless, @sonorant, @front, @back, @high-front, @high-back, @non-front and @non-back.
Every other capital letter is a sound. S > Z / V_V looks for the letter S, and N > m / _p for the letter N, whatever N means in a syllable template. A letter outside the IPA has no features, so C and V don't match it: C > x leaves Sa as it is.
Features
Square brackets pick sounds by feature. In a target they change those features and keep the rest. Right after a sound, they change that sound.
| Rule | Word | Result |
|---|---|---|
[-voice] > [+voice] / V_V | apa | aba |
[velar] > [palatal] / _i | ki | ci |
[stop, +voice] > [-voice] / _# | ad | at |
V > [+long] / _CV | pata | paːta |
t > p[+voice] | ta | ba |
V[-stress] > ə | ˈpata | ˈpatə |
[stop] > [fricative, +continuant] / V_V | aka | axa |
[+voice] covers vowels and sonorants as well: [+voice] > [-voice] / _# turns ta into tḁ, so narrow it as in the table.
A change has to end in a sound the IPA has. The field refuses [stop] > [fricative] / V_V with Every [fricative] sound is also +continuant. and gives the rewrite, [stop] > [fricative, +continuant] / V_V. Where no added feature gives a sound, as in a > ɓ[-voice], it says that no IPA sound has those features.
Feature names
| How to write it | Names |
|---|---|
[+x] or [-x] | voice, nasal, lateral, continuant, sonorant, syllabic, long, back, round, high, stress |
[+x] | aspirated, labialized, palatalized, velarized, pharyngealized, rhotic |
[x], place | bilabial, labiodental, dental, alveolar, palatoalveolar, postalveolar, alveolopalatal, retroflex, palatal, velar, uvular, pharyngeal, epiglottal, glottal, labiovelar, labiopalatal, postalveolar-velar |
[x], manner | stop, nasal, trill, tap, fricative, approximant, click, affricate |
[x], height | close, near-close, close-mid, mid, open-mid, near-open, open |
[x], other | consonant, vowel, central, pulmonic, velaric, implosive |
Commas join names: [stop, +voice]. stress belongs to a vowel: a vowel after a stress mark is +stress, and every other vowel -stress.
A plain sound has no -aspirated, so the field refuses [+aspirated] > [-aspirated] with No sound is −aspirated. To take that mark off, write the sound: pʰ > p. Marks for tone, phonation, release and ejectives have no name a rule can use, and the same goes for them: pʼ > b.
place, manner and height head the rows above and work with α alone. The field refuses one written by itself, as in [place] > x, and names features of the group to write in its place. It does the same with segment, which groups consonant and vowel, and airstream, which groups pulmonic, velaric and implosive. tone, phonation and release are refused too. A saved rule with one of these still opens and changes no word, and the list of idle changes says why. Other names are refused, with Unknown phonological feature "voiced".
| Refused | Write |
|---|---|
[+voiced] | [+voice], or @voiced for consonants alone |
[plosive] | [stop], or @plosive without implosives |
[+front] | [-back] or @front |
[+velar] | [velar] |
Boundaries and marks
. in an environment matches only a . written in the word, since rules don't work out syllables. For a sound before one consonant and a vowel, write C_CV.
| Rule | Word | Result |
|---|---|---|
o > oː / C_.C | to.pa | toː.pa |
o > oː / C_.C | topa | topa |
o > oː / C_CV | topa | toːpa |
o > oː / C_CV | tosta | tosta |
t > d / V_V | pa.ta | pa.da |
t > d / V_V | paˈta | paˈda |
o > u / _+ | to+na | tu+na |
A rule looks past the stress marks and boundaries it doesn't name. + is a morpheme boundary, which seam rules use.
Optional sounds and captures
(C) may be left out. C* is zero or more, and belongs in an environment. C$1 names a sound so the target can move or copy it, and C₁ means the same. [αplace] makes two sounds agree. \ reads the next character as a sound.
On a plain sound a digit belongs to the sound. h₂ is one consonant that C matches, and h, h₁ and h₂ are three sounds, so h₂ > x leaves the other two as they are. A plain sound takes $1 to name it, as in a$1. The IPA keyboard has h₁ h₂ h₃ under Other symbols.
| Rule | Word | Result |
|---|---|---|
o > oː / _C(C)V | tosta | toːsta |
a > e / _C*i | apki | epki |
C$1 V$2 > V$2 C$1 | pa | ap |
C$1 C$1 > $1 | appa | apa |
h₂ > x | h₁ah₂eh₃ | h₁axeh₃ |
a$1 > $1$1 | ta | taa |
n > [αplace] / _C[αplace] | anka | aŋka |
\C > x | Ca | xa |
a > e / _(C)C# | at | et |
a > e / _(C)C# | atk | etk |
(C)x > y | ax | ay |
(C)x > y | tx | y |
(C) matches with or without its sound, whichever lets the rest of the rule fit, so a > e / _(C)C# finds one or two final consonants. Where both fit, the longer match wins: (C)x > y replaces both sounds of tx. Braces take the first option that lets the rest fit.
A class in the target
A class in the target is the sound the same class matched in the source.
| Rule | Word | Result |
|---|---|---|
Vx > Vː | tax | taː |
Vx > V[+long] | tax | taː |
Vh₂ > Vː | suh₂kéh₂ | suːkéː |
Cx > C | tkxa | tka |
aC > C | tak | tk |
CxC > CaC | tkxta | tkata |
VC > CV | tak | tka |
With one such class in the source, every one in the target is that sound, so Cx > CC turns tkxa into tkka. With several, both sides need the same count, and they pair up in order.
Number the sounds where the sides don't pair up. A number can stand beside a class without one: in V₁xV > V₁V the plain V is the second vowel. The IPA keyboard types ₁.
| Rule | Word | Result |
|---|---|---|
C₁C₂ > C₁ | takta | taka |
V₁CV₂ > V₁ | taka | ta |
V₁xV > V₁V | axe | ae |
The field refuses a target it can't read and gives the rewrite where there is one.
| Refused | Write |
|---|---|
CC > C | C₁C₂ > C₁ |
VCV > V | V₁CV₂ > V₁ |
ex > V, x > V, V > C, {a, b} > V | The sound itself, as in ex > e |
∅ > V / _x | The sound to add, as in ∅ > ə / _x |
Vx > [+long] | Vx > V[+long] |
(C)x > C | Two rules, as Cx > C and x > ∅ |
{Cx, b} > C | One target per option, as in {Cx, b} > {C, d} |
a > # | a > ∅ / _# |
A rule of this kind that is already saved still opens. A word it fits keeps its form at that rule, the next rules run on, and pointing at the ! on its row says what to fix.
What a rule compares
A rule compares sounds. One sound spelled two ways matches, as g and ɡ, ʧ and t͡ʃ, or ʦ and t͡s do, and the word keeps its spelling of the sounds the rule leaves alone. g > ɡ evens out a spelling.
| Rule | Word | Result |
|---|---|---|
g > k | ɡaɡa | kaka |
ʧ > ʃ | t͡ʃa | ʃa |
t͡s > s | ʦa | sa |
g > k / a_# | ɡaɡ | ɡak |
g > ɡ | gaga | ɡaɡa |
Spellings that read as other sounds stay apart: ts is two sounds and t͡s one affricate, and a: is a and a colon where aː is a long vowel. A sound with a diacritic is a sound of its own, so name both to take both.
| Rule | Word | Result |
|---|---|---|
ts > x | t͡sa | t͡sa |
a: > x | aːt | aːt |
a: > aː | a:t | aːt |
a > e | aːt | aːt |
{a, aː} > e | aːta | ete |
A word's ' and : are read as sounds too, since the catalogue doesn't know them. So is ’, which phones and Macs type in place of '. A rule looks past a stress mark and not past a sound, so a ' written for stress stops every environment that spans it. Write ˈ for stress, ʼ for an ejective and ː for length. The lexicon marks a form holding a symbol the catalogue doesn't know.
| Rule | Word | Result |
|---|---|---|
x > ∅ / V_C | sux'kex | sux'kex |
x > ∅ / V_C | suxˈkex | suˈkex |
oː > u | to:ma | to:ma |
oː > u | toːma | tuma |
Idle changes
A change is idle when it changed no word of the line. Each row carries a count such as 1/1/3, and pointing at it says the same in words.
| Number | What it counts |
|---|---|
| First | Words the change changed among the columns shown |
| Second | Words it changed on the whole line |
| Third | Words the line carried when it ran |
0/4/12 is a change at work on words you aren't showing. 0/0/12 is idle.
Under the line's name, a count such as 2 idle opens the list of idle changes, each with its reason.
| Reason | Next step |
|---|---|
No word here has S. | Compare the rule with the forms above its row. An earlier change may have taken the sound. |
Spaces don't separate sounds. | Use the rewrite in the message. |
Write the arrow as >: a > e | Use the rewrite in the message. |
C in the target could be any C in the source. | Use the rewrite in the message. The words the rule fits keep their form until then. |
[place] is a group of features. | Write one of the features the message names. |
No sound is −aspirated. | Write the sounds themselves, as the message shows. |
Words here have ', which is not in the catalogue | In the words, write the symbol the message offers in its place. |
No effect on the 6 words here. | No word fits the rule as written. Check the environment first, including any . the words don't carry. |
Rules and inventories
An inventory has no effect on a rule. A rule reads the words of its line, so it can match and write sounds the inventory doesn't list, and editing a chart can't fix an idle rule. Change the rule or the words. A stage's Sounds tab sets the sounds of its words against the inventory.
Sound change patterns
A pattern adds a few sound changes of one kind from the history of real languages, such as Grimm's law or umlaut. Its rules are written with features and classes, so it fits any set of sounds.
- Point at a row of the matrix, press + and choose Add changes from a pattern.
- Pick a pattern. Each of its changes shows how many of the line's words it changes at that point, with a few of them before and after. A change that changes no word is marked skipped.
- Press Add and the number of changes.
Each change arrives as a sound change of one rule, which you edit, move and delete like any other. The skipped ones are left out, and the plan's limit on changes counts the rest. Lemula has eight patterns, and Generate a language describes how they are picked and run.