Conlang template

[Qwynegold]

I while ago I started thinking that it would be good if I had some kind of template I could use when conlanging, because I always get stuck somewhere in the process and never finish anything. A template should go through one area at a time, present what possible options there are, and ask you to choose one. I think this could speed up the process. Then I saw the first episode of LangTime, and saw that David Peterson already has a template for himself.

So I was thinking like first it should present some very basic questions like:

Choose a morphological type, or a combination of several types.
1) isolating
2) fusional
3) agglutinative
4) polysynthetic
5) oligosynthetic
6) non-concatenative

And then there would be questions about increasingly detailed stuff. Of course you can choose to offroad at some places and add stuff that's not in the template.

I haven't really decided yet if I'll attempt to create a template like this. It seems like it's a lot of work, and lately I have not had any energy. But I thought I'd just ask here if anyone has any comments or suggestions. Is it a good idea or a bad idea? Is anyone else using templates for conlanging?

[bradrn]

The main reference grammar template that I’m aware of is the Lingua Questionnaire, an amazingly thorough set of questions covering pretty much every aspect of grammar.

(A little rant: The Questionnaire was originally intended as a template for structuring reference grammars, such grammars being immediately recognisable because they inevitably start with Section 1.1.1: Direct and indirect speech — something which no sane grammar-writer would usually do. You may not be surprised to hear that I find such grammars incredibly awful to read through, and a pain in general to use. Please, I beg of you, if you do end up using the Lingua Questionnaire, please do not structure your grammar following the Questionnaire!)

Apart from that one, another template I’m aware of is Describing Morphosyntax: the workshop approach, an adaption of Describing Morphosyntax for conlangers. As a shorter list of questions, possibly this might be more useful to you than the Lingua Questionnaire (though that one is of course more thorough).

(Not that I’ve actually used any of those templates, those are just the ones I’ve heard of… but I probably should use them, given that I often struggle to figure out which bits of my languages I still need to do…)

[Vardelm]

Is it a good idea or a bad idea?

This might be one of those things that is a really good idea.... only. Once you get down to the practicalities of documenting a language, can you really shoe-horn all languages into 1 template? That might depend on what purpose the grammars that come out of said template are for. If it's a reference grammar, that probably helps. If it introduces people to the language and builds each topic on top of the ones that precede, that might be difficult. I have found with the languages I'm working on that what gets introduced when might vary. How flexible the template is and how rigidly one sticks to it would play a role as well.

Is anyone else using templates for conlanging?

Not right now, although I think I tend to think of my languages in the same systematic way. For instance, I tend to think of adjectives, genitive constructions, and adjunct phrases as being somewhat similar in function (qualifying an argument), so maybe that's template-ish?

Down the road here, I'd like to get my conlangs documented using VulgarLang. I have followed it's development for a while now, and I think it's useful beyond the simplistic random generation tool that it was a couple years ago. The capabilities for evolving descendant languages still needs work. However, I'd like to have something like this and use a template for the documentation so that all of my languages have a fairly consistent structure to their grammars, just to make everything easier to manage. That might be where templates are useful, although I haven't done anything with managing an entire language family yet. Just having consistency across grammar seems like a good idea, and might offset any downsides to making grammars with a structure more customized for a particular language.

[bradrn]

Is it a good idea or a bad idea?

This might be one of those things that is a really good idea.... only. Once you get down to the practicalities of documenting a language, can you really shoe-horn all languages into 1 template? That might depend on what purpose the grammars that come out of said template are for. If it's a reference grammar, that probably helps. If it introduces people to the language and builds each topic on top of the ones that precede, that might be difficult. I have found with the languages I'm working on that what gets introduced when might vary. How flexible the template is and how rigidly one sticks to it would play a role as well.

My impression was that Qwynegold wanted a template just to give a guide for which areas to look at next when developing a conlang, an application of templates which seems fine to me. (Although I vehemently agree with you on the inadequacy of templates for structuring reference grammars!)

[Vardelm]

My impression was that Qwynegold wanted a template just to give a guide for which areas to look at next when developing a conlang, an application of templates which seems fine to me. (Although I vehemently agree with you on the inadequacy of templates for structuring reference grammars!

Makes sense. In that sense, I haven't really used a template at all, although I have referred to the Wikipedia article on inflection quite a bit (among other articles) to help remind myself of what possibilities there are.

[Kuchigakatai]

A little rant: The Questionnaire was originally intended as a template for structuring reference grammars, such grammars being immediately recognisable because they inevitably start with Section 1.1.1: Direct and indirect speech — something which no sane grammar-writer would usually do. You may not be surprised to hear that I find such grammars incredibly awful to read through, and a pain in general to use. Please, I beg of you, if you do end up using the Lingua Questionnaire, please do not structure your grammar following the Questionnaire!

Yes, reference grammars are that way, but that's because you're supposed to know the language already! It doesn't ultimately matter much what order a reference grammar puts things in in the end, as you should pretty much jump to any chapter and find more or less the same level of detail and complexity. That's what makes them reference grammars.

You seem to be confusing them with textbooks, which are easier to read from the beginning, and continuously build on the previous topics. You can in fact write a textbook in a way that it can also serve as a reference grammar (Haywood and Nahmad's A New Arabic Grammar of the Written Language (1965) would be an example), but then you have the problem that a bunch of the information will be disjointed across different lesson units / chapters (and I've heard people complaining about that regarding Haywood and Nahmad's grammar as a work of reference).

By the way, I think the Lingua Questionnaire starts with the syntax section as a way to combat some still-existing tendency to concentrate on phonology and morphology, skipping a great number of things about syntax. Notice, too, how section 1.1 makes some particular emphasis on documenting echo questions (a phenomenon rarely included at all in grammars) and minor imperatives (often skipped as well, or mentioned in too little detail). The Lingua Questionnaire tries to be a counterweight to deficiencies in the trends of reference grammars.




The topic about what order to create things in in a conlang is interesting, because a language sprawls in many directions, and yet you have to make the whole thing coherent. It seems to me a basic phonological inventory is in fact necessary at the beginning, along with basic inflections and word order, but then the whole thing starts growing and there might be different ways to keep track of it. Zompist has long recommended making many sections and jumping between them up and down, or alternatively, using loose sheets that you keep reordering (see Online LCK > Writing > Basic outline), and I'd add, you could also try an electronic equivalent of the latter.

Regarding the topic of this thread, I've made my own chapter template to guide the order of topics, but I have no lists of options inside (I added some clarifications in my translation):

Code: Select all

Rasgos (features of interest)
Fonología (phonology)
Ortografía (orthography)
Morfología flexional (inflectional morphology)
Morfología derivacional (derivational morphology, includes numbers and indefinite/interrogative pronouns)
Orden de palabras básico (basic word [and constituent] order)
Uso de casos (uses of cases)
Posesión (possesion)
Aposición (adposition)
Uso de aspectos (uses of aspects)
Tiempo (tenses or the expression of time)
Pronombres personales (personal pronouns)
Demostrativos (demonstratives)
Determinantes y pronombres indefinidos (determiners and indefinite pronouns, including quantifiers like 'all' and 'most')
Números (numbers, how they're used)
Sintagmas nominales (noun phrases: structure with the internal word order and grouping possibilities, effects of subconstituent weight)
Sintagmas verbales (verbal phrases: structure with verb auxiliaries and arguments, predicative/locative/presentative copulas)
Sintagmas adjetivales (adjetival phrases: structure with degree adverbs, complements as in 'hard to say' and 'similar to me')
Sintagmas adverbiales (adverbial phrases, leaving conditions aside: temporal, locative, concessive, causal, purposive/final)
Subordinadas completivas (noun clauses)
Subordinadas relativas (relative clauses, including "headless" (or "fused-relative") ones)
Condiciones (conditions: if)
Conjunciones (conjunctions)
Parataxis (parataxis vs. hypotaxis: choosing coordination vs. subordination)
Topicalización (topicalization)
Preguntas (questions: polar, biased polar, content, echo, rhetorical: 'wouldn't it be...?', '(how) could it be that...?', 'is it not...?')
Exclamaciones (exclamatives)
Negación y restricción (negation and restriction: 'only', 'except for')
Verbos modales (modal [and other such auxiliary] verbs)
Partículas modales (modal [sentential] particles)
Imperativo (the imperative)
Pasivo (the passive)
Reflexivo y recíproco (expression of reflexives and reciprocals)
Comparación (comparison)
Foco (focus, combining phonological and syntactic marking)
Cambios de valencia (valency changes)
Reducción contextual (contextual reduction, i.e. dropping constituents to avoid repetition)

You can see that this list is biased towards the conlangs I like to make though: presence of case (if only direct vs. genitive), some absence of syntactic markers for nouns (which produces ambiguity between possession and adposition sometimes, that's why the adposition chapter appears where it does), aspect is more important than tense, conditions are complex enough that they merit their own chapter, a relatively high use of parataxis, sentential particles... Also, from chapter 6 to 37 the discussion is all about syntax (only 4 and 5 are about morphology), which again reflects where my interests are.

[bradrn]

Yes, reference grammars are that way, but that's because you're supposed to know the language already! It doesn't ultimately matter much what order a reference grammar puts things in in the end, as you should pretty much jump to any chapter and find more or less the same level of detail and complexity. That's what makes them reference grammars.

You seem to be confusing them with textbooks, which are easier to read from the beginning, and continuously build on the previous topics. You can in fact write a textbook in a way that it can also serve as a reference grammar (Haywood and Nahmad's A New Arabic Grammar of the Written Language (1965) would be an example), but then you have the problem that a bunch of the information will be disjointed across different lesson units / chapters (and I've heard people complaining about that regarding Haywood and Nahmad's grammar as a work of reference).

An interesting perspective! I do agree that reference grammars are most useful when you already know the language — and that I would never try to learn a language from a reference grammar — but that’s not incompatible with a reference grammar being easily readable from beginning to end, without making it into a textbook. (A particularly good example of this is, say, Hyslop’s grammar of NE Ambae.) I read a grammar in that way primarily when I want to get a good sense of how a language works in general, what constructions it uses etc., without wanting to find out about any specific construction I want to know about. And such considerations are particularly important for conlang grammars, which are normally read beginning to end to get a good overview of the language.

[Kuchigakatai]

An interesting perspective! I do agree that reference grammars are most useful when you already know the language — and that I would never try to learn a language from a reference grammar — but that’s not incompatible with a reference grammar being easily readable from beginning to end, without making it into a textbook. (A particularly good example of this is, say, Hyslop’s grammar of NE Ambae.) I read a grammar in that way primarily when I want to get a good sense of how a language works in general, what constructions it uses etc., without wanting to find out about any specific construction I want to know about. And such considerations are particularly important for conlang grammars, which are normally read beginning to end to get a good overview of the language.

Hmm, right. Maybe we could say there's also a difference between reference grammars meant for language learners, and those meant for linguists. A linguist (or conlanger) has a much easier time using a reference grammar when less knowledge is assumed, and glosses are provided (I can't help but remember clawgrip's eternal gripe about no one publishing grammars of languages like Latin, Old Church Slavonic and Standard Arabic with glosses, which meant he had to learn a bit of languages like those in order to explore the reference grammars).

By the way, I think it should be possible to write a reference grammar where section 1.1.1 is direct vs. indirect speech generalities in a way you (or most conlangers) would enjoy. You mentioned that:

You may not be surprised to hear that I find such grammars incredibly awful to read through, and a pain in general to use. Please, I beg of you, if you do end up using the Lingua Questionnaire, please do not structure your grammar following the Questionnaire!

but I find that when *I* experience this, it's because said section 1.1.1 has way more detail than I could possibly care about, or because it's anxiously concerned with showing how the assumptions of some random linguist's theoretical model fits its data well (and both problems are independent of the question of chapter order).

[bradrn]

An interesting perspective! I do agree that reference grammars are most useful when you already know the language — and that I would never try to learn a language from a reference grammar — but that’s not incompatible with a reference grammar being easily readable from beginning to end, without making it into a textbook. (A particularly good example of this is, say, Hyslop’s grammar of NE Ambae.) I read a grammar in that way primarily when I want to get a good sense of how a language works in general, what constructions it uses etc., without wanting to find out about any specific construction I want to know about. And such considerations are particularly important for conlang grammars, which are normally read beginning to end to get a good overview of the language.

Hmm, right. Maybe we could say there's also a difference between reference grammars meant for language learners, and those meant for linguists.

Yes, I definitely agree. I tend to find the former grammars much more difficult to use, because on those occasions when I do want to find something specific, I usually can’t. The latter are much better for that purpose, but then tend to be harder to read from front to back.

In general, I think there is a trade-off between two ways of using a grammar. You can optimise a grammar for being read from beginning to end, but that makes it harder to find specific things; or you can optimise a grammar for finding specific things, but that makes it harder to read it from beginning to end. The trick is trying to find a way to structure a grammar such that it makes both uses easier, which is what Hyslop has managed to do. (And of course, then you get the reference grammars which seem to be optimised for neither use!)

By the way, I think it should be possible to write a reference grammar where section 1.1.1 is direct vs. indirect speech generalities in a way you (or most conlangers) would enjoy. You mentioned that:

You may not be surprised to hear that I find such grammars incredibly awful to read through, and a pain in general to use. Please, I beg of you, if you do end up using the Lingua Questionnaire, please do not structure your grammar following the Questionnaire!

but I find that when *I* experience this, it's because said section 1.1.1 has way more detail than I could possibly care about, or because it's anxiously concerned with showing how the assumptions of some random linguist's theoretical model fits its data well (and both problems are independent of the question of chapter order).

No, I don’t agree with that — if I’m trying to read a grammar from start to finish to get a good idea of a language, I am not terribly interested in topics such as direct and indirect speech, question formation and imperatives. What I am interested in is topics like: what is the overall typology of the language? How do I read the orthography? What is the structure of basic clauses? How are the noun and verb phrases structured? And what categories do nouns and verbs inflect for? The Questionnaire tends to bury such questions in the middle of grammars, in what tends to be the least convenient place for readers.

[Ares Land]

I haven't really decided yet if I'll attempt to create a template like this. It seems like it's a lot of work, and lately I have not had any energy. But I thought I'd just ask here if anyone has any comments or suggestions. Is it a good idea or a bad idea? Is anyone else using templates for conlanging?

I use the outline of reference grammars to check that I do cover everything. I use both zompist's conlang outlines (they cover a very broad range of subjects, so they're very good for avoiding stupid mistakes like 'Oops, I forgot to cover questions and reported speech') and natlang grammars. (To check if I haven't missed some possibilities.)

Other than that... I don't rely that much on linguistic typology, really.)
Over the years, I've come up with a process of my own.

- I work out a proto-language first. The grammar for that one is fairly rough which I think is only naturalistic. (We don't know that much about PIE or proto-Semitic after all).
- I try to come up with a realistic scenario of how it evolves. (For instance, in the Helian family, the proto-language is losing a lot of morphological complexity to sound change, but gained new affixes at the same time. In the Tarandim family, the proto-language has many features of creoles, but grammaticalized a lot of construction, turned most particles into bound morphemes).
- I come up with sound changes. What I do is come up with a realistic target phonology, and try to figure out how to go there from what I have. I also like to add some of those sound changes that really mess up the way the language works (I'm fond of the way Old French dropped pretty much all intervocalic stops, for instance).

Then to work out grammar, I select a topic that interests me, run a great deal of possible proto-language constructions through the SCA and figure out grammar out of the results, trying to come up with plausible regularizations and odd cases/irregularities.

Sometimes I'm bored with the results, or I figure I'd like to play with some particular idea, so I adjust the sound changes accordingly.
With the Tarandim family, I was bored with most of the words I had and I wanted to play with root and pattern morphology, so I added a syncope rule.

With Erdanila, for instance, I wanted to work out the proverbial polysynthetic conlang; the results always felt too artificial to me.
Then I started working with sound changes, not bothering much with typology or features, but just trying to make sense of what came out of the SCA, and I was a lot more satisfied with the results.

I don't know if I can recommend the process, because it's absurdly long.

But I like the way you have to think both like a native speaker and like a linguist documenting an unknown language.
And you get nice surprises too. Sometimes I discover unexpected things in Tarandim morphosyntax; which is pretty thrilling. When working on Erdanila, sometimes I solved a problem in morphosyntax only to find out that Mohawk solves the same issue in the same way.

As for writing grammar... I deliberately moved away from academic models. If I write for myself, anyway, tables and a telegraphic style are more than enough. If I write for others, ie. if I'm sharing it here, using a formal academic style would only bore the reader to death, so I try to keep a clear and attractive style, and I deliberately avoid the traditional order. You can spend pages describing phonology, for instance, so I tend to skip that part. I think I'd like to try a 'language lessons' format someday too.

[Qwynegold]

The main reference grammar template that I’m aware of is the Lingua Questionnaire, an amazingly thorough set of questions covering pretty much every aspect of grammar.

(A little rant: The Questionnaire was originally intended as a template for structuring reference grammars, such grammars being immediately recognisable because they inevitably start with Section 1.1.1: Direct and indirect speech — something which no sane grammar-writer would usually do. You may not be surprised to hear that I find such grammars incredibly awful to read through, and a pain in general to use. Please, I beg of you, if you do end up using the Lingua Questionnaire, please do not structure your grammar following the Questionnaire!)

Yeah, I have downloaded that one a long time ago. I find that after you have filled in like 1% of the questionnaire, you get so fed up with it that you never return to it again. So the Lingua Questionnaire does not suit my purpose; I want a tool that helps you actually get things done and not get bogged down in the small details.

Apart from that one, another template I’m aware of is Describing Morphosyntax: the workshop approach, an adaption of Describing Morphosyntax for conlangers. As a shorter list of questions, possibly this might be more useful to you than the Lingua Questionnaire (though that one is of course more thorough).

Oh thanks, I wasn't aware of that one. Though some of those questions I have no idea what they're about, like "What are the distributional properties of nouns?", "What are the structural properties of nouns?" and "Dative of interest". I would like to have something that presents all the possible options for any given question (though of course it should not deter you from coming up with creative ideas that are not attested in any natlangs).

[bradrn]

I would like to have something that presents all the possible options for any given question (though of course it should not deter you from coming up with creative ideas that are not attested in any natlangs).

In that case, you might find the WALS chapters useful (though I’m sure you’re aware of that already).

[Qwynegold]

This might be one of those things that is a really good idea.... only. Once you get down to the practicalities of documenting a language, can you really shoe-horn all languages into 1 template?

Probably not, so you'd have to be aware that a template might lead you to think in one specific way. And you should not be afraid to break free from the template if you feel that there's some part of your conlang that the template is not suitable for.

That might depend on what purpose the grammars that come out of said template are for. If it's a reference grammar, that probably helps. If it introduces people to the language and builds each topic on top of the ones that precede, that might be difficult. I have found with the languages I'm working on that what gets introduced when might vary. How flexible the template is and how rigidly one sticks to it would play a role as well.

I was thinking like if you want to quickly develop a conlang to such a level that it's somewhat usable, you should first make decisions about the biggest, most important things, like morphosyntactic alignment, etc. After that, the second most important things, e.g. are your verb gonna have any tenses? If so which ones? Which voices will you have? Any aspects? Are you going to use noun cases or adpositions? And so on. So this would not lead to a well structured reference grammar, because information about a specific topic would be scattered throughout. Maybe the template could have suggestions on how to group different parts together in a grammar?

Down the road here, I'd like to get my conlangs documented using VulgarLang.

Hmm? I'll have to play with that trial version and see what it does...

[Qwynegold]

Makes sense. In that sense, I haven't really used a template at all, although I have referred to the Wikipedia article on inflection quite a bit (among other articles) to help remind myself of what possibilities there are.

Yeah, I often look at various pages on Wikipedia to see what options there are. If I were to make this template, I would for example present a big list of noun cases with a small explanation of what each case does. For those cases that I don't really understand I would just make links to Wikipedia articles. I would also include information like "38% of natlangs have no case, 9% have only two cases, 3% have three cases" etc, according to WALS. So you'd get a sense of what's normal and what's exceptional.

[Qwynegold]

[long post]

Thanks for telling about your process! Do you have the problem that, before you've even realized, you have stopped working on a conlang because it's tedious or because there's some specific hurdle that you need to solve before you can move on? Do you reach a satisfying level of "completeness" in your conlangs?

[Vardelm]

Down the road here, I'd like to get my conlangs documented using VulgarLang.

Hmm? I'll have to play with that trial version and see what it does...

The grammar editor with the use of Regex to create your morphological forms is what's interesting to me. It allows for the "translator", where you can basically input a gloss and get the resulting text in the language.

I was thinking like if you want to quickly develop a conlang to such a level that it's somewhat usable, you should first make decisions about the biggest, most important things, like morphosyntactic alignment, etc. After that, the second most important things, e.g. are your verb gonna have any tenses? If so which ones? Which voices will you have? Any aspects? Are you going to use noun cases or adpositions? And so on.

Definitely agree here.

As I described in this post, deciding on a few key features of the language really helps. If certain parts of the language aren't working out, then you have a better idea of what's important & should be kept vs redesigning or discarding. I actually think that picking the key features at the start will lead to less roadblocks overall.

So this would not lead to a well structured reference grammar, because information about a specific topic would be scattered throughout. Maybe the template could have suggestions on how to group different parts together in a grammar?

This leads to an interesting thought: should the "template" for designing a conlang be different from the "template" used to document that conlang? I think probably "yes".

[long post]

Thanks for telling about your process! Do you have the problem that, before you've even realized, you have stopped working on a conlang because it's tedious or because there's some specific hurdle that you need to solve before you can move on? Do you reach a satisfying level of "completeness" in your conlangs?

I know you addressed this to Ares Land, but the point above about the key features of a conlang addresses this. I have also found that - for me - having a specific sort of cultural aesthetic to the language helps maintain my interest through. On the old board, I got bogged down with my Devani language because it had the personality of wet cardboard. I had a lot more fun with Jin at that point, so my current work in my scratchpad follows that same path. Dwarven is now strongly Celtic flavored and Devani is very Indic. The other thing too is just having multiple conlang projects going at the same time allows me to move back & forth such that when I do run into a hurdle, I can just set it aside & work on something that I do have inspiration for at the moment. So multiple conlang projects are easier in some ways that focusing on 1. I can't answer the "completeness" question as of yet, but it's going well.

[Ares Land]

[long post]

Thanks for telling about your process! Do you have the problem that, before you've even realized, you have stopped working on a conlang because it's tedious or because there's some specific hurdle that you need to solve before you can move on? Do you reach a satisfying level of "completeness" in your conlangs?

Oh yes, sure. I think these issues are just a normal part of any artistic project. (I write short stories and novels too -- none of them readable, I'm afraid -- and the tediousness or hurdles are definitely there too).

My main problem, though, is lack of time. Real life tends to inferfere (I also make things harder for myself by pursuing way too many projects).

Classical Tarandim was pretty "complete"... but I'm a lot better at conlanging now than I was when I worked on it, so of course II'm reworking the entire family. Sigh. I can't recommend that kind of perfectionism, but I guess I'll have to learn to live with it.

[bradrn]

So this would not lead to a well structured reference grammar, because information about a specific topic would be scattered throughout. Maybe the template could have suggestions on how to group different parts together in a grammar?

This leads to an interesting thought: should the "template" for designing a conlang be different from the "template" used to document that conlang? I think probably "yes".

Yes, I completely agree! I’ve long thought that the structure of a reference grammar should depend on the language being described, rather than just using a template.

Classical Tarandim was pretty "complete"... but I'm a lot better at conlanging now than I was when I worked on it, so of course II'm reworking the entire family. Sigh. I can't recommend that kind of perfectionism, but I guess I'll have to learn to live with it.

Sounds familiar! I just tell myself that once a conlang is reasonably complete, I don’t go back to it. That’s worked so far, since all my conlangs so far have been intended as protolangs so I don’t want to spend an excessive amount of time on them… not sure how that will work out with non-protolangs. (Of course, I’m far more linguistically knowledgable now than I was with those other languages, so hopefully I won’t need to rewrite these newer ones too much.)

[Moose-tache]

I think the only universal template for grammars is this:

Micro-Syntax
Meso-Syntax
Macro-Syntax

...where each section is informed by how the language builds phrases. In English, Micro-Syntax would be broken into sections about morphology and parts of speech, Meso-Syntax would describe the internal structure of noun phrases, verb phrases, prepositional phrases, etc., and Macro-Syntax would deal with clause structure and coordination, and transformations. In a language with no formal noun-verb distinction, no morphology, no topical fronting, etc., the general sections would still apply, even though their content would be completely different.

[bradrn]

I think the only universal template for grammars is this:

Micro-Syntax
Meso-Syntax
Macro-Syntax

...where each section is informed by how the language builds phrases. In English, Micro-Syntax would be broken into sections about morphology and parts of speech, Meso-Syntax would describe the internal structure of noun phrases, verb phrases, prepositional phrases, etc., and Macro-Syntax would deal with clause structure and coordination, and transformations. In a language with no formal noun-verb distinction, no morphology, no topical fronting, etc., the general sections would still apply, even though their content would be completely different.

Interestingly enough, I find that I absolutely hate reading grammars which are structured in that order. I much prefer the opposite order: macro, then meso, then micro. (With a phonology section at the beginning, of course.) And of course it’s dependant on the language: you could for instance make a good argument for putting either morphology or syntax at the beginning, depending on how agglutinating the language is.

In general, I think this is my ideal grammar structure:

Phonology
Parts of speech
Basic clause syntax
Complex clause syntax
Morphology
Noun phrase
Verb phrase

…where you can rearrange the last five sections pretty much as needed, although you should put either ‘basic clause syntax’ or ‘morphology’ first. I prefer reading grammars where I can understand each glossed example using only the material before that example, so I try to rearrange the sections so I can achieve that.