[The following is an edited version of my contribution to a discussion
on the conlang email list.  The discussion was about opposites and about
how to deal with them in constructed languages.]



                     Designing an Artificial Language

                               Opposites

                            by Rick Morneau

                             July 9, 1992
                        Revised: July 19, 1994


Questions about the implementation of opposites in artificial languages
(henceforth ALs) are really questions about the interaction between
morphology and lexical semantics - two of my favorite subjects.

Linguists have written quite a lot on the subject of opposites.  For a
highly readable reference, I recommend "Lexical Semantics" by D.A.
Cruse, Cambridge University Press, 1986.  The book contains three
chapters just on opposites.  Here's a partial breakdown of the kinds of
opposites you can run into:

    Complementaries: mutually exclusive - no gray area, such as
	true/false, dead/alive, hit/miss

    Antonyms: gradable, such as long/short, good/bad, hot/cold.  This
	group has several sub-groups.

    Directional opposites: north/south, up/down, forwards/backwards

    Antipodals: cellar/attic, head/toe, full/empty

    Counterparts: male/female, ridge/groove, heaven/hell

    Reversives: rise/fall, enter/leave, tie/untie

    Converses: before/after, above/below, in front of/behind

    Relational: doctor/patient, predator/prey, parent/child

and much more.

Fortunately or unfortunately, I've never heard of a natural language
that marks its words consistently and equally for polarity.  When it
occurs at all, there is always an imbalance, as with the English "un-"
(For example, "marked" is unmarked, while "unmarked" is marked.)

However, both natural and artificial languages mark their words for many
other things, and I see no reason why we can't extend this idea to cover
polarity.  For example, in many Malayo-Polynesian languages (Indonesian,
Tagalog, Fijian, et al.) verbs are derived from primitives to mark their
transitivity and whether they are active or passive.  Or, consider
Iroquoian langauges, such as Mohawk and Cherokee, which mark their verbs
for the semantic roles agent and patient played by the arguments.  As
for nouns, consider the Bantu langauges which mark for class (which is
often semantically relevent).

What if we extended these morphological derivations to indicate their
polarity by doubling the number of marker morphemes?  In other words,
where you may now have one morpheme to mark a specific feature, instead
have two: one for positive and one for negative.  For example, let's say
your AL marks its verbs for transitivity, like this:

     ke- = positive, transitive (X does something to Y)
     bu- = negative, transitive	   "
     di- = positive, intransitive (something happens to X)
     sa- = negative, intransitive	    "

If the root for "free, non-captive" is "booga", then we could create the
following words:

     kebooga = to free, to release
     bubooga = to capture
     dibooga = to escape
     sabooga = to surrender

This approach is okay for a crude first pass, but it is unsatisfying as
a real solution because transitivity is a *syntactic* phenomenon.  I
feel that a *semantic* solution based on thematic relations would be
much more flexible, and I'll have LOTS to say on this in my essay on
lexical semantics.  Also, this solution is too simplistic, because it
fails to deal adequately with gradeable concepts, like "hot", "warm",
"lukewarm", "ambient", "cool", "cold", "frigid".  For these, you could
start with derived primitives to handle basic (+), (0) and (-) concepts,
and then add fine-tuning morphemes to cover the cases in between.  For
example:

    positive (+) primitives:  CVNC, where N is a nasal, homorganic
				with the C that follows it.

    neutral (0) primitives:   CVLC, where L is a liquid, such as /l/
				or /r/

    negative (-) primitives:  CVSC, where S is either /s/ or /z/,
				agreeing in voicing with the C
				that follows it.

Next, add the fine-tuning morphemes:

    ki- = to move in a positive direction
    no- = to move in a negative direction

So, to put it all together, let "pend" = "hot", "peld" = "ambient", and
"pezd" = "cold".  With the fine-tuning morphemes, you can now get:

    kipend = scalding, red-hot
      pend = hot
    nopend = warm, not too hot
    kipeld = lukewarm, tepid
      peld = ambient, room-temperature
    nopeld = slightly cool
    kipezd = cool
      pezd = cold
    nopezd = frigid

You can, of course, add more fine-tuning morphemes to handle finer
distinctions or more extreme ones, such as "super hot" and "super cold".

The only problem I have with all of this is that it is definitely
unnatural, and I would normally be reluctant to include a feature in an
AL that could potentially go against some linguistic universal.
However, I doubt if this is a serious problem, since linguistic purists
could always regard the combination of a "primitive + polarity marker"
as a single morpheme.

I will say no more about this because I am currently working on an essay
that goes into considerably more detail on the whole topic of lexical
semantics for ALs, and I prefer to say it all there rather than here.
Besides, the approach I discuss in that essay is considerably different
than the one discussed here, and, in my opinion, considerably better as
well.  :-)


			   End of essay