While I'm at it, why does camxes think "djoan" is invalid? "djo'an" is accepted, but "Johann" and "Joanne" are quite different. There's no chance of "djoan" breaking up into "djo an", since "an" requires a leading pause. "djoan" is accepted by jbofi'e.

You are delving into the murky areas of the phonotactics. For some of your questions there are no definite official answers at this point.

The usual take has been that "ndj" and the others are not allowed in cmevla, despite there being an example in CLL ("djandjonz" if I recall correctly). Personally I find the prohibition of ndj/ndz/ntc/nts in general doesn't make much sense, I would remove it for all words, but I don't really see a reason to treat cmevla differently from other words at the phonotactics level.

Stress in cmevla and cmavo is free, so it doesn't really matter how you count syllables. My interpretation would be that "iy" and "uy" work as Cy syllables and they would not normally carry stress. You can stress any number of syllables in cmevla and cmavo, even none. Two consecutive stressed syllables are very hard to detect as being stressed though, since stress is really a contrast between adjacent syllables. In any case, stress is only relevant for brivla.

There is no way to indicate that a cmevla "should" have no stress, but then that is never the case, since stress is not a relevant feature for cmevla. There is no way to indicate that your preferred pronunciation is without stress. Similarly there are many other features for which there is no way to indicate one's preference, for example tone.

>>1459

The reason camxes doesn't like "djoan" is that it doesn't allow syllables without an onset (as would be ",an" in this case. Similarly it would reject "djaan" or "djean". "djouan" would be OK.

>>1460

> I don't really see a reason to treat cmevla differently from other words at the phonotactics level.

Cmevla are already treated differently from other words at the phonotactics level, in that they're allowed to have non-initial consonant clusters at the beginning, and have no constraints on consonant triples or larger clusters. (Not to mention allowing 'iy' and 'uy'.) I believe the intent is to place as few restraints as possible on names imported from other languages. If I understand the CLL correctly, fu'ivla aren't supposed to have any constraints on internal consonant clusters either. Only lujvo get forced into consonant triples where the second pair must form an initial pair, and ndj|ndz|ntc|nts are forbidden. At least, that's what I think the CLL says. I agree that the prohibition against those four triples doesn't seem to make any sense.

>>1461

... and why doesn't camxes allow syllables without an onset? I don't see any reason for that restriction in cmene or fu'ivla.

Is there a version of camxes that shows the morphology output? The java version only outputs the syntax tree, with full words as leaf nodes.

>>1462

It seems to me that if a rule like "ndj" is not to be enforced for cmevla, then there's no point in enforcing "kb" either. Either allow any combination of phonemes, or maintain the same constraints as for the rest of the language. (The case of iy and uy is slightly different because the reason it doesn't appear in fu'ivla is morphological, for word determination purposes, not strictly phonological.)

The case of initial clusters is indeed something I have a problem with too. My personal "solution" is to say that cmevla can have an ".y" added in front, and then this added sylable can take care of the non-initial part of the cluster.

What camxes does is separate cmevla into two classes, which I call "jbocme" and "zifcme". "jbocme" are fully assimilated names, and they must consist of pure lojbanic syllables. "zifcme" contain all the rest. Perhaps "oa" could be allowed in zifcme too.

As for why the onset requirement, this has to do with the history of Lojban. In Loglan, the combinations aa, ae, ao, ea, ee, eo, oa, oe, oo are all valid, Loglan doesn't have the apostrophe. Personally I wouldn't have a problem with these, all of them except for aa occur in Spanish, so they are very natural to me. But Lojban introduced the apostrophe precisely to split these appart. It seems a bit perverse to allow them again in fu'ivla when the only reason Lojban has the apostrophe is to prevent these sequences.

I don't know about camxes implementations. I normally check with the formal grammar itself, although I understand it can be a bit daunting if you are not very familiar with it:
http://www.lojban.org/tiki/BPFK+Section:+PEG+Morphology+Algorithm

>>1463

You make a good point about phonological vs morphological rules. In my head, I'd been dividing the rules into pair vs triple rules instead. However, your "solution" to the "problem" that initial consonant clusters are relatively unconstrained in cmevla seems somewhat silly to me. Is it really a problem if phonological as well as morphological rules are relaxed for cmevla? Speakers can already insert buffer vowels between any two consonants, effectively creating another syllable in their pronunciation. The stricter phonological rules in gismu and lujvo are there for the convenience of speakers who have difficulty with certain consonant clusters, yes? Cmevla, on the other hand, and fu'ivla, to a certain degree, should be allowed greater phonological flexibility, so that the original word can be recognizable. I don't really understand why there's a division in camxes between "lojbanic" names and "non-lojbanic" names, when it would be simpler to have a division between "name-words" and "non-name-words".

It seems a bit strange to accept "li,on" ("Leon") but not "djo,an". If I'm reading the morphology correctly, camxes completely ignores commas, and thinks "li,on" is a one-syllable word. Shouldn't both be treated as two-syllable words?

>>1464

>> The stricter phonological rules in gismu and lujvo are there for the convenience of speakers who have difficulty with certain consonant clusters, yes?

It's hard to say, some of the rules seem just random (whyever fobid "mz"?) The rules are a mixture of ease of pronunciation for the speaker, ease of differentiation for the listener, and some randomness thrown in. But "ease" is very relative to someone's native language and natural ability, so we might just as well treat the rules as just arbitratry. Some are reasonably general ("no double consonants") some are unreasonably particular ("mz").

>> Cmevla, on the other hand, and fu'ivla, to a certain degree, should be allowed greater phonological flexibility, so that the original word can be recognizable.

Personally, I disagree. I think that when you fully lojbanize a word, you have to adapt it to Lojban's phonotactic rules, not just approximate it, otherwise it remains somewhat foreign. But the position that borrowed words need to be kept as true to origin as possible rather than as true to Lojban as possible is also there. For a recognizable original, there's always the foreign quoting method with ZOI.

>>I don't really understand why there's a division in camxes between "lojbanic" names and "non-lojbanic" names, when it would be simpler to have a division between "name-words" and "non-name-words".

jbocme and zifcme together constitute cmevla, so you don't really need to be aware of the division. I put it there simply for my peace of mind, because I wanted a class of what I consider truly lojbanized names, consisting purely of lojbanic syllables, while still catering to the more official tradition. I still find it unreasonable that "ktktptktis" could be considered more acceptable than, say, "akbar". The jbocme/zifcme division doesn't really play any grammatical role.

>>It seems a bit strange to accept "li,on" ("Leon") but not "djo,an".

Since both "l" and "n" can be syllabic, "lion" can be syllabified in many different ways: "lion", "l,ion", "li,on", "lio,n", "l,i,on", "l,io,n", "li,o,n" or "l,i,o,n". It's your choice how many syllables you use for the actual realization, but one syllable would be the underlying form. Saying it as "li,on" runs the risk of confusing it with "liion", which does have two syllables, but any syllabification is allowed.

Some people may want to argue that "lion" and "li,on" are two different words, rather than the same word. It makes very little difference in practice. The only place where it matters would be when using it as the opening and closing word of ZOI. It's probably not a good idea in that case to rely on them as being different words (so that one can be used to quote the other) nor to rely on them as being the same word (so that one form can be used as the opening and the other form as the closing word).

>>1465

If every syllable must have an onset, then are there now some situations where words beginning with a vowel do not require a pause in front? E.g., "la dinko a la mruli"?

>>1466

The glottal stop is one of the possible onsets, so ".a" does have an onset. Again, it wouldn't make sense to allow "oa" between words without an intervening glottal stop if "oa" was considered so potentially problematic that the apostrophe had to be introduced.

(Personally, I don't think a glottal stop is required in front of an i or u onset, but you probably won't find many people that agree with me about that.)

Is there a good list of which ones are disallowed? I had thought that "la singpolymas." would be a valid lojban rendering, but the jbofihe app hates it

>>1468

See: http://jbotcan.org/docs/cll/c3/s6.html

The problem with singpolymas is the "gp" combination, because "g" is voiced and "p" unvoiced.

The reason that clusters like ndj and nts are forbidden is they are highly confusable to the ear with nj and ns, for example, and it interferes with audio-visual isomorphism.

lenu cusku zoi ler ndj ler kei cu simsa lenu cusku zoi ler nj ler .iri'abo lo tance cu zvati le selzva be me'o dy

>>1474

But consider these analogous pairs:

bdj / bj
gdj / gj
ldj / lj
mdj / mj
rdj / rj
vdj / vj

None of those triples is forbidden, so unless ndj/nj is more highly confusable than all of those, for a majority of the world population, the restriction for "ndj" seems just arbitrary.

I would consider most of those to be considerably less ambiguous because the place of articulation is different enough between the first and second consonants, whereas with ndj and nts the n and d/t are in the same spot, which causes there to be a slight d/t sound even when that phoneme isn't actually there, at least to my ears. R also has that problem because for me it has an alveolar articulation (I natively have an alveolar approximant, and speak Lojban with an alveolar tap), but less so than N, which is itself more or less just D with a vibrating nose.

>>1477
Seconded. For the same reason, words like “pinxe” can be problematic when you're in the habit of pronouncing n as /ŋ/ before velar consonants. It comes out as “pinkxe”.

>>1477

OK, so why "n" but not "r" and "l", which are also alveolar? If the issue is point of articulation, then lts, ltc, ldz, ldj and rts, rtc, rdz, rdj should be as impermissible as nts, ntc, ndz, ndj.

And what about ntr, ndr?

>>1479

Well, as I said, for R I do kind of have a similar problem. I think L is generally lateral enough that it sticks out more from the D/T, and I think R still works because as an approximant I have no problem, and as a tap (which is what I use for Lojban) I hear one 'D'-ish sound if it's just rj and two of them if it's rdj.

I think the issue might come down to the fact that r and l are pure liquids, whereas N is literally really just a nasalized D.

Maybe it's worse for us lax-mouthed native anglophones so this could be me malglicoing it up, but as an example for me "France" and "frants" are only distinguished by really way over-emphasizing the T in the second one, like "fran-t-s," which I wouldn't really think to do in normal discourse, and which borders on using the vowel buffer rule just to be able to distinguish.

I tell you though, if you want to get into stupid banned consonant clusters, why is mz banned? I can't really see how that's too confusable with nz if words like "finpe" are allowed without "mp" being banned, let alone both "ns" and "ms," but again maybe I'm biased because of English's many -s endings that make one have to carefully distinguish.

Oh I almost forgot, for ntr and ndr, it's exactly for the reason that all of those consonants are alveolar that it's not ambiguous. It's because you have to move your tongue from N to J that the intrusive alveolar stop leaks in; here, you never move your tongue from the alveolar ridge except to strike it, so it doesn't happen (or not as much if anyone out there indeed does have trouble with those clusters).

>>1480

The ntc/nc distinction shouldn't be especially difficult for English speakers, since it occurs in English (I don't suppose English speakers would confuse "gun shop" and "gun chop" for example). France/frants shouldn't really matter because in Lojban a vowel has to follow (cmevla aside). The problem might be that "nj" doesn't occur in English (as far as I know), so it might be difficult for English speakers to distinguish it from "ndj", but it still seems pretty arbitrary to me.

I agree that "mz" is the worst as far as aribitrariness goes. My grading for least to most arbitrary banning is:

1- double consonants
2- sibilant-sibilant pairs
3- voiced-unvoiced, unvoiced-voiced pairs
4- xc, cx, xk, kx
5- nts, ntc, ndz, ndj
6- mz

>>1479
I think it's the difficulty of releasing a stop without emitting a plosive, or at least without transforming a following fricative at the same place of articulation into an affricate.

As for ntr, ndr, at least in my American English dialect, the movement is sufficiently different from that of releasing a stop that there is no confusion—my tongue slides back over the roof of my mouth.

>>1482

> I don't suppose English speakers would confuse "gun shop" and "gun chop" for example

I think English speakers may pronounce “chop” orally, but “shop” nasally. Pronouncing “nj” nasally makes it much easier for me not to produce an affricate.

>>1482

The one problem I can perceive here is that these, if you change them to Lojban morphology, might be closer to gyn. cap. and gyn. tcap. than GYNcap. and GYNtcap. When I say either phrase I insert a glottal stop before the fricative/affricate which clearly differentiates if there's a T-sound associated with the second word.

My list would look almost exactly the same, except I'd swap #3 and #5 (probably a given, from the topic of our discussion).

>>1484

A glottal stop there? That doesn't sound very Englishy to me. Think of "gunship" and a potential "gunchip" if you prefer. Wouldn't English speakers distinguish those easily? The point is that both Lojban "nc" and "ntc" occur in English, but whereas "ndj" also occurs, "nj" doesn't.

The reason I put (3) as less arbitrary than (5) is not because the combinations may be easier or more difficult to produce or distinguish, that will vary wildly depending on one's native language, but because (3) involves two whole classes of phonemes and (5) a single one "n" (Or a single-member class.)

The reason I put mz last is because it involves just two particular phonemes, and only in one particular order, since the order zm is accepted.

I'm not completely sure whether the x-rules are more or less arbitrary than the n-rules. I can sort of justify xc and cx by some kind of extension to the sibilant-sibilant ban (but then how come xs, sx, xf, fx, fs, sf, vz, zv are ok?) but I can't really justify xk, kx except as "difficult for English speakers". xk and kx should be treated like fp, pf, vb, bv, st, ts, zd, dz.

The phonotactics of natural languages often have gaps and some unsystematicity, but there's no reason for the phonotactics of Lojban not to be as systematic as possible.

>>1485

> xk and kx should be treated like fp, pf, vb, bv, st, ts, zd, dz

Seconded. If we have ct and tc, xk and kx should logically be permissible as well. They may fall in the "hard for English speakers" category, but not any moreso than x itself, I should think.

Also strange is the omission of a voiced counterpart for x and the impermissability of initial pairs like mb, nk, and all the variants on that theme.

>>1486

> Also strange is the omission of a voiced counterpart for x

Some people have proposed using "q" for that. It's probably too late for that though.

> and the impermissability of initial pairs like mb, nk, and all the variants on that theme.

Those don't really fall into the pattern of permissible initials, which is:

(s,c,z,j)(p,b,t,d,k,g,f,v,x,m,n)(l,r)

plus the affricates (and with several exceptions).

>>1487
I can't say I understand why the pattern is as it is. Certainly too late to change it though.

>>1488

The pattern is greatly influenced by English.

The permissible initials in Lojban are mostly those of English with some generalization. The restrictions to the pattern are all things that don't occur in English initially: sx, cx, zn, jn, zl, jl, zr, jr, tl, dl, nl, nr

The ones that don't occur in English are basically those with c, z, j in the first spot where English only has s, plus vl, vr, xl, xr, ml, mr, and the affricates ts, dz.

>>1485

It's very English-y, just like how all syllable-final consonants in English have a glottal stop before them as well. It's a pseudo-secret of English phonology that, like that the Germanic language it is, it's full of glottal stops that speakers don't even perceive, as you may or may not have already known.

I decided to test these words, "gunship" and "gunchip" because I figured maybe I'm biased. So I went up to my roommate and said, slowly, "Define 'gunchip,'" and without any hesitation he said "I guess a ship with guns on it." When I said "no, it's with a CH," he actually said to me "come on, there's like no difference there," and I mentioned France-frants etc. and he agreed. Now maybe that's just us, and obviously that's just a crappy little anecdote that, as far as you know, I could have theoeretically pulled out of my ass completely, but I still figured I'd at least mention it.

>>1486

Agreed, based on that xk and kx should definitely be allowed. I can't see any reason for their being banned other than the fact that for native English speakers (i.e. the vast majority of Lojban speakers) it's (slightly) hard to differentiate, though personally I've never understood how anyone has trouble with /x/, since it's never presented a problem for me.

>>1490

OK, I accept I may be wrong about the nc/ntc distinction in English then, I'm certainly not an expert on English phonetics. So "unshare" and "unchair" are homophones?

In any case, they are not homophones in (my variety of) Spanish, which happens to have a sh sound. For example "manyar" (slang word for "to eat", from Italian "mangiare") vs "manchar" ("to stain") differ exactly in nc/ntc and don't sound the same.

>>1458

".akybar." would work, I think.

What if you capitalize a "y", does the stress-indicating shift override the skipping rule? For example, is ".mumynic." different from ".mumYnic."?

>>1458

What if my name is Myrtle and I want to translate it as ".mrtl."- where does the stress fall? Would there be none? Or would I HAVE to capitalize the initial syllabic consonant as ".mRtl."?

>>1463

"kb" cannot be allowed because it is stringing an unvoiced consonant with an adjacent voiced consonant. "ndj" has no such restrictions as nasals (although typically voiced) can be either unvoiced or voiced and the other two are voiced.

>>1468

Remove the "g", since "n" can operate both as an alveolar nasal or a velar nasal. I disagree with this (I have always agreed with those who advocate the use of eng or another letter representing it for Lojban), but the official rules say that such interpretation is allowed.

>>1483

My "ch" actually occur in a totally different region of the mouth from my "sh" (as do "dj" from "zh"), although I understand why Lojban represents "ch" as "tc".