Setting aside the "free modifiers" of R176-9, which are quite differently distributed and composed, the (unfree) sentence modifiers which occur in Loglan utterances are of two sorts. They are either (1) phrases, such as pa da ('Before X'), in which some PA-word functions as a preposition and some argument, in this case a very simple one, functions as its "object" or operand; or they are (2) argument-less PA-words, such as pa itself, which function in the sentence like English "adverbs". Thus pa without an argument is an ellipsis for 'Before some assumed time, say the moment of speech' and thus means simply 'Earlier'. Group F is exclusively concerned with the construction of these two kinds of modifiers. Its first grammar rule, , gives the basic architecture of Loglan modifiers, which is exceedingly simple. Notice that both the pa and the pa da varieties of modifiers may be optionally followed by the triplet, the choice of gu, pause/comma or nothing at all, which is one of the optional punctuators developed in Group A.
=> PA gap
Mu titci pa = We eat earlier.
=> PA argument gap
Mu titci pa la Ven = We eat before nine. (When a modifier comes at the end of an utterance an explicit gap would be redundant and so is not used.)
Fa = Afterwards. (Later/then/etc.)
=> M4 NO mod
No fa = Not afterwards.
The parser needs to know what kind of no this is; so the preparser has looked ahead of this no (as the parser can't) and found that it is negating a modifier. So the preparser erects the signpost M4 before the no to inform the parser of its discovery. (The preparser is thus a kind of forward scout for the machine's parser…which snuffles along blindly—but very mentally—with its nose to the ground.)
=> KA modifier KI mod
Ke fa ki pa da = Both afterwards and before X.
=> M4 NO kekmod
No ke fa ki pa da = Not both afterwards and before X.
Pa da = Before X.
Kanoi fa ki pa = If later then earlier.
=> modifier M5 A mod
Fa noa na = Later only if now.
Two machine lexemes are introduced in this short group, M4 which marks those no's that negate modifiers, and M5 which marks the eks that connect them. The keks of R63 do not need to be marked.
Group G. Arguments
This is the longest rule group in the grammar, comprising nearly a quarter of all its rules. But the argument rules nevertheless constitute a coherent group, with few gramemes used elsewhere until we get to itself. Once developed, arguments are then used very widely throughout the grammar…for example, in the preceding group to make modifying phrases.
The question may well arise: Why should not arguments precede modifiers in their development? The answer is that arguments use modifiers in their development as well, and that arguments unlike modifiers have a great variety of backward references. They are, recall, used as early as Group B, Linked Arguments. It seemed best to put them in a position where all they will use has already been built; and then let them have all the backward references they require. In short, arguments are the most circular structures in Loglan grammar in that they are more frequently embedded in themselves.
An argument is a designation, and in Loglan as in the natural languages there are many kinds of designations. A preview of what is to be found in this large and important rule group may therefore be in order. We will commence with "names" (R68-9, 80), and move on through "definite descriptions" (R72-6, 83) and the various kinds of "quotations" (R84-6). Then we will provide for "event descriptions"—called "indirect quotations" in European grammar—whose handling in Loglan (see R87) expresses one of the fundamental philosophical orientations of the language (namely that objects of thought and imagination are best treated as a subset of definite descriptions). Then we provide for the "indefinite descriptions" (Ne mrenu vs. Le mrenu) in R97-100 and 102; and finally the various ways of connecting arguments are given toward the end of the group; R103-16.
La Djan, ditca = John is a teacher.
=> name DJAN
La Djan Djonz, ditca = John Jones is a teacher.
The commas in S68-9 represent "morphemic pauses", that is, the pauses required after names which permit their resolution. But this is a morphological matter, and the grammar pays no attention to such pauses. In fact the preparser eliminates all morphemic pauses before presenting a string to the parser. The morphemic pauses between the parts of a multiple name are necessary in careful speech—otherwise the lexer would hear them as single names, e.g., Djandjonz—but such internal pauses are not represented by commas in text. It would distract the eye to do so.
Ra da ditca = All of them are teachers.
Ne da ditca = One of them is a teacher.
The grammar rule will be much more elaborate once the mathematicians set to work on it. But I have thought it best to keep rudimentary until the rest of the grammar is functioning smoothly.
=> LE descpred
Le ditca pa gudbi = The teacher was good.
=> LE mex descpred
Le te ditca = The three teachers.
=> LE arg1 descpred
Le le ditca gu bukcu = The teacher's book.
=> LE mex arg1
Le to da ditca = The two of them teach.
=> LA descpred
La Ditca ga sadji = Teacher is wise.
Note the "possessive" provided in R74. An alternative possessive form will be provided in R88 in which the order will be possessor-possession, as in the Romance languages: 'El libro de la maestra'. Like English ('The teacher's book' and 'The book of the teacher') Loglan has both possessive orders. As usual our policy is to provide ways of speaking every common natural word-order and let usage decide between them.
Note also, from R76, that descriptions may be used as names. When this is done the predicate words in the la-marked description are all capitalized in text, and the expression has, like other names, the sense of being a unique designation in that context. Thus there is only one la Farfu in the Smith family, just as there is only one Rat in The Wind in the Willows. Names may not be used as predicates, however. Thus *Da Djan cannot be used to mean 'X is a John' because this expression is not parsed as a sentence. (It is *-ed because it lacks a pause-comma before the name.) The Djan in Da, Djan as parsed by our grammar is neither a name nor a predicate but one of the free modifiers. It is a vocative, in fact, which modifes Da. So the proper rendering of Da, Djan into English is 'X, John' as if one were answering a question asked by John about the identity of something. The proper rendering of of English 'X is a John' into Loglan is Da mela Djan. In this expression the predifier me turns la Djan into a predicate (see R24); and Da is the first argument, or "subject", of that predicate.
Da ditca = X teaches (is a teacher).
Hu ditca = Who teaches?
Tai ditca = T teaches.
=> LA name
La Tam, bi ba = Tom is (that) someone.
=> LIO TAI
Lio nei cmalo = The number n is small.
=> LIO mex
Lio te cmalo lio nei = The number three is smaller than the number n.
=> descriptn gap
Le ditca ga gudbi = The teacher is good.
=> LI (utterance LU)
Li, Le ditca ga gudbi, lu steti = Quote Le ditci ga gudbi close-quote, is a sentence. (The commas here are stylistic only, i.e., neither morphemic nor lexemic.)
=> LIU (word)
Liu lu purda = The word lu is a word.
=> LIE (X, string, X)
Lie Dai, Deutschland, Dai dotca purda = The string 'Deutschland' is a German word. (These commas are obligatory. 'X' is any Loglan word that does not appear in the quoted string. I usually use the initial letter of the Loglan word for the language in which string is written, in this case D for dotca.)
=> LEPO sentence gap
Lepo da pa cluva de, viekli = That X loved Y is clear.
The parenthetic portion of R84 is carried invisibly through the first parsing of the utterance and then parsed "re-entrantly" as an independent utterance. The parenthetic portions of R85-6 are not even seen by the parser. It is assumed that they either do not need to be (R85) or cannot be (R86) parsed.
=> JI argument
Da ji la Djan, ditca = The X who is John, teaches. (Ji identifies.)
=> JI modifier
Da ja vi le hasfa ga kicmu = X, who is in the house, is a doctor. (Ja predicates.)
=> JIO sentence gap
Da jio brudi de, murpeu = The X who is a brother of Y, is a seaman (seaperson). (Jio identifies.)
Pe is also an allolex of JI and, like the 'de' of Romance-style possessives, puts the possessor last. Thus the phrase pe le ditca in Le bukcu pe le ditca ('El libro de la maestra') is also a production of R88.
Da jia sucmi = X, who (incidentally) swims. (Jia predicates.)
=> argmod M6 A argmod1 gap
Da jia sucmi, e jia prano - X, who (incidentally) swims and runs.
Da = X.
=> arg1 argmod gap
Da ji de = The X who is Y.
Sai = S.
=> mex arg2
Te Sai = Three of the S's.
=> mex descpred
Te solda pa kamla = Three soldiers came.
=> indef1 gap
Te solda ga ditca = Three soldiers are teachers.
Either the comma or the gu of can be used here in place of ga to separate the first argument from its predicate. In fact *Te solda, ditca parses in substantially the same way as Te solda ga ditca does. But I have starred it because it is probably not a good idea to adopt this usage, despite its tempting economy. The reason is that if the first argument ends in a linked argument, will not successfully separate it from an upcoming predicate; instead will allow the intended predicate to be absorbed into its first argument. For example, Le farfu je le botci, talna does not say that the father of the boy is an Italian, as a careless speaker might have intended; instead it is a designation of some boy-fathering Italian: 'The father-of-the-boy (type of) Italian'. Replacing the comma with ga conveys the intended meaning unequivocally: Le farfu je le botci ga talna - 'The father of the boy is Italian'. Thus the stronger marker ga is necessary in some cases. Since ga will work in all cases and only in some, it seems wiser for loglanists to habituate themselves to the always-successful marking move…even though ga is less economical than the pause/comma permitted by that would often do the job.
The frowned-upon uses of form another unused portion of a superset of grammatical possibilities.
Te solda = Three soldiers.
=> indef2 argmod gap
Te solda jia nigro = Three soldiers, who are (incidentally) black.
Bai = B.
Ne fumna = A (one) woman.
=> arg4 ZE arg3
Bai ze Cai = B and C jointly.
=> arg4 ZE indefinite
Bai ze ne fumna = B and a woman jointly. (Ze among arguments has the effect of forging single, teamlike entities.)
Bai = B.
=> KA argument KI arg6
Kanoi Bai ki Cai = If B then C.
bei groda cei = b is bigger than c.
=> DIO arg6
Mau cei gi bei groda = Than c, b is bigger. (Mau is one of the optioned case-tags of the DIO-Lexeme. Derived from cmalo it marks the lessers in "greater-lesser than" relationships. For the uses of the fronting operator gi, see R165-6.)
=> IE arg6
Mau ie cei gi bei groda = Than which c is b bigger? (DIO-words are always omissible; Ie cei gi bei groda means the same thing. Yet mau is not redundant. It suggests in advance the kind of predicate that is coming up.)
=> LAE arg6
Donsu da lae bei = Give X to whoever has address b (or of whom b is a sign in some other sense).
bei groda cei = b is bigger than c.
=> arg6 ACI arg7
bei groda cei, e dei, onoici fei = b is bigger than c, and than d or f but not both. (Onoi is "exclusive or" and -ci turns its connectands into the right connectand of e, i.e., it hyphenates them.)
bei groda cei = b is bigger than c.
=> arg8 A arg7
bei groda cei, e dei, onoi fei = b is bigger than c and d, or f but not all three. (Without -ci these "afterthought" connectives group left.)
bei groda cei = b is bigger than c.
=> arg8 AGE argument
117; 60, 24, 14, 8.
bei groda cei, ege dei, onoi fei = b is bigger than c, and d or f but not both. (The -ge suffix groups all that follows it to the right, and thus has the same effect as a -ci suffix on the 2nd connective; cf. S112.)
Notice that has only one forward reference, and that is the first rule in the next group. Most references to are backward, which only means that arguments have already been widely used. Indeed, we may recall that they have been used to construct one class of modifiers, namely prepositional phrases (R60); they have been used to build one kind of predunits, namely those composed of arguments preceded by the predifier me (R24); and of course arguments figured in linked arguments (R8 and 14).
The single forward reference to R117 leads of course to the most common use of arguments in Loglan utterances, their contribution to the "terms" and "termsets" which are major constituents of nearly every Loglan utterance.
Group H. Terms & Term Sets
A term is either a modifier or an argument. So terms are strings of one or more arguments and/or modifiers in any order. Term sets are strings of none or more terms and include the connections of such strings with one another. Group H is really two groups, both of them very small. One of them, R117-20, constructs ; the other, R121-7, constructs . The uses of these two gramemes are as follows:-
Terms are used as the first arguments ("subjects") of sentences, and include any immediately following modifiers. Thus the argument-modifier pair La Djan, na la Fomen in La Djan, na la Fomen, traci ti = 'John, in May, travels here' is an instance of . But so also are the strings of fronted modifiers and/or arguments which come before the subject of a sentence. The only difference is that terms that come before a subject must be set off from it by gi. For example, Ti na la Fomen in Ti na la Fomen, gi da traci = 'Here in May, X travels' is also an instance of . Gi is the fronting operator; it announces that a string of such fronted terms has been concluded.
Termsets, in contrast, are always deployed after their predicates. For example, ti in Da traci ti is a . Indeed, all sets of "sutori" (second-and-subsequent) arguments, including any sentence modifiers with which they may be mixed (ti fa in Da traci ti fa - 'X travels here later'), or any strings of modifiers, or of arguments and modifiers mixed together, when they follow their predicates—for example, as fa, na la Fomen in Da traci fa, na la Fomen ('X travels later, in May') follows traci—are instances of .
As we will see, termsets are made of terms. But there are two differences in the resulting structures. First, termsets, but not terms, may be null, i.e., represented by the right boundary marker gu or by nothing at all; see R121. Second, termsets (but again not terms) may be connected to each other. Just why these two manoeuvres are useful can probably only be discovered by using them. But I can give you a preliminary glimpse of their utility for termsets by pointing out that they (and not terms) are grammatically attached to sentence predicates; see R128. These predicate-containing constructions then, together with their possibly null termsets, become the "predicates" of Loglan sentences; and such predicates may be connected to each other in every possible way…including ways which allow them to have "joint termsets". For example, in the English sentence 'John loved and hated Mary', 'Mary' is probably intended by the speaker to be a joint termset of the connected pair of predicates 'both loved and hated', that is, to be the "direct object" of both "verbs" (to use an older grammatical terminology). Of course we cannot be sure of this in English; but in Loglan we can be. It is just this kind of potentially ambiguous construction that in Loglan necessitates either nullifying or truncating individual termsets in order to make room for joint ones. For example, the two possible renderings of 'John loved and hated Mary' into Loglan are (i) La Djan, pa cluva, e tsodi la Meris, in which la Meris is an individual termset, and (ii) La Djan, pa cluva, e tsodi gu la Meris, in which it is a joint one. What has happened in (ii) is that in it e and gu have nullified the individual termsets of cluva and tsodi, respectively, and made room for a joint one.
Such manoeuvres are never required of terms. But if this structural distinction seems too intricate for easy recall, you may prefer to remember a simpler, positional one: terms occur ahead of their predicates; termsets occur after them. This surface property will take you a long way toward managing their uses properly.
Da pa cluva de = X loved Y.
Da pa cluva fa = X loved later.
Da pa cluva de = X loved Y.
=> terms term
155, 157-8, 165-66, 171.
Da pa cluva de fa la Ven = X loved Y after nine. (The order of terms is syntactically unimportant. Thus Da pa cluva fa la Yen, de means approximately the same thing.)
Notice that all these references are forward. In effect, they list the many uses of . R155 deploys it as the delayed subject of "goa sentences" (sentences in V-O-S word order); R157 uses it as the fronted modifiers of a goa-sentence; R158 uses it as the subject, accompanied by any following modifiers, of a declarative sentence in Loglan-normal (S-V-O) word order; R165-6 arranges for its use as the fronted arguments, with accompanying modifiers, of sentences in O-S-V word order, but also as the fronted modifiers or prenex quantifiers of any utterance; and in R171 accommodates fragmentary utterances, for example, answers to Hu, Vihu and Nahu ('Who?', 'Where?' and 'When?') questions.
We now use to develop .
Da pa cluva de, e tsodi gu fa la Ven = X loved Y and hates (someone), (both) after nine. (Gu represents the null termset and turns fa la Ven into a joint termset. A comma is insufficient here; therefore this allogram is not but .)
=> terms gu
Da pa cluva de fa la Ven, e tsodi di gu va do = X loved Y after nine, and hates W, (both) near Q. (Here gu truncates the 2nd termset and prepares for a joint termset, va do, a modifier of both predicates, to be spoken.)
Notice that it is and not that is used to truncate or nullify an individual termset. This is because the use of a comma or a pause at such points can lead to ambiguities. So it is either gu itself or, when it is not required, nothing at all that terminates a termset.
Da pa cluva de fa = X loved Y later.
=> termset2 A termset1
Da pa cluva de fa, e di pa = X loved Y later and W earlier.
=> KA termset2 KI termset1
Da pa cluva ke de fa ki di pa = X loved both Y later and W earlier.
Da sanpa de di vi do = X is a sign of Y to W at Q.
=> PAUSE termset2
128, 139, 141, 148, 154.
Da sanpa, de di vi do = X is a sign, of Y to W at Q. (This discretionary pause-comma in a place where does not occur must be explicitly provided for.)
Predicates are equipped with individual termsets in R128 and 148, and with joint termsets in R139, 141 and 154. All these applications of take place in the next group, which constructs predicates. Termsets are used for no other purpose in the grammar than to construct predicates.
Group I. Predicates
In this group, the predicate expressions that make the claims of sentences are constructed. Basically they consist of a sentence predicate from Group E coupled with a termset, possibly null, from Group H; this coupling is accomplished in R128. In the sequel we'll call such expressions simply "predicates". Predicates may be "marked" or "bare”, that is, prefixed by tense or abstraction operators (which includes the left-marker ga) or not so prefixed; see R129-31. This distinction is fundamental; it involved building two tracks through the grammar of predicates, one for bare forms, one for marked ones. For example, in a long sequence of rules, R132-42, a distinction between "backpreds" and the two kinds of "front" predicates, "barefronts" and "markfronts", is gradually developed; and in R137 and 140, this distinction is finally used. It turns out that the forms called backpreds are going to be used as the right connectands of both kinds of ekked connections, the "bareekpreds" and "markekpreds". These are the bare and marked versions, respectively, of ekked predicates, that is, of connected predicates in which the connections are made with afterthought or A-form connectives. Barefronts and markfronts, of course, are destined to be the left or leading connectands of these same ekked predicates; and the ekked connections themselves will of course reflect the "bareness" type of their leading elements.
By R143-54, the bare vs. marked distinction is no longer important. Identity predicates ("identpreds") are now developed—BI and kin may not be treated as "just another PREDA" because they enter into little word compounds and must be recognized by the lexer—and finally kekked, or forethoughtfully connected, predicates ("kekpred") are developed. Along the way, in R135-6 and 146-7, the ACI- and AGE-forms of afterthought connections are developed in the same way that they were developed for arguments.
=> sentpred termset
Da fumna = X is a woman. (Here we are using the null termset , which, because it is final, may be expressed by , i.e., omitted altogether.)
=> M7 PA barepred
Da pa fumna = X was a woman.
=> PO gap sentence gap
Da po de fumna = X is a case of Y's being a woman. (Both s may be null here, the 2nd because it is final, the 1st because there is no following
for the normally close-binding 'po' to bind to. Po will not "stick" to de, and so does not need to be separated from it.)
Da fumna, e ditca = X is a woman anda teacher. (Both these connectands are barepreds.)
Da fumna, e pa ditca = X is a woman and was a teacher.
=> M8 NO backpred1
Da fumna, e no ga blanu marpi = X is a woman and not a blue snake. (Here ga extends the scope of 'no' over the whole backpred. Ge would also work but is bad usage.)
Da fumna, e ditca = X is a woman and a teacher. (Backpreds are always right connectands.)
=> backpred1 M9 ACI backpred
Da fumna, e ditca, aci stude = X is a woman, and a teacher or student. (Aci binds ditca aci stude into the right connectand of e, thus altering the normal left-grouping of afterthought connection.)
=> barefront M10 A backpred
Da fumna, e ditca, aci stude = X is a woman, and a teacher or student. (This is a bare "ekpred" because its leading element is bare.)
Da ditca lo dotca = X is a teacher of German.
=> bareekpred termset
Da ditca, e stude gu lo dotca = X is a teacher of, and a student of, German. (Again gu prepares for the joint termset.)
=> markfront M10 A backpred
Da pa ditca, e cluva = X was a teacher and a lover. (This is a marked "ekpred" because its leading element is marked.)
Da pa ditca lo frasa = X was a teacher of French.
=> markekpred termset
Da pa ditca, e cluva gu lo frasa = X was a teacher of, and a lover of, French.
Da ditca = X is a teacher.
Da pa ditca = X was a teacher.
=> M8 NO predicate2
Da no ga blanu tcaro = X is not a blue car. (This is long-scope negation. X may not be a car of any kind.)
Ditca = Be a teacher!
=> predicate2 M11 AGE predicate1
Gudbi, ege ckano, a briga = Be good, and be kind or brave! (If e replaced ege, this would read 'Be good and kind, or be brave!')
=> BI termset
Da bi le mrenu jio pa godzi la Frans = X is the man who went to France.
=> NO identpred
Da no bi le mrenu jio pa godzi la Frans = X is not the man who went to France.
=> M3 KA predicate KI predicate
Ke ckano ki briga = Be both kind and brave!
=> NO kekpred
No ke ckano ki bunbo = Don't be both kind and a fool!
Da mrenu = X is a man.
Da bi le mrenu = X is the man.
=> kekpred termset
155, 158, 162.
Da ke briga ki bunbo gu raba = X is both brave (about everything) and a fool about everything.
Predicates, once developed, are used in constructing sentences. All three of these references are to the next group, in which sentences are made. In R155
is used to make the Goa-sentence; in R158 it is part of the declarative sentence in normal word order; and in R162 it is the whole of the Loglan imperative.