Yale Colloquium Ruth Kramer 1 THE MORPHOSYNTAX OF GENDER: EVIDENCE FROM AMHARIC Ruth Kramer, Georgetown University / November 28, 2011 1 INTRODUCTION 1 Gender = the sorting of nouns into two or more classes, as reflected in agreement 2 Gender ≠ declension class (Harris 1991, Aronoff 1994, Wechsler and Zlatić 2003, Alexiadou 2004) Genders often correlate with biological sex (male vs female) or some other nominal feature (animacy). There is significant literature on o …the sociolinguistics of gender (e.g., Hellinger and Bussmann 2001) o …the acquisition and processing of gender (e.g., Franceschina 2005) o …the typology of gender systems (e.g., Corbett 1991) Within the morphosyntactic literature, though, fundamental questions about the morphosyntax of gender remain controversial. What is the syntactic locus of gender? (N, GenderP) What are the properties of gender features? (especially wrt interpretability) Focus: What is the relationship between ‘grammatical gender’ and ‘natural gender’? Grammatical gender = idiosyncratic gender most often on inanimates (masculine, feminine, neuter) (1) a. French b. Hausa c. Russian matin sáafiyáa utro morning.M 3 morning.F morning.NEUT ‘morning’ ‘morning’ (Corbett 1991:53) ‘morning’ Natural gender = biological gender or sex (male, female, N/A) Key questions: how do the same morphological resources express both types of gender? Do both types of gender have the same syntactic configuration and/or use the same set of features? The relationship between grammatical gender and natural gender has been explored for only a handful of related languages (Spanish, Italian, Greek, etc.) …which all have similar gender systems heavily based on grammatical gender. Today: Using evidence from the language Amharic (Ethiosemitic), I argue that natural gender should play a central role in the morphosyntax of gender (cf. Kramer 2009). 1 Many thanks to Sandy Chung, Jorge Hankamer, Jim McCloskey, Héctor Campos, Mark Norris and audiences at the University of Pennsylvania and the University of California, Santa Cruz for highly useful comments and questions. Special thanks to the Amharic consultants whose judgments shaped this work, especially Senayit Ghebrehiywet, Betselot Teklu, Hileena Eshetu, Harya Tarakegn and Mehret Getachew Tadesse. All errors are my own. 2 This definition is taken from Hockett 1958:231; cf. Corbett 1991:1. 3 Gloss abbreviations: 3 – third person, AUX – auxiliary, DEF – definite marker, F – feminine, M – masculine, NEG – negation, NEUT – neuter, PF – perfect aspect, PL –plural, S – singular.
Transcript
Yale Colloquium Ruth Kramer
1
THE MORPHOSYNTAX OF GENDER:
EVIDENCE FROM AMHARIC Ruth Kramer, Georgetown University / November 28, 2011
1 INTRODUCTION1 Gender = the sorting of nouns into two or more classes, as reflected in agreement2
Gender ≠ declension class (Harris 1991, Aronoff 1994, Wechsler and Zlatić 2003, Alexiadou 2004)
Genders often correlate with biological sex (male vs female) or some other nominal feature (animacy).
There is significant literature on o …the sociolinguistics of gender (e.g., Hellinger and Bussmann 2001) o …the acquisition and processing of gender (e.g., Franceschina 2005) o …the typology of gender systems (e.g., Corbett 1991)
Within the morphosyntactic literature, though, fundamental questions about the morphosyntax of gender remain controversial.
What is the syntactic locus of gender? (N, GenderP)
What are the properties of gender features? (especially wrt interpretability) Focus: What is the relationship between ‘grammatical gender’ and ‘natural gender’?
Grammatical gender = idiosyncratic gender most often on inanimates (masculine, feminine, neuter) (1) a. French b. Hausa c. Russian
Natural gender = biological gender or sex (male, female, N/A)
Key questions: how do the same morphological resources express both types of gender? Do both types of gender have the same syntactic configuration and/or use the same set of features?
The relationship between grammatical gender and natural gender has been explored for only a handful of related languages (Spanish, Italian, Greek, etc.)
…which all have similar gender systems heavily based on grammatical gender. Today: Using evidence from the language Amharic (Ethiosemitic), I argue that natural gender should play a central role in the morphosyntax of gender (cf. Kramer 2009).
1 Many thanks to Sandy Chung, Jorge Hankamer, Jim McCloskey, Héctor Campos, Mark Norris and audiences at the University of Pennsylvania and the University of California, Santa Cruz for highly useful comments and questions. Special thanks to the Amharic consultants whose judgments shaped this work, especially Senayit Ghebrehiywet, Betselot Teklu, Hileena Eshetu, Harya Tarakegn and Mehret Getachew Tadesse. All errors are my own. 2 This definition is taken from Hockett 1958:231; cf. Corbett 1991:1. 3 Gloss abbreviations: 3 – third person, AUX – auxiliary, DEF – definite marker, F – feminine, M – masculine, NEG – negation, NEUT – neuter, PF – perfect aspect, PL –plural, S – singular.
Yale Colloquium Ruth Kramer
2
The Amharic gender system relies heavily on natural gender, and is difficult for previous analyses to account for.
Proposal: gender feature located on nominalizing head n (see e.g., Ferrari 2005, Acquaviva 2009) o Two kinds: interpretable (natural gender) and uninterpretable (grammatical gender) o Licensing conditions pair up roots with different types of n’s
This proposal avoids previous analytical pitfalls, is supported by independent evidence within Amharic and shows some promise in accounting for different types of gender systems.
Plan
Gender in Amharic (Section 2)
Previous accounts of the morphosyntax of gender (Section 3)
The proposal (Section 4)
Extension of the analysis beyond Amharic (Section 5)
Conclusions (Section 6)
2 GENDER IN AMHARIC Amharic
Member of the Ethiosemitic branch of Semitic language family (along with Tigrinya, Tigre, Harari, etc.)
National language of Ethiopia, ~15 million monolinguals, lingua franca
Like Semitic: root and pattern morphology, similar basic vocabulary (bet = house, ras = head)
Unlike Semitic: SOV, postpositions = head-final (influence of the neighboring Cushitic languages, e.g., Somali, Afar, etc.)
The Amharic gender system recognizes two genders: masculine and feminine.
No consistent morphophonological correlates of gender (Leslau 1995:161, Cohen 1970:71; one exception below)
Gender indicated by agreement on e.g., the definite marker, demonstratives, verbs (2) -u ‘the (m. sg.)’
-wa ‘the (f. sg.)’ The Amharic system for assigning gender is heavily reliant on natural gender.
The gender of an animate noun is assigned exclusively according to its natural gender (Leslau 1995: 161ff., Hartmann 1980:278ff., Appleyard 1995:33).
o Some male-female pairs have different roots (mostly kinship terms, domesticated animals). (3) Different Root Nominals
abbat ‘father’ ɨnnat ‘mother’
bal ‘husband’ mist ‘wife’ säw ‘man’ set ‘woman’ wäyfän ‘bull calf’ gidär ‘heifer’
o But the vast majority have the same root.
Yale Colloquium Ruth Kramer
3
(4) Same Root Nominals a. tämari-w tämari-wa student-DEF student-DEF.F the (male) student the (female) student
b. muʃɨrra-w muʃɨrra-wa
wedding.participant-DEF wedding.participant-DEF.F groom bride c. hakim-u hakim-wa doctor-DEF doctor-DEF.F the (male) doctor the (female) doctor d. halafi-w halafi-wa person.in.charge-DEF person.in.charge-DEF.F the (male) person in charge the (female) person in charge Walta hed12a2 Walta hed01a2
e. wɨʃʃa-w wɨʃʃa-wa
dog-DEF dog-DEF.F the (male) dog the (female) dog
The default (unmarked) gender is masculine (cf. Sauerland 2008). o If the natural gender of the referent is unknown, then the nominal is masculine.
(5) hɨs’an-u wänd näw set?
baby-DEF.M male is female? ‘Is the baby a he or a she?4’ (Leslau 1995:164)
o The nominal nobody takes masculine agreement (cf. Roca 1989)
If the natural gender of the referent for one of these animal nouns is known, though, it ‘overrides’ the feminine default, showing the dominance of natural gender within the system.
4 The noun hɨs’an ‘baby’ is a same-root nominal, i.e., it can be either masculine or feminine depending on whether it
refers to a male or female infant. 5 Other such animals include ɨbab ‘snake,’ asa ‘fish’ and nɨb ‘bee.’
Yale Colloquium Ruth Kramer
4
(8) ayt’-u mouse-DEF.M ‘the male mouse’
This type of nominal (= feminine default animal) will be crucial in the analysis! As for inanimate nominals, almost all of them are masculine (Leslau 1995:161, Cohen 1970:71). (9) Masculine Nouns (inanimate)
mot ‘death’ wɨdɨddɨr ‘competition’
kɨbɨr ‘honor’ bet ‘house’
wänbär ‘chair’ dɨmmɨr ‘total, sum’
dɨngay ‘stone’ wäräda ‘district’
kɨbab ‘circle’ gazet’a ‘newspaper’
Only a handful of inanimate nouns are feminine.6 (10) Feminine Nouns (inanimate)
As noted earlier, there is no consistent morphophonological correlate of gender across nominals.
The only exception is that in some same-root nominals, the female form has the suffix –it.7
(11) a. lɨdʒ lɨdʒ-it b. aroge arog-it8
boy, child girl old man old woman c. mänäkwse mänäkws-it d. t’ot’a t’ot’-it monk nun ape female ape
The suffix –it is neither a necessary nor a sufficient condition for feminine gender in Amharic, regardless of animacy (Leslau 1995:163-164, Cohen 1970:74).
6 It is difficult to calculate the exact number of feminine nouns since nouns are not listed in Amharic dictionaries with their gender (that in itself is an indication of how small a role grammatical gender plays in Amharic). After informally surveying the gender sections of three grammars (Leslau 1995, Hartmann 1980 and Cohen 1970), as well as performing some basic searches in the Walta Information Center Tagged Amharic News Corpus (Demeke and Getachew 2006), my best estimate is that there are about 20-30 feminine nouns (including toponyms). 7 Probably a descendent of the proto-Afroasiatic suffix *-at that indicated feminine gender; there are cognates in many other Semitic and Afroasiatic languages (e.g., Hebrew –et, -it, Ancient Egyptian –t ). See Zaborski 1992:37. 8 The final vowels in these nouns are deleted when the -it suffix is added in order to avoid hiatus. This is similar to other nominal suffixes, which also trigger deletion of the final vowel on the stem which they attach to (Leslau 1995:36).
Yale Colloquium Ruth Kramer
5
Upshot:
Grammatical gender is only relevant in Amharic for the small number of feminine inanimate nouns, and the even smaller number of feminine default animals.
Otherwise, the natural gender (or lack thereof) determines the gender of a nominal in Amharic. o If a nominal refers to a male referent, the nominal is masculine. o If a nominal refers to a female referent, the nominal is feminine. o If a nominal refers to a referent whose natural gender is unknown, or which does not have
natural gender, the nominal is masculine (by default).
Amharic is a language that has almost lost grammatical gender (cf. Ge’ez; Lambdin 1978); ‘residue’ of grammatical gender remains.
3 THE MORPHOSYNTAX OF GENDER: PREVIOUS APPROACHES Gender is not straightforward: Gender is not just a feature on N that’s idiosyncratically listed for each noun (see e.g., Carstens 2011).
(13) NP
g N [+FEM]
Many languages contain nouns that can have either masculine or feminine gender depending on the natural gender of the referent (Corbett 1991:181-2, Wechsler and Zlatič 2003, Alexiadou 2004).
o ‘Common gender’ nouns / same-root nominals (14) a. Amharic (tämari ‘student,’ hakim ‘doctor’; see (4))
b. Spanish (estudiante ‘student,’ patriota ‘patriot,’ testigo ‘witness;’ Harris 1991) c. Archi (Caucasian) (lo ‘child,’ misgin ‘poor person’ Corbett 1991:181)
d. Greek (odigos ‘driver,’ musikos ‘musician,’ ipurgos ‘minister’ Alexiadou 2004:40)
It would be undesirable to have two homophonous, synonymous nouns for each of these cases, one with masculine gender and one with feminine gender.
Gender is also not the head of its own projection, similar to NumP (Picallo 1991, see also Bernstein 1993). (15) GenP
3 Gen NP [+FEM] g N
o Unlike number, gender is not consistently phonologically expressed as a separate morpheme
from the nominal, nor is it consistently interpretable. o In fact, Gen would be (for some nouns) a syntactic head comprised only of uninterpretable
features, which is undesirable from a minimalist standpoint (cf. Chomsky 1995 on Agr nodes). o The original evidence for GenP from Catalan is not compelling (Kramer 2009), and the proposal
has been generally disputed on empirical and theoretical grounds (Ritter 1993, Alexiadou 2004).
Natural gender cannot simply be set aside in a morphosyntactic investigation of gender (see e.g., Bernstein 1993:117, Picallo 2008:50)
Yale Colloquium Ruth Kramer
6
This is not a useful approach if the theory of gender is meant to encompass languages like Amharic that have almost nothing but natural gender.
A handful of morphosyntactic analyses of gender do explicitly treat both natural and grammatical gender.9
A noun with unspecified gender receives gender via a lexical rule that refers to the natural gender of its discourse referent.
Lexical redundancy rule : assign feminine gender if referent is female (Roca 1989, Riente 2003)
Morphological feature co-occurrence rule : a rule that says if sex is male, then gender is masculine and likewise for feminine (Ralli 2002)
A noun with unspecified gender enters into an agreement/concord relationship with its referent (mechanics unclear; Alexiadou 2004)
These analyses contribute significantly in that they address both grammatical and natural gender.
However, they are not without drawbacks. Drawback 1: These analyses do not generalize well to Amharic.
Feminine default animals have a fixed default gender that is ‘overridden’ with natural gender. (17) a. ayt’-wa b. ayt’-u
mouse-DEF.F mouse-DEF.M ‘the mouse, the female mouse’ ‘the male mouse’
Their gender thus seems to be simultaneously specified (feminine default) and unspecified (corresponds to the natural gender of the referent).
These analyses generally ‘convert’ natural gender (male/female) to the gender feature used in the syntax (masculine/feminine, i.e., equivalent to grammatical gender).
o However, in Amharic, this ‘conversion’ would have to happen for almost every animate noun. This is overly complicated – why not have the natural gender feature simply be the gender of a given nominal?
Drawback 2: Many of the analyses assign the gender of a nominal from the sex of its discourse referent via a lexical (presyntactic) rule.
9 See also Pollard and Sag 1994:Ch.2 and Wechsler and Zlatić 2003:Ch. 4 for HPSG analyses that capture both grammatical and natural gender.
Yale Colloquium Ruth Kramer
7
However, it seems implausible for the discourse referents of a given derivation to be accessible to operations in the lexicon, in a standard minimalist syntax.
To put it another way, information does not usually flow from the syntax/semantics derivation to the lexicon.
Drawback 3: There is little discussion of the syntax of gender in these analyses.10
Gender features must be in the syntax (agreement, pronominal reference; see e.g., Picallo 2008).
Puzzle for minimalist syntax in particular: gender is sometimes an uninterpretable feature, sometimes an interpretable feature, but it does not cause a crash (pace Chomsky 2000, 2001)
Another Perspective: much morphological research of the last twenty years has been conducted in the non-lexical framework Distributed Morphology (DM; Halle and Marantz 1993, Harley and Noyer 1999, among many others)
In DM, there are no pre-syntactic morphological rules – morphological operations occur in the syntax (e.g., head movement) or during post-syntactic PF (e.g., affixation, cliticization, etc.)
The lexicon itself is distributed: there is a pre-syntactic lexicon contains feature bundles, which are given morphophonological content at PF and interpreted at LF(/the Encyclopedia).
Gender is often seen as a quintessentially lexical phenomenon (so much idiosyncrasy) and the analyses above all rely on lexical rules of one type or another.
Question: can we capture the above facts without pre-syntactic lexical rules and with a distributed lexicon?
Overall: The previous analyses that address both natural and grammatical gender (while insightful)…
…run into difficulties with Amharic (because of its heavy reliance on natural gender).
…rely on a spurious lexicon-discourse referent connection.
…are rooted in lexicalism and do not address the syntax of gender. Perhaps a new, non-lexical approach to the morphosyntax of gender can help address these problems.
4 THE MORPHOSYNTAX OF GENDER: A NEW APPROACH 4.1 Narrowing Down the Options A Starting Point: a productive line of morphological research within Distributed Morphology has been to decompose lexical categories (like N) into a category-neutral root (√) and a categorizing head (Marantz 1997, 2001, Arad 2003, 2005, among many others).11
10Gender features must be present in the syntax not only for agreement purposes, but also for pronominal reference(cf. Picallo 2008:47-48). Even in Amharic, a nominal with feminine grammatical gender like the city Asmara below can be referred to by a feminine pronoun (bolded).
in-Asmara-and vicinity-DEF.F 3-live.IMPF-3PL C-were-3PL 172 Ethiopia-n-PL-TOP ‘172 Ethiopians who were living in Asmara and its (lit. her) vicinity…’ (Walta tik23a11)
11 This idea existed before Distributed Morphology (see e.g., van Riemsdijk 1990 on n), and Distributed Morphologists are not the only ones who subscribe to it (see e.g., Lowenstamm 2008). Cf. also Borer 2005 for a similar approach, although Borer (2005:20-21) argues against the specific Distributed Morphology analysis adopted here.
Yale Colloquium Ruth Kramer
8
(18) vP
3 v √P = ‘ (to) hammer’ g √HAMMER
(19) nP
3 n √P = ‘(a) hammer’ g √HAMMER
There is now more ‘space’ for gender features in the syntax. Is the gender feature of a nominal on the root or on n? Gender on the Root? No. Same undesirable repetitions as gender on N in Section 3.
Same-root nominals (e.g. tämari ‘student’) would have to have two homophonous, synonymous roots. Gender on n? Now we’re talking. Perhaps a gender feature comprises part of the nominalizing head n.
Roots have licensing conditions, such that they are licit in the context of a n[-FEM] or n[+FEM] (Acquaviva 2009).
This allows for a simple treatment of same-root nominals (tämari ‘student’): no licensing conditions!
What are these licensing conditions? Easiest to think of them as conditions on morphophonological insertion of a particular root, assuming that morphophonological material is inserted post-syntactically.
However, these analyses also have drawbacks.
They fail to discuss how the gender features on n relate to natural vs grammatical gender. o This means that they have difficulty accounting for Amharic. Most pressingly, the feminine
default animals remain problematic.12
The n analyses also fail to discuss the syntactic properties (e.g., interpretability) of the gender features.13 Upshot: n analyses are a step forward, but more needs to be done. 4.2 The Proposal We’ve seen two strands of research:
Strand 2 (Section 4.1): do not address both natural and grammatical gender, non-lexicalist
Neither discuss interpretability, and neither can quite cover Amharic’s gender system.
12 If the feminine default animals are treated like same-root nominals, then it’s unclear how they receive feminine gender as a default. If they are treated as licensed only in the context of n[+FEM], then it is unclear how they could ever have masculine gender. 13 The exception is Kihm 2005; see Appendix 2 for some criticism of his approach to gender and interpretability.
Yale Colloquium Ruth Kramer
9
Let’s try to take the best from each strand, and develop an analysis that also accounts for all of Amharic gender and deals with interpretability. Ingredient 1 (of 3): Natural gender is an interpretable gender feature housed on some types of n.14 (20) Types of n (incomplete list)
a. n i [+FEM] Female natural gender b. n i [-FEM] Male natural gender c. n No natural gender (or natural gender irrelevant/unknown) = ‘plain’ n
Ingredient 2: Licensing conditions determine which roots combine with which flavor of n (Acquaviva 2009).
Different root nominals (bal ‘husband’ and mist ‘wife’) are licensed under either n[+FEM] or n[-FEM]. (21) nP
3
√P n i[+FEM] = ‘wife’ g √MIST
(22) nP
3
√P n i[-FEM] = ‘husband’ g √BAL
Same root nominals (tämari ‘student’): licensed under any n in (20) (23) nP
3
√P n i[+FEM] = ‘(female) student’ g √TÄMARI
(24) nP
3
√P n i[-FEM] = ‘(male) student’ g √TÄMARI
(25) nP 3
√P n = ‘student (gender unknown or irrelevant)’ g √TÄMARI
14 The interpretation of these features is perhaps something as simple as [λx.x is male] and [λx.x is female]. They take an entity/individual and return true if that entity is female for [+FEM] or male for [-FEM]. Alternatively, the gender feature could trigger a presupposition that the discourse referent associated with the nominal is female or male, similar to how gender features work in pronouns (see e.g., Heim and Kratzer 1998 among many others).
Yale Colloquium Ruth Kramer
10
The nominal in (25) will be realized with masculine agreement, since masculine is the default.
Recall that morphophonological exponents are inserted after syntax.
These exponents (Vocabulary Items) are pairings of bits of morphophonology and a set of features.
There may not be a one-to-one match between the feature bundle in the syntax and the features of the Vocabulary Item inserted to realize it – and that’s ok.
o Morphophonological exponents are often underspecified wrt syntactic information (e.g., agreement).
Vocabulary Items compete for insertion at a particular feature bundle according to the Subset Principle (Halle 1997): insert the Vocabulary Item that matches the most features on a feature bundle, without containing any features not present in the bundle.
(26) Vocabulary Items for the Definite Marker15
a. D, [DEF], [+FEM] ↔ -wa b. D, [DEF] ↔ -u
Assume that definite D agrees in gender with n (however this is accomplished).
A definite D with a [+FEM] feature will be realized as (26)a –wa.
A D with a [-FEM] feature will be realized as (26)b –u and so will a D with no gender feature ((26)a -wa has a feature not present on the feature bundle!)
Overall: anything but feminine gender = ‘masculine’ exponents (elsewhere) Upshot: simple way to capture the generalizations that if a nominal is female, it has feminine gender; if it is anything else (male, unknown natural gender), it has masculine gender But what about the inanimates and the feminine default animals? The masculine inanimates in fact come for free given what has already been sketched.
They are licensed under ‘plain’ n since they don’t have natural gender. (27) nP
3
√P n = ‘house’ g √BET
This will result in masculine gender as a default, exactly like when the natural gender is unknown/irrelevant for an animate nominal.
We just need the final ingredient for the feminine default animals and the feminine inanimates. Ingredient 3: The feminine grammatical gender on inanimates and feminine default animals is an uninterpretable [+FEM] feature on n.
15 The definite marker is always –u when the noun is plural regardless of gender (Leslau 1995:155). In terms of the Vocabulary Items for the definite marker in (26), this can easily be captured by having a [-PL] feature as part of the features of the Vocabulary Item –wa. Then, -wa will be inserted for the definite determiner associated with singular feminine nominals, and –u for the definite determiner for all other nominals.
Yale Colloquium Ruth Kramer
11
(28) Types of n (complete) a. n i [+FEM] Female natural gender
b. n i [-FEM] Male natural gender c. n No natural gender (or natural gender irrelevant/unknown) d. n u [+FEM] Feminine grammatical gender
Language in general often has uninterpretable and interpretable versions of the same feature (e.g., number on nominals and on T), even on the same head (e.g., Pesetsky and Torrego 2007: Q feature on C). Feminine inanimate nominals are only licensed under n u[+FEM]. (29) nP
3
√P n u[+FEM] = ‘sun’ g √S’ÄHAY
They surface with the same kind of agreement as a female nominal (same [+FEM] feature on the same head – it’s just not interpreted).
Feminine default animals are licensed under the interpretable n’s ((20)ab), similar to a same-root nominal. (30) nP
3
√P n i[+FEM] = ‘(female) mouse’ g √AYT’
(31) nP
3
√P n i[-FEM] = ‘(male) mouse’ g √AYT’
However, when their natural gender is not known/relevant, they are licensed under the uninterpretable n[+FEM] (and not a ‘plain’ n). (32) nP
3
√P n u[+FEM] = ‘mouse’ g √AYT’
o Thus, when the natural gender of a feminine default animal is known, the nominal will have the
natural gender of the referent. o When the natural gender is unknown, it will have feminine gender.16
16 This requires same-root nominals in Amharic to have licensing conditions, because they must not be licensed under n u[+FEM] (so that they do not have a feminine default). See the conclusion for some thoughts on this.
Yale Colloquium Ruth Kramer
12
Why no uninterpretable [-FEM]? At the most basic level, there is no need for such a feature. Masculine inanimates trigger ‘masculine’ agreeing forms by default (lack of gender). o In languages with more than two genders, this feature may be needed (see Section 5.1). Overview and Assessment:
Gender assignment system that is almost entirely based on natural gender as an interpretable feature on n, and ‘masculine’ forms as a default for anything that does not have a [+FEM] on n17
The system also accounts for the residue of feminine grammatical gender in the language using uninterpretable version of the same feature on the same head: n u[+FEM]18
More successful than previous analyses in that… o All the Amharic facts are accounted for o There’s no need for a discourse referent-lexicon connection o No more ‘calculation’ of gender from natural gender o What about interpretability?
o The analysis is explicit about the interpretable vs uninterpretable properties of the gender feature o Lingering issue: why doesn’t an uninterpretable feature cause a crash? See the Appendix!
4.3 Additional Evidence for the Analysis The analysis is independently supported by additional evidence from Amharic including…
the feminine suffix –it
the interaction of gender and number
diminutives Evidence 1: -it. If n comes in four different flavors (feature bundles), we might expect there to be a distinct morphophonological realization for at least one of these flavors.
Indeed, there is. The female natural gender suffix –it is a realization of the interpretable n[+FEM] in the context of particular roots.
(33) a. lɨdʒ lɨdʒ-it b. aroge arog-it
boy, child girl old man old woman
(34) n i [+FEM] ↔ -it /__ [√LƗDƷ , √AROGE…]19
Evidence 2: Gendered Plurals. Amharic has a regular plural suffix and a set of irregular pluralization strategies (different suffix, partial reduplication, phonotactic changes, etc.)
17 The relation between gender and the root nominal also has several of the special, local characteristics of a categorizing head/root relationship (Marantz 1997, 2001), e.g., root-specificity (different roots take different genders = types of n’s) and p aradigmatic gaps (not all roots are possible with all genders = types of n’s) 18 Why are there no human nominals licensed under uninterpretable n[+FEM]? Perhaps this is because natural gender is more relevant/detectable for humans than animals. Cf. Spanish where the majority of the animals have fixed gender, but only four human nominals do (Harris 1991)) 19 It may be that PF does not ‘see’ interpretability, since it is not relevant for PF operations. In that case, -it would be the realization of n[+FEM][+ANIMATE] in the context of particular roots.
Yale Colloquium Ruth Kramer
13
o In Kramer 2009, I show how irregular plural morphology does not compete with the regular plural suffix for morphophonological insertion at the same node.
o There are double plurals: näfs-at-otʃtʃ ‘souls’
o Every nominal with an irregular plural can be alternatively regularly pluralized: näfs-otʃtʃ. o Irregular plurals also show characteristics of the local relationship between a categorizing head and its
root (Marantz 1997: root-specificity, paradigmatic gaps, idiosyncrasies), whereas regular plurals do not. This led me to propose that irregular plural morphology is a realization of n (cf. Lecarme 2002), whereas regular plural morphology is the realization of Num. (36) NumP 3 nP Num Regular plural
3 otʃtʃ
√P n Irregular plural g at
√NÄFS If n has a gender feature and a plural feature, it is predicted (ceteris paribus) (i) that irregular plurals in Amharic will be capable of varying with gender, and (ii) regular plurals will not be (since they do not have gender). Both predictions are borne out! (i) Certain irregular plurals are gendered: they take separate masculine and feminine suffixes.
(37) Gendered Irregular Plural 1 : k’ɨddus
k’ɨddus ‘saint’
k’ɨddus-an ‘saints’ (masc. pl. or mixed group)
k’ɨddus-at ‘saints’ (fem. pl.)
(38) Gendered Irregular Plural 2: ityop’p’ɨyawi
ityop’p’ɨyawi ‘Ethiopian (person)’
ityop’p’ɨyawi-yan ‘Ethiopians (masc. pl. or mixed group)’
ityop’p’ɨyawi-yat ‘Ethiopians (fem. pl.)’
(ii) No regular plurals vary with respect to gender; both masculine and feminine nominals take -otʃtʃ. (39) Masculine Feminine
The restriction of gendered plurals to irregular plurals is puzzling unless gender is a feature on n.
Yale Colloquium Ruth Kramer
14
Evidence 3: -it and Plurals. There is a curious asymmetry in the behavior of the feminine suffix -it wrt plurals.
Nominals ending in -it are freely regularly pluralized.
(40) a. mänäkws-it-otʃtʃ b. arog-it-otʃtʃ nuns old women
c. muʃɨrr-it-otʃtʃ d. t’ot’-it-otʃtʃ brides female apes
However, they cannot be irregularly pluralized. (41) Singular Irregular Plural *Feminine Irregular Plural
a. mänäkwse(-it) mänakos-at *mänakos-it-at, *mänäkos-at-it monk(-FEM) monk-PL
b. mämhɨr(-t) mämhɨr-an *mämhɨr-t-an, *mämhɨr-an-t
teacher(-FEM) teacher-PL This asymmetry is predicted if gender features are on n.
The feminine suffix and the regular plural suffix are independent heads in the syntax (n and Num, respectively) and don’t compete for morphophonological insertion at the same slot.
However, the feminine suffix and any irregular plural affixes compete for insertion at the n node. o Only one Vocabulary Item (either the feminine suffix or an irregular plural marker) may be
inserted. o Why does the irregular plural ‘win’ the competition for n when the nominal is plural? Appendix!
The contrast between (40) and (41) falls out if gender is on n. Evidence 4: Diminutives Amharic has a highly productive diminutive formation for both inanimate and animate nominals (Leslau 1995:167-169).
There is not a consistent phonological correlate of diminutivization (occasionally –it is used), but all diminutivized nominals trigger feminine agreement.
(42) a. bet-u Non-Diminutive
house.DEF.M ‘the house’
b. bet-wa t-amɨr-allätʃtʃ Diminutive
house-DEF.F 3FS-be.cute-AUX.3FS ‘The (adorable little) house is cute’
Diminutives (and other types of evaluative morphology) are a rich area of morphological investigation (see e.g., Scalise 1986, Stump 1993, Steriopolo 2008, Fortin 2011, and many others).
However, it is not entirely clear what category diminutivizing morphology should have (or even whether diminutives are inflectional or derivational morphology).
It is likely there are several different kinds of diminutives, both potentially within a language and across languages (see e.g., Anderson 2005, Steriopolo 2008, De Belder et al. 2009)
Some possibilities for the syntactic locus of diminutivization o SizeP (Steriopolo 2008, De Belder et al. 2009)
Yale Colloquium Ruth Kramer
15
o LexP (between n and the root; De Belder et al. 2009) o lP (where l = phonological exponent of diminutive; Lowenstamm 2008) o nP (Wiltschko 2006, Steriopolo and Wiltschko 2007, Steriopolo 2008)
Regardless, the fact that the diminutive triggers feminine agreement in Amharic (without any interpretation of female natural gender) indicates that it has an uninterpretable [+FEM] feature.20
An uninterpretable [+FEM] feature was needed to analyze the feminine inanimates and the feminine default animals.
The diminutive thus presents independent evidence for an uninterpretable feminine gender feature in Amharic, which may even be on n.21
5 EXTENSION BEYOND AMHARIC 5.1 Back to Romance One of the original shortcomings of earlier analyses was that they were difficult to extend to Amharic. Can the present analysis cover the languages treated in earlier work? Focus: Spanish (Roca 1989, Harris 1991, Alexiadou 2004) (43) ANIMATES: (cf. (16))
Different-root nominals (mujer ‘woman’ vs hombre ‘man’)
Animate nominals that have a fixed gender (persona ‘person.F’) = epicenes INANIMATES
Inanimates (domicilio ‘house.M’ vs residencia ‘residence.F’)
DEFAULT: masculine (Roca 1989) Different-root nominals are licensed under a particular interpretable n (e.g., n[+FEM] for mujer, n[-FEM] for hombre). Same-root nominals are licensed under any of the n’s in (19) (male, female, ‘plain’).
So far, the same as Amharic! Inanimates: some are licensed under plain n (domicilio), and some are licensed under n u[+FEM] (residencia)
Spanish has more of the latter than Amharic = more feminine inanimate nouns.
This is the source of the intuition that grammatical gender plays a bigger role in Spanish than in Amharic.
20 As well as an interpretable ‘diminutive’ feature (corresponding to the interpretation of cuteness, smallness, etc; Jurafsky 1996.). 21 This analysis of diminutive gender raises the interesting question of what happens when a given nominal has more than one gender feature, e.g., if a different-root nominal like bäre ‘ox,’ which is licensed under a n[-FEM] is diminuized.
Cohen (1970:77) reports that the resulting gender is feminine (yɨtʃtʃ bäre ‘this (cute) ox’). In fact, it seems to be a robust
cross-linguistic generalization that the gender feature on a diminutive morpheme always determines the gender of the nominal, whether feminine as in Amharic, neuter (as in German and Yiddish), or masculine (as in some diminutives in Serbian-Croatian; Wechsler and Zlatić 2003). It may be that the structurally highest gender feature is simply what agreeing elements first encounter when/if they probe down the tree, and thus this feature ends up as the gender of the nominal; see Kramer 2009, Steriopolo and Wiltschko 2010.
Yale Colloquium Ruth Kramer
16
Epicenes: many animals and a few humans are licensed only under n u[+FEM]. (44) nP
3
n u[+FEM] √P = ‘person’ g √PERSONA
Discussion
It is straightforward to use the ingredients of the previous analysis (interpretable and uninterpretable gender features on n, licensing conditions) to analyze the Spanish gender system.
But why does Spanish lack feminine default animals, i.e., animals where natural gender ‘overrides’ grammatical gender?
o Other languages have nominals like the feminine default animals often called hybrid nouns (French, Russian, etc.; Corbett 1991:Ch. 8, Matushansky to appear)
Nouns that agree either in grammatical gender or in natural gender
e.g., Russian, vrač ‘doctor’ referring to a woman can be either masculine according to its declension class ( = grammatical gender) or feminine when it refers to a female doctor
o What distinguishes Spanish from Amharic, French and Russian? 5.2 Other Types of Gender Systems So far, the investigation has been limited to:
two-gender systems (masculine vs feminine)
sex-based gender systems (grammatical gender correlates with natural gender to some degree)
Does the gender assignment system developed here generalize across different types of gender systems?
Will look at just a sampling of gender systems for space/time reasons Three-gender sex-based systems (feminine, masculine, neuter):
Examples: German, Greek, Russian, Serbian-Croatian, etc.
A crucial empirical question: what is the default gender in these languages?
If the neuter is default (at least for inanimates), then these languages provide evidence for an uninterpretable masculine feature on n (not nec in two gender languages) and thus that both types of gender features are binary.
o n u [+FEM] feminine o n u [-FEM] masculine o n neuter
Sex-based gender systems that lack grammatical gender (e.g., English)
These languages will lack n’s that have uninterpretable gender features.
There are therefore three types of n in these languages: interpretable female gender on n, interpretable male gender on n, and plain n (for any nominals that lack natural gender).
Some of these languages have all three types: English, Tamil (Dravidian, Corbett 1991:8-10), Defaka (Niger-Congo, Corbett 1991:12)22
22 Technically, in Tamil and Defaka, male humans have one gender, female humans another, and everything else is neuter. The interpretable gender features on n’s in Tamil and Defaka are thus bundled with a human feature: n[+FEM][+HUMAN], n[-FEM][+HUMAN], and ‘plain’ n[-HUMAN].
Yale Colloquium Ruth Kramer
17
Some of these languages just contrast one natural gender against all other nominals. o Female gender on n and plain n: Diyari (Australian; Corbett 1991:10), Dizi (Omotic; Corbett
1991:11), Halkomelem (Salish; Corbett 1991:11) o Male gender on n and plain n: Kala Lagaw Ya (Australian), Kolami, Ollari, and Parji (all
Dravidian) (Corbett 1991:10-11)
The analysis here covers sex-based systems without grammatical gender very well. Non-sex-based gender systems (many Bantu languages, many Algonquian languages, etc.)
What do they base gender assignment on?
Animacy, type of animal, size, edible-ness, etc. – semantic properties of a nominal (Corbett 1991:31)
Also, morphophonological form (Qafar, Hausa, Bantu partially; Corbett 1991:47)
A way forward: o The relevant semantic properties are all potential interpretable features on n o The relevant morphophonological exponents are realizations of (arbitrarily) different n’s
(Lowenstamm 2008)
Overall: The analysis shows promise in generalizability beyond Amharic, although this is of course the (tip of the) tip of the iceberg.
6 CONCLUSIONS Summary
The Amharic gender system is highly dependent on natural gender.
Previous gender analyses have difficulty extending to Amharic and have some analytical drawbacks.
Proposal: natural gender is interpretable gender features on n, grammatical gender is an uninterpretable gender feature on n, and licensing conditions determine which roots combine with which n
This proposal is supported by independent evidence within Amharic and has the potential to be extended to other types of gender systems.
A Diachronic Perspective
Ge’ez (ancient Ethiosemitic language; indirect ancestor of Amharic) had a robust grammatical gender system and much more widespread gender agreement than Amharic (Lambdin 1978)
The loss of most gender agreement in Amharic could have triggered a shift away from grammatical gender (cf. the loss of grammatical gender in English; Curzan 2003, Platzer 2005)
Evidence for this: for younger speakers (<25), the shift is nearly complete. They tend to treat any feminine inanimate noun as a diminutive, and treat feminine default animals like same-root nominals (natural gender only).
Open Questions
What does a n-based analysis of gender predict about nominalizations of other categories? How is gender used to nominalize other lexical categories within Amharic and in other languages? (Ferrari 2005)
This analysis places the bulk of the work of gender assignment on the licensing conditions. Is this appropriate/ideal? (Alternative: grammatical gender on the root, natural gender on n; Kramer 2009)
REFERENCES Acquaviva, Paolo. 2009. Roots and lexicality in Distributed Morphology. In Alexandra Galani, Daniel
Redinger and Norman Yeo, eds., York-Essex Morphology Meeting 2. http://www.york.ac.uk/language/ypl/ypl2issue10.html
Yale Colloquium Ruth Kramer
18
Alexiadou, Artemis. 2004. Inflection class, gender and DP-internal structure. In Gereon Müller, Lutz Gunkel, and Gisela Zifonun, eds. Explorations in Nominal Inflection. Berlin: Mouton. 21-50.
Alexiadou, Artemis. 2009. Ro[u:]ting the interpretation of words. Paper presented at Root Bound, USC, Los Angeles, CA.
Anderson, Stephen R. 2005. Morphology. In Lynn Nadel, ed. Encyclopedia of Cognitive Science. Hoboken: Wiley. Appleyard, David. 1995. Colloquial Amharic: A Complete Language Course. New York: Routledge. Arad, Maya. 2003. Locality constraints on the interpretations of roots. Natural Language and Linguistic Theory 21.
737-778. Arad, Maya. 2005. Roots and Patterns: Hebrew Morpho-syntax. Dordrecht: Springer. Aronoff, Mark. 1994. Morphology by Itself: Stems and Inflectional Classes. Cambridge, Mass.: MIT Press. Bernstein, Judy. 1993. Topics in the syntax of nominal structure across Romance and Germanic languages.
Doctoral dissertation, City University of New York. Borer, Hagit. 2005. Structuring Sense. Volume 1: In Name Only. Oxford: OUP. Borer, Hagit. 2008. Notes on late insertion. Keynote talk presented at WCCFL 27, Los Angeles, CA. Borer, Hagit. 2009. Very late insertion. Paper presented at Root Bound, USC, Los Angeles, CA. Carstens, Vicki. 2011. Hyperactivity and hyperagreement in Bantu. Lingua 121. 721-741. Chomsky, Noam. 1995. The Minimalist Program. Cambridge: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries, the framework. In Roger Martin, David Michaels and Juan
Uriagereka, eds. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge: MIT Press. 89-155.
Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed. Ken Hale: A Life in Language. Cambridge: MIT Press. 1-52.
Cohen, Marcel. 1970. Traité de Langue Amharique. 2nd ed. Paris: Institut d’Ethnologie. Corbett, Greville. 1991. Gender. Cambridge: CUP. Curzan, Anne. 2003. Gender Shifts in the History of English. Cambridge: CUP. De Belder, Marijke, Noam Faust and Nicola Lampitelli. 2009. On an inflectional and a derivational
diminutive. Paper presented at Roots, Stuttgart. Demeke, Girma A. and Mesfin Getachew. 2006. Mannual annotation of Amharic news items with part-of-
speech tags and its challenges. Ethiopian Languages Research Center Working Papers 2. 1-16. Dowty, David and Pauline Jacobson.1988. Agreement as a semantic phenomenon. In Proceedings of ESCOL 5.
95-108. Epstein, Samuel., H. Kitahara, and D. Seely, 2010. Uninterpretable features: what are they, and what do they
do? In M. Putnam, ed. Exploring Crash-Proof Grammars. John Benjamins, Amsterdam. 125–142. Ferrari, Franca. 2005. A syntactic analysis of the nominal systems of Italian and Luganda: how nouns can be
formed in the syntax. Doctoral dissertation, New York University. Fortin, Antonio. 2011. The morphology and semantics of expressive affixes. Doctoral dissertation, Oxford
University, Oxford, England. Franceschina, Florencia. 2005. Fossilized Second-Language Grammars: The Acquisition of Grammatical Gender.
Amsterdam: Benjamins. Halle, Morris. 1997. Distributed morphology: impoverishment and fission. In Benjamin Bruening, et al., eds.,
MIT Working Papers in Linguistics 30: Papers at the Interface. Cambridge: MITWPL. 425-449. Halle, Morris and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In Ken Hale and
Samuel Jay Keyser, eds. The View from Building 20. Cambridge: MIT Press. 111-176. Harley, Heidi. 2009. Roots: identity, insertion, idiosyncrasies. Paper presented at Root Bound, USC, Los
Angeles, CA. Harley, Heidi and Rolf Noyer. 1999. Distributed morphology (State-of-the-Article). Glot International 4. 3-9. Harris, James. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22. 27-62. Hartmann, Josef. 1980. Amharische Grammatik. Wiesbaden: Steiner. Heim, Irene and Angelika Kratzer. 1998. Semantics in Generative Grammar. Oxford: Blackwell. Hellinger, Marlis and Hadumod Bussmann, eds. 2001. Gender Across Languages: The Linguistic Representation of
Women and Men. 3 vols. Amsterdam: Benjamins. Hockett, C.F. 1958. A Course in Modern Linguistics.. New York: Macmillan.
Yale Colloquium Ruth Kramer
19
Jurafsky, Daniel. 1996. Universal tendencies in the semantics of the diminutive. Language 72. 533-578. Kihm, Alain. 2005. Noun class, gender and the lexicon-syntax-morphology interfaces: a comparative study of
Niger-Congo and Romance languages. In Guglielmo Cinque and Richard S. Kayne, eds. The Oxford Handbook of Comparative Syntax. Oxford: OUP. 459-512.
Kramer, Ruth. 2009. Definite markers, phi-features, and agreement: a morphosyntactic investigation of the Amharic DP. Doctoral dissertation, UC Santa Cruz.
Lambdin, Thomas O. 1978. Introduction to Classical Ethiopic (Ge’ez). Ann Arbor, Michigan: Scholars Press. Lecarme, Jacqueline. 2002. Gender “polarity:” theoretical aspects of Somali nominal morphology. In Paul
Boucher and Marc Plénat, eds. Many Morphologies. Somerville, MA: Cascadilla Press. 109-141. Legate, Julie Anne. 2002. Phases in ‘Beyond Explanatory Adequacy.’ Ms., MIT. Lowenstamm, Jean. 2008. On little n, √, and types of nouns. In Jutta Hartmann, Veronika Hegedűs, and
Henk van Rimesdijk, eds. Sounds of Silence: Empty Elements in Syntax and Phonology. Amsterdam: Elsevier. 105-144.
Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Marantz, Alec. 1997. No escape from syntax. University of Pennsylvania Working Papers in Linguistics 4. 201-225. Marantz, Alec. 2001. Words Ms., Massachusetts Institute of Technology. Matushansky, Ora. to appear. Gender confusion. In Lisa Cheng and Norbert Corver, eds. Diagnosing Syntax. Pesetsky, David and Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In
Simin Karimi, Vida Samiian and Wendy K. Wilkins, eds. Phrasal and Clausal Architecture. Amsterdam: Benjamins. 262-294.
Picallo, M. Carme. 1991. Nominals and nominalization in Catalan. Probus 3. 279-316. Picallo, M. Carme 2008. Gender and number in Romance. Lingue e Linguaggio 7. 47-66. Platzer, Hans. 2005. The development of natural gender in English, or: sex by accident. In Nikolaus Ritt and
Herbert Schendl, eds. Rethinking Middle English. Frankfurt: Peter Lang. 244-262. Pollard, Carl and Ivan Sag. 1994. Head-Driven Phrase Structure Grammar. Stanford: CSLI. Ralli, Angela. 2002. The role of morphology in gender determination: evidence from Modern Greek. Journal of
Linguistics 40. 519-551. van Riemsdijk, Henk. 1990. Functional prepositions. In H. Pinkster and I. Genée, eds. Unity and Diversity:
Papers Presented to Simon C. Dik on his 50th Birthday. Foris: Dordrecht. 229-241. Riente, Lara. 2003. Ladies first: the pivotal role of gender in the Italian nominal inflection system. McGill
Working Papers in Linguistics 17.2. 1-54. Ritter, Elizabeth. 1993. Where’s gender? Linguistic Inquiry 24. 795-803. Roca, I. M. 1989. The organisation of grammatical gender. Transactions of the Philological Society 87. 1-32. Sauerland, Uli. 2008. On the semantic markedness of phi-features. In Daniel Harbour, David Adger, and
Susana Béjar, eds. Phi Theory. Oxford: OUP. 57-82. Scalise, Sergio. 1986. Generative Morphology. 2nd ed. Dordrecht: Foris. Steriopolo, Olga. 2008. Form and function of expressive morphology: a case study of Russian. Doctoral
dissertation, University of British Columbia, Vancouver. Steriopolo, Olga and Martina Wiltschko. 2007. Parameters of variation in the syntax of diminutives. In: Milica
Radisic, ed. Proceedings of the 2007 Canadian Linguistics Association Annual Conference. Steriopolo, Olga and Martina Wiltschko. 2010. Distributed GENDER hypothesis. In G. Zybatow et al., eds.
Formal Studies in Slavic Linguistics: Proceedings of the Formal Description of Slavic Languages 7.5. New York: Peter Lang GmbH. 155-172.
Stump, Gregory T. 1993. How peculiar is evaluative morphology? Journal of Linguistics 29. 1-38. Wechsler, Stephen and Larisa Zlatić. 2003. The Many Faces of Agreement. Stanford: CSLI. Wiltschko, Martina. 2006. Why should diminutives count? In H. Broekhuis, et al., eds. Organizing Grammar.
Linguistic Studies in Honor of Henk van Riemsdijk. Berlin: Walter de Gruyter: 669-679. Zaborski, Andrzej. 1992. Afro-Asiatic languages. In William Bright, ed. The International Encyclopedia of
Linguistics. Oxford: OUP. 36-37.
Yale Colloquium Ruth Kramer
20
APPENDIX Appendix 1: Irregular Plural Affix Beats Feminine Suffix. When given a feature bundle for n that contains [+FEM] and [+PL] features, an irregular plural affix is inserted and not the feminine suffix (see (41)). This is because the Vocabulary Item for the feminine suffix is slightly more complex than described above. Specifically, it contains a [-PL] feature. (45) n, [+FEM], [-PL] ↔ -it This means that it does not match one of the features of our given feature bundle, and thus it cannot be inserted as per the Subset Principle (Halle 1997). The [-PL] feature in the Vocabulary Item for -it is independently needed for the case of a root that typically takes -it but is not irregularly pluralized, e.g., t’ot’a ‘ape.’
A n feature bundle that contains [+PL] in the context of t’ot’a should not be capable of being realized and should crash the derivation.
But if this feature bundle also contains [+FEM], it is predicted that -it should be inserted and thus t’ot’it should be the feminine irregular plural of t’ot’a (it is not).
This scenario is prevented if the Vocabulary Item for -it has a [-PL] feature -- -it cannot be inserted in the first place since it does not match the features of the feature bundle.
Appendix 2: The Uninterpretability Puzzle It is often assumed that gender is part of the interpretable bundle of phi features on a DP (Chomsky 2000, 2001 and other minimalist work).23
But what about grammatical gender on feminine inanimates – that’s uninterpretable(not just in Amharic)! Option 1: Grammatical gender is actually interpretable (see e.g., Dowty and Jacobson 1988, but see also Pesetsky and Torrego 2007: fn. 31 and references therein).
However, the actual semantic interpretation of grammatical gender is unclear at best.
See also Legate 2002:2-3 for some criticism of this idea. Option 2: Grammatical gender is checked/deleted by entering into an Agree relation with an interpretable version of the same feature (Kihm 2005, Picallo 2008)
It remains unclear what the interpretation of that feature is (it is not natural gender) – somehow related to nominal classification?
Option 3: perhaps uninterpretable features do not cause a crash (Legate 2002, Pesetsky and Torrego 2007:fn.15, Epstein et al. 2010, Carstens 2011).
In Pesetsky and Torrego 2007:fn.15 and Legate 2002, unvalued features cause a crash, and uninterpretable features are simply ignored by the semantics.
Grammatical gender is uninterpretable but valued, so it would not cause a crash in this approach. Ruth Kramer / Georgetown University / 37th and O St., NW / Washington, DC 20057 / [email protected]
እግዜር ይስጥልኝ!
23 Chomsky does not mention gender directly, but defines phi features as agreement features (Chomsky 2000:101) and many languages have gender agreement.
