1.1 Definitions and Conventions: We require a small technical vocabulary. A phoneme is a class of one or more speech-sounds all of which are regarded as “instances of the same sound” by the speakers of some language. The individual sounds so classified are called phones. All the phones which are members of a given phoneme are called its allophones. I shall use strokes [/] to mark off phonemes, sets of phonemes, and phonemic transcriptions. Thus /a/ is a phoneme, /a e i o u/ is a set of phonemes, and /eAmuGODzi/ is a phonemic transcription. I shall use parentheses to mark off phones, sets of phones, and phonetic transcriptions. Thus (ah) is a phone, (ah aa) are the two allophones of phoneme /a/, and (ey-AH-moo-GOHD-zee) is the phonetic transcription of the utterance phonemically transcribed as /eAmuGODzi/. The utterance itself is Ea mu godzi; it means ‘Let’s go!’ Phonetic transcriptions of utterances will sometimes be called guides; they are phonetic guides to at least one way of pronouncing those utterances. Phonemic transcriptions are often referred to simply as productions. They exhibit one of the ways an utterance might be produced in speech.
MRH: I have no problems with the terminology in this paragraph. It should be noted that my preferred phonetic notation is different, being a variation on Loglan standard spelling.
Stressed syllables are shown in uppercase letters in both the guides and productions. Hyphens are used to show syllabification in guides. Stress rise and fall, shown by the shift from lower to upper case characters, are normally sufficient to show syllabification in productions. But when there are strings of unstressed vowels it is sometimes necessary to show syllable breaks in a production by using close-commas (commas without an adjacent space), e.g., /i,aimiGODzi/. This production shows us that the utterance involved was I ai mi godzi = ‘And yes I intend to go’. The close-comma between them puts /i/ and /ai/ in separate syllables. Without the close-comma, the default syllabification rule would pair the vowels from the left. Thus /iaimiGODzi/ syllabifies as (yah-ee-mee-GOHD-zee), and this gives quite a different resolution, namely *Ia i mi godzi = ‘Certainly; and I go’.
MRH: For reference to current materials (I am not at this time going to make systematic changes here), I use hyphens to denote syllable breaks (and regard their inclusion as optional) and use apostrophe at the end of a syllable (not accompanied by a hyphen in contexts between syllables, but replacing it) to denote stress, again optionally. My current parser will recognize these things. The optionality (necessitated by my desiring the parser to handle both regular Loglan text and phonetic transcriptions) means that I do not have a means of indicating explicitly that a syllable is not stressed. I could add a notation for this, though. I do provide * as notation for emphatic stress. I make no use of close commas.
The asterisk [*] on *Ia i mi godzi shows that this utterance is wrongly pronounced. I shall use leading asterisks [*] to mark all expressions which are illegitimate at some level of correctness, whether that level is phonological, morphological or grammatical. Even bad usages are so marked. *Ia i mi godzi is morphologically incorrect because it is missing an obligatory pause before the little word I; such pauses are sometimes necessary for resolution. I shall use leading question-marks [?] to show that an expression is of unknown legitimacy—unknown, that is, to the reader. Thus ?Ia i mi godzi is how we would mark this utterance before the reader is expected to know that it is malformed.
Pauses are shown in written specimens either by a comma-space, e.g., Ia, ice mi godzi (‘Certainly; and I go’), or by period-spaces which require two following spaces, e.g., Ia. I mi godzi (‘Certainly. I go’). The second marking is equivalent to the full-stop in English. Both pauses are shown in productions by close-periods, e.g., /ia.imiGODzi/, and in the guides by open-periods (space-period-space). e.g, (yah . ee-mee-GOHD-zee). So the correct pronunciation of these last two specimens is given by the pair of guides (yah . ee-mee-GOHD-zee) for the one that means ‘Certainly. I go’ and (yah . EE-sheh-mee-GOHD-zee) for the one that means ‘Certainly; I go’. A silence is the long pause before an utterance begins or after one ends; it is not usually useful to show silences in productions, but when it is useful to do so they are shown by pound-signs [#]. A breathgroup is a pauseless string of phonemes lying between pauses, silences, or pauses and silences in either order.
MRH: I use commas to signal explicit pauses. I also use periods (and other closing punctuation) in specific contexts. These generally end certain kinds of utterances. Both commas and periods are largely controlled by grammatical rather than phonetic rules, which is in itself instructive. I must admit that I do not require two spaces after a period (I probably don’t require any, I ought to do something about that).
I shall use square brackets to enclose textual characters, or sets or strings of characters. Thus [a] and [l] are visual characters; [, ] is the comma-space and [. ] is the period-space; [ . ] is the open-period used in guides; while [,] [.] and [#] are the close-comma, close-period and silence, respectively, used in phonemic transcriptions. Although we will rarely have occasion to use bracketed utterances, we can also say that [Ea mu godzi] and [Let us go.] are visually perceived utterances, that is, pieces of text.
Sometimes I shall wish to refer to characters that are not on the fonts available to me for printing this document. I shall say, for example, that in lieu of the proper characters the printed expression [alpha] stands for the Greek lower-case character called alpha while [Alpha] refers to the upper case version of that character.
MRH: I have no particular use as yet for quotation conventions of this sort, but I might in future!
I shall, as is customary in Institute publications, use bolding to mark Loglan specimens whenever their status as text or speech is immaterial. Thus Ea mu godzi is the utterance variously represented above. I shall also use bolding to indicate emphatic stress in both phonemic transcriptions and guides: /eAmuGODzi/ = /ey-AH-moo-GOHD-zee/ = ‘Let’s go!’ I shall use single quotes to mark non-Loglan (usually English) specimens when the mode of delivery is immaterial, and I shall using underlining for emphasis in both such non-Loglan specimens and in ordinary text. Thus ‘Let’s go!’ is a translation of Ea mu godzi pronounced as /eAmuGODzi/ with no special stress on any word, while ‘Let us go!’ is a translation of the same utterance when pronounced /eAmuGODzi/ with godzi emphasized. (Bolding is also used for titles, technical terms about to be defined, and for chapter, page and section headings in this document…I trust without confusion.) I shall use double-quotes for English expressions in the text which are either not being used literally or which have not been defined yet. Thus the allophones of a phoneme may be said to be “instances of the same sound” but are not literally.
MRH: I note the convention which I have been trying to follow in email of enclosing Loglan text appearing in an English context in angle brackets.
The Loglan words which I customarily use as English words when I am addressing loglanists remain unmarked. For example, ‘sutori’, which means ‘at least second’ or ‘second and subsequent’, as in ‘the sutori places of the predicate’, has become a quasi-English word in my idiolect and so is undistinguished in my English prose…just as the French, German or Latin words and phrases which have been taken into English are now usually unmarked in written English (‘That’s a priori reasoning’ ‘He's a gestalt psychologist’). I expect loglandical readers to know the meanings of these borrowings from Loglan, as from German or Latin, or to catch up by looking them up. Besides, they impart a usefully loglandic flavor to one's English thought and speech. The borrowings from Loglan which I notice I use most frequently are ‘da’ and its kin. I treat these words as genderless, numberless, and caseless 3rd person English pronouns.
MRH: This is a good place to note our need of further “native” grammatical vocabulary.
I occasionally use the sign ‘=>’ to mean ‘is the source of’, ‘<=’ to mean ‘is derived from’, and ‘=’ between a specimen and its translation to mean ‘has approximately the same meaning as’. I use parentheses for other uses than Loglan phonetics, of course, as in the next paragraph. I trust this will cause no confusion.
I have used least equivocal spellings in ordinary English letters (a procedure made familiar by the Berlitz people) to represent all the Loglan phones we’ll need. Thus, I've used (sh) to represent the sound of Loglan /c/, (ee) for the principal sound of Loglan /i/, (oo) for the main value of its /u/, (igh) for the otherwise hopelessly equivocally-spelled ‘eye/aye/I’-sound of English, which in Loglan is the sound of the diphthong /ai/, and I've used (eigh) for the sound it invariably spells in such words as ‘eight’ and ‘freight’ for one of the two allophones of Loglan /e/; and I have chosen to do this rather than use the special symbols of the International Phonetic Alphabet. The IPA symbols are not widely understood except among linguists; changing fonts to print them would slow down the production of this notebook by about a factor of ten; and our discussion of the sounds of Loglan loses little in technical precision by our adoption of this humbling convention. Phonetic issues in Loglan are not so wide-ranging that they require the full armamentarium of modern phonetic scripts.
MRH: My inclination is to explain the Loglan pronunciation of individual sounds, then use Loglan spelling with phonetic annotations to handle phonetic transcriptions.
1.2 Two Types of Phonemes: There are two types of phonemes in Loglan: the regular phonemes that are used throughout the language, and the irregular ones that may be used only in (proper) names. The term ‘name’ will always mean a proper name in our discussion of Loglan.
MRH: it is currently on the agenda of the Institute to remove the irregular phonemes.
1.3 Regular Phonemes: There are 23 regular phonemes in Loglan: /a b c d e f g h i j k l m n o p r s t u v y z/; so to represent them, Loglan uses the entire English alphabet less the letters [q w x]. Two of these phonemes, /h y/, are new since 1975. /h/ was added in 1981 because it was present in all the target languages and adding it made some much needed word-space available for new little words. Unexpectedly, incorporating /h/ also increased the average recognition scores of Loglan composite primitives (C-Prims) due to the large improvement in the recognizability of the 100 or so primitives that were remade with /h/. Schwa was added in 1981 as the buffer in buffered dialects and was assigned the unused English letter [y]. In 1986 the sound /y/ was given the even more important function of serving as the intraverbal hyphen in regular words; see Sec. 1.8.
MRH: nothing here requires any change.
MRH: sections 1.4-1.8 on vowel phonetics are just fine as they are.
1.4 The 6 Regular Vowels: The 6 regular vowels are pronounced as in the following table. The primary allophone is given first, the second, if one is recognized, appears on following lines:
As in ‘father’ in the Germanic dialects of Loglan, but as in S. ‘casa’, F. ‘la’ and E. ‘palm’ in the Romance dialects. In the "standard" (Gainesville) dialect, the Romance pronunciation is preferred.
Romance /a/, as in S. ‘casa’, F. ‘la’ 6c E. ‘palm’
Germanic /a/, as in G. ‘vater’ 6c E. ‘father’
Problems: In unstressed syllables, /a/ is occasionally but incorrectly pronounced as in ‘about’ or ‘sofa’, which permits it to be confused with the (uh) sound of /y/; see below.
As in ‘met’
(e) or (eh)
Before vowels, the sound of ‘eigh’ in ‘freight’ or ‘a’ in ‘ate’; it is also the first part of the diphthong ‘ey’ in ‘grey’
Problems: There is a strong tendency for English-speaking persons to use either the monophthong (eigh) as in ‘eight’ or the diphthong (ay) as in ‘day’ for /e/ in V-final monosyllables, i.e., to say ‘day’ for de and ‘say’ for se. This must be resisted, if only because Loglan words like dei and sei also exist in which the diphthong (ay) is genuinely present. This tendency can be resisted by making sure that dese are pronounced like the first parts of ‘debt’ ‘set’. These then contrast sharply with deisei as properly pronounced (day say).
As in ‘machine’
Before vowels, ‘y’ as in ‘yet’
As in ‘note’ but shorter; resist the tendency to say ‘oh-oo’
Before /r i/, as in ‘more’ or ‘noise’
As in ‘lute’ but shorter; resist the tendency to say ‘oo-oo’
Before vowels, ‘w’ as in ‘woo’
‘e’ as in unstressed ‘the’, ‘a’ in ‘sofa’ and ‘above’, ‘u’ in ‘up’ and ‘under’
1.5 The Advantages of Romance (aa): Loglan /a/ has two permissible sounds, the Germanic (ah) of 'father' and the Romance (aa) of 'palm' and Spanish 'casa'. I shall usually use the Romance (aa) of 'casa' in the guides simply because I wish to remind the reader that, in the regular words of Loglan (although not necessarily in its names), the Romance pronunciation is preferred over Germanic (ah). Thus, I will usually guide the reader toward the preferred pronunciation (MAAT-maa) of matma = 'mother', despite the fact that the Germanic pronunciations (MAHT-mah) and *(MAHT-muh), the second one erroneous, are much more common among current loglanists, who are largely anglophones. (English, along with Dutch, German and the several Scandinavian languages, is a Germanic language.) But remember that you may always substitute the more familiar Germanic sound of (ah) for any of my (aa)s. *(MAHT-muh), however, is not an acceptable pronunciation of matma. If you're going to use Germanic (ah), you must to do so in both syllables, saying (MAHT-mah) with the (ah) of 'father' in both places.
The chief difficulty with Germanic (ah) is that it is a slack sound that easily degenerates into (uh) in unstressed syllables…especially in the mouths of native speakers of Germanic or Slavic languages. Speakers of such languages regularly and unconsciously allow this slackening to happen to many of their unstressed vowels. So, if you do elect to try to pronounce matma consistently as (MAHT-mah), the second (ah) is very likely to slip into a grunted (uh) when you're not listening. Unfortunately, hard as you try, you'll probably end up saying *(MAHT-muh)…which may cause trouble for your human auditors and eventually for your computer. If, on the other hand, you at least try for the tense Spanish (aa) at the outset, trying to say (MAAT-maa) with the stretched lips, wide-open mouth and higher pitch that the production of this Romance (aa) requires—think of a Spaniard saying ‘casa’—then your Loglan /a/ is much less likely to slump off into a grunted (uh) even when it is unstressed.
That's the main reason, then, why you should consider trying to speak the Romance dialect of Loglan: your unstressed /a/s are much more likely to remain intelligible. If you know any Romance languages, you can easily find your models for this (aa). Spanish especially is noteworthy for them. But if you are a monolingual speaker of American English, then the Boston Irish version of ‘faather’—if you can conjure it up from a recent movie—is probably your best model…unless, of course, you have The Institute cassettes, which abound in Romance (aa)s.
1.6 The TWO Spellings of ‘e’ of ‘met’: Notice that the principal allophone of /e/, which is the ‘e’ of ‘met’, has two spellings in the pronunciation guides, namely (e) and (eh). These expressions spell exactly the same sound, namely that same ‘e’ of ‘met’. Two spellings are necessary because there is a pair of phonetic contexts in which each spelling would be misleading if it were used in both contexts, yet each is a good guide in one of them. So both spellings are required, each in its own context. Thus (e) is used to guide pronunciation when /e/ just precedes a consonant, as it does in 'met' itself, unless that consonant is /r/. Thus the 'e' of 'met' in metro 'meter' is spelled (e) in the guide (MET-roh). But the same 'e' of 'met' when found with following /r/ is spelled (eh), as it is in (KEHR-tee), for example, which is a guide to the pronunciation of kerti = ‘air’. The same sound is also spelled (eh) when /e/ is final in a word or syllable. This doesn't happen in English, so we can give no English examples. But the 'e' of 'met' is final in many Loglan words. Thus, nesepe all end with the 'e' of 'met'. So the guides for these three words are (neh seh peh), which happen to be the sounds of English 'net set pet' with the 't's omitted.
To take another example, the pronunciation of the Loglan word ie—which means 'Which?'—has the value of the first two sounds in English 'ye(t)'. As /e/ is final in ie it will be written with (eh) in the pronunciation guide, which is therefore (yeh). Notice that using (e) alone in the guide to ie—that is to say, giving its pronunciation as *(ye)—would create quite a different, and for some speakers an erroneous, impression of the sound of this Loglan word. Once again, ie is pronounced (yeh), and this is identical to the first two-thirds of the word 'ye(t)'. This is a very un-English way of pronouncing a final /e/, but with a little effort you can learn to do it.
Loglan /e/ has a second allophone, of course. This allophone is the sound of (eigh) in 'eight' which you may have been tempted to give /e/ in words like nesepe, thus pronouncing them like 'nay' 'say' 'pay' incorrectly. But in Loglan, the (eigh) allophone of /e/ is reserved for positions before vowels, and it never occurs anywhere else. Thus (eigh) occurs in words like English 'Mayo' (MEIGH-oh) and Loglan eo (EIGH-oh) ('Please') because this /e/ precedes a vowel; but (eigh) is never correct for /e/ when /e/ precedes a consonant or is final in a word or syllable. Native Romance speakers are inclined to misuse Loglan (eigh) in just this way.
1.7 The Odd Spellings of /i/ and /y/: Two of the 6 vowel phonemes will seem oddly-spelled to the English-reading eye, namely /i/ and /y/. These phonemes have been given letters which commonly have quite different phonetic values in English. But the sounds of these Loglan phonemes—which are (ee y) in the case of /i/ and the "grunted" (uh) in the case of /y/—are certainly not odd to the English ear. The sounds (ee y) are in fact very common in English. Indeed, the letter [i] has just these values in many European languages, including but not only the Romance ones. On the other hand, the short, "grunted" vowel (uh) that is spelled with the letter [y] in Loglan is not consistently represented in any European spelling, the (uh)-sound being usually regarded as a "slighted", or even a "degenerate", version of some other vowel. Thus, the German letter [e] when final is always sounded as (uh), even though [e] has other values in other contexts. In English, (uh) is sometimes spelled with [e] as in 'the' (thuh), sometimes with an [a] as in ‘sofa’ (SOH-fuh), sometimes with a [u] as in 'upon' (uh-PAHN), and sometimes with an [o] as in 'phonetic' (fuh-NEH-tik); so it does not appear to "have a letter of its own" in English. On the other hand, the sound (uh) is extremely common in English. So having such a familiar sound uniquely represented by the same single letter, even an odd one, may actually be something of a relief for English-reading eyes. Note that the work the letter [y] most commonly does in English is performed in Loglan by another letter, namely [i]. This frees the letter [y] to do its present work in Loglan as (uh).
1.8 /y/ as a Hyphen: As we have seen, the sound of Loglan /y/, which is often called schwa (shvuh) by linguists, is a very short, usually unstressed, "grunted" vowel that occurs in all Germanic and Slavic languages, and in many other consonant-rich languages as well, /y/ has a very odd distribution in Loglan in that in the regular words of unbuffered dialects—see the next section for its role in the buffered ones—it occurs only as an intraverbal hyphen. In these dialects—which include the standard one—/y/ is the pronunciation of the hyphen-like letter [y] that is used to "glue" the parts of complex words together. An example is mekykiu (MEK-uh-kyoo), which means 'ophthalmologist' or 'eye-doctor'. In this word the hyphenated elements /mek/ and /kiu/ are not words, but affixes. An affix is a usually shortened representative of one of the words in the defining metaphor used to build a complex word; see Sec. 2.55 on Making Complexes. In this case, the defining metaphor is menki kicmu (MENG-kee-KEESH-moo), and mek and kiu are derived from these words by processes that we will consider later; Sec. 2.52. The segment /y/ between them turns them into a single word. The phrase menki kicmu also means 'eye doctor' but this time without the English hyphen. When /y/ is used as a hyphen in regular words, it is always a never-stressed sound between affixes. Because the sound (uh) is a hyphen it does not even count as a syllable in locating stress; see Sec. 1.11.
In the regular words of unbuffered dialects the phoneme /y/ has no other use than as a hyphen. That is to say, /y/ is always the sign of the intraverbal hyphen in these dialects whenever it appears in a non-name; see Sec. 1.12 for the uses of schwa in names and below for its uses in buffered dialects. MRH: I do note the remark of John Cowan that the more definite sound in “look” might be useful for y.
MRH: as I remark above, section 1.4 through 1.8 are fine as they are.
1.9 /y/ as a Buffer: In the buffered dialects of Loglan, /y/ may be used as a buffer between any pair of "difficult" consonants, even when that pair is wholly within an affix or a simple word. For example, in the Japanese dialect of Loglan, the word for 'mother' will probably be buffered. It will become matyma, a 3-syllable word pronounced (MAAT-uh-maa). Here the phoneme /y/ is not used as a hyphen. Indeed, matma is a simple word without any constituent affixes, so it could hardly contain a hyphen. Instead the /y/ in matyma is a buffer between the two "difficult" consonants /t/ and /m/, that is, it makes them pronounceable. (Consonant pairs are difficult when the speakers of a buffered dialect choose to buffer them by introducing /y/ between the members of the difficult pair.) If this happens, matyma will become a dialect word, that is, a harmless variant of some standard word. It will have exactly the same meaning as the standard word matma. Moreover, we expect (MAAT-uh-maa) to be as easily recognized by a speaker of the standard dialect as (MAAT-maa) is likely to be recognized by speakers of the Japanese dialect as a "variant" of their own matyma.
1.10 /iy/ as a Hyphen in Buffered Dialects: Notice that the speakers of a buffered dialect may not use /y/ as their intraverbal hyphen. /y/ has already been preempted as their buffer, and to use it for the hyphen as well would lead to confusion for the resolver as to where the "joints" in a complex word actually were. This is contrary to the spirit of Loglan, in which not only are all words uniquely resolvable, but also all terms within words. Using /y/ for both hyphen and buffer in a buffered dialect would make the term-resolver's task impossible in that dialect.
So the hyphen adopted for use in these dialects is /iy/ (pronounced (yuh)). Thus the Japanese speaker of Loglan is obliged to use mekiykiu pronounced (MEK-yuh-kyoo) for mekykiu (MEK-uh-kyoo) ‘ophthalmologist’ if da buffers any consonants at all. Of course, if dab doesn't buffer da's consonants, da is in that respect speaking the standard dialect. In that case da may use the standard hyphen /y/ without fear of spoiling the resolvability of his speech; see Sec. 2.1 on resolution.
1.11 The Effect of Hyphenating and Buffering on Stress: In the regular words of any dialect the buffer /y/ and the hyphens /y/ and /iy/ are always unstressed. Hyphens and buffers are "so unstressed", in fact, that they do not even count as syllables in locating stress. This rule permits the same syllable to be stressed in both the buffered and the standard versions of a given word, e.g., (MAAT-uh-maa) and (MAAT-maa). Indeed, because of this rule both versions of the word matma may be said to be "penultimately stressed”; the penultimate stress rule is discussed in Sec. 2.37. Although mekykiu and mekiykiu are both stressed on the syllable (MEK) in (MEK-uh-kyoo) and (MEK-yuh-kyoo), they may both be said to be penultimately stressed; for neither hyphen counts as a syllable.
MRH: I have not interested myself in hyphenated or buffered dialects.
1.12 /y/ in Names: The phonology of /y/ in names is, as usual, a different matter. In names /y/ may occur in either stressed or unstressed positions in all dialects. For example, /y/ appears in a stressed syllable in 'Hunter' => Hyntr (HUHN-trr) and in an unstressed one in 'Washington' => Ua'cintyn (WAH-sheen-tuhn); see Secs. 2.4-12 on making names.
But what are we going to do with the Japanese version of the name for Amsterdam, which my English-Japanese dictionary lists as 'Amusuterudamu'? Like all loglanists, Japanese loglanists will no doubt be making the effort to pronounce the name of this city in the way its inhabitants do. In standard Loglan, that effort produces A'msterdam (AHM-stehr-dahm), for we anglophones who speak standard Loglan are as tolerant of consonant-clustering as the Dutch are. But not so the Japanese. The habit of buffering, which is encouraged elsewhere in their dialect of Loglan, may well cause them to buffer their Loglan names too. In that case, the word for A'msterdam in the Japanese dialect of Loglan is likely to come out A'mysyterydam (AH-muh-suh-tehr-uh-dahm) or even A'mysyterydamy (AH-muh-suh-tehr-uh-dah-muh). Unfortunately this seems to generate two, or even three, different Loglan names for the same place. But the "misfortune" is only apparent. Having one, two or a dozen variant names for Amsterdam, along with the Loglan standard one with which all the others may be equated by an easy algorithm, is as unlikely to be troublesome for the mutual intelligibility of the two dialects as the fact that buffering creates two words for mothers. Each phonological variant is a synonym of the standard, and once the pattern of buffering in any given dialect has stabilized, each member of such synonym pairs will easily be recaptured from the other. So each dialect word will be as easily understood by the speakers of the standard dialect as the dialect speakers will understand the standard word. Mutual intelligibility of dialects is the aim. We think that consonant-buffering will advance this aim rather than retard it.