Return-Path: Message-Id: <9110240902.AA11539@relay1.UU.NET> Date: Thu Oct 24 15:58:31 1991 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: more response to And X-To: lojban@cuvmb.cc.columbia.edu To: John Cowan , Eric Raymond , Eric Tiedemann Status: RO X-From-Space-Date: Thu Oct 24 15:58:31 1991 X-From-Space-Address: cbmvax!uunet!CUVMA.BITNET!LOJBAN You write: >Did you consider that _kandi_ requires 25% more memory, effort in >learning, or whatever, than _kanda_, with final vowel copied from the >stem, and then decide that the _di_ calling to mind _dim_ (!!) was a >greater asset to memory? Did >I< consider it? No. Did JCB do so when inventing the language? Who knows, but I doubt it. We did not consider a change in the basic morphology or word-making algorithm to be within our charter in redesigning Loglan as Lojban. JCB never has recorded exactly what was considered and rejected and why in originally developing same (for which I've criticized him, but this is spilt milk). JCB claims to have done 'engineering tests' with several different word making strategies, with his students at U. of Florida. The results are unpublished and his discussion of this sounds like he did not even retain notes of the experiments and methodology, much less consider publishing them. >I did indeed intend it mainly just as a different way of looking at >current Lojban, but unless I am taught (by teacher or experience) to >pronounce /-nb-/ as /-nVb-/ it will come out as /-mb-/. /tcVr/ and >/tVcr/ may well not be distinct unless buffered. I've never been to Canberra Australia, but would never pronounce it Camberra, nor do I feel a need to buffer it. Now I MIGHT mishear one as the other, but I'm not sure. Remembering that Lojabn words end in a vowel, as do most Lojban syllables, I ask you to compare "matcVri" and "matVcri" for each of the vowels - I don't have any problems. Now I can see that British and some American speakers might have trouble with any vowel before "r" in an unstressed syllable, but that is a different problem, and one we emphasize in teaching pronunciation: no schwa or centralizing for non-descript unstressed vowels. Can this be taught successfully? I don't know. >This level of phonology isn't pragmatic. Once you are fluent, you will >always use the same buffer vowel. Phonological processes are done >unconsciously (although I can deliberately start speaking with, say, a Probably true, but I am talking about a situation where you are talking to someone in either a difficult environment, or to someone you don't normally talk to who is evidencing some misunderstanding of your normal speech. In such a case, most people make some adjustments in enunciation and emphasis, and in Lojban, this will include buffering adjustments. I suspect there is no totally universal phonology of this many sounds that will not confuse someone of some culture. > [buffering] another factor in buffering is length. The buffer sound will be MUCH shorter than any vowel, including the hyphen schwa, which itself is normally shorter duration than other vowels in my speech. (I think Lojban will eventually evolve towards being syllable-timed rather than stress-timed as is English, which will aid this. But thus far, all the poets have been English speakers and people trying to translate English songs whose rhythm is stress-timed, thus fighting this trend.) >It's not just the buffer vowel that makes the phonology wobbly: other iffy >elements include: > >(a) /x/ contrasting with /h/ when adjacent to /i u o/ remember that Lojban "h", the apostrophe, is a devoiced vowel glide - it should never be velarized or even particularly fricative. Calling the thing an "h" sound is to make it easier for English speaking non-linguists, who hear "h" (and are never satisfied by our explanation of why we use apostrophe instead of "h" for the sound) >(b) /e/ versus /ei/ and /o/ versus /ou/. I can perfectly well hear the >difference between _fed_ versus _fade_, and when I speak Italian, Italians >can tell whether I'm saying /se/ or /sei/, but when I hear Italians, I >can't tell whether they're saying /se/ or /sei/. The former is why we now teach Lojban 'e' as "fed". To an Italian who is having trouble being understood, we might even teach as the vowel of "fad", since an Italian will not be making an /a/ anywhere near that sound. /ou/ is not standard to Lojban without the apostrophe. What people do in Lojbanizing names is at their own risk. >(c) contrasts between a sonorant (glide, liquid, nasal) in syllable nucleus >and syllable onset, in particular: > >(d) /ii/ v. /i/; /uu/ v. /u/. The divowels are not used in Lojban except in names (normal caveat) and in VV attitudinals where there is usually a glottal stop preceding and the pair is pronounced as a semi-vowel/vowel. This criticism was one of the things that led to our change from JCB's Loglan versions, in adding the devoiced glide apostrophe in nearly all VV combinations. >(e) /s z/ versus /c j/ before /i/. I can see some language speakers (japanese?) having problems here, but then Chinese has even more phoneme density in the fricatives, so we had to compromise somewhere. I actually have more problems with tongue twister words with both /c/ and /s/ in them (or adjacent ones: we have a lot of "she sells sea shells" in Lojban, and this IS a problem). >(f) possibly the following non-salient contrasts are forbidden or don't arise: > /mps/ v. /ms/ /mbz/ v. /mz/ /mpt/ v. /mt/ /mbd/ v. /md/ > /mpc/ v. /mc/ /mbj/ v. /mj/ /nts/ v. /ns/ /ndz/ v. /nz/ > /ntc/ v. /nc/ /ndj/ v. /nj/ /tcr/ v. /tr/ /djr/ v. /dr/ Almost all of these triplets are explicitly forbidden by the morphology. /tcr/ is not. But note that the normal syllabification of a triplet is C/CC, and the CC must be permissible initial (a very restricted set). The main risk here is with people for whom an affricate like /tc/ or /ts/ is considered a single phoneme in the native language. The solution is to syllabify between the two phonemes comprising the affricate in a triplet. >(g) There are a number of further problems that couldn't be avoided without a >drastic reduction of the segmental inventory: > > /r/ versus /l/ > /b/ versus /v/ > /p t k/ versus /b d g/ In gismu all of the above are never in sole opposition. In lujvo there is a designed workaround (use longer forms) that avoids sole oppositions when needed. This leaves only cmavo as places where confusion can occur on the basis of single consonants, and we tried to ensure some grammatical or usage clues in these, where possible, to minimize confusion. >But given that Lojbanists are keener on getting the language spoken than >perfected, my criticisms do not of course warrant changing the system. After 36 years, isn't it about time to see if it works rather than to endlessly theorize how it might be better when there is so little data on speech problems in planned languages (or second languages in general). Lojban is a scientific tool. We will learn from those who speak it. My complaint with conlang developers who do not intend their language to be spoken is that there truly is no way to evaluate the result aesthetically or linguistically, except to apply theories that aren't all that solid in the irregular NL world. >We'll each speak with a different accent, with slightly different >phonemic inventories, be consequently exposed to a confusing signal, but >probably understand each other perfectly well by relying on (pragmatic) >context: British and Americans speaking English understand each other >okay, and I don't think Lojban would be worse. I understand that >J.C.Brown invented the phonology, for which I cannot think he is to be >thanked; his claim for audiovisual isomorphism and an unambiguous signal >(for Loglan, but I think its phonology is pretty much the same) are >spurious. I've heard data that suggests that even in optimal conditions, our hearing is so inaccurate that we may miss half the phonemes, and reconstruct the rest thru context. A computer won't miss phonemes, but will have different problems. So the question "unambiguous to observer ___?" is in order before evaluating. Lojban has made changes from JCB's original design, and indeed JCB has made his own changes, not always for the better. Lojban is still algorithmically audiovisually isomorphic, although JCB's current language is not. In pragmatic situations, we will not find out until we have fluent speakers. In any event though, I think Lojban will turn out to be closer to the ideal than most or all NLs or other ALs that are actually spoken by fluent speakers. >I think it's reasonable to conclude (4) on the basis of (1-3): > (1) Lojban values a robust and homophone-less signal. > (2) Most people can learn a non-native phonology only imperfectly. > (3) All or most Lojban speakers will be non-native. > (4) Therefore the phonology should be resistant to errors made by non-natives. > >I think Lojban has tried for (4) but not done nearly as well as it might >have - probably because it was invented by largely monolingual English >speakers. Still, no-one is attracted to Lojban for its phonology - it's >the unambiguous syntax & the semantics that are its real assets. So >long as Lojban doesn't make exaggerated claims about the phonology and >excessive prescriptions for it, I shouldn't criticize (& will therefore >endeavour to gag myself on the subject henceforth). The unambiguous morphology also counts for something. JCB has emphasized the phonology, which has raised expectations. But then almost every language instruction program tends to start with pronunciation, thereby raising expectations and dashing hoped when people realize after years of study that they can never truly sound like a native. But an AL has no native speakers to hold up as a standard or ideal. I try to de-emphasize pronunciation with beginners, presuming that long before people become fluent in the grammar and vocabulary, they will have developed an accent that is understandable to others. I should note in passing that I have a totally different accent in my bits of fluent Lojban speech than in my normal pedagogical speaking style. I discovered this while working with Logflash. When I said the words fast as they flashed by, I started developing what was even to me a noticeable eastern European voice quality. I seldom achieve this speed even in fast reading of text because the language is still clumsy for me, but the faster I go, the more Slavic I sound, even to me. I should then note that Slavic language have many more consonant cluster than Lojban or English, and yet are well understood. The bottom line: let's find out. We will avoid the exaggerated claims for the most part, and if anything you are accusing us of insufficient prescription on vowels and buffers, so that seems to be a non-problem. But please do not gag yourself should you have new ideas or insights as you learn more, or of course as others after you raise the same questions over and over. lojbab@grebyn.com