from the paper archives - pc on abstractors and tense

[Logical Language Group <lojbab@ACCESS.DIGEX.NET>]

On a hunch, I wandered over and looked into the archives, and found
something relevant - the birth of the Aristotelian abstractors in NU, as
well of the current Lojban tense system.  This may explain a lot of
questions, and incidentally give Cowan some good words for the reference
grammar, because I don't think even he has read this before.

Over the course of 1987, most of our work was the production of the
gismu list for what was then called Loglan-88.  However, based on notes
from pc about each TLI cmavo, I also prepared a list of desired
Loglan-88 cmavo to be assigned where possible based on the new gismu,
and also to set of systems for memory hooks e.g. se/te/ve.  This was
obviously an iterative process until the gismu list was completed just
before Xmas 1987.  Many of the cmavo, were however, assigned before my
wedding to Nora which used vows written using a hybrid with mixed TLI
and Loglan-88 words.  For the most part, all of the early cmavo were
either direct crossovers from TLI lists, or were derived from proposals
made by Jim Carter and by others during the TLI period.

Around January 1988, shortly after the gismu list was finished, I
prepared a first draft cmavo list (reported in JL5).  I don't see a
printed copy of this in the archives, though pc may have it, since he
commented on it.  It may also be on my long-unused 8086-based computer,
which I was still using back then.  Maybe I'll be able to offload the
files from that machine for historical purposes.

His extensive comments on that list were received in a letter dated 28
Jan 1988.  In the discussion of abstractors, his comments indicate that
"jei" was already there, and we had added "ce" for "process
abstractions".

The latter was added on the basis of a then-recent paper pc had read and
sent me a copy of:  _Principle of Abstraction for Events and Processes_
Peter Roeper, Journal of Philosophy of Logic 16 (1987) 273-307.  I
hadn't been able to understand the paper, and asked pc to explain it to
me on the phone.  I don't remember the details but I have the paper
(which I still don't think I can understand %^).  The paper makes a
distinction between event clauses and process clauses, and on the basis
of pc's explanation, I misunderstood that we needed to make the
distinction between "nu" and something else which I assigned to "ce".

pc's comments in his letter (edited, * indicates my emphases):

>I am also doubtful about the need for a special operator for process
>abstraction.  Processes are just events with certain special attributes.
>Could we have the same sentence abstracted by ce and nu to different
>results?  The closest I can come to an affirmative answer is those cases
>where English uses almost the same words to describe a proces and the
>state of habitually being involved in that process:  "He is building a
>house."  (process) and "He builds houses" (state - probably) I think
>Loglan has (*well, should have) better ways of dealing with this
>difference.

[pc has stated the problem - it was up to me to find a solution, is
basically the way I responded to this]

For the record (because it will turn up again when we get to tenses
(properly speaking), the types of event usually distinguished (the set
goes back to Aristotle; I'm not sure how I-E it is, rather than neutral -
the usual tests are linguistic.) are states, activities, achievements,
and processes.

[pc then went into a lengthy definition of each of the 4 types, with
examples.  Given the length he went to, his statement that this
classification had been standard since Aristotle, and that he believed
it to be neutral, rather than I-E biased, my brain started working
feverishly %^)]

>States start fully in place, last for a period without any internal
>differences (so far as the state is concerned) and end without any
>change either.  Activities can't be said to be going on (and they do,
>rather than just last) until they have already been going on (huh?!  You
>can't say "I am running" until you can already say "I have run" - think
>about it) and they stop in a similar way.  Activities tend to be
>composed of cyclic events often with the same name as the longer
>stretch, though otherwise identifiable - running so long as you keep
>repeating the little-running pattern.  Acheivements are instantaneous,
>they can't go on at all, they tend to be the beginning or end of another
>kind of event, often with some other conditions present (e.g. winning a
>race is the end of running a race plus a further condition about being
>the first to do so.)  Processes have a definite beginning and a definite
>finish, though they may end before they are finished, and they go on by
>a series of different subprocesses and activities, not cyclic with
>respect to the overprocess.  So building a house may encompass building
>a wall (another process) which encompasses the activity of laying
>bricks.  A process is not perfected while it is still going on (you
>can't say "I have built this house" at the same time as you can say "I
>am building this hous" - in contrast to an activity.)  The goal is
>decisive for processes, apparently, since adding one can make an
>activity into a process - running is an activity, running a mile is a
>process (and running a mile every day is an ac tivity again - the
>element of repetition.)  The thing about processes is that they can be
>finished, or they can stop short of finishing, or they can be carried on
>- or rather the corresponding activity can - beyond the goal (He stopped
>running/ran the mise/kept on running.)

[Several pages later, he did get to tense, and gave me the following
brief course in his professional specialty.  I know I had a phone con
with him discussing all this, and am quite sure that I suggested having
cmavo for the 4 Aristotlian event types by the end of that conversation,
or perhaps the next day, as a solution to the "problem" he cited above.]

[I know that as soon as the idea of putting the 4 types into the
language came into my head, I saw that by the standard rules of Loglan,
any predicate could be abstracted using any of the four types of
Arisotelian abstractors.  I asked pc, in the course of discussing
aspect, whether this would work, whether you could use an arbitrary
Aristotelian abstractor with an arbitrary predicate.  We discussed "run
a race" and "beat the dog" as specific predicates because these came up
in the aspect discussion, and by the end of the discussion, the
Aristotelian operators were solidly ensconced in the language design as
the rational basis for both our tense and our event operators.]

[Here is that all-inspiring lecture on tense, from which so much of the
Lojban design (especially the parts that differ from TLI Loglan) spring.
Note that is comments about the then-Lojban system how to say various
things may not be completely accurate because in 1988 our tense grammar
was very close to TLI's, and had no contours,and a totally different
system of expressing distance/interval/direction than at present.]

>A Lecture on Tense (Oh Goody, goody!)  Although we keep talking about
>tense being about time, many of the features of the actual tense system
>(however it may be manifested in some actual language) can be most
>easily accounted for by thinking of psychological operations:  recall
>and anticipation, remembering and expecting.  These pairs differ in
>"liveliness" and permanence of focus:  recall may involve recalling not
>only the external events of a time but the internal as well, including
>remembering or expecting other events at that recalled time.  And
>similarly for anticipation.  However, because of the holistic nature of
>these acts, a recall of a recall collapses into a simple recall of the
>second (recalled recalled) event.  Anticipation does not seem to bear
>repetition at all, since the anticipated event is not yet real and so
>does not allow of complete richness (though we can still remember and
>expect in anticipation, we cannot recall nor anticipate, except to move
>to a leter anticipated avent.)  In short, we have four possible points
>of refernce - the present, an anticipated one, a recalled one, or one
>that we recall anticipating.  At each of these, we can look at what is
>going on or remember or expect other events.  The backcurving cases -
>anticipated remembering, recalled expectations and especially recalled
>anticipations and their complexes - are expecially untidy from the point
>of view that takes tense to be about time.  For, in most cases, the
>event turns out to be prior to the present time.  I can, in English, say
>"He will have come by tomorrow" only if I do not know that he has come
>already - even if it later turns out that he had indeed already come.
>For the opposite direction, note the difference between "He said he
>would be here by now" and "He said he will be here tomorrow", passing
>through not cancels out the use of retro form.
>
>Of course, one of the things about a recalled anticipation is that the
>anticipated events never come to pass (in our time line at least, if we
>want to preserve the time talk) and so these locutions slide over in
>many languages into ways of talking about things that do not happen even
>now or after - English "would" for contrary to fact conditions.
>
>In a language with obligatory tense markers, where every sentence is
>marked for tense, life is fairly easy.  We set a time reference and all
>subsequent sentences carry that mark, at least relative to the now.  If
>a new time is set, it must be done fairly explicitly and we immediately
>get a new mark (or the old one clearly referred to a new time in the
>same general direction.)  We can wander from that time by remembering or
>expecting fairly easily, for we always have the flag to remind us where
>we are based.
>
>Loglan doesn't have obligatory tense markers.  This creates no problem
>so long as we keep to running narrative.  We start at a time, given
>implicitly or explicitly at the beginning, and we stay there (or we move
>somewhat with the flow of the story) to the end.  But, if we need to do
>flashbacks or even to mention a recollection or expectation relevant to
>that action, we do have to put in some signs for the moves.  With
>literal remembering or expectation, we could do the whole thing within
>the ongoing narrative, in a "lepo" clause attached to the appropriate
>predicate, for example.  And the longer flashback - or flashahead -
>could be done with an explicit change of time base.  But the first often
>is narratively incorrect - no person in the story actually recalls the
>relevan fact but the narrator needs to stick it in to explain the next
>event.  And the second presents real problems in getting back tidily to
>the old point after the flash is over.  But this second problem is one
>that no language hanles well (since none carries more than one referent
>and now in full form) and so it should not be taken as a special problem
>for Loglan or other untensed languages (but it would be nice to solve it
>somehow.)
>
>We have in Loglan two structures, simple tense worlds an compounds - of
>any level, but, in fact, nothing more than three occurs in natural
>languages.  We also have a number of functions to perform:  setting a
>reference, moving from a reference briefly, changing a reference, and,
>perhaps, stressing a reference.  We would also like to be able to move
>off and return to a reference fairly efficiently.  And we would like to
>do these in such a way that a person may choose to tense almost every
>sentence (untense is not obligatory either - though preferred style,
>perhaps - and since we usually have to have something there, why not
>"pu" instead of "cu"?) without wandering off in unexpectied ways.
>
>If we take this last requirement seriously, we have to allow a running
>tense marker.  And here we have two choices:  either we use the marker
>appropriate for that point as referring to the fixed point and adding
>where that lies relative to now or we use "ca" to be a running marker to
>indicate "at the established reference point".
>
>In the latter case, we would need a more complex form to spring from
>some remote axis to the present (a not uncommon move in many kinds of
>discourse).  We could then also use the simple forms for quick moves to
>remembered or expected events without losing the continuing reference.
>This, in turn, would requiremore complex forms for shifting to new axes,
>since a simple form would not be unambiguous.  Of course, axis shifts
>can be done by means other than the tense forms - by tense markers as
>discursives or by explicit reference "when he did so and so" "shortly
>thereafter", and so on, but we are presently dealing only with
>systematics.  The use of simple forms for remembering/expecting is not
>necessary, we could use "capu" and "caba" for these, as well, and thus
>use simple forms for most axis shifts.  But this would not fit in well
>with the (preferred) tenseless form, for all vectors would then need a
>"ca" which was not required ordinarily in ongoing axis reference.
>
>The other possibility is that "pu" (and the same gois for "pa" or "ca",
>of course) is the form used simply ongoing past reference when a tense
>is used (either habitually or in the habitual renseless forms for
>emphasis.)  In this case, the small shifts, vectors, would require a
>>complex form - the original reference andthe vector, "pupu" for the
>pluperfect, even in an ongoing past context.  An ongoing discussion at
>the recalled anticipated axis (a rare thing, to be sure) would require
>an ongoing "puba".  The jump from remote to present would take only a
>simple shift to "ca".  But other shifts would again be complex.
>
>Since within the body of text the frequency of various uses is clearly
>mainly ongoing axis, some vectoring and only an occasional axis shift
>(with shifts to the present no obviously more common than other moves),
>the first discussed system seems the most efficient:  "ca" or unmarked
>continuing axis, single tenses for vectors, and compound forms for axis
>shifts.
>
>This leaves three questions:  what compound forms for shifts? how do we
>introduce axes originally (if different from how we shift)? and can we
>reclaim old axes rapidly after a while on a shift?  Every way I have
>tried for this last question has failed, so, unless you - or someone
>else can see a cute way - the answer is just NO.  To move back, we need
>a "Back at the time when..." explicitly.
>
>For the shift of axes, the obvious solution is "-ca", the vector that
>never really gets used (unlike "-ba" and "-pu", which appear, minus
>their prefixes, as the normal vectors.)  So, "puca" sets up a new past
>axis.  But, past to what - the old axis (wherever it was) or the
>present?  I think we usually want to do the former of these when the
>narrative axis was past already (and similarly for futures from futures)
>and, when the new axis is oppositely placed form the old axis as the old
>axis is from the present, the new axis is still placed - at least
>ideationally (regardless of actual time) in the same direction form the
>present as the old axis:  "baca" from a past axis gois to a point
>between the old axis and the present (at least not known to be later
>than the present.)
>
>(This reminds me, Loglan doesn't really seem to have a subjunctive of
>the contrary-to-fact sort, since the "suppose that" carries no
>commitment of this sort (of course neither does the English
>necessarily).  I don't know whether we need this or not.)
>
>The jump to the present is simply "caca".  And, based on that, we could
>do more complicatedjumps to pints related in unusual ways to the present
>- to a future axis from an old past by "cabaca" or some such thing.
>
>The introduction can be done even more simply, assuning we don't do it
>just by implication or explicit "when the icehouse burned" type things.
>The simple form will do, either as a tense or as a sentence modifier.
>We can even start off on a vector, attached to an axis:  "The bot had
>("pupu") been runnning for miles" and then the story picks up at that
>time.
>
>As I said, this is not the only way to do these things, but it looks the
>most efficient and involves a fairly thorough use of the resources.  It
>also accommodates a lot of old-language habits from a variety of old
>languages - I think.
>
>How come the place "tenses" aren't allied with predicates?  Surely there
>is a "here/there/yonder" set (memory presents me with something like
>"zvati" and "darno" though nothing in the middle - and I haven't checked
>these guys out for Loglan form, though I'm sure that the semes are
>primitivized.)
>
>I'm not going to get into the remoter regions of PA - remote fom tense
>properly speaking, as it were.  But I do want to say a bit about
>aspect(s) here and, in the process ask about some missing forms (well,
>forms I don't see - they may not be possible and so not missing.)  These
>are the forms for certain dinds of distribution aspects, of the type
>NIPA and SAPA:  "once", "never", "often in the past" and so on.
>
>I don't know a good definition of "aspect".  It covers a whole slough of
>thinks that people want to say that are more or less time related and
>that often are dealt with in places like those used for tenses.  For the
>moment, I want to deal with two and a half such notions, contour aspect
>- both extensional and intensional - and distribution aspect.
>
>My question above is about the latter, which talks about how various
>occurrences of what can be considered the same event (type) are
>distributed (when the same sentence is true.)  These range form Never
>("noca") on the one hand to Always ("roca") but cover more specialized
>cases, like "at least once in the past", which might be just "pu" or
>"roipu" to be sure.  It also includes complexes of these, the famous
>"pujeba" of Arthur and so on.
>
>Distribution also has an intensional side to it - the repeated events in
>the past, for example, ("ro'opa", say) may be just something that
>happens over and over or it may be a habit or discipline.  (We also, at
>this point have to distinguish between two kinds of repetitions, loosely
>cyclic or on occasion ones and non-cyclic or various occasions.  "He
>struck her repeatedly" - sorry about the old example - may be "flailed
>away at her" or "struck her once just about every time he saw her".
>Note the ever popular - with linguists - "he struck her repeatedly
>repeatedly."  We are concerned only with the second here, though the
>first needs some representation as well.  We have the same problem with
>the superfective - overdoing it on an occasion or overdoing occasions:
>"He kept on running after the race was over" and "He kept on running
>after he turned 95".  And the flatly ambiguous "He kept on hitting her
>after he had been told to stop."  I wonder why linguists haven't gone in
>for "He kept on keeping on hitting her"?  I have no ideas about this
>first problem.)  Even more precise distributions seem to have two sides:
>"always" may be just a factor it may be a law, and similarly never and,
>of course, a unique event may simply happen only once (as all events
>strictly do) or it may be an Occasion (theology gets into this one as
>badly as science into some of the others.)
>
>But the real intensional stuff is in countour aspect.  Most events
>extend through time a bit (last, tike, go on, ... different locutions
>for different cases, perhaps.)  Extensionally, an event begins at a
>point (or not quite one) before which it isn't and after which it is
>(and what it is AT that point varies, perhaps) and it ends at a mirror
>image point.  Between these points, the event is "going on" and can be
>referred to with present tensed forms; outside it is not and can only be
>referred to with approprioate non-present tensed forms.  Strictly
>speaking, from almost everywhere on the inside, we can also refer to the
>event with non-present tensed forms as well, since it is there before
>and after, except at the very ends.  And at the very ends it is often
>not clear just what to say:  as he crosses the finish line it seems
>wrong to say either that he is still running the mile or that he ran the
>mile.  The appropriate thing is the middling "he has run the mile",
>where this is no longer just a matter of axis and vector.  Or perhaps we
>need specila "begin" and "end" marks.
>
>But what can be done "strictly speaking" is not what gets done in
>natural languages.  I've already mentioned the four kinds of events and
>womething about the differences among them.  And all of these factors
>and more come into play at the intensional level.  First, before some
>events - especially processes and achievements - they are viewed as
>already present ("in their causes" as it is said.)  This inchoative
>aspect may even constitute an event - a state or activity which goes on
>for while or it may be a simple matter of referring to the coming event.
>Similarly, after the event it may continue to be present in its effects,
>again constituting a state (at least) as well as a back reference to
>event itself.  And, of course, there are the several sorts of
>transitions into and out of events.  States just switch on and off
>conceptually (even though becoming blue - sliding up the fuzzy value
>curve - may take some time, the final step from non-blue to blue is
>treated as instantaneous.)  Activities may take longer to start up.
>Being cyclic, it can be argued that the activity isn't really engaged in
>until at least one cycle is done.  So really getting to running takes at
>least the one process of putting a foot down on the the ball and rolling
>in that certain way and...  They peter out in a reverse fashion:  one
>foot coming down without the other already in its phase of the cycle or
>so.  Of course, an activity may also stop abruptly (he smashes into the
>wall or jsut keels over or ...) and this may be seen as a different
>ending.  Processes, of course, can end in two clearcut ways - stopping
>and finishing - depending upon whether the goal has been achieved.  The
>big one is abrupt, often an achievement.  The other, merely stopping,
>reduces the process to an activity (for there is no longer a goal), and
>so may be more slow or not.  The superfective ("keep on") also reduces a
>process to an activity and so can go on.
>
>And in the middle, we don't even talk of states as going on, they just
>are.  A progressive, if they had one, would perhaps guarantee that it
>was neither the beginning nor the end of the state, that there was a bit
>left and had already been a bit but that is already risky, since states
>don't have natural limits.  Activities do, so that they can progress,
>with the propriety - as noted - that there already has been a bit - the
>first cycle - before we can say it is going on presumably - barring a
>crash - a bit more to wind down.  But processes don't look back at the
>whole until it is finished, so when it is going on, you can say it has
>been going on but not that it has been ("He has been building the house
>and still is" but not "He is building the house and has built it" or "he
>is building the house and built it" - where all are possible with "He is
>running".)  Loglan note:  if "-e-" is at all intentional, it goes only
>with activities and processes - achievements don't extend in time at all
>and states don't go on.  And so we also need at least two ending words
>and maybe two beginnings as well.

lojbab (with thanks to Nora for typing all this in,and pc for writing it)