There are very few new real-world orthographies that are really "deep" —
although Korean is one.  This is probably because speakers — perhaps
especially those learning to read — find such orthographies hard to use.
After all speakers "know" the morphophonemic rules, and using that
knowledge plus context clues to "back-derive" the underlying form may
actually be easier for average humans than the other way around. French and
English (and Tibetan!) and other orthographies with much historical depth
would probably not have worked as well as they do if there had been (lots
of) anaptyxis and not just loss and merger of phones. In other words I'd go
for a pretty surfacey romanization with some morphophonemic info thrown in
for disambiguation if and when needed and possible.

IMO ‹y› for /ɨ/ is perfectly cromulent. I would probably have used it for
Sohlob if there had been a precomposed capital ‹J› with caron, freeing up
‹j› for /j/. In the end having similar notations for ‹ä› /æ/ and ‹ï› /ɨ/ is
rather convenient though. Of course I have no tones to worry about. I could
never do a tone language because I'd never be able to speak it. Even though
I do manage the word accents in Swedish (being my L1) I'm totally tone deaf

>/mîr?rh/ (underlying form; ? means I have no idea how to transcribe an
>underlying empty vowel slot)

I can think of a few:

- Apostrophe
- Middle dot
- ‹ə›
- ‹V›
- ‹v› (if it's free)

Of these ‹ə› and ‹V› are what I use for an "undetermined" vowel which takes
all its features except [+vowel] from vowel harmony (essentially a short
copy of the nearest preceding long vowel). When citing actual words as
opposed to affixes I just write the surface vowel though.

For the //n// -> [θ] thing you might do what Irish orthography sometimes
does and write both the underlying and the surface consonant: ‹nth› or
‹thn›, at least if those don't occur as clusters, or don't in the relevant

For upstep/downstep you might pick any convenient punctuation character —
it's especially convenient to use punctuation which normally only occurs at
the end of words at the start/inside of words, e.g. ‹. :› before the vowel

Den mån 29 apr. 2019 19:55Aidan Aannestad <[log in to unmask]> skrev:

> This has gotten me thinking about some things, especially in light of
> the fact that I just this morning read a chapter about orthography
> design for natlangs. It's really quite interesting how different the
> concerns are.
> For natlangs, almost the only concern is making sure that native
> speakers can see a word (in context) and recognise it as quickly as
> possible. Exactly how best to do this varies by language (especially in
> regards to tone, which is what I was reading about specifically), and
> the exact decisions of how to represent things may also have to do with
> nearby major languages / areal orthography conventions. (Speaker
> communities' preferences, though, trump all other considerations,
> including readability concerns.)
> None of this matters at all for conlangs, unless you're creating a
> conlang which uses Roman letters in its world-internal setting. For all
> other conlangs, you have a number of competing factors, which may have
> different weights under different circumstances:
> - Ease of typing
> - Usefulness as a guide to pronunciation for linguistically uninformed
> readers (eg of a novel)
> - Usefulness as a guide to pronunciation for linguistically informed
> readers (eg other conlangers - or yourself!)
> - Usefulness as a guide to the underlying mechanics of the language's
> phonology and morphophonemics
> - Usefulness as a representation of the language's own script (if it has
> one)
> - Aesthetic / 'flavour' properties of particular graphemes and
> combinations of graphemes
> Indeed, depending on how you intend to use your conlang, you may want
> more than one romanisation scheme, with each meant for different
> purposes; though it may be possible to do a good job of more than one at
> a time (particularly 1, 2 and 6 above probably go pretty well together).
> ]
> The unnamed polysynthetic language I'm working on is an interesting case
> study, maybe; and I've been thinking about it a bit recently, so I'm
> going to talk about it anyway :P It has six vowels (the canonical five
> and /ɨ/), a complex set of consonants including two voiceless liquids,
> tone with upstep and downstep processes, and a pile of morphophonemic
> things going on under the surface.
> - I can't imagine that a linguistically uninformed reader would too ever
> really get what's going on here, and I don't intend to use it in
> contexts where that matters, so I can set that concern aside.
> - Typing can be a pain. Tone diacritics aren't too bad to type with the
> keyboard I've got, but representing upstep and downstep is possibly
> tricky. Spelling /ɨ/ as <y> solves half of the problems on its own;
> writing /l̥ r̥/ as <lh rh> isn't too bad either. There's not a lot
> stopping me from using <th dh> for /θ ð/, if I want to.
> - Linguistically informed readers can probably guess at /most/ of the
> above, though <y> isn't too clear. I've been writing pretty much
> straight IPA, modulo tones, for posts here though, just to get around
> that kind of an issue.
> - There's no native script (yet), and I'm not super concerned about
> romanisation aesthetics.
> - Morphophonemics is where things get weird. There are not just piles of
> epenthetic vowels, but two different kinds; there's two different
> dissimilation processes; tones can do odd things in combination; and the
> topic marker exists (almost) exclusively as a process where the last
> consonant of a word turns into a voiceless continuant. I've got a
> variety of options, then:
> [mírɨ́tɨ́] 'said' < /mir[HL] /+ /-t
> /Options:
> /mírɨ́tɨ́ /(straight surface IPA)
> /mírýtý /(straight surface romanised)
> /mîryty /(surface segments with phonemic tone pattern; leaves tone
> realisation ambiguous due to differences between underlying /ɨ/ and
> epenthetic /ɨ/)
> /mîrt/ (straight underlying form, completely unindicative if you aren't
> familiar)
> [mírìl̥ɨ̀] 'want to say' < /mir[HL]/ + /-Vr̥
> /Options:
> /mírìl̥ɨ̀/ (straight surface IPA)
> /mírilhy/ (surface with easy romanisation, leaving low tone unmarked)
> /mîrilhy/ (surface segments, phonemic tone)
> /mîr?rh/ (underlying form; ? means I have no idea how to transcribe an
> underlying empty vowel slot)
> [θò] '1sg[TOP]' < /no[L]/ + /[-voice +cont]
> /Options:
> /θò/ (straight surface IPA)
> /tho/ (surface easy romanisation)
> /nho/ (morphophonemic-y spelling more clearly derived from the root;
> causes <th> to be questionable for underived /θ/ and <dh> to mean /θ/ as
> well)
> /ńo/ (morphophonemic-y spelling leaving the root even more clear and
> avoiding problems with <th dh>)
> TL;DR, romanisation is /really hard/ when you have complex
> morphophonemic stuff going on.
> On the subject of fictional historical reconstructions, that seems like
> an unusual subcase of the 'flavour' consideration, where the goal is not
> so much to give text a purely aesthetic flavour, and more about giving
> an extra degree of in-world realism to the reconstructions being discussed.
> On 2019/04/29 11:41, BPJ wrote:
> > The snag is of course that as conlangers we usually know what those
> > romanizations are supposed to stand for, so maybe our non-use of IPA is
> > rather part of the fictional setting where we pretend to be historical
> > linguists who are describing real-world (pre)-historical languages, or at
> > least languages which are real in the worlds where those languages are
> > real, perhaps leaning on a scholarly tradition which is also real in
> those
> > worlds. There may of course be different opinions on how legitimate that
> > behavior is.
> >
> > A more unquestionably legitimate concern is ease of typing, and perhaps
> > ease of doing grammar rather than phone*ics. As Logan Kearsley wrote in
> > another thread recently:
> >
> > LK >And I know that this is not true for some other people, but one thing
> > LK >that is fairly consistent about my process (and has actually stalled
> > LK >progress on many projects I would like to finish at some point) is
> > LK >that I cannot make very much progress without a good romanization
> > LK >system. If I cannot fluently type an unambiguous representation of
> the
> > LK >language, then I cannot work on it, because the details of
> > LK >transcribing it will get in the way of thinking about how it actually
> > LK >works. Usually that means that most of the phonetics / phonology do
> > LK >indeed have to be figured out fairly early on, to make it possible to
> > LK >devise a reasonable romanization system, but it's not necessarily the
> > LK >*first* thing.
> >
> > (It feels good to know that I'm not the only one! :-)
> >
> > It certainly is a factor that I have a key for ‹ä› and dead keys for some
> > of the most usual Latin alphabet diacritics readily available on my
> Swedish
> > Linux keyboard, so that I can readily type the "standard" Sohlob
> > romanization in any application, while I only would type IPA in Vim using
> > my own CXS-to-IPA keymap.
> > On my handheld devices where I use Multiling O Keyboard[^1] switching
> > to/from the IPA mode takes only one touch more than typing a letter with
> > diacritic, so that's hardly it.
> >
> > Some of us wouldn't want a romanization using any diacritical marks, or
> > rather not any requiring any special devices, and for some of them the
> main
> > problem with IPA may well be that it requires special devices, sensu
> lato.
> > In fact I can very much sympathize with that. After I decided that I
> wanted
> > to replace my old Latin-1 based romanization of Sohlob with something
> less
> > weird I literally hesitated for years about how to write the non-Latin
> > consonants. Not only was/am I not a big fan of tacking on a ‹+h› to any
> > consonant to indicate any of a number of modifications, but I already
> had a
> > number of ‹h› + resonant digraphs for voiceless resonants/voiceless
> lateral
> > fricative which I didn't want to give up. But how then distinguish
> between
> > e.g. [xl] and [kɬ]? I indeed considered using ‹hk hg hc hs› etc. but felt
> > that (1) Teonaht has prior art and (2) it would hardly be helpful. I
> ended
> > up with a "multigraded" romanization — or rather four different
> > romanizations for different purposes.[^2] The "narrow" version using tons
> > of diacritical marks is really mostly indulgence of the "imaginary
> > historical linguist" variety. The "broad" version is what I mostly use
> for
> > language description, except that I usually use ‹hl› etc. for the
> voiceless
> > resonants. This means that all the letters really needed to romanize
> Sohlob
> > languages
> >   can be typed on my Swedish Linux keyboard using only precomposed
> > characters. Incidentally the extra consonant letters needed to type
> > Classical Sohlob and Chidilib are the same as Lakhota uses, which is a
> Good
> > Thing because it means there are many fonts which support the set. Now
> I'm
> > even happy to use ‹ȟ› rather than ‹ǩ› for /x/ since I regard [h] as a
> > better potential mispronunciation than [k] or (horror!) [ks].
> >
> > The original reason to exist for the "simplified" version was that I
> > thought that that was what I was going to use for language names and
> proper
> > names in English (and Swedish) text and in narrative fiction, should I
> > write any again. In fact I'm using the "ASCII" version (potentially plus
> > the middle dot) for names in English text, even though _Hïlïb_ probably
> > would stand a better chance than _Hielieb_ of being pronounced close to
> > [hɨlɨb̥].  Such is the power of the notion that English doesn't use
> > diacritical marks — which actually isn't quite true anyway!
> >
> > My opinion on digraphs has actually loosened up considerably over the
> last
> > your or two, especially since I realized that if used in a consistent and
> > principled way letters and punctuation characters can in fact be
> > diacritics. I came to that realization when it hit me that that caron ~
> ‹h›
> > and umlaut ~ ‹e› are actually freely interchangeable in the current
> > romanization of Sohlob, except that ‹ȟ› is preferred to ‹ǩ› and ‹h› is
> > omitted after ‹j› to the extent that I use ‹ǰ› at all.
> > As you can see my "ASCII" version even uses ‹r› as a diacritic for
> > "retracted", and I've even used ‹td› for dental _t_ in a file name! In
> fact
> > I'm currently mulling on what other letters than ‹+h +y/+j +e +r/r+ h+›
> may
> > have an "obvious" diacritic use. Doubling may also be regarded as a
> > diacritic when used in a principled way.
> >
> >
> > [^1]:
> >
> > [^2]:
> >
> > Den sön 28 apr. 2019 21:50Jörg Rhiemeier <[log in to unmask]> skrev:
> >
> >> Hallo conlangers!
> >>
> >> On 22/04/19 18:28, BPJ made the following digression in a longer post:
> >>
> >>> {begin digression}
> >>> Indeed one of my reasons for not using IPA is that I feel that using
> it,
> >>> especially between slashes or brackets is or may look like a stronger
> >>> commitment to phonemehood or a particular phonetic value than I want to
> >>> make at least at this point. E.g. using _ȟ_ skirts the question of
> which
> >>> Sohlob varieties, communities etc. have [x] or [χ] at any point in
> time,
> >>> space or the speech stream. (Also I happen to think that e.g. [pahad]
> is
> >> a
> >>> better mispronunciation of [paxad] than [paksad] or [pakad], not least
> >>> because that is how an articulatorically unskilled Chidilib or Linjieb
> >>> speaker would render it!)
> >>> {end digression}
> >> Right! IPA is a *phonetic alphabet*, it is made for discussing phonetic
> >> values. Reconstructed languages at most have *phonemes*. Using IPA
> >> creates the impression that one can nail down the sounds written that
> >> way to some more or less exact phonetic value, which usually is unknown
> >> and often unknowable in a reconstructed language. Also, IPA is a *big*
> >> chest of letters, and with most languages, you won't need more than a
> >> few of them, and it is often more convenient to use some other alphabet
> >> which does not have so many superfluous letters.
> >>
> >> This is, for instance, part of the reason why Indo-Europeanists don't
> >> transcribe PIE in IPA. (The bigger reason of course is that there is a
> >> long-standing tradition of transcribing the PIE phonemes as they are
> >> done, and nobody feels like changing it more than necessary to
> >> accommodate newer insights.) Consider the "laryngeals". Just what are
> >> they? If you transcribe them as, say, *h, *χ, *χʷ, you have cast a vote
> >> in the endless debate about their phonetic values even if that wasn't
> >> your intention, and people who maintain a different position will attack
> >> you. So everybody just uses *h1, *h2, *h3, which says just what is known
> >> about them, namely "three vaguely h-like sounds which differ from each
> >> other in some unknown way". The same thing holds, of course, for other
> >> language families such as Uralic, Semitic, Sino-Tibetan or Austronesian.
> >>
> >> In a similar way, ancient written languages are not transcribed in IPA,
> >> either; rather, linguists render the *graphemes* of the language - which
> >> is the *only* thing we know for sure about them! - in a convenient
> >> alphabet, and IPA usually is a bad choice. That's, for example, why you
> >> don't see vowel letters in transcriptions of languages like Egyptian or
> >> Phoenician - the ancient Egyptians and Phoenicians did not write the
> >> vowels, either, so we can't be sure about them.
> ---
> このEメールはアバスト アンチウイルスによりウイルススキャンされています。