luggifer4

The Method

Contents

Introduction

This is all a work in progress. Suggestions and feedback are appreciated: please email me or send them to my tumblr askbox. If you're offering feedback on the process, please only do so after actually trying it out. Also, none of this has been tested quite yet. I aim to create a language to test the Phonology section once I've finished it and another to test the Grammar section once that's done.

This page assumes familiarity with some linguistic concepts. I've linked to relevant Wikipedia articles when I can.

This method is based on what I'm calling the "standard diachronic method", which seems to be in very common use. A good beginner's introduction is the YouTube video series How to Make a Language by Biblaridion. That method is good, great even, if you just want to get a language out the end. But I want more than that. I want a language and a record of its history that's detailed enough to know exactly what the language looked like (what constructions were available, how marked each one's use was, what the exact nuances of each one's meanings were, precisely how words were pronounced, and so on and so forth) at every point in time in the language's history, rather than just the end.

Aims

  1. This method should be able to describe naturalistic constructed languages with a wide variety of language features. Specifically:

  2. Using a description of a language created with the method, it should be easy to derive the state of a language at any point in its history within the time interval described, for the purposes of:

    • Creating and interpreting texts in the language
    • Translating into or out of the language
    • Discussing the language from a synchronic perspective

  3. It should be easy to further develop a language in the future direction.

Non-aims

I do not aim to model any of the following, though they may be explored in future iterations.

Glossary

Some important terms you might not know, whose definitions are simple enough that they don't need a whole Wikipedia article.

Diachronic: considering development over time.

Productive: (of a construction, rule, affix, etc.) able to be applied to novel situations, as opposed to being arbitrarily limited in scope.

Synchronic: considering only the instantaneous state of a system; i.e., not diachronic.

Phonology

Phonemic inventory

Typically, synchronic representations of a language's phonology include an inventory of phonemes and a list of allophonic rules. The phonemes are normally given in tables, often using the International Phonetic Alphabet (IPA). This is a very effective way of keeping track of phonemes, so we'll be doing pretty much the same thing. Create a phoneme table for the start of your language's history. For example, here's a table of consonants in Australian English:

Labial Dental Alveolar Post-alveolar Palatal Velar Glottal
Nasal m n ŋ
Plosive fortis p t k
lenis b d g
Affricate fortis tʃ
lenis dʒ
Fricative fortis f θ s ʃ h
lenis v ð z ʒ
Approximant central (w) ɹ j (w)
lateral l
Adapted from: Wikipedia, Australian English phonology (accessed 23 September 2026)

Don't forget about vowels, or tones if you have them.

Over time, allophonic rules (see below) will decay into changes at the phonemic level. There are three kinds of changes we're interested in:

Notice that simple shifts result in phonemes being relabelled. Remember that a phoneme is just a label we give to a cluster of related physical sounds. The label itself doesn't technically matter: as a matter of convention and convenience, we just like to give phonemes labels that correspond with the general idea of the sounds they represent. So if the sounds a phoneme represents change significantly, that could be cause for a relabelling.

We'll keep a changelog of all the changes that have happened since the last phoneme table. Each entry in the changelog will record:

A relabelling caused by a simple shift won't necessarily be associated with the end of an allophonic rule. The other two kinds of changes are likely to be associated with allophonic rules, however, because, as far as I know, splits and mergers are generally apparent in phonetic realisations before becoming phonemic.

I use notation involving +, - and >, but you can use whatever you want, as long as it works for you. "X > Y" is just shorthand for "-X, +Y".

For example:

Time Phoneme table change
50 +kʷ
60 -y
75 pʰ > ɸ

The point of these changelog entries isn't to describe the actual sound changes. (We'll get to that later.) It's simply to enable you to construct a phoneme table at whichever point in time you want. Which brings us to the last step.

If you're tracking a sufficiently large portion of your language's history, you may find that the changelog becomes exhaustively long. Not to worry. As soon as that happens, or before, create a new phoneme table with all the changes applied. Then, if you want a table of phonemes at any point in time -- for example, to coin a new word, or determine the form of a loanword, or check that your orthography covers all the phonemes -- just take the most recent phoneme table, and apply all the changelog entries since then.

Allophonic rules

In the standard diachronic method, sound changes are normally thought of as instantaneous. That isn't precise enough for us. We need to recognise that sound changes typically happen in several steps:

  1. Speakers start unconsciously pronouncing words differently, usually because the new pronunciation is easier (involves less movement of articulators) than the old one.
  2. Children acquiring the language only have access to the new pronunciation, so they internalise it as an allophonic rule of the language.
  3. Children acquiring the language inevitably simplify its phonology, interpreting what were previously allophonic distinctions as phonemic ones. The sound change is no longer productive in new words (though it may remain present allomorphically, as in the English plural ending -s).

For written languages, there is also a fourth step, spelling reform, wherein the spellings of words are updated to reflect the change. However, that is beyond the scope of this project.

Some sound changes are instead caused by stratum conditions. That is also beyond the scope of this project.

Each of these steps will take some time to fully propagate through a speech community. However, in the interest of simplicity, we will not be modelling (1), as it will vary greatly between individual speakers, and we will model (2) and (3) as each occurring instantaneously, because that provides a level of temporal resolution I'm happy with, and I'm driving this train.

The period of time between (2) and (3), when the sound change manifests as an allophonic rule, productive in new words, is called the lifetime of that rule. At the end of a rule's lifetime, it decays and may produce a phoneme table change (see above).

Simple allophonic rules, which just universally shift a phoneme's pronunciation, are likely to decay immediately: that is, (2) and (3) are likely to occur at basically the same time. In this case, don't bother recording it as an allophonic rule at all.

Whenever I come up with a new allophonic rule, I generally like to give it an identifier: a number or a short description that I can use to refer to the rule in a way that's abstracted away from its phonetic effect. (If you like, you can skip this step and just refer to a rule by a formal expression of its effect.) For example:

Name Effect
Intrusive /r/ # > r / V_V
/t/-glottaling t > ʔ / !_V
/k/-labialisation k > kʷ / _[+round]
Train-changing t > tʃ / _[+postalveolar]
Y-I merger y > i
Credit is due to Dr. Geoff Lindsey for the incredible name "train-changing" for this allophonic rule in some English dialects

I'm unsure whether it's better to record the lifetime of an allophonic rule in the rule table or, as I've done here, in the same list that contains our phoneme changes. Recording them in the rule table would make it easier to see the lifetime of a single rule; but recording the start and end of of each rule's lifetime as an event in the same table as the phoneme changes makes it easier to see which rules are associated with which phonemes. It's almost certainly better to pick one or the other; it's inconvenient to have to remember to change a value everywhere it's recorded if you record it in multiple places. Testing is probably required.

For now, we'll be recording the lifetimes of our allophonic rules with the list of phoneme changes. For example:

Time Allophonic rule event Phoneme table change
30 +"/k/-labialisation"
40 +"Y-I merger"
50 -"/k/-labialisation" +kʷ
60 -"Y-I merger" -y
75 pʰ > ɸ
85 +"Train-changing"

As you can see, here, the allophonic rule "/k/-labialisation" appeared at time 30, introducing a new allophone [kʷ], then decayed at time 50, leaving behind an extra phoneme /kʷ/, which is simultaneously added to the phoneme table. Likewise for the Y-I merger, except its decay instead removed the phomeme /y/.

Note that if there had already been a phoneme /kʷ/ before the decay of /k/-labialisation, then the "phoneme table change" cell at time 50 would be empty, as the phoneme table would not have changed.

To do