Академический Документы
Профессиональный Документы
Культура Документы
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Springer is collaborating with JSTOR to digitize, preserve and extend access to Natural Language & Linguistic Theory.
http://www.jstor.org
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
JERROLD M. SADOCK
1. BACKGROUND
(1) A
C D
b c
* I wish to thank the Spencer Foundationand the University of Chicago for financial
support that enabled me to spend a year at the Center for Advanced Studies in the
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 381
2 David Perlmutternotes that this name is inappropriatein at least two ways. First, it
suggeststhat thisis a theoryof syntax,which it is not, andsecondly,it suggests that there is
somethinglexical - in the sense of listing - about the theory, which there isn't. While I
agree with both of these nomenclaturalcomplaints,the title of this paper is by now well
enough dispersedand has become so familiarto me that I will retainit, thus following the
venerablepracticein linguisticsof misnamingtheories.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
382 JERROLD M. SADOCK
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 383
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
384 JERROLD M. SADOCK
is just what a prose description would make it, an element with the
morphological character of an affix but the syntax of a complement-
taking verb. In an autolexicalmodel we may assign 's two quite separate
structuralpositions in the morphology and the syntax and thereby des-
cribe directly some of its properties without recourse to postsyntactic
rules of cliticization or the like. From a syntactic perspective, 's in
English figures in structureslike (4), but from a morphologicalpoint of
view, it occurs in structureslike (5).
(4) S
NP VP
N V ADV
(5) W
N CL
John s
3There is a grammaticalsentences 'S John coming but it is clear that this is the productof
a rapidspeech rule and not the cliticizationprocessthat I am talkingaboutwhichcan occur
in careful,slow styles of speech.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 385
N v ADV
John here
John s here
N ~~~V ADV.
NP VP
S
Note that by separatingthe morphologicaland syntacticdescriptionsof a
sentence we not only make it unnecessary to assume a special cliti-
cizationcomponent(Zwicky 1982), we also obviate the need to separate
inflectional morphology from derivational morphology, as Anderson
(1982) does in assigning inflectionalmorphologyto the syntax. Suppose'
that purely inflectionalmaterialis representedin the morphologyalone,
where it arguably belongs (Lapointe 1980, 1983). Separating syntactic
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
386 JERROLD M. SADOCK
(7) V
N V Af DET N
DET N
V ~~NP
NP ~~~VP
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 387
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
388 JERROLD M. SADOCK
(8) V
V INF
NP N V N
V NP
NP VP
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 389
(9) S
VP
V NP
V INF
NP N V N
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
390 JERROLD M. SADOCK
(10) g
B C D
, 2 ,- D
D- C B A
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 391
(11) W
N DET A N V
N A N
DET N
V x NP
NP VP
What is wrong with (11) can not be localized in either the syntax or
morphology,but rather must be sought in the manner in which the two
structuresare associated. Therefore in what follows I will seek to tighten
the frameworkin both substantive and formal ways. I will be especially
concerned with seeking general correlations that govern the possible
relationshipsbetween syntactic and morphologicalrepresentationsand to
state these as universalpfinciples of autolexicalsyntax.These rules of the
interface have much the same status in the present context that Gold-
smith's (1976) association principles have in autosegmental phonology.
The interactionof these principleswill account for the ungrammaticality
of *John a good linguist's without requiringany further language parti-
cular statementeither in the grammarof English or in its lexicon.
The data with which I am mainly concerned are the facts of noun
incorporationin West GreenlandicEskimo and SouthernTiwa, a Tanoan
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
392 JERROLD M. SADOCK
3. WEST GREENLANDIC
b. Kaali pisuppoq.
Karl(ABS) walk-INDIC/3s
Karl is walking.
In the verb phrase we find one or more subcategorizednominalswhich
6
I use the following abbreviationsin morphemeglosses: ABS = absolutivecase, ERG
ergative case, ALLAT = allative case, INST = instrumentalcase, PERL= perlative case,
EQUAL = equalis case, and INDIC indicative mood. A slash separating two such
abbreviationsindicates that there is one morphemethat representsboth categories. For
transitiveverb inflections,the feature complex preceding the slash is taken as referencing
the ergative argument, and the one following the slash is taken as referencing the
absolutiveargument.
I use upper case letters, sometimesprecededby '+', in the text in talking aboutfeatures
representingthese categories.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 393
can be in any case but the ergative, and the verb itself. The order of
these elements is once again quite free.
(13) Nuummut ingerlavoq.
Nuuk-ALLAT go-INDICI3s
He is going to Nuuk.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
394 JERROLD M. SADOCK
7 These rules are intended to conform to the principlesoutlined in Gazdar and Pullum
(1982) and Gazdaret al. (1985), though the notation is different.In particular,the linear
precedence rules are to be interpretedas implyinga set of phrase structurerules derived
from a set of immediatedominanceprinciples.Thus the ordering statementsnecessarily
refer only to sisters generatedby a single phrasestructurerule. For present purposesit is
importantto note that (G2c) will constrainthe orderingof a verb of being or namingand its
absolutivecomplementintroducedby one varietyof (Glc), but will not constrainthe order
of an absolutivesubjector object with respect to a verb, since these are introducedby Gla
or Glb and are not sisters of the verb. This produces exactly the right result for West
Greenlandic.(In these rules, superscriptsare used, in the standardmanner,to indicate bar
level. So NOrepresentslexical N, N' is the notationused for N, N2 is N and so on.)
8 I will assume in section 4.3 below that these agreement rules are the Greenlandic
instantiationsof a grammaticaluniversal that falls under some version of the Control
Agreement Principleof Gazdarand Pullum (1982) (cf. (I) below), and thus need not be
stated in the particulargrammarof this language at all. They are includedat this point in
the text for the sake of clarity.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 395
3.2. A SimpleMorphology
Though very productive and very rich, the basics of West Greenlandic
morphology are also quite simple. A word must consist of stem and
inflection, in that order. In the vast majority of cases, a stem is either
basic, or consists of another stem followed by a derivational affix, the
class of the resulting stem being in general a function of lexical proper-
ties of the affix.For the most part, the semanticeffect of such affixationis
quite predictable,the affixhaving semantic scope over the stem that it is
attached to. (But see Fortescue 1980 for a discussion of exceptions to
this principle.)
(23) illorsuaqarfimmut
((((illu) rsuaq)qar) fik) -mut
house big have place ALLAT
to the place with big houses
(24) illogarfissuarmut
((((illu) gar) fik) rsuaq)-mut
house have place big ALLAT
to the big town
A few derivational affixes can productively derive a stem from an
inflected word. This stem requiresits own inflectionto be complete. The
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
396 JERROLD M. SADOCK
(28) Ajorporaasiinngooq.
ajor-poq -aasiit -gooq
bad INDIC/3s as-usual it-is-said
They say that as usual, it's bad.
These rudimentaryprinciplesof Greenlandicmorphologyare formalized
in (G4)-(G8), where (G4) is a general rule for forming a new stem from
another stem, (G5) forms stems from inflected words, (G6).is a rule for
inflectinga stem, and (G7) forms a word of unspecifiedgrammaticalclass
by attachinga clitic to a word of any class. The firstfour rules specify the
hierarchical structure of words. The only linear ordering principle
needed in this entirely suffixinglanguage, is (G8), which places an affix9
after all its sisters.
In this microgrammar,I have adopted the convention'0 that syntax is
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 397
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
398 JERROLD M. SADOCK
input to (G4) since the bar level of the result of applying (G6) is not
suitable as Y in (G4), and the output of (G7) cannot be the input to any
of the rules (G4-G6) because of a similar bar-level discrepancy. Fur-
thermore, since the segmental phonology of all of the morphology of
West Greenlandicis highly concatenative, we automaticallyaccount for
the commonplace observation that derivational affixes generally occur
inside inflectional affixes, and inflectional affixes inside of clitic mor-
phemes.
Were it not for the existence of (G5), a kind of morphologicalrule that
is almost without parallel in more widely studied languages, it would be
the case that all inflectionalmorphemeswould have to occur outside all
derivationalmorphemes,and all clitics would have to occur outside all
inflectional morphemes.These results follow from the phrase-structure
frameworkof the morphology suggested above and do not require the
postulation of either separate levels of morphology or separate small-
scale modules of grammar.12
12 This system alone will not, however, account for the bracketingfacts concerning so
called Level I and Level II affixationthat have been so much discussedin the frameworkof
'lexical'phonology(Mohanan1982, Kiparsky1982).
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 399
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
400 JERROLD M. SADOCK
how they combine with the units at or below the X-zero level), as I
argued in Sadock (1983, 1984) that lexical items in general must. The
'strange' facts of noun incorporationthen fall out as the only possible
results for the case where the syntax and morphology of a lexical form
disagree.
Consider the case of the object-incorporatingaffix-qar 'to have '.
Morphologically,it has just the properties of an ordinary, derivational
suffixthat participatesin (G4) above. Unlike run-of-the-millderivational
morphemesin other languages,which do not have any syntacticfunction
at all, this particularmorphemeis subject to syntactic subcategorization
requirements,indeed the very same requirementsas those of an ordinary
independent, formally intransitive verb of West Greenlandic like
sanavoq 'to build'. As with any formally intransitive verb in this lan-
guage that is semanticallyassociatedwith a patient thematic role, it may
take an instrumental-case argument in the syntax, representing an
indefinite patient. Let us then take its lexical representationto include
specificationsalong the lines of (29). In the notation I have adoptedhere,
the first term of the bracketedexpressionfollowing the equal sign refers
to the rule that introducesthe lexeme, and the matrices that follow give
furtherfeature requirementsof the variablesin that rule, startingwith the
symbol to the left of the arrow,in the order in which they occur.
(29) qar: semantics= 'have'
morphology= [G4, [-N, +V], [+N, -V]]
syntax= [Glc, [-TRANS], [INST]]
For the simplest sort of sentence presenting an example of noun in-
corporation,such as Hansi illoqarpoq'Hans has a house', we would have
the two-partrepresentationin (30), where, as in section 2, the trees in the
upper part of the diagramrepresent the morphology of the expression,
and the tree in the lower part representsits syntax. The well-formedness
of this example consists in the fact that the morphologicaltree satisfies
the morphologicalrequirementsof every lexeme, and the syntactic tree
satisfies all of their syntactic requirements.In particular,the syntactic
frame of the crucial lexeme -qar is satisfied by the existence of an
instrumental N2 complement in the syntax, and the morphological
requirement that it be attached to a noun stem is satisfied in the
morphologicalrepresentation.'3
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 401
(30)
N-1 V INFL
N INFL N V
N42
[INST] V
N2 VI
v2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
402 JERROLD M. SADOCK
N
</MOD
N2
[INST] V
N2 V1
v2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 403
Several comments are in order at this point. For the moment I will
concentrate on the principles that govern feature distributionsin and
between the two diagrams,neglecting the very importantquestion of the
crossing association lines that are found in (34), a subject that I will
investigate in detail below.
I assume that the distributionof features of nodes in the syntacticpart
of this diagramconforms to the following well-establishedprinciples:
(I) The Head Feature Convention of Gazdaret al. (1985) applies
to the syntactic tree: the features of a phrasal node must be
the same as those of the head of that phrase.14
(II) The Control Agreement principle of Gazdar et al. (1985)
applies to the syntactic tree: roughly, functions take on the
agreement features of their arguments.In West Greenlandic,
in particular
a. The head agrees with its sisters in V2
b. The modifieragrees with the head in N1
c. The possessumagrees with the possessorin N2
(III) Lexical subcategorizationfeaturesof,a lexeme must be met in
the syntactic tree.
Thus the N2 complement of -qar must be instrumentalby (III) because
-qar lexically subcategorizes a syntactic complement with that feature;
the noun qamut must be instrumentalby (I) because it is the head of an
instrumentalN2, and ataaseq is instrumentalby (Ilb) because it modifies
(i.e., is a function of) an instrumental noun. Furthermore,qamut is
intrinsically(lexically)plural, so both the mother node N2, of which it is
the head, and its modifierataaseq must also be pluralaccordingto (I) and
(JIb), respectively. Finally, V1 must be third person singular by (11a)
(assumingVs to be functionsand subjects to be arguments),and V must
be third person singularby (I).
The second point is that I have assumedthat purelyinflectionalaffixes
are not mentioned in syntactic structureat all: they are entities that find
their properplace in the morphologyalone. The inflectionof a word may
well reflect features determinedin the syntax by principles (I)-(III), but
'4 The HFC as stated in Gazdaret al. (1985) does not requirestrict identitybetween the
featuresof the mothernode and the featuresof the head, but only identityinsofaras it is
not overridenby other feature specificationdevices. This detail is irrelevantas far as the
phenomenadealt with in this paperare concernedand has consequentlybeen suppressedin
the statementof (I).
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
404 JERROLD M. SADOCK
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 405.
In the simplest cases, such as-that of the proper noun Hansi in (34),
where there are no derivationalaffixes,the featuresof the root are spread
through the word to its inflection (in this case 0, the inflection of
unpossessed,third singular, absolutive nouns). But consider the case of
the word qamuteqarpoqin (34). The incorporatednominal qamut must
be instrumentalplural according to (IV). But by (I'), the features of the
derived stem qamuteqarreflect only the features of the head (here the
affix-qar), namely third person, singular,which have been passed to -qar
from the syntax by (IV). These are also the features of the inflection,
-poq, the third singularindicative suffix,as predicted by (II'd).
Let us now consider examples which reveal some importantfactors
concerningthe order of words and morphemes.
l5 We may actually distinguishtwo types of inflection, one where the features that are
spreadin accordancewith (II'd) are passed to the stem from the syntax (e.g., agreement
features on adjectives and verbs), and one where they reflect either intrinsic,or freely
instantiatedfeatures (e.g., conjugational,and declensional class features, and tense and
numberfeatures,respectively).The firstclass generallymeets Anderson's(1982) suggested
definitionof inflection, while the second class - as emphasizedrecently by Jensen and
Stong-Jensen(1984) does not. This latter class may well include semanticallymeaningful
expressionsand thus correspondto semanticfunctionsother than identity.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
406 JERROLD M. SADOCK
N-1 V INFL
N INFL N V
V N2[+ABS]
2 ,, V V1
v2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 407
(37)b. V-1
N-1 V INFL N-
N INFL N V N INFL
Hansi u
nunuippiaraq mikisoq
N MOD
N'
vt N2 [+ABS]
N2
V2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
408 JERROLD M. SADOCK
Consider again example (33), the tree for which was given in (34). Here
again there are crossing lines of association because -qar is mor-
phologically a suffix, a requirementfrom which the language allows no
deviation. The modifier ataatsinik can also occur after the verb in a
grammatical sentence: Hansi qamuteqarpoq ataatsinik. When the
modifier follows, there is once again one pair of crossing association
lines, this time the lines associating-qar and ataaseq is the two structures.
But there is no grammaticalexample related to (35b) with the modifier
preceding the verb: *Hansimikisoq nukappiaraavoq.16
This cannot be because of the rule requiringmodifiersto follow their
heads, as (33) shows, but can only be a function of (G2c), the rule
ordering absolutive complements after the verb that subcategorizes
them. If the modifier mikisoq were to precede the verb in the mor-
phological representation,then all of the lines associating nukappiaraq,
-u, and mikisoq would cross. The structurein (37b) is thus the closest to
one without crossing branches that is consistent with the morphological
and syntactic requirementsof the various lexemes in it.
Finally,consider the case of a possessorstrandedby incorporation,17 as
in (38), whose autolexical structureis (39).
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 409
(39) V-1
N-1 V INEL
N INFL N v
N2[ERG] N
N2 V
V1
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
410 JERROLD M. SADOCK
18 See Allen, Gardiner,and Frantz(1984) and Sadock (to appeara) for details.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 411
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
412 JERROLD M. SASDOCK
b. Wisi u- k'uru-k'euwe-m20
two PREF dipperold PRESENT
The two dippersare old.
(49)a.
v-I
DET N
V2l
N2 V
N2 V
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 413
b.
DET N
N2 V
V2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
414 JERROLD M. SADOCK
(T4) V2 N2, V1
(T5) V'IN2,V
(T6) VI-V
(50) k'ar: semantics='eat'
syntax = [T5]
(51) k'euwe: semantics='be old'
syntax = [T6]
Rule (T3) can then convert the strictlysyntactic lexical specificationsof
SouthernTiwa verbs into 'mixed', incorporatingspecifications (52) and
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 415
(53), which are quite similar to the basic, lexical specifications of the
incorporatingaffixes of West Greenlandic. Since (T3) is a rule, rather
than a feature default or cooccurrence restriction, this conversion is
optional.
(52) k'ar: semantics='eat'
syntax = [T5]
morphology= [T2, [-N, +V], [+N, -V]]
(53) k'euwe: semantics='be old'
syntax = [T6]
morphology= [T2, [-N, +V], [+N, -V]]
It is these derived lexical forms that figure in the diagrams (49a) and
(49b). These structuressatisfy both the morphologicaland the syntactic
requirementsof the derived, incorporatingverb forms, and also meet the
requirementson goodness of fit between the morphologyand syntaxof a
naturallanguage expressionembodied in (IV)-(VII).
Note also that under this autolexicaltreatmentnothing special needs to
be said about the agreementfacts of SouthernTiwa. As Allen, Gardiner,
and Frantzpoint out, the verb agrees with its subject and object whether
or not that element is incorporated.Since agreement is determined by
the syntax in the theory I am advocating, and since in this theory the
syntax of incorporated and unincorporatedexpressions is exactly the
same, then the agreement facts are accounted for without further
specification.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
416 JERROLD M. SADOCK
1
(55)
N INFL N INFL V V
N N ~ ~~~~~~~~~~~~~ I
N2N2 VI
v2 ~ ~ V
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 417
(56) (UpperSorbian)
zonina drasta
woman-ADJ-NOMIFEMIsg dress-NOMIFEMIsg
a woman'sdress
But remarkably,the noun from which the adjective is formed retains its
referentiality,and can be modified,possessed,or subsequentlyreferredto
with personalpronouns:
(57) (UpperSorbian)
stareje zonina
old-GENIFEMIsg woman-ADJ-NOMIFEM/sg
drasta
dress-NOMIFEMIsg
dress of an old woman
(58) (UpperSorbian)
mojeho bratrowe dzeci
my-GENIMASCIsg brother-ADJ-NOM/plchild-NOMIpl
my brother'schildren
(59) (LowerSorbian)
to su nasogo nanowe
thoseare our-GENIMASCIsg fatheri-ADJ-NOMIpl
crejeje, won jo je zabyl
shoes hei is them forgotten
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
418 JERROLD M. SADOCK
Al INF A2 N2 INF
N1
A [XFSG]]
[GEN]] A
2 N
A [F,SG]]
[NOM]
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 419
A-' A / N-1
A A\AA
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
420 JERROLD M. SADOCK
following three ways: (1) unlike all other tense markers,which occur with
the bare stem of the verb, -zoo- requiresan infinitiveform of the verb if
the verb-stembegins in a vowel; (2) the aspect markermust be -a, just as
it must in an independentinfinitive;and (3) if the verb-stem is defective
and lacks an infinitive, it also lacks a future form. Yet it is clear that
-zoo- is morphologicallypart of the verb, not just because it fills the
tense-morphemeslot in the canonical verb template,but also for various
phonological reasons. For one thing, no word of Kirundi can end in a
long vowel as -zoo- does..
These facts indicate that the syntaxof the marker-zoo- is just that of a
verb that takes an infinitivecomplement,while its morphologyis just that
of an ordinaryverb-internalmorpheme. Therefore Goldsmith suggests
an autolexical representationfor a word like bazookuandika 'they will
write' as in (63).
(63) V-1
ba zoo ku andik a
zoo andik a
V ASP
' '.~VI
V [TNS] 1NF]1
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 421
This case also falls under (VII), so that -zoo-, which has distinct syntax
and morphologyis requiredto combine with the head of its complement,
which it clearly does in this example and even more clearly does in
correspondingtransitivesentences.
The last example I wish to offer is a case of nominalization.In a careful
and insightful recent treatment of morphological and syntactic inter-
actions, Sugioka (1984) argues that several derivational suffixes of
Japanesetake phrasesas complementsin the syntax.The clearest exam-
ple is perhapsthe deverbal nominalizer-sa.
The syntax of noun phrases in Japanese is in general extremely
restricted,complementsof the head noun being confinedto genitive-case
N2s. But just in case -sa nominalizes a verb of strong desire, the full
syntax of the verb-phrase is preserved as the following two examples
from Sugioka (1984, p. 138) show.
Note that due to the existence of (I) (the HFC), (IV) (the principle
requiring identity of lexical categories in the syntax and morphology),
and (VII), (the principlethat only heads are incorporated)there is some
redundancy between these two specifications. Since the verb inSthe
morphologicalspecificationmust be the head ot the complement phrase
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
422 JERROLD M. SADOCK
6. NON-HEAD INCORPORATION
23
Other generalprinciplesof syntaxwill presumablypredict at least part of the bar level
facts involved here too, but I neglect this refinementhere.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 423
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
424 JERROLD M. SADOCK
A N A N
CONJ _
N2 N2
N2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 425
(74) W W W W
I
DEM '
I
OBJ DEM
\ ~~~DET N DET N
V pro V D N N2
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
426 JERROLD M. SADOCK
N INFL N N
N MOD
N2[Ioc] N
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 427
7. CONJOINABILITY
Principles (VII') and (VIII') are useful in accounting for facts concern-
ing the incorporationof conjuncts. This is possible to some extent in
SouthernTiwa, but not at all in West Greenlandicexcept for a sporad-
ic phenomenonthat appearsto be the polysyntheticanalogue of gapping
(Sadock to appear b). On the face of it, where (VII') applies, it ought to
bar the incorporationof conjuncts altogether,at least if we interpret'the
syntactic head' to mean 'the unique syntactic head', and if we assume
that conjoined phrases are multiply headed. It is therefore of no little
interestthat an incorporatedconjunct in SouthernTiwa behaves as if it is
the unique head of the noun-phraseobject.
The evidence for the incorporatedconjunct'sbeing the unique head in
SouthernTiwa is quite clear. The verb in SouthernTiwa must agree with
both its subject and object phrases, whether these are incorporatedor
not. When a conjoined object is external to the verb, it forces plural
(here type B) agreement on the verb, but when one of the conjuncts is
incorporated,the agreement is only with the incorporatedentity (here
singular,type A agreement).
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
428 JERROLD M. SADOCK
N1 CL V
N P
P2
N2 V
VI
24
This exampleoccurs in Allen, Gardiner,and Frantz(1984) as
(i) Tikantawibanba bakade'an.
Given the freedom of word order that Southern Tiwa displays, I believe (79) to be
grammaticalas well.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 429
N-'
N P N P
P2 P
p2 pV
p2V
VI
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
430 JERROLD M. SADOCK
(83)
V
V V
V1 CONJ V1
vi ~V
v V
However there is very persuasive evidence that (83) is not the correct
structurefor (82), and that in the appropriatestructure,tusar is in fact
the unique head of the VI that -umaar combines with. The evidence is
that the contemporative mood ('CONTEMP' in the gloss in (82)) is
otherwise a subordinate clause form. It requires that its subject be
coreferent with the subject of the clause to which it is subordinate.This
is a reflection of a general feature of West Greenlandicsyntax, namely
the fact that the straightforwardconjunction of verb phrases or clauses
with like subjects is generally avoided. What is found instead is that one
of the verb phrasesmust appearin the contemporativemood, a formally
subordinateform of the verb. Thus the correct structurefor (82) actually
includes somethinglike (84), which does conform to principle (VII').
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 431
(84) V-2
V INF V INF
V COMP V V
N2' 2'
S COMP
s / CON
VI V
VI V
vi
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
432 JERROLD M. SADOCK
with example (68) above, does not seem to attach to only one memberof
a conjunct (CarolynJenkinsp.c.):
(87) *Nuummullu Sisimiunukassaanga.
Nuuk -mut -lu Sisimiut-mut -kar -ssa -vunga
ALLAT and ALLAT go FUT INDIC/ls
I am going to Nuuk and to Sisimiut.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 433
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
434 JERROLD M. SADOCK
(89) X
V Af
N-' V V
N Af N V
N2 V
N2 V1
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 435
shown that the basic facts of one very common sort of cliticization in
natural language fall out as the unspecifiedtype of autolexical process,
and have suggested how aspects of derivational and inflectional mor-
phology in several very different languages might also receive a natural
and empiricallysuccessful treatmentin an autolexicalframework.
Thirdly,it seems to me that a grammarconsistingof a numberof fully
autonomousstructuralsubsystemsrelatedto each other by principlesthat
universally require a certain degree of conformitybetween them has a
great deal of psychological plausibility as compared with competing
models for dealing with transmodularphenomena. This plausibilitylies
partiallyin the great simplificationof the individualcomponents-that an
autolexicalsyntaxis capable of achieving, and partiallyin the naturalness
of the principlesof the interface grammarthat makes that simplification
possible.
In the case of West Greenlandic,the language that I have discussed
most carefullyhere, the normalizationthat resultsfrom the separationof
syntax and morphologyis striking. Not only are several special rules of
both syntax and morphology that would otherwise be required to deal
with noun incorporationentirely eliminated, but the language comes to
resemble more familiarlanguages in the details of these two systems to
an extent that is remarkable in view of the typological extremity of
Eskimo. The syntactic and semantic properties of lexemes of West
Greenlandic also turn out to be not very different from those of their
counterparts in English (or Southern Tiwa or Upper Sorbian). What
distinguishesWest Greenlandicfrom less 'exotic' languages is only the
relatively trivial fact that there are more lexemes in West Greenlandic
with independent syntax and morphology than we are accustomed to
finding.
Several of the principles that serve to limit the degree of mismatch
between syntactic and morphologicaltrees need to be in place anyway,
and are simply generalizationsto both componientsof the grammarof
principles suggested independently by syntacticians (I', II', and III').
Principle (IV) merely requires that lexemes be representedin the same
way undersyntacticor morphologicalanalysis,an assumptionthat figures
tacitly in every system of grammar.The remainingprinciples,(V), (VI),
(VII') and (VIII') are unique to autolexicalsyntax,but each has a ring of
empirical and psychological plausibility.It is not only true that mor-
pheme ordering is more rigid than phrasal ordering, it is also quite
reasonablethat this be the case since words are often learnedindividually
and hence characterizedby great internalfixity.Principle(VI) insuresthe
maximumdegree of conformity between the independentsyntactic and
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
436 JERROLD M. SADOCK
25In unpublished work, Steven Lapointe (ms.) has quite independently suggested a treat-
ment of French du that is strikingly similar to the general style of autolexical syntax.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 437
REFERENCES
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
438 JERROLD M. SADOCK
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions
AUTOLEXICAL SYNTAX 439
Departmentof Linguistics
Universityof Chicago
Chicago, IL 60637
U.S.A.
This content downloaded from 128.135.12.127 on Tue, 12 May 2015 03:21:39 UTC
All use subject to JSTOR Terms and Conditions