A NOTE ON THE LOCUS AND FUNCTION OF FORMAL GENDER

This note tries to assess the syntactic locus of grammatical Gender in the nominal structure with evidence mainly drawn from Spanish. Data from ellipsis resolution and its licensing in a variety of contexts, and for different types of syntactic objects, suggest the conjecture that noun classifying devices, such as grammatical Gender, may constitute the inflectional correlate of hierarchically organized, and possibly universal, nominality descriptors. Such descriptors would be coded in the structural space that, under different theoretical proposals, has been labelled as ‘lexical phase’, ‘first phase’ or ‘little-n’ space.


A linguistic imperfection?
This paper examines some aspects of nominal inflection.It focuses in particular on noun classification with evidence drawn mainly from Spanish where noun classification surfaces as formal Gender.Under a minimalist lens, this feature is a puzzling grammatical element because it seems uncongenial to the idea of optimal design.Unlike the so-called edge/EPP feature, formal Gender appears to be computationally inert as it does not block or trigger constituent displacement.Unlike structural Case, it does not render an argument active (or visible) for syntactic operations, and it appears to be parasitic on Number for Agreement/checking relations.In Spanish, Gender is also largely unpredictable on the basis of either interpretive or purely formal criteria (see Harris 1981;Roca 1989).The label 'imperfection', in the sense given in Chomsky (1995), appears to fit such a grammatical category.The present study constitutes an attempt to inquire the possible function of such apparent misfit from optimal design desiderata.
From the point of view of acquisition, an interesting aspect of grammatical Gender is its early learnability in Spanish, which is unexpected given its apparently low, or null, computational profile.The literature on acquisition shows that, despite its apparent lack of computational import, and its very limited semantic content, children are aware of Gender marking quite early in the acquisition stage.By age 2;7/3 on average (Lew-Williams & Fernald 2007), Spanish children know that grammatical Gender is a characteristic trait of nominals.Note that their awareness of the category cannot be confused with, and is independent of, some morphophonological errors they may make in its instantiation with respect to the normative, or adult-like form.Children do not simply take one of the two possible available forms and over-regularize its use for each and every noun.They neither omit grammatical Gender altogether by filtering it out of their production.Its formal relation, albeit limited, with the nouns for animates showing sexual bi-morphism does not foster its acquisition.Franceschina (2005) points out that one clear trend that emerges cross-linguistically is that children do not begin by analyzing the set of sex-related Gender marking and proceed to extend their command to inanimates.Sex-related Gender is not acquired earlier than the purely formal one and, moreover, children ignore this extra-linguistic information, paying more attention instead to syntactic and morpho-phonological cues (Pérez-Pereira 1991).
In the discussion that follows, I examine some syntactic evidence to assess the syntactic locus of Gender features in nominal structures, and conjecture that noun classification simply externalizes, in the inflectional domain of nominals, some basic properties of the linguistic system.I motivate my conclusions on the basis of empirical evidence and recent theoretical proposals that argue for the adoption of a much more abstract conception of syntactic constructs than those we have generally been considering.

Gender marking in Spanish. A short survey.
As is well known all nouns are assigned one of two possible formal Genders in Spanish, either the so-called 'masculine' or the 'feminine', which are more or less evenly distributed within the set of nouns (Fuchs et al. 2015).One of the traditionally assumed functions of Gender, or noun class, is that of concord controller (see Hockett 1958;Corbett 1991;Corbett & Fraser 2000).Together with grammatical Number, concord is reflected in the categories syntactically related to the noun, namely determiners, adjectives and some verb forms, as well as anaphoric or bound interpreted pronouns.Consider example (1) in Spanish:1 (1) Todas las puertas traseras fueron decapadas para poderlas pintar all F.PL the F.PL doors F.PL back F,PL were scraped F.PL to can-them F.PL paint 'All the back doors were scraped in order to be able to paint them' Grammatical Number, the other inflectional element involved in the alliteration of concord in (1) is a more common and widespread category than Gender -or noun class-in the languages of the world, although natural languages use only a few of the possibilities that could logically be used to express Number.The grammatical expression of values for grammatical Number is quite limited, ranging from just the two values singular/plural (as in Spanish or English) up to five (according to Corbett 2006).The possible ways to grammatically express Number can cross-linguistically be formalized by different combinations of features.Harley & Ritter (2002) for example, have proposed that the possibilities can be captured by combining the features [plural], [augmented] and [group].Be as it may in the case of Number, a cross-linguistic formalization in terms of abstract features is not straightforwardly conceivable in the case of Gender or noun classification systems, for various reasons.A significant percentage of languages of the world do not classify nouns into grammatical types at all, and among the 'noun classifying/gendered languages' there is a great deal of variation in the number of classes or types into which languages may classify nouns, ranging from a minimum of two to a dozen or more (Corbett 1991(Corbett , 2011)).
Formal Gender surfacing with /-o/ /-a/ terminations can also be related to a non-linguistic property or the entity, such as the natural sex of the named individuals (ex.gato/gata 'cat M/F '; juez/jueza 'judge M/F '; león/leona 'lion/lioness').However, the relation between this biological trait and formal Gender is not systematic.In a number of cases, nouns for sexed individuals are grammatically either masculine or feminine, as in the so-called epicene nouns: ballena 'whale', criatura 'creature/baby', hiena 'hyena', jirafa 'giraffe', oveja 'lamb', pantera 'panther', or víctima 'victim', among many others, are grammatically feminine, whereas hipopótamo 'hipopotamus', orangután 'orangutan', topo 'mole', tiburón 'shark', avestruz 'ostrich', among many others, are masculine.Similar examples of Gender marking idiosyncrasies can be provided for many other languages having formal Gender systems.Without unnecessarily extending the lists or regularities and exceptionalities, it can be said that grammatical Gender is in large measure unpredictable on the basis of interpretive criteria.As a result, in a large number of cases the grammatical class that a noun belongs to has to be inferred either by morpho-phonological cues or should simply be stored in memory.It may be facilitated in the acquisition stage by cues provided by determiners (la testuz 'the F forehead F (of an animal)' / el avestruz 'the M ostrich M '), diminutive or augmentative morphemes (valle M 'valley'/vallecito 'valley-DIMIN.';calle F 'street'/callecita 'street-DIMIN'; mano F 'hand' / manaza 'hand-AUGM'), or by modifying adjectives (diente blanco 'white M tooth M ' / frente ancha 'wide F forehead F ').
This brief presentation of idiosyncrasies raises questions of how a theory of grammar may account for such absence of systematicity and, in more general terms, how to account for the pervasive variation in noun classifying devices.Either a language may lack such a device (Basque, for example), or nominal classification may surface in a variety of forms.They range from either a purely formal system, as in Spanish, to systems that sort into several domains comprising the types of entities nouns may denote (expressing biological sex or animacy, as well as being a vegetable, a deity, an animal, an object made with a specific component, an atmospheric event or an edible item, among other classifications that, for a number of possible reasons, may historically have been, or still are, relevant to a community of speakers. 3The overarching question is why languages use a grammatical device such as nominal classification, independently of whether or not it can be realized with such a variety of means.Within a Minimalist approach to grammar, there seem to be two major approaches to account for how variation phenomena should be conceived, leading -roughly speaking-to two major research strategies and procedures.One line adopts the view that Principles of UG allow for spaces of variation by leaving possible options for their realization.Variation is encoded in the properties of features hosted in a universally fixed structure of functional projections.Under that approach, what we may call a Gender/Noun class parameter could perhaps be posited, together with the boundaries within which it may vary.A second approach takes the view that the principles of grammar are not only few but also invariant (that is, not parameterized).Under this second view, language diversity is the result of different externalization paths or realizations within the core components of the system.I will consider grammatical Gender, or noun classification systems, under this second approach and conjecture that noun classifying devices, together with grammatical Number, possibly are the inflectional correlates of hierarchically organized nominality descriptors coded in the structural space that, according to different theoretical approaches, has been labelled as 'lexical phase', 'first phase', or 'little-n' space.My tenet that the expression of formal Gender is inflectional, as grammatical Number is, appears to be at odds with some empirical facts brought up by recent approaches to the lexicon.The following section addresses that issue.

Gender as a nominalizer
Constructionist approaches have proposed an analytic view of the lexicon by arguing that syntax operates on sub-word elements (Halle & Marantz 1993 et seq.).Under the Distributive Morphology (DM) approach, some studies have suggested relating formal Gender to the requirement of having to assign a categorial label to an otherwise acategorial root.According to this general proposal, categorial labels have the function of assigning an interpretation to the conceptual space encoded by a bare root, which is an element devoid of grammatical properties.Roots are assumed to be unmanageable by the syntactic system unless they are grammaticalized or lexicalized by a categorizer that imposes an interpretive perspective on them.Under that assumption, Gender has been assigned the role of nominalizer, that is, the main constituent feature, or property, of a so-called 'little-n'.Glossing over different implementations, what we may call 'the categorizer hypothesis' can be represented as in (3).It recasts under a DM approach the traditional tenet that Gender is an inherent nominal feature, without considering it a property of roots or of noun stems (see, among others, Ferrari-Bridgers (2005, 2008); Alexiadou et al. 2007;Lowenstamm 2008;Acquaviva 2009;Kramer 2014;Merchant 2014;Saab (2010, to appear) and Panagiotidis 2015 for different implementations of that general hypothesis). ( This proposal has been adopted to explain the different behavior of Gender and Number in ellipsis contexts.It is well known that nominal ellipsis allows Number mismatches (see Leonetti 1999;Depiante & Masullo 2001;Brucart 2004;Eguren 2010;Nunes & Zocca 2010;Saab 2010;Gallego 2011;Merchant 2014, among others).The examples in (4), where the deleted fragment is within brackets, show that Number mismatches are allowed under deletion: 4 (4) a. Los relojes analógicos y el <reloj> digital the clocks analogical and the <clock> digital 'The analogical clocks and the digital (one)' b.Una lámpara de madera y tres <lámparas> de metal a lamp of wood and three <lamps> of metal 'A wooden lamp and tree metal (ones)' c.María admira a su hermana pero Pedro envidia a las <hermanas> Maria admires to POSS SING sister but Pedro envies to the PLUR <sisters> suyas POSS PLUR 'Maria admires her sister but Pedro envies his (sisters)' Exponents of grammatical Number behave like those of Tense and Mood in some deletion contexts in the sense that the deleted material may not inflectionally coincide with that of its phonologically specified antecedent, as in the examples (5).The data observed in ( 4) and ( 5) follow the generalization that ellipsis resolution appears to be blind to inflectional morphology (Lasnik 1995 'Juan admires his brother but Pedro envies his <sister> ' c. (*)?El gato de pelo largo y la <gata> de pelo corto estaban the MASC cat MASC of hair long and the FEM <cat> of hair short were en la azotea on the roof 'the long-haired cat MASC and the short-haired <cat.FEM> were on the roof' 5 An anonymous reviewer points out that speakers marginally favor the alternation abstractly represented in (i a), and exemplified in (6 b, c), over the alternation (i b), exemplified in (6 a, d).The difference in acceptability is probably related to the grammatical fact that masculine is the unmarked Gender in Spanish: (i) a.Under the assumptions represented in the schema (3) for nominal constructions, Gender is not purely inflectional but a constitutive feature of the 'little-n' category.Any element, such as grammatical Gender, contained in this syntactic space cannot undergo ellipsis if strict identity for the value of Gender is not respected, as is abstractly represented in ( 7): The representation depicted above, with its corresponding assumptions, descriptively covers the contrasts observed in ( 4) and ( 6).Some proponents of that approach include the additional requirement that ellipsis is licensed by a diacritic syntactic feature [E] in the functional head [Num] selecting the 'little-n' space (see Merchant 2014;Saab (2010, to appear)). 6Some analyses of constructions containing bare count nouns, like the one exemplified in (8), put into question whether or not these elements are licensed for ellipsis under the diacritic [E] feature hypothesis just mentioned. (8) Necesitas bolígrafo need.2.SG ballpen Bare count nouns appear in the unmarked singular and do not denote individual objects but name kinds -or types-of entities.To denote tokens (or instantiations) of a type, the syntactic category minimally hosting a bare noun merges with Number and with a Determiner, phonologically null or overt, among other possible functional elements. 7Bare count nominals like the one exemplified in (8) are not licensed by a null Determiner, have no quantificational force, and are not interpreted as singular because they are underspecified for Number.They are subject to very strict distributional conditions in Spanish as they can only be in object position of a restricted class of categories: that of intensional verbs of the types necesitar 'to need', buscar 'to look for', desear 'to desire', ofrecer 'to offer', poseer 'to possess', utilizar 'to use', pedir 'to ask for', contratar 'to contract' or encontrar 'to find'; the object of light verbs like tener 'to have', llevar 'to wear', hacer 'to do' or dar 'to give'; and the complement of a handful of prepositions (en 'in', a 'to', sin 'without' or con 'with', among some others).Examples of such a distribution are in (9 a-c): 8 6 Such contextual formulation of the ellipsis operation is reminiscent of the notion of filter proposed in Chomsky & Lasnik 1977.The assumption that there is diacritic syntactic feature in Num licensing ellipsis departs from accounts treating this phenomenon as an interface (PF) effect triggered by deaccenting (Tancredi 1992;Chomsky & Lasnik 1993), or from a sidewardlike copy and deletion mechanism (Ticio 2016). 7See Longobardi 1994;Szabolcsi 1994;Cinque 2005, among others. 8See, for references and discussion, the papers in Bosque (ed) 1996.If bare nouns truly lack an inflectional tier, the syntactic feature [E] that purportedly licenses ellipsis and assumed to be in the functional head [Num] cannot be selected.Even cases like (11) -with no Gender alternation presentshould arguably be disallowed.They are, however, grammatical:9 (i) a. Quiere llevar maleta wants 3.SING to carry suitcase 'S/he wants to be a suitcase carrier' b.No tenemos coche not have 1.PLUR car 'We aren't car owners' c.Utilizan pincel repetidamente/muchas veces (para restaurar este tipo de objetos) use 3.PLUR brush repeatedly /many times (to restore these types of objects) 'They are recurrent brush users (to restore these types of objects)' Moreover, and unlike the bare mass nouns and the bare plurals examined in Campos 1986, the nominal gap can appear inside an island such as a complex NP or a sentential subject, as shown in (ii a, b) respectively: (ii) a. María decía que Juan aún busca piso mientras Ana hacía correr el rumor de M. was saying that J. still looks for appartment while A. was spreading the rumor (of que ya tiene <piso> that) already has b.Juan presume de no necesitar corbata pero que nunca lleve <corbata> cuesta de creer J. boasts about not needing tie but that never wear 3.SING is hard to believe We can additionally add that bare count nouns cannot undergo processes of backwards anaphora, which are only possible with grammatically referential expressions (see Vergnaud & Zubizarreta (1992) and references cited there): (11) a.No llevas corbata, pero tampoco necesitas <corbata> (para asistir a not wear 2.SING tie, but neither need 2.SING <tie> (to go to la ópera) the opera) 'You are not wearing (a) tie, but (you) neither need (one) to go to the opera' b.Buscan piso aunque ya tienen <piso> seek 3.PLUR apartment although already have 1.PLUR <apartment> '(They) are looking for (an) apartment although (they) already have (one)' c.Juan busca compañera pero no encontrará <compañera> porque Juan looks for companion but not will find 3.SING <companion> because es un pelmazo is 3.SING a bore 'Juan is looking for (a) companion but (he) will not find (one) because (he) is a bore' Bare noun ellipsis may be licensed by a variety of factors.Some researchers have related it to interpretive requirements on discourse linking, which can be triggered either by adverbs highlighting contrastivity or by application of a notion of "extended partitivity" (Eguren 2010). 10Independently of these types of constructions and the conditions for their licensing, other questions remain related to how one would account for a number of idiosyncrasies in the expression of Number, as well as some other types of Gender mismatches in ellipsis contexts.I will consider each of these issues in turn.

Inherent/lexical Number
As is well known, the morphological realization of Number does not always correspond either to the optional [±plural] realization of countable nouns or to the 'numeral classifier' reading of pluralized mass nouns (such as three wines or several coffees).Consider the cases of singularia tantum in the list (12) or pluralia tantum in (13 a, b): (12) caos, grima, sed, tez, salud, basura, a.o.'chaos ', 'disgust'/'shivers', 'thirst', 'complexion', 'health', '  The clitic en/ne pronominalizing a bare noun is invariable, that is uninflected for Gender and Number.Note however that a bare count noun must be realized in one of the two possible Gender forms (i.e.corbata 'tie FEM ' , pis 'appartment MASC ', companya 'companion FEM ' in (i a-c).Under the hypothesis being examined, the obligatory expression of Gender in these types of bare count nouns can be attributed to a morphological well-formedness condition blocking bare stems when vocabulary items are inserted (ex.*/corbat-/).
(13) a. albricias, antípodas, bártulos, celos, modales, víveres, a.o.'congratulations ','opposites', 'gear', 'jealousy', manners','provisions' b. tijeras, gafas, pantalones, tenazas, ojeras, a.o 'scisors', 'glasses', 'trousers', 'pliers', 'bags (under someone's eyes)' The nouns in (12) can be described as defective for Number just as one could arguably describe epicene nouns like jirafa 'giraffe FEM ' or hipopótamo 'hippopotamus MASC ' as being 'masculine defective' or 'feminine defective' respectively.Singularia or pluralia tantum nouns do not form a natural class, as they display a variety of denotations: among other things, they may refer to collectives, groups of entities lacking discernable boundaries, or items whose integrating parts have either some sort of internal cohesion or form a unique object (Acquaviva 2008).Whatever their denotations, Number cannot be considered purely inflectional in these cases, but must be seen rather as 'lexical' or inherent in a similar way to grammatical Gender.Hence, Number is also a good candidate for the 'little-n area' in these cases, and we could assume that inherent (or 'lexical') Number is an abstract independent category in the little-n space, as represented in ( 14).11I have labelled it 'NUMBER' (small capitals) to facilitate the discussion at this point.
This analysis allows us to compare certain cases of Number mismatch with those of Gender under the 'little-n' hypothesis we are still considering.In Spanish, some pluralia tantum expressions of the types (13 b) that denote objects composed of integrating parts such as tijeras 'scisors' or pantalones 'trousers', among a few others, have a singular version (ex. tijera and pantalón, lit. 'scisor' and 'trouser' respectively).Both versions of the noun do not induce any change in meaning with respect to Number, and one or other may be used according to the dialect, the idiolect, or the register used by the speaker.Ellipsis with alternating Number may show some degrees of unacceptability: (15) a.We could assume that lexical/inherent Number may have a matching correlate in the inflectional domain, as represented in ( 16): The hypothesis that an abstract feature, or an abstract projection corresponding to lexical or inherent Number -in the 'little-n area'-has to be related to an inflectional correlate has already been suggested in Acquaviva 2008 for singularia and pluralia tantum cases exemplified in ( 12) and ( 13 According to Aquaviva 2008, cases of these types conform to a denotation somewhat similar to the inherent or 'defective' Number exemplified above, as the expressions invoke collections, accumulations, aspectual readings, or groups of entities whose occurrence or appearance is indistinguishable.He notes that the alternation does not change the referential properties of these nouns or their countability (because they are all mass nouns), and suggests that Number can be realized in two structural positions.One of them would correspond to the inflectional area and the other would be realized as a constitutive element of the 'little-n' space, as we have informally represented in (16).One can see now that Number-alternating mass nouns exhibit the unacceptable 'Gender behavior' under ellipsis observed in (6), as those nouns do not allow deletion under alternation, as shown in ( 19).It must be pointed out that such ellipsis constructions have a lesser degree of unacceptability as compared to the types of Gender alternations exemplified in (6), which are considered generally worse that the alternations in (15).It is possible that the difference in acceptability is related to the experimental evidence showing that that concordmatching violations in Gender appear to be more disruptive or conspicuous than those involving Number:15 (19) a. ??(*) El agua del río y las <aguas> del mar the SING water of the river and the PLUR <waters> of the sea b. ??(*) Las nieves del Kilimanjaro y la <nieve> de Baqueira the PLUR snows of the Kilimanjaro and the SING <snow> of Baqueira c. ??(*) Les gusta la lluvia de Bilbao pero no las <lluvias> de Ranchipur (they) love the SING rain of Bilbao but not the PLUR <rains> of Ranchipur d. ??(*) Como abono, utilizaban las cenizas volcánicas y la as fertilizer,(they)usedthe PLUR ashes volcanic PLUR and the SING <ceniza> vegetal <ash> vegetable SING e. ??(*) No puedes dormir con este calor pero puedes not can 2.SING sleep with this SING heat SING but can 2.SING soportar los <calores> del verano cope with the PLUR <heats> of the summer The above examples show that Number can behave in a similar fashion to Gender in some types of ellipsis constructions.Acquaviva's 2008 assumption that the lexical/inherent Number of singularia and pluralia tantum and Number alternating mass nouns must have a correlate in the inflectional area, as suggested in ( 16), may suggest a partial explanation for the behavior of singular collectives of the types committee, couple, faculty, family, or team, among others, that are known to show quirky Number behavior with verbal concord, pronouns anaphorically interpreted, and distributive anaphors: 16(20) La familia ha/han expresado su desacuerdo y no firmarán the family has/have expressed POSS (its/their) disagreement and not will 3.PLUR sign el documento the document (21) (?) El equipo se estaba/estaban insultando unos a otros the team was/were insulting each other A detailed analysis for these constructions is beyond the scope of this note.17One may speculate however that the alternating verbal concord and the plural anaphora effects exemplified above may be understood under the distributed Number analysis ( 16).These types of collective expressions (or 'pluringulars') show a formal singular in the functional/inflectional projections superordinate to the n category (cf.*Estos/los comité 'these/the PLUR committee') but they appear to contain a semantic plural in the 'lexical' or 'first phase' projection.The verb, or the anaphoric expressions related to them, may be sensitive to such interpreted (or semantic) plurality.
The preceding discussion suggests that there would be no inconsistency in assuming a similar 'split' analysis for Gender/noun class, as in ( 22), where lexical projections correlate with independent inflectional exponents.The quoted "GENDER" and "NUMBER" in the little-n area are for expository purposes so far: The assumption that the locus of formal Gender/noun class and grammatical Number is in the inflectional area has already been proposed in Picallo (1991Picallo ( , 2008)); Bernstein 1993;Fábregas & Pérez-Jiménez 2010.Evidence that both Gender and Number must be constituent elements of the inflectional area in Romance is provided by the behavior of pronouns in deletion contexts, to which I turn in the next section.

Pronouns
Pronominal categories have been assumed to be complex structures that display (in full or in part) the functional architecture of nominals (Postal 1969;Abney 1987;Cardinaletti & Starke 1999;Déchaine & Wiltschko 2002, among others).They may be conceived as syntactically rootless elements formed by hierarchically organized functional projections for Person, Number and Gender (or a subset of these).Their value for Gender in particular may induce presuppositions about their possible denotations.For pronominal anaphora, we should assume that the denotation of pronouns is resolved in concord with the inflectional -in other words, formal-content of their assumed antecedents.Pronouns are blind to the possible lexical idiosyncrasies of the expressions naming the entities they denote.They appear to be only sensitive to the specification of values expressed in the functional layers of their intended co-referring nominals, independently of whether those expressions denote animates or inanimates (23 a, b) A mismatch in the feature values of the pronouns and those of the inflectional content of their co-referring antecedents leads to ungrammaticality, if anaphora is intended, or to a free interpretation of the pronoun in the same contexts.The indices in the unmatching pronouns exemplify this state of affairs: (28 a) shows a mismatch in Gender in a nominal that can alternate for this feature (i.e., el/la mar) and (28 b) to a mismatch in Number in mass nouns that can also show alternation, such as ceniza/cenizas 'ash/ashes': (28) a.El estado del mar i es peligroso y se recomienda no the state of the MASC sea is dangerous and it is recommended not entrar en ella *i/j to enter in PRON FEM b.Recogí la ceniza i cuando vi que ella no las *i/j (I) picked up the FEM,SING ash FEM,SING when (I) saw that she not PRON FEM,PLUR quería recoger wanted to pick up Summarizing, anaphorically interpreted pronouns appear to only be sensitive to the inflectional or functional phase of their co-referring nominals. 19Such a syntactic space includes Person, but also Gender and Number.If -by assumption-pronouns are purely inflectional elements, anaphoric pronouns should allow Gender mismatches in ellipsis contexts, because only inflectional categories allow it.This is indeed the case.Consider (29) where two anaphoric interpretations for the pronoun in the deleted fragment are allowed: the interpretation corresponding to the local (or sloppy) reading and the strict (or non-local) one, as shown in the reconstruction of the deleted fragments (i) and (ii).( 29 Regardless of whether it is the phonologically realized one or the one deleted with the VP, the anaphorically interpreted pronoun remits to two entities (animate or inanimate) that are salient in the linguistic environment, and the antecedence condition on ellipsis requires the pronoun to match the Gender content of its short-or long-distance intended co-referring expression.The interpretive ambiguity is possible if both formal Gender as well as formal Number are constitutive elements in the inflectional layers, which is what an anaphoric pronoun is sensitive to.Gender alternation under ellipsis also obtains in the much stricter condition of binding.Pronouns bound to quantified expressions can also display alternations under ellipsis, as we see in (30 a, b): (30) a. Bound pronouns only allow the sloppy reading represented within the brackets because they are variables and can't be bound outside their own sentence.The constituent values of a bound pronoun may be different from that of a coreferring one, although their form may morpho-phonologically be the same.Heim 2008 andKratzer 2009 have suggested that bound pronouns (i.e., pronominal variables) reflect the inflectional content of the quantified expressions binding them, and such inflectional content is what determines the form the related pronouns take.
The ungrammaticalities in ( 6) and ( 19) can now be assessed under a different perspective.They are ungrammatical because they show Gender and Number alternations, but such alternations may overtly manifest different combinations of notions coded in the most embedded (lexical) area with which the functional elements Gender and Number relate.We have already shown that alternations under deletion are possible when they are purely inflectional.Alternations are not possible when they externalize lexical idiosyncrasies.Next section will consider some conjectures on the possible components of the so-called 'little-n' space alluded to in the representation (22).

On the lexical content of nominals
Following some insights in the proposal put forth in Ramchand (2001Ramchand ( , 2008)), I suggest that the lexical content of nominal constructions that I have been referring to as 'little-n', following DM terminology, should be built up with recursively embedded conceptual categories that express structured facets or nominal meaning.These could include categories related to sortality, meronomy, cohesion, boundedness, unity, identity, or others that may characterize nominal denotations.Such notions should be universal and possibly rooted in more general cognitive processes related to the perception and conception of entities, as well as general reasoning about the physical world, which infants are known to perform at a very young age independently of whether or not they already recognize particular objects as entities of specific kinds (see Spelke 1990).The structure and the categorial/conceptual content of the lexical projections corresponding to nouns or verbs allude to either static entities or to dynamic states of affairs respectively.Note that being compositionally and conceptually different, the theory of grammar does not necessarily have to posit 'nominalizer' or 'verbalizer' elements (i.e. 'litte-n' or 'little-v'), in particular if they are devoid of other content than categorial specification.
The notions that compose the lexical projection of nouns may overtly be 'externalized' in the form of correlate categories in the functional domain, which may encyclopedically surface in a variety of forms.The set of superordinate inflectional elements (or extended projections) that dominate the lexical (or 'first phase') space should not be necessarily universal, and languages may vary with respect to which functional projections they select and the overt forms they may take.Basque, for example, does not select a formal Gender/noun class inflectional category but Spanish does and can also recruit it to express mass/count distinctions, and non-linguistic perceptual properties of entities such as size, collective readings or sexual bi-morphism (see section 2). 20The absence of realization of functional correlates in the inflectional space does not imply absence of the conceptual/interpretive categories on which nominal denotations are built because inflection and the abstract conceptual notions that it may make overt should be independent.Such independence is suggested by Li et al. 2009, who have shown that young infants are aware of the conceptual difference related to the singular/plural distinction, irrespectively of whether their native language morphologically expresses grammatical Number, or it does not.I suggest that the same independence may apply to Gender/noun class categories.
Mientras María admiraba a las actrices cómicas Pedro alababa a while Maria admired to the FEM actresses comic FEM.PLUR Pedro praised to los <actores> trágicos the MASC.PLUR tragic MASC.PLUR 'While Maria admired the comic actresses Pedro praised the tragic <actors> ' (*)? ... [... ANTECEDENT MASC ] ... [ ... <deletion FEM > ...] b. * ... [... ANTECEDENT FEM ] ... [ ... <deletion MASC > ...] d. * garbage' (??)? Sus pantalones están limpios pero el <pantalón> tuyo está POSS.3.PLUR trousers .are clean PLUR but the SING yours SING is arrugado y sucio 12 wrinkled SING and dirty SING 'His/her trousers are clean but yours is wrinkled and dirty'.b. (*)?? Saca esta tijera de la mesa y pon estas <tijeras> en el cajón remove 2.SING this scissor from the table and put 2.SING these in the drawer ), as well as for a handful of mass nouns that can display Number alternations.A list of such items in Spanish is provided in (17).Similar examples hold for other languages (see, for example, Tsoulas 2009 for Greek).The constructions in (18) from different languages -and taken from various sources-exemplify the use of the plural version of such mass nouns.Dans nos obscurités, allume le feu qui ne s'éteint jamais [Com.Taizé] (with)in POSS 1.PLUR obscurities light up the fire that extinguishes never b.SP.Las cenizas de Gabo se repartieron entre México y Colombia [RPP.4/4/2014] the PLUR ashes of Gabo were divided between Mexico and Colombia Los humos de soldadura son una mezcla de partículas y gases 13 the PLUR smokes of welding are a mixture of particles and gases c.CAT.No s'ha menjat les sopes del plat (s/he) has not eaten the PLUR soups of the dish , count or mass items (24 a, b), or have nominal antecedents with lexical/inherent number, as shown in (25 a, b): PRON MASC,SING submerge in water FEM hot FEM , the MASC sugar MASC PRON FEM,PLUR submerge in water FEM hot FEM , the FEM,PLUR chaquetas se encogen jacquets FEM,PLUR SE shrink 'If you submerge them in hot water, the jacquets shrink' PLUR.MASC gear PLUR.TANTUM.MASC and load-PRON MASC,PLUR in the truck b.Toma la basura y cárgala en el camión take the SING,FEM garbage SING.TANTUM.FEM and load-PRON FEM, SING in the truck It was shown in (17) that some mass nouns can alternate in Number.A few others can also vary with regard to Gender assignment (ex.el/la mar 'the masc/fem sea').18Inboth cases, the anaphoric pronouns related to them must match the inflectional layer of the expressions denoting such entities.he) picked up the FEM,SING ash FEM,SING to reuse-PRON FEM,SING as fertilizer b.Recogió las cenizas para reutilizarlas como abono (s/he) picked up the FEM,PLUR ashes FEM,PLUR to reuse-PRON FEM,PLUR as fertilizer Cualquier hombre quiere conocer a alguien que lo quiera y cualquier mujer también <quiere conocer a alguien que la quiera> 'Any man wants to know somebody who loves him and any woman too <wants to know somebody who loves her> ' b.Cada libro tiene un número que lo identifica y cada ficha también <tiene un número que la identifica> 'Each book MASC has a number that identifies it MASC and each card FEM too <has a number that identifies it-FEM>