G aph Theo y and Uni e sal G amma
Tesi di Do o a o in Linguis ica
Uni e si à o Fi enze - Facol à di Le e e e Filoso ia/
Dipa imen o di Linguis ica
In associazione con CISCL,
Cen o In e dipa imen ale di S udi Cogni i i sul Linguaggio, Siena
Ciclo XX – AA 2007/08
Se o e disciplina e L-Lin/01
Candida o: Ludo ico F anco
Supe iso e: Coo dina o e del do o a o:
P o . Luigi Rizzi P o . Leona do Ma ia Sa oia
2
3
Ring aziamen i
Deside o ing azia e anzi u o il P o . Luigi Rizzi, mio ad iso a Siena e la P o .
Ri a Manzini, mia u o io en ina. Ri engo che sia un p i ilegio eno me pe un
linguis a quello di po e coglie e s imoli da pe sone così piene di senso ed
en usiasmo. E che an o hanno da o alla linguis ica con empo anea.
Un ing aziamen o pa icola e a alla P o . Ad iana Belle i. Sono ce o che senza
la sua capaci à di coin olge e e appassiona e nelle lezioni pe il co so di
Mo osin assi a Siena dell’A.A. 2003/2004 non mi sa ei innamo a o di ques a
disciplina (almeno non ino al pun o da spende e le mie ene gie in un do o a o
senza la bo sa). E ai i colleghi del CISCL, in pa icola e Enzo, Giuliano e Ida.
Al a ci azione d’obbligo pe la mia amiglia e pe l’aiu o mo ale e, sop a u o,
ma e iale nel co so di mesi di icili (a comincia e da Tempo Reale in ca i e acque e
le conseguen i, ine i abili di icol à economiche occo se).
Tu a ia, non la ing azio come spesso edo si usa a e negli Acknowledgmen s
“pe a e eso possibile la s esu a della esi”, bensì pe a e pe messo che non mi
indebolisse u o quan o è passa o in o no, in pa icola e la mia assu da que elle con
l’Uni e si à degli S udi di Fi enze con annessa sospensione dal co so di lau ea in
Medicina. Una cosa che mi ha a o male, che anco a s en o a capi e e che ha
ine i abilmen e condiziona o il mio la o o pe il do o a o.
Ring azio poi Gigapedia.o g, la mia Biblio eca d’Alessand ia con igu abile e le
p oxy lis s in gi o pe la e e.
Scuse sen i e al P o . Leona do Sa oia pe a e agi o d’is in o e pe non a e a o
i e imen o a lui quando sa ebbe s a o do e oso.
In ine un ing aziamen o al P o . And ea Mo o, anche se non lo conosco
pe sonalmen e e l’ho incon a o solo ad un paio di semina i, pe una ase che ho
le o in un suo bel lib o di ulga i o ecen e in me i o al appo o a neu oscienze e
linguis ica eo ica:
“…Fo se non u o il male iene pe nuoce e, una isione pa ziale non è di pe sé
nega i a, bas a che sia esplici a e non abbia la p e esa di esau i e l’a gomen o […] Senza poi
esclude e che siano i mes ie i di neu oscienzia o e di linguis a a esse e ecchi: non è de o che
non ne debba nasce e uno nuo o”.
Un bello s imolo. Una bella s ida possibile.
Da ul imo una spe anza: mi augu o che il mio pseudo-inglese, da qui in a an i,
assomigli il più possibile ad una lingua na u ale.
Ques a esi è dedica a a mia moglie Salomé, s ao dina ia ogni gio no di più:
senza di lei a ei pe so il il-o.
4
ABSTRACT
In he las ew yea s, Noam Chomsky (1994; 1995; 2000; 2001) has gone qui e a in
he di ec ion o simpli ying syn ax, including elimina ing X-ba heo y and he le els
o D-s uc u e and S-s uc u e en i ely, as well as educing mo emen ules o a
combina ion o he mo e p imi i e ope a ions o Copy and Me ge. Wha emain in
he Minimalis P og am a e he ope a ions Me ge and Ag ee and he le els o LF
(Logical Fo m) and PF (Phonological o m).
My doc o al hesis a emp s o o e an economical heo y o syn ac ic s uc u e
om a g aph- heo e ic poin o iew (c . Dies el, 2005), wi h special emphases on he
elimina ion o ca ego y and p ojec ion labels and he Inclusi eness Condi ion
(Chomsky 1994). The majo in luences o he de elopmen o such a heo y ha e
been Ch is Collins’ (2002) seminal pape “Elimina ing labels”, John Bowe s (2001)
unpublished manusc ip “Syn ac ic Rela ions” and he Ca og aphic Pa adigm (see
Belle i, Cinque and Rizzi’s olumes on OUP o a s a ing poin ega ding his
pa adigm).
A syn ac ic s uc u e will be ega ded he e as a g aph consis ing o he se o
lexical i ems, he se o ela ions among hem and no hing mo e.
5
Quo es
“Any in e ence ule can be non- i ially e ised so ha one ails o accep i
whe e one once accep ed i , o ice e sa.”
W.V. Quine
"G aphics e eal da a."
E. Tu e

6
7
P e ace / Guidelines
This wo k consis s o six chap e s.
In he i s chap e I in oduce how opologic ne wo ks could be use ul o
a wide ange o cogni i e ma e s, om a heo e ic pe spec i e, gi en some
basic p inciples. I in oduce as well he ascina ing hypo hesis o a opological
and i al na u e o language (sp eading linguis ic a ia ion) made up by
Pia elli Palma ini & U iage eka in 2004.
Then, I e iew he idea o Da id B. Sea ls who claims ha many
echniques used in bioin o ma ics and gene ics, e en i de eloped
independen ly, may be seen o be g ounded in gene a i e linguis ics and,
symme ically, he disco e o a “language” gene: FOXP2. These disco e ies
(o belie s) a e epo ed he e o show how ui ul could be a c oss-
disciplina y a i ude in he sea ch o he inne mos na u e o Language.
Tha ’s also he main eason o which I ha e ied o apply some basic
p inciples o G aph Theo y o Minimalis p inciples.
The las sec ion o he chap e is a ough his o ical su ey (mainly based
on Wildgen, 1994, 2000) o a “ opological” way o hinking among human
sciences, om Lullus, B uno and Leibniz o Fillmo e, Minsky and Langacke .
No e ha he i s chap e is in ended o be qui e ligh and does no equi e a
na ow compe ence in he ield o gene a i e linguis ics.
In he second chap e “G aphs and T ans o ma ions” I gi e a su ey o
some essen ials o G aph Theo y om a ma hema ical poin o iew, wi h
majo emphases on g aph ans o ma ions including a ansla ion o
Chomsky g amma s in o g aph g amma s, showing he compu a ional
comple eness o g aph ans o ma ion.
Un o una ely, g aphs a e qui e gene ic s uc u es ha can be encoun e ed
in many a ian s in he li e a u e, and he e a e also many ways o apply
ules o g aphs. One canno deal wi h all possibili ies in a basic su ey and i
would also go beyond he scope o my wo k. I ocus on di ec ed g aphs
(syn ac ic ees a e a special kind o di ec ed, label-edged g aphs), because he
di ec ed g aphs can be specialized in o many o he ypes o g aphs, and I
de ine g aph g amma s as a language-gene a ing de ice wi h he mo e
gene al no ion o a ans o ma ion uni ha models bina y ela ions on
g aphs.
Chap e h ee “F om ba e ph ase s uc u e o syn ac ic g aphs” is he co e
o he wo k wi h he g aph heo e ic ( e)de ini ion o in e nal and ex e nal
MERGE and he explana ion o he way lexical g aphs could cap u e
8
cons i uency, C-command and o he impo an syn ac ic ela ions exp essed
in s anda d ee ep esen a ions (assuming o example ha di e en PF
in e p e a ions could be elegan ly de i ed ia syn ac ic g aphs using a e sal
algo i hms de eloped by Yasui (2004a,b)).
I also in oduce in his sec ion o he heo e ical issues ega ding a label-
ee syn ax, such as he ones de eloped by Ch is Collins (2001; 2002), John
Bowe s (2001), and Joan Chen-Main (2006). Finally, I ou line some ligh
simila i ies o B ody’s Mi o Theo y (1997) wi h my p oposal.
Chap e ou “Topics o Pe sian Syn ax and a g aph based analysis o
Pe sian Eza e” in oduces some challenging issues o Pe sian g amma and
de elops a g aph based accoun o he “Eza e puzzle” in Wes e n Indo-
i anian languages, in which NP modi ie s s anda dly occu pos nominally
and “link” o he noun head ia an Eza e pa icle (Ez), which may be
in a ian (Pe sian, So ani), o ag ee wi h N in φ- ea u es (Ku manji, Zazaki).
We will use in ou esea ch he p ecious c oss-linguis ical da a collec ed by
La son and Yamakido (2005; 2006) and Sam elian (2006).
A e a basic (bu a icula ed) ske ch o Pe sian syn ax, wi h majo
emphases on some syn ac ic aspec s ha I ha e al eady conside ed in
p e ious wo ks, such as wo d o de and spli -headness, In e se Case
A ac ion o Pe sian ela i e clauses and Pe sian ligh e b cons uc ions
which seems o be an ou e ins ance o Hale and Keyse (1993; 2002)
composi ional syn ac ic analysis (wi hou he a- oles) o A gumen S uc u e
(see also Ha ley, Folli, Ka imi, 2003), I gi e a de ailed e iew o p e ious
analyses o he Eza e mo pheme in he gene a i e amewo k (c . Samiiam,
1983; 1994; Ghomeshi, 1997; Kahnemuyipou , 2000; F anco, 2004; La son and
Yamakido, 2005; Sam elian; 2006 among o he s) and I de elop he e a g aph
analysis o he Eza e phenomenon based on Den Dikken and Singhap eecha
(2004), whe e he au ho s gi e a c oss-linguis ic accoun ( he poin o
depa u e o hem was he compa ison be ween F ench and Thai) o he
noun ph ases in which linke s occu , in e ms o DP-in e nal P edica e
In e sion (see also Mo o, 1997; 2000). This app oach p omp s an analysis o
ela i e-clause cons uc ions ha ecognizes ela i e clauses as p edica es o
DP-in e nal small clauses, combining he a ac ions o he adi ional
app oach and he Ve gnaud/Kayne aising app oach by assigning ela i e
clauses an in e nal s uc u e simila o he adi ional one while gi ing i he
ex e nal dis ibu ion o a p edica e by ea ing i as he p edica e o a noun
ph ase-in e nal small clause. In o he wo ds, we could say ha a so o
Gene alized P edica e In e sion in he Pe sian complex noun ph ase is ma ked
by he p esence o he Eza e mopheme.
9
In a g aph heo e ic pe spec i e he assump ion ha a syn ac ic elemen
could be in e p e ed as a linke , implies he heo e ical necessi y ha ce ain
linguis ic i ems could be selec ed by he Lexicon as Edges (E), ins ead o
e exes (V) and his is a s imula ing ac .
The i h chap e is a g ound o some (possible and ex eme) heo e ical
consequences o a g aph heo e ic analysis o language acul y.
The Conclusion ollows.
This wo k also has h ee appendixes. The i s in ol e he ield o logics
and is a ough e iew o Cha les Sande s Pei ce’s exis en ial g aphs (based
mainly on P oni, 1992); he second is a mo e nea o he ques ions and
answe s aised in his wo k, and has a s aigh linguis ic opic, showing some
o he mos in e es ing dynamics o he Rela ional G amma Pa adigm
es ablished in he se en ies by Pe lmu e and Pos al. Finally he hi d
appendix is a b ie applica ion o a G aph heo e ic de i a ion o an
Aus onesian Language (Tagalog).
16
a ise in he indi idual, and, ideally, how hey a ose in he e olu iona y
his o y o ou species. A ew yea s ago, Ma c Hause , Noam Chomsky and
Tecumseh Fi ch published an in luen ial pape , i led “The acul y o
language” (Hause e al., 2002). They ask which componen s o ou linguis ic
abili ies a e uniquely human, and which componen s, o close analogues o
hose, can be ound in o he species (in a b oad sense). In a alk du ing he
Basque Coun y encoun e wi h Noam Chomsky, held in 2006 in San
Sebas ian, Ma k Hause de ines language:
“as a mind-in e nal compu a ional sys em designed o hough and o en ex e nalized in
communica ion. Tha is, language e ol ed o in e nal hough and planning and only
la e was co-op ed o communica ion, so his se s up a dissocia ion be ween wha
we do wi h he in e nal compu a ion as opposed o wha he in e nal compu a ion
ac ually e ol ed o . We [Hause , Chomsky and Fi ch] de ined he acul y o
language in he b oad sense (FLB) as including all he men al p ocesses ha a e bo h
necessa y and su icien o suppo language. The eason why we wan o do i in ha
way is because he e a e nume ous hings in e nal o he mind ha will be in ol ed
in language p ocessing, bu ha need no be speci ic o language. Fo example,
memo y is in ol ed in language p ocessing, bu i is no speci ic o language. So i ’s
impo an o dis inguish hose ea u es ha a e in ol ed in he p ocess o language
compu a ion om hose ha a e speci ic o i . Tha ’s why we de eloped he idea o
he acul y o language in he na ow sense (FLN), a acul y wi h wo key
componen s: 1) hose men al p ocesses ha a e unique o language, and 2) hose ha
a e unique o humans. The e o e, i se s ou a compa a i e phylogene ic agenda in
ha we a e looking bo h o wha aspec s a e unique o humans, bu also wha
aspec s a e unique o language as a acul y8”.
The e o e, a e y good summing-up o he obse a ions made in ha
seminal a icle is he one made by Dennis O (2007):
“The pic u e ha Hause , Chomsky and Fi ch o e is ha o ‘blind’ ecu si e
gene a ion, coupled wi h in e ace componen s ha ans o m he ou pu o na ow
syn ax in o ep esen a ions encoding sound and meaning, and a ange o modali ies
o pe o mance, o al speech being one o hem, ha lie ou side he na ow-syn ac ic
sys em. sugges ing ha hese pe o ma i e componen s—concep ual-in en ional and
senso imo o sys ems—a e p esen in o he species, such as highe p ima es—a
conjec u e ha goes back o Desca es” (O , 2007: 78).
8 F om M. Pia elli-Palma ini, J. U iage eka and P. Salabu u Eds. (in p ess) O Minds and
Language: The Basque Coun y Encoun e wi h Noam Chomsky, Ox o d Uni e si y P ess.

17
I hink ha i he unc ion o language ( he na ow one) e ol ed only in
humans, his sugges s ha i s e olu iona y his o y is inc edibly sho in
biological e ms. Indeed, he e is e idence ha language was bo n in humans
abou 100,000 o 40,000 yea s ago (Bicke on, 1990; Lass, 1997).
The ecen pa adigm shi in e olu iona y biology, commonly labelled he
E o-De o pa adigm, has gi en wide c edibili y o claims o sal a ional
eme gence o a ai , igge ed by some minimal mu a ion, o as a by-p oduc
o a unc ionally un ela ed change, such as, o example, b ain g ow h. Fo an
in oduc ion, one o he many wo ks o di ulga ion w i en by S. J. Gould
could be an in e es ing eading.
Acco ding o Ande son and Ligh oo (2003), which in e es ingly i led
hei book The Language O gan – Language as Cogni i e Physiology, he heo y o
(uni e sal) g amma mus hold uni e sally such ha any pe son’s g amma
can be a ained on he basis o na u ally a ailable igge expe iences.
“The ma u e g amma mus de ine an in ini e numbe o exp essions as well-
o med, and o each o hese i mus speci y a leas he sound and he meaning. A
desc ip ion always in ol es hese h ee i ems and hey a e closely ela ed; changing
a claim abou one o he i ems usually in ol es changing claims abou he o he wo.
The g amma is one subcomponen o he mind, a men al o gan which in e ac s wi h
o he cogni i e capaci ies o o gans. Like he g amma , each o he o he o gans is
likely o de elop in ime and o ha e dis inc ini ial and ma u e s a es” (Ande son
and Ligh oo (2003: 45).
Mode n physiology has disco e ed ha he isual sys em ecognizes
iangles, ci cles o squa es h ough he s uc u e o he ci cui s ha il e and
ecompose he e inal image (see he opog aphical model o Hubel and
Wiesel, 1962). Ce ain ne e cells espond only o a s aigh line sloping
downwa d om le o igh wi hin a speci ic, na ow ange o o ien a ions;
o he ne e cells o lines sloped in di e en di ec ions. The ange o angles
ha an indi idual neu on can egis e is se by a gene ic p og am, bu
expe ience is needed o ix he p ecise o ien a ion speci ici y (c . he Nobel p ize R.
Spe y, 1961).
In Ande son and Ligh oo (2003) is depic ed his in e es ing expe imen :
“In he mid-1960s Da id Hubel, To s en Wiesel, and hei colleagues de ised an
ingenious echnique o iden i y how indi idual neu ons in an animal’s isual sys em
eac o speci ic pa e ns in he isual ield (including ho izon al and e ical lines,
mo ing spo s, and sha p angles). They ound ha pa icula ne e cells we e se
wi hin a ew hou s o bi h o eac only o ce ain isual s imuli, and, u he mo e,
18
ha i a ne e cell is no s imula ed wi hin a ew hou s, i becomes o ally ine in
la e li e. In se e al expe imen s on newbo n ki ens, i was shown ha i a ki en
spen i s i s ew days in a dep i ed op ical en i onmen (a all cylinde pain ed
only wi h e ical s ipes), only he neu ons s imula ed by ha en i onmen
emained ac i e; all o he op ical neu ons became inac i e because he ele an
synapses degene a ed, and he ki en ne e lea ned o see ho izon al lines o mo ing
spo s in he no mal way” (Ande son and Ligh oo , 2003: 51. C . hei wo k o o he
in e es ing in e disciplina y da a).
We commonly see he p ocess o language lea ning as a simila ly selec i e
p ocess: pa ame e s a e p o ided by he gene ic equipmen , and ele an
expe ience ixes hose pa ame e s (Pia elli-Palma ini and U iage eka, 2004).
No ice ha a ce ain ma u e cogni i e s uc u e, in he sense o Ande son and
Ligh oo , eme ges a he expense o o he possible s uc u es, which a e los
i e ie ably as he inac i e synapses degene a e.
Following Chomsky (2004), o a good app oxima ion, he in e nal s uc u e
o he language o gan is de e mined by ou ac o s. The i s ac o is he
gene ic endowmen , exp essed in he ini ial s a e o he Language Facul y. The
second ac o is ex e nal linguis ic da a, which shapes he de elopmen o he
language o gan wi hin na ow bounda ies (c . Guas i, 2003). The hi d ac o
comp ises wha is called “de elopmen al cons ain s” in heo e ical biology
and gene al p inciples o biophysics. The ou h ac o conce ns he embedding
o he Language Facul y wi hin he mind, ha is, he way i in e aces wi h
o he componen s (O , 2007).
Anyway, i is in e es ing o epo he e ha S e en Pinke and Ray
Jakendo (2005) disag ee in many espec s (dicho omies) wi h he a icle by
Hause , Fi ch and Chomsky. These include:
i) The Na ow/B oad dicho omy, which makes space only
o comple ely no el capaci ies and o capaci ies aken in ac om
non-linguis ic and non-human capaci ies, omi ing capaci ies ha
may ha e been subs an ially modi ied in he cou se o human
e olu ion;
ii) The u ili y/o iginal- unc ion dicho omy, which conceals
he possibili y o capaci ies ha a e adap a ions o cu en use;
iii) The human/non-human dicho omy, which ails o
dis inguish simila i y due o independen ly e ol ed analogous
unc ions om simila i y due o inhe i ance om a ecen common
19
ances o ;
i ) he co e/non-co e and syn ax/lexicon dicho omies,
which omi he as se o p oduc i e linguis ic phenomena ha
canno be analyzed in e ms o na ow syn ax, and which hus
inco ec ly isola e ecu sion as he only unique de elopmen in he
e olu ion o language.
My pe sonal opinion is ha we acqui e a p oduc i e sys em, a g amma , in
acco dance wi h he equi emen s o a geno ype. We mus assume ha he
linguis ic geno ype yields ini e g amma s, because hey a e ep esen ed in
he ini e space o he b ain, bu ha hey ange o e an in ini y o possible
sen ences. This ( i ual o eal) geno ype may ac as a “pa se ” which in e ac s
wi h he g amma o assign s uc u es and meanings o incoming speech
signals, and cap u es ou capaci y o unde s and spoken language in eal ime.
I belie e ha he linguis ic module con ains abs ac s uc u es which a e
composi ional (consis ing o uni s made up o smalle uni s) and which i a
na ow ange o possibili ies.
A his poin , specula ions conce ning language e olu ion become highly
ele an o linguis ic heo y and I suppose ha he p oposal o Pia elli
Palma ini and U iage eka (2004), o which language has a i al o igin, could
be seen as mo e han a simple p o oca ion.
Pia elli Palma ini and U iage eka (hence o h: PP&U) a gue ha such a
apid ansi ion (I mean, om non-language o language) canno be easily
accoun ed o in cus oma y “adap i e” e olu iona y e ms and hey p opose
ha only a b ain eo ganiza ion “o a d as ic and sudden so ” could ha e
gi en aise o such a s a e o a ai s. This could be ealis ic simply assuming
he E o-de o pa adigm ci ed abo e, bu PP&U conside ha wo main ac s
sugges ha his eo ganiza ion o human cogni i e unc ions may ha e been
“epidemic” in o igin:
i) Recen e olu iona y accoun s o he eme gence o b oad
sys ems in o ganisms poin in he di ec ion o “ho izon al”
ansmission o nucleic ma e ial, o en o i al o igin9 ( om i uses
o pa asi es o ansposable elemen s: he example pa excellence is
9 In si ua ions o his so , e olu iona y apidi y is a consequence o boos ing ele an
numbe s by ha ing no indi iduals, bu en i e popula ions, ca y ele an ly mu a ed gene ic
ma e ial. C . Mic obiology and Immunology On-Line Tex book: USC School o Medicine.
20
he o igin o he Adap i e Immune Sys em);
ii) When he o mal p ope ies behind con ex -sensi i i y in
g amma ( o example as displayed in syn ac ic displacemen ) a e
s udied in Minimalis e ms, a su p ising pa allelism su aces wi h
he wo kings o he Adap i e Immuni y.
The poin o depa u e o PP&U is he conside a ion o he impo ance o
“ho izon al” ansmission o mobile DNA sequences - called ansposable
elemen s - ha a e pe asi e in he genomes o bac e ia, plan s and animals.
These elemen s eplica e as and e icien ly and i is common o ind
hund eds o housands o copies o such elemen s in one single genome.
Indeed, ini ial sequencing o he human genome e ealed ha as much as
45% o he o al is cons i u ed o DNA ha o igina ed om ansposable
elemen s10.
T ansposable elemen s can some imes mo e “la e ally” be ween species, a
phenomenon known as ho izon al ans e . Once hese ho izon al ans e s o
gene ic ma e ial ha e success ully aken place, hen o dina y “ e ical”
ansmission pe pe ua es he new genome: o example, some esea che s
ha e sugges ed ha he immune sys em o highe e eb a es is he p oduc
o he ac i i y o a ansposable elemen ha was “domes ica ed” ollowing
ho izon al ans e om a bac e ium millions o yea s ago (c . Hiom e al.,
1998 ci ed in PP&U, 2004; and o a “b oad spec um” analysis he a icle o
1987, “Mi ochond ial DNA and human e olu ion” by Cann e al.).
Then, he au ho s y o jus i y he assump ion ope a ed by Chomsky o
he s uc u al pe ec ion o he “language o gan”, also conside ing he biza e
ac ha c ea ionis s ha e used Chomsky’s ideas as “e idence” agains
e olu iona y heo y.
PP&U sea ch o biological coun e pa s ha could be conside ed al leas
quasi-pe ec and in oduce he e m “modula imp o emen ”. One example
gi en o modula imp o emen is he ca dio ascula sys em o e eb a es
conside ed as a ac al space illing ne wo k o b anching ubes, unde he
assump ion ha he ene gy dissipa ed by his anspo a ion sys em is
minimized.
I is an inspi ed accoun , in my opinion he one who conside he e olu ion
o an en i e mechanism (speci ically chomskian Na ow Syn ax) which
10 S able inse ion o ansposons, ha e ol e new coding and/o egula o y unc ions,
some imes occu s, wi h d ama ic e olu iona y consequences (Ka p, 2004).
21
es ablishes one o mo e in e aces o be mos likely epigene ic in na u e: i al
in e ac ions p o ide he igh le el o complexi y11.
Fu he mo e, complex co-e olu ions be ween i uses and hos s a e known
o ha e happened, wi h s uc u al changes in he hos , which in addi ion ge
ansmi ed o i s o sp ing. Fo example he exp ession o cap u ed genes
encoding immuno- egula o y p o eins is supposed o be one o he
mechanisms used by i uses in hei in e ac ion wi h he hos ’s immune
sys em (c . Hsu e al. 1990 ci ed in PP&U, 2004).
Along he lines o his biological subs a e, PP&U p oposal de elops:
“Suppose ha , a some poin , humans only had some p imi i e o mal, pe haps a
o m o p o o-language in he sense o Bicke on (1990) o maybe e en a sys em
un ela ed o symbolic communica ion. Na ow Syn ax in he sense ha conce ns
mos syn ac icians would no ha e a isen ye . Then a majo mind/b ain
eo ganiza ion would ha e aken place, which one hopes he de ec ion o he
mo phological i us may be ela ed o. The echnical ques ion is: supposing we ha e
an o ganized elemen a y syn ac ic s uc u e, and u he mo e an alien elemen
which in some sense does no belong, wha can he hos do in o de o elimina e i ?
Fi s o all, i mus de ec he in ude . This is no i ial ask in a se o mechanisms
which, by all accoun s, has i ually no holis ic cha ac e is ics. One possibili y is o
he hos o de ec he in ude on he basis o no being able o in eg a e i
seman ically. Nex , he e has o be some so o “immune esponse”, whe eby he
in ude is somehow elimina ed. The issue he e is “who” elimina es he i us, and
“how”. One mus bea in mind ha all o his has o be done wi h sys emic
esou ces. One o he ew simple ways ha a se o mechanisms o he assumed
complexi y would ha e o p oceeding wi h he immuniza ion ask would be o ma ch
he i us elemen in ca ego ial ype”. This is a kind o p esupposed s uc u e (non-
e minal symbols, ph asal nodes) in ph ase-s uc u e g amma s. I is as i a
mo phological “an igen” we e de ec ed and elimina ed by a syn ac ic “an ibody”. As
o how he elimina ion p oceeds, one has o allow he se o mechanisms he abili y
o dele e he i us ma ched by he an ibody, unde a s ong e sion o he ma ch: ull
ca ego ial iden i y. In u n, i he hos beha es as immune sys ems do, i should keep
a memo y o he p ocess (a e a single exposu e o a i us, immune cells memo ize
he in ude and p o ide esis ance o li e)” ( om Pia elli Palma ini and U ige eka
2004: 361-362).
11 Vi uses a e exquisi ely species and issue speci ic, hey code o s uc u al p o eins and
can in ec an en i e popula ion, and impo an ly o ou pu poses, unlike bac e ia o o he
pa asi es, hey can in eg a e in o a genome. Unlike maliciously buil compu a ional i uses,
biological i uses don’ ha e a pu pose, hus may a p io i esul in a a ie y o consequences
o an o ganism. G an ed, he no mal esul is a mo e o less majo dis up ion o he
o ganism’s unc ions o s uc u e, due o he apid mul iplica ion o he in ec ing i us a he
expenses o he hos ’s own machine y, bu his is no ine i able, and in p inciple a i us may
some imes be in eg a ed s ably, and inhe i ably, in o he genome o i s hos (c . Ka p, 2004).

22
P esumably, hen - i I unde s ood co ec ly he assump ions made by he
wo au ho s - in he p esence o a de ec ed i us o ca ego y X, he hos will
sys ema ically espond wi h ma ching an ibody ca ego y X, and he elimina ion
o unde comple e ea u al iden i y wi h he pa icula ca ego ial alues ha
X happens o exhibi o , o he wise he ele an hos (de i a ion) would die
( e mina e). I epo below an example o a i al linguis ic de i a ion aken by
PP&U (2004: 363):
(1-1) a. [ _ [ T-ag seem [ [Jack] [ o be …]]]]12
a ge sou ce
b. Vi us () de ec ion: [ T-[ag ] [seem [ [Jack] [ o be …]]]]
c. Sea ch o ca ego ical ma ch [ T-[ag D] [seem [ [DPJack] [ o be …]]]]
wi h an ibody ()
d. Elimina e i us unde [ T-[ag D] [seem [ [DPJack] [ o be …]]]]
ca ego ical iden i y D alue = DP alues
e. Sys ema ize he sequence < b; c; d> as ypical o he language
As a esul o he ans o ma ional p ocess, he e is a demons able sense in
which he o mal objec is mo e complex han i was p io o he
“immuniza ion”, in ha he basic “ ee” ela ions a e wa ped. This can be
illus a ed as in (2-1), aken om PP&U (2004: 363):
(2-1)
12 In his ins ance he c ucial ea u e in he a ge (o mo emen ) a e ag eemen ea u es in
Tense (T), and he sou ce o he mo emen is Jack, which can app op ia ely check hose
unin e p e able ea u es in e ms o i s own in e p e able ones. In he p ocess, he sou ce
elemen becomes accessible o he compu a ion by way o Case alua ion, which he a ge
ende s (Chomsky, 1995).
23
So, i is possible o a gue ha he wa ped objec esul ing om associa ing
he an ibody DP o he T wi h a i al ea u e - which happens o be o he
De e mine so ( o example ag eemen in pe son/numbe ) - c ea es new
local ela ions (Collins, 1997).
In pa icula , he i al an igen-an ibody ela ion es ablishes a “chain”. Fo
ins ance, TP1 in (2-1) ( he mo he o he Jack node) es ablishes he con ex o
he lowe link in he chain, while he T’2 in he example ( he mo he o he T2
node hos ing he an igen) es ablishes he con ex o he highe link in he
chain. The chain linking he wo ele an si es o he immuniza ion is {{Jack,
T’2 }, {Jack, TP1}}, o {T’2, TP1} ac o ing ou Jack.
Fo PP&U a syn ac ic chain is analogous o seconda y s uc u ing in nucleic
acids, ha is, he es ablishmen ( h ough some hing like pseudo-kno s, which
a e pai s o s em-loop elemen s in which pa o one s em esides wi hin he
loop o he o he ) o ela ions be ween bases u he apa in he linea
sequence: ela ions o he han he mos elemen a y pai ings which p ima y
s uc u e yields. Jus as RNA seconda y s uc u es ha e nume ous
consequences ( h ough he abili y o in o ma ion sha ing o a so which,
wi hou he pseudo-kno , would be oo long-dis ance o be iable) so oo chains
ha e consequences (see pa ag aph 1.3 and ig. 1-1).
Thus, in he PP&U iew, he complex objec in (2-1) is simply a
ea angemen o mo e elemen a y lexical ea u es, no hing mo e holis ic
han ha .
The abo e esul is de ini ely opological:
“A e he immuniza ion akes place, a (new) linguis ic opology eme ges, and he
esul lends i sel o o he wise impossible in e p e a ions” (Pia elli Palma ini and
U iage eka, 2004: 365).
I ealize ha PP&U conjec u e is essen ially me apho ical. None heless, I
hink his me apho is p oduc i e and wo h pu suing o i s se e al
in e es ing consequences. Indeed, he e a e easons (hin s) o belie e ha i
may be mo e han jus a me apho . The nex wo pa ag aphs should be
explica o y.
1.3 The language o genes (Da id B. Sea ls, 2002)
In he 1980s, se e al esea che s began o ollow a ious h eads o
Chomsky’s legacy in applying linguis ic me hods o molecula biology. The
24
main esul s included he undamen al obse a ion ha o mal
ep esen a ions could be applied o biological sequences — he ex ension o
linguis ic o malisms in new, biologically inspi ed di ec ions — and he
demons a ion o “ he u ili y o g amma s in cap u ing no only
in o ma ional bu also s uc u al aspec s o mac omolecules” (c . Sea ls, 1992).
Da id B. Sea ls in an in e es ing a icle published on Na u e in 2002
demons a ed in which way is possible o apply linguis ics o he s uc u e o
nucleic acids. In b ie , Sea ls showed ha a olded RNA seconda y s uc u e
en ails pai ing be ween nucleo ide bases ha a e a a dis ance om each o he
in he p ima y sequence, es ablishing (in some way) hose ela ionships ha
in linguis ics a e called dependencies.
The mos basic seconda y-s uc u e elemen is he s em-loop13, in which he
s em c ea es a succession o nes ed dependencies ha can be cap u ed in an
idealized o m by he ollowing con ex - ee base-pai ing g amma , as
epo ed in (Sea ls 2002: 211) (see Chap e wo o a su ey on Chomsky’s
hie a chy).
(3-1) [SgSc; ScSg; SaSu; SuSa; Sε ]
(Whe e he ε in he las ule indica es ha an S is simply e ased.)
Fo Sea ls his g amma a o ds any and e e y de i a ion o “hai pin”
sequences o a o m such as he ollowing:
(4-1) SgScgaSucgauSaucgaucgaSucgaucgaucgaucgauc
De i a ions om his g amma g ow ou wa d om he cen al S, c ea ing
he nes ed dependencies o he s em (Fig. 1-1a), analogous o such
phenomena as nes ed ela i e clauses in na u al language.
We ha e o say ha in a ealis ic s em-loop, he de i a ion would
e mina e in an unpai ed loop o a leas se e al bases and migh also con ain,
o example, non-Wa son–C ick base pai s. Bu such ea u es a e easily added
o he g amma wi hou a ec ing he undamen al esul ha any language
consis ing o RNA sequences ha old in o hese basic s uc u es equi es
con ex - ee exp ession.
13 The s uc u e is also known as a hai pin o hai pin loop. I occu s when wo egions o
he same molecule, usually palind omic in nucleo ide sequence, base-pai o o m a double
helix ha ends in an unpai ed loop. The esul ing lollipop-shaped s uc u e is a key building
block o many RNA seconda y s uc u es (c . Wa son JD, Bake TA, Bell SP, Gann A, Le ine
M, Losick R. (2004). Molecula Biology o he Gene. 5 h ed. CSHL P ess)
25
Then Sea ls no e ha , in addi ion o s em-loop s uc u es, a bi a ily
b anched olded s uc u es may be cap u ed by simply adding o he
g amma abo e a ule SSS, whose applica ion c ea es bi u ca ions in he
de i a ion ee (Fig. 1-1b).
The base-pai ing dependencies emain non-c ossing, al hough mo e
complica ed. The esul ing g amma is o mally ambiguous, meaning ha
he e a e gua an eed o be sequences in he language o which mo e han
one de i a ion ee is possible.
Thus, he s ing gaucgaucgauc can be de i ed as a single hai pin o as a
b anched s uc u e (Fig. 1-1a,b). This linguis ic p ope y o ambigui y,
e lec ed in na u al languages in sen ences ha can be syn ac ically pa sed in
mo e han one way ( o example, “She saw he man wi h he elescope”), di ec ly
models he biological phenomenon o al e na i e seconda y s uc u e.
Fu he mo e, inding ha he language o RNA is a leas con ex - ee has
ma hema ical and compu a ional consequences, o example, o he na u e
and inhe en pe o mance bounds o any algo i hm dealing wi h seconda y
s uc u e14 (Knudsen and Hein, 1999).
These consequences show he impo ance o cha ac e izing linguis ic
domains in he common e minology and me hodology o o mal language
heo y, so as o connec hem immedia ely o he weal h o ools and
unde s anding al eady a ailable15.
In he ligh o hese p ac ical consequences o linguis ic complexi y16 (c .
Shiebe , 1985), a signi ican inding is ha he e exis phenomena in RNA ha
in ac aise he language e en beyond con ex - ee.
The mos ob ious o hese a e so-called non-o hodox seconda y
s uc u es, such as pseudokno s (Fig. 1-1c). This con igu a ion induces c oss-
se ial dependencies in he esul ing base pai ings, equi ing con ex -sensi i e
exp ession
Sea ls a gue ha , gi en his u he p omo ion in he Chomsky hie a chy,
14 Fo ins ance, he as , egula exp ession sea ch ools used commonly in bioin o ma ics
(such as hose in he popula Pe l sc ip ing language) a e uled ou , as in hei s anda d o m
hey speci y only egula languages (Knudsen and Hein, 1999).
15 Fo his eason, om ea ly nine ies bioin o ma ics ex books ha e de o ed whole
chap e s o he ela ionship o biological sequences o he Chomsky’s hie a chy.
16 Na u al languages seem o be beyond con ex - ee as well, based on linguis ic
phenomena en ailing c oss-se ial dependencies, al hough in bo h domains such phenomena
seem o be less common han nes ed dependencies. Thus, by one measu e a leas , nucleic
acids may be said o be a abou he same le el o linguis ic complexi y as na u al human
languages. Fo some obse a ions on his issue see Shiebe , S. 1985. “E idence agains he
con ex - eeness o na u al language”. Linguis . Phil. 8, 333–343,
32
On op o all he mo e speci ic concep ual ields (a ays o nine concep s),
s ands a uni e sal ield, which con ains hose quali ies o God ha a e a he
o igin o all u he en i ies and hei concep s. This seman ic sys em has an
on ological and me aphysical ounda ion in he adi ion o A is o elian and
medie al logic ( igu e 2-1).
Fig. 2-1. Lullus’ A s Magna
The idea ha concep s/wo ds o m linea a ays, ha he ex emes may be
linked oge he , and ha a hie a chy o such a ays exis s, is a so o i s
ealiza ion o “ ield-seman ics” (c . Wildgen, 1998).
Anyway, he in e es ing hing he e, is ha Lullus did no s op a he s a ic
idea o a (ci cula ) ield o concep s: he p oposed a combina o y mechanism
which may ha e been mo i a ed by he “machine y” o medie al syllogis ics,
bu which con ained a new ma hema ical impulse which allowed he la e
de elopmen o compu ing machines by Leibniz, Pascal, and o he s (c .
Wildgen, 1994).
Leibniz ga e o he Lullus’ idea he name A s Combina o ia, by which i is
now o en known. I ’s in e es ing o no ice ha some compu e scien is s ha e
adop ed Lullus as a so o ounding a he , claiming ha his sys em o logic
was he beginning o in o ma ion science.
This me hod was an ea ly a emp o use logical means o p oduce
knowledge. Lullus hoped o show ha Ch is ian doc ines could be ob ained
a i icially om a ixed se o p elimina y ideas. Fo example, one o he ables
lis ed he a ibu es o God: goodness, g ea ness, e e ni y, powe , wisdom,

33
will, i ue, u h and glo y. Lullus knew ha all belie e s in he mono heis ic
eligions - whe he Jews, Muslims o Ch is ians - would ag ee wi h hese
a ibu es, gi ing him a i m pla o m om which o a gue. A hie a chy o
linea (and ci cula ) ields, and a combina o ial dynamics on i , al eady
cons i u e a powe ul heo e ical ins umen o o ganizing he uni e se o
concep s, and ep esen a so o ga e o he lexicon24.
In he la e six een h cen u y, Gio dano B uno (1548-1600) began o
elabo a e he Lullian sys em, app oaching o a new sys em o concep ual
o ganiza ion based on he analogy be ween he uni e se and he mind. He
eplaced Lullus’s closed linea ield wi h a egula , bidimensional pa e n
ex ending o in ini y. B uno’s iew o he undamen al wo kings o language
based on his s udies o ancien languages, app oaches ha o con empo a y
semio ics: “Images do no ecei e hei names om he explana ions o he
hings hey signi y, bu a he om he condi ion o hose hings ha do he
signi ying25”. A iew which was epea ed by Fe nand de Saussu e nea ly 300
yea s la e in he 1890’s wi h his idea ha a wo d is composed o wo pa s:
he “signi ied” and he “signi ie ”, and is a he ounda ion o con empo a y
linguis ics.
F om a ma hema ical poin o iew, Lullus’s ield is a ci cula segmen
di ided in o nine sub-segmen s. Wolgang Wildgen no es ha :
“I ins ead o linea segmen s he basic uni is a egula su ace, we may conside
ei he he illing o an (in ini e) a ea by ci cula su aces (sphe es), o i s illing by
egula su aces (polygons) o bodies (polyhed a). The co esponding ma hema ical
p oblem is ha o an (op imal) package o ci cles/sphe es o polygons/polyhed a”
(Wildgen 1998: 216).
B uno decided exac ly ha he illing o a su ace by squa es is mos
adequa e: his seman ic uni e se is cons uc ed on he basis o a squa e g id.
No e ha B uno’s sys em as a so o in e nal dynamic ha conce ns wha
now we call “ he ille s o he memo y pa e n26”. E e y wo d in a pa e n
may be eplaced by i s me apho o i s me onym. Thus, o B uno, a ex
24 S udies upon Lullus o e e y a ie y appea now h oughou he wo ld, he majo i y o
which a e e iewed in he bibliog aphical bulle in o he jou nal S udia Lulliana, a ailable on-
line.
25 Gio dano B uno, On he Composi ion o Images, Signs & Ideas, (p. 31), ed. by Higgins, Dick
New Yo k: Willis, Locke & Owens, on-line a ailable.
26 C . McEl ee, B ian, Fo ake S ephani and Dye Lisbe h. 2003. “Memo y s uc u es ha
subse e sen ence comp ehension” in Jou nal o Memo y and Language 48, 1 : 67-91.
34
which was i s gene a ed along he lines o basic meanings associa ed wi h
he e ms illed in o he sys em can p oduce a o ally di e en ex o ex
in e p e a ion by using a se o me apho ical and me onymical p ocesses.
Some au ho s hink ha he consequences o B uno’s pa allel wo k on
cosmology and a i icial memo y a e a new model o seman ic ields which
was so adical in i s ime ha he i s ollowe s (al hough igno an o his
adi ion) a e he Von-Neumann au oma a (c . Spo ns and Alexande , 2002)
and he neu al ne sys ems o he la e 1980s (c . Wildgen 1998: 237).
The adi ion o Lullus and B uno was s ill ali e when Leibniz (1646-1716)
designed his “De Syn hesi e Analysi uni e sali seu A e in eniendi e judicandi”.
Leibniz’ solu ion is an a i hme ic one and can be in e p e ed as a p ecu so o
ea u e-seman ics27. I associa es cogni i ely p imi i e (i.e. non-de inable)
concep s wi h p ime numbe s. All de inable concep s co espond o non-
p ime numbe s, which can be decomposed in o p ime numbe s. Leibniz
elimina es (as Lullus did) he basic dis inc ion be ween subjec and p edica e,
and p ac ically conside s only wo le els: p imi i e and (by de ini ion)
composi e concep s.
Then, Leibniz ske ches a cons uc i e de ice, which gene alizes he
me hods o Euclid and applies hem o concep ual sys ems28. The ansi ion
om he a i hme ical o he geome ical cha ac e is ic co esponds o he
ansi ion be ween possible (concei able) wo lds, as pu e in en ion, o he
eal wo ld, o he spa ializa ion and empo iza ion o in en ional concep s.
Cen al no ions a e geome ical cong uence and he in e sec ion o
geome ical igu es29.
27 A ough p esen a ion: a seman ic ea u e is a no a ional me hod which can be used o
exp ess he exis ence o non-exis ence o seman ic p ope ies by using plus and minus signs.
(i)
Man is [+HUMAN], [+MALE], [+ADULT]
Woman is [+HUMAN], [-MALE], [+ADULT]
Boy is [+HUMAN], [+MALE], [-ADULT]
Gi l is [+HUMAN], [-MALE], [-ADULT]
In e sec ing seman ic classes sha e he same ea u es. Some ea u es need no be
speci ically men ioned as hei p esence o absence is ob ious om ano he ea u e: This is a
edundancy ule.
28 Fo a de ailed explana ion see Ru he o d, D., 1998. Leibniz and he Ra ional O de o
Na u e. Camb idge Uni e si y P ess.
29 No e ha Leibniz was he i s o use he e m “analysis si us” la e used in he 19 h
cen u y o e e o wha is now known as opology. The e a e wo akes on his si ua ion. On
he one hand, Ma es (1986: 240), ci ing a 1954 pape in Ge man by F euden hal, a gues:
"Al hough o [Leibniz] he si us o a sequence o poin s is comple ely de e mined by he
dis ance be ween hem and is al e ed i hose dis ances a e al e ed, his admi e Eule , in he
35
In b ie , Leibniz demons a es only how he simples no ions like space,
poin , line, plane, ci cle, posi ion in space a e cons uc ed. This ype o
concep ual cha ac e is ic has he me i ha all en i ies de ined can be
cons uc ed and Leibniz imagines how his sys em i elabo a ed can be used o
desc ibe plan s and animals and o in en machines. The geome ical
cha ac e is ic would allow man o do his wi h symbolic echniques in his
imagina ion wi hou he help o conc e e igu es and models30.
1.5.2 Ges al heo y and he opological psychology o Ku Lewin
A he o igin o Ges al heo y s ands philosophy and psychology, which
we e no ye ins i u ionally sepa a ed in Ge many in ea ly 1900s. In he
a ious schools o Ges al psychology (Be lin, G az, Leipzig) di e en aspec s
we e o eg ounded: psychophysiological aspec s in Be lin (e.g., We heime ,
Ko ka, Köhle , Lewin), in ellec ual o ces as Ges al - ounda ion in G az (e.g.,
amous 1736 pape sol ing he Königsbe g B idge P oblem and i s gene aliza ions, used he
e m geome ia si us in such a sense ha he si us emains unchanged unde opological
de o ma ions. He mis akenly c edi s Leibniz wi h o igina ing his concep . ...i is some imes
no ealized ha Leibniz used he e m in an en i ely di e en sense and hence can ha dly be
conside ed he ounde o ha pa o ma hema ics."
Bu Hi ano (1997) a gues di e en ly, quo ing Mandelb o (1977: 419):
"...To sample Leibniz’ scien i ic wo ks is a sobe ing expe ience. Nex o calculus, and o
o he hough s ha ha e been ca ied ou o comple ion, he numbe and a ie y o
p emoni o y h us s is o e whelming. We saw examples in 'packing,'. A “ opological”
in e es o Leibniz is u he ein o ced by inding ha o one momen i s he o a ached
impo ance o geome ic scaling. In "Euclidis P o a"..., which is an a emp o igh en Euclid's
axioms, he s a es,...: 'I ha e di e se de ini ions o he s aigh line. The s aigh line is a
cu e, any pa o which is simila o he whole, and i alone has his p ope y, no only
among cu es bu among se s.' This claim can be p o ed oday."
Thus he ac al geome y p omo ed by Mandelb o d ew on Leibniz's no ions o sel -
simila i y and he p inciple o con inui y: na u a non aci sal us.
We also see ha when Leibniz w o e, in a me aphysical ein, ha " he s aigh line is a cu e,
any pa o which is simila o he whole..." he was an icipa ing opology by mo e han wo
cen u ies (c . wikipedia.o g-> Leibniz).
30 Leibniz' c i ique o image-like models can be gene alized o all oo speci ic and ad hoc
pic o ial desc ip ions. In an in e es ing pape ha ace he linguis ic in luences o Leibniz,
Wol gang Wildgen, who says:
“A s a egy al eady condammed by Leibniz 300 yea s ago is sys ema ically ied by
cogni i e seman ics which wo k wi h ad hoc igu es and wi h pic u es which ha e no
heo e ical s a us. Seman ics o his ype will soon accumula e a chao ic uni e se o ad hoc
igu es and will loose he capaci y o ind gene al and s able egula i ies which is a cen al
aim o any scien i ic en e p ise. Thus Leibniz geome ical cha ac e is ic is a kind o
decons uc ion o cogni i e seman ics in he s yle o Lako ”. (Widgen 1998; p. 224)
36
Meinong, Benussi), emo ional and symbolic aspec s in Leipzig (e.g.,
Co nelius, Bühle ) (c . Casa i and Va zi, 1999).
Ges al heo y, in he mos gene al e ms, is a heo y o mind and b ain
ha p oposes ha he ope a ional p inciple o he b ain is holis ic, pa allel,
and analog, wi h sel -o ganizing endencies; o , ha he whole is g ea e han
he sum o i s pa s. The classic Ges al example is a soap bubble, whose
sphe ical shape is no de ined by a igid empla e, o a ma hema ical o mula,
bu a he i eme ges spon aneously by he pa allel ac ion o su ace ension
ac ing a all poin s in he su ace simul aneously. A good summing up is he
one made by Casa i and Va zi:
“Ges al heo is s emphasise he igu e-g ound a icula ion o pe cei ed
con igu a ions. Some po ions o a pe cei ed scene, due o hei in insic wholeness,
a e pe cei ed as igu e, and a e delinea ed agains a backg ound which is pe cei ed
as comple ing i sel behind he igu e. Now, he isual bounda y ha sepa a es
igu e om g ound is "o ien ed": i belongs o he igu e and no o he g ound31”.
Fo linguis ics issues, he mos p ominen igu e o Ges al is p obably
Ku Lewin (1890-1947). As ea ly as 1912 Ku Lewin o esaw ha a scien i ic
psychology would ha e o make use o “ opology” and o “ he dynamics
which could be concei ed in a opological s uc u e” (c . Lewin 1969: 9).
Many o Lewin’s ideas ecall p inciples o “ o ce dynamics” wo ked ou in
g ea e sophis ica ion in he linguis ic sphe e by Talmy (1988), and Talmy,
along wi h Pe i o (1989) and o he s, ied o demons a e he impo ance o
opology o he unde s anding o a a ie y o di e en so s o linguis ic
s uc u ing (c . pa ag aph 1.5.4).
As Talmy no es, he concep ual s uc u ing e ec ed by language is
illus a ed mos easily in he case o p eposi ions. A p eposi ion such as' in is
magni ude neu al (in a himble, in a olcano), shape neu al (in a well, in a
ench), closu e-neu al (in a bowl, in a ball); i is no howe e discon inui y
neu al (in a bell-ja , in a bi d cage). Wo k on e b-aspec and he mass-coun
dis inc ion, oo, has p o i ed om a opological o ien a ion in he pa adigm o
Cogni i e Linguis ics (c . pa ag aph 1.5.4).
Lewin’s cen al idea was ha o a “psychological li e-space”
(psychologische Lebens aum): li e-space is cons i u ed by he indi idual and a
si ua ion ele an o he indi idual a a gi en momen . The li e-space o an
indi idual has wo aspec s: e e y pa ial domain o an indi idual's li e-space
31 F om Casa i, Robe o, and Va zi, Achille., 1999, Pa s and Places: The S uc u es o Spa ial
Rep esen a ion. Camb idge, Mass.: MIT P ess.
37
co esponds o a “psycho-domain”, con aining he pe son, s uc u es o he
li e-space speci ically ele an o he pe son (indi idual si ua ions) and
s uc u es o he li e-space which a e cons i u ed independen ly om he
pe son (s anda d si ua ions)32 and psychic “locomo ion”, such as pa hs in he
li e-space wi h p e e ed ou es, ba ie s and obs acles.
So, language has a majo unc ion in he Lewin amewo k: i ans e s
in e nal s a es o he pe son o his/he en i onmen . Language enables a ype o
indi ec locomo ion in psychic space: o example, su oga e locomo ion as in
he speech ac o o de ing o in social coo dina ion ia language, change o
social and pe sonal ela ions and cogni i e in luence, which c ea es new
possibili ies o psychic locomo ion ia lea ning.
F om a na ow psychological poin o iew, Lewin begins wi h he
opposi ion hing (in ui i ely: a closed connec ed uni y) and egion (in ui i ely:
a space wi hin which hings a e ee o mo e). As Lewin poin s ou , wha is a
hing om one psychological pe spec i e may be a egion om ano he :
“A hu in he moun ain has he cha ac e o a hing as long as one is ying o
each i om a dis ance. As soon as one goes in, i se es as a egion in which one
can mo e abou .” (Lewin, 1969: 116)
He de ines he no ion o a bounda y zone z be ween wo disconnec ed bu
p oxima e egions m and n, as he egion, o eign o m and n, which has o be
c ossed in passing om one o he o he . The whole m + n + z is hen
connec ed in he opological sense. (Lewin 1936: 121)
The concep o a ba ie he de ines as a bounda y zone which o e s
esis ance o passage o hings be ween one egion and ano he (Smi h, 1994).
32 Fo Lewin, he de elopmen o a child o an adul may be desc ibed as a change in li e-
space. The li e-space o a child al e s as soon as i lea ns o g asp, o con ol, o walk, o
speak, and so on. A p isone has a d ama ically educed li e-space and some si ua ions may
con ain a ac o s (c . emo ional a ac o s, sympa hy, lo e, e c.) o epelle s (si ua ions o
us a ion, ange ), which p o oke eac ions o escape.
Thus a we ha e conside ed he pe son o be an in eg al componen o a li e-space. Bu
pe sons may hemsel es be ega ded as a opological ield wi h an inne a ea (in ape sonal
domain), a pe iphe y o his domain, and a senso -mo o ical domain, which lies be ween he
pe son and his/he con ex ( he si ua ion).
The opological psychology o Lewin was la e elabo a ed by F i z Heide in his
“a ibu ional psychology”. Heide s eng hened he ela ion be ween 'li e span' ca ego ies
and seman ic ca ego ies, o example, pe cep ual, expe imen al, a ec ing, causing,
e alua ion, pa -whole ela ions (possessi e), can, ying, wan ing, e c. (c . Heide , 1958).
Thus, he psychic and he symbolic wo ld ha e he s uc u e o a ield, and opological and
dynamic ( ec o ial) no ions om con empo a y ma hema ics a e used in Heide (1958) o
speci y hese ields.

38
Such esis ance may be asymme ic; hus i may be g ea e in one dimension
han in he opposi e di ec ion.
Ba ie s e ec he deg ee o communica ion be ween one egion and
ano he , o in o he wo ds he deg ee o in luence o he s a e o one egion
on ha o ano he egion. Hence he no ion o deg ee o in luence, oo, need
no be symme ic (Smi h, 1994). Fo example, he ac ha A is in a ce ain
deg ee o communica ion wi h B does no imply ha B is in equally close
communica ion wi h A.
In Lewin’s pa adigm wo egions A and B a e said o be pa s o a
dynamically connec ed egion i a change o s a e o A esul s in a change o
s a e o B. The no ion o dynamic connec edness, oo, is by wha was said
ea lie a ma e o deg ee. In ac we can dis inguish a hie a chy o deg ees o
in e -linkage be ween egions, and he e Lewin echoes discussions in he
Ges al - heo e ical li e a u e o he no ions o “s ong” and “weak” Ges al en
(1969: 173).
A s ong Ges al may be de ined as a complex wi h a high deg ee o
dynamic connec edness be ween i s pa s (examples gi en: an o ganism, an
elec omagne ic ield).
A weak Ges al , o example a chess-club, has a lesse bu s ill non-ze o
dynamic connec edness be ween i s pa s, while a pu ely summa i e whole
(an Und-Ve bindung in Ges al e minology) is such ha i s sepa a e uni s
mani es a ze o deg ee o dynamic connec edness33 (Smi h, 1994).
We ha e in oduced he basic concep s o Lewin’s opological psychology
in a a he gene al way, abs aining om any speci ic applica ions o
psychological ma e s. I will ex end oo much he scope o ou discussion; i
is in e es ing o no ice ha Lewin’s c i ics, howe e , igh ly d ew a en ion o
a ce ain c ucial sho all in his use o ma hema ical no ions in his w i ings. As
was co ec ly poin ed ou by his c i ics, Lewin a ely makes he ma hema ical
heo y o no ions such as connec edness, bounda y, sepa a eness, and so on,
do any subs an ial wo k wi hin he amewo k o his in es iga ions34.
Anyway ce ain aspec s o Lewin’s gene aliza ions o s anda d opology
ha e since shown hemsel es o be highly ui ul i applied o a linguis ic
ield. In my opinion, hese gene aliza ions include:
33 In e es ingly, in he ligh o ou discussions, he no ions cen al o Ges al heo y can be
de ined no me ely on he basis o he no ion o dynamic connec edness, bu also in e ms o
s uc u e-p ese ing ans o ma ions.
34 This c i icism was pu o wa d in an in luen ial a icle by London (1944), an a icle which
did much o hwa he u he de elopmen o opological psychology (o o a opologically
ounded cogni i e science) as Lewin had concei ed i .
39
i) he ecogni ion ha i is possible o cons uc opology on
a non-a omis ic, me eological (c . Casa i and Va zi, 1999) basis which
wo ks in e ms o wholes ( egions) as well as pa s;
ii) he sys ema ic employmen o he no ion o asymme ic
bounda y, a no ion which u ns ou o be c ucial in many cogni i e
sphe es (c . Smi h, 1994);
iii) he employmen o opological ideas and me hods also in
ela ion o ini e domains o objec s.
1.5.3 Husse l’s opology
Ano he p ecu so o he idea o using opology as a ounda ion o human
sciences (and, in nuce, cogni i e sciences) is Edmund Husse l (1959-1938).
Husse l’s Logical In es iga ions (1900-01) con ain a o mal heo y o pa ,
whole and dependences ha is used by he au ho o p o ide a amewo k
o he analysis o mind and language o jus he so ha is p esupposed in
he idea o a opological ounda ion o cogni i e science.
The i le o he hi d o Husse l’s Logical In es iga ions is “On he Theo y
o Wholes and Pa s” and i di ides in o wo chap e s: “The Di e ence
be ween Independen and Dependen Objec s” and “Though s Towa ds a
Theo y o he Pu e Fo ms o Wholes and Pa s”. Husse l’s heo y is
conce ned also wi h he ho izon al ela ions be ween he di e en pa s
wi hin a single whole, ela ions which se e o gi e uni y o in eg i y o he
wholes in ques ion (c . Smi h, 1994).
In o he wo ds, in Husse l’s heo y some pa s o a whole exis me ely side
by side, hey can be des oyed o emo ed om he whole wi hou de imen
o he esidue. A whole all o whose pa s mani es exclusi ely such “side-by-
sideness” ela ions wi h each o he is called a heap o agg ega e o , mo e
echnically, a pu ely summa i e whole (Und-Ve bindung). In many wholes,
howe e , and one migh say in all wholes mani es ing any kind o uni y,
ce ain pa s s and o each o he in ela ions o wha Husse l called necessa y
dependence (which is some imes, bu no always, necessa y in e dependence).
Such pa s, o example he indi idual ins ances o hue, sa u a ion and
b igh ness in ol ed in a gi en ins ance o colou , canno , as a ma e o
necessi y, exis , excep in associa ion wi h hei complemen a y pa s in a
whole o he gi en ype (c . Smi h, 1994).
The e is a huge a ie y o such la e al dependence ela ions gi ing ise o
40
co espondingly huge a ie y o di e en ypes o whole which mo e
s anda d app oaches o “ex ensional me eology” (c . Casa i and Va zi, 1999)
a e unable o dis inguish. The connec ion be ween pa and whole on he one
hand and dependence on he o he may be seen in he ac ha e e y whole
can be ega ded as being dependen on i s own cons i uen pa s. I is one no
inconside able ad an age o Husse l’s heo y ha i allows a p ecise
o mula ion o hese and a ange o ela ed heses wi hin a single amewo k.
Roman Jakobson (1966) applied Husse l’s ideas on pa s, wholes and
ca ego ies om he Logical In es iga ions in di e en b anches o linguis ics,
in he ea ly de elopmen o ca ego ial g amma and o phonology, espec i ely.
Thus, Jakobson’s accoun o dis inc i e ea u es is as he himsel admi s an
applica ion o Husse l’s idea35.
1.5.4 Fillmo e’s ames and he “nou elle ague” o opological
seman ics
Lullus’s ela ional concep is elabo a ed by he concep o alence. Cha les
Fillmo e would la e (well, nine cen u ies la e ) call hese schema a o he
o ganiza ion o concep s in o a uni a y mac o-concep : ames (c . Fillmo e,
1977).
Fillmo e has been ex emely in luen ial in he a eas o syn ax and lexical
seman ics; he was one o he ounde s o cogni i e linguis ics, and de eloped
he heo ies o Case G amma (1968), and F ame Seman ics (1976). In all o his
esea ch he has illumina ed he impo ance o seman ics, and i s ole in
mo i a ing syn ac ic and mo phological phenomena.
The amewo k o Case G amma is a sys em o linguis ic analysis,
ocusing on he link be ween he alence o a e b and he g amma ical
con ex i equi es, c ea ed by Fillmo e in he con ex o ea ly
T ans o ma ional G amma . This heo y analyzes he su ace syn ac ic
s uc u e o sen ences by s udying he combina ion o deep cases (i.e.
hema ic oles) -- Agen , Pa ien , Bene ac o , Loca ion o Ins umen -- which a e
35 The opological backg ound o Husse l's wo k makes i sel el al eady in his heo y o
dependence. I comes o he o e abo e all howe e in his ea men o he no ion o phenomenal
usion (a so o non-disc e e Me ge?): he ela ion which holds be ween wo adjacen pa s o
an ex ended o ali y when he e is no quali a i e discon inui y be ween he wo. Fo example,
adjacen squa es on a chess-boa d a ay a e no used oge he in his sense; bu i we
imagine “a band o colou ha is subjec o a g adual ansi ion om ed h ough o ange o
yellow”, hen each egion o his band is used wi h i s immedia ely adjacen egions (Smi h,
1994).
41
equi ed by a speci ic e b. Fo ins ance, he e b "gi e" in English equi es
an Agen (A) and Pa ien (P), and a Bene icia y (B); e.g. “Gianni (A) gi es
money (P) o Ma ia (B)”. Ob iously, he he i age o Fillmo e is s ill ali e in
con empo a y gene a i e g amma esea ches.
Acco ding o Fillmo e, each e b selec s a ce ain numbe o deep cases
which o m i s case ame. Thus, a case ame desc ibes impo an aspec s o
seman ic alency, o e bs, adjec i es and nouns. Case ames a e subjec o
ce ain cons ain s, such as ha a deep case can occu only once pe sen ence.
Some o he cases a e obliga o y and o he s a e op ional. Obliga o y cases
may no be dele ed, a he isk o p oducing ung amma ical sen ences.
A undamen al hypo hesis o case g amma is ha g amma ical unc ions,
such as subjec o objec , a e de e mined by he deep, seman ic alence o he
e b, which inds i s syn ac ic co ela e in such g amma ical ca ego ies as
Subjec and Objec , and in g amma ical cases such as Nomina i e, Accusa i e,
e c. Fillmo e (1968) pu s o wa ds he ollowing hie a chy o an uni e sal
subjec selec ion ule:
(5-1) [Agen < Ins umen al < Objec i e]
Tha means ha i he case ame o a e b con ains an agen , his one is
ealized as he subjec o an ac i e sen ence; o he wise, he deep case
ollowing he agen in he hie a chy (e.g. Ins umen al) is p omo ed o
subjec .
F ame seman ics is a opological heo y ha ela es linguis ic seman ics o
encyclopaedic knowledge de eloped by Fillmo e, and is, in many espec s, a
u he de elopmen o his Case g amma (c . also Eco, 1975).
The basic idea is ha one canno unde s and he meaning o a single wo d
wi hou access o all he essen ial knowledge ha ela es o ha wo d. Fo
example, one would no be able o unde s and he wo d “sell” wi hou
knowing any hing abou he si ua ion o comme cial ans e , which also
in ol es, among o he hings, a selle , a buye , goods, money, he ela ion
be ween he money and he goods, he ela ions be ween he selle and he
goods and he money, he ela ion be ween he buye and he goods and he
money and so on.
Thus, a wo d ac i a es, o e okes, a ame o seman ic knowledge ela ing
o he speci ic concep i e e s o (o highligh s, in ame seman ic
e minology). A seman ic ame is de ined as a cohe en s uc u e o ela ed
concep s ha a e ela ed such ha wi hou knowledge o all o hem, one
does no ha e comple e knowledge o one o he ei he , and a e in ha sense
48
2.2.3 Connec ed and disconnec ed g aphs
A g aph G is said o be connec ed i he e is a pa h be ween any wo dis inc
e exes o G, else i is said o be disconnec ed. A mul ig aph is one in which
he e exis s mul iple edges be ween he same pai o e exes. The mul iplici y
o he edge be ween he pai o e exes is gi en by he numbe o mul iple
edges be ween he pai o e exes unde conside a ion.
2.2.4 Subg aphs
A subg aph G1= (V1, E1) is de i ed om a di ec ed o undi ec ed g aph G =
(V, E) such ha V1 and E1 a e subse s o V and E and nei he a e null se s39. I
V1=V hen he subg aph is called a spanning subg aph.
An induced subg aph has a se o e exes (U) which is a subse o he se o
e exes (V) o he g aph (G) and con ains all he edges p esen in G ha a e
inciden wi h he chosen subse o e exes (c . ig 2-5).
The cons ain is ha U ≠ 0. Such a subg aph o G is said o be induced by
U (Gibbons, 2002).
ig. 2-5 subg aph de i a ion
2.2.5 Comple e g aphs, isomo phism and deg ees
A comple e g aph (Kn) on V, a se o ‘n’ e exes, is a loop- ee undi ec ed
g aph whe e o all e exes x, y ∈ V, and x ≠ y, he e is an edge. Fo a chosen
alue o ’n’, he e exis s only one comple e g aph.
39 I ’s in e es ing o no ice ha ins ead o emo ing nodes and edges, one may add some
nodes and edges o ex end a g aph such ha he gi en g aph is a subg aph o he ex ension.
The addi ion o nodes causes no p oblem a all, whe eas he addi ion o edges equi es he
speci ica ion o hei labels, sou ces, and a ge s, whe e he la e wo may be gi en o new
nodes (c . K eowski e al. 2006).
.

49
Fig. 2-6 comple e g aph
When we ha e wo undi ec ed g aphs G1= (V1, E1) and G2= (V2, E2), a
unc ion ƒ: V1
→
V2 is called a g aph isomo phism i ƒ is one- o-one and on o and
o all e exes x, y ∈ V1 and edges {x, y} ∈ E1 i and only i {ƒ (x), ƒ (y)} ∈ E2.
So, G1 and G2 a e called isomo phic g aphs.
ig. 2-7 hese g aphs a e isomo phic, Indeed, he equi ed isomo phism
is gi en by 1 -> 1; 2->2; 3->3; 4->4; 5->5
The deg ee o any e ex ‘ ’, deg ( ), is he numbe o edges in he g aph G
ha a e inciden wi h . A loop a any e ex is coun ed as wo edges. A
deg ee o 1 o any e ex makes i a pendan e ex. An undi ec ed g aph (o
mul ig aph40) wi h he same deg ee o each e ex is called a egula g aph.
The incoming, o in, deg ee o is he numbe o edges in G ha a e inciden
in o . I is deno ed by id ( ). The ou ing, o ou , deg ee o is he numbe o
edges in G ha a e inciden om . I is deno ed by od ( ). A loop a a e ex
con ibu es a coun o one o bo h in-deg ee and ou -deg ee.
2.2.6 Plana and bipa i e g aphs
The g aph G is called a plana g aph i G can be d awn in he plane o a
pape wi h i s edges in e sec ing only a e exes o G; else i is called a
40 So, he Eule ci cui in an undi ec ed g aph (o mul ig aph), G = (V, E), wi h no isola ed
e exes, is one ha a e ses e e y edge o he g aph exac ly once. Since he Eule ci cui
a e ses e e y edge o he undi ec ed g aph exac ly once, i is no possible o ha e wo
dis inc Eule ci cui s o he same g aph. An open ail ha a e ses each edge only once is
called an Eule ail (c . Gibbons, 2002)
50
nonplana g aph. A g aph G = (V, E) is said o be bipa i e, i V = V1 U V2 wi h
V1∩V2 = 0, and e e y edge o G is o he o m {x, y} wi h x ∈ V1 and y ∈ V2. I
each e ex in V1 is connec ed o e e y e ex in V2, hen i is a comple e
bipa i e g aph. I |V1| =m and |V2|=n hen he g aph is deno ed by Km, n.
ig. 2-8 A comple e bipa i e g aph
2.2.7 Elemen a y subdi isions
In G = (V, E) a loop- ee undi ec ed g aph, he ope a ion o emo ing an
edge e = {a, c} and adding he edges {a, b}, {b, c} o G-e, whe e b
∈
V is called
elemen a y subdi ision.
The loop- ee undi ec ed g aphs G1= (V1, E1) and G2= (V2, E2) a e called
homeomo phic i hey a e isomo phic o i bo h o hem can be ob ained om he
same loop- ee undi ec ed g aph by a sequence o elemen a y subdi isions41.
2.2.8 Hamil on G aphs
The g aph G = (V, E) wi h |V| ≥ 3, has a Hamil on cycle i he e is a cycle in
G ha con ains e e y e ex in V. A Hamil on pa h is a pa h in G ha con ains
each e ex42.
41 The Ku o owski’s heo em s a es ha a g aph is nonplana i and only i i con ains a
subg aph ha is homeomo phic o ei he K5 o K3, 3 (C . G imaldi, 2005).
42 In a amous opog aphic p oblem, he a elling salesman p oblem a pe son is supposed o
isi each own in his dis ic , and his he should do in such a way ha sa es ime and
money. Ob iously, he should plan he a el so as o isi each own once, and so ha he
o e all ligh ime is as sho as possible. In e ms o g aphs, he is looking o a minimum
weigh ed Hamil on cycle o a g aph, he e ices o which a e he owns and he weigh s on
he edges a e he ligh imes. i is widely belie ed ha no p ac ical algo i hm exis s (c .
Dies el, 2005).
51
ig. 2-9 An Hamil on cycle
2.2.9 Colou ing a g aph
A p ope colou ing o an undi ec ed g aph G= (V, E) occu s when each o
he e exes o G a e colou ed so ha i {x, y} is an edge in g hen x and y a e
colou ed wi h di e en colou s, ha is adjacen e exes ha e di e en
colou s. The minimum numbe o colou s needed o p ope ly colou G is
called he ch oma ic numbe o G (c . oo no e 37).
2.3 T ees
Le ’s now - gi en some basic p inciples and a e minology - y o de ine a
(syn ac ic) ee in o g aph heo e ical e ms. In gene al e ms, a ee is use ul
as a pic o ial ep esen a ion o s uc u e. As a de ice o ep esen ing
s uc u e, i is applicable o any si ua ion whe e a hie a chy o choices is made.
A de i a ion is a pe ec example o a hie a chy o choices, and as such a ee is
ideal as a isual ep esen a ion o a de i a ion. To cons uc a de i a ion ee,
we s a wi h a ee con aining only he oo node. Fo each s ep o he
de i a ion, he ee is co espondingly ex ended. Tha is, e e y ime a
p oduc ion is used o eplace a non- e minal in he cu en sen en ial o m by
he s ing on he igh hand side o he p oduc ion, lines a e d awn om he
co esponding non- e minal in he ee o each symbol in he eplacemen
s ing (c . Ha ju, 2005). A each s age in he cons uc ion o he ee, eading
om le o igh , he lea nodes will be he cu en sen en ial o m. In o mal
language heo y, and speci ically in gene a i e linguis ics, ee diag ams a e
much used, and a ee diag am is an acyclic, connec ed g aph. He e’s a
de ini ion:
(3-2) A loop- ee, undi ec ed, connec ed g aph G= (V, E) which has no cycles is
called a ee and is deno ed by T= (V, E).
52
ig. 2-10 A ee o ope a ions o he a i hme ic o mula x(y+z)+y
The spanning subg aph, o a connec ed g aph, which is also a ee is he
spanning ee43. I p o ides minimal connec i i y be ween all e exes o he
g aph. The e a e wo algo i hms, acco ding o Dies el (2005) o sea ch o he
spanning ee in a g aph:
i) Dep h Fi s Sea ch algo i hm44
ii) B ead h Fi s Sea ch algo i hm45.
I G is a di ec ed g aph and he undi ec ed g aph associa ed wi h G is a
ee, hen G is a di ec ed ee. I is a oo ed ee i he e is a unique e ex ‘ ’
( oo ) such ha he in-deg ee o , id( ) =0 and o all o he e exes ( ) he in-
deg ee, id ( ) =1. The e ex wi h ou -deg ee, ou ( ) =0 is he lea o e minal
e ex. All o he e exes a e b anch nodes o in e nal e exes. The sub ee a
43 Spanning ees a e o en op imal solu ions o p oblems, whe e cos is he c i e ion. We
may also wish o cons uc g aphs ha a e as simple as possible, bu whe e wo e ices a e
always connec ed by a leas wo independen pa hs. These p oblems occu especially in
di e en aspec s o aul ole ance and eliabili y o ne wo ks, whe e one has o make su e
ha a b eakdowno one connec ion does no a ec he unc ionali y o he ne wo k. Simila ly,
in a eliable ne wo k we equi e ha a b eak-down o a node (compu e ) should no esul in
he inac i i y o he whole ne wo k (c . Ha ju, 2006)
44 In ui i ely, in a deep- i s sea ch one s a s a he oo (selec ing some node as he oo in
he g aph case) and explo es as a as possible along each b anch be o e back acking.
Fo mally, a deep- i s sea ch DFS is an unin o med sea ch ha p og esses by expanding he
i s child node o he sea ch ee ha appea s and hus going deepe and deepe un il a goal
node is ound, o un il i hi s a node ha has no child en. Then he sea ch back acks,
e u ning o he mos ecen node i hadn' inished explo ing. C T.H. Co men, C.E.
Leise son And R.L. Ri es , In oduc ion o Algo i hms, MIT P ess, 1993.
45 B ead h- i s sea ch algo i hm begins a he oo node and explo es all he neighbou ing
nodes. Then o each o hose nea es nodes, i explo es hei unexplo ed neighbou nodes,
and so on, un il i inds he goal. C T.H. Co men, C.E. Leise son And R.L. Ri es , In oduc ion
o Algo i hms, MIT P ess, 1993.
53
any e ex is he subg aph induced by ha e ex as oo and all o i s
deceden s. I he edges o b anches o he oo ed ee a e o de ed hen i is
called an o de ed oo ed ee. Gi en m ∈ Z+, a oo ed ee T = (V, E) is an m-a y
ee i o all e exes he od ( ) ≤ m. When m =2 i is a bina y ee. I od ( ) = 0
o m o all ∈ V hen T is called a comple e m-a y ee. Fo m=2, i is a comple e
bina y ee and can be used o ep esen bina y ope a ions. In a comple e m-
a y ee each in e nal e ex has exac ly m-child en. I T is a comple e bina y
ee o heigh h and all he lea es in T a e a le el h hen, T is called a ull
bina y ee.
The le el numbe is he numbe o pa hs om he oo e ex o he e ex
unde conside a ion. I h is he la ges le el numbe achie ed by a lea o T=
(V, E), hen, T is said o ha e a heigh h. A oo ed ee o heigh h is balanced i
he le el numbe o e e y lea in T is (h-1) o h. A e ex in a loop- ee
undi ec ed g aph G = (V, E) is called an a icula ion poin i he subg aph G-
has mo e componen s han he gi en g aph G.
The emo al o he a icula ion poin s disconnec s he g aph. In e ms o
communica ions sys ems and ne wo ks he a icula ion poin s indica e
loca ions whe e he sys em is mos ulne able.
A loop- ee connec ed undi ec ed g aph wi hou any a icula ion poin s is
called biconnec ed and is said o be a “nonsepa able g aph”. A g aph wi h
weigh s assigned o i s edges is called a weigh ed subg aph. Weigh s a e
posi i e eal numbe s a ached o he edges. The migh signi y pa ame e s
such as cos , ime, leng h e c. on each edge conside ed as a link. I x, y ∈ V,
bu (x, y) ∉ E hen w (x, y) = ∞. A se P o bina y sequences ( ep esen ing a
se o symbols) is a p e ix code i no sequence in P is he p e ix o any o he
sequence in P. When a sequence o bina y cha ac e s (0 and 1) a e used o
ep esen symbols, hen, ca e should be aken ha he bina y code o one
symbol does no o m a p e ix o he code o ano he symbol, else decoding
will be di icul and migh be inco ec .
2.3.1 The Sho es pa h and he Minimal spanning ee p oblems
In ou discussion, I hink ha is ele an o in oduce his wo p oblems
and hei solu ions.
The sho es pa h p oblem a ises whene e he e is a need o de e mine he
sho es , cheapes , o mos eliable pa h be ween one o many pai s o nodes
in a ne wo k.
The minimal spanning ee p oblem occu s when we need o design he
simples ne wo k (a spanning ee) ha will connec opologically dispe sed

54
sys em componen s so ha hey can communica e wi h each o he and he
o al cons uc ion cos is minimized.
The sho es pa h p oblem di e s om he minimum spanning ee
p oblem in ha , he o me is used o ind he cheapes pa h be ween some
nodes, while he la e is o ind he cheapes ee ha connec s e e y node
(c . Dies el, 2005).
Many o he mos salien co e ing edien s o ne wo k echniques a e
cap u ed by he sho es pa h p oblem. I is o en encoun e ed in he
anspo a ion and communica ion ne wo ks.
2.3.1.1 Dijks a’ Sho es Pa h algo i hm
In a gi en weigh ed di ec ed g aph G= (V, E) he sho es pa h be ween
any wo e exes is ha pa h in which he sum o weigh s o all he
cons i uen edges is he minimum. The main idea o he algo i hm is o
change he empo a y labels associa ed wi h e exes in o pe manen ones.
The pe manen label o a e ex deno es he sho es pa h weigh om he
sou ce e ex o he cu en e ex.
The sou ce e ex is gi en a pe manen label (0,-) and each o he o he
e exes is gi en a empo a y label (∞,-). Le P and P’ (= V-P) be he se s
con aining e exes wi h pe manen and empo a y labels espec i ely. A
each s ep, he algo i hm chooses he e ex x ∈ P’ wi h he minimum
empo a y label, and makes i pe manen , eco ds i s p edecesso ’s index and
upda es he empo a y alues o all he e exes. I epea s his p ocedu e ill
all nodes ge pe manen labels (c . Ha ju, 2005).
2.3.1.2 K uskal’s Minimal Spanning ee algo i hm
The main idea o his algo i hm is o add a each s ep he cheapes edge
(lowes weigh ) om he se o emaining edges. The subg aph de eloped a
each s ep should no con ain any cycles. Ini ially he coun e i is se o 1 and
an edge e1 in G= (V, E) is selec ed wi h he lowes weigh .
I edges e1, e2…ei ha e been selec ed o 1≤ i ≤n-2 whe e n = |V|, hen, ei+1
is selec ed such ha i has he lowes weigh among he emaining edges and
he subg aph being o med does no con ain any cycles. Subsequen ly, i is
eplaced by i+1.The p ocedu e is epea ed ill i< n-1.I i = n -1 hen he
subg aph o G de e mined bys he edges e1, e2…en-1 is connec ed wi h n
e exes and n-1 edges and is he minimal spanning ee o G (c . Ha ju,
2006).
55
2.3.1.3 P im’s Algo i hm
This gi es he “op imal ee” o a g aph. Ini ially he coun e i is se o 1
and an e ex 1 ∈ V is placed in se P. Now N and T a e de ined as N = V-{ 1}
and T =0. Fo 1≤ i ≤n-2 whe e n = |V|, P is he se o e exes and T is he se
o edges and N = V-P. The edge o minimal weigh in G ha connec s a e ex
x in P o a e ex y (= i +1) in N is added o he se T. Now, y is placed in se P
and dele ed om N. The coun e is inc emen ed by 1. The p ocess is epea ed
ill i=n. I i = n hen he subg aph o G de e mined by he edges e1, e2…en-1 is
connec ed wi h n e exes and n-1 edges and is he op imal spanning ee o
G. (c . Dies el, 2005).
2.4 G aph ans o ma ions
G aph ans o ma ion is a ule-based me hod ha pe o ms local changes
on g aphs (Dies el, 2005). Wi h g aph ans o ma ion ules i is possible o
speci y o mally and isually o ins ance he seman ics o ule-based sys ems
(like he seman ics o unc ional languages), speci ic g aph languages, g aph
algo i hms (like he sea ch o all Eule ian cycles in a g aph), and many mo e.
The idea o a g aph ans o ma ion ule is o exp ess which pa o a g aph
is o be eplaced by ano he g aph. Unlike s ings, a subg aph o be eplaced
can be linked in many ways (i.e., by many edges) wi h he su ounding
g aph. Consequen ly, a ule also has o speci y which kind o links a e
allowed; his is done wi h he help o a hi d g aph ha is common o he
eplaced and he eplacing g aph and equi es ha he su ounding g aph
may be linked o he eplaced g aph only wi h edges inciden o his hi d
g aph (c . K eowski e al. 2006).
He e is a o mal de ini ion o a ule o g aph ans o ma ion, aken om
K eowski e al. (2006: 6):
(4-2) A ule = (L ⊇ K ⊇ R) consis s o h ee g aphs L; K; R such ha K is
a subg aph o L and R. The componen s L, K, and R o a e called le -hand
side, gluing g aph, and igh -hand side, espec i ely.
He e I show, ollowing K eowski e al. (2006), a simple example,
conce ning sho es pa hs. The igu e below shows he wo essen ial ules o
he compu a ion o sho es pa hs in dis ance g aphs, ha is g aphs labeled
wi h non-nega i e in ege s. The i s ule adds ( add) a di ec connec ion
56
be ween each wo nodes ha a e connec ed by a pa h o leng h 2 and sums
he dis ances up.
Using his ule, one can compu e he ansi i e closu e o he gi en
dis ance g aph. I one applies he second ule ( selec ), which chooses he
sho es connec ion o wo di ec connec ions as long as possible, one ends up
wi h sho es connec ion be ween each wo nodes.
Fig. 2-11. G aph ans o ma ion ules o he compu a ion o sho es pa hs, aken
om K eowski e al. (2006; p. 7)
2.5 G aph Languages
Analogously o Chomsky g amma s (see pa ag aph 2.7) in o mal
language heo y, g aph ans o ma ion can be used o gene a e g aph
languages. A g aph g amma consis s o a se o ules, a s a g aph, and a
e minal exp ession ixing he se o e minal g aphs (c . Eh ig e al. 1999).
Such a e minal exp ession may consis o a se Δ ⊆ ∑ o e minal labels
admi ing all g aphs ha a e labeled o e Δ.
He e is o mal de ini ion o a g aph g amma :
(5-2)
A g aph g amma is a sys em [GG] = (S; P; Δ) whe e S is he ini ial g aph o
GG, P is a ini e se o g aph ans o ma ion ules, and Δ ⊆ ∑ is a se o
e minal symbols. The gene a ed language o GG consis s o all g aphs G ha
a e labeled o e Δ and ha a e de i able om he ini ial g aph S ia
successi e applica ion o he ules in P (K eowski e al. 2006: 14).
I gi e an example o a g amma gene a ing connec ed g aphs (see
pa ag aph 2.2.).
Conside connec ed = {°; P; *} whe e he s a g aph consis s o a single node
57
and he e minal exp ession allows all g aphs labeled only wi h * . No e ha
he symbol * deno es a special label in ∑ s anding o unlabeled and being
in isible in displayed g aphs (c . Gibbons, 2002).
The ules in P = {p1; p2; p3} a e depic ed in Figu e 2-12. The ule p1 adds a
node and an edge e such ha is he a ge o e, and akes as sou ce o e an
al eady exis ing node. The ule p2 is simila , he only di e ence being ha he
di ec ion o he new edge e is in e ed. The hi d ule p3 gene a es a new
edge be ween wo exis ing nodes. The new edge can also be a loop i he wo
nodes in he le -hand side o p3 a e iden i ied, o example, i hey a e one
and he same node in he ma ch o he le -hand side. I can be shown ha he
gene a ed language o connec ed, L(connec ed), consis s o all non-emp y
connec ed unlabeled g aphs.
Fig. 2-12. Rules o he gene a ion o a (connec ed) g aph
2.6 Fo mal g amma s and linguis ic heo y
I gi e in his sec ion a ough p esen a ion o Chomsky G amma s, be o e
applying a ansla ion o Chomsky G amma s in o G aph G amma (see
abo e), in he inal pa ag aph o his chap e conce ning g aph heo y’s
o malisms.
In linguis ics a o mal g amma is a p ecise desc ip ion o a o mal
language — ha is, o a se o s ings o e some alphabe . The wo main
ca ego ies o o mal g amma s a e gene a i e g amma s, which desc ibe how o
w i e s ings ha belong o a gi en language (gene a e), and analy ic g amma s,
which desc ibe how o ecognize when s ings a e membe s in he language
64
diag am has been assumed (c . Haegeman, 1996) o ha e p ope ies such as:
(a) The e is one e ex, called he oo e ex, ha is domina ed by no
e exes and om which he e is a pa h o e e y e ex, whe e a pa h is any
linea subse o a ee (c . Chap e 2).
(b) E e y e ex o he han he oo has exac ly one e ex ha
immedia ely domina es i .
(c) The e exes ha each singula e ex immedia ely domina es a e
o de ed om he le (o be e , empo ally); as a co olla y o (c):
(c1) says ha a well- o med sen ence needs o cons i u e a (single)
connec ed g aph wi h one special e ex as i s oo .
(c2) need no (o should no be e ained wi hin he minimalis p og am),
whe e linea o de is assumed o play no signi ican syn ac ic ole.
The pos ula es (c1) and (c2) has been adop ed in i ually e e y heo y o
ph ase s uc u e, and i speci ically excludes a diag am (c . Yasui, 2004a) wi h
a closed ou e such as (3-3):
(3-3)
1
2 3
4 5
6
The p oblema ic node is clea ly 6, which is domina ed by wo nodes, 4 and
5; hus, he s uc u e gi en in (3-3) iola es wha assumed in (c2).
Whe he (c2) should be assumed o no depends me ely on o he
assump ions on ph ase s uc u e. I y o show in he p esen con ibu ion
ha a g aph-based analysis o some o he undamen al assump ions in he
minimalis p og am leads o he ejec ion o (c2) (o he g ow h o a linea
model).
In he syn ac ic g aphs, o be p oposed below, in ac he ou pu o ex e nal
MERGE is a ee wi h he p ope y (c2) bu ha o in e nal MERGE (o
mo emen ) is a g aph ha has a “closed ou e”47.
47 This dis inc ion could o e a na u al explana ion o he pa ame ic di e ence in wh-
mo emen ; he PF equi emen o linea izing lexical i ems o ces a g aph wi h a closed ou e
o be changed in o a ee in ei he o he wo possible ways, which co espond o ou e and
co e mo emen .

65
3.2 G aphs and Ba e ph ase s uc u e
One impo an simpli ica ion o ph ase s uc u e p oposed since Chomsky
(1994, 1995: Chap e 5) is he elimina ion o ca ego y and p ojec ion labels by
he ex ensi e use o lexical i ems hemsel es, which is mo i a ed by he
Inclusi eness Condi ion48. To mee he Inclusi eness Condi ion, (4-3) is o be
assumed ins ead o (1-3) (=(2-3) in ou e ision om a g aph heo e ic poin
o iew).
(4-3) will
i will
will be
be snowing
V = {i , will, will, will, be, be, snowing}
E={<will, i >, <will, will>, <will, will>, <will, be>, <be, be>, <be,
snowing>}
The s uc u e gi en in (4-3) con ains nodes wi h he same labels will and be.
I nodes wi h he same label a e o be iden i ied as one (c . Yasui, 2004a), he
se V in (4) is non-dis inc om {i , will, be, snowing}.
Then, (4-3) is o ced o be eplaced by (5-3):
(5-3) will
i be
snowing
V={i , will, be, snowing}
E={<will, i >, <will, will>, <will, will>, <will, be>, <be, be>, <be,
snowing>}
48 The Inclusi eness Condi ion says ha he ou pu o a sys em does no con ain any hing
beyond i s inpu . I was i s p oposed in Chomsky (1995: 225) as a condi ion me by he
compu a ional sys em o human language, and aken o imply ha he in e ace le els
con ain no hing mo e han a angemen s o lexical ea u es. In o he wo ds, a language
which mee s he inclusi eness condi ion canno con ain aces o indices le a e mo emen .
66
The s uc u e in (5-3) con ains h ee loops (c . pa ag aph 2.2), <will, will>,
<will, will> and <be, be>, which exp ess no hing o he han in e media e
p ojec ions (so hey can be ejec ed; e ased). The e o e, (6-3) is o be inally
assumed he e:
(6-3) will
i be
snowing
V={i , will, be, snowing}
E={<will, i >, <will, be>, <be, snowing>}
The s uc u e in (6-3) migh no look like a syn ac ic ee, bu he se o
nodes V is essen ially he enume a ion in he sense o Chomsky (1995: 225-
227), and E in (6-3) seems o mee he Inclusi eness Condi ion jus because no
p ojec ion and ca ego y labels a e added (a simila conclusion is ound in
Yasui, 2004a).
Fu he mo e, he o de pai <α, β> is gene ally de ined as {{α}, {α, β}}, and
i looks qui e close o Chomsky's (1995:244-245) de ini ion o he objec
o med om α and β o he ype α: {α, {α,β}}. I {α, β} is adop ed ins ead o {α,
{α,β}} as he de ini ion o he objec o med om α and β, he disc epancy
be ween he o mal de ini ion and i s g aphical ep esen a ion can disappea ,
which seems o be an in e es ing esul (c . Collins, 2002).
In his g aph heo y based accoun o syn ac ic s uc u e, i is e iden ha
my p ima y in luence (my deb ) is he pape o Collins (2002), ha is a so o
ques o a label ee syn ax.
3.3 Ex e nal and In e nal Me ge
Fi s , le ’s conside ex e nal MERGE, which is applied o wo subs uc u es
α and β and p oduces a la ge s uc u e only i some syn ac ic ela ion holds
be ween α and β (c . S a ke, 2001). To pa aph ase i in he e ms o a linguis ic
g aph heo y, MERGE is a kind o “linking ope a ion” o e wo g aphs49, wi h
49 Acco ding o he s anda d concep ion o syn ac ic s uc u e, me ging lexical i ems, α and
β, is ep esen ed by he ee in (i), wi h no o de assumed.
(i)
67
an o de ed pai added o he linking p ocess. I is o mally de ined as (7-3):
(7-3) MERGE de ini ion in opological ne wo k e ms:
Gi en wo g aphs G1 =(V1, E1) and G2=(V2, E2),
MERGE (G1, G2) gi e a g aph G=(V, E) such ha
(i) V = V1 ∪ V2 and
(ii) Fo some 1∈ V1, 2 ∈ V2, E = E1 ∪ E2 ∪ {< 1, 2>} o
E= E1 ∪ E2 ∪ {< 2, 1>}
The esul ing se o e exes is simply he union o he E se s o he wo
inpu g aphs ( his ac is c ucial o he d i age o Chap e , 5). Some e ex in
one g aph en e s in o a local syn ac ic ela ion wi h some e ex in he o he
g aph, whe eby he wo g aphs a e combined50. Ten a i ely, a de i a ion
s a s wi h a se o minimal g aphs, each o which consis s o a single lexical
i em, chosen om he lexicon be o e he connec ion among he i ems ake
place. Fo ins ance, he example in (4-3) could s a s wi h (8-3)
(8-3) . i . will . be . snowing
G1=(V1, V2): V1={i }, E1=∅
G2=(V2, E2): V2={will}, E2=∅
G3=(V3, E3): V3={be}, E3=∅
G4=(V4, E4): V4={snowing}, E4=∅
Since he auxilia y be selec s he p og essi e o m o a e b (snowing), he
o de ed pai <be, snowing> is in oduced as in (9-3):
(9-3) . i . will . be
. snowing
The new node γ is in oduced along wi h wo di ec ed edges connec ing i wi h α and β. α
and β a e p onounced bu γ has no phone ic alue. I γ is he same ype as α, α is he head o
he s uc u e labeled as γ, and α selec s o ag ees wi h β. The syn ac ic ela ion is e e sed i γ
is he same ype as β (c . Yasui, 2004b).
50 I assume ha is he i s membe o each o de ed pai o selec he second membe o o
ag ee wi h i wi h i s EPP ea u e (c . Bosko ic, 2002).
68
MERGE(G3, G4) = G5=(V5, E5)
Whe e V5=V3∪V4={be, snowing}
And E5=E3∪E4 {<be, snowing>}={<be, snowing>}
Then, he i em Will selec s51 e bal and also case-checks o nomina i e;
hus, he ecu si e applica ion o MERGE will con e (9-3) in o (10-3) and
hen (11-3) (=(6-3)):
(10-3) . i . will
. be
. snowing
MERGE (G2, G5)=G6=(V6, E6)
V6=V2∪V5={will, be, snowing}
E6=E2∪E5∪ {<will, be>}={<will, be>, <be, snowing>}
(11-3) . i
. will
. be
. snowing
MERGE (G1, G6)=G7=(V7, E7)
V7=V1∪V6={i , will, be, snowing}
E7=E1∪E6∪ {<will, i >}={<will, i >, <will, be>, <be, snowing>}
I does no ma e which o he o de ed pai s in E7 is added i s . Fo
example, i is easy o e i y ha adding he pai <will, i > be o e <will, be> o
<be, snowing> will make he same esul .
I is impo an o no ice ha MERGE as de ined in (7-3) does no p e en a
g aph om being combined wi h i sel . This is, in ac , an ins ance o an
51 I is possible o ind some simila i ies wi h some assumpion made in Bowe s (2001). C .
pa ag aph 2. o a discussion o his p oposal.
69
in e nal MERGE ope a ion (o mo emen ). To illus a e his poin , conside
(12-3), an example in ol ing wh-mo emen :
(12-3) [I wonde ] wha Gianni will say.
The ba e ph ase s uc u e analysis o (12-3) p obably could be he s uc u e
ep esen ed in (13-3)52.
(13-3)
[WH]
wha [WH]
[WH] will
will
Gianni
will say
say wha
Wi h a g aph heo e ic ep esen a ion, (12-3) would ha e a s uc u e like
(14-3) be o e he in e nal MERGE o he i em wha :
(14-3)
. [WH]
Gianni . . will
. say
. wha
V={[WH], will, Gianni, say, wha }
52 No e ha he P s uc u e and he mo emen (in e nal me ge) o he subjec /objec a e
igno ed he e.

70
E={<say, wha >,<will, say>,<will, Gianni>,<[WH], will>}
MERGE {G, G} will p oduce (15-3), whe e he wh-checking ela ion o
<[WH], wha > is inse ed:
(15-3) MERGE (G, G)=G’=(V’, E’):
a.
V’=V ∪ V=V={[WH], will, Gianni, say, wha }
E’=E ∪ E ∪ {<[WH], wha >}={<say, wha >,<will, say>,
<will, Gianni>, <[WH], will>, <[WH], wha >}
b.
. [WH]
Gianni . . will
. say
. wha
The ep esen a ion gi en in (15-3) clea ly iola es one o he de ining
p ope ies o ee om a classic pe spec i e, (see c2): wha is immedia ely
domina ed by say and [WH] (c . also Chen-Main (2006) p oposal, e iewed
below in pa ag aph 3.9.3). Mo e gene ally, in e nal MERGE on a ee always
in oduces one closed ou e, and he esul is no a ee by de ini ion ( his ac
has been no ed o i s by Yasui, 2004a).
The example (15-3) migh look oo ou ageous, gi en he widely accep ed
iew ha a sen ence has a ee s uc u e. No hing, howe e , seems o be
w ong wi h his kind o ep esen a ion as a syn ac ic s uc u e.
We may ask i e e y lexical i em has a “ e ex” na u e. We will add ess
his ques ion in Chap e 4 and 5.
In b ie , I ha e a gued ha in e nal MERGE is no di e en om ex e nal
MERGE in adding one o de ed pai o E. The di e ence is ha in ex e nal
MERGE, one membe o a pai is a new lexical i em in oduced om he
lexicon, while wo o he lexical i ems al eady in oduced o m a pai in
in e nal MERGE.
71
3.4 Cons i uency
I belie e ha lexical g aphs can cap u e cons i uency and o he impo an
syn ac ic ela ions exp essed in s anda d ee ep esen a ions. Take again (1-
3) o example. I s adi ional and minimalis ep esen a ions a e showed in
(16-3 a,b), while i s lexical g aph is (16-3c).
(16-3)
The non-b anching nodes in (16-3a) a e elimina ed in (16-3b), and he
emaining p ojec ion nodes a e ep esen ed by lexical i ems as labels, which
a e ma ked o dis inguish hem om hose ha ha e a phone ic ealiza ion.
72
The node be* exp esses impo an syn ac ic ela ions: be and snowing a e
sis e s; hey o m a cons i uen ; and he cons i uen hey o m is he same
ype as be a he han snowing. The lowe will* has a simila unc ion. The
uppe will* cap u es he ac ha i o ms a la ge s uc u e wi h he
cons i uen ha is o he same ype as will, and ha he esul an s uc u e is
also he same ype as will. The ma ked e exes a e emo ed in (5c), whe e
he ela ions o selec ion and ag eemen a e ep esen ed by di ec ed edges.
The ou nodes in (16-3c) co espond o he ou minimal p ojec ions in
(16-3b). The uppe will* and be* in (16-3b) co espond o he subg aphs
oo ed by will and be in (16-3c), espec i ely. Mo e gene ally, a cons i uen o
he ype α can be de ined as α i sel o a sub-g aph ha consis s o all he
nodes α domina es. One cons i uen ha alls ou o his de ini ion is he
lowe will* in (16-3b), which co esponds o he in e media e p ojec ion I’ in
(16-3a).
This is a welcome esul , since an in e media e p ojec ion is syn ac ically
and seman ically in isible as Chomsky (1994:10) claims.
Head-mo emen , which is s ic ly local, a ec s wo nodes connec ed by a
single edge. In a lexical g aph, he subjec is adjacen o he In l jus like he
objec is adjacen o he e b.
Assuming he s anda d ee ep esen a ion like (16-3a), Kayne (1984) and
Haegeman (1996: 485) a gue ha cli iciza ion o he objec o he e b is
upwa d and syn ac ic, bu cli iciza ion o he subjec o he In l is downwa d
due o he in e media e node I’ and hence mus be analyzed as a PF-
ope a ion. I a lexical g aph is adop ed, his conclusion can be ci cum en ed
since he p oblema ic in e media e node I’ is absen , as discussed abo e (c .
also Yasui, 2004a).
The con igu a ion o a lexical g aph alone does no dis inguish a speci ie
om a complemen . I syn ac ic s uc u e is buil i s by sa is ying selec ional
equi emen s, ollowed by ag eemen o o mal ea u e checking, a speci ie
and a complemen can be dis inguished based on hei de i a ional his o ies.
Al e na i ely, a speci ie , i i is no an exple i e, esul s om Mo e o
in e nal Me ge; i is connec ed in o he ca ego y selec ing i and also o he
one inducing ag eemen wi h i . I will assume ha a speci ie can be
iden i ied by i s de i a ional his o y o i s double connec edness in a lexical
g aph.
I hink ha i is possible o gi e a basic condi ion o de i a ional s eps:
A de i a ion g aph o an exp ession is a eco d o one possible sequence o s eps
aken o de i e he exp ession in ques ion om lexical i ems.
73
3.5 G aph heo y and C-command
Nex , conside how c-command can be de ined in a lexical g aph. The
example in (17-3a; adap ed om Yasui 2004a, who discusses he same
a gumen in a sligh ly di e en manne ) shows a ypical con as in e lexi e
binding, and i s ba e ph ase s uc u e and lexical g aph a e (17-3b,c),
espec i ely.
(17-3) a. he mo he o he boy alked abou he el /*himsel
c. [pas ]
he alk
mo he abou
o he sel /*himsel
he
boy
In (17-3b), he uppe he* is immedia ely domina ed by he oo , which
domina es he e lexi e, bu he lowe he* is immedia ely domina ed by o *,
which does no domina e he e lexi e. Applying he same de ini ion o c-
command o (17-3c) can accoun o he con as ; he uppe he c-commands
80
In my opinion, he e y in e es ing hing in Yasui accoun 59 is he
demons a ion ha one way o de i e mul iple PF-in e p e a ions is o choose
a a b anching node ei he o i s child nodes a he han always gi ing
p io i y o i s le child, while ano he possibili y is o s a a pos -o de
a e sal om any lea node and p oceed owa ds he oo gene ally in he
ascending o de o he nodes. This is easonable since all he lea nodes a e
p onounced be o e he in e nal nodes in pos -o de .
The oo o a g aph is unique in domina ing all he o he e exes and ha
lea e exes a e also unique in he opposi e sense; hey domina e no o he
nodes. I a pos -o de a e sal s a s om a lea node, he e should be as
many head- inal p onuncia ions as he numbe o lea nodes.
3.9 O he leading hypo hesis o a label- ee syn ax
I in oduce in his pa ag aph o he in luen ial heo e ical issues ega ding
a label- ee syn ax, such as he ones de eloped by Ch is Collins (2001; 2002),
John Bowe s (2001), and Joan Chen-Main (2006). I ha e o admi ha he
majo in luence o he de elopmen o my g aph-based label(and le el)- ee
syn ax has been Ch is Collins’ (2002) seminal pape “Elimina ing labels”-
3.9.1 Collins (2001-2002): elimina ing labels and p ojec ions
The main poin o Collins’ analysis is o sugges ha i may be possible o
elimina e labels in he minimalis amewo k. In o he wo ds, he ope a ion
Me ge(V, X) yelds (b) a he han (a):
(25-3)
a. Me ge (V, X) = {V, {V,X}}
b. Me ge (V, X) = {V, X}
This is also he basic assump ion o my wo k. Collins a gues ha gi en a
p inciple o lexical access (Chomsky 2000) ha he calls “The Locus P inciple”,
he labels in he heo y o Ba e Ph ase S uc u e can be elimina ed en i ely,
lea ing ba e Me ge as he only ope a ion o syn ax apa om Ag ee.
He e is Collins’ Locus P inciple:
(26-3)
59 I is in e es ing o no ice ha she also de eloped a C++ p og am o he de i a ion o
lexical g aphs (c . Yasui 2003).

81
“Le X be a lexical i em ha has one o mo e p obe selec o s. Suppose X is chosen
oma lexical a ay and in oduced in o he de i a ion. Then he p obe/selec io o X
mus be sa is ied be o e any new unsa u a ed lexical i ems a e chosen om he
lexical a ay. Le us call X he locus o he de i a ion”. (Collins 2002: 46)
Following Collins’ e minology, a lexical i em is said o be sa u a ed i all o
i s selec o s ha e been sa is ied and unsa u a ed i one o mo e o hem is s ill
unsa is ied.
When all he selec o s o e e y lexical i em in he a ay ha e been sa is ied,
he lexical a ay is sa u a ed and he p ocess o o ming ela ions is comple e.
I any selec o o any lexical i em has no been sa is ied, hen he lexical a ay
is unsa u a ed and he p ocess o o ming ela ions is incomple e. I is
possibly o assume ha when a selec o o a lexical i em has been sa is ied, i
is dele ed om he lexical en y. A sa u a ed lexical i em hus con ains no
selec o s, while an unsa u a ed lexical i em con ains a leas one unsa is ied
selec o .
The Locus P inciple ensu es ha no ela ion can be o med be ween a
lexical i em and ano he unsa u a ed lexical i em.
Then, Collins a emp s o ex end he Minimal Link Condi ion o
subca ego iza ion in a label- ee heo y o Me ge. One s aigh o wa d way o
s a e he Minimal Link Condi ion is as ollows (see Chomsky, 2002 and Rizzi,
1990):
(27-3) Le P be a p obe. Then, he goal G is he closes ea u e ha can en e in o
an ag eemen ela ion wi h P.
Collins p oposes o accoun o he ac ha subca ego iza ion/selec ion
condi ions a e se e ely cons ained o apply o he nea es c-commanded
ca ego y o he app op ia e ype by ea ing he subca ego iza ion/selec ion
ea u e as a kind o p obe, hence subjec o he Minimal Link Condi ion
(MLC); in o de o explain he ac ha he unc ional p ojec ion in an
example such as he ollowing doesn’ block subca ego iza ion:
(28-3) John looks oo happy o lea e.
Collins s ipula es ha he MLC applies o subca ego iza ion in such a way
ha i is blocked he s ipula es ha he MLC applies o sub-ca ego iza ion in
such a way ha i is blocked jus in case he e is an in e ening lexical
ca ego y ([+/-V, +/-N]). The p oblem is why a p enominal adjec i e doesn’
82
block selec ion o N by a D elemen such as he:
(29-3) a. [ he [sma [s uden ]]]
b. [ he [ e y sma [s uden ]]]
Conside ing ha his is no a p oblem in he case o a b anching AP such
as e y sma in (29-3b), Collins specula es ( ollowing Rubin, 1996) ha
pe haps p enominal adjec i es a e always b anching ca ego ies, hough he
doesn’ eally a gue e y s ongly o such an app oach.
In any case, i is clea ha Rubin’s heo y, while in oducing a new
unc ional ca ego y Mod which selec s Ad P and AP complemen s, is jus
ano he means o ge ing a ound he ac ha X’- heo y doesn’ p o ide a
na u al way o dis inguishing modi ica ion om ela ions such as sub-
ca ego iza ion and selec ion (c . also Bowe s, 2001).
3.9.2 A basic ope a ion o syn ax: Fo m Rel. (Bowe s, 2001)
In an in luen ial ye s ill unpublished pape (book?) om 2001 John
Bowe s ies o show ha he only ope a ion needed in he syn ax is “Fo m
Rela ion” (Fo mRel), which combines pai s o lexical i ems, o ea u es o
lexical i em, and o ms o de ed pai s in acco dance wi h speci ic p ope ies o
hose lexical i ems. Thus, Bowe s a gues ha he e is a e y small se o
o de ed pai s ha cons i u e he undamen al ela ions (in he ma hema ical
sense) o na u al language syn ax.
Bowe s, ollowing Collins (2001; 2002) assumes in his wo k jus ou basic
linguis ic ela ions: selec ion, subca ego iza ion, modi ica ion, and ag eemen .
In addi ion, he au ho a gues ha each ime an o de ed pai is o med,
he e is an immedia e e lex in bo h a he phonological in e ace and he
seman ic in e ace (wi h p inciples called “Immedia e Spell-Ou ” and
“Immedia e In e p e a ion,” espec i ely).
Gi en hese assump ions, Bowe s ies o show ha he no ions o
cons i uen s uc u e (and so he ee adjoining model) and mo emen a e
simply a e ac s o he undamen al legibili y condi ions ha hold a he
seman ic and phonological le els, oge he wi h a small numbe o
compu a ional p inciples ha ei he limi he sea ch space o Fo mRel o limi
he possible ou pu s o Spell-Ou and In e p e a ion60.
60 Nei he he idea ha he p imi i es o syn ac ic heo y should be ela ions a he han
cons i uen s no he idea ha Spell-Ou and In e p e a ion should be immedia e a e o ally
new and unique o Bowe s’ heo y. Va ious simila p oposals ha e been sha ed in
83
Bowe s assumes, as men ioned abo e, ha he e is jus one basic ope a ion
in na ow syn ax, Fo m Rela ion (Fo mRel). Fo mRel is a bina y ope a ion ha
applies o lexical i ems a and b and o ms an o de ed pai (a, b).
Gi en his ope a ion, a ne wo k o syn ac ic ela ions is buil up in he
ollowing way. Fi s , an a ay A o lexical i ems is chosen om he Lexicon.
Second, Fo mRel applies successi ely o pai s o lexical i ems, selec ed om A
and om p e iously o med o de ed pai s, con inuing un il all he selec ion
and subca ego iza ion (c . Collins, 2002) ea u es o e e y lexical i ems a e
sa is ied and none a e le unsa is ied (c . Bowe s, 2001: 16).
The de i a ion in Bowe s’ pape is egula ed by he ollowing
compu a ional p inciple, a sligh ly modi ied e sion o Collins’ (2003) Locus
P inciple (see abo e (26-3)):
(30-3) The Locus P inciple acco ding o Bowe s:
“Suppose a lexical i em l, called he Locus, con aining unsa is ied selec ion and
subca ego iza ion ea u es, is selec ed om a lexical a ay. Then all he
subca ego iza ion condi ions and selec ional equi emen s o l mus be sa is ied
be o e a new lexical i em can be selec ed as he Locus” (Bowe s, 2001: 18).
Bowe s ag ees wi h Collins, conside ing he ac ha o him a lexical i em
all o whose sub-ca ego iza ion condi ions and selec o s ha e been sa is ied is
said o be sa u a ed; i any o hem ha e no been sa is ied, i is said o be
unsa u a ed.
Indeed, he Locus P inciple ules ou he possibili y o a lexical i em A
o ming a ela ion wi h an unsa u a ed lexical i em B61.
I gi e an example below o make clea e Bowe s’ iew. Conside he
ph ase ead he books. The Locus P inciple equi es ha he ela ion selec ion
RSel ( he, books) be es ablished be o e he ela ion subca ego iza ion RSub
( ead, he). I he la e was o med i s , he Locus P inciple would be
iola ed, since he i em he would be unsa u a ed (c . Collins, 2002) a ha
poin 62.
amewo ks such as Pe lmu e and Pos al Rela ional G amma (see Appendix B).
61 As shown in Collins and U a 2001, his imposes an inhe en o de on he p ocess o
o ming a ne wo k o ela ions be ween lexical i ems.
62 I is impo an o no e ha he e a e no cons i uen s in a heo y o his so . In he
example jus discussed, he e is no cons i uen [ he book] in na ow syn ax, no is he e one
o he o m [ ead [ he books]] (wi h o wi hou labels). Ins ead, he e a e simply wo ela ions
( ead, he) and ( he, books).
84
F om a heo e ical poin o iew, i ’s in e es ing o say ha , hough he e is
a supe icial simila i y be ween an accoun based pu ely on ela ions and one
ha inco po a es he ope a ion Me ge o i s equi alen , due o he ac ha
bo h in ol e he cons uc ion o se s, i mus be said he ope a ion Me ge
goes a beyond wha is in ol ed in a ela ional heo y. In he sen ence abo e,
he ou pu o he (canonical) Me ge ope a ion would be a new syn ac ic objec
o he o m: { ead, { he, books}}.
Despi e he ac ha he ou pu s o successi e applica ions o Me ge a e
only uno de ed se s, each ope a ion esul s in a new syn ac ic objec which
inco po a es he esul s o all he p eceding ope a ions: i is clea ly a heo y
ha inco po a es a no ion o cons i uen s uc u e.
In he ela ional heo y o John Bowe s, on he o he hand, no new
syn ac ic objec s o his so a e p oduced. Ins ead, s ill conside ing he
sen ence abo e, he e a e jus he wo o de ed pai s ( ead, he) and ( he,
books).
Bowe s’ accoun would be a e olu iona y one, bu I suppose ha he no ion
o (a simpli ied) Me ge is an axiom o syn ac ic heo y63, and I ha e ci ed
Bowe s’ pape in my wo k mainly because is ano he impo an a emp o
elimina e labels and o simpli y he whole syn ac ic de i a ion p ocess.
Indeed, Bowe s gi es a simple and economical accoun o he possible
linea iza ion p ocess in syn ax (Bowe s, 2001; p.27):
(31-3) “Suppose a and b a e lexical i ems and he o de ed pai (a, b) is a membe
o he selec ion ela ion RSel. The linea iza ion unc ion (FL) ope a es on RSel(a,b) as
ollows:
FL (RSel(a, b)) = a-b”.
Thus Bowe s’ FL is a e y simple and gene al unc ion which ensu es ha
he phone ic o m o he “ i s coo dina e” o a sub-ca ego iza ion ela ion
p ecedes he phone ic o m o he “second coo dina e”.
He e is an example o see how FL wo ks. Le ’s s a by choosing om he
lexicon he i ems ead, he, and book. Assuming ha he selec s nouns and ead
subca ego izes de e mine s, he wo ela ions ( he, books) and ( ead, he) can be
63 The syn ac ic ope a ion Mo e could dissol e in o he simple , mo e economic, and
indispensable syn ac ic ope a ion Me ge in he simpli ied way shown in pa ag aph 3.3, and
X-ba heo y could be simpli ied om a g aph heo e ical poin o iew; in my idea o syn ax,
Me ge is a necessa y “linking ope a ion” o e wo g aphs, wi h an o de ed pai added o he
linking p ocess and, in many espec s I ag ee wi h S a ke (2001) o e he unnecessa ily o
ope a ion Mo e by he ac ha Mo e equi es Me ge, bu no ice e sa.
85
o med. The Locus P inciple, in he o m o Bowe s, equi es ha hey a e
o med in ha o de . The ela ion ( ead, he) couldn’ be o med i s , because
he would be unsa u a ed a ha poin . He e is a schema a o he p ocess,
which is a simpli ied e sion o he one in Bowe s (2001: 29)
(32-3) Bowe s’ linea iza ion p ocess
i)
a. Selec he, books om A [Locus] > he (unsa u a ed); books is
sa u a ed.
b. Fo mRel( he, books)=( he, books) : FL(( he, books))= he-books
ii)
a. Selec ead om A; selec he om ( he, books) o med a s ep ib) [Locus] >
ead (unsa u a ed); he is sa u a ed.
b. Fo mRel( ead, he)=( ead, he) : FL(( ead, he)) = he-books- ead
3.9.3 Mul i-dominance and Lexical g aphs (Chen-Main, 2006)
Finally, conce ning linea iza ion, I discuss b ie ly some o he hings pu in
e idence by Chen-Main PhD disse a ion (2006), ha explo es o mal and
linguis ic consequences in a “mul i-dominance” sys em ha esul om
aking linea izabili y o be a p ope y o well- o med syn ac ic s uc u es.
He wo k is ano he in en ional e u n o he no ion ha syn ac ic
s uc u es should be ep esen ed a he mos basic le el wi h nodes ( e exes)
and edges, and an in i a ion o impo use ul ideas and esul s om he s udy
o g aphs.
Fo Chen-Main, he poin o depa u e is he conside a ion o how some
common cons uc ions such as wh-ques ions and coo dina ed cons uc ions
seem o allow lexical i ems o play mul iple g amma ical oles ypically
associa ed wi h dis inc posi ions. As a p o o ypical example, in sen ences
like “Wha did Gianni ea ?”, wha is usually assumed o unc ion as he objec
o ea , e en hough i appea s sen ence ini ially a he han in he canonical
objec posi ion.
In ano he Chen-Main example “Joe bakes and Sam sells cookies”, a single
noun ph ase, cookies, sa is ies bo h e bs’ need o an objec . T adi ionally,
his appa en “mul iple-linking” is a ibu ed o co-indexing dis inc elemen s
illing mul iple posi ions, only one o which is p onounced.
Al e na i ely, Chen-Main a gue ha a mul iply-linked elemen can be
concep ualized as an elemen immedia ely domina ed by mul iple pa en

86
nodes. Unde such a mul i-dominance app oach, ees no longe su ice o
ep esen ing he immedia e dominance ela ion. Ra he , he se o syn ac ic
s uc u es is expanded o include non- ee g aphs in a shape simila o hose
examples I ha e gi en in sec ions 3.2 and 3.3.
The in e es ing ac abou Chen-Main hesis is o examine how such mul i-
dominance s uc u es64 a e gene a ed and how hei e minals a e linea ized.
Tha ques ion is answe ed in he wo k by adop ing he node-con ac ion
ope a ion, o iginally in oduced in o he T ee Adjoining G amma o malism
o allow o coo dina ion (c . also Ci ko, 2005). Chen-Main conside s node-
con ac ion o be a gene al mechanism in he T ee g amma sys em.
He e is my schema ic esume o he p oposal: node-con ac ion is in ol ed
no only in gene a ing coo dina ion, bu also o cases adi ionally deal wi h
ia mo emen . The exis ence o island e ec s ha p ohibi mo emen om
ce ain domains indica es ha one mus speci y when mul i-dominance
canno occu (by placing ce ain locali y es ic ions on node con ac ion a
he de i a ional le el, Chen-Main a gues ha a numbe o hese island e ec s
can be de i ed).
The linea iza ion ques is a ma e o eal in e es (c . Obs iously Kayne,
1994 and Chap e 5) and, as I ha e al eady men ioned, sha es some
simila i ies wi h my app oach65. This p oposal oo lea es behind he one-
pa en -pe -node es ic ion ha cha ac e izes ees (nodes a e allowed o be
immedia ely domina ed by mul iple pa en nodes), bu i di e s om my
iew in he ac ha he possibili y o he elimina ion o labels is no
conside ed he e.
Below, in (33-3) I show Chen-Main de ini ion o a lexical g aph:
64 To summa ize he ma e , mul i-dominance has been explo ed by a numbe o
esea che s: Pe e s and Ri chie’s (1981) “ph ase linking g amma ” is a a ie y o mul i-
dominance syn ax in which wo ypes o immedia e dominance a e possible and a node is
allowed o ha e a pa en in bo h ela ions. The s uc u es used by Goodall (1987) o analyze
coo dina ion allow a single lexical i em o be pa o mul iple conjunc s. La e , Gä ne ’s
(1997) close examina ion o he widely ollowed Minimalis P og am (Chomsky 1995) led o a
p oposal o eplace he ope a ions Me ge, which combines wo syn ac ic objec s and o ms a
single combined objec , and Mo e, which duplica es pa o a syn ac ic objec and me ges i
wi h he o iginal objec , wi h a single hyb id ope a ion whose applica ion allowed mul i-
domina ed s uc u es. In 2001, bo h S a ke and Chomsky ecas Mo e as a special case o
Me ge. S a ke (2001) a gued ha Mo e could be educed o a special case o Me ge applied
o non-adjacen nodes, and Chomsky (2001) in oduced he e ms Ex e nal Me ge, which
me ges nodes ha ha e no been me ged be o e, and In e nal Me ge, which e-me ges a node
ha has p e iously been me ged, esul ing in mul iply domina ed nodes. (see also Ci ko,
2005).
65 Howe e , he p oposed p ocess o de i ing o de ing in o ma ion does no gua an ee a
linea iza ion o e minals o e e y g aph. A g aph may be unlinea izable due o ei he lack o
o de ing in o ma ion o con lic ing o de ing in o ma ion ( o a de ailed accoun c . Chen-Main,
2006; p. 56-62)
87
(33-3) De ini ion o a syn ac ic g aph (Chen-Main, 2006; p. 53)
A syn ac ic g aph is a i e- uple <N, Q, ID, SP, L>, whe e
N is a ini e se , he se o nodes,
Q is a ini e se , he se o labels,
ID is an i e lexi e, in ansi i e, asymme ic ela ion in N × N, he
immedia e dominance ela ion
SP is an i e lexi e, in ansi i e, asymme ic ela ion in N × N, he sis e
p ecedence ela ion
L is a unc ion om N in o Q, he labelling unc ion, and such ha he
ollowing condi ions hold:
a. Single Roo Condi ion
∃ X ∈ N such ha
∀ Y ∈ N, (X, Y) ∈ ID*
b. Non-O e lapping Condi ion
∀ X, Y ∈ N,
i. i (X, Y) ∈ ID, hen (X, Y) ∉ SP, and
ii. i (X, Y) ∈ SP, hen (X, Y) ∉ ID.
c. Acyclici y Condi ion66
∀ X, Y∈ N,
i (X, Y) ∈ ID+, hen (Y, X) ∉ ID+.
A co olla y o his de ini ion gi en is ha syn ac ic ees could be
conside ed as special cases o g aphs. They a e subjec o an addi ional
condi ion ha a oid om no linea ized s uc u es.
(34-3) Single Pa en Condi ion o syn ac ic ees (based on Chen Main
66 No e ha Chen-Main acycli y condi ion is a c ucial poin o depa u e om my p oposal
(c . Pa ag aph 3.2; 3.3).
88
de ini ion o a syn ac ic g aph see abo e (33-3))
∀X, Y, Z ∈ N,
i (X ≠ Y)
hen ¬ (((X, Z) ∈ ID) and ((Y, Z) ∈ ID))
Resuming all hese a gumen in jus wo wo ds we may say ha Chen-
Main (2006) is ano he in e es ing p oposal ha conside s syn ac ic s uc u es
as di ec ed g aphs ha mee ce ain well- o medness condi ions, and ha hese
condi ions allow some non- ee syn ac ic s uc u es.
3.10 Ano he way o simpli y hings: Mi o Theo y (B ody, 1997)
Mi o Theo y is a syn ac ic amewo k de eloped in (B ody, 1997), whe e
i is o e ed as a consequence o elimina ing pu po ed edundancies in
Chomsky’s minimalism (Chomsky, 1995). A undamen al ea u e o Mi o
Theo y is i s equi emen ha he syn ac ic head-complemen ela ion mi o
ce ain mo phological ela ions (such as cons i uency).
This equi emen cons ains he ypes o syn ac ic s uc u es ha can
exp ess a gi en ph ase; he mo phological cons i uency o he ph ase
de e mines pa o he syn ac ic cons i uency, he eby uling ou o he ,
weakly equi alen , al e na i es. Ano he undamen al ea u e o B ody (1997)
is he elimina ion o ph asal p ojec ion. Thus he X-ba s uc u e on he le
becomes he mi o heo e ic s uc u e on he igh :
(35-3)
XP X
------>
YP X’ Y Z
X ZP
(B ody, 1997) calls his sys ema ic collapse o X, X’ and XP nodes
“ elescope”. E e y node may now ha e phone ic con en , and child en a e
iden i ied as speci ie s o complemen s depending on hei di ec ion o
b anching; le -daugh e s a e speci ie s and igh -daugh e s a e complemen s
(p e iously, as we know speci ie s we e child en o XP, and complemen s
we e child en o X’). Fu he mo e, he complemen ela ion is a “wo d-
89
o ming” ela ion, whe e acco ding o he “mi o ing” ela ion, he phone ic
con en o each head ollows he phone ic con en o i s complemen . Fo
example, mi o heo y can gene a e ees like he ollowing, which gi en he
“mi o ” ela ion be ween mo phology and syn ax, is p onounced John sleep -
s:
(36-3)
-s
Johni Sleep
i
Kobele e al. 2002 in a wo k based on Joshi (1987)67 gi e a o mal
ep esen a ion o a Mi o g amma . A mi o heo e ic ee can be iewed as
a s anda d bina y b anching ee oge he wi h wo unc ions; one, a unc ion
om b anches o a wo elemen se { igh ; le }, he o he , a unc ion g om
nodes o a wo elemen se {s ong; weak}. I a is he pa en o a’, hen a’ is a
speci ie (o le child) o a i ((a; a’)) = le and a complemen (o igh child)
o a o he wise.
In basic e ms, a mi o heo e ic exp ession is de ined o be a mi o
heo e ic ee along wi h a labelling unc ion om he nodes o he ee o a
se o labels. A label consis s o a phone ic pa (which is opaque o he
syn ax) and a ini e sequence o syn ac ic ea u es. A mi o heo e ic
g amma consis s o a ini e lexicon o ‘basic’ exp essions, oge he wi h wo
s uc u e building ope a ions, me ge and mo e, which build exp essions om
o he s ei he by adjoining s uc u es, o by displacing sub-pa s o s uc u es.
Each ope a ion in Mi o Theo y is ea u e d i en, and ‘checks’ ea u es
(and hus a de i ed exp ession will ha e ewe ea u es han he sum o al o
he ea u es o he exp essions ( okens) used o de i e i ). The exp essions
gene a ed by he g amma a e hose in he closu e o he lexicon unde he
s uc u e building unc ions.
A comple e exp ession is one all o whose ea u es ha e been checked, sa e
o he ca ego y ea u e o he oo , and he s ing language a a pa icula
ca ego y is simply he yields o he comple e exp essions o ha ca ego y.
67 I ankly sugges o ead Joshi, A a ind K. 1987. An In oduc ion o T ee Adjoining
G amma s. In A. Manas e -Rame , edi o , Ma hema ics o Language. John Benjamins,
Ams e dam.
96
desc ibes phases as sel -con ained componen s o de i a ion, and asse s ha
in e nal elemen s o a gi en phase mus be on i s phase’s edge, be o e mo ing
ou o ano he phase. Wi hin a adi ional pa adigm, i is a mus o desc ibe
Pe sian CP as head-ini ial, because his is an easy way o accoun o many
ac s conce ning CPs in Pe sian:
a) complemen clauses ollow ma ix clauses;
b) ela i e clauses ollow ma ix clauses;
c) he in e oga i e pa icles aya/maga (i ) a e he “ opmos ” i ems in
in e oga i e sen ences;
d) al hough wh-wo ds do no necessa ily mo e, when hey mo e, i is o he
beginning o an u e ance.
See he examples below:
(10-4)
a. ne-mi-dun-e [CP ke a da mi-yam]
NEG-DUR-know-3S ha omo ow DUR-come-1S
‘He doesn’ know I’m coming omo ow.’ (Mahoo ian 1997: 90)
b. un ma d-o [CP ke uzname mi-xund] peyda ka d
ha man-OM ha newspape DUR- ead isible did
‘He ound he man who was eading he newspape .’ (Mahoo ian 1997: 34)
c. aya in go be-ye-shoma-s ?
INTER his ca -Ez-you-is
‘Is his you ca ?’ (Mahoo ian 1997:9)
d. ce a ma sake be-man-im?
Why we quie SBJN- emain-1P
‘Why do we emain quie ?’ (Dehda i 2006: 45)
Conce ning spli headedness an implied s a emen was made by Ka imi
(2005) assuming he ollowing clause s uc u e o Pe sian:
(Ka imi, 2005: 7)
(11-4)

97
I p opose he ollowing able (adap ing and e ising he one p oposed in
Dehda i, 2006) o summa ize he e idences gi en by he empi ical da a:
(Tab. 4-1)
Rela ionship
Head ini ial
Head inal
Pe Aux - VP
√
V – manne ad e b
√
V – Obj
√
Ve bal Copula – P ed
√
V – PP
√
Passi e Aux – V
√
P eposi ion – N
√
De – N
√
Num – N
√
N – Rela i e encli ics
√
N – Gen (Ez)
√
Adj – supe la i e (es. a in)
√
Complemen ize – S
√
In e oga i e – S
√
Tense (es. mi pa icle) – VP
√
Ad e bial Subo dina o - S
√
98
I would be logic o assume ha human pa se s a e quicke when he
b anching is consis en in one di ec ion. Mixed-b anching seems ha d o
p ocess. Howe e , Pe sian da a seems o e eal a mixed model.
Kayne (1994) challenges he abo e iew. Acco ding o Kayne's heo y,
linea o de ing is mapped om asymme ic c-command ela ions ha hold
be ween non- e minal nodes; hus, dis inc wo d o de ings should no e lec
dis inc hie a chical s uc u es (c . Chap e 5). Kayne (1994) a gues ha head
inal languages ha e a head-ini ial unde lying s uc u e wi h abs ac
unc ional ca ego ies, o which mo emen ope a ions apply so as o de i e he
appa ean head- inal o de . I is in e es ing o no ice ha Chomsky (1995)
poin s ou some weaknesses o Kayne’s heo y such as i s c ucial eliance on
non-b anching nodes o deduce su ace o de .
Ano he ype o wo d o de a ia ion ha is no s ongly linked o clea
syn ac ic (o seman ic) di e ences has been e e ed o sc ambling, he
widesp ead phenomenon o Pe sian syn ax. See he ollowing example om
Japanese, which is accep able, wi hou a hea y s ess on he ini ial cons i uen
o a pause a e i :
(12-4) Ma y-o John-ga mi- a
Ma y-Acc -Nom go-Pas
'Ma y, John saw.'
One dominan app oach ep esen ed by Sai o (1985) and subsequen wo k
is o ega d (12-4) as de i ed by he syn ac ic ope a ion o sc ambling (c . also
Ka imi, 1999 o Pe sian da a). Sc ambling, in a adi ional amewo k,
p oduces an adjoined s uc u e and he mo ed cons i uen lea es a ace in i s
o iginal posi ion. So we ha e o assume hie a chical p ope ies ha di e
om he non-sc amblel (base-gene a ed) ins ance.
As we ha e seen in chap e 3 a syn ac ic g aph heo y allows o o mula e
algo i hms ha can deduce mul iple PF in e p e a ions om he sha ed
syn ac ic s uc u e.
The lexibili y in wo d o de o he mul iplici y o PF in e p e a ion
appea s o be a es ed in head- inal languages (Fukui (1993); Yasui (2004)).
Bu also head ini ial languages71 ha e a iable deg ees o wo d o de
eedom.
71 Fo ins ance, English, which is head-ini ial, allows a ce ain amoun o wo d o de
eedom by shi ing a hea y NP igh wa d as shown in (ia,b), bu a ligh cons i uen like he
p onoun i canno be shi ed igh wa d, as shown in (id):
99
I can be said ha head- inal languages allow wide a ia ion in wo d
o de han head-ini ial languages, and he a ia ion in he o me is always
le wa d; igh wa d wo d lexibili y is highly limi ed. Ano he
impo an asymme y be ween head-ini ial and head- inal languages was
poin ed ou by B esnan (1972).
B esnan (1972: 42) s a es ha only languages wi h clause-ini ial
Complemen ize pe mi a Complemen ize -a ac ion ans o ma ion72.
Fukui (1993) p oposes a s imula ing heo y o he co ela ion be ween a
alue o he head-pa ame e and wo d o de lexibili y based on a
g amma ical ope a ion (Mo e alpha) ha c ea es a s uc u e ha is inconsis en
wi h he alue o a gi en pa ame e in a language is cos ly in he language,
whe eas one ha p oduces a s uc u e consis en wi h he pa ame e alue is
cos less (Minimalis economy is well in e p e ed in his way). Acco ding o
his idea, sc ambling in head inal languages is o no cos since i mo es a
cons i uen le wa d and does no des oy he head- inali y.
He e is Fukui's accoun (1993: 400):
a. A language has a cos less op ional mo emen (o shows lexible wo d o de )
only i i is head- inal, and he ope a ion p oduces a s uc u e consis en wi h he
head- inal alue (i.e., i is le wa d).
b. A language has a cos ly obliga o y mo emen only i i is head-ini ial, and he
ope a ion p oduces a s uc u e inconsis en wi h he head-ini ial alue (i.e., i is
le wa d).
(i)
a. They b ough he beau i ul d ess in o my oom,
b. They b ough in o my oom he beau i ul d ess. (Fukui (1993: 410))
c. They b ough i in o my oom.
d. *They b ough in o my oom i .
On he o he hand, o example, Japanese sc ambling mo es a cons i uen le wa d,
whe he i is ligh o hea y, as shown in (iia,b):
(ii)
a. sono u ukusii do esu-o ka e a-wa wa asi-no heya ni mo eki a.
ha beau i ul d ess-Acc hey-Top I-no oom o b ough (Fukui (1993: 410))
b. so e-o ka e a-wa wa asi-no heya ni mo eki a.
i -Acc hey-Top I-no oom o b ough
72 Logically possible bu non-exis en would be languages wi h a clause- inal
Complemen ize ha a ac s a wh-ph ase igh wa d.
100
C ucially, i is possible o obse e ha in Pe sian long dis ance sc ambling
occu when in ol e head- inal elemen s73, while, in e es ingly, i is blocked
by any in e ening NP wi h he same case as he NP being long dis ance
sc ambled74 (c . Ka imi, 1999: 174- 176; Richa ds, 2002: 240).
_________________
| |
(13-4)a. *Sasan Kimea go [ke _ ke ab-â- o az Sepide kha ide]
Sasan Kimea said ha book-Pl-Spec.Acc om Sepide bough
“Sasan, Kimea said [ ha _ bough he books om Sepide]”
________________________________
| |
b. *be Ali Sasan be Kimea go [ke ke ab- o _ dâde]
o Ali Sasan o Kimea said ha book- Spec.Acc ga e
“To Ali, Sasan said o Kimea [ ha he ga e he book _ ]
_________________________________________________
| |
c. *dokh a -â- o Kimea pesa -â- o ash ig ka d [ke _ bebusand]
gi l-Pl-Spec.Acc Kimea boy Pl-Spec.Acc encou agemen did ha kiss
“ he gi ls, Kimea encou aged he boys [ o kiss _]
73 Following Yasui (2004a), I hink ha an elegan accoun o Pe sian wo d o de may be
gi en i we simply admi he possibili y o an algo i hm shi in he cou se o he de i a ion (c .
Chap e 3).
Indeed, he PF-in e p e a ion o a s anda d syn ac ic ee is ob ained by igno ing i s non-
e minal nodes and p onouncing i s e minal nodes om le o igh , and he o de ing o
e minal nodes comes om a alue o he head pa ame e se o he language in ques ion,
while, ob iously, a lexical g aph equi es a di e en (shi able) PF-in e p e i e algo i hm,
since all i s nodes need o be p onounced.
Does a lexical g aph possess he linea i y equi ed o i s PF in e p e a ion? I claim ha i
o igina es in i s o e all con igu a ion. Lexical i ems a e in oduced in o a syn ac ic de i a ion
one by one, p oducing a la ge s uc u e a e e y s ep. F amp on and Gu mann (1999, 2002)
make his poin clea in hei heo y o c ash-p oo compu a ion: lexical i ems a e in oduced
au oma ically in he igh o de , and no c ash caused by inco ec o de o selec ion is
possible. Then, i is na u al o expec he igh o de o s uc u e-building o be e lec ed in
he PF wo d o de .
As I ha e al eady poin ed ou in he p e ious chap e , ollowing Yasui (2004), ee a e sal
algo i hms can be classi ied in o wo majo ca ego ies: dep h-p io i y and wid h-p io i y
a e sals. Wha seems o be ele an o a e sal o na u al languages is he o me : s a ing
om he oo , we go as deep as possible un il eaching some lea node, ypically he le mos
one ( o spa ial- empo al easons); we mo e back o i s mo he node and isi he o he
child en i any; he emaining nodes a e a e sed in he same manne .
74 No ice ha his ac s seems o hold in Japanese as well (c . Sai o, 1985:185, ci ed in
Richa ds, 2002), whe e sc ambling o a subjec pas ano he subjes is impossible:
________________
| |
(i) *Sono Okasi-ga John-ga [ _ oisii- o] omo ei u
his candy-Nom John-Nom as y ha hinks
“This candy, John hinks [ _ is as y]
101
4.1.3 Some obse a ions on Pe sian ela i es: a Case
A ac ion phenomenon
Pe sian ela i e clauses a e usually in oduced by he complemen ize ke
( ha ), which is used ega dless o he animacy, gende o unc ion o he
head noun (Ka imi, 2001). In non- es ic i e ela i e clauses, he head noun
o en ca ies an encli ic mo pheme (Encl) which links he noun o he
ollowing ela i e clause (14-4). I he ela i ized noun is he objec o he
main sen ence, hen i may appea wi h he objec ma ke â as illus a ed in
(15-4). Tha ’s an in e es ing empi ical obse a ion. The ollowing examples
a e om Mege doomian (2001).
(14-4) zan-i ke injâ neshas e as hamsa -e Nâde as
woman-Encl ha he e si -Pa is spouse-Ez Nade is
`The woman ha is si ing he e is Nade ’s wi e”
(15-4) ke âb-i- â ke di uz kha ide budam em uz- sobh
amâm ka d-am
book-Encl-Obj ha yes e day bough was-1sg oday-mo ning
inish did-1sg
“This mo ning, I inished he book ha I had bough yes e day.”
The ela i e clause may be sepa a ed om he head noun by he main e b
as illus a ed below (see Mege doomian, 2001 and F anco, 2004 o a mo e
a icula ed discussion). In addi ion, se e al ela i e clauses could ollow a
head noun. The ollowing example in aken om Ghomeshi (2002).
(16-4) mâ pesa -ân-i- â en ekhâb mi-kon-im ke da jang
she ka na-ka de-and
we boy-Plu -Encl-Obj choosing Imp-do-1pl ha in wa
pa icipa ion neg-done-3pl
“We choose ( he) boys ha ha e no pa icipa ed in he wa .”
I he head noun is he subjec o di ec objec o he ela i e clause, i is
o en le as a gap as was shown in he examples in (14-4) and (15-4).
Howe e , e en in such cases, he ela i ized noun may be eplaced by a
esump i e p onoun in he clause i o igina ed om. Thus, in (17-4), an
example aken om Mege doomian doc o al hesis (2001), he head noun

102
plâk-e kuchak (small plaque) is he subjec o he ela i e clause; i is
subs i u ed by he esump i e p onoun ân (i ).
The use o he esump i e p onoun usually occu s when he head noun is
sepa a ed om he ela i e clause by an in e ening e b (c . McCloskey,
1992). In his example, he e b pey bo de-and (ha e ound) p ecedes he
ela i e clause.
(17-4) dâneshmand-ân be plâk-e kuchak-i da maqz pey-bo de-and ke
ân niz âkonun nâshenâx e mânde bud.
scien is -Plu o plaque-Ez small-Encl in b ain ound-3pl ha
i also un il now unknown emained was
“Scien is s ha e ound a small plaque in he b ain ha un il now had
emained undisco e ed”.
Thus, when he head noun is he indi ec objec o is ex ac ed om a
P eposi ional Ph ase adjunc in he clause, a esump i e p onoun is used. In
o he wo ds, he posi ion om which he head noun o igina es is subs i u ed
by a p onoun ha ag ees wi h he head noun. This is exempli ied in he
sen ences below:
(18-4) in bache-hâ ke az ânhâ âd es mi-po sid-i...
his kid-Plu ha om hem add ess Imp-ask-2sg
“These kids om whom you asked o he add ess...”
(19-4) shah -i ke da ân azâho â shode bud ...
ci y-Encl ha in i demons a ions become was
“The ci y in which demons a ions ook place...”
(20-4) zan-i ke ba ây-ash ke âb kha id-i ...
woman-Encl ha o -Cli ic(3sg) book buy-Pas -2sg
“The woman o whom you bough a book...”
As shown in F anco (2006) and as al eady men ioned abo e (15-4), he
mo pheme / aa/; / o/ in spoken o m) - as he speci ic ma ke o accusa i e
case - can accompany, a leas in spoken language, he head noun o he
ela i e clause ha is he subjec o he main clause and he objec o he
ela i e clause (21-4) in Pe sian.
(21-4) Zan-i- o [ke did-i] inja-s
103
woman-Acc ha saw-2sg he e-is
“The woman whom you saw is he e”
While his phenomenon conce ning he Pe sian language, known in he
li e a u e as Case A ac ion, esembles o he “In e se A ac ion” discussed
in Bianchi (1999) o La in and Ancien G eek, i has i s own peculia
cha ac e is ics:
a) i is qui e op ional
b) i blocks ex aposi ion, as shown in (22-4),
c) i is always he nomina i e case ha is a ac ed o he accusa i e case
(23-4a,b,c).
(22-4). *Zan-i- o inja-s [ke did-i]
(23-4)
a. pesa -i- o [ke … ]
boy-Encl-obj ha
NOM ⇔ ACC
b. *be pesa -i- o [ke … ]
o boy-encl-obj ha
DAT ⇔ACC
c. * az pesa -i- o [ke … ]
om boy-encl -obj ha
ABL ⇔ACC
Then, his a ac ion only applies o he head noun o he es ic i e ela i e
clause. Since he head noun o he non- es ic i e clause lacks he es ic i e
mo pheme /-i/, i canno a ac he ma ke o he accusa i e case (24-4).
(24-4) * an ma d-e mosen- o [ke di uz did-am] em uz
a
ha man-EZ old-obj ha yes e day saw-I
oday wen -3sg
“Tha old man, whom I saw yes e day, wen oday”.
104
Ka imi (2001), ins ead, a gues ha / aa/,/ o/ as he speci ici y ma ke o
accusa i e case in Pe sian, canno be gene a ed wi h he ela i e head in he
ela i ized posi ion (subjec posi ion o he ela i e clause) (25-4 a;b). Hence,
ejec ing Kayne’s (1994) aising analysis, she sugges s ha he head noun is
base-gene a ed in he Spec o he la ge DP and hen i mo es o he spec o
KsP ( he sugges ed p ojec ion ha has he ma ke in i s head) (26-4).
(25-4). a. Kimea un pesa -i- o [ke inja neshas e bud] be
man mosa e i ka d.
Kimea ha boy-Acc ha he e si ing was-3sg o
me In oduc ion do-pas -3sg
“Kimea in oduced o me he boy who was si ing he e”
b. un… [ CP [ C’ ke [-i pesa - o] inja neshas e bud] ]
ha ha encl boy-Acc he e si ing was-3sg
(26-4) [ksP [un-pesa -i] i [Ks’ – o] [DP i [ D’ ] [CP ] ] ]
(Ka imi, 2001)
Gi en he possibili y o he occu ence o he accusa i e case ma ke / aa/
wi h he subjec o he main clause (when is he objec o he ela i e clause), I
p opose ha , a some poin in he compu a ional p ocess o de i a ion, he
case ma ke / aa/; / o/ was p esen inside he ela i e clause.
Adop ing a aising analysis - as sugges ed by Kayne (1994) and discussed
in Bianchi (1999) and Bha (2002) among o he s - o ela i e clauses in
Pe sian we will come up wi h a s uc u e as in (27-4) o sen ences like he
one in he example (21-4). The case ma ke o he accusa i e oge he wi h
he head noun mo es o he posi ion o Spec o CP and hen he head noun
u he mo es o spec o DP.
(27-4) [ CP [ DP [ zan ] K –i K - o] i [C’ ke i did-i] inja as
I my p oposal is igh , so he empi ical ac s ha I ci ed abo e a e
e idences o he possibili y o a aising analysis o explain he syn ax o
ela i e clauses in Pe sian. The impossibili y o gene a ing he accusa i e
ma ke wi h he head o he ela i e clause in a ela i ized subjec posi ion in
Ka imi (2001) appea as an inadequa e a gumen o ejec ing a aising
105
analysis75: as we ha e shown, he appea ance o he same case ma ke wi h
he subjec o he main clause is an impo an empi ical obse a ion ha
suppo s a aising analysis.
4.1.4 Pe sian Noun Ph ases [a ough guide o he Eza e domain]
The head o a noun ph ase could be a noun o an in ini i al e b. P onouns
and p ope names may also head noun ph ases and hey unc ion as
possesso s in o ming complex noun ph ases (such as possessi e
cons uc ions: ke âb-e Saloomeh (Salome’s book)).
Pe sian head noun is p eceded (a he su ace) by he de e mine s, he
nume al cons uc ions and he quan i ie s, and i is ollowed by he
modi ie s, which usually consis o an adjec i al ph ase (AP). Supe la i e
adjec i es, howe e , do no appea in he AP; ins ead, hey p ecede he head
75 In his no e I show he basic syn ac ic in e p e a ions o he wo majo compe ing
analyses o ela i e clauses: he aising analysis and he ma ching analyses. The head aising
analysis was o iginally p oposed by B ame (1968), Schach e (1973), and Ve gnaud (1974).
Recen e sions include he ele an one o Kayne (1994), among o he s. Unde he head
aising analysis ha we a e adop ing, he head NP o igina es inside he ela i e clause CP, as
shown in (i).
(i) he [book]j [CP [which j]i John likes i]
The ma ching analysis was o iginally p oposed in Lees (1960) and Chomsky (1965) and has
been discussed and ex ended in Saue land (1998). The ma ching analysis pos ula es ha
co esponding o he ex e nal head he e is an in e nal head which is phonologically dele ed
unde iden i y wi h he ex e nal head. Howe e , he in e nal head and he ex e nal head a e
no pa o a mo emen chain. In ac Saue land a gues ha in ce ain cases, he lexical
ma e ial o he in e nal head does no need o be he same as he lexical ma e ial o he
ex e nal head. I jus needs o be simila enough.
(ii) he [book] [CP [which book]i John likes i]
A mo e a icula ed discussion is impossible he e, conside ing he aim o his wo k.
Anyway, I wan o show he consequences o an in e es ing obse a ion made by Richa d
La son (1985). La son obse ed ha headed ela i e clauses con aining a ace in adjunc
posi ion, bu nei he a ela i e ad e b o a s anded p eposi ion, a e g amma ical only i he
ex e nal head o he ela i e clause is a ba e-NP ad e b.
(iii) a. he way [Opi ha you alk i] (La son, 1985, om Bha , 2002)
b.* he manne / ashion [Opi ha you alk i]
c. You alk ha way.
d.*You alk ha manne / ashion.
The well- o medness o he ope a o - a iable chain in (iiia) depends upon wha he head
NP is. In o ma ion abou he head NP is equi ed in e nal o he ela i e clause. Unde a head
aising o a ma ching analysis, he ill- o medness o (iiib) di ec ly ollows om he
ung amma icali y o (8d). This explana ion is no di ec ly a ailable unde he head ex e nal
analysis, and La son, who is assuming he head ex e nal analysis, has o in oduce a no
economical ea u e ansmission mechanism which makes he ele an in o ma ion abou he
head NP a ailable in e nal o he ela i e clause.
112
h ee main kinds o non- e bal cons i uen : ba e noun heads, adjec i al small
clauses, and p eposi ional small clauses. Hale and Keyse analysis d aws i s
p ima y inspi a ion om English (bu includes an inc edible se o da a om
na i e Ame ican languages), whe e he ca ego ial s a us o adjec i al and
nominal e b oo s is e y clea . I gi e a ep esen a ion o Hele and Keyse ’s
unde lying s uc u es o denominal (une ga i e and loca ion/loca um) and
deadjec i al e bs:
(40-4)
This app oach makes he di e ence be ween une ga i e and unaccusa i e
e bs depend on mo e han he X-ba no a ion. I explains he “seman ic-
mo phological” p ope ies o e bs o hese classes. In many languages, he
e balizing pa o he s uc u e is isibly mo phologically ealized as an a ix
o ligh e b, as in hese examples om Hale and Keyse (2002).
(41-4)
a. nega egin “ o c y” (Basque)

113
c y do
jolas egin “ o play” …and many mo e
play do
b. di-yin “ o b ea h” (Na ajo)
do b ea h
di-zheeh “ o spi ” … and many mo e
do spi
On such an app oach, he hema ic p ope ies o a pa icula e b a e
dependen on he syn ac ic and seman ic p ope ies o he e balizing
unc ional elemen and o he non- e bal cons i uen which make i up.
Changing he p ope ies o he e balizing elemen — he ligh e b — esul s
in a change in Agen selec ion: he ligh e b is esponsible o he p esence o
absence o an ex e nal a gumen . Simila ly, he causa i e/ inchoa i e
al e na ion in pai s like John opened he doo /The doo opened is also he esul o
a ying he ligh e b, al hough he mo phological consequences o his
a ia ion a e in isible in languages such as English o I alian. Each o Hale
and Keyse p oposed unde lying s uc u es o English e bs, abo e, ha e
na u al non-inco po a ed coun e pa s in Pe sian complex p edica e
cons uc ions, whe e he ligh e b and non- e bal elemen a e ealized
sepa a ely. Fu he mo e, he agen i i y o a pa icula complex p edica e is
dependen on he ligh e b in ol ed, and he elici y o he complex
p edica e is dependen on he non- e bal elemen in ol ed, in a e y
anspa en way (Ha ley, Folli, Ka imi, 2005).
Pe sian is a language in which he complex syn ac ic na u e o e bs is
e y easily disce ned, and in which Hale and Keyse ’s p oposals conce ning
he s uc u e o he e b ph ase ind wide con i ma ion.
4.1.5.1 De i ing Pe sian a gumen s uc u es (Ha ley, Folli, Ka imi 2003)
We a gued ha une ga i es a e o med when a nominal elemen is
inco po a ed in o a ligh e b which selec s o an ex e nal a gumen .
Simila ly, inchoa i es esul when an adjec i al elemen is inco po a ed in o a
ligh e b which does no selec o an ex e nal a gumen . These s uc u es
ansla e na u ally o Pe sian complex p edica es (Haely, Folli, Ka imi, 2003).
Conside he ep esen a ion o a complex p edica es like ge ye ka dan,
“weeping doing” ha ansla es as a ypical une ga i e like c y:
(42-4)
114
Simila ly, conside he syn ax o a Complex p edica e ha ansla es as a
ypical inchoa i e, like bidâ shodan “awake becoming”:
(43-4)
Jus as hypo hesized by Hale and Keyse o he English
causa i e/inchoa i e al e na ion, he al e na ion be ween he inchoa i e and
he causa i e o awake in Pe sian is accomplished by changing he ligh e b
om he equi alen o 'become' (shodan) o he causa i e 'make' (ka dan).
P obably, he Pe sian case cons i u es he s onges possible e idence o
he syn ac ic na u e o l-syn ax as p oposed by Hale and Kaise .
4.2 The Eza e puzzle
Now, I gi e a de ailed e iew o p e ious analyses o he eza e
phenomenon in he gene a i e amewo k (c . Samiiam, 1983; 1994;
Ghomeshi, 1997; Kahnemuyipou , 2000; F anco, 2004; La son and Yamakido,
2005; Sam elian; 2006 among o he s) and I de elop a g aph analysis o he
Eza e based on Den Dikken and Singhap eecha (2004), whe e he au ho s
gi e a c oss-linguis ic accoun ( he poin o depa u e was he compa ison
be ween F anch and Thai) o he noun ph ases in which linke s occu , in
e ms o DP-in e nal P edica e In e sion (see also Mo o, 1997 and Den
Dikkken 2006).
115
4.2.1 An in oduc ion o Eza e
A numbe o Wes I anian languages - Pe sian, di e en Ku dish dialec s,
Haw ami, Zazaki (La son & Yamakido, 2005) - sha e se e al aspec s in hei
noun ph ase s uc u e:
i) The su ace wo d o de pa e n is s ongly head-ini ial.
Adjec i al modi ie s, he possesso NP, p eposi ional ph ases and
he ela i e clause ollow he head noun, which may only be
p eceded by some de e mine s (in example demons a i es,
ca dinals and quan i ie s), and in e y ew cases by an adjec i e (c .
pa ag aph 4.1.3).
ii) Possession is exp essed by means o a ba e NP (DP)
which ollows adjec i al and some p eposi ional modi ie s80.
iii) Elemen s occu ing be ween he head noun and he
possesso NP a e linked o he head and o one ano he by he
Eza e, ealized as an encli ic mo pheme.
i ) P eposi ional complemen s appea ou side he Eza e
domain and ollow he possesso NP (DP)81
The ollowing Pe sian NP exempli ies hese poin s:
(44-4) in lebâs-e se id-e bi âs in-e Ma yam
his d ess-EZ whi e-EZ wi hou slee e-EZ Ma yam
“ his Ma yam’s slee eless whi e d ess”
(Sam elian, 2006)
80 The exp ession o possession by means o an NP in close cons uc ion wi h he head
noun, and, o some ex en , he Eza e cons uc ion is eminiscen o he Semi ic cons uc s a e
cons uc ion (Bo e , 1988). Despi e he ac ha he possesso NP is no cons ained o be
s ic ly adjacen o he head noun, as i is he case in cons uc s a e nominals, he cons i uen s
occu ing be ween he head noun and he possesso NP ha e been ne e heless assumed o
be subjec o some signi ican cons ain s, leading o he analysis o he Eza e domain as a
domain o ba e heads (X°s) adjunc ion in Pe sian (Ghomeshi, 1997). This is eminiscen o he
wo d-like p ope ies o cons uc s a e nominals (Bo e , 1988).
81 I ’s in e es ing o no ice ha he wo d o de pa e n is iden ical o ha o he Cel ic noun
ph ase. Howe e , unlike Cel ic languages, he wo d o de in he noun ph ase does no
pa allel he one in he clause s uc u e. Al hough Pe sian is e b inal, he e e sed o de
wi hin he noun ph ase has p o ided mo i a ion o he applica ion o he head mo emen
analysis o he nominal domain in some wo ks (Kahnemuipou 2000; F anco, 2005).
116
The Eza e cons uc ion has been a pa icula ocus o in e es in di e en
ecen s udies (Samiian, 1983; 1994; Ghomeshi 1997; Kahnemuyipou , 2000;
La son and Yamakido, 2005 among o he s). Ac ually his cons uc ion aises
se e al issues in syn ax (and also mo phology, see Sam elian, 2006) mainly
he s a us o he Eza e i sel .
The Eza e has gene ally been assumed by Pe sian g amma s (see Laza d
1992) o be seman ically acuous. Fu he mo e, i can be i e a ed wi hin he
NP, occu ing as many imes as he e a e modi ie s. Bu , a leas in Pe sian, i
is no he exp ession o a conco d be ween he head noun and i s dependan s
( his is no a case o some Ku dish dialec s; see below).
On he basis o hese obse a ions, Samiian (1983) and Ghomeshi (1997)
p opose no o iew he Eza e as a mo pheme a all, bu a he as an elemen
inse ed in Phonological Fo m (see Chomsky, 1981). Fo Ghomeshi (1997) -
maybe he leading analysis o his phenomenon - he need o he Eza e
owel esul s om he ac ha nouns being non-p ojec ing in Pe sian, a
“phonological linke ”, in example he Eza e, mus be p esen in o de o
indica e ph asing wi hin he nominal cons i uen .
This iew o he Eza e has been ejec ed in subsequen s udies and a ious
al e na i e analyses ha e been sugges ed. Eza e has been seen:
i) as a Case-ma ke (Samiian 1994, La son and Yamakido
2005; 2006).
ii) as a ma ke associa ed wi h he syn ac ic mo emen o
he noun and ealizing a “s ong ea u e” (Kahnemuyipou , 2000).
iii) as a he mo phological ou e i em o a unc ional head in
he domain o AP, in a “ca og aphic app oach” (F anco, 2005,
ollowing Cinque, 1994).
i ) as a su ix a aching o he head and o some o i s
in e media e p ojec ions, and ma king hem as awai ing a modi ie
o a complemen . (Sam elian, 2006, ollowing Nichols, 1986).
In he ollowing sec ions wo k, a e an his o ical excu sus and he e iew
o he mos in e es ing analysis o he Eza e phenomenon in he gene a i e
pa adigm, I will made a new g aph based p oposal ollowing some
obse a ions de eloped by Den Dikken and Singhap eecha and Den Dikken
(2006) o Chinese – Manda in, which may lead us o conside Eza e as a
linke indica ing subjec p edica e in e sion.
This iew o he Eza e is qui e he opposi e o he Case-ma ke analysis
sugges ed by Samiian (1994) and La son and Yamakido (2005; 2006),
117
acco ding o which he Eza e is a he a dependen ma king de ice.
4.2.2 The Eza e cons uc ion: a ough his o ical pe spec i e and a basic
o e iew o Eza e domain
The Eza e cons uc ion is a speci ici y o hose languages ha display a
head-ini ial wo d-o de pa e n wi hin hei DP (e.g. Pe sian, Ku dish
dialec s, Haw ami, Zazaki, Ke manian dialec s, e c). The co ela ion be ween
he head-ini ial wo d o de pa e n and he a ailabili y o he Eza e may be
accoun ed o on his o ical g ounds82 (c . Sam elian, 2006). The encli ic Eza e
has p obably i s o igin in a demons a i e- ela i e mo pheme in Old I anian.
In Pe sian, i can be ela ed o hya ( ya), a demons a i e, linking he head
noun o adjec i al modi ie s, o he possesso NP and also o a ela i e clause
in Old Pe sian:
(45-4) Kā a hya manā [Da mes e e (1883) om Sam elian
(2006)]
‘my a my; he a my which is mine’
(46-4) kāsaka hya kapau aka [Meille (1931) om Sam elian
(2006)]
‘ he blue s one’
(47-4) i ānam ja ā u ā a am kā am hya dā aya ahauš xšāyahiyhyā.
[Meille (1931) om Sam elian (2006)]
‘Bea Vi âna and his a my which decla es i sel as a p oponen o he
king Da ius.’
I ’s in e es ing o no ice ha hya ( ya) is no a simple linke , bu ha i
u he has a demons a i e alue. The demons a i e hya ( ya) can unc ion
as a head by i sel :
(48-4) ima ya adam akuna am [Meille (1931) om Sam elian (2006)]
‘This is wha I did’
82 Howe e , he e is no necessa y co ela ion be ween he head ini ial wo d o de pa e n
wi hin he NP and he Eza e. Some Sou h-wes e n I anian languages, which a e e y close o
Pe sian, dispense wi h he Eza e. In di e en Ke manian dialec s, o ins ance, he Eza e,
al hough a ailable, due o a close and longs anding con ac wi h Pe sian, is ha dly e e used
o is op ional (Rebuschi, 2002). Simila ac s a e also obse ed in some No h Wes e n I anian
languages, such as Tâ i dialec s (Lecoq 1989) o in Sou he n Ku dish dialec s (Fa ah 2000).

118
Hya ( ya-) becomes –i in Middle Pe sian and p og essi ely looses i s
demons a i e alue o end up as a simple linke . Con a y o Pe sian,
Ku dish and Zazaki (c . La son & Yamakido, 2006) ha e s ill a so-called
“Demons a i e Eza e”, di e en om he a ixal Eza e, which unc ions as a
demons a i e p onoun heading nominal ph ases, as shown in he ollowing
examples.
(49-4) yê dwê … yê sêyê
EZ second EZ hi d
“ he second one, he hi d one”
(Ku manji Ku dish, om La son & Yamakido, 2006)
(50-4) ki êb-î min o hî o
book-EZ my and EZ-you
‘my book and you s
(So ani Ku dish om Sam elian, 2006)
Nowadays, when a ailable, in Wes e n I anian languages he Eza e
gene ally links he head noun o i s adjec i al modi ie s and o he possesso
NP, as illus a ed by he ollowing examples o Pe sian aken om
Kahnemuyipou (2000) and o So ani Ku dish aken om Sam elian (2006):
(51-4) lebâs-e se id-e donya (Pe sian)
d ess- EZ whi e- EZ Donya
“Donya whi e d ess”
(52-4) ki asêk-î hin-î Na mîn (So ani Ku dish)
d ess-EZ blue-EZ Na mîn
“a blue d ess o Na min’s”
Fu he mo e, he complemen o an adjec i e can also be in oduced by he
Eza e:
(53-4) Saloomeh mašqul-e kâ as (Pe sian)
Salomé occupied-EZ wo k be-3sg-p es
‘Salomé is busy wo king’
119
The ac ha adjec i es can ake he Eza e is no su p ising. Indeed,
adjec i es beha e like nouns in many espec s, so ha in Pe sian, o ins ance,
i is qui e impossible o es ablish a dis inc class o adjec i es, and se e al
i ems a e indis inc ly used as nouns o adjec i es, depending upon he
con ex (Laza d, 1992).
A mo e in e es ing empi ical obse a ion is he use o he Eza e wi h some
p eposi ional heads. This occu s in Pe sian, o ins ance, as illus a ed by he
ollowing examples:
(54-4) a. ba â-ye Saloomeh
o -EZ Salomé
b. aleyh-e Saloomeh
agains -EZ Salomè
“ o Salomé” ; “agains Salomé”
I has been a gued by Ghomeshi (1997) ha p eposi ions occu ing wi h
he Eza e a e no in ac p eposi ions, bu nouns.
Real p eposi ions such as bâ “wi h”, az “ om”e c., by con as , ne e occu
wi h he Eza e (c . also Chap e 5). This assump ion is a guably app op ia e
o loca i e p eposi ions, such as zi “unde ”, poš “behind”, which a e
o iginally nouns and display a ange o nominal p ope ies (Samiian, 1994;
La son and Yamakido, 2005): he ph ase hey head can unc ion as he subjec o
he di ec objec o a sen ence.
I aces howe e se ious p oblems when applied o i ems such as ba â
“ o ”, alâ aqm “despi e”, and aleyh “agains ”, which ha e none o he
dis ibu ional and mo phological p ope ies o nouns83. Consequen ly, he
analysis o hese i ems as nouns is exclusi ely mo i a ed by he ac ha hey
can be ma ked by he Eza e. Anyway, no only is i unclea wha such an
analysis could gain om i , bu also a mo e mys e ious is he way i could
wo k!
Con a y o Ghomeshi (1997), and acco ding o La son and Yamakido
(2005), I hink ha is possible o conside hese i ems as p eposi ions, which
implies ha , wi h ega d o he Eza e, p eposi ions ha e, simply, wo dis inc
sub ypes: hose which ake he Eza e and hose which do no ake he Eza e ( his ac
will ha e some consequences o some u he discussions aised in he
ollowing Chap e ).
83 Fo ins ance, he cons i uen s hey head can ne e be he subjec o he di ec objec o a
sen ence (c . Sam elian, 2006).
120
Apa om i s ypical uses o in oduce adjec i es and he possesso NP,
he Eza e mo pheme can in oduce p eposi ional ph ases and ad e bial
ph ases. The possibili y o p eposi ional ph ases o be in oduced by he
Eza e in Pe sian is illus a ed by he ollowing example akan om he no el
Yek uz mânde be eyd-e pâk by Z. Pi zâd and ci ed in Sam elian (2006)
(55-4) ne-mi a ânes -am asmim begi am sobh-hâ-ye [PP bâ mâda - â]
biš a dus dâ -am yâ sobh-hâ-ye [PP bâ kabu a -hâ- â].
NEG-can.Pas -1sg decision ake-1Sg mo ning-pl-EZ wi h mo he -Acc
mo e like.P es-1sg o mo ning-Pl-EZ wi h pigeon-pl-Acc
“I could no decide whe he I lo ed be e he mo nings wi h mo he o
heones wi h he pigeons”
Once again, con a y o wha has been claimed in some p e ious s udies
(Ghomeshi 1997, La son and Yamakido 2005), he e is no ban on he p esence
o PPs wi hin he Eza e domain, wha e e be he ype o he head p eposi ion.
The si ua ion is mo e con as ed wi h espec o ela i e clauses (c .
pa ag aph 4.1). In Pe sian, only educed ela i es can be linked o he head
noun by he Eza e (c . Sam elian 2006, om which a e aken he ollowing
examples. I ecommend o e e o he wo k o a c oss-linguis ic su ey and
o mo e de ails84):
(56-4) in ja ân-e [az Suis ba gaš e]
his young-EZ om Swi ze land e u n
“ his boy e u ned om Swi ze land”
84 Sam elian also obse es ha Ku dish, by con as , allows o all ela i e clauses o be
in oduced by he Eza e:
(i) mi o -ê ku min dî -î (Ku manji)
man-EZ.M.SG ha I see.PAST
‘ he man whom I saw’
(ii) aw şâ -a-y (ka) dî -mân (So ani)
ha own-DEF-EZ ( ha ) see.PAS-1.P
‘The own ha we isi ed’
Fu he mo e in he Ku manji dialec , e en a non- ela i e subo dina e clause which is a
dependen o a noun can be in oduced by he Eza e, as illus a ed by he ollowing example:
(iii) bi xayâl-â [ko aw ji bâjê da ka i bûn]
a imagina ion-EZ ha hey om ci y ou we e
‘Imagining ha hey we e ou side he ci y...’
[ aken om Bedi Khan and Lesco , 1970]
121
(57-4) * ke âb-i-e ke u-ye miz as
book- el-EZ ha on-EZ able be.PRES
“The book ha is on he able”
In he ollowing sec ion I will y o ou line, om a diach onical pe spec i e,
he analyses o he Eza e-mo pheme in he gene a i e li e a u e o he las
wo decades.
4.2.3 Res ic ion on Eza e (Samiian, 1983)
Vida Samiian (1983) is he i s de ailed s udy on Eza e in Pe sian wi hin a
mode n syn ac ic amewo k, namely X-ba heo y. The empi ical ac s
men ioned by Samiian ha e been aken up in subsequen wo ks (Ghomeshi
1997, Kahnemuipou 2000, La son and Yamakido, 2005), al hough hey ha e
been accoun ed o in a adically di e en way. In his sec ion, I will conside
di e en es ic ions on he Eza e cons uc ion poin ed ou by Samiian (1983),
hen, in he ollowing one I will hen gi e a de ailed accoun o Ghomeshi
(1997).
Samiian (1983) poin s ou wo majo ypes o es ic ions on he Eza e
cons uc ion: he cons i uen s occu ing wi hin he Eza e domain a e s ic ly
o de ed and cons ained wi h ega d o hei dis ibu ion. In dep h, acco ding
o Samiian, di e en elemen s linked by he Eza e o he head noun occu in
a ixed o de :
(58-4) ke âb-e â iz-e sabz-e bija zes-e Saloomeh
book-EZ his o y-EZ g een-EZ wi hou alue-EZ Salome
“Salome’s g een his o y book wi hou any alue”
In he example (58-4) he head noun is ollowed by an a ibu i e noun, an
adjec i e modi ie , a p eposi ional modi ie and a possesso NP:
(59-4) Head Noun – A ibu i e noun – Adjec i e modi ie – P eposi ional
Modi ie – Possesso NP
Al hough Samiian (1983) conside s his o de o be a s ic one, i mus be
no ed ha he only eal cons ain conce ns he placemen o he possesso
NP, which mus occu in he inal posi ion wi hin he Eza e domain, any
o he posi ion being excluded o he possesso (c . also Kahnemuipou ,
128
accoun o he Eza e owel be o e a ibu i e adjec i es (as I ha e a gued
abo e i is no clea why a ibu i e adjec i es need case).
Following he p inciples o Case Theo y (c . Haegeman, 1996) and
S owell’s Case Resis ance P inciple (1981), Samiian assumes ha non-case-
assigne s need case, whe eas Case-assigne s do no . The igge o he
inse ion o Eza e is, hen, he lack o Case-assigning p ope ies. Hence, Eza e
appea s only on ca ego ies ha canno assign case, such as [+N] ca ego ies,
and doesn’ appea on case-assigning ca ego ies, such as [-N]. Gi en his, a
ques ion a ises: why mos p eposi ions in Pe sian ake hei complemen ia
Eza e?
Since e bs and p eposi ions a e bo h case-assigne s by i ue o hei [-N]
ea u e, he la e a e no expec ed o need some special de ice o aking a
complemen . Howe e , he ac emains ha he only ph asal ca ego y whe e
Eza e is no ound is he e b (bu see pa ag aph 4.3 o an in e es ing
empi ical ac ).
Samiian di ides he p eposi ions in Pe sian in wo g oups — hose which
do no ake Eza e (Class 1; C1) and hose which ake Eza e (Class 2, C2). C1
p eposi ions possess all he p ope ies associa ed wi h p eposi ions, including
he abili y o di ec ly assign case. C2 P eposi ions, on he con a y, exhibi
some nominal p ope ies including he inabili y o assign case. In o de o
explain he syn ac ic beha iou o C2 P eposi ions, Samiian adop s he
Neu aliza ion Hypo hesis o Ge man adjec i es, o which she ci es he wo k
o an Riemsdijk (1991).
Unde his p oposal, Ge man adjec i es a e neu alized in hei [+N]
ea u e, ha is, hey a e speci ied only o he [+V] ea u e, a he han ully
speci ied [+V,+N] elemen s. Consequen ly, as a [+V] ca ego y hey a e non-
dis inc om he [+V,-N] ca ego y, om e bs. This p o ides an explana ion
o he ac ha adjec i es and [-N] ca ego ies in Ge man sha e some
p ope ies: a leas , hey all can assign case.
Following he same line o easoning, Samiian sugges s ha C2
P eposi ionss a e neu alized wi h espec o hei [-N] ea u e, hus hey ha e
only he ea u e speci ica ion [-V]. The ull pa adigm o Pe sian lexical
ca ego ies acco ding o Vida Samiian is gi en as ollows in (71-4).
(71-4) N: [-V,+N]
V: [+V,-N]
A: [+V,+N]
C1 P: [-V,-N]

129
C2 P: [-V]86
(Samiian 1994: 38)
Building on Samiian (1994) and ollowing he La sonian DP-shell
s uc u e87 (c . La son 1991-2006), La son and Yamakido (2005) sugges oo
ha Eza e is a case ma ke .
Howe e , hei accoun di e s om he one sugges ed by Samiian. La son
and Yamakido make he in e es ing assump ion ha C2 p eposi ions a e
nouns and hus, by elimina ing he dis inc ion be ween mos Pe sian
p eposi ions and nouns, hey a e able o p o ide a uni ied syn ac ic accoun
o all nominal modi ie s. They p opose ha nominal modi ie s a e gene a ed
as a gumen s o D pos -nominally, in he posi ion o ela i e clauses. As [+N]
elemen s hey equi e Case, hence, in English, hey mo e up o ge case-
licensed by he De e mine . Pe sian, howe e , has a i s disposal he Eza e
ma ke , which acco ding o hem is a special “de ice” o making Case
a ailable in he base posi ion. Thus, Eza e allows he unde lying pos -
nominal posi ion o nominal modi ie s o eme ge, since hey a e case-licensed
in hei base posi ion.
In o he wo ds, La son and Yamakido p opose an analysis in which
(p edica i e) nominal modi ie s o igina e pos nominally as inne a gumen s
o D, which only a e wa ds combines wi h NP and a subsequen ly me ged
ligh de e mine (d) a ac s he D head, de i ing D–N wo d o de .
Fu he mo e, DP modi ie s a e conside ed o be lowes complemen s o he
head. Unde his accoun , nominal modi ie s, in all languages, o igina e as
a gumen s o D in a pos nominal posi ion. Those DP-modi ie s ha do no
ha e Case ea u es o be checked, PPs and CPs o ins ance, emain in si u.
86 I ’s in e es ing o no ice ha C2 Ps a e le wi h only he ea u e [-V] which makes hem
unable o assign case, since acco ding o Samiian only ca ego ies speci ied o [-N] a e case-
assigne s. The e o e, C2 Ps ha e o make use o a special de ice, mo e speci ically he
dummy case assigne Eza e, in o de o be able o ake complemen s.
87 He e is a b ie explana ion o Richa d La son’s DP heo y. In a pape om 1991 La son
discussed he syn ac ic p ojec ion o DP om he s andpoin o gene alized quan i ie heo y,
and a gued ha , unde he la e , he mos app op ia e analogy is no be ween DP
and CP/TP (c . Abney, 1987; Szabolsci, 1983), bu a he be ween DP and VP. Speci ically,
La son sugges s ha :
(i) DP can be unde s ood as p ojec ing a gumen s acco ding o a
hema ic hie a chy ha is pa allel o (bu di e en in ole-con en om) ha
ound in VP,
(ii) Tha De e mine s so hemsel es in o in ansi i e, ansi i e and
di ansi i e o ms, much like Ve bs, and
(iii) ha nominal modi ie s, including ela i e clauses, p ojec in he DP
e y much like ad e bial elemen s in VP.
130
On he con a y, hose modi ie s ha bea Case ea u es ( o ins ance APs)
a e equi ed o mo e o a si e whe e hey can check Case. The hypo esis o
La son and Yamakido is ascina ing. The p enominal posi ion is assumed o
be such a si e and he e a e languages such as Pe sian howe e ha ha e in
hei D-sys em an i em ha can be inse ed o check Case on [+N] de e mine
complemen s, allowing he la e o emain in si u: Eza e would be he Case-
ma king de ice.
The analysis o Eza e as a Case ma ke , anyway, aces se e al p oblems.
The da a on which i elies is no well g ounded. The majo a gumen
in oked by bo h Samiian (1994) and La son and Yamakido (2005) o suppo
his analysis is he ac ha cons i uen s such as PPs and ela i e clauses,
which no no equi e o be Case-ma ked, a e excluded om he Eza e
domain, bu o he au ho s (c . Sam elian, 2006 o ano he c i ical su ey)
ha e empi ically demons a ed ha in Pe sian PPs headed by P1s, when hey
a e modi ie s, as well as ad e bial ph ases occu wi hin he Eza e domain.88
Fu he mo e, since unde La son and Yamakido’s accoun , Eza e is
supposed o enable [+N] modi ie s, which o he wise should mo e o a
p enominal posi ion, o emain in si u, i is unclea why in o he languages,
o ins ance, educed ela i es a e allowed o emain in si u wi hou being Case-
ma ked, while in Wes I anian languages such a Case-ma king de ice is
necessa y in o de o le hem emain in a pos -nominal posi ion.
The iew o Eza e as a Case-ma ke becomes e en mo e p oblema ic when
o he Wes I anian languages a e aken in o accoun . Fi s , as epo ed in
Sam elian (2006), in almos all g oups o Ku dish dialec s es ic i e ela i e
clauses may be in oduced by Eza e; In addi ion, in Ku manjî dialec s, which
ha e gene ally main ained mo phological case ma king, he Possesso NP
linked o he head noun by Eza e appea s in oblique case, as shown by he
ollowing example, aken om Sam elian (2006):
88 Fu he mo e, hough ela i e clauses canno be linked o he head noun by Eza e in
Pe sian, educed ela i es, on he con a y, can be in oduced by Eza e. This is shown by he
ollowing a es ed examples:
(i) ...nazm-e os ân a asso -e in ja ân-e [az suis ba gash e]RRC bish a hâsel xâhad
âmad
o de -EZ p o ince by means-EZ his young-EZ om Swi ze land back u n.PP mo e
gained AUX come.PAS
‘(...) he p o ince’s o de will be be e es ablished by his young man came back om
Swi ze land’
(ii) aks-e [châp shode da uznâme]RRC aks-e â i-e dâs ân as 17
pho o-EZ publica ion become.PP in newspape pho o-EZ na a o -EZ s o y be.PRES
‘The pho o published in he newspape is he pho o o he s o y’s na a o ’
I is unclea why cons i uen s such as PPs and educed ela i es would equi e o be Case-
ma ked in Pe sian.
131
(72-4) mashl-ah na mîn-ê
house-EZ.Fem Na min-Obl.Fem
“Na min’s house”
Viewing Eza e as a Case ma ke would imply ha he possesso NP
Na min is case ma ked wice o he same unc ion (c . also Kahnemuyipou ,
2000) and his is p oblema ic wi hin any heo y o Case ma king.
Gi en his, ei he he Pe sian Eza e is a adically di e en i em om i s
Ku dish coun e pa , o Eza e is no a Case-ma ke , nei he in Pe sian no in
o he I anian languages. The ac ha PPs and educed ela i e clauses a e
linked o he head noun by Eza e (a leas in Pe sian) s ongly suppo s he
la e conclusion.
4.2.5 Eza e and mo emen (Kahnemuyipou , 2000)
Kahnemuyipou (2000) p o ides an explana ion o Eza e inse ion based
on syn ac ic mo emen . He no es ha i Eza e we e a ma ke inse ed only o
iden i y cons i uen -hood, as p oposed by Ghomeshi (1996), hen he o de o
he modi ie and he noun would be i ele an . Howe e , he e a e cases in
Pe sian whe e he adjec i e p ecedes he noun and no Eza e is inse ed (c .
(73-4)). Mo eo e , he Eza e owel is ung amma ical in his con ex .
(73-4) a. gol-âb lowe -wa e “ ose-wa e ” s. ?âb-e gol
b. bozo g-ma d big-man “g ea man” s. ma d-e bozo g 'big man'
c. ke âb-xune book-house “lib a y” s. ?xune-ye ke âb
(Kahnemuyipou 2000:3)
Kahnemuyipou akes his ac o sugges ha he Eza e cons uc ion is
associa ed wi h syn ac ic mo emen and ha he Eza e owel is he
ealiza ion o a s ong [Mod] ea u e bo ne by modi ie s. He assumes a le -
b anching s uc u e and a p enominal Me ge posi ion o all noun-modi ying
elemen s.
Thei pos nominal su ace posi ion, hen, is de i ed by mo emen o he
noun. Re e ing o Cinque’s (1990) p oposal abou he base posi ion o
adjec i es in he noun ph ase, namely in he speci ie o unc ional ph ases
abo e he NP, and acco ding o my p oposal, independen ly de eloped in
F anco (2004), Kahnemuyipou sugges s ha modi ie s in Pe sian oo head
132
unc ional p ojec ions abo e he noun ph ase and, u he mo e, ha hey
bea he s ong ea u e [Mod]. The noun, which also bea s he ea u e [Mod],
mo es up and head-adjoins o he modi ie , hus checking i s [Mod] ea u e
agains he [Mod] ea u e o he modi ie . The [Mod] ea u e is hen
mo phologically ealized as Eza e on he noun. Kahnemuyipou doesn’
ackle he issue o hose p eposi ions ha ake hei complemen ia Eza e. I
emains unclea whe he hey a e conside ed o o igina e below he G ound
complemen (c . F anco, 2004) and hen mo e up o head-adjoin o i , which
would mean ha he p eposi ion is modi ied by i s complemen .
To esume in some poin s his p oposal, Kahnemuyipou (2000) sugges s
ha he Eza e ma ke is associa ed wi h syn ac ic mo emen and suppose o
Pe sian:
i) Righ b anching s uc u e.
ii) P enominal Me ge posi ion o adjec i es, modi ying nouns, possesso s,
e c.
iii) Basing on Cinque (1990) and Rubin (1997), he assumes ha adjec i es
a e loca ed in he heads o unc ional p ojec ions ModP abo e NP.
i ) The adjec i es (and all elemen s ha modi y nouns) bea a ea u e
[Mod].
) The ea u e [Mod] is s ong and igge s o e mo emen o he noun.
i) The noun mo es up, head-adjoins o he adjec i e and checks he s ong
ea u e [Mod] agains he [Mod] ea u e o he Adjec i e89.
ii) The s ong ea u e [Mod] is mo phologically ealized on he noun
elemen by Eza e.
iii) In his way, he pos nominal su ace posi ion is de i ed by successi e
head adjunc ion o check he s ong ea u e [Mod].
The ep esen a ions o he de i a ional s ages o his analysis a e shown
below:
(74-4) sag-e siah-e gonde
dog-Ez black-Ez big
“big black dog”
89 No e ha he o de o adjec i es in Pe sian doesn’ appea o be as s ic as his p oposal
would p edic (c . F anco, 2004).
133
a. ModP
Adj0 ModP
gonde
[Mod]
Adj0 NP
siâh-e
[Mod]
N0 (CP)
sag-e
b.
c. ModP
Adj0 ModP
Adj0j Adj0 j NP
gonde
Ni0 Adj0 [Mod] i (CP)
sang-e siâh-e

134
4.2.6 An in lec ional a ix? (Sam elian, 2006)
Ano he iew o he Eza e phenomenon is he one o Sam elian (2006)
al eady ci ed he e o many in e es ing c oss-linguis ic examples, who a gue
ha he Eza e in Pe sian could be ega ded as an in lec ional a ix, a aching o
a head noun and o some o i s in e media e p ojec ions and ma king hem as
awai ing a modi ie o a single NP complemen .
Sam elian also demons a e ha he a ixal analysis o he Eza e could be
applied o Ku manji, hough he Eza e cons uc ion does no display he same
ange o p ope ies as in Pe sian.
The analysis ou lined in Sam elian’s pape en ails ha he Eza e pa icle,
whose o igins in Mode n Pe sian can be aced back o he Old Pe sian
ela i e/demons a i e hya/ ya (see abo e pa ag aph 4.2.1), has unde gone a
p ocess o eanalysis-g amma icaliza ion, being hus ein e p e ed as a pa o
he nominal in lec ion.
Gi en i s encli ic s a us (a leas in Mode n Pe sian), i could be assumed
ha he con lic a ising om he equi emen o wo opposi e di ec ions o
a achmen – mo pho-syn ac ic a achmen o he igh and phonological
a achmen o he le – has been esol ed by he eanalysis o he Eza e as a
nominal in lec ional a ix, hus aligning mo pho-syn ac ic a achmen wi h
phonological a achmen .
Such a iew assumes ha Eza e pa icle has ceased o unc ion as a ela i e
pa icle, specializing as a de ice o nominal a ibu ion.
Sam elian hink ha , on he basis o dis ibu ional, p osodic and
mo phological c i e ia, wo majo se s o in lec ional a ixes wi hin he
Pe sian NP will be es ablished. The membe s o he i s se , i.e. he de ini e
su ix –(h)e and he plu al su ix hâ, may be conside ed as wo d-le el
in lec ional a ixes: hey a ach o he head (i.e. he noun) wi hin he NP and
canno be sepa a ed om i by any o he in lec ional a ix. Fu he mo e, hey
bea lexical s ess and canno ha e wide scope o e he coo dina ion o wo
nouns. The membe s o he second se , i.e. he Eza e, he de e mine –i and
pe sonal encli ics, a e a gued o be ph asal a ixes: hey occu a he igh
edge o nominal non-maximal p ojec ions, and a e loca ed a e Se (1)
in lec ional a ixes. See he examples below, aken om Sam elian (2006; p.
13).
(75-4)
a. in bâ âhang hamân-i bud ke pesa -e-ye ilm-e hendi
135
ba â-ye dox a -e mi-zad
his ime melody same-encl be.Pas ha boy-De -EZ ilm-EZ Indian o -
EZ gi l-De Imp-bea .Pas
“This ime he melody was he same as he one he boy in he Indian ilm
was playing o he gi l”.
b. * ke âb-i-e Ma yam
book-Ind-EZ Ma yam
“a book o Ma yam”
c. * hamsâye-ye nega ân-e bachecehâ-yash-e Ma yam
neighbou -EZ wo ied-EZ child-P-PAF.3.P-EZ Ma yam
“Ma yam’s neighbou who is wo ied abou his child en”
In b ie , Sam elian’s wo k sugges s a mo phological accoun o hese
es ic ions in e ms o posi ion class mo phology (S ump, 2001) whe e
collec ions o i ems compe e o ealiza ion in a single posi ion.
In his pe spec i e, he da a in 75-4) ecei e a pu ely su ace-based
accoun , in e ms o cons ain s on a ix s acking, wi h no need o syn ac ic
cons ain s. I his accoun is app op ia e, i is expec ed ha such examples
would become g amma ical in case o a eo de ing o he cons i uen s wi hin
he NP so ha a ix s acking is a oided. This indeed seems o be he case,
gi en he con as be ween (76-4a) and (76-4b):
(76-4) a. *qah emân-e [ ânde shode az mihan-ash]-e in
oman
he o-EZ d i e.Pas become.Pas om homeland-p on.3.S]-
his no el
“ he he o o his no el, who is d i en away om his homeland”
b. qah emân-e [az mihan-ash ânde shode]-ye in
oman
he o-EZ om homeland-PAF.3.S d i e.Pas become.Pas]-EZ his
no el
“ he he o o his no el, who is d i en away om his homeland”
136
4.3 An in e es ing ac : he ense Eza e o he Behdînî-Ku dish Geo ey
Haig (2005)
An in e es ing empi ical ac is epo ed in a pape by Geo ey Haig, who
akes a look a he Eza e pa icle in an ano he Wes e n I anian language, he
Behdînî dialec o Ku dish (BK), spoken in No h I aq. In his language, one
exponen o he Eza e has unde gone a so o di e en de elopmen : i is
a guably no longe pa o an NP (o DP) a all, bu is now a pa icle wi h a
pa icula ense/aspec alue ( ea u e), p esumably pa o a Tense p ojec ion
in he clause. Typical examples ha Haig has aken om MacKenzie (1961),
a e he ollowing:
MacKenzie (1961)
(77-4) xusk-a min ya çuy-î sîk-ê
sis e -Ez.F 1S.Obl Ez.F go:Pas-P cpl ma ke -Oobl
“My sis e has gone o he ma ke ”
(78-4) go -ê ku šah-ê wan yê mi -î
say:Pas- o.him ha king-Ez.M 3Pl.Obl Ez.M die:Pas-P cpl
“(He) said o him ha hei King had died.”
The pa adigm o o ms a ailable o his pa icle is iden ical o he
pa adigm ound o he NP-based Eza e; MacKenzie (1961) had al eady
poin ed ou ha he wo a e e ymologically iden ical. Haig (2005) e e s o he
la e as he Tense Eza e. In MacKenzie’s da a, he ense Eza e was la gely
es ic ed o co-occu ence wi h s a e o loca ional p edica es, and pa icipial
e b o ms. Bu he mo e ecen ieldwo k o Haig shows ha he Eza e
pa icle now egula ly occu s wi h ini e p esen ense e b o ms. Wi h ini e
p esen ense e b o ms he Tense Eza e con as s wi h clauses lacking such
pa icles, indica ing ha he pa icle is indeed now pa o he sys em o
ense/aspec dis inc ions in he language, adding a pa icula sense o
immediacy o cu en ele ance:
(79-4) Ez yê xwa in-ê çê.di-k-im
1Sg Ez.M meal-Obl Imp-do-P es-1Sg
“I am making/p epa ing a meal ( igh now)” (Haig, 2005)
137
s.
(80-4) Ez xwa in-ê çêdi-k-im
1Sg meal-Obl imp-do-P es-1Sg
“I (gene ally) make Ku dish ood” (Haig, 2005)
Then, Haig ies o econs uc he mos plausible diach onic scena io ha
led o he cu en si ua ion, no ing a s iking (and inspi ing, o me) pa allels
o he de elopmen s o copula elemen s om nominal linke s / p onouns in
Heb ew and Manda in90 (c . Li & Thompson 1977 and, also, Den Dikken,
2005) and sugges ing ha he pa hway conce ned, in ol ed he eanalysis o
an o iginally hyb id elemen , he Old I anian ances o o he Eza e (c .
pa ag aph 4.2.1), which con la ed a C and a D p ojec ion, wi h he la e
ul ima ely leading o he Tense-p ojec ion posi ion o he BK Tense-Eza e.
The pa allels I ha e ound in Haig (I mean he ones conce ning he
“e olu ion o he copula”, especially in Chinese Manda in), eminded me he
encoun e s wi h Den Dikken (1995; 2005) and Mo o (1993; 1997; 2000) heo ies
abou copula and, in some ways, ha e been an unconscious igge o he
p esen analysis o Eza e.
4.4. Eza e: Linking he Complex (ne wo k o he) Noun Ph ases
The Eza e mo pheme in ou analysis, pa ly based on Den Dikken and
Singhap eecha (2004) is a linke which, in a adi ional pa h, is in oduced in
he cou se o a P edica e In e sion ope a ion ha in e s a noun-ph ase-
in e nal p edica e wi h i s ph asal subjec .
The cons i uen p eceding he Eza e mo pheme could be in e p e ed, in
ac , as he subjec o he in e ed p edica e, which ells us ha he wo d-o de
e ec o P edica e In e sion is undone la e in he de i a ion ia aising o he
subjec o he Speci ie o a highe unc ional p ojec ion (in a ein o a
Ca og aphic app oach) whose head is emp y in he base, bu ge s illed by a
linke as i aises. This is he mos plausible accoun in a adi ional
pe spec i e.
90 To explain how a demons a i e p onoun (shi) de eloped in o a copula, Li & Thompson
(1977: 420) sugges ha his change came om a opic mechanism: “ he subjec p onoun
which is co- e e en ial wi h he opic in he commen o a opic-commen cons uc ion is
eanalyzed as a copula mo pheme in a subjec p edica e cons uc ion. As Li and Thompson
(1977) ha e poin ed ou , he copula o Manda in shi, o igina ed om he demons a i e
p ononun, which was eanalyzed as a copula when he opic g amma icalized in o a subjec ,
he opic-commen cons uc ion hus becoming a subjec -p edica e con uc ion (c . also
Bake , 2003)
144
could ha e been - gi en an “edge s a us” o some i ems - be ween e ex-i ems
and edge-i ems (in ui i ely de e mine s, e bs and complemen ize s would be
edges; e e en ial i ems e exes), a oiding e en he need o a se o ela ions
( he ela ions would be subsumed in he Lexicon in his way). I hink ha he
second op ion is i a ional, and no language (na u al o a i icial) can be
compu ed in ha way.
On he con a y, he i s possibili y is somewha s imula ing and I will y
o make an a icula ed discussion o he consequences o a p oposal o his
kind in he ollowing chap e .

145
5. Some heo e ical issues and implica ions
In his chap e I will lay some possible heo e ical ounda ions, based upon
he ep esen a ional in es iga ions made in he hi d chap e o his wo k,
and assumed o o he analysis o Eza e in he ou h chap e .
5.1. On he possible “Edgeness” o unc ional i ems
As I ha e pos ula ed in he las pa ag aph o he p e ious chap e - mainly
based on he syn ac ic analysis o Pe sian - some i ems o he lexicon may be
selec ed as edges ins ead o e exes. The p esence o he Eza e mo pheme as a
linke in Pe sian noun ph ases has been ou i s empi ical example
(in e p e ed in he ein o Den Dikken and Singhap eecha (2004)). This
obse a ion leads o a s ong simila i y wi h he ep esen a ion o chemical
s uc u es (going a beyond he me apho o Bake (2001)), as we will see
(lying) pe asi ely in he p esen sec ion.
Howe e , he e y in e es ing ac is ha his kind o obse a ion could be
assumed as he g ound o an empi ical (weak) gene aliza ion ha has se e al
consequences o he s udy o syn ax in a g aph heo e ic pe spec i e:
The only i ems ha may be s uc u ally (no p agma ically) dele ed in syn ax a e
unc ional i ems (in example complemen ize s, p eposi ions o pos posi ions,
de e mine s, copulas, linke s). Func ional i ems a e ep esen ed as edges. Thus, while
a e ex (a lexical i em) has necessa ily a phonological con en , an edge may be
p onounced o no .
I assume ha , in a g aph heo e ic pe spec i e, his is one o he
undamen al obse a ions conce ning he empi ical pa ame ic di e ences
ac oss na u al languages. Many examples will ollow, wi h he aim o
con ex ualize his simple, ye hope ully no i ial deduc ion.
An unconscious igge o his issue has been ep esen ed by he powe ul
Lewis o mulas (de ini ely, a so o di ec ed g aphs), commonly used o he
schema a o chemical elemen s and molecules93. As we will see in dep h as a
me apho , bu as we may in ui i ely in oduce he e, lexical i ems (such as
nouns, e bs o adjec i es) may be seen as p o ons inside a nucleus, while
unc ional i ems as elec ons ha exis ou side o he nucleus in a eas o high
p obabili y (bu wi hou spa ial subs ance) called o bi s, o shells. Again,
93 No ice ha he disco e o he deeply impo an chemical phenomenon o isome ism has
been an icipa ed by he g aphical no a ions o C um B own in he 1850s.
146
( uzzy) opology plays a c ucial ole.
Ou hypo hesis ep esen s an economical way o explain some o he
pa ame ic (o , say, ypological) di e ences o many languages (in example
he absence ( s. p esence) o de e mine s (as in he Sla ic languages o
Chinese Manda in) o he absence ( s. p esence) o copula (as in Russian o
Chinese Manda in).
5.1.1. On he di e en s a us o lexical s. unc ional i ems
In some way, a p elimina y heo e ical X’—Theo y g ound o my
assump ions could be ound in he seminal wo k o Fukui and Speas (1986).
They sugges ed ha he p ojec ions o lexical heads L° a e ecu si ely i e able
L’s94, while he p ojec ions o unc ional heads F°s a e non i e able F’ ( hese
p ojec ions a e d i en in example by he discha ge o F°s unique
subca ego iza ion ea u e on o he complemen s o F°s) and closed F’’ (being
his p ojec ions d i en by he discha ge o a unique ag eemen ea u e on o a
maximal p ojec ion ha mo es in o he (necessa ily) unique Spec posi ion).
The e o e, gi en he ela i ized X-Theo y o Fukui and Speas (1986) any
unc ional head has a lonely complemen (sis e o F° and domina ed by F’))
and (a mos ) one speci ie (sis e o F’ and domina ed by F’’), who ag ees
wi h F° and closes o he F’’ p ojec ion.
5.1.2. The on ology o unc ional i ems
Func ional i ems a e necessa ily non-p oduc i e closed classes. Taking
hei inne mos na u e as an issue o he discussion, i is possible o say ha
hey lack a desc ip i e con en . As a gued in he g oundb eaking wo k o
Abney (1987):
“Thei seman ic con ibu ion is second o de , egula ing o con ibu ing o he
in e p e a ion o hei complemen . They ma k g amma ical o ela ional ea u es,
a he han picking ou a class o objec s”. (Abney 1987: 65).
I hope his poin could be use ul o my p oposal. Func ional i ems a e
edges in a g aph based syn ac ic model jus because hey a e ela ional i ems,
wi hou a speci ic weigh (o , a some le el, a e e en ial na u e)95.
94 In Fukui ans Speas (1986) ein e p e a ion o X’-Theo y hese p ojec ions a e d i en, o
example, by he discha ge o he a ea u es o subca ego iza ion ea u es.
95 I hink ha ca ego ies a e no uni a y no ions, bu eme ge a he in e ace o di e en
147
Take Pe sian Eza e as an example. As a unc ional mo pheme in wha is
canonically assumed o be he nominal “domain”, such an i em is use ul o
speci y a e e ence. In mo e gene al e ms, a noun p o ides a p edica e, while a
de e mine “picks ou a pa icula membe ’s o he p edica e’s ex ension”
(Abney 1987: 76).
Fu he mo e on Fin el (1995) a gues ha unc ional mo phemes ha e a
logical- ela ional ( o example pe mu a ion-in a ian ) seman ics:
“Logicali y means insensi i y o speci ic ac s abou he wo ld, sugges a pu ely
ma hema ical ela ionship” ( on Fin el, 1995: 179)
5.2. Case: om Go e nmen and Binding o G aph Theo y
E en hough only p onouns show o e mo phological case in languages
such as I alian (i.e io s. me) o English (I s. me) - aking as a s a ing poin
he Go e nmen and Binding (GB) Pa adigm (c . Chomsky, 1981; Haegeman,
1996) - i is widely assumed ha all NPs ha e Case (called abs ac Case) ha
ma ches he mo phological Case ha shows up on p onouns. Appeal is made
o o he languages wi h much iche case sys ems han I alian o English o
“back up” his claim.
Re iewing he basis o GB Case Theo y we may obse e ha (in ph ase
s uc u e e ms):
a. Nomina i e Case is assigned o he NP speci ie o I[+ in] such
as in (5-1).
(5-1)
componen s o human cogni i e and communica i e capaci ies (c . Muysken, 2005).
Ca ego ies a e a mo e ambiguous han a commonsensical iew could assume. C oss-
linguis ically, he e is an in ini e se ies o examples conce ning ca ego ies unde -
speci ica ion, o e lapping phenomena (c . Com ie e al. 2005) and he e is also an Indonesian
language claimed o be (almos ) mono-ca ego ial (c . Gil, 2001). No e also ha in ecen yea s,
wi hin he gene a i e amewo k, Ma an z and o he esea che s coming om Dis ibu ed
mo phology pa adigm ake a s ong s and (in many, s ill unpublished pape s) and p opose
ha basically all lexical ca ego ies (nouns, e b, adjec i es) a e o med in a simila manne
om ca ego y-less oo s. This claim p edic s ha we should ind p oduc i e iple s (noun,
adjec i e, e b) o e e y ca ego y-less oo (c . Ma an z, 2001).
148
b. Accusa i e case is assigned o he NP sis e o V o P. The C[ o ]
which is homophonous wi h he p eposi ion o ac s like P o Case
assignmen . No e ha he subjec o a non- ini e clause could no ecei e Case
om I[- in] since only I[+ in] assigns Nomina i e Case.
(5-2)
c. Geni i e Case, inally, is simply assigned o he speci ie o N.
Wha is he same abou hese posi ions ha ecei e Case and he posi ions
ha assign Case? Chomsky obse ed ha e e y maximal p ojec ion (=XP) ha
domina es he NP ha ecei es Case also domina es he head ha assigns i
(a leas , i we do no coun he IP ha in e enes be ween he C[ o ] and he
NP).
A ( ough, a oiding he no ion o c-command) de ini ion o go e nmen
comes om his obse a ion:
a GOVERNS b i
149
a. a is a head [±N,±V] o I[+ in] o C[ o ], and
b. e e y XP ha domina es a also domina es b, and
c. e e y XP (o he han IP) ha domina es b also domina es a.
In his de ini ion, a and b s and o pa icula ca ego ies. P oposi ion (a)
equi es ha a is one o he heads N, V, A, P, I[+ in] o C[ o ]. Almos always,
b is an NP, since a e nouns ha need Case (which is assigned unde he
go e nmen ela ion). P oposi ion (b) de e mines how high up he ee a head
may go e n: i e e y maximal p ojec ion abo e he head mus also domina e
he NP in ques ion, hen he NP mus be below he maximal p ojec ion o he
head (e.g. VP o V, IP o I[+ in]). P oposi ion (c), inally, p o ides he lowe
limi o go e nmen by no allowing he head o go e n down in o ano he
maximal p ojec ion o he han IP.
Toge he , P oposi ion (b) and (c) es ablish locali y cons ain s on he
go e nmen ela ion o each head (c . Rizzi, 1990; Manzini, 1992).
The Case assignmen ules in e ms o go e nmen a e qui e simply ad
elegan . Re e ing o he s uc u es ep esen ed abo e we may say ha :
a. I[+ in] assigns nomina i e case o he NP speci ie ha i go e ns.
b. N assigns geni i e case o he NP speci ie ha i go e ns.
c. V, P, C[ o ] assign accusa i e case o he NP ha hey go e n.
GB equi es ha all NPs mus ha e Case a S-s uc u e by he Case Fil e
(C . Haegeman, 1996).
CASE FILTER: *NP i i does no ha e Case a S-s uc u e.
Wi h one u he assump ion, we will ha e he mo i a ion o A-
mo emen in passi e, unaccusa i e, and aising cons uc ions (C .
Zampa elli, 1995-2000). The amous Bu zio’s Gene aliza ion (Bu zio, 1986)
s a es ha p edica es which do no assign a seman ic ole o hei ex e nal
a gumen canno assign Case o hei complemen (s). This p o ides he
answe o why he passi e objec mus (c oss-linguis ically, can) mo e.
In GB, passi e e bs do no assign a seman ic ole o hei ex e nal
a gumen posi ion, so hey ha e los he abili y o assign Case o hei
complemen .
The e o e, he NP objec canno emain in place a S-s uc u e and mus
mo e o a posi ion whe e i can ge Case: he speci ie o I[+ in] whe e
nomina i e case is assigned. The same obse a ion accoun s o he

150
unaccusa i e and aising cons uc ions: he NP which canno ecei e Case in
i s (Deep s uc u e) posi ion is he one which mus mo e; and i may only
mo e o a posi ion which does assign Case, he speci ie o I[+ in].
Now, i is in e es ing o no ice ha , con a y o GB assump ions, mo e
ecen esea ches (Bi ne and Hale, 1996; Chomsky, 1998) assume s uc u al
Case as a syn ac ic ea u e ha is incapable o inducing mo emen . Thus, Case
ge s checked (e ased) as a esul o s uc u al ac o s ha exis independen ly
o Case i sel . In he Minimalis P og am, Chomsky a gues ha Case-
checking is “ancilla y” o o he ea u e-checking mechanisms.
I assume ha , c oss-linguis ically, Case mo phemes, e ealing a ela ion
be ween i ems in wha we may con inue o call a clause (o in a ph ase: see
Case ma ked adjec i es), a e unc ional elemen s ha may be ep esen ed as edges.
A i ial p oo is ha na u al languages make ex ensi e use o p eposi ions
(o he a ailable unc ional elemen s) whe e Case mo phology is no a gi en
syn ac ic possibili y.
Thus, mo phological Case on nominals is a common de ice o exp ess he
syn ac ic (and seman ic) ela ionships be ween clausal cons i uen s. Howe e ,
i is impo an o ema k ha he languages o he wo ld ha use his s a egy
a y g ea ly wi h espec o he numbe o Case ca ego ies ep esen ed in
hei in lec ional sys em.96
A p oblem, conce ning he GB pa adigm, is ha i we assume a ine
s uc u e o he in lec ional ph ase (c . Pollock 1989; Belle i, 1990), and
acco ding o Chomsky and Belle i we e ine he obse a ion in (5-1) saying
ha “NP is nomina i e i go e ned by Ag ”, many c oss-linguis ical da a,
such as Japanese and Icelandic qui ky subjec s (U a, 2000; Sigu dsson, 2002) o
spli -e ga i i y pa e ns (i.e. ag eemen s. Case), such he ones in Geo gian
ao is (c . Ha is, 1981; Ma an z, 1984; Com ie, 1989) aise objec ions. See he
examples below:
96 The minimal Case pa adigm con ains wo membe s, since pa adigma ic ela ionships
be ween wo d o ms a e ul ima ely based on bina y opposi ions (minimal pai s). This implies
ha whene e a language has an o e ly ma ked Case ca ego y exp essing a speci ic
unc ion, a co esponding ze o-ma ked base o m is coun ed as a Case ( he “de aul case” o
“di ec case”) e en i i has non-speci ic unc ion desc ibable in posi i e e ms. In such
ins ances, he base o m ecei es i s case s a us only by i ue o con as ing wi h a
unc ionally (and o mally) ma ked Case ca ego y. An example, gi en in he Wo ld A las o
Language S uc u es (edi ed by Com ie e al. 2005), is Mapudungun (A aucanian; Chile),
which has only one o e case su ix -mew ~ -mu exp essing di e se oblique unc ions such as
place, cause, and ins umen . On he o he e ex o he con inuum he e a e languages which
show e y la ge pa adigms. The languages wi h he la ges pa adigms, again acco ding o
he Wo ld A las o Language S uc u es, a e Hunga ian wi h (unde some analyses) 21
p oduc i e cases, ollowed by Kaya dild (Tangkic; Queensland, Aus alia) wi h 20, and Lak
(Nakh-Daghes anian; eas e n Caucasus, ex ensi ely s udies a Max Plank Ins i u e
Depa men o E olu iona y An h opology) wi h 19 cases. (Iggesen, 2005).
151
Japanese
(5-3) a. Ta oo-ni hebi-ga kowa-i (U a, 2000)
Ta oo-Da snake-Nom ea ul-P es
“Ta oo is ea ul o snakes”
b. Ta oo-ni eigo-ga deki -u
Ta oo-Da English-Nom unde s and-P es
“Ta oo unde s ands English”
As we may obse e easily, da i e NP in he examples abo e in (5-3a,b)
beha e as subjec s (in example, hey can bind subjec o ien ed anapho s).
A simila qui kyness is sha ed by Icelandic:
Icelandic
(5-4) Mé a hjàlpaò (Sigu dsson, 2002)
Me-da was helped
“I was helped”
(5-5) a. Aku leiddis
Aki-da bo ed
“Aki was bo ed”
b. Aku i is ha a leiddis
Aki-da seem o-ha e bo ed
“Aki seemed o ha e been bo ed”
c. Viò öldum Aku ha a leiddis
We belie e Aki-da o-ha e bo ed
“We belie e Aki o ha e been bo ed”
Icelandic qui ky Subjec s, summa izing a su ey made up by Sigu dsson
(2002), ha e been es ed o e lexi isa ion, ECM in ini i es, Raising, Con ol
e c. Some examples ha e been gi en in (5-5).
Fu he mo e, a so-called spli -e ga i i y due o he ense/aspec speci ica ion
o he clause is ound somewhe e. Basic ac s conce ning he ela ion be ween
Case and ag eemen a e illus a ed in he ollowing examples. Geo gian is
152
well-known o i s spli -e ga i i y o his kind97. In Geo gian, o example, he
ao is ense sys em demands e ga i i y and he p esen ense sys em
demands accusa i i y (da a a e om Com ie, 1978 and Lyle, 1997, ci ed in
U a, 2005; c . also Ha is, 1981).
Geo gian
(5-6)
a. PRESENT
SUBJ in ansi i e: Nomina i e + subjec ag eemen
SUBJ in une ga i e: Nomina i e + subjec ag eemen
SUBJ in unaccusa i e: Nomina i e + subjec ag eemen
OBJ in ansi i e: Accusa i e + objec ag eemen
b. AORIST
SUBJ in ansi i e: E ga i e + subjec ag eemen
SUBJ in une ga i e: E ga i e + subjec ag eemen
SUBJ in unaccusa i e: Absolu i e + subjec ag eemen
OBJ in ansi i e: Absolu i e + objec ag eemen
Geogian
(5-7)
a. PRESENT
a’. S uden -i midis. In ansi i e
s uden -NOM go(PRES)
‘The s uden goes.’
a’’. S uden -i ce il-s ce s T ansi i e
s uden -NOM le e -ACC w i e(PRES)
‘The s uden w i es he le e .’
97 Besides Geo gian, Hindi and many o he Indo-A yan languages, Bu ushaski, Tibe an,
Nepali, Samoan, e c. show simila spli -e ga i i y due o he ense/aspec speci ica ion o he
clause (see Com ie 1978; Dixon, 1994, and Palme , 1994 o a lis o such languages).
153
b. AORIST
b’. S uden -i mi ida. In ansi i e
s uden -ABS go(AOR)
‘The s uden wen .’
b’’. S uden -ma ce il-i dace a. T ansi i e
s uden -ERG le e -ABS w i e(AOR)
‘The s uden w o e he le e .’ ( om U a, 2005)
Al hough i is a well-known ac ha ag eemen in Geo gian is
ex ao dina ily puzzling and highly esis an o a sys ema ic explana ion, he
da a in (5-6) and (5-7) e eal he weakness o GB assump ions abou Case,
showing a disc epancy be ween abs ac Case and Case mo phology /
ag eemen .
In b ie , he se o da a p esen ed he e means ha Case mo phology is no
a sys ema ic e lex o abs ac Case.
We may a gue ha Case has no o be iewed unde licensing, o il e ing,
bu i is ins ead Case ha in e p e s syn ax and exp esses a ela ional ea u e
(a PF le el).
Anyway he ea u e alues a e la gely sel -explana o y. In languages lacking
mo phological case98, g amma ical ela ions a e exp essed by wo d o de
and/o mo phologically and p osodically independen unc ion wo ds
(gene ally, p eposi ions and pos posi ions), and pa ly also by mo phological
de ices on he e b. He e a e some c oss-linguis ic examples conce ning wha ,
ollowing and pa ially e ising Com ie (1989), we may call he Alignmen o
(S uc u al) Case Ma king:
i. No mo phological case
Thai
(5-8) phom kin khaw
I ea ice
“I ea ice”
98 Remembe ha in hese languages, he de aul o m ecei es i s Case s a us by pai ing i
wi h unc ionally and o mally Case-ma ked ca ego ies.