Professional Documents
Culture Documents
Neuvel 2011 No more phology! West Greenlandic evidence against a morphological tier of linguistic representation (Pragmatics and Autolexical Grammar)
Neuvel 2011 No more phology! West Greenlandic evidence against a morphological tier of linguistic representation (Pragmatics and Autolexical Grammar)
Neuvel 2011 No more phology! West Greenlandic evidence against a morphological tier of linguistic representation (Pragmatics and Autolexical Grammar)
Sylvain Neuvel
1. Introduction1
. I would like to express my gratitude to Jerry for expanding my mind, making me read
Greenlandic children’s books, and for the most intellectually challenging cigarette breaks in
recent history.
Sylvain Neuvel
2. West Greenlandic
–– The vowels /i/, /a/ and /u/ are respectively realized as [e], [#] and [o] before pho-
nemic /r/ or /q/ (/r/ is a uvular continuant and /q/ a uvular stop). ex: /qimmiq/
‘dog’ → [qimmeq].
–– Except for the sequence /ai/ which only occurs word finally, a vowel assimilates to a
preceding short /a/. ex: /qimmiarauvuq/ ‘he/she is a young dog’ → [qimmi#raavoq].
–– A consonant completely assimilates to a following consonant, except for the
sequence /ts/. ex: /sinikta/ ‘let us sleep’→ [sinitta].
–– All non-nasal consonant clusters are voiceless. ex: /sinikvikmi/ ‘in the bed’ →
[siniffimmi].
–– /t/ is affricated before a phonemic /i/. ex: /qamutit/ ‘sleds’ → [qamutsit].
3. Representations
NP V NP
[pl., instr.] [3sg.]
NP N
[pl., instr.]
The word qamuteqarpoq ‘he has sled(s)’ corresponds in the syntactic representation
above, not only to a V node, but to part of the object NP as well. In other words,
qamuteqarpoq is not a simple syntactic verb in this structure, but corresponds to a
complex non-constituent unit. Given the evidence just presented, I also assume that
qamuteqarpoq is not a syntactic atom and includes an NP, but I give the sentence
ataatsinik qamuteqarpoq a slightly different syntactic structure shown below.
(3) S
V NP
[3sg.]
NP V
N NP
[pl., instr.]
The differences between the two structures are minimal, but the representation I have
chosen makes this sentence syntactically intransitive. The subject NP in this sentence
appears in the absolutive case (Hansi(*p) ataatsinik qamuteqarpoq ‘Hans.abs (*erg)
has one sled’; and the ending of the verb is that of an intransitive verb. Moreover,
before it is combined with other words to form a sentence, the verb qamuteqarpoq
Sylvain Neuvel
3.2 E
vidence in favor of one or many semantic representations
at the word level
3.2.1 Function/Argument structure
Function/Argument Structure: F(unction)/A(rgument) structure is a semantic represen-
tation that reflects the logical configuration of a given construction. F/A structure
is articulated around the basic categories of argument and proposition, roughly
corresponding to entity expressions and truth-valued expressions. These two basic
types serve as the building blocks for other recursively defined categories, as per the
axiom in (4) (Sadock in preparation).
(4) a. There are two basic categories: Arg(ument) and Prop(osition)
b. If A and B are categories, then there is a category C such that C(A) = B
The categories defined in this manner are of two general types, functors and
modifiers. Functors are categories that combine with arguments and/or propositions
one by one to produce, in the end, a proposition. Following Sadock (in preparation),
I symbolize functors as Fψ, where ψ is a string of a(rguments) or p(ropositions) with
which the functor must combine (in the given order) to yield a proposition. A transi-
tive verb like kiss, as in Moya kisses Maggie, would belong to the category Faa since it
first combines with one argument (corresponding to the object) to form a predicate of
category Fa, which takes another argument to form a proposition. A verb like believe,
as in Maggie believes that Leeloo is a cat, is of category Fpa; it first combines with a
proposition to form a one place predicate (Fa). The F/A representations of both sen-
tences are given below. Note that these trees represent only logical hierarchy and that
the direction of branching is irrelevant.
Arg Fa Arg Fa
moya maggie
Faa Arg Prop Fpa
kiss maggie believe
Arg Fa
leeloo cat
The other type of semantic category that lexical items can belong to is modifiers (Mx).
A modifier is a function from a given category (x) to a more complex function of the
same category. A lexical item of category Mfa is thus a predicate modifier (slowly, for
example), one of category Mp, a propositional modifier, etc.
No more phology!
For our purposes, it is sufficient to point out that the verb in (1) is a predicate that has
two semantic arguments, the meaning ‘sled’, and a 3rd person argument, and that the
word qamut ‘sled’ has a simpler meaning and a simpler role structure. Any description
of the relation between these two words will therefore involve knowledge which, in
Autolexical grammar, is described in F/A structure and in role structure.
4. Morphological representations in WG
Arg Fa Arg Fa
3sg. 3sg.
Fpa Prop ModFa Fa
cause habit.
Fa Arg Fpa Prop
eat 3pl. cause
Fa Arg
eat 3pl.
b. [sanioqquttoq] [sanioqqutt#ttoq]
/saniuqquttuq/ /saniuqquttaqtuq/
‘one who goes by’ ‘one who habitually goes by’
Fa Fa
GO_BY Modfa Fa
habit.
GO_BY
Any description of the morphological phenomenon involved above will have to account
for the difference in Function/Argument structure described above, either relationally,
or compositionally.
No more phology!
When it comes to Role structure, the relation between the words in (7a) does
not involve a difference in the number of roles associated with each meaning. The
truth conditional meanings of the words involved simply differ in that one includes
the habitual ‘sense’ and the other does not. Based on the words in (7) alone, one could
represent the Role Structure correspondence between these words as (8) (where υ
represents the list of roles associated with the meaning θ):
This formulation is insufficient. Since the habitual can occur more ‘deeply’ into the
semantic structure of a word and modify a subordinate event, it will not in all cases
simply affect the entire truth conditional meaning of a word in the manner suggested.
In the next examples, the habitual modifies not the entire meaning of the word, but
only the act of winning.
(9) a. [ajugaanir#ppaa]
/ajugaaniraqpaa/
‘he/shei said that he/shej wins’
b. [ajugaas#nnir#ppaa]
/ajugaasaqniraqpaa/
‘he/shei said that he/shej often wins’ (Bittner 1995)
The RS correspondence must account for the fact that the habitual event may be
included in a larger meaning. In a relational system, we can represent the modified
meaning (θ) as included in a set (ψ) (possibly equal to θ), as in (10) below.
The words in 7 also differ phonologically in that the habitual forms include the
sequence of phonemes /taq/. Again, any account of the habitual in WG will have to
include that difference in form.
In a truly word-based theory of morphology where the word is the smallest
unit of interest, the relation between the words in (7) can simply be represented as
a bidirectional relation describing the contrast between every word pair involved in
this relation. I refer the reader to Ford & Singh 1991, 1996; Ford et al. 1997; Neuvel &
Singh 2002 and Neuvel 2003, in press, for a description of Whole Word Morphology, a
genuine example of such a word-based theory. In Whole Word Morphology, the rela-
tion between verbs and their habitual form would, using an autolexical representation,
be represented as (11).
(11) MSyn: α V αV
F/A: ϕ[Fa ρ] ↔ ϕ[Fa [MOD habit.] [Fa ρ]]
RS: ‘θ’ ∈ ψ, υ ~‘habitually θ’ ∈ ψ, υ
Form: /XC CY/ /XC taq CY/
Sylvain Neuvel
The lexical correspondence in (11) describes the contrast between verbs and their
habitual form on every level. Variables on every level can be filled with information
from any word that fits on one of the two lexical patterns.3 In a theory such as WWM,
anything that can be said morphologically of the relation between two words has to be
described in terms of contrasts between representations. The very concept of a mor-
phological level of representation is therefore illogical, since morphology is simply a
contrast, the expression of a productive difference on every level of representation.
On the other hand, one could ascribe to a more traditional theory of morphology
where words are formed compositionally using morphemes. In such a case, the first
thought that comes to mind is to give morphemes their own level or representation.
Under a compositional view, the difference between the words in (7) would be
accounted for by the addition, to the verb, of the habitual morpheme -taq-. The caus-
ative in (7a) would come from the morpheme -sit-, the eating part from the morpheme
niri-, and the third person with the morpheme -pai.
If we add a morphological level to the representation of the words in (7), the word
nerisippai could be represented as something like (12).
(12) [nerisippai]
Syn: V NP
[3sg.]
NP V
[3pl.]
F/A:
Prop
Arg Fa
3sg.
Fpa Prop
cause
Fa Arg
eat
3pl.
RS: ‘cause to eat’, Ag, theme
Morph: niri + sit + pai
(whether concatenated or in a tree structure)
Form: /nirisitpai/
While this representation may seem “normal” at first glance, it is interesting to ask
what exactly we’ve just added on this level. The “niri”, on the morphological level, is
. See Neuvel 2003 for a description of how the variables X and Y on the Form level can be
filled and resolved.
No more phology!
not the sequence of phoneme /niri/ since that is part of the formal representation, it
is not a syntactic verb, since that is part of the syntactic representation, and it has no
meaning, since that information is represented on the remaining levels of represen-
tation. In fact, the combination of the morphemes niri-, -sit-, and -pai, is the sum
total of the representation of the word. In terms of information, all we are adding in
the morphological representation is the fact that there are three morphemes, so the
morphological representation could very well be written as M + M + M. For this to
be remotely interesting, we need to know which part of the syntax, the meaning and
the form belong to each morpheme. Let us then look at the representation they would
receive in an Autolexical account.
Based only on the information we’ve used so far, the representation of these
morphemes would have to include something like (13).
(13) a. niri-
Syn: V
F/A: Prop
Fa Arg
eat
RS: ‘eat’, Ag, Th
Form: /niri/
b. -sit-
Syn: [V [V]]
F/A:
Prop
Arg Fa
Fpa Prop
cause
RS: ‘cause’, Ag, Event
Form: /sit/
under that view, exhaustively composed of these smaller units that each have their own
representation on the syntactic, semantic and formal levels of representation. In other
words, if we were to build a lexicon of WG, it would, for all intents and purposes, only
contain units of this type, making their identification as such fairly pointless.
In fact, besides the fact that such an analysis will run into all the well documented
problems faced by an Item-and-Arrangement approach to morphology, it also ironi-
cally makes most of what we conceive of as grammar completely redundant since
100% of the syntax and semantics of WG will be contained in the combinatorics that
must be specified for each of these mini lexical entries.
The very point of having a multi-level theory of grammar is that there can be
mismatches between levels. If we assume the existence of morphemes as the smallest
lexical entries, mismatches will only occur within the representations of morphemes.
Every word is exhaustively composed of morphemes on every level of representation,
and as such, the correspondence between levels beyond the morphemic level is by
definition 1 to 1.
The idea each level being the product of its own generative device also becomes
fairly difficult to maintain since, in the end, it must generate in parallel precisely what
is described in the lexical entries of morphemes.
5. Conclusion
I would suggest that there is an essential difference between morphology and the other
components of grammar. While each component corresponds to actual properties of
language utterances, morphology is simple the expression of certain regularities in the
properties of words on EVERY level of representation at once. In other words, mor-
phology, whether seen in relational or compositional terms, is not a level of represen-
tation but a set of generalizations about all the formal aspects of any given language.
Furthermore, the very existence of the morpheme, when pushed to its natural
conclusion, seems to deny the properties that make Autolexical grammar what it is.
One could conclude that in order to exist, Autolexical grammar has to evacuate the
concept of the morpheme and rely on a relational word-based theory of morphology
to account for the regularities between lexical units in any language.
The arguments made here naturally lead to the discussion of other linguistic
phenomena, like clitics, whose analysis often involves some sort of morphological
representation.4 While a detailed analysis of clitics is far beyond the scope of this paper,
. In his review of this paper, Anthony Woodbury rightfully points out that I have avoided
what is traditionally described as incorporation in this paper (Woodbury 1996). Much like
No more phology!
the logic remains essentially the same. In a theory of grammar in which each level of
representation is independent, there simply cannot be any evidence of a particular
structure or of anything else on a morphological tier: form is evidence of what goes
on the formal level, anything involving scope is either syntactic or semantic or we
wouldn’t know it’s there. A morphological representation is, however, a great place to
hide exceptions. But if the content of this representation is useless or entirely redun-
dant, except when it “explains” an exception to a syntactic or semantic generalization,
and if it is only arrived at using syntactic or semantic evidence, I would just call it for
what it is.
References
Anderson, Stephen R. 2001. Lexicalism, incorporated (or incorporation, lexicalized). In CLS 36:
The panels,Akira Okrent & John Boyle (eds), 13–34. Chicago IL: Chicago Linguistic Society.
Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago IL:
University of Chicago Press.
Baker, Mark. 1995. Lexical and nonlexical noun incorporation. In Lexical Knowledge in the
Organization of Language, Urs Egli, Peter Pause, Christoph Schwarze, Arnim von Stechow &
Gotz Wienold (eds), 3–33. Amsterdam: John Benjamins.
Bergsland, Knut. 1955. A Grammatical Outline of the Eskimo Language of West Greenland.
Ph.D. Dissertation, University of Oslo.
Faarlund, Jan Terje. 1989. Autostructural analysis. Paper presented at the Workshop on Auto-
lexical Syntax, University of Chicago, April 16. (Also published in Autolexical Syntax: Ideas
and Methods, Eric Schiller, Elisa Steinberg & Barbara Need (eds). Berlin: Mouton. 1995).
Ford, Alan & Singh, Rajendra. 1991. Propedeutique morphologique. Folia Linguistica 25: 549–575.
Ford, Alan & Singh, Rajendra. 1996. Quelques avantages d’une linguistique débarrassée de la
morpho(pho)nologie. In Trubetzkoy’s Orphan, Proceedings of the Montreal Roundtable on
Morphonology: Contemporary Response [Current Issues in Linguistic Theory 144], Rajendra
Singh (ed.), 119–139. Amsterdam: John Benjamins.
Ford, Alan, Singh, Rajendra & Martohardjono, Gita. 1997. Pace Panini: Towards a Word-Based
Theory of Morphology. New York NY: Peter Lang.
Fortescue, Michael. 1984. West Greenlandic. London: Croom Helm.
Jackendoff, Ray. 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: OUP.
clitics, incorporation seems best handled with some sort of morphological apparatus. There
simply isn’t enough space in this paper to give my reviewer the explanation he deserves, but, in
a nutshell, once you accept that a single word can have a complex syntactic representation, one
or many complex semantic representation, and a phonemic representation, there will always
be enough “stuff ” at these levels to describe a morphological pattern in a relational way. Other
problems will arise, however, and that’s why I chose to make my point with simpler examples.
For a VERY long description of these other problems and the solution to all of this, I will
apologetically refer to Neuvel 2003.
Sylvain Neuvel
McCawley, James D. 1971. Prelexical syntax. Report of the 22nd Annual Roundtable Meeting on
Linguistics and Language Studies, 19–33. Washington DC: Georgetown University Press.
Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60: 847–894.
Mithun, Marianne. 1986. On the nature of noun incorporation. Language 62: 32–37.
Morgan, Jerry. 1969. On arguing about semantics. Papers in Linguistics 1: 49–70.
Neuvel, Sylvain. 2003. Metamorphology: A Word-based Analysis of Polysynthesis and Other
Multivalent Morphological Relations. Ph.D. dissertation. University of Chicago.
Neuvel, Sylvain & Singh, Rajendra. 2001. Vive la difference! What morphology is about. Folia
Linguistica 35(3–4): 313–320.
Pustejovsky, James. 1995. The Generative Lexicon. Cambridge MA: The MIT Press.
Rishel, Jørgen. 1974. Topics in West Greenlandic Phonology. Copenhagen: Akademisk Forlag.
Sadock, Jerrold M. 1980. Noun incorporation in Greenlandic: A case of syntactic word formation.
Language 56: 300–319.
Sadock, Jerrold M. 1983. The necessary overlapping of grammatical components. In CLS 19(2),
Amy Chukerman, Mitchell Marks & John F. Richardson (eds), 198–221. Chicago IL: Chicago
Linguistics Society.
Sadock, Jerrold M. 1986. Some notes on noun incorporation. Language 62: 19–31.
Sadock, Jerrold M. 1991. Autolexical Syntax: A Theory of Parallel Grammatical Representations
[Studies in Contemporary Linguistics]. Chicago IL: University of Chicago Press.
Sadock, Jerrold M. 2003. A Grammar of Kalaallisut or West Greenlandic Inuttut. Munich: Lincom.
Sadock, Jerrold M. In preparation. How to say things about verbs. ms.
Woodbury, Anthony C. 1996. On restricting the role of morphology in Autolexical syntax. In
Autolexical Theory: Ideas and Methods, Eric Schiller, Elisa Steinberg & Barbara Need (eds),
318–363. Berlin: Mouton de Gruyter.