1.13 The 25 Vowel-Pairs: Loglan abounds in vowel-pairs. Not being allowed to enter the syllables of regular words, schwa or /y/ does not figure in any of them. But all possible pairings of the other five vowels /a e i o u/ occur in some words of the language. This creates the 25 vowel-pairs shown in Table 1, which provides a “phonological map” of their pronunciations. The upper entries in each cell of the table are the monosyllabic pronunciations of that particular pair if such a pronunciation is permitted. A ‘—’ is entered on the first line if it is not. The lower entries in each cell are that pair’s pronunciation as a disyllable, again if that is permitted.
Notice that some pairs in the table may be pronounced both monosyllabically and disyllabically. Thus /ii/ may be (yee) or (ee-ee). Others, like /aa/, may only be pronounced disyllabically. Still others, like /ai/, apparently occur only monosyllabically. That is, /ai/ may be pronounced only as the (igh) of hai and sai and never as *(ah-ee). Incidentally, these two Loglan words are identical in sound to English 'high' and 'sigh'.
In using this table, try to pronounce the pairs with hyphenated guides as two distinct syllables with a "glide" between. It is easier to do this if you stress one of the two syllables relative to the other. While both stress patterns are possible, stressing the first syllable in a vowel-pair is a bit more common in Loglan. As an anglophone, you are likely to find stressing the first syllable a bit more "natural" as well.
We observe that the always-monosyllabic pairs are the four "natural" diphthongs /ai ao ei oi/. These can easily be pronounced monosyllabically as (igh ow ay oy) in any context; and so they are.
The next group are the eleven always-disyllabic pairs. These are /aa ae au ea ee eo eu oa oe oo ou/. Except for /au/ and /ou/ these pairs cannot be pronounced as diphthongs; and so they never are. /au/ and /ou/ are included among the pairs that are always disyllables to prevent their monosyllabic pronunciations from being confused with one another and with the natural diphthong /ao/ = (ow). So /au/ and /ou/ are always pronounced disyllabically in Loglan, that is as (aa-oo) and (oh-oo) (or (ah-oo) and (oh-oo) if the Germanic /a/ of ‘father’ is preferred); and /ao/ is always pronounced (ow). This keeps them acoustically quite separate from one another.
The remaining ten pairs are the optional disyllables and are discussed separately in the next section.
1.14 The 10 Optionally Disyllabic Vowel-Pairs: Notice that the 10 vowel-pairs which have both monosyllabic and disyllabic pronunciations in the Table 1 are all those that begin with either /i-/ (ia ie ii io iu) or /u-/ (ua ue ui uo uu). (In English, we would think of them as the 'y'- and 'w'-words, respectively.) They are therefore quite easy to remember. In certain contexts these optional disyllables are all easily pronounced as monosyllables. For example when they stand alone they are always given the values (yah yeh yee yoh yoo) and (wah weh wee woh woo). But there exist contexts in which these vowel-pairs are difficult to pronounce monosyllabically. Following /r l m n/ seems to be four of these contexts. So in these four contexts, and perhaps some others not yet identified, the loglanist is free to pronounce the i- and u-words either as monosyllables or as disyllables, that is, as (ee-ah) and (oo-ah), etc. Thus, the Loglan word lui ('to please…') may be pronounced either as the English name 'Louie' (LOO-ee) or as the French name 'Louis' (l'wee); and which is chosen is largely a matter of personal preference. The word will resolve as lui either way. In the Gainesville dialect, the swift French monosyllables seem currently to be preferred.
Uncertainty about the pronunciation of the optional disyllables reflects the fact that The Institute does not know yet how loglanists of a sufficiently wide variety of native-language backgrounds are going to treat them. So we will wait to describe a general pronunciation pattern here, preferring to observe the speech of active loglanists a little longer. In the meantime, no morphological confusion is generated by keeping these options open.
MRH: The treatment of vowel pairs is good; my parser prefers the monosyllabic pronunciation of these syllables, but will accept the other if it is forced. I believe that while the monosyllable pronunciation is occasionally forced by the context, the disyllabic one can only be forced by explicit hyphenation.
1.15 The Pair-from-the-Left-Rule: The default rule for grouping a written string of vowels into pairs is to start pairing them from the left and to restart the pairing process at any marked pause or syllable-break encountered. Once this is done we have to examine the pairs so made to see whether they are obligatory monosyllables, obligatory disyllables, or optional disyllables. For example, suppose we encounter the written word [aiuiaoea], which is an implausibly long “compound attitudinal”; see Lexeme UI. How do we pronounce it? The first part of the answer is given by the pairing rule. Starting from the left we can try first to syllabify the string as ?/ai,ui,ao,ea/. (We don't know yet whether this is a correct transcription of this new word, so we mark it with leading [?].) We now see that /ui/ is an optional disyllable, and so may be pronounced either (wee) or (oo-ee), and that /ea/ is an obligate disyllable that must be pronounced either (eigh-aa) or (eigh-ah) depending on our personal choice between Romance or Germanic /a/. The other two pairs are the obligate monosyllables (igh) and (ow). We choose to pronounce /ui/ as (wee). This gives us our second transcription, namely ?/ai,ui,ao,e,a/. But what about stress? We decide to use penultimate stress for this new word (which we probably haven't heard yet), for this is another good default rule; see Sec. 2.37. So we produce ?/ai,ui,aoEa/ as our next transcription, this time exploiting the economizing convention that wherever there is a stress-rise or -fall in a transcription, a syllable break is inferrable. Is this the most economical way we can transcribe the pronunciation (igh-wee-ow-EIGH-aa) phonemically? No; it is not. The very pair-from-the-left rule we have been studying allows us to remove the two remaining close-commas. Thus /aiuiaoEa/ transcribes the pronunciation (igh-wee-ow-EIGH-aa) accurately and compactly. It does so because we can infer that there are syllable breaks between the vowel-pairs as counted from the left.
What if someone elects to exercise da's privilege of pronouncing the optional disyllable [u] in the word [aiuiaoea] disyllabically? This rather odd pronunciation— odd because this [ui] is not in the difficult context of leading [l r m n] in which we might expect some readers to prefer the disyllable—can nevertheless be shown by the insertion of two close-commas, one on either side of the second [i]: /aiu,i,aoEa/. This does indeed give (igh-oo-ee-ow-EIGH-aa). One close-comma would not be sufficient here; the pairing rule starts over again with any pause or inferrable syllable break. As a consequence, the transcription /aiu,iaoEa/ would have to be pronounced (igh-oo-yah-oh-EIGH-aa). The interested reader is invited to work out other details of this transcription system.
MRH: this rule is usually superseded by a rather different rule under my parser. The rule given here is applied in parsing compound attitudinals. I also note that JCB himself does not use this rule in names. Where long strings of vowels are found in predicates or names and explicit hyphens are not supplied, one works from the left and chooses according to a hierarchy: choose an obligatory monosyllable if it is present; otherwise choose a single vowel if it is followed by an obligatory monosyllable (which will then be chosen in its turn; otherwise choose an optional disyllable pronounced monosyllabically; otherwise choose a single vowel. A different grouping can be enforced if desired using hyphens (instead of close-commas).
1.16 Indications of Syllabicity: The novice is of course not expected to remember which pairs of vowel letters in a given string are to be pronounced monosyllabically, or to carry out all the above inferences. So at first we need to tell da. In the pronunciation guides the "syllabicity" of a vowel-pair—i.e., whether it is to be pronounced as a monosyllable or as a disyllable—is always plain. All syllable-joints within a breathgroup are indicated by hyphens [-]. So if a pair of vowels is shown with a hyphenated joint, it's a disyllable; if it is not, it's a monosyllable. Thus, (LOO-ee) and (lwee) are plainly the disyllabic and monosyllabic productions, respectively, of lui. (They are also the pronunciations of English 'Louie' and French 'Louis'.)
But in phonemic transcriptions, where syllable structure has been played-down so as to lift other features of Loglan speech into prominence, we have seen that certain inferences may be drawn from the transcriptions themselves that will satisfactorily distinguish disyllabic from monosyllabic productions of all vowel-pairs. By way of summary, these inferences proceed as follows.
First, if the two vowels of some pair are shown as having different levels of stress—as they are shown to have in both /eAmuSUCmi/ Ea mu sucmi = ‘Let’s swim’ and /muSUCmiEa/ Mu sucmi ea = ‘We swim, I suggest’—then they are obviously disyllabic in both productions. Indeed ea is one of those eleven vowel-pairs, see Sec. 1.13, that are always disyllabic. So what we are doing in these two transcriptions is showing which of the two vowels in ea is being stressed: (eigh-AA-moo-SOOSH-mee) vs. (moo-SOOSH-mee-EIGH-aa).
Second, if both elements of an identifiable vowel-pair are shown as having stress—both emphatic or both normal stress—then that pair of vowels is to be pronounced as a monosyllable. Thus in /AImuSUCmi/ Ai mu sucmi = 'Yes, we'll swim' the /AI/ is obviously being pronounced monosyllabically, that is, as the (IGH) of 'high'. Loglan ai is the 'Aye' of intention, compliance or consent…as in the nautical 'Aye, aye, sir!'. We can of course write the pronunciation guide for this utterance as (IGH-moo-SOOSH-mee). (But is it possible that you are beginning not to need the guides?)
This leaves pairs of unstressed vowels as the only unexamined Case. For example, in /eituMREnu/ = (ay-too-MREH-noo) for Ei tu mrenu = 'Are you a man?'—literally, 'Eh, you man?'—both phonemes in the interrogative ei (which is pronounced, remember, as the English diphthong (ay) in 'say') are shown to be without stress. The interrogative ei is often unstressed when it is in utterance-initial position. In utterance-final position, however, ei is usually stressed. We can show this by writing /tuMREnuEI/ (too-MRE-noo-AY). We know from Sec. 1.13 that /ei/ is one of the four vowel-pairs that are always pronounced monosyllabically, namely /ai ao ei oi/. But even if we didn't know this, or had forgotten, we could infer the monosyllabicity of ei from the transcription /eituMREnu/ itself. On the other hand if we wish to show that two adjacent unstressed vowels are being pronounced disyllabically, as for example if someone spoke the invitation Ea mu sucmi without stressing either syllable of the disyllabic /ea/, then we could show that production with a close-comma in the transcription: /e,amuSUCmi/. So we can report the stress possibilities for ea in this sentence as the set of three productions: /eAmuSUCmi/ /EamuSUCmi/ and /e,amuSUCmi/. Make sure you can pronounce them all. The fourth logical possibility, /E,AmuSUCmi/, is not really a lively phonological one.
We can now write the third inference as follows:-
Third, if two adjacent unstressed vowels are separated by no mark, and they are indeed members of the same vowel-pair, then they are to be pronounced monosyllabically. Thus, /ei/ in /eituMREnu/ can be inferred to be a monosyllable because no mark separates the unstressed phonemes /e/ and /i/ from each other. (Note that from this point of view, a stress-rise, such as occurs in /eA/ (eigh-AA), and a stress-fall, such as occurs in /Ea/ (EIGH-aa), are both "marks".) If and only if the two unstressed adjacent vowels are separated by a close comma /,/—as in /e,amuSUCmi/—are they intended to be pronounced disyllabically. Such a close-comma is not a pause; but it does indicate elongation of the time interval occupied by these two unstressed phonemes just as pausing between them would do.
It is important that we transcribe these ambiguous unstressed cases in this explicit way and not rely on the reader's memory of what kind of vowel-pairs they are. Besides, we have those ten optional i- and u-initial pairs to worry about, those for which pronunciation has still not settled down. So we must be especially careful to convey in our guides and transcriptions just which of the several options we are asking the reader to consider.
We have now finished with the Loglan vowel sounds and can address the consonants.
1.17 The 17 Regular Consonants: There are 17 consonants among the regular phonemes, which means the sounds used in non-names. Their pronunciations are given here by English examples. Fortunately, all regular Loglan sounds occur in English, so good examples can always be found. The secondary and tertiary allophones of some phonemes are given on the sutori lines of their entries. The parenthetic expressions in the right hand column of the table are phones. The spellings given in this column are the least equivocal English spellings used in the pronunciation guides:
as in 'boy'
as 'sh' in 'sheep'
as in 'dog'
as in 'fat'
as in 'goat'
as in 'hat'
as 'z' in 'azure', '-ge' in 'garage'
as in 'king'
as in 'lake'
'-le' in 'kettle' when vocalic
as in 'make'
as in 'chasm' when vocalic
as in 'net'
before /k g/, as in 'sank'
'-ain' in 'certain' when vocalic
as in 'pot'
as in 'rot'
'-er' in 'father' when vocalic
as in 'sat'
as in 'top'
as in 'vet'
as in 'zoo'
All phones are single sounds even if the guides to them are written between parentheses with two letters, e.g., (ng). Also, note that the sounds represented by (ll mm nn rr) are what are produced when the normally consonantal continuants (l m n r) are sustained, e.g., as in 'brrr' where (r) is sustained to become the vowel-like (rr). Observe that the sound after (b) in (brr) is long but that it is a single sound. That is to say, no articulatory change occurs within it. When used in this way a consonant is said to be vocalic. We shall have more to say on the vocalic consonants in Sec. 1.19.
MRH: we now prefer to say ``syllabic consonant”
Notice that, except for /c/ and /j/, the letters associated with these 17 Loglan consonants are pronounced exactly as they most commonly are in English. Thus [b d f g h k l m n p r s t v z], while sometimes ambiguous in English, have a most common sound. That most common sound is always at least one of the sounds that letter will have in Loglan. This is true of allophones as well. Thus Loglan /n/ has the three allophones (n ng nn), all of which occur in similar contexts in English. English speakers who repeat the phrase 'sing out for sin gout in the fountain' will hear all three sounds of /n/. The difference between Loglan and English /n/ is that (ng) is one of its allophones in Loglan, whereas /ng/ is phonemic in English. Thus the 'sin/sing' distinction exists in English but not in Loglan. But in both languages the sound that is written as [n] is pronounced (ng) whenever it immediately precedes a /g/ or a /k/. Observe that the English word 'sink' has the sounds of 'sing' in it. The same is true of Loglan. Thus manko = 'mouth' is standardly pronounced (MAHNG-koh) in Loglan because the (ng) allophone of /n/ is required before /k/.
MRH: there is nothing new to say about consonant pronunciation here.
1.18 The Odd Sounds of [C c] and [J j]: Only the letters used to represent /c/ = (sh) and /j/ = (zh) seem to be oddly paired with their sounds in Loglan. The letters [c j] have been assigned to these two phonemes, first, because we need these sounds in this international language—(sh zh) are especially common in Chinese, for example—and, second, because no "digraphs" are permitted in Loglan. (A digraph is a sequence of two letters used to represent a single sound; thus [ck kn ng ph pn sh th] are all English digraphs.) But the Loglan rule is that each simple sound must be represented in written Loglan by a single letter. The letter [c] happened to be unemployed, [k] and [s] were doing all its (English) work in Loglan; and [c] is at least weakly associated with the (sh) sound in other languages. For example, French 'chat' is (shaa) and Italian 'ciao' is (chow). It turns out that the (ch)-sound in this Italian word may be further analyzed as (t) + (sh). So (ch) is in fact written in Loglan phonetics as (tsh), and phonemically this is of course /tc/. Thus, with [c] given the value (sh), the Italian word ‘ciao’ can now be written in Loglan phonemics as /tcao/, and English ‘chew’ as /tcu/.
Loglan /j/ is parallel to Loglan /c/. Thus (zh) is the voiced, that is, the unwhispered, version of unvoiced or whispered (sh), and it occurs commonly in French as the French /j/ of ‘Jean’ (zhaan). (We shall show the nasalization of a preceding vowel by superscript n.) (zh) also occurs in certain words of French origin in English, such as 'azure' and 'garage'. But (zh) is also one component of a very common consonant of English, one that is also represented by the letter [j]. For just as the (ch) sound may be analyzed as (t) + (sh) in Loglan, so the sound of English [j] may be analyzed as (d) + (zh). Thus English 'joke' may be written (dzhohk) in Loglan phonetics, and /djok/ in Loglan phonemics. By a precise parallel, 'choke' may be written (tshohk) phonetically and /tcok/ phonemically. This explains why the Loglan word for 'chain' is written tcena and that for 'judge' is written djudi (Try to pronounce these as (CHEH-naa) and (JOO-dee), not as (CHEIGH-naa) and (JUH-dee).)
1.19 The Four Vocalic Consonants: The vocalic values of the four consonants /l m n r/, namely (ll mm nn rr), are virtually confined to borrowed names in Loglan. Examples are English ‘Earl’ which is respelled in Loglan as R1 but is still pronounced (RR-ll); ‘Burton’ => Brtn pronounced (BRR-tnn) or sometimes (BRR-tyn); and ‘Herbert’ => Hrbrt pronounced (HRR-brrt). In such English names, as in similar names in other consonant-rich languages, these four vocalic consonants play the roles of vowels. So they have vocalic values in the Loglan loan-words Rl, Brtn and Hrbrt which are derived from these natural names, and which thus require no other vowels.
These same four consonants, however, are seldom pronounced as vowels when they occur in the regular words of Loglan, i.e., in its non-names. Thus the instances of /r/ in the predicates rodlu (ROHD-loo) 'road', farfu (FAAR-foo) 'father', and brudi (BROO-dee) all have consonantal values. In contexts like mrenu 'man', however, the /r/ is sometimes vocalic, especially in the speech of newcomers to the language. That is, mrenu is sometimes pronounced as the trisyllable (mrr-EN-oo) rather than as the disyllable (MREH-noo). But the word is still spelled mrenu and transcribed as /MREnu/ in Loglan phonemics; for in regular words (rr) is simply an allophone of /r/. That is to say, (rr) is a legitimate, contextual variation of /r/ but does not have separate phonemic status in the language.
What this means is that hearing (mrr-EN-oo) rather than (MRE-noo) does not cause a loglanist to think da's heard a different word; but only that da is listening to someone who is not comfortable with this Loglan consonant combination yet.
MRH: There is a further use of syllabic consonants in 1989 Loglan I and in the current language. Syllabic consonants are sometimes used as consonantal hyphens in the formation of borrowings. It should also be noted that (following a note in L1 which suggests this usage) I now require that all syllabic consonants be doubled in writing. This causes the spelling of some names to change; the consonantal hyphens are already written doubled. The fact that rr can occur in predicates has the effect that the distinction between the consonantal and syllabic uses of a consonant becomes phonemic ( vs re troviri> , and the remarks about safety of pronouncing consonantal occurrences as if they are syllabic are at the very least doubtful.
1.20 The Unfamiliar Consonant Pairs: /mr/ is just one of a handful of Loglan consonant combinations which will be unfamiliar to monolingual anglophones. Some others are /dz/ as in dzoru (DZAW-roo) = 'walk'; /ts/ as in tsero (TSEH-roh) = 'error'; /cl/ as in cluva (SHLOO-vaa) = 'love'; /ct/ as in ctifu (SHTEE-foo) = 'stuff/matter/material'; /ck/ as in ckozu (SHKOH-zoo) = 'cause'; /cm/ as in cmeni (SHME-nee) = 'money'. Despite their curious appearance, none of these consonant-pairs will be difficult for an anglophone to pronounce.
The two pairs /dj/ and /tc/ look difficult but are not. The sounds they represent are actually very common in English, being usually spelled [j] and [ch] as discussed above.
1.21 The Three Irregular Phonemes /q w x/: These three sounds—one of them, /w/, a vowel, and the other two, /q x/, being consonants—occur only in names and letter-words, and then only rarely. All three are, by world standards, "difficult" phonemes. That is to say, each is found in only a small subset of the world's languages and is therefore unfamiliar to most human ears and tongues. But the letters assigned to these three sounds in Loglan, namely [Q q W w X x], occur as letter-words (see Sec. 2.21) or as parts of acronyms (2.29) in many scientific contexts—for example, the chemical symbol for tungsten is 'W', and the 'X' in 'X-ray' must somehow be accommodated—and they are also employed, although not frequently, in writing one of the most extensive vocabularies of science, namely the Linean binomials of biology. These two-term species names, like 'Homo sapiens' and 'Escherichia coli', are by international convention always either italicized or bolded in text and spelled identically in all languages…even in Chinese, Japanese, Hindi, and Russian, languages which do not use the Latin alphabet in any other context. It is therefore advisable that Loglan, too, provide the means, at least, for the transcription of these binomials, and in ways that preserve as much as possible of their standard visual character. To this end, once the 23 regular Loglan sounds had been given letters, the 3 letters left over from the standard Western European alphabet were given to these three fairly commonly used European sounds. Their phonetic values are as follows:
'th' as in E. 'thin' (unvoiced)
'th' as in E. 'then' (voiced)
'eu' as in F. 'bleu', umlaut 'u' in G. 'München'
'ch' as in G. 'Bach', 'k' in R. 'Kruschev'
As an anglophone, you will know the sounds (th) and (dh), of course; both are spelled in English with [th]. But you may balk for awhile at spelling them with [Q q]. One useful mnemonic is that upper case [Q] looks quite a lot like upper case Greek theta [θ] but with its bar slipped down and tipped a little, whence [Q]. Theta is, of course, the letter used in Greek for its (th) sound. The other two irregular sounds are non-English, but probably familiar enough to those linguaphiles who make an effort to pronounce foreign words correctly.
MRH: It is the view of the Institute that these letters should be dropped completely, though names for them should be retained.
1.22 The Use of Irregular Phonemes in Names: As suggested by the examples given above, the three irregular sounds of Loglan may also be used, at the name-maker's discretion, to produce better imitations of natural place and person names than would otherwise be possible. Thus German 'Bach' may now be imported into Loglan without change as Bax (bahkh). Similarly, 'München' (Munich) may be exactly reproduced as Mwnxen (MEUN-khen). The sounds of English 'Theodore' may be exactly reproduced as Qi'ydor (THEE-uh-dohr), and 'Kruschev' may be well-approximated as Xrucyf (KHROO- shuhf). On the other hand, if the word-maker is concerned about minimizing pronunciation difficulties for the large number of persons, on a world-wide basis, to whom one or more of these three irregular sounds is completely alien, and if the respelling of the natural name in regular Loglan sounds is both sufficiently distinct from other names and a recognizable transformation of its source word, the builder may opt for less than perfect imitations of the natural language sources and use regular sounds entirely. This would give Bak (bahk), Muncen (MOON-shen), Tiador (TYAH-dohr), and Krucyf (KROO-shuhf) for the four natural names in question; and at present this is perfectly acceptable.
At the moment, a mixed strategy seems to be most attractive. Thus with the arrival of /x/, 'Bach' is now being enthusiastically rendered by Bax and 'Kruschev' by Xrucyf. On the other hand, Tiador and Muncen still seem to be acceptable variants of 'Theodore' and 'München' which avoid the formidable oddities of Qi'ydor and Mwnxen. It may be that these oddities will turn out to be mainly visual. In that case, they are likely to become less objectionable with the passage of time. So the policy of whether or not to use the irregular sounds of Loglan in contriving the Loglan forms of imported natural names is still an open one. It is likely to remain so for some time.
1.23 Three Stress Phonemes: We have already indicated how stress is written in the guides and productions, namely no stress by lower-case letters, normal syllable stress by upper-case letters, and emphatic stress by both bolding and upper-casing the normally stressed syllable of the emphatically stressed word. It remains to mention that these three values are sufficient to describe all meaningful differences in Loglan stress contours.
For example, the pronunciation guide for the Loglan equivalent of 'You too came!' is
(too-SWEE . paa-KAAM-laa)
The Loglan sentence whose production is here being guided is of course Tu sui pa kamla. Word-for-word this means 'You also before come'. As we will see in the chapter on grammar, the Loglan tense operator is a separate word; it precedes the expression to be tensed.
Acoustically, stress is an increase in both the amplitude (loudness) and the duration of the sounds that make up a stressed syllable or syllables. Both are measured relative to the other sounds of its utterance. Thus (SWEE) is not only the loudest syllable in the above utterance, it is also the longest.
MRH: as noted above, I denote stress by an apostrophe and emphatic stress by an asterisk, following the syllable, serving also as a syllable separator if there is a following syllable.
1.24 One Pause Phoneme: The pauses that occur in Loglan, as in any language, vary markedly in both length and significance: from long hesitations to brief phrasing pauses. Nevertheless, for morphological and grammatical purposes, it is sufficient to classify all Loglan pauses as members of a single phoneme. This makes the Loglan pause phoneme a single one with many allophones. For example, the glottal stop in Loglan is an allophone of pause. So those instances of exceedingly brief glottal stops that occur, for instance, between the vowels of a disyllabically pronounced vowel-pair are not counted as pauses at all. In short, as in most languages, Loglan pauses are those silences that occur in the midst of speech that are perceptible as such. In acoustic practice this will mean that there is a number of milliseconds above which a pause, say a glottal stop, is perceptible by most auditors as such and below which it is usually not.
A pause is always marked by a period, or full stop, in the guides and productions. In the guides, open-periods [ . ] are used; in the productions close-periods [.] are used. The silences that precede and follow utterances are not pauses; they are either not marked at all or marked by the pound-sign [#] in the productions. An utterance so marked is an isolated utterance. Thus
is the full phonemic transcription of one way in which the utterance Tu sui pa kamla may be produced, namely as an isolated utterance with emphatic sui and an optional pause after it. Contrast the phonemic transcription of this production with the phonetic guide to the pronunciation of the same utterance in the previous section. Notice that the phonetic detail has disappeared, leaving the stress/pause "contour" of the production plainly revealed. Because stress and pause together describe a pattern that overlies the whole utterance, the stress and pause phonemes of a language are often called its suprasegmental phonemes. Its vowels and consonants are called its segmental phonemes.)
MRH: We have restored the single pause phoneme, reversing an earlier decision which created a shorter pause phoneme appearing between components of serial names. Explicit pauses are denoted by a comma followed by at least one space. In some contexts, this must be written; we have tried to make sure that it can always be optionally written in any place where a pause is actually required, though in many places a pause can be deduced that is not actually written.
MRH: I have no comments on 1.25 or 1.26.
1.25 Intonation: Intonation is not a phonemic quality of Loglan speech, and so is not indicated in our phonemic transcriptions. Intonation in Loglan is in fact free to vary in any way a speaker or a dialect group finds natural. It is true that in the Loglan speech we have observed so far questions are generally accompanied by rising tone, and the approaching end of a declaration by falling tone. But this may only be because the Loglan speakers we have observed so far are native speakers of English; and in English this is what happens. Still, in deference to the tonal habits of this potentially "ancestral" group of speakers, a questioning or exclamatory intonation is sometimes shown in the pronunciation guides by the use of [?] or [|].
It will be interesting to see how the tonal dimension of Loglan develops as speakers become more fluent. Will it only reflect the dialectical origins of different groups of speakers in different natural languages? Or will other, language-wide tonal features eventually develop? If so, will these be the result of "universal mechanisms", that is to say, of mechanisms which are inferrably fixed in the human gene pool and so operate in human languages generally? Or will they be only "founder effects", that is, the accidental consequences of the fact that the language started in a certain place—Gainesville, Florida, U.S.A., to be exact—and was first taught by a certain retired college professor who happened to be a native speaker of English, but who also happened to be foreign-born, to have lived many years abroad, and to have a small acquaintance with certain other European languages? Complete tonelessness, such as often characterizes the speech of badly programmed computers, is probably not even a possibility for human speakers.
Incidentally, toneless computer speech—of the sort that can now be produced by Vocoders and other devices—is surprisingly intelligible when the language is Loglan. This is probably because all the syntactical work normally done by tones in human languages is done by particles in Loglan. Thus, we use the question particle ei (ay) (as in 'lay') to transform the declarative sentence Tu mrenu /tuMREnu/ ('You are a man') into the question Tu mrenu ei /tuMREnuEI/ ('Are you a man?'). (Ei tu mrenu /eituMREnu/ works just as well of course, and is slightly to be preferred grammatically.) Notice that the phonetic difference between the two English utterances is far greater than that between the two Loglan ones, and that the English difference is largely in the word-order and intonation. In Loglan, our recognition of the fact that the second utterance is a question could be made to depend entirely on our hearing and understanding the extra word ei. So both questions /tuMREnuEI/ and /eituMREnu/ may be spoken quite tonelessly and still be recognized as questions by a loglanist…or by a suitably programmed loglaphone computer.
1.26 Buffered Dialects: Anticipating that many who will eventually speak Loglan will find some of the consonant clusters in its standard, anglophone dialect awkward to produce, we expect the sixth vowel, schwa or what is now Loglan /y/, to be used for "buffering" any or all of those clusters: that is, to separate difficult consonant pairs by interposing this neutral vowel between them. Dialects which used /y/ in this way would then be known as the "buffered" dialects of Loglan.
It is likely that the Japanese dialect of Loglan will be buffered. While Japanese has some consonant pairs, they together comprise only a small fraction of those that occur in standard Loglan. Here's a partial list of permissible consonant-pairs in Japanese: dj tc ts; n followed by c d g k r s t z and by dj tc ts; mb mp. Nearly any other consonant-pair is going to be awkward for a Japanese-speaking loglanist to produce.
As an example of what the Japanese already do about such matters, here is their rendering of Dutch 'Amsterdam': 'Amusuterudamu'. [u] is obviously the letter by which they represent their buffering vowel in Latin transcription. But phonetically the Japanese buffering vowel has a value much closer to schwa /y/ than to our Loglan /u/, being approximately the sound of English 'u' in ‘full’. Thus a Japanese dialect of Loglan could well employ /y/, or some suitably neutral local allophone of it, to stand between any pair of consonants that was not on some official list of "pronounceable pairs", a completion of the list begun above. The result would be easy for a Japanese to produce and yet be fully understandable to speakers of the standard, European dialect of Loglan…as 'Amusuterudamu' is understandable to us now. Thus with a little practice in listening to them, ma'tyma, fa'ryfu, so'ryme, and byru'di would soon be intelligible Loglan words to English-speaking loglanists, whether in speech or in print. Many predicate words would be the same in both dialects, e.g., ti'tci, da'mpa and tse'ro; and of course none of the Loglan little word sequences—which offer few if any pronunciation difficulties to the Japanese—would require buffering; and so this extensive region of the language would also remain substantially invariant across dialects.
The question arises: If we are going to use /y/ for buffering consonants in the buffered dialects, what are we going to use as the intraverbal hyphen in those dialects? The two functions must be kept separate; to let one operator serve them both would be to give wrong information about term-joints to the resolver, /iy/ pronounced (yuh) has suggested itself as the hyphenating syllable to be used in buffered dialects; see Sec. 1.10. /iy/ is easy to pronounce, phonemically similar to /y/, and can easily—one might say harmlessly—be used in all positions in which /y/ is used in the standard dialect. As an example of a word that would then contain both /y/ and /iy/, take sanpyse’nsi, the new Loglan word for ‘semiotics’ in the standard dialect. The consonant pair /ns/ is permissible in Japanese and needs no buffering; but /np/ is not, and so requires buffering. Replacing /y/ with /iy/ and buffering /np/ with /y/ produces sanypiyse’nsi (saan-uhp-yuh-SEN-see), an eminently pronounceable word for 'semiotics' in Japanese Loglan…and one which is recognizably the "same" as (saan-puh-SEN-see) in standard Loglan.
MORPHOLOGY (WORDS & WORD-FORMS)
2.1 Design Objectives: The principal objective to be met in designing Loglan morphology was to ensure the audio-visual isomorphism of the language. By this is meant that the written and spoken forms of every well-formed Loglan utterance, as those forms would be seen and heard by their readers and listeners, respectively, would in all non-trivial respects be reciprocally inferrable from one another. In short, that the spoken string /la.ARqr.paceNOInaBRAgai/ be convertible always and everywhere into the written string La Arqr, pacenoina bragai ('Arthur is no longer king'), and vice versa. It was hoped that this feature would contribute not only to Loglan's utility as a man-machine interface language, where the contribution of isomorphism is obvious, but also to its learnability as a second language by adults, in which the transition from writing to speaking is often a difficult one, and finally also to its acquisition as a second tongue by children, in which the reverse transition is often the difficult one.
Perhaps the most important task to be performed in insuring audio-visual isomorphism is that of word resolution, a task of the listener. For the listener to perform this task as swiftly as speech is uttered, and also impeccably, there must be a reasonably simple resolving algorithm—learnable by humans and programmable for computers—that is capable of transforming any heard utterance which has been grammatically composed of well-formed but largely pauselessly-spoken Loglan words, into just one string of properly spaced, capitalized and punctuated written words. Loglan morphology makes such an algorithm both easy to write and to execute in real time. Even a small computer can execute it routinely.
We must note one exception to Loglan's audio-visual isomorphism. If an utterance being read aloud from text involves a Linnaean binomial, that is, one of the standardly spelled pairs of genus and species names which are universally used as naming terminology in biological science (e.g., Australopithecus afarensis and Escherichia coli), then that utterance will not always be simultaneously transcribable in a form that matches the original text in all particulars. The Linnaean binomials are regularly pronounceable in Loglan phonology; see Sec. 2.13. As such they may be regularly resolved and transcribed; but those transcriptions will not be infallible reproductions of the original Linnaean words. To recapture those originals exactly, thus completing the isomorphism, will require in most cases a tabular lookup that will probably not be executable in real time.
MRH: it is the case that from the pronunciation, the orthography can be inferred, mod such things as optional uses of punctuation and phonetic notation. The opposite is not the case, because of the optional disyllables, but this is in my opinion a harmless variation. The word resolution algorithm is fully implemented in my parser. I am not nearly as concerned with the Linnaean nomenclature as JCB was.
2.2 Definitions and Conventions: First, we need a small technical vocabulary and a number of expository conventions. A morph is a simple word or a simple component of a larger word that has a distinct meaning or makes a distinct contribution to the meaning of the embedding word. A morph, in short, is the smallest meaning-bearing component of speech. For example, pa is both a simple word meaning (roughly) 'previously' and a component of compounds like pacenoina, where it means 'before'. A morpheme is a set of one or more morphs with the same or related meanings. If a morpheme has two or more morphs as members, they are said to be its allomorphs. For example, both no and -noi(-) convey the meaning of negation in Loglan ('no' 'not' 'non-' etc.), and the second morph is plainly a derivative of the first; so each is an allomorph of the morpheme no. A morpheme is named for its free allomorph, i.e., the one that is derivationally most basic and also capable of functioning as a simple word. In Loglan, these are always the same morph; and in this case it is no. Morphs that are not free are bound. Thus -noi(-) is the bound allomorph of no because it occurs only in compounds. The leading ‘-’ in the expression ‘-noi(-)’ means that -noi(-) is never initial in its embedding word; the trailing ‘-’ means that -noi(-) is sometimes medial and sometimes final. In an alternative terminology, the markings on -noi(-) show that it is either an infix or a suffix but never a prefix. In this terminology a bound morph is called an affix.
A word is a string of one or morphs that is treated as a single element by the grammar, just as 'nevertheless' is treated as a single element by English grammar. Words are simple if they contain exactly one morph, and compound otherwise. Boundaries between words are called junctures. It is the business of the resolver—a computer program or some subsystem of the human central nervous system—to discover the junctures in a string of pauselessly pronounced words. When it is important to show junctures, we will represent them by the boldface vertical bar [|]. The joints between morphs in a compound or complex word will be shown by the boldface slash [/].
A set of morphs or morphemes whose phonological constituents may be specified in a relatively compact formula is said to be a form-class or simply a form. We shall use the following conventions to represent the constituents of such forms: 'X' will represent any (segmental) phoneme, that is, any vowel or consonant including the irregular ones /q w x/ and /y/. 'V' will represent any of the primary vowels /a e i o u/ or any vowel except /y w/. 'C' will represent any regular consonant, that is, any consonant except /q x/. If it is unimportant whether a pair of primaries is pronounced monosyllabically or not, we shall write its formula as 'VV'; if an obligatory monosyllable is called for, we shall write it 'vv'; if we require an obligate disyllable, we shall write 'V,V'.
MRH: I note that it is a matter of current conversation in TLI that we should use current linguistic terminology. Some changes are in order. I note that the use of ``affix” which we now deprecate is much more special than the one defined here. I am not a linguist, so I will leave it to others to analyse what appears above.
A permissible initial consonant pair is a pair of regular consonants that may be initial in a morph, i.e., in an affix or a word; a permissible medial is a pair of consonants that may be adjacent in a morph. Clearly, the permissible initials are a proper subset of the permissible medials. When it is unimportant whether a pair of adjacent consonants is a permissible initial, we shall write it as 'CC'. When a consonant pair in a given formula must be a permissible initial, we will write it 'cc'.
As before, we shall mark proscribed words or forms with a leading asterisk; for example, *spai is not a permissible word because *ccvv is not a permissible form. In a demonstration sequence, if the reader is not expected to know yet whether a given form is acceptable, for example, ?ccV,V, it will be marked by a leading question-mark. ?Spea pronounced (SPEIGH-aa) is one of the instances of this questioned form. Actually, this form too is unacceptable…for reasons that we will find out later.
Formulas and their ingredients are usually distinctive enough to require no special marking. Thus VV is the formula for the 25 possible pairs of the 5 primary vowels. Silences, pauses, syllable breaks and stresses on preceding vowels or vowel pairs are shown in the formulas by their phonological marks [# . , ’]. If a specific phoneme is required in some formula, it will be shown in the formula by the same lower-case letteral that represents it phonemically. Thus the formula dV is the set of 5 morphs generated by putting each of the 5 primaries after phoneme /d/, in particular, it is the set da de di do du. The ambiguity this convention seems to create between 'vv', which could mean a pair of /v/s but doesn't, and 'vv', a monosyllabic vowel-pair, is only apparent. If it were a pair of v's, they would be bolded, stroked or bracketed. Besides, "double consonants", or adjacent instances of the same consonant, do not exist in Loglan. [MRH: momentary amnesia about syllabic consonants!] This renders 'cc' unambiguous as well.
To specify a set of equally permissible alternatives in a formula we shall separate them by strokes and enclose the set in square brackets. Thus, [CCV'/CV'C] says that either of these two kinds of stressed syllables is acceptable here. The order in which the alternatives are listed is unimportant. A plus-sign is used as a sign of concatenation between the parts of a word. Thus [CCV'/CV'C]+CV is the complete formula for the form-class to which both brudi and matma belong. (They are pronounced (BROO-dee) and (MAAT-maa) and mean 'brother' and 'mother'.) Parts of a form which may be iterated one or more times—that is to say, may be present in a larger form as a string of one or more instances of that part—are shown in parentheses in morphological formulas. Thus, (X)+[C/q/x]+. is a formula for the form composed of a string of one or more consonants and vowels of any sort and in any order followed by either a regular consonant or one of the two irregular consonants /q x/, the whole followed by a pause.
MRH: I do not take exception to this; I might often use PEG expressions for the same purposes.
Table 2.1 The Two Partitions and Three Classes of Loglan Words
Always C + /./
/q w x/
Penultimate or none
As hyphen only
At least one
2.3 Two Major Partitions and Three Word-Classes: As Table 2.1 shows, the two major partitions of Loglan morphology are (1) between regular and irregular words, and, among regular words, (2) between structure words and predicates. There is only one kind of irregular words, namely names. Thus the two morphological partitions generate three major word classes: (i) names, (ii) structure words, and (iii) predicates.
Names are the relatively permanent designations of places, persons and things, which, in any given context, are always meant to be unique designations. In Loglan most but not all names are borrowed from the natural languages. The structure words of any language establish the grammatical structure of its utterances and are generally quite short. They tend to be "little words", like 'the' and 'now' in English and le and na in Loglan, but structure words may also be compounds of short words, like 'nevertheless' in English and Lena = 'the-present…' in Loglan. Compound structure words may sometimes be quite long, for example, pacenoina = 'before-and-not-now', or 'no longer'. Occasionally we will speak of simple structure words as little words. The words in the third class, predicates, are the words of reference in any language. For example, predicates are the nouns, verbs and adjectives of English. Unlike English predicates Loglan predicates are grammatically undifferentiated and have a simple, readily identifiable form.
Table 2.1 gives the features that form the two partitions and identify and characterize the three classes of words. The first two features differentiate irregular from regular words. All irregular words are names, and these and only these may contain the irregular sounds /q w x/. Furthermore, only names end with consonants followed by an obligatory pause; all other Loglan words end in vowels. The next feature, stress, also helps to differentiate names from non-names. Stress may occur anywhere in a name, and if stress in a name is "irregular", that is, if it falls on any syllable except the penultimate one, the stressed vowel is marked with an apostrophe in the written form, e.g., Pari's. The next feature also helps to differentiate names from non-names. The grunted vowel schwa or /y/ may occur anywhere in a name and even be stressed, e.g., Hyntr (HUHN-trr); but it does not occur in structure words at all and occurs only as a hyphen in predicates, e.g., mekykiu (MEK-uh-kyoo). The final feature, the presence or absence of consonant-pairs, differentiates the two kinds of regular words from one another. Structure words never contain adjacent consonants; predicates always contain at least one pair. Names, in keeping with their general flexibility, may or may not contain adjacent consonants.
Because of these deep morphological "furrows" in the vocabulary of Loglan, the learner can tell the broad morphological class of each new Loglan word da encounters at a glance. Take the sentence La Arqr, pacenoina bragai (laa . AAR-thrr . paa-ceh-NOY-naa-BRAA-gigh). La must be a little word; it is V-final and contains no CC. Arqr must be a name; that it is C-final is sufficient to tell us this, but there is another clue. Arqr also contains the irregular phoneme /q/ (th). Pacenoina, although clearly too long to be a "little" structure word, must nevertheless be a compound one, i.e., one that is composed of the little words pa + ce + noi + …, etc. For it, too, is V-final and free of CC's. Only bragai (BRAA-gigh) is a predicate; while it is also V-final, it contains the consonant-pair /br/. Eventually you will be able to translate this sentence as 'Arthur is no longer king'. But right now we are concerned only with its morphology.
The forms of names are discussed in the next section. The morphology of structure words begins with Sec. 2.18, and of predicates, with Sec. 2.34.
MRH: the qualification of stress in structure words (cmapua) as penultimate or none is striking. I have always understood that stress in structure words is simply free. In fact, there is s stress rule for what to do when stress in a cmapua is final, so I think this is an inconsistency. I view stress in cmapua as free, subject to the relevant stress rule. Otherwise, this section is current. One further remark is that in the current state of the language it makes less sense to distinguish names as irregular; there are no longer any irregular phonemes, and there are phonetic rules for the formation of names similar to those for the formation of borrowings.