From newshub.ccs.yorku.ca!torn!utcsri!rutgers!uwm.edu!news.bbn.com!ingria Wed Sep 16 21:23:29 EDT 1992
Article 6914 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca sci.bio:3920 sci.skeptic:20032 soc.men:5020 soc.women:4728 comp.ai.philosophy:6914
Path: newshub.ccs.yorku.ca!torn!utcsri!rutgers!uwm.edu!news.bbn.com!ingria
>From: ingria@bbn.com (Bob Ingria)
Newsgroups: sci.bio,sci.skeptic,soc.men,soc.women,comp.ai.philosophy
Subject: Re: missing verbs
Message-ID: <lba9b6INNec0@news.bbn.com>
Date: 14 Sep 92 23:51:02 GMT
References: <1992Sep9.034138.15488@news.media.mit.edu> <dgsnzqd@lynx.unm.edu>
	<1992Sep9.162211.11503@lmpsbbs.comm.mot.com>
	<1992Sep9.230021.5182@news.media.mit.edu> <BILL.92Sep9232609@ca3.nsma.arizona.edu>
Reply-To: ingria@BBN.COM
Lines: 113
NNTP-Posting-Host: bbn.com
In-reply-to: bill@nsma.arizona.edu's message of 10 Sep 92 06:26:09 GMT

In article <BILL.92Sep9232609@ca3.nsma.arizona.edu> bill@nsma.arizona.edu (Bill Skaggs) writes:
   minsky@media.mit.edu (Marvin Minsky) writes:

      > The problem was to replace phrases like "A defeated B" by a
      > homologous phrase with the same meaning, like "B <verbed> A",
      > where A and B are of equivalent status.  Really, half of the
      > verbs seem to be missing, in this sense.  

   Yeah, but I'm not convinced that there's anything sinister going on.

That's probably true, but I think Marvin Minsky has pointed out a very
interesting phenomenon, in any case.

   It seems to me that some sort of agent-patient stuff might explain
   what's going on -- in English, agents (entities that cause the state
   of affairs to come about) tend to get active verbs, and patients tend
   to get passive verbs.  If "A defeated B", presumably this state of
   affairs was not desired by B, so B was not the agent, so B should not
   be placed in subject position with an active verb.

Well, I first thought something along these lines was the explanation,
but I'm not so sure now.  While the non-existence of the dual of
``defeat'' seems to be explainable on thematic grounds, there is more
to it than that.  ``defeat-inverse'' presumably doesn't exist because
it would be a verb that has a non-agent subject AND an agent object.
Though there are verbs which have non-agent subjects (contrast ``look
at'', which has an agent subject, with ``see'', which does not), as
far as I know, there are no verbs (in any language) with agent
objects.  However, it is curious that there is no simple transitive
verb with a non-agentive object that indicates the subject was not the
victor.  Look at the example in Marvin's first message: ``lose to'',
which is a prepositional verb.  Why isn't there a simple transitive
verb with this meaning?  At least one?  I also can't think up any off
the top of my head in any of the languages I know.  (It would be
interesting to look at Latin or Classical Greek and see if there are
such verbs that take an accusative object, or if they require an
oblique object of some sort.  My lexical grasp of these languages is
mostly recognition-level at this point.)  If there is a systematic gap
along these lines, that's an interesting datum.

   Anyway, English (and other languages) are full of curious asymmetries.
   For example, it's okay to say "The bicycle is next to the house", but
   it's not okay to say "The house is next to the bicycle", though
   logically they ought to mean the same thing.

This isn't the same thing.  "The house is next to the bicycle" is
sayable, it's just weird---presumably for pragmatic reasons; if you're
in a context where the house is movable (e.g. it's a trailer home, or
a model, or the great 21st century flying house) it makes sense.  On
the other hand, what we seem to have in the verb case is the
apparently _systematic_ absence of an entire _class_ of lexical items.
And if this is true across several languages, then there is probably
something quite interesting going on here.

Note, as a possibly related case, an interesting phenomenon in Navaho.
Transitive sentences in Navaho are of the form:

noun-phrase-1 noun-phrase-2 verb+marker

where MARKER indicates whether noun-phrase-1 is the thematic subject
of the sentence or noun-phrase-2 is the thematic subject of the
sentence (or the ``logical subject'' to use a term from pedagogical
grammar; or the ``deep subject'' to use an old transformational grammar
term).  I'll call the marker that indicates that noun-phrase-1 is the
thematic subject ``A'' and the marker that indicates that
noun-phrase-2 is the thematic subject ``P'' (in reality, these are
``yi'' and ``bi'', but it's been a long enough time since I looked at
Navaho that I can't remember which is which, mea culpa).  This system
interacts interestingly with the Navaho animacy hierarchy, which runs,
in descending order:

humans
domesticad animals
undomesticated animals
anything else

i.e. humans are ``more animate'' than domestic animals, and so on.

As long as NP1 and NP2 are of the same animacy, they can appear in any
order:

(1) boy girl kicked-A (``The boy kicked the girl'')
(2) boy girl kicked-P (``The girl kicked the boy; the boy was kicked by the girl'')
(3) girl boy kicked-A (``The girl kicked the boy'')
(4) girl boy kicked-P (``The boy kicked the girl; the girl was kicked by the boy'')

However, if they are of differing animacy, NP1 must be the entity that
is higher on the animacy hierarchy:

(5) boy mule kicked-A (``The boy kicked the mule'')
(6) boy mule kicked-P (``The mule kicked the boy; the boy was kicked by the mule'')
(7) #mule boy kicked-A (``The mule kicked the boy'')
(8) #mule boy kicked-P (``The boy kicked the mule; the mule was kicked by the boy'')

(7) cannot be used as synonymous with (6), nor can (8) be used as
synonymous with (5).  The only context in which (7) or (8) can be said
is one in which the mule is regarded as of the same animacy as the boy
(e.g. a science fiction story about intelligence manipulation
experiments or about the alien mules from Alpha Centauri).
Apparently, in the Navaho view of the world, a creature of lesser
animacy does not have the power to act on one of higher animacy.
Hence, if a person is kicked by a mule, it must be because the person
somehow allowed this to take place, by negligence, for example.  So,
even in the case where a creature of higher animacy is adversely
affected by one of lower animacy, it still retains its agency in some
sense.  (This, along with other material, is discussed in a book by
Gary Witherspoon.)

All this seems tantalizingly similar to the lexical gap in English
discussed above.

-30-
Bob


