/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BASB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRKB30



␈↓ α∧        BARWIS[W80,JMC]␈α↓ contains comments␈α↓ on Barwise's␈α↓ lecture  and
␈↓ α∧his notes.  Let me␈α↓ suggest that others attending the␈α↓ lectures  might
␈↓ α∧want to make similar files for exchanging comments.  It seems␈α↓ to  me,
␈↓ α∧and others have complained, that␈α↓ we have slowed Barwise too␈α↓ much  by
␈↓ α∧our comments in class and have made it difficult for the␈α↓ students  to
␈↓ α∧follow his train␈α↓ of thought.  We give␈α↓ him printouts of␈α↓ the  comment
␈↓ α∧files,  and  this  may  be  of  use  to  him.

␈↓ α∧        Incidentally,  the  E  command  αβXpo  is  useful  for  reading  files
␈↓ α∧referred to in other files.␈α↓  The effect of the command is␈α↓ to  search
␈↓ α∧from the␈α↓ present location of␈α↓ the pointer for␈α↓ text that␈α↓ looks  like
␈↓ α∧aaa[bbb,ccc] or aaa.bbb[ccc,ddd], interpret that text as a␈α↓ file  name
␈↓ α∧and  switch  to  that  file.
␈↓ α∧␈↓␈↓ u2


␈↓ α∧␈↓␈↓ ∧ Notes on Barwise's course - draft of February 10, 1980

␈↓ α∧␈↓[NB. The prudent won't look at drafts dated before the course ends].

␈↓ α∧␈↓␈↓ αTAt␈αthe␈αfirst␈αmeeting,␈αB␈αsaid␈αthat␈αscenes␈αwould␈αnot␈αbe␈αindividuals␈αin␈αhis␈αtheory␈αbut␈α
part␈αof
␈↓ α∧␈↓its␈αmodel␈αtheory,␈αand␈αthat␈αthe␈αtheory␈αcouldn't␈αbe␈αdone␈αin␈αfirst␈αorder␈αlogic.␈α It␈αcan␈αbe␈αdone␈αin␈αfirst
␈↓ α∧␈↓order␈αlogic,␈αe.g.␈αby␈αmaking␈αscenes␈αindividuals.␈α Moravcsik␈αcalled␈αthat␈αbloating␈αthe␈αontology,␈αbut␈αit
␈↓ α∧␈↓can␈α
be␈α
regarded␈α
as␈α
enriching␈α∞it.␈α
 Barwise␈α
and␈α
I␈α
later␈α
agreed␈α∞that␈α
the␈α
issue␈α
is␈α
whether␈α∞there␈α
are
␈↓ α∧␈↓useful statements whose expression requires functions from objects to scenes.

␈↓ α∧␈↓␈↓ αTIn␈α∞the␈α∞jan␈α∞23␈α∞lecture␈α∞in␈α∞Carolyn's␈α
understanding,␈α∞B␈α∞said␈α∞that␈α∞events␈α∞would␈α∞be␈α∞entities␈α
in
␈↓ α∧␈↓model␈α⊂theory␈α∂but␈α⊂not␈α∂in␈α⊂the␈α∂language.␈α⊂ Useful␈α∂event␈α⊂valued␈α∂functions␈α⊂or␈α⊂functions␈α∂comparing
␈↓ α∧␈↓events seem even easier.

␈↓ α∧␈↓Notes on Barwise's lecture notes.

␈↓ α∧␈↓p.3␈α-␈α
He␈αis␈α
unfair␈αto␈α
naive␈αrealism␈α
or␈αat␈αleast␈α
to␈α"Dick␈α
sees␈αJane".␈α
 There␈αis␈α
no␈αneed␈α
to␈αsuppose
␈↓ α∧␈↓that␈α⊂"Dick␈α⊂sees␈α⊂Jane"␈α⊂is␈α⊂the␈α⊂only␈α⊂thing␈α⊂that␈α⊂can␈α⊂be␈α⊂said␈α⊂in␈α⊂order␈α⊂to␈α⊂justify␈α⊂treating␈α⊂it␈α⊂as␈α⊂an
␈↓ α∧␈↓example␈α⊂of␈α∂the␈α⊂simplest␈α⊂thing␈α∂that␈α⊂can␈α⊂be␈α∂said␈α⊂and␈α∂from␈α⊂which␈α⊂the␈α∂others␈α⊂can␈α⊂sometimes␈α∂be
␈↓ α∧␈↓deduced.

␈↓ α∧␈↓The␈α
simplest␈α
deduction␈α
is␈α
to␈α
deduce␈α
from␈α
"John␈α
saw␈α
Mary"␈α
and␈α
"Mary␈α
was␈α
running"␈α
to␈α
"John␈α
saw
␈↓ α∧␈↓Mary␈αrunning".␈α As␈αBarwise␈αpoints␈αout,␈αthis␈αdeduction␈αisn't␈αgenerally␈αvalid.␈α I␈αthink␈αwe␈αwill␈αhave
␈↓ α∧␈↓to␈α
want␈α
to␈α
say␈α
"If␈α
John␈α
saw␈α
Mary␈α
and␈α
Mary␈α
was␈α
running,␈α
then␈α
John␈α
saw␈α
Mary␈α
running␈α
and␈α
knew
␈↓ α∧␈↓that␈α∪Mary␈α∀was␈α∪running␈α∪unless␈α∀something␈α∪prevented␈α∪it".␈α∀ We␈α∪will␈α∪appeal␈α∀to␈α∪non-monotonic
␈↓ α∧␈↓reasoning.␈α
 The␈α
reason␈αfor␈α
doubting␈α
the␈αappropriateness␈α
of␈α
taking␈α"John␈α
saw␈α
Mary␈α
running"␈αas
␈↓ α∧␈↓basic␈α
is␈α
that␈αit␈α
is␈α
only␈αthe␈α
first␈α
step␈α
down␈αthe␈α
slippery␈α
slope␈αof␈α
infinite␈α
qualification.␈α However,␈α
the
␈↓ α∧␈↓non-monotonic␈αreasoning␈α
hasn't␈αbeen␈αdeveloped␈α
in␈αthe␈αintensional␈α
contexts␈αthat␈αwill␈α
be␈αrequired
␈↓ α∧␈↓to use it to solve the problem.

␈↓ α∧␈↓␈↓ αTMaybe scenes will solve the problem, but I doubt it.

␈↓ α∧␈↓Remarks on the second installment of the notes

␈↓ α∧␈↓␈↓ αT1.␈αIt␈αshould␈αbe␈αpointed␈αout␈αthat␈αthe␈αassociationist␈α"theory"␈αwasn't␈αa␈αtheory,␈αbecause␈αit␈αdidn't
␈↓ α∧␈↓present␈α
a␈α
mechanism␈αbut␈α
merely␈α
a␈α
metaphor.␈α The␈α
cited␈α
arguments␈αagainst␈α
it␈α
are␈α
also␈αsomewhat
␈↓ α∧␈↓metaphoric for that reason.

␈↓ α∧␈↓␈↓ αT2.␈α∞Helmholtz␈α∞had␈α∞a␈α∞better␈α
idea␈α∞(as␈α∞usual)␈α∞than␈α∞most␈α
of␈α∞his␈α∞successors␈α∞in␈α∞his␈α
"unconscious
␈↓ α∧␈↓inference",␈αbut␈αit␈αseems␈αlikely␈αthat␈αthe␈αpre-conscious␈αprocessing␈αof␈αvisual␈αinformation␈αis␈αlocalized,
␈↓ α∧␈↓and␈αthat␈αthe␈αresults␈α
of␈αthe␈αearly␈αstages␈αare␈α
not␈αavailable␈αthroughout␈αthe␈αbrain.␈α
 Therefore,␈αwhile
␈↓ α∧␈↓we␈αmight␈α
say␈αthat␈αthe␈α
first␈αvisual␈α
cortex␈αknows␈αsomething,␈α
we␈αwon't␈α
be␈αable␈αto␈α
say␈αthat␈αthe␈α
person
␈↓ α∧␈↓knows␈α⊂it.␈α⊂ This␈α∂can␈α⊂be␈α⊂discussed␈α∂along␈α⊂the␈α⊂lines␈α∂of␈α⊂Dennett's␈α⊂discussion␈α∂of␈α⊂the␈α⊂localization␈α∂of
␈↓ α∧␈↓various aspects of feeling pain.  Barwise says the same thing at the top of p. 14.

␈↓ α∧␈↓␈↓ αT3.␈α∂I␈α∂haven't␈α∂seen␈α∂Bruner's␈α⊂paper␈α∂yet,␈α∂but␈α∂if␈α∂it␈α∂is␈α⊂like␈α∂the␈α∂book␈α∂␈↓↓A␈α∂Study␈α∂of␈α⊂Thinking␈↓␈α∂by
␈↓ α∧␈↓Bruner,␈α∞Goodnow␈α∞and␈α∂Austin␈α∞also␈α∞written␈α∂in␈α∞the␈α∞fifties,␈α∂there␈α∞is␈α∞a␈α∂major␈α∞error.␈α∞ I␈α∂can't␈α∞decide
␈↓ α∧␈↓whether Barwise makes this mistake or not.  His examples suggest that he does.
␈↓ α∧␈↓␈↓ u3


␈↓ α∧␈↓␈↓ αTThe␈α∞trouble␈α
with␈α∞the␈α∞1950s␈α
view␈α∞of␈α∞pattern␈α
recognition␈α∞is␈α∞that␈α
it␈α∞is␈α∞based␈α
on␈α∞the␈α∞idea␈α
of
␈↓ α∧␈↓categorization␈α
into␈α
one␈α
of␈α
a␈α
preassigned␈α
finite␈α
set␈α
of␈α
categories␈α
and␈α
that␈α
a␈α
category␈α
is␈α
regarded␈α
as␈α
a
␈↓ α∧␈↓boolean␈α
combination␈α∞of␈α
elementary␈α
categories.␈α∞ At␈α
least␈α
this␈α∞is␈α
the␈α
view␈α∞taken␈α
by␈α
Bruner,␈α∞et.␈α
 al
␈↓ α∧␈↓and in Earl Hunts ␈↓↓Concept Learning␈↓.

␈↓ α∧␈↓␈↓ αTThis␈α∞is␈α∂the␈α∞error␈α∂that␈α∞Russell␈α∂points␈α∞out␈α∞pervades␈α∂philosophy␈α∞and␈α∂logic␈α∞before␈α∂Frege.␈α∞ It
␈↓ α∧␈↓amounts␈α
to␈α
using␈α
only␈αthe␈α
monadic␈α
predicate␈α
calculus␈α
or␈αeven␈α
just␈α
the␈α
propositional␈α
calculus.␈α It
␈↓ α∧␈↓doesn't␈αsee␈αthat␈αconcepts␈αrequire␈αrelations␈α
for␈αtheir␈αdefinitions.␈α Let␈αme␈αdefine␈αa␈α
two␈αdimensional
␈↓ α∧␈↓pattern␈αas␈αan␈αexample:␈αOne␈αof␈αthem␈αconsists␈αof␈αfour␈αparts.␈α Three␈αof␈αthe␈αparts␈αare␈αhorizontal␈αline
␈↓ α∧␈↓segments␈α∀of␈α∪equal␈α∀length␈α∪and␈α∀the␈α∪fourth␈α∀is␈α∪a␈α∀vertical␈α∪line␈α∀segment␈α∪a␈α∀bit␈α∪longer␈α∀than␈α∪the
␈↓ α∧␈↓horizontals.␈α The␈αleft␈αends␈αof␈αthe␈αthree␈α
horizontals␈αco-incide␈αwith␈αthe␈αbottom,␈αmiddle␈αand␈α
top␈αof
␈↓ α∧␈↓the␈α∂vertical.␈α∂ Given␈α∂this␈α⊂description,␈α∂the␈α∂pattern␈α∂can␈α∂be␈α⊂drawn.␈α∂ The␈α∂way␈α∂I␈α∂have␈α⊂described␈α∂it,
␈↓ α∧␈↓functions␈α⊂from␈α⊂segments␈α⊂to␈α∂points␈α⊂and␈α⊂functions␈α⊂from␈α⊂segments␈α∂to␈α⊂numbers␈α⊂are␈α⊂used␈α⊂and␈α∂the
␈↓ α∧␈↓equality␈α⊂relation␈α⊂for␈α⊂points␈α⊂and␈α⊂numbers␈α⊂are␈α⊂used.␈α⊂ Of␈α⊂course,␈α⊂the␈α⊂description␈α⊂could␈α⊃be␈α⊂done
␈↓ α∧␈↓entirely with relations, but it could not be given only monadic predicates.

␈↓ α∧␈↓␈↓ αTThe␈αearly␈αartificial␈αintelligence␈α
work␈αon␈αvision␈αwas␈α
motivated␈αby␈αthis␈αdisagreement␈αwith␈α
the
␈↓ α∧␈↓"pattern␈α∂recognizers"␈α∂all␈α∂of␈α∂whom␈α∂took␈α∂the␈α∂boolean␈α∂view.␈α∂ Minsky␈α∂and␈α∂I␈α∂pointed␈α∂out␈α⊂that␈α∂this
␈↓ α∧␈↓approach␈α∂would␈α∂be␈α∂hopeless␈α∂for␈α∞controlling␈α∂a␈α∂robot␈α∂that␈α∂could␈α∂be␈α∞told␈α∂to␈α∂pick␈α∂up␈α∂the␈α∂doll.␈α∞ It
␈↓ α∧␈↓cannot␈α
merely␈αcategorize␈α
the␈αscene␈α
as␈α␈↓↓doll-containing␈↓;␈α
it␈αmust␈α
locate␈αthe␈α
doll␈αand␈α
the␈α
shapes␈αand
␈↓ α∧␈↓attitudes␈α
of␈αits␈α
parts.␈α
 The␈αslogan␈α
was␈α
"description␈αnot␈α
just␈α
discrimination".␈α In␈α
the␈α
1970s,␈αAzriel
␈↓ α∧␈↓Rosenfeld␈αand␈αothers␈αhave␈αdeveloped␈αpicture␈αgrammars␈αand␈αhave␈αtried␈αto␈αmake␈αpattern␈αmatchers
␈↓ α∧␈↓like␈α∞grammatical␈α∞parsers.␈α∞ Their␈α∞success␈α∞has␈α∞been␈α∞less␈α∞than␈α∞that␈α∞of␈α∞the␈α∞linguists,␈α∂because␈α∞visual
␈↓ α∧␈↓patterns␈α∞are␈α
more␈α∞complicated,␈α
and␈α∞the␈α
articles␈α∞I␈α∞have␈α
seen␈α∞don't␈α
seem␈α∞to␈α
recognize␈α∞that␈α∞we␈α
are
␈↓ α∧␈↓really␈αinterested␈αin␈αrecognizing␈α
a␈αscene␈αor␈αan␈α
object␈αas␈αa␈αthree␈α
dimensional␈αthing␈αthat␈αgives␈αrise␈α
to
␈↓ α∧␈↓two dimensional images by projection and that these images are partially obscured by occlusion.

␈↓ α∧␈↓␈↓ αT4.␈αI␈αshould␈αmention␈αthat␈α
Marr's␈αprimal␈αsketch␈αis␈αquite␈α
a␈αlot␈αlike␈αthe␈αdrawings␈α
generated␈αat
␈↓ α∧␈↓the␈α⊃Stanford␈α⊃AI␈α⊃Lab␈α⊃in␈α∩the␈α⊃1960s␈α⊃by␈α⊃Manfred␈α⊃Hueckel.␈α∩ Do␈α⊃Marr␈α⊃et.␈α⊃al.␈α⊃refer␈α∩to␈α⊃Hueckel?
␈↓ α∧␈↓Binford says no and mentions similar work by himself and Horn.

␈↓ α∧␈↓␈↓ αT5.␈α∞I␈α∞should␈α∞like␈α
to␈α∞see␈α∞a␈α∞definition␈α
of␈α∞categorization␈α∞to␈α∞see␈α
if␈α∞its␈α∞present␈α∞use␈α∞by␈α
Johnson-
␈↓ α∧␈↓Laird is subject to the above criticism.

␈↓ α∧␈↓␈↓ αT6.␈α∞On␈α∞page␈α
16␈α∞Barwise␈α∞remarks␈α
that␈α∞he␈α∞doesn't␈α
"want␈α∞to␈α∞build␈α
any␈α∞accidental␈α∞features␈α
of
␈↓ α∧␈↓human␈α∂or␈α∂animal␈α⊂perception␈α∂into␈α∂the␈α⊂logic␈α∂of␈α∂perception".␈α⊂ I␈α∂agree,␈α∂but␈α⊂I␈α∂am␈α∂surprised␈α⊂at␈α∂his
␈↓ α∧␈↓eagerness␈α
to␈α∞build␈α
what␈α
seem␈α∞to␈α
be␈α
accidental␈α∞features␈α
of␈α
English␈α∞treatment␈α
of␈α∞perception␈α
verbs
␈↓ α∧␈↓into his logic.

␈↓ α∧␈↓␈↓ αT7.␈α
Another␈α
question␈α
about␈α
the␈α
logic␈α
of␈α
perception.␈α
 One␈α
may␈α
say␈α
"I␈α
see␈α
an␈α
Englishman"␈αor␈α
"I
␈↓ α∧␈↓hear␈αan␈αEnglishman",␈αand␈αthe␈αgiant␈αsaid␈α"I␈α
smell␈αan␈αEnglishman".␈α At␈αthis␈αlevel␈αall␈αthe␈α
modalities
␈↓ α∧␈↓of␈α∞perception␈α∞are␈α∞treated␈α∂the␈α∞same.␈α∞ Is␈α∞it␈α∞Barwise's␈α∂intention␈α∞that␈α∞the␈α∞whole␈α∞logic␈α∂of␈α∞perception
␈↓ α∧␈↓should␈αbe␈αindependent␈αof␈αthe␈αmodality?␈α Will␈αhe␈αisolate␈αthe␈αpart␈αthat␈αcan␈αbe␈αtreated␈αthis␈α
way?␈α I
␈↓ α∧␈↓haven't␈α∂thought␈α⊂about␈α∂it␈α⊂much,␈α∂but␈α∂I␈α⊂suppose␈α∂that␈α⊂I␈α∂have␈α∂always␈α⊂thought␈α∂that␈α⊂each␈α∂modality
␈↓ α∧␈↓requires␈α∂its␈α∞own␈α∂treatment␈α∞in␈α∂order␈α∞to␈α∂treat␈α∞the␈α∂relation␈α∞between␈α∂appearance␈α∞and␈α∂reality.␈α∞ This
␈↓ α∧␈↓treatment␈αfor␈αvision␈αinvolves␈αthe␈αfact␈αthat␈αobjects␈αhave␈αa␈αthree␈αdimensional␈αstructure␈αand␈αthat␈αwe
␈↓ α∧␈↓input␈α∂a␈α∂sequence␈α∂of␈α∂two␈α∂dimensional␈α∂projections␈α⊂of␈α∂parts␈α∂of␈α∂an␈α∂object.␈α∂ Also␈α∂objects␈α⊂hide␈α∂one
␈↓ α∧␈↓another.
␈↓ α∧␈↓␈↓ u4


␈↓ α∧␈↓␈↓ αT8.␈α⊃Barwise␈α⊃characterizes␈α⊃the␈α⊃task␈α⊂of␈α⊃a␈α⊃vision␈α⊃system␈α⊃as␈α⊂that␈α⊃of␈α⊃describing␈α⊃the␈α⊃scene␈α⊂or
␈↓ α∧␈↓answering␈αquestions␈αabout␈αit.␈α As␈αI␈α
mentioned␈αbefore,␈αwe␈αoriginally␈αthought␈αabout␈α
AI␈αperception
␈↓ α∧␈↓of␈α⊂scenes␈α⊂in␈α⊂terms␈α⊂of␈α∂getting␈α⊂the␈α⊂information␈α⊂required␈α⊂to␈α∂manipulate␈α⊂the␈α⊂scene.␈α⊂ The␈α⊂logic␈α∂of
␈↓ α∧␈↓perception␈αin␈α
the␈αservice␈α
of␈αmanipulation␈α
may␈αbe␈α
different␈αand␈α
perhaps␈αmore␈α
fundamental␈αthan
␈↓ α∧␈↓that of perception in the service of conversation.

␈↓ α∧␈↓␈↓ αT9.␈αThis␈α
correct␈αdescription␈αof␈α
the␈αanalogy␈αbetween␈α
the␈αdescription␈αof␈α
a␈αscene␈α
(noticing␈αthat
␈↓ α∧␈↓the␈α
scene␈α∞is␈α
taken␈α
as␈α∞three␈α
dimensional)␈α
is␈α∞not␈α
really␈α
consistent␈α∞with␈α
the␈α
earlier␈α∞quotations␈α
from
␈↓ α∧␈↓Bruner.

␈↓ α∧␈↓␈↓ αT10.␈α∪Marr␈α∪is␈α∪looking␈α∪for␈α∪the␈α∪"universals␈α∪of␈α∪human␈α∪vision"␈α∪as␈α∪Chomsky␈α∪looks␈α∀for␈α∪the
␈↓ α∧␈↓universals␈αof␈αhuman␈αlanguage.␈α Neither␈αdistinguishes␈αvery␈αclearly␈αwhat␈αconstraints␈αare␈αpeculiar␈αto
␈↓ α∧␈↓humans,␈αand␈αwhat␈αare␈α
imposed␈αby␈αthe␈αnature␈α
of␈αlight.␈α A␈αperception␈α
system␈αthat␈αused␈αultra␈α
sound
␈↓ α∧␈↓that␈α
penetrates␈α
solid␈αobjects␈α
and␈α
is␈αpartially␈α
reflected␈α
by␈α
internal␈αsurfaces␈α
would␈α
be␈αquite␈α
different.
␈↓ α∧␈↓Its␈α∞images␈α
might␈α∞be␈α∞three␈α
dimensional␈α∞from␈α∞the␈α
start␈α∞since␈α∞relative␈α
range␈α∞information␈α∞might␈α
be
␈↓ α∧␈↓available.

␈↓ α∧␈↓␈↓ αT11.␈α∂I␈α⊂repeat␈α∂that␈α∂I␈α⊂am␈α∂unimpressed␈α∂by␈α⊂the␈α∂linguistic␈α∂peculiarities␈α⊂of␈α∂perception␈α⊂verbs␈α∂in
␈↓ α∧␈↓English.␈α∃ Of␈α∃course,␈α∃it␈α∃might␈α∃be␈α∃taken␈α∃as␈α∃a␈α∃clue␈α∃or␈α∃surface␈α∃manifestation␈α∃of␈α∃some␈α∀more
␈↓ α∧␈↓fundamental␈αproperty␈α
of␈αthe␈α
logic␈αof␈α
perception,␈αbut␈α
these␈αproperties␈α
will␈αhave␈α
to␈αstand␈α
on␈αtheir
␈↓ α∧␈↓own feet.  I also doubt that anaphora is related to the logic of perception in any fundamental way.

␈↓ α∧␈↓␈↓ αTAll␈αthe␈α
above␈αare␈α
haggles␈αwith␈α
Barwise's␈αtreatment␈α
of␈αpreliminaries␈α
to␈αhis␈α
semantics.␈α The
␈↓ α∧␈↓proof␈α∂of␈α∂the␈α∞pudding␈α∂for␈α∂me␈α∞will␈α∂be␈α∂the␈α∞expressiveness␈α∂of␈α∂his␈α∞formalism␈α∂and␈α∂the␈α∂reasoning␈α∞it
␈↓ α∧␈↓permits.␈α∞ The␈α∂idea␈α∞that␈α∂perception␈α∞can␈α∂be␈α∞usefully␈α∂formalized␈α∞by␈α∂taking␈α∞scenes␈α∂as␈α∞fundamental
␈↓ α∧␈↓seems quite plausible to me.