The theoretical fallout of non-systematic sex denotation in French

BLS41 – Feb. 7, 2015
1
The theoretical fallout of non-systematic sex denotation in
French gender∗
Christopher Hammerly ([email protected])
Department of Linguistics, University of Maryland
1
Introduction
The distribution of gender in French, as well as other Romance languages, appears to be at once
systematic and arbitrary:
Systematic
(Semantically) Arbitrary
(1)
a. le chat
b. la chatte
(2)
(3)
‘the (male) cat’
‘the female cat’
(4)
a. le papillon
b. *la papillon
a. l(e)’homme ‘the man’
b. *l(a)’homme
(5)
a. *le sentinelle
b. la sentinelle
‘the guard’
a. *le femme
b. la femme
(6)
a. *le maison
b. la maison
‘the house’
‘the woman’
‘the butterfly’
Systematicity occurs when gender is consistent with the biological sex of the referent.
ã Nouns referencing animate male beings are specified with masculine gender (1a & 2a)
ã Nouns referencing animate female beings are specified with feminine gender (1b & 3b)
In lieu of semantic cues, systematicity occurs according to morphophonological properties of the
nominal, such as word final phoneme strings (Tucker et al. 1977).
ã Words ending in /EzÕ/, /sjÕ/, /zjÕ/, /ZjÕ/, and /tjÕ/ are feminine
ã All other nominals ending in /Õ/ are masculine
This rule accounts for as many as 98.2 percent of nouns ending in /Õ/, including (4a) and (6b).
Tentative claim: The gender specification of any given nominal is predictable from properties
of its referent and/or its morphophonological form.
However, this claim cannot be maintained, as neither semantic nor phonological features serve as
perfectly reliable cues to the gender of any given noun.
ã Not all nouns referencing male beings are masculine
(7)
∗
a. *le sentinelle mâle
b. la
sentinelle mâle
the.fem guard.fem male
‘the male guard’
Acknowledgments: Without Claire Halpert, my advisor at University of Minnesota, and Omer Preminger, my
advisor at University of Maryland, this project would not exist. Endless footnotes would be required to properly
acknowledge their contributions. I would also like to thank Ruth Kramer, Alexander Williams, and everyone at
S-Lab at UMD for discussion and comments. All claims (and any errors) are my own.
2
ã Not all nouns referencing female beings are feminine
(8)
a. le
papillon
femelle
the.masc butterfly.masc female
‘the female butterfly’
b. *la papillon femelle
ã Few, if any, morphophonological rules are exceptionless. What about the other 1.8 percent
from the rule for /Õ/?
These inconsistencies lead to a revision of the above claim:
Final claim: Properties of the referent and morphophonological form as indicators of gender
specification are probabilistic trends rather than deterministic certainties.
However, kids are able to use distributional information to acquire gender systems with ease (e.g.
Gagliardi & Lidz 2014) and adults make minimal speech errors and agree on the assignment of
gender to novel words (Tucker et al 1977; Karmiloff-Smith 1979).
O Gender systems remain learnable and usable despite imperfections in systematicity
The goal of this talk is to work towards a representation of the patterns of gender
specification in French in light of learnable non-systematicity
Proposals
• There are four sub-genders in French based on semantic (un)interpretability
• Both interpretable and uninterpretable gender features are housed on the nominal category
assigning head n (Kramer 2009, 2013). This leads gender specification to be a function of
nominalization in the syntactic component
• Theories of First-Phase Syntax (Marantz 2000, 2001; Arad 2003) can account for the
distinction between derivational versus inflectional morphology in French nominalizations
• Though specification occurs in syntax, patterns of gender specification are partially nonsystematic. No grammatical restrictions (i.e. licensing conditions) can be formulated to
account for the distribution of gender
• Restrictions on patterns of specification are restrictions at LF interpretation
2
The gender feature on n
2.1
Feature attributes
There are four feature attributes necessary to represent gender
1. Specification: presence (or absence) of a feature or attribute
BLS41 – Feb. 7, 2015
3
2. Type: relevant and sufficient conceptual-semantic information based on markedness
ã fem; feminine gender is marked in French
(9) Elles
aiment la
musique.
pro.3.fem.pl like.3.pl det.fem music.fem
‘They (group of all females) like music.’
#They (mixed group of males and females) like music.
(10) Ils
aiment la
musique.
pro.3.masc.pl like.3.pl det.fem music.fem
‘They (group of all males) like music.’
‘They (mixed group of males and females) like music.’
3. Valuation: agreement condition for PF interpretation
ã Positive (+) valuation marks feminine gender, negative (–) marks masculine, and unvalued ( ) features cause the derivation to crash at PF and must be valued via Agree
4. Interpretability: readability at the LF interface
ã Interpretable (i ) features are assigned a semantic signature at LF according to the type
and valuation of the feature, uninterpretable (u) features are unassigned
Note: Valuation and interpretability are independent attributes (Pesetsky & Torrego 2007)
This results in the following feature paradigm:
F
M
f
Controller/Valued [i: + fem] [i: – fem] [u: + fem]
Target/Unvalued
[ fem]
Table 1. Gender features in French
2.2
m
[u: – fem]
Housing the gender feature
I assume a distributed lexical framework, Distributed Morphology (DM; Halle & Marantz 1993,
1994; Harley & Noyer 1999)
ã Allows for a more direct exploration of the link between nominalization and gender specification (Ferrari 2005; Lowenstamm 2008)
ã Provides a platform to eschew entirely idiosyncratic explanations of gender specification
through the use of syntactic combinatorics rather than lexical sub-categorization
Principles of DM
ã Syntax ‘All The Way Down’: Morphological and syntactic elements share identical constituent structure formed with a shared operation (i.e. Merge)
ã Late Insertion: Morphological and syntactic elements are underspecified for phonological
and conceptual-semantic information in syntax
4
Elements of DM
ã Encyclopedia Entries: Conceptual-semantic signature of morphemes exponed via Encyclopedia Access at LF
ã Vocabulary Items: Phonological signature of morphemes exponed via Spell-Out at PF
√
ã L-Morphemes ( ): open-class, language-specific, indices representing the minimal shared
conceptual core independent of syntactic (sub)-categorization
ã F-Morphemes: closed-class, functional, syntactic feature bundles. The category assigning
heads v, a, and n c-command and categorize roots
O The nominal category assigning head n is the locus of valuation and interpretability of
gender (Kramer 2009, 2013)
Types of n in French:
n = [nml] [i :+fem] = F = feminine interpretable (refers to female sex property)
n = [nml] [i :–fem] = M = masculine interpretable (refers to male sex property)
n = [nml] [u:+fem] = f = feminine uninterpretable
n = [nml] [u:–fem] = m = masculine uninterpretable
In the next section, I support this assertion by showing that nominalization patterns, as
apparent through semantic and phonological suppletion, vary with changes in gender.
3
Root-derived and category-derived nominalization
There are two pathways for morphological expression:
ã Derivational: non-compositional and “idiosyncratic”
F Phonological suppletion: When a root is associated with a single Encyclopedia Entry,
but multiple Vocabulary Items
(11)
a. la chèvre [la SEvö]
‘the goat’
b. le bouc [l@ buk]
‘the male goat’
c. la bique [la bik]
‘the female goat’
F Semantic suppletion: When a root is associated with a single Vocabulary Item, but
multiple Encyclopedia Entries1
1
Note this is distinct from homophony, were multiple roots have coincidentally coinciding phonological forms.
BLS41 – Feb. 7, 2015
(12)
5
a. la chèvre [la SEvö]
‘the goat’
b. le chèvre [l@ SEvö]
‘the goat cheese’
ã Inflectional: compositional and productive
(13)
a. la chèvre [la SEvö]
‘the goat’
b. le chèvre -au [l@ SEvö-o]
‘the young goat’
How do we account for these differences in morphological expression?
ã Consider how the nominalization in (14) is distinguished from that in (15).
(14) Root-derived nominal (derivational)
(15) Nominal-derived nominal (inflectional)
DP
DP
D
la
nP
√
P n [u:+fem]
∅
√
SEvö
nP
D
l@
nP
√
n [u:–fem]
-o
P n [u:+fem]
∅
√
SEvö
First-Phase Syntax (Marantz 2000, 2001; Arad 2003) provides a theory to make this distinction
√
ã The first category assigning head (n, v, or a) c-commanding is a phase head
√
√
F The initial Merge between and n, [ n], schematized in (14), triggers an interpretive
cycle where phonological
and semantic interpretations are assigned to the constituent of
√
the phase-head ( )
F This domain, the ‘first phase,’ is where Idiosyncrasies occur in the form of phonological
and semantic suppletion of roots varying as a function of morphosyntactic context
√
ã Further Merging of “root-derived” forms results in “category-derived” forms, [[ n] n]
F The phonological and conceptual-semantic exponence of the root-derived form is fixed
F Additional categorizing projections (e.g. the higher nP in (15)) build on the form and
meaning of the lower category projection
6
O Tests must be used to distinguish between nominal- versus root-derived forms
ã Neither shared meaning nor phonological overlap alone allow us to detect these differences.
F Homophonous nouns can be false alarms for semantic suppletion
F Two forms with independent roots with closely related meanings are false alarms
for phonological suppletion
F Nouns historically derived with inflectional morphology can become root-derived
O For the remainder of the talk I focus on identifying root-derived forms varying
as a function of gender
ã Phonological suppletion is identified using idioms
ã Semantic suppletion is identified through identical forms with overlapping reference
4
Specification of root-derived forms
With this in mind, we can return to the data from examples (11-12), repeated below
(16)
a. le
chèvre
[l@ SEvö]
√
the.m GOAT.m
‘the goat cheese’
b. la chèvre
[la SEvö]
√
the.f GOAT.f
‘the goat’
c. le
bouc
[l@ buk]
√
the.m GOAT.M
‘the male goat’
d. la bique
[la bik]
√
the.f GOAT.F
‘the female goat’
Claim: All nouns in (16) share a common root but differ in specification of gender
ã (16a-b) are semantically suppletive
F Test: is the shared phonological form [SEvö] between (16a) and (16b) an instance of
homophony?
No! They are significantly overlapping in meaning—both are related,√approximately
speaking, to “things to do with goats.” This suggests a shared root, GOAT, whose
interpretation changes as a function of the gender specified on n
This is a productive process with numerous examples:
(17) a. le
finale
[l@ final]
(18)
√
the.m FINAL.m
‘the finale of a concert’
b. la finale
[la final]
√
the.f FINAL.f
‘the final match (sports)’
a. le
diesel
[l@ djezEl]
√
the.m DIESEL.m
‘the diesel fuel’
b. la diesel
[la djezEl]
√
the.f DIESEL.f
‘the diesel car’
BLS41 – Feb. 7, 2015
7
ã (16b-d) are phonologically suppletive
F Test: Are (16b-d) mutually replaceable in an idiom in the sense that ‘go’ and ‘went’ are
replaceable in ‘go bananas’ and ‘went bananas’ ?
Yes! chèvre, bouc, and bique preserve identical interpretations in idioms:
(19) devenir chèvre “to go crazy” (literal translation: “to become a goat”)
a. “...je vais devenir bouc...”2
“...I’ll go crazy...”
b. “je vais devenir bique...”3
“...I’ll go crazy...”
– This usage is not meta-linguistic—it is used to emphasize the sex of subject
– The added interpretation of ‘male’ in (16c) and ‘female’
in (16d) comes from changes in
√
gender rather than a change in interpretation of GOAT
Interim Take-Home Message
O All n’s (F, M, f, m) can be Merged with a given root resulting in semantic or phonological
suppletion varying as a function of the morphosyntactic environment provided by gender.
Caveat: just because all n’s can be Merged does not mean they are ‘licensed’ !
What is meant by ‘licensing’ ?
ã Not all roots combine with the entire inventory of n’s
(20)
a. le
papillon
√
the.m BUTTERFLY.m
‘the butterfly’
√
b. *the.m BUTTERFLY.M
Hypothetical interpretation: ‘the male butterfly’
√
c. *the.f BUTTERFLY.f
Hypothetical interpretation: ‘object related to butterflies’
√
d. *the.f BUTTERFLY.F
Hypothetical interpretation: ‘the female butterfly’
There are endless examples of this across all conceptual categories
√
ã Licensing must provide a way to restrict the n- pairings
2
http://nokiaphones.fr/forum/viewtopic.php?f=36&t=2015&start=170#p32430 (Accessed 11/25/14)
http://forum.doctissimo.fr/loisirs/buvette-dede/this-only-friend-sujet_13_627.htm#t200147
(Accessed 11/25/14)
3
8
5
Accounting for patterns of specification
√
Past attempts to account for specification patterns have used ‘licensing conditions’ to restrict npairings
ã What we definitely don’t want in licensing:
F Cannot amount to a list—this would result in a system no different than a lexical theory
ã However, the nature of ‘licensing conditions’ have been left up to interpretation
F Harley & Noyer (2000) and Acquaviva (2009): licensing is a restriction on Vocabulary
Insertion (i.e. Spell-Out), but the precise mechanism is unclear
F Kramer’s (p.c.) analysis of Amharic maintains and clarifies licensing as a restriction
on Vocabulary Insertion, but adds a semantic licensing condition. Different roots are
subject to different conditions.
ã Where past accounts fall short:
F Roots are still divided into categories (lists?) to dictate how licensing conditions apply
to match each root with the ‘appropriate’ n.
F Licensing conditions are a grammatical mechanism for Merging a root with a particular
sub-set of n’s or an n with a particular sub-set of roots
I argue that these licensing conditions are not at all a component of the grammar.
ã Do the starred sentences in (20) tell us about grammaticality or acceptability?
F Mismatches in ‘grammatical’ gender are treated as instances of ungrammaticality. In
reality, there is no reason to think that the grammar encodes any restriction of this sort.
Claim: All n-
√
pairings are licensed by the grammar, but not all pairings receive
sensible interpretations at the interfaces
What is the basis for this claim?
ã It is well-known that we can have perfectly grammatical, but unacceptable sentences:
(21) #Colorless green ideas sleep furiously. (Chomsky 1957)
(22) #The cheese the mouse the cat chased ate stinks.
At the level of the “word,” this is what is occurring with gender unacceptability
ã We know that phonological and semantic information play key roles in gender processing and
acquisition from experimental and computational research
F To maintain Late Insertion,
this information cannot be available in narrow syntax—the
√
point at which the [ n] Merge occurs.
ã Even given semantic information, not all roots of the same natural class use the same patterns
of gender
BLS41 – Feb. 7, 2015
9
F Numerous examples have shown how roots used to make reference to animate beings do
not always make regular use of the interpretable gender to denote the sex of the referent
(23)
a. le chat
b. la chatte
‘the (male) cat’
‘the female cat’
(24)
a. le papillon
b. *la papillon
‘the butterfly’
F An analogous argument holds for phonological cues.
ã The gender specification patterns of any given root changes over time
F We find a previously ‘ungrammatical’ form becoming ‘grammatical’
(25) le/la
ministre
√
the.m/the.f MINISTER.M/F
‘the male/female minster’
In fact, this is a previously ‘unacceptable’ form becoming ‘acceptable’
F The frequency with which this type of change occurs suggests this fact is not a change
in the basic structure of the grammar, but a change in the interpretive system
Which interpretive system restricts patterns of specification?
ã PF (Spell-Out/Vocabulary Insertion) and LF (Encyclopedia Access) are the two candidates
ã LF is the most likely candidate
F Unacceptable forms are articulable phonologically, but non-sensical semantically
F At their core, all gender systems are based in some semantic classification coded in the
interpretability of the feature (Corbett 1991)
F The focus of future theoretical, computational, or experimental work in this area will
expand the nature of this restriction
6
Conclusions
O First phase syntax proved necessary to preserve the idiosyncrasies of root-derived nominalization and the systematicity of category derived nominalization
O Within the root-derived nominalization, semantic and phonological suppletion was shown
to vary as a function of gender corroborating the idea that the locus of gender is on n.
√
O With those examples I argued there is no grammatical restriction on n- pairings
O However, as evident in differences in gender specification patterns across nouns referring to
animate beings, we know not all roots can be paired with every gender. This fact is due to
acceptability rather than grammaticality and is not encoded in the syntactic computation.
O The most likely location for the restriction on specification is in the interpretation of roots
at LF.
10
References
[Acquaviva, 2009] Acquaviva, P. (2009). Roots and lexicality in distributed morphology. In Alexandra Galani, Daniel Redinger and Norman Yeo (eds.). YPL2-Issue 10 (May 2009) Special IssueYork-Essex Morphology Meeting (YEMM). University of York. Department of Language and
Linguistic Science.
[Arad, 2003] Arad, M. (2003). Locality constraints on the interpretation of roots: The case of
hebrew denominal verbs. Natural Language & Linguistic Theory, 21(4):737–778.
[Chomsky, 1957] Chomsky, N. (1957). Syntactic structures. Mouton.
[Corbett, 1991] Corbett, G. G. (1991). Gender. cambridge textbooks in linguistics. Cambridge:
Cam.
[Ferrari, 2005] Ferrari, F. (2005). A syntactic analysis of the nominal systems of Italian and Luganda: How nouns can be formed in the syntax. PhD thesis, New York University, Graduate
School of Arts and Science.
[Gagliardi and Lidz, 2014] Gagliardi, A. and Lidz, J. (2014). Statistical insensitivity in the acquisition of tsez noun classes. Language, 90(1):58–89.
[Halle and Marantz, 1993] Halle, M. and Marantz, A. (1993). Distributed morphology and the
pieces of inflection. In Hale, K. and Keyser, S. J., editors, The View from Building 20, pages
111–176. MIT Press, Cambridge, Massachusetts.
[Halle and Marantz, 1994] Halle, M. and Marantz, A. (1994). Some key features of Distributed
Morphology. MIT Working Papers in Linguistics, 21:275–288.
[Harley and Noyer, 1999] Harley, H. and Noyer, R. (1999). Distributed morphology. Glot International, 4:3–9.
[Harley and Noyer, 2000] Harley, H. and Noyer, R. (2000). Formal versus encyclopedic properties of
vocabulary: Evidence from nominalizations. The lexicon-encyclopedia interface, pages 349–374.
[Karmiloff-Smith, 1979] Karmiloff-Smith, A. (1979). A functional approach to child language: A
study of determiners and reference, volume 24. Cambridge University Press.
[Kramer, 2009] Kramer, R. (2009). Definite markers, phi-features, and agreement: a morphosyntactic investigation of the Amharic DP. PhD thesis, UC Santa Cruz.
[Kramer, 2013] Kramer, R. (2013). Gender in amharic: A morphosyntactic approach to natural
and grammatical gender. Georgetown.edu.
[Lowenstamm, 2008] Lowenstamm, J. (2008). Little n, root, and types of nouns. The Sounds of
Silence: Empty Elements in Syntax and Phonology, pages 105–143.
[Marantz, 2000] Marantz, A. (2000). Roots: the universality of root and pattern morphology. In
conference on Afro-Asiatic languages, University of Paris VII, volume 3, page 14.
[Marantz, 2001] Marantz, A. (2001). Words. In 20th West Coast Conference on Formal Linguistics,
University of Southern California, pages 23–25.
BLS41 – Feb. 7, 2015
11
[Tucker et al., 1977] Tucker, R. G., Lambert, W. E., and Rigault, A. (1977). The French speaker’s
skill with grammatical gender: An example of rule-governed behavior. Mouton The Hague.
A
Additional examples
(26)
[l@ gaösÕ]
a. le
garçon
√
the.m BOY.M
“the boy”
b. le
garçon(n)
-et [l@ gaösOn-E]
√
the.m BOY.M -M
“the small boy”
-ne [la gaösO-n]
c. la garçon
√
the.f BOY.M -F
“the tomboy”
(27)
a. la bique
[la bik]
√
the.f GOAT.F
“the nanny goat”
b. la biqu
-ette [la bik-Et]
√
the.f GOAT.F -F
“the young female goat”
c. (tentative analysis)
le
biqu
-et [l@ bik-E]
√
the.m GOAT.F -M
“the young male goat”
(28) devenir chèvre “to go crazy” (literal translation: “to become a goat”)
a. “j’vais devenir biquette...”4
“...I’ll go crazy...”
b. *devenir biquet5
(29)
4
5
[l@ Sa]
a. le
chat
√
the.m CAT.m
“the cat”
b. le
chat
-∅ [l@ Sa]
√
the.m CAT.m -M
“the male cat”
-te [la Sa-t]
c. la chat
√
the.f CAT.m -F
“the female cat”
(30)
[l@ mutÕ]
a. le
mouton
√
the.m SHEEP.m
“the sheep”
b. le
belier
[l@ belije]
√
the.m SHEEP.M
“the ram (male sheep)”
c. la brebis
[la bö@bi]
√
the.f SHEEP.F
“the ewe (female sheep)”
https://twitter.com/vansofgusgus/status/481216466596155393 (Accessed 11/25/14)
No results matching a string search on http://www.google.fr