Construcciones lingüisticas relativas tempranas.pdf

8/12/2019 Construcciones lingisticas relativas tempranas.pdf

1/35

STARTING SMALL EFFECTS IN THE ACQUISITION OF EARLYRELATIVE CONSTRUCTIONS IN SPANISH

Cecilia Rojas-Nieto

Instituto de Investigaciones Filolgicas,Universidad Nacional Autnoma de Mxico

Circuito Mario de la Cueva s/nCiudad Universitaria 04510

Mxico, DF, Mxico

[email protected]


2/35

AbstractThis paper builds on previous Usage-based accounts of developing sentence complexity

(Diessel 2004; Diessel & Tomasello 2000, 2001), considering early relative constructions(RC) in Spanish. RCs development shows various starting small processes (Elman,1993): Most CRs show no embedding; they are dialogical co-constructional results or takean absolute position and not intonation integration to any verbal frame. When embedded,constructional frames are lexically biased with an open slot for Head RC insertion. CRsinternal structure is mostly similar to independent clause type, with no gap nor genuinerelative function for the relative pronoun. In sum, CRs show an exemplar basedacquisition and individually preferred constructional frames. All these phenomena pointtowards a non linear, experience based learning, affected by frequency and oriented byfunction.


3/35

STARTING SMALL EFFECTS IN THE ACQUISITION OF EARLYRELATIVE CONSTRUCTIONS IN SPANISH

1. PresentationSpanish relative constructions are basically produced with a relative pronoun strategy(RCs), and are unanimously recognized as complex structures. 1 RCs are acknowledged tocombine in a reduced constructional space a set of operations: embedding; headdependence, and possible head anaphoric marking; focusing of an internal constituent;coreference calibration of the focused constituent; coreference marking by means of arelative pronoun (REL) in an initial position, keeping internally a constituent gap (S and Orelatives) or a resumptive pronoun (RES) (f.i., IO relatives) (Brucart, 1999). According tothis set of properties, shared across languages, relative constructions have been consideredto be a late developmental achievement (Echeverra, 1978; Hurtado, 1984).

Syntactic renditions of RC acquisition tend to propose a default entrance, selected among possible options in terms of simplicity, markedness and lower processing cost, togetherwith a linear developmental path, going from simpler to complex, unmarked to marked,accessible to inaccessible. So, RC studies have based their developmental hypotheses uponthe relativization accessibility hierarchy: subject > object > oblique > possessive (Keenan& Comrie 1977; Barriga 2002); a processing motivated preference (cf. Prideaux & Baker1986) to have the same function in the antecedent NP and REL: S[S] and O[O], rather thanS[O], or O[S] (Echeverra 1978; Hurtado 1984; Sicuro-Corra 1995). A simpler syntacticstructure, with adjoined (Hale 1976) or conjoined types of RC structures, is supposed to be

preferred over an embedded one (Tavakolian 1981). Lexically headed rather than

determiner headed RCs are considered as basic, since various models would takedeterminer headed RC to be textually dependent and/or anaphoric reductions of lexicallyheaded constructions (Bello 1847/1988: 323-325; Brucart 1999).

Despite the appealing and elegant predictions that might point towards a converging defaultas a starting point for children to develop RCs, it is a well known fact that childrens dataon various constructions and categories development have proved to be reluctant to showinitial defaults (Dabrowska 2001; Gathercole, Sebastin & Soto 1999; Rojas 2004), or tofollow abstract principles of grammatical models (Dabrowska & Lieven 2005; McClure,Pine & Liven 2006; Lieven, Pine & Baldwin 1997; Tomasello 2000; 2003; etc.).

This is one of the foundational points in recent studies on RCs development inspired byUsage based theory of language acquisition, where no RC default is even mentioned(Diessel 2004). 2 According to this view it is argued that children start to produce early RCs,not necessarily following the abstract predictions of syntax, but rather revealing the effectsof experienced familiar use from which children adopt selective and lexically specific

1 Gerundive constructions are heavily restricted and normatively stigmatized . 2 Cf. also, Diessel & Tomasello 2001; Diessel & Tomasello 2005; Kidd, Brandt, Lieven &Tomasello 2007; Tomasello 2003.


4/35

construction frames i.e., form-function pairs, which despite their apparent complexity,are in fact monopropositional in nature, and expose a unified communicative intent. In hisfundamental study, Diessel (2004) elaborates extensively this point. Early RCs are mainly

presentational constructions ( Here is a rabbit that Im patting : Diessel 2004: 3), whichDiessel considers to be syntactically simple despite their complex appearance, since they

correspond to a single assertion, and are dedicated to introducing new referents in discourse(Lambrecht 1988; Moreno Cabrera 1999).

The evidence there presented points towards one of the main proposals in Usage basedresearch: that early RCs are based upon concrete exemplars experienced by a child. Thisexperienced usage, by definition situated, particular and individual, is the rough materialfrom which the child extracts chunks and pieces not necessarily atomic, whose possibleformants are latter analysed and gradually organized, when every child finds internal

patterns and regularities, builds analogies and relations, and organizes a network amongthem (Tomasello 2003). We must be aware that these early selected exemplars neitherreproduce nor obliterate by definition the abstract regularities supposed to define linguisticfacts; but they perform-expose those regularities in a probabilistic way, with all thehaphazard and vagaries, also preferences and dominance, that real, situated, dialogicallyframed, concrete language use has in a particular ecological niche (Givn 2009).

In effect, as Givn (2009) emphasizes, early relative constructions present from the startvarious pragmatic functions related to reference building: reference introduction, referencetracking, and reference negotiacion. Formally RCs may present various ways to do thisfunctions, not necesarrilly embedded frames, nor only and strictly the existencial

presentative, but a bunch of sequencial, paratactic structures, constructions whosemotivation is mainly functional and whose various sources are present in parental use.

On the other hand, Diessels interpretative proposal that supposedly complex constructionslike the presentational ones (or for that matter, other constructions in the space of complexclause constructions, as Diessel argues), are in fact simpler, monopropositional ones, may

be considered as a starting small type of argument.

Effectively, evidence has been obtained in other developmental and problem-solvingspaces, that starting small may be a way to enter complex systems (Elman 1993; Newport1990; Seidenberg 1999). Under this view, complex tasks may not be detected as complex

but reanalyzed as simpler and, once so reduced, they may be solved by undemanding means(Newport 1990; Rojas submitted; Seidenberg 1999). This would be a plausible case forearly RCs in apparently complex frames that may be worked as simple ones.

Building upon some aspects of these proposals Usage-based, functional orientation, andstarting small, this study will undertake the analysis of RC development in Spanish.Following classical (Bowerman 1979; Braine 1976; Limber 1973), and recent childlanguage studies on clause combining (Diessel 2004; Diessel &Tomasello 2000 and 2001;Givn 2006 and 2009; Rojas submitted), we can argue that children have their own,concrete and simple way to enter complex constructions based on their individualexperience. This piece of research on RCs development would rely, hence, on theexpectations that, guided by functional motivations on reference building i) children willadopt particular frames with lexical specificities, not for their markedness or complexity


5/35

properties, but closely affected by experienced use, ii) with the effects of a percolation ofcomplexity through their own processing resources.

Being the focus of this work the emergent structure in children situated interaction, and theway children solve a supposed complex task without recruiting abstract, exotic or exuberant

types of proposals, the following research questions will guide this analysis: How complex are relative constructions in early language? Do we get evidence of the initial selection of less complex structures sort of

defaults? Can we trace back starting small effects in childrens data?

2. The dataThe data to be considered here comes from the corpus ETAL : Etapas tempranas en laadquisicin del lenguaje (Early stages in language acquisition), pertaining to the Institutode Investigaciones Filolgicas, at the Universidad Nacional Autnoma de Mxico (Rojas

2007). Attention will be directed to three subjects: two girls and one boy. They are allSpanish monolingual children of urban educated families. Age range considered goes fromlast observation (4;02 ~ 4;00) down to first two videos with no attested RC documentation(2;3 ~ 3;04). Usual criteria for data selection have been followed: only spontaneous child

produced constructions have been considered, and no successive reiterations have beencounted.

- Insert here Table 1. -

Analysis will consider only RCs marked by the relative pronoun que (1i-1v), which sum atotal of 312 tokens. Other RCs marked by quien , who, cuyo whose, are absent in

childrens data. The infrequent cases with other relative markers, like donde where (2a),and cuando when (2b) or cual which, with or without exposed head, have not beconsidered and were not counted in data presentation in table 1.

Que -relatives may be lexically headed by a noun phrase or a bare noun (1i ); they also takea pronoun (1ii-1iii) or a determiner (1iv-1v) as a head. These last ones determiner headedrelatives (DET REL) present a definite determiner marked for gender and number ( el, la,lo the m/f/n singular; los, las the m/f plural) plus a REL (1iv-1v). Various arguments aremade on DET being a derived head or anaphoric trace of the lexical omitted head (Brucart1999). Since this DET was historically a demonstrative, it is possible that it will keep someindexical force, and it need not by definition be anaphoric, but deictic. In our analysis and

for data presentation, we will take DET to be an anaphoric/indexical head. But its status inchildrens grammar will be kept as an open question needing further and specific research.

1) i. una casita que tiene mucho espacio one house-DIM REL have-PR.3S much rooma little house that has lot of space.

ii. ese que tienesthat REL have-PR.2Sthat one that you have.


6/35

iii. vi una, una que tena espumasee-PST.1S one one REL have-IMPF.3S foamI saw one, one that had foam.

iv) estamos viendo lo que salpica

be-PR.1P see-GER DET REL splash-PR.3Swe are looking what (=the that) splashes.

v. dame la que tiene puntitos give-IMP=IO.1S DET REL have-PR.3S spot-DIM-M-Pgive me the one that (lit. = the that ) has little spots.

(2) i. Al cuarto a donde se fue Kiso to=the room to where RFL.3S=go.PST.3S KisoTo the room where Kiso (dogs name) has gone to .

ii Un da cuando estaba pequea me pona

One day when be-IMPF.3S small O1S=put-IMPF.1SOne day when I was small I used to put on me (cream).

3. Analysis

3.1 Starting points: Default or CRs diversity?From the first RC documentations, it becomes clear that individual children show a varietyof RCs. Besides, each child may prefer different constructional frames for her earliestrelatives. Just taking in consideration the first five RC produced by each child we get ahandful of different exemplars exposing various constructional frames. a) Both: FN headedand DET-headed RCs (3i-3iv vs. 3v). b) RCs with no internal predication (3i) or with anovert internal predication (3ii-3v), which may be ritual or formulaic (vgr. encontrar find,for Julio: 4.i-4iv). c) The relative pronoun (REL) may have no internal function (3i), orhave a function lexically determined by the preferred/ritual predicate (f.i. encontrar finddetermines an Object REL in 4i-4iv). d) As for embedding, RCs may be in constructionwith a free head, or inserted in a preferred/ritual verb frame: like mira look (imperative)for Julio (4i-4iv), and ste es this/it is for Elia (5i-5iii). On the contrary, in Flors data,RCs mainly adhere to syntactically free NPs (3i-3iii).

Consider then the following first five exemplars from every child, which show all thesesyntactically variegated constructional exemplars.

3) F LOR (2;03- 2;04) (M = mother)i) M: qu es eso?

whats that?F: la cama que agua, que aqu

the bed REL water, REL aqu the bed that water, that ... here.

ii) M: qu estas viendo, Flor?what are you looking at, Flor?


7/35

F: el coche que maneja the car REL drive-PR.3S (=moves)the car that moves.

iii) F: ese seor que s tiene huevo that man REL AFF have-PR.3S egg

that man that really has (an) egg.

iv) F: esa es la ropa this be.PR.3S DET clothesthis is the clothes

M: cul? which one?

F: que de ese nene esREL of that baby be.PR.3Swhich , belongs to that baby.

v) F: ma e que se cay

look-IMP DET REL RFL.3S=fall.down-PST.3Slook the one which (lit. the that ) fell down.

4) J ULIO (3;01- 3;02)i) J: m(r)a que cont(r)

look-IMP REL find-PST.3Slook that (=what) he found.

ii) J: una, m(r)a, que cont(r)! one, look-IMP REL find-PST.1Sone, look that (=what) I found.

iii) J: m(r)a que cont(r), velo look-IMP REL find-PST.1S, see-IMP=O.3Slook that (=what) I found, see it.

iv) J: m(r)a la ata que cont(r) look-IMP the tire REL find-PST.1Slook the tire that I found.

v) J: m(r)a- que me qued look-IMP REL IO.1s=result-PST.1Slook that (=how ) it resulted to me.

5) E LIA (3;04)i) E: esta es la casa que estaba

this be.PR.3S the house REL be-IMPF.3Sthis is the house that was.

ii) E: esta es mi diadema que me trajo una amiga this be.PRS.3S my diademREL IO.1S=bring-PST.3S a friend'this is my diadem that a friend brough to me.

iii) E: tu pjaro es, las cosas horribles que no me gustan


8/35

your bird be.PR.3S the things horrible REL NEG IO.1S=please-PR.3Pyour bird is, the horrible things that do not please me.

iv) E: no te enseo ste, mi premio que me regal Tana NEG IO.2S=show-PR.1S this, my price REL IO.1S=give-PST.3S TanaI wont show you this one , my price that Tana gave to me

v) E: muerde it bites

M: a quin? whom?

E: a unas personas que estn llorando to some person-P REL be-PR.3P cry-GERto some persons that are crying.

Individual preferencesIn sum, every child shows a different profile for her earlier RCs: Flor and Elia prefer lexicalheads + RC. Julio selects no overt head with a dominant verb frame: mira look. Eliaadopts a different constructional frame: ste es this is NP REL, and also appends RCs tosyntactically free NPs. Flor clearly prefers RCs with free NPs, and only shows a singleframe with mira look (see Table 2 as a summary).

Insert here Table 2

These differences among early exemplars, which show every childs selection, are only to be expected from probabilistic recurrent encounters with family usage data, and ratify thatas for syntactic complexity, there is not an initial default with a less complex status: nounique entrance to RCs, but rather a handful of individual preferences, lexically driven and

partially ritual, which cut across the criteria credited to define a possible syntactic default.

Emergent regularities?The earlier RCs, which clearly expose individual differences and no default, lead us toexplore whether we could find latter some syntactic factors affecting childrens RCs

production, looking for possible structural patterns, which may emerge in the course ofdevelopment.

In order to test RCs development, and the emergent regularities they might present, we willexplore various syntactic variables that would define RCs complexity (Givn 1990):- RCs external syntax: In this section the syntactic freedom or integration of CRs will beconsidered.- RC internal syntax. Here we will focus on the adaptations presented by the RCs, in termsof the relative pronoun (REL) expected properties, to test its genuine pronominal characterand syntactic function, concurrent or not with a gap presence.

In every case, we will track whether or not the possible regularities have a lexical source asa foundation; we will pay attention to frequency and, if relevant, refer to developmentalchronology.


9/35

3.2 External syntax

Freedom and Embedding

Across the observed period and unexpectedly from any syntactic perspective, most RCs produced by our children do not clearly qualify as sentence embedded constructions (N 170= 55%). They appear as isolated fragments, in an absolute position with no overt syntacticdependence but REL fronting and sequencing. RCs are produced in an independentintonation unit (6i) and they may even occur in a different conversational turn, one turndistant from their possible head (6ii-6ii).

6) i. Julio (28) painting with watercolour markers which he has been asked to coverJ: ayer yo tap los otos,

yesterday I cover-PST.1S DET other-Pyesterday I put the lid to other ones que me comp mi pap*DET REL OI.1S=buy-PST.3S my Dadthe ones that my Dad bought for meayer, que me copyesterday REL OI.1S=buy-PST.3Syesterday, that he bought for me.

ii. Mother and Flor (31), engaged in book readingF: esa es la ropa

this be.PR.3S DET clothingthese is the clothing

M: cul? which one

F: que de ese nene esREL, of that baby be.PR.3Sthat, belongs to that baby.

iii. Flor (41) is asking the mother a particular toyF: juete nenes

toy babiesbabies toy

M: cul juguete de los nios?, cul ?which childrens toy? which one?

F: que se mueve REL RFL.3S=move-PR.3Sthat it moves.

This lack of integration of RC from any syntactic frame even a head, as in the previousexemplars may also involve [head RC], not just the RC. In these cases both head and RCkeep apart from any main clause. This is a normal and frequent case in conversation, where[head RC] gives an answer to a recurrent identification question: cul? which one? (7i-ii).


10/35

7) i. Julio (26) has demanded the aunt (A) some waterA: a ver ensame, cules vasos?

see, show me, which glasses?J: (l)os vasos que son (r)icos.

DET glasses REL be.PR.3P tastythe glasses that are tasty.

ii. Elia (25) wants a toy from the upper shelf, which Observer (O) tries to reach.E: la sirena

the mermaidO: ay! no alcanzo, cul sirena?

auch!, I cant reach up, what mermaid?E: la de la que tiene este... cola.

DET of DET REL have-PR.3S umh tailthe one with the one that has , uh tail

O: la que tiene cola, ah! the one that has tail, oh!

E: sa.

that one.

These and similar cases of [Head RC] with no syntactic integration can be argued to be aconversational result, prompted by the question asking for a referent identification: cul? which one? Such questions do settle a context for a free NP-RC to occur in anindependent conversational turn (Brucart, 1999).

But free [Head RCs] may also occur in the course of the childs own discourse, not onlyacross dialogue turns. Children built by themselves free [Head RCs], with noconversational support, when involved in description activities (8i-ii) or when quarrellingor negotiating on reference (8iii), in a sort of referential competition situation (Givn,

2009).

8) i. Elia (27), asking for a boy she had seen at Cs homeE: y y el hijo que tenas ?

and and the son REL have-IMPF.2S?and what about the son you had?

ii. Flor (37) in a book looking activity, considering an image.F: un sodo que (es)t mu(y) bonito

a fox REL be-PR.3S very nicea fox that is very nice.

iii. Flor (35) discusses with Mom about the brush to be used for cumbing.F: no es mi cepillo

its not my brushete mi, ete mi cepillo, es mi ce- ete es mi cepillothis is my, this is my brush is my br- this is my brush

M: me dejas peinarte? will you let me to comb you?

F: este que no es ese cepillothis REL NEG be-PR.3S that brushthis (one) that is not that brush.


11/35

Various discourse situations ask for elaborated reference. They are the ecological niche ofnoun phrases with an RC expansion, which add to bare heads a specific information thatcharacterize or help to identify the item the head refers to. These complex noun phrases donot need to be integrated to any predicate to do this task.

However, adding to the point that free RCs or [Head-RC] are not exclusively a dialogicalco-constructional result, lets consider more elaborated sequences, where children keep

both the lexical head and the RC joined together, but with an intonation break parting themfrom the previous and supposedly main clause (9i-v).

Among them, the clearer cases have both, an intonation break and a functional closure(CL): a pronoun or a noun phrase which takes in the main clause the position whichotherwise [Head RC] might have taken. Then, after the intonation break, the phrasefunctioning as CL is reformulated and elaborated by means of [Head-RC], as an appositiveclarification (9i-iii) cf. similar previous cases in (1iii) and (5iv), not repeated here foreconomy. These constructions overtly expose and mark that [Head + RCs] areindependent constructions, not integrated to any previous main clause, but sort of antitopic

phrases, which months later will occur preposed, with similar marking conditions (i.e. a pause and a closure) in a topic position (9iv-9v).

9) i. Elia (23)E: dame ese , ese cuado, ese que lo tengo , que lo tengues

give-IMP=OI.1S that.CL that square that REL O.3S=have-PR.1S REL O.3S=have-PR.2Sgive me that one, that square, that one that I have, that you have.

ii. Flor (34) with Granny (G)

G: voy a sacar una ropa que dej en la lavadoraIm going to take out some clothes that I left in the washing machineF: me quiedo tae la (=traerla)

IO.1S=want-PR.1S bring-INF=O.3FS.CLI want to bring themla la la opa que que dejates en la bebaloda.the the the clothes REL REL leave-PST.2S in the washing machinethe clothes that you left in the washing machine.

iii. Julio (25) and his aunt (A) are looking some imagesJ: aqu, la eata (=piata) (pointing to a stick)

here, the piata

A: qu?what?J: a, eso, lo que tiene .

look-IMP that DET REL have-PR.3Slook, that, what (lit. the that) he holds

A: un palo.a stick.

iv. Flor (52)la que me la dio mam , ahora la voy a poner


12/35

DET REL IO.1S=O.3S=give.PST.3S Mum now O3S=go-PR.1S to put-INFthe one Mum gave it to me, now I will put it.

v. Flor (64) talking about putting music in the tape-recorderuna que t no te sabes, pnme laone REL you NEG RFL.2S=know-PR.2S put-IMP=O.3S

one you dont know, put it for me.

To consider the overall presence of free [Head-RCs] in childrens data, see in table 3 therelative proportion of free constructions versus [Head RC] integrated to a predicativeframe.

Insert here Table 3

Table 3 shows that more than half of RCs (=.577) do not fit a criteria used to attribute themcomplexity; notably embeddedness in a verb frame. And this is the normal situation but forJulio, who as we have already seen, slightly prefers from his earlier records RCs integrated

to a particular predicative frame with mira look. The rest of the children produce more[Head-RC] without any external syntactic function; they are not inserted in any clause, andare not syntactically dependent on any predicate. We do not have anything to say aboutsentence embedding here, since the frames are free NP plus RC, and in the extreme casesisolated RCs in an absolute position. We have anything to say either about the supposedfunctional parallelism between head and REL, since heads have no syntactic relation. Whatwe have instead is a set of referential or deictic forms (pronominal, nominal, DET) with anappended RC, which jointly constitute a basic construction frame. Head RCs have the typeof relation that a complement has to the complemente, a sort of topiccomment relation,CRs are NP expansions, rather than embedded clauses. Moreover, early RCs are notnecessarily appended to NP or PRO in the same turn. In effect, RCs themselves and [DET

RC] or [NP RC] frames have in early child language a freedom that RCs alone will scarcelykeep in adult language (Brucart 1999). In early child language, RCs seem to be parsed as

possible independent frames that may occur alone, or associated to a NP, with which astable frame [Head RC] is early established.

Consider now table 4 to have an idea about how often free [Head-RC] are prompted byconversation, against the frequency of free constructions that rely on childrens ownadaptations: when producing and absolute [NP RC] in one turn, or when an intonation

break or a syntactic closure part [NP RC] from a possible main clause.

Insert here Table 4

Here we can attest that important as it is the conversational support in early child language,dialogue is just one type of context, though an early one, which only accounts for about aquarter of free [Head RC] constructions (.267). Children can and do produce isolated[RCs], and independent [Head-RC] by themselves, with an intonation break or a closure

pronominal or lexical , an important and frequent preamble of integration (.328). In fact, jointly considered, childrens own free [Head RCs] with no dialogic support are the mostfrequent and accessible way for children to add RCs in their discourse.


13/35

We have again positive evidence that [RCs] are first parsed as isolated pieces, linked withmore or less fluency to a noun to form [Head RCs] frames. These are early, self-containedand independent constructions, which serve as a basic niche for children to adopt the use ofRCs. Although the ties between Head and RC may initially be also loose and Head and RCmay be flanked by an intonation break or even occupy each a different conversational turn.

These data put an accent in an important fact which Givon states in his 2009 paper (in press): RC emergence in childrens use is to be understood in the context of referencetracking. Head RC constructions serve reference building, which does not need any

predicative frame to be realized. So, isolated RC or absolute Head RC are doing thiscommunicative task in early child language, inserted in unfolding conversation or as a piecein the still fragmentary and unintegrated sequences of child speech, where every part heads and verb embedding frames to be, are mainly kept apart from the unintegrated,

paratactic adjunction of RCs.

Construction frames Notwithstanding that isolated [RCs] and independent [Head RCs] tend to be dominant inchildrens early production as we have seen in table 3, we already know that from thevery first moment children also insert [Head RCs] in various predicative frames, which inthe earliest data (table 2) are in fact item based (mainly, mira look and este es this is).As previous work has argued for their English parallels (Diessel 2004), we can expect these

particular frames will keep its initial readiness, and turn in time to be dominant.

The analysis of the lexical frames preferred by children across the period under study proves that in fact the frames first adopted mira look, and este es this is become forchildren the main discursive niches for RCs to occur. But childrens individual preferences

emerge early and keep operating across the period. And we attest in our data a wide framediversity, more than previous studies would lead us to expect (see table 5 below).

Insert here Table 5

Julio, who has shown from his first RCs a preference for the frame mira look, continuesto use this frame as his dominant one (12/24 =50%). But he also has a set of secondaryframes: hay-haba there is/was [], este es this is [] , quiero I want [], and ahandful of constructions with various verbs, none of them particularly prominent: ver see,tapar cover, prestar lend, prender turn on (see 10i-10iii).

10) i. Julio (25), ritually starting to tell Little Red-hood storyba-una vez una capeuceta. que se llama Ju Juya.

be-IMPF.3s one time a little-hood, REL RFL.3S=call-PR.3S Ju JuliaOnce upon a time there was a Little Red Hood whose name was Julia.

ii. Julio (26) telling a riddleetes una seora que se va llevar uos huevosthis is a lady REL RFL.3S=go.PR.3S take.away-INF DET eggsthere was a lady that was going to take away some eggs.

iii. Julio (26), playing with a lamp


14/35

pende la luz que s esturn.on-IMP the light REL AFF be.PR.3Sturn on the light that its the right one.

Elia continues to select as a main frame the first one she uses: the identificational presentative ste es [] this is (14/27 =52%); but she also adds other presentativeconstructions: aqu hay [] there is and tengo [], I/we have . Together againwith a handful of activity verbs: ensear show, dar give, buscar look for, quitar takeout (with one occurrence per verb type) (11i-11iii).

11) i. Elia (23)es unos juguete que son para jugar

be.PR.3S DET.P toy.S REL be.PR.3P to play-INF(those) are some toys which are for playing.

ii. Elia (24)(d)me la mochila que tiene...give.IMP=IO.1S the backpack REL have-PRS.3Sgive me, give me the backpack containing

iii. Elia (25)voy a buscar unos zapatos que son as, mira, stosgo.PR.1S to look.for-INF DET shoes REL be.PR.3S like.this look theseIm going to look for some shoes which are like this, look, these ones.

Flor, the child with a wider RCs production and not a clear dominant frame, adopts theidentificational frame ste es [] this is as her preferred one (14/81 =.173), togetherwith other frames with ser to be: the ritual era una vez once upon a time and a cleft type

(lo que pasa / quiero es what I want/ it happens is) , though none of them was presentamong her earliest productions. The preferred frame ( ste es identificational) emerges at(F42: 2;06,4), two months after Flors initial verb framed RCs: mira look, whichcontinues to be present. Besides, Flor uses two more frames built around tengo [] I have... and hay [] there is/are which become partly prominent. She incorporates gradually aconsiderable number of other lexically free and rather diverse verb frames: dar give,querer want, poner put, sacar take out, etc. (12i-iv).

12. i. Flor (34) from a window sees somebody has entered the courtyardhay una sudadera azul que entr a la casathere.is a t-shirt blue REL enter-PRT.3S to the housethere is (a person with) a blue t-shirt coming into the house.

ii. Flor (34) asking for some toys in a birds cageme das os nenes que tene? IO.1S=give-PR.2S DET babies REL have-PR.3Swill you give me the babies that (the cage) has inside?

iii. Flor (37) while eating and mentioning dirt things yo ten(g)o p(l)ato que no tiene .... mugueI have-PRS.1S dish REL NEG have-PST.3S dirtI have a dish that does not have dirt.


15/35

iv. Flor (44)me cuentas un cuento que traiga un libro ? OI.1S=narrate-PR.2S a story REL bring-SBJ.PR.3S a book?will you tell me a story that is on a book?

All this individual differences and preferences do not exclude some emergent regularities; but in no way point towards a unique presentational construction to include early RCs. Aset of four verb frames, which most children present, account for half of the embedded RCstokens ( mira look, ste es this is, hay there is, and yo tengo I have). A rich verbdiversity is only present in Flors data; Julio and Elia have only started their way into it.

In adition, we have to consider that besides the particular verb frames included in table 5,which tend to present an Object (O) slot for [Head RCs] insertion, the set of verbs jointly

presented under various are mainly transitive activity verbs which have also an O as anelaboration site for a [Head RC] to occur ( buscar look for, cantar sing, cazar hunt,

comer eat, contar narrate, dejar leave, ensear show, escarbar dig, llevar take,necesitar need, pasar pass, pegar hit, prender turn on, prestar lend, quitar takeaway, sacar take out, tapar cover, tirar throw away). Only a few and infrequentintransitive verbs ( estar be.loc, llamarse be called, ir go, llegar arrive) offer aSubject position as a [head RC] elaboration site.

These data lead us to conclude that there is no specific construction being a general defaultto anchor a [Head RC] insertion. Although some constructions are first used and tend to be

preferred as sort of attractor-frames, every child may select particular constructions, whichmay be similar to the ones other children prefer. But every childs selection for some

particular frame has to be determined in close inspection to the childs own data.

Despite this frame diversity it is not out of question that there may be a reason for theconfluence and/or on the opposite individual preferences we attest; a possible functionalassembly that may joint together a set of frames, which in some not overtly marked way arehaving the same effect, and probably doing the same operation.

In fact, it seems to be the case that what all these frames have in common is a slot position[] in the focus domain, where a NP can be inserted. This position tends to be the Objectfor most attested frames ( mira look, tengo I have, quiero I want, etc.); but it may also

be a Predicative position, if a particular child adopts ste es this is identificational-equative construction focusing a descriptive a NP as her preferred frame. And it could

equally possible had been an S, if the locative presentative frame aqu est /n lit. here be.loc-PR.3s/p, meaning sort of here you are, here it is /they are had been preferred bysome child; which has not been the case in these data, but could still be possible for anotherchild. In fact most cases where the slot position is an S, are on charge of various intransitiveverb frames, well known for the focal informational properties their subject present.

Our data do not point towards a unique syntactic position, nor an item-based unique frame, but towards a set of constructions with an elaboration site: a slot where a prominentreferential or descriptive NP is inserted; Objects being a well known position to put new


16/35

focal information, as Subjects in intransitive verbs are (Clancy 2003; Du Bois 2003; Givn1984). Not to insist on the informational prominence the construction este es un/el this isa/the projects upon the identificational noun phrase that takes the Predicate position.What we find across the various syntactic frames where [Head CRs] occur in discourse is a

NP position inside the focus domain that functions as an elaboration site. And the same

informative position may be credited to [Head CRs] even when occurring isolated, in anabsolute constructions: both, when prompted by WH-questions, which build a focus position for their answers, or as absolute NPs which are by themselves focal (Zubizarreta1999). 3

As for the various constructional frames involved in this sort of focus elaboration operation,this communality could be generalized by considering them presentative ones (f.i. Alfonso& Melis 2007). I would rather insist on the focus side of the generalization, since dialogue

prompted cases, or absolute [Head RC] frames give the same focusing result thatconventional presentative frames do, despite not being inserted in any overt presentativeframe.

On the grammar side, it is true that when we consider this elaboration site in syntacticterms, we see that OBJ and PRED-NOM are the preferred syntactic position. They jointlyrepresent the 86% of [Head-RCs] inserted in a predicate frame, with a frequency ranging

between 96 ~ 84 % in our childrens data (see Table 6 below for their independent proportions). However, this preference does not seem to depend upon some abstract properties of Objects or Nominal Predicates. We can trace back the dominance of particularsyntactic position to an emergent effect of the lexical predicates in childrens preferredconstructions (Clancy 2003; Du Bois 2003), which include a slot to be worked out as anelaboration site.

Insert here Table 6

At the end, we can make the generalization that it is not the deterministic effect of presentative frames adoption, nor any syntactic variable which define childrens particularsites for [Head RC], but the joint effect of childs adoption of selective construction and theinformational properties of those constructions, that have a focus position to elaborate andsolve reference building scenaries: either in reference presentation or in referencenegotiation, by means of a RC which expands a referential or descriptive noun phrase.

3.3. Internal syntax

Clause adaptationsOne of the main sources of complexity in canonical RCs refers to the internal adaptationsthe RCs expose: particularly the empty space or syntactic gap RCs have, the resumptive

3 This elaboration site, which I consider to be in the focus domain and even focal is not meant to bethe focus in Lambrechts terms; but in a way it prefigures what would be this position in theinformation structure, itself an expected childs developemental achievement.


17/35

pronoun they may include, and the dual binding relation which relative pronouns operate: backwards referring to the head, onwards associated to the syntactic gap.

From a wider perspective, this problem is connected to a general issue, the modificationsany dependent clause might present in its internal structure, qua dependent clause. We

know well that clause linkage tends to be marked by clause internal adaptations, whichrange in a cline from nullto maximal adaptations, so that a clause may not have any markto indicate a dependent status or present various types of dependency marking, and a looseror tighter integration: from boundary marking, to illocutionary force integration,informational structure restrictions, and argument sharing or integration. In the positive sideof this marked dependence cline, heavy reductions and internal marking are expected:uninflected verbs or subjunctive verb inflexion, argument sharing reductions or dependentargument forms, and the like. The more marked adaptations side is expected to align withthe more clearly embedded and dependent clauses (Aissen 2004; Givn 1980; Givn 2006;Lehman, 1988; Van Valin & La Polla, 1997).

A question to be explored next is whether childrens RCs do, or do not present anyadaptation that marks them to be dependent. For RCs the expected internal adaptations aremainly related to REL properties: coreference and syntactic function. REL marks acoreferential link between the head and the internal constituent in the RC, whose syntactic

position REL is credited to occupy. Relativization accessibility (Keenan & Comrie 1977)considered from a Usage perspective leads us to expect Subject REL will be the mostfrequent ones, followed by Object, Indirect Objects, and Oblique RELs. The point is a little

bit tricky in Spanish, a subject dropping language, since any subjectless RC need not beconsidered to have the subject represented by REL; the very same subjectless sentencecould occur outside any dependent clause context and independently of REL presence.

Suspending for a while the supposed truth that internal subject omission in RCs is a proofof the binding and syntactic properties of REL in early child language, we will look for theinternal adaptations that RCs may present, which will prove them to be dependent. But weexpect RCs in child language be more similar to independent main clauses, despite REL

presence. In effect, some recent experimental work on childrens RCs has shown the impactof RCs similarity to independent clauses on acquisition processes (Diessel & Tomasello,2005). But previous studies of spontaneous data have not particularly emphasized this typeof comparison.

Given that REL function and binding properties are the most conspicuous aspect of RCcomplexity, we will consider next the syntactic properties of REL, in order to evaluate howadapted are Childrens RCs to its dependent status, as compared to an independent clause.

REL function in questionThematic association. In a similar vein to the embeddedness problem, the analysis of REL

position and function in childrens RCs gives us the unexpected result that REL do notnecessarily have any syntactic function to fill. RCs may have no clear syntactic gap norconstituent omission whose function REL pronoun will take. This is clearly seen in caseslike (13i-13iii).


18/35

13) i. Julio (28) is telling Goldilocks and the three bears storyhaba tles ositos que se cai una nia la silla exist-IMPF.3S three bear-DIM-P REL RFL.3S=fall-PR.3S one girl the chairthere were three bears that a child falls down from the chair.

ii) Elia (24)

me voy a sabe(r) una cancin que una casita es bonitaIO.1S=go-PR.1S to know-INF a song REL a house-DIM be-PR.3S niceI will learn a song that a little house is nice.

iii) Flor (67) telling a storyera una nia que el abuelito se meti a un abujero de ratones

be.IMPF.3S a girl REL the grandfather RFL.3S=go-PST.3S to a hole of micethere was a girl that her grandfather went into a mice hole.

Head NP are expanded in these cases by means of a RC configuration which has no internalgap, nor trace, nor any evidence of structural dependence: only discourse continuity is keptin a lax way and the relation between Head and RC is a thematically supported one.These RCs, sort of syntactic anacoluthon , if taken seriously and not merely consideredanomalous, permit us to envisage a possible and early way for children to build RCs.Children may work on the linking side of these constructions on the basis of a thematicassociation a well known procedure to keep discourse continuity, with no syntacticconditions to bind the RC to its external context and no modification on its internal form.REL would be in these cases just a sort of local continuity mark.

Resumptive linking. Together with these thematically linked RCs, with no internaladaptation, we have the same effects in RCs resulting from a Resumptive PRO strategy(RES). A pronoun takes the syntactic position expected to be associated with REL, leavingRCs with no internal gap. So children have a Possessive RES in CR instead of a genitiveREL (14i); an IO-RES in (14ii), instead of a case marked IO-REL.

(14) i. Elia (29) is narrating the life of a national hero when he was a child.un pobre que se muri sus papsa poor REL RFL.3=die-PST.3S POS.3P parentsa poor (child) that his parents got dead (=whose parent died).

ii. Flor (44)la nena que, la que le pusiste la piyamathe baby REL, DET REL OI.3S=put.on-PST.2S the pajamasthe girl that, the one you put her the pajamas.

In colloquial adult Usage, RES-strategy is normal for IO-RELs (as in 14ii) and is quitefrequent for GEN-RELs (as in 14i), though grammarians proscribe them from writing. Butchildren also produce Object-RES (15i-iii) and Subject-RES RCs (18i-ii), which in adultUsage are almost absent (O-RES), or plainly ruled out (S-RES).

(15) i. Elia (29) asking for some nasal drops (Object RES)unas, unas que las tengo aqusome, some REL O.3P=have-PR.1S heresome, some that I have them here.


19/35

ii. Flor (52) (Object RES)la que me la dio mam, la voy a ponerDET REL OI.1S=O.3S=give-PST.3S Mum O.3S=go.PR.1S to put-INFthe one that my Mum gave it to me, Im going to put it.

iii. Julio (29), excited tells about a baloon his father just bought (Object RES)mo, mi (gl)obo que me lo comp(r)mine, my balloon REL OI.1S=O.3S=buy-PRT.3Smine, my balloon that he bought it for me.

(16) i. Julio (30) looking a book (Subject RES)te voy a ensear uno que se es un caballoIO.2S=go.PR.1S to show.INF one REL that be.PR.3S a horseI will show you one which that.one is a horse.

ii. Flor (63) (Subject RES)

es una nia que ella se duma be.PR.3S a girl that she RFL.3S=sleep.IMPF.3S(this) is a girl that she got slept.

What we want to emphasize here is that in both cases, Thematic (THM) and RESconditions, RCs do not have any empty place for REL. In THM, the internal site does notexist; in RES the internal function is filled by a pronoun (RES). Although RCs with RES

pronouns are considered to be the result of a secondary relativization strategy, and may beconsidered more elaborated types of RCs than simple REL constructions (Comrie &Kuteva, 2007), the fact is that they are simpler. Even if we have an internal link in RCswith a RES, these constructions are in all respects similar to an independent clause.

Relative-word need not to be here but a linking mark, and internal syntax of RC is similarto an independent clause.

Again, RCs with a RES pronoun just manifest a topic continuity procedure, which being ananaphoric operation may apply all over the grammar, and by itself do not define structurallydependent clauses but only topic continuous ones (Givn 1984; Givn 1990). We propose,hence, to consider RES constructions as a starting small effect: on the formal side, there is

just a local relation between a Heads and a RC marked by REL as a topic continuity mark;and a sequence of thematically or topic continuous constructions, on the functional one(Kuno 1987).

But again we could also suppose a Usage effect of experienced adult models. It is in fact possible in adult Spanish Usage to have a RES pronoun inside the RCs, down from IOrelativization as a necessary condition, and in GEN-REL, as a generalized one.Effectively, but for written texts, the RES-strategy has practically replaced in adult Spanishthe marked Genitive-REL cuyo whose, with a REL + internal possessive (= que su RELPOS). RES-strategy also emerges in various peripheral syntactic positions, but it isinfrequent and restricted almost anomalous with O-REL constructions, and not at all

permitted in adult S-REL (Lope Blanch, 1984; Palacios, 1983). However, our childrensdata exhibit what would count as a RES strategy even in S and O positions. So we can not

but insist that this is a childs way to build relatives: putting as a RC a main clause frame,


20/35

keeping topic continuity by means of RES, and adopting as a surface mark a REL, with noother syntactic integration at all. These RCs all by themselves could be produced asindependent clauses, and REL will be a sort of topic continuity mark. 4

Hanging relativesWe can add on this line of argumentation various constructions initiated by the child, andthen interrupted, after REL production. Here the REL produced by child does not arrive tohave an internal function, since the construction is not finished, but only announced (17i-17iii).

(17) i. Elia (23)tenemos muchas cosas que have-PST.1P many things RELwe have many things that

ii. Flor (61), doing some gardeningt escarbas, las plantas que, que...you dig-PST-2S, the plants REL, RELYou dig the plants that, that

iii. Julio (29)mira, aqu hay un cuento que look-IMP, here be-PST-3S a story RELlook, here it is a story that

These fragmented productions make clear that a REL may be produced after a nominalhead, without the childs having planned the internal configuration of RCs. REL presencecan only be explained here if we credit childs adoption of a frame N+REL, where N RELkeep a sequential relation, not necessarily a pronominal or syntactic one, in a stillunplanned RC construction.

Childrens RCs that are similar to independent clauses, with no internal function for REL but just a thematic link or a resumptive pronoun, are mapped in Table 7. They jointlyrepresent a 17% of RCs. with very similar numbers for every child: Flor has 18%, Julio17% and Elia 15% of these types of RC constructions, where REL is not associated to anyinternal gap.

Insert here Table 7

Topic continuity or Subject gaps in RCsFor a subject dropping language, like Spanish, the absence of S in RCs can not be used todetermine a subject gap, since a gap i. e., subject drop is possible outside RCs. Subject

4 In our childrens data, Case + REL is just starting to emerge, but we do not have any Case + RELcombined with RES. This fact may be taken as a secondary evidence that REL does not fill anysyntactic position in RCs when it coocurs with RES .


21/35

internal REL constructions (S-REL) have in Spanish the same type of structure that anindependent clause has; particularly in the context of situated reference, a normal conditionfor child speech (Givn 2009), or referential continuity, a discourse condition for RCs. So,even if RCs configuration seems to have a canonical S-REL, we can argue that there is noneed to consider a gap to be present in a RC when considering Subject continuous RCs.

(18) i. Floryo tengo un perro en mi casa que, que hace guau guauI have-PR.1S a dog at my home REL, REL make-PR.3S arf arfIve got a dog in my house that, that barks arf arf.

ii. Julio :sta es a mam que est enojadathis be.PR.3S the Mum REL be.PR.3S angryThis one is the mother that is angry.

iii Eliaeste amarillo es el hijo que est con todos sus papsthis yellow be.PR.3S the son REL be-PR.3S with all his parentsthis yellow one is the soon that stays with all his parents.

This we also have seen in table 7: Subject internal continuity in childrens RCs is mostfrequent. This option represents a total of 58% among all RCs. Besides, Elia shows even amore radical preference for S-RELs (=.737) face O-RELs (=.113). Flor has lower rates forS-RELs (=.545), since she produces substantially O-RELs (= .277), and even a couple orIO-RELs (= .011) (See 16ii above). Finally, Julio has the closest relation between S-REL (=.477) and O-REL (= .341), since he has adopted from the first moment some ritual internal

predicates in RCs, which ask for an O-REL ( encontr I found, tengo I have). We can seethis even distribution, overall favorable to S-REL, more clearly in Fig. 1, where the

previous cases of clear no gap condition have been left aside, in order to focus S-RELand O-REL selection.

Insert here Figure 1

We find here, that S accessibility is higher than O in RCs binding; as it is expected here andoutside RCs conditions. And since Subject continuity in RCs is not marked, we could addS-REL to RCs that keep the form of an independent clause; which amounts to more than ahalf of all RCs. Topic continuity supported in S-REL, with no marking inside RC make theuniverse of simple unmarked RCs, a dominant one. The few cases with a RES subject (like16i-16ii, above) show even more clearly an independent clause frame in a RC position.Henceforth the set of RCs internally marked as dependent will only be Object internal REL,which we turn to consider next.

Hard cases made easier: Object RELsO-REL constructions are the main and almost undisputed candidates to illustrate thecomplexities attributed to RCs in general. Spanish Object presence for transitive verbs isconsidered categorical, but under certain conditions it is possible to have Object omissions


22/35


23/35

giving and object manipulation verbs (give, receive, buy, bring, leave, put) with a 1S benefactive. But the constructional diversity she develops latter leaves out of the questionthe productivity that O-REL constructions get for her (See Table 8).

(21) Flor O-RELme das os nenes que tene ? will you give me the babies she holds?la opa que, que dejates en la bebaloda the clothes that you left in the laundry. la capeta que yo te taje the folder I brought to you.mis pasas que me compr Ins the raisins Ines bought for me.

Once more, the landscape that our data designs on the domain of O-REL is an accidentedone. Childrens O-REL constructions present the individual differences that result whenfocus of attention, processing preferences (Bates, Bretherton & Snyder, 1988) andindividual experiences are jointly played. But we can make anyway some generalizations.

On the verb side of the corner we see recurrent predicates of particular semantic frames:object manipulation, objects transfer and bare object contact, which are used for agrounding expansion, anchored in speech act participants. This regularity gives verbs liketener have- hold, give , receive , find, and the like, the lions share among O-RELconstructions (See Table 8, below).

What we see in table 8 is a more or less extended verb inventory in every child, which isconfidently on a pair with a more or less rich and diverse object representation. (Considerthat only Flor refers to her own perceptions and desires with RCs internal verbs: which Iwant, which I saw). There is a main grounding verb, semantically general and

polisemous: tener have, as a sort of O-REL attractor. But the verb inventory includesconcrete manipulative actions put, move, bring ; concrete-social relations give, present,buy-, and internal experiences see, want , all mostly related to what sounds like a natural

history of childs relations to concrete objects.

The semantic frame motivation for early verbs involved in O-REL constructions is furthersupported when we see some specific pairings among activity verbs and entities referred asOs in O-REL constructions. Although children first and mainly use general verbs framesfor grounding concrete objects in discourse, they add latter specific verbs for specific itemsin O-REL constructions: stories are narrated, books tell things; songs are played (lit. poner put) in a recorder; paintings and designs are done and erased; and many objects arereported as made (22).

(22) quelo lo que bol Cami I want the one that Cami erased (a design).

oyes la que pus did you listen the one I put (some music).ola una que no hemos cantado now one that we have not sang (a song).ma lo que dice look what it says (the book). mam, ma lo que hizo pap Mom, look what Daddy has made (a design).

Insert here Table 8

A second type of generalization concerns the semantic category of the items that take theO-REL position. They are all concrete objects. This is true not only for Noun headed RCs,


24/35

but also for DET-REL constructions, that indexically refer to situated items by means of ageneric determiner plus a relative: lo que (lit. DET REL the that= what). This is whynoun, pronoun and DET headed RCs, all share the same types of internal predicate.

(23) i. lo que tienes DET.N REL have-PR.2Swhat you hold.

ii. mira lo que encontrlook-IMP DET.N REL find-PST.1Slook what I found.

iii. etes lo que quierothis is DET.N REL want-PR.1Sthis is the one I want

iv. mila lo que me compr mi paplook DET:N REL IO.1S=buy-PST.3S my Dadlook what my Dad bought for me.

Supporting the complex syntax of an O-REL, we find concrete objects surroundingchildrens space, and a rich knowledge of general and specific relations and activitiesobjects are involved in, which are at childrens disposal to help them to identify or groundconcrete items in discourse situation.

Even if O-REL show a syntactic gap in their internal configuration, O-REL the children produce are the joint result or a conspiracy effect from rich Object representation, whichmay be implanted on the basis of recurrent discourse practices, commenting on objectscloser vector the one who holds them, caused motion (take, in-out), and social transfer(giving, receiving, and the like).

At the same time, O-REL constructions co-occur with O-RES the most, and among theirinternal predicates, at least tener have, and querer want are mentioned by grammariansto accept O-dropping. We cannot add these facts as an argument here, because we wouldneed as a basis to know the syntax of childs Object dropping in main clauses, which to myknowledge has not been studied yet in Spanish acquisition.

4. Conclusion

Looking towards a possible learning explanation of what otherwise would be considered to be unlearnable and developmentally unexplainable, this paper has argued that early relativeconstructions do not necessarily have the syntactic properties that would define them ascomplex constructions outside childrens reach.

We have found in early production data that RCs main niche is not embedding; RCs tend to be adjoined to plain and syntactically independent heads, and suppose no external functionfor the head. Our data is congenial to Givns proposal that a sort of condensation is on the


25/35

way from these non integrated frames to their integrated counterparts, which emerge laterand have as a preamble, isolated production in different intonation units, hanging relatives,and pronominal and lexical closures around the verb frame.

Despite the growing importance of lexically specific verb frames for Head RC insertion, we

do not have a single or main presentative frame as the starting or turning point to get RCembedding. The set of verbal frames we have as Head RC niches define theircommonalities not as presentational frames unless we loose the identification of

presentational frames to the point where it losse it formal significance but in terms of theopen slot they include in their focus domain: an O for transitive verbs or an S for anintransitive verb frames. There, children can build an elaborated reference caracterization;

both, to present a new participant, or to settle an argument on reference negotiation.

We also have found a significant absence of internal function for REL, and clear evidence pointing towards REL not necessarily being a pronominal form. RCs may not have anempty space to be filled by the relative word: either because RCs is only thematicallylinked to the head, and there is not a correferential argument to be binded by REL, or

because the correferential argument is overtly exposed in the RC by means of a RES pronominal, leaving no need to consider REL as its anaphoric exponent. Subject internalRELs are a particular and dubious frame in Spanish, a subject dropping language whichdoes not need not to have an S, nor, as a consequence to have an S-exponent coded as aREL in RCs.

In every case, RCs tend to have the same structure that a main clause may have.This is particularly the case for S-REL, for RES RCs and for THM linked RCs.

In a parallel to the functional motivation to use a Head RC in the landscape of referencetracking, for reference building, presentation and negotiation, we infer a functionalmotivation for RC internal configuration: notably REL marking, no gap tendency, preferredlexical verbs. What on the surface seems to be an adult type of structure unfolds underdetailed scrutiny as a sort of main clause insertion in an RC position. So, the mainfunctional effect of a REL is topic continuity marking. REL only asks to be considered a

pronoun when marked by case, which in early child language does not yet occur, and it willnot for many months after RC configurations first emerge.

So, although the main source of REL binding would refer to RC subject as therelativization hierarchy predicts this prominence does not guarantee we face arelativization scenario, but simply a subject based topic continuity effect. Cases whererelative word can be credited an internal function reduces to Object RCs.

Though we need to know more of the Object dropping conditions in early child languageand in adult Usage, possibly dropping with tener have and querer want and we couldeasily add, ver see , and hay there is opens the scenario for unmarked topic continuityadvancing to a second syntactic position O, and not just to S.

Even if O-REL would end up as clear modified RCs, with an internal gap, O-REL casesshow some simple ways to make out of a complex structure a simpler one. Here it is not thecase that the structure has not been modified although O-REL frames are the main source


26/35

of RES, but the point is that internal configurations of RCs profits from a handful ofsemantic frames, whose reduced types and centrality make possible for the child to applythem routinely and have a high memory access.

So, we can conclude that as far as syntactic criteria may characterize early RCs as complex

frames (lets say embeddedness, REL dual function external-internal and functional parallelism between head and REL), childrens data does not quite comply with it.Certainly it is not the case that there are no regularities in childrens early RCs, but theseregularities do not conform to the abstract predictions on complex dependence of relativesor functional parallelism. Only the relativization accessibility hierarchy agrees with thecontinuity based on the subject of RCs we have observed. But even in this case, we cannotdefinitively state that REL is taking the role of an S, nor that RC is exposing a definitedependency link towards a main clause, which may not even exist if the RC occurs in ahead free situation.

Hard and disputable as it may be attempting to characterize in its own terms theorganization of childrens entrance to complex syntactic facts, we consider that children areoperating in a way that can be described as a starting small landscape. Children treat RCsas chunks that may be produced alone; they recruit some frames, not necessarily ones witha verb, and they may locally insert a RC to a referential or descriptive NP, without takingcare of more.

The fact that some predicative frames recur from child to child and emerge as preferredones for individual children points towards a preferential syntactic position to insert alexical NP and a successive RC. It is more an information effect that an abstract and a prioricondition for RCs to occur.

We have considered all these syntactic twists having a RES, forgetting the gap, taking aHEAD + RCs alone, with no syntactic frame, or rather selecting some particular syntacticframes as starting small effects. We want to emphasize that what we have is a set ofways to act and reduce the complexity of RCs. Sort of divide and you will vanquish(Elman 1993; 2005), adopt a construction frame (Diessel y Tomasello, 2001), put a mainclause as an RC (Diessel y Tomasello, 2005), which make easy to get a RC result, underthe functional drive of complex reference building.


27/35

List of abbreviations

AFF AffirmationCL ClosureDET Determiner

DIM DiminutiveGER GerundIMP ImperativeIMPF Imperfect

NEG NegationPR PresentPST PastREL Relative pronounRFL ReflexiveRES Resumptive pronounSBJ SubjunctiveF FeminineM Masculine

N NeuterP PluralS Singular


28/35

References

Aissen, J. 2006. Taller de Complementacin (agosto, 2006). San Cristbal de las Casas,Chiapas: CIESAS-Suroeste.

Alfonso, M. & Melis, Ch. 2007. La alternacia en el orden del sujeto en la oracinintransitiva del espaol. Primer Encuentro de Estudios de Gramtica del Espaol. Mayo 2007. Mxico: El Colegio de Mxico.

Barriga, R. 2002. La produccin de oraciones de relativo en nios de seis aos. In Estudios sobre habla infantil en los aos escolares. un solecito calientote . Mxico: ElColegio de Mxico.

Bates, E., I. Bretherton & L. Snyder (eds) 1988. From first word to grammar: individualdifferences and dissociable mechanisms. Cambridge: Cambridge University Press.

Bello, A. 1847/1988. Gramtica de la lengua castellana destinada al uso de losamericanos. Ed. crtica de Ramn Trujilo con las notas de R. J. Cuervo (1988).

Madrid: Arco Libros.

Bowerman, M. 1979. The acquisition of complex sentences. In Language acquisition.Studies in first language development , P. M. Fletcher & M. Garman (eds), 285-305.Cambridge: Cambridge University Press.

Braine, M. D. S. 1976. Childrens first words combinations. Monographs of the Society for Research in Child Development 41.

Brucart, J.M. 1999. La estructura del sintagma nominal: las oraciones de relativo. InGramtica descriptiva de la lengua espaola, vol 1, I. Bosque & V. Demonte

(coords), 395-522. Madrid: Espasa. Campos, H. 1986. Indefinite object drop, Linguistic Inquiry 17(2): 354-359.

Comrie, B. & Kuteva, T. 2007. Complexity and relative clauses in creole languages,Seminario de complejidad sintctica . Noviembre 2007. Hermosillo, Son.:Universidad de Sonora.

Clancy, P. M. 2003. The lexicon in interaction. Developmental origins of preferredargument structure. In Preferred argument structure: Grammar as Architecture for

function, J. Du Bois, L. Kumpf & W. Ashby (eds), 81-108. Amsterdam: JohnBenjamins.

Dabrowska, E. 2001. Learning a morphological system without a default: The Polishgenitive. Journal of Child Language 28: 545-574.

Dabrowska, E. & Lieven, E. 2005. Towards a lexically specific grammar of childrensquestion constructions. Cognitive Linguistics 16(3): 437-474.

Diessel, H. 2004. The acquisition of complex sentences in English. Cambridge: CambridgeUniversity Press.


29/35

Diessel, H., & Tomasello, M. 2000. The development of relative constructions in earlychild speech. Cognitive Linguistics 11: 131-152.

Diessel, H., & Tomasello, M. 2001. The acquisition of finite complement clauses inEnglish: A usage based approach to the development of grammatical constructions.Cognitive Linguistics 12: 97-141.

Diessel, H., & Tomasello. M. 2005. A new look at the acquisition of relative clauses. Languag e 81(4): 882-906.

Du Bois, J. 2003. Argument structure. Grammar in use. In Preferred Argument structure:Grammar as Architecture for Function , J. Du Bois, L.-E. Kumpf & W. J. Asby(eds), 11-60. Amsterdam: John Benjamins.

Echeverra, M. 1978. Desarrollo de la comprensin infantil de la sintaxis espaola.Concepcin, Chile: Universidad de Concepcin.

Elman. J. L. 1993. Learning and development in neural networks: The importance of

starting small. Cognition 48(1): 71-99.Elman, J. L. 2005. Learning and generalizations. Lesson from neural networks. Plenary

Lecture. X International Congress for the study of child language. Freie UniversittBerlinIASChL (26 July, 2005). Berlin.

Fernndez Soriano, O. 1999. El pronombre personal. Formas y distribuciones. Pronombrestonos y tnicos. In Gramtica descriptive de la lengua espaola vol. I, I. Bosque& V. Demonte (coords), 1209- 1274. Madrid: Espasa.

Gathercole, V. M., Sebastin, E. & Soto, P. 1999. The early acquisition of Spanish verbmorphology. Accross the board or piece meal knowledge. The International Journalof Bilingualism 3: 183-182.

Givn, T. 1980. The binding hierarchy and the typology of complements. Studies in Language 4(3): 333-377.

Givn, T. 1984. Syntax . A Functional-Typological Introduction , vol. I.Amsterdam/Philadelphia: John Benjamins.

Givn, T. 1990. Syntax. A Functional-Typological Introduction, vol. II. Amsterdam/Philadelphia: John Benjamins.

Givn, T. 2006. Multiple routes to clause union: the diachrony of syntactic complexity.Seminario de complejidad sintctica , Universidad de Sonora, Mxico (13-14 nov.2006).

Givn, T. 2009. The ontogeny of relative clauses. How children learn to negotiate complexreference. In The Genesis of Syntactic Complexity , T. Givn (in press), Chap. 5.Amsterdam: John Benjamins.

Hale, Ken. 1974. The adjoined clause in Australia. In Gramatical categories in Australianlanguages, R. M. Dixon (ed), 78-105. Canberra: Australian Institute of AboriginalStudies.

Hurtado, A. 1984. Estructuras tardas en el lenguaje infantil. Mxico: SEP-OEA.


30/35

Keenan, E. & Comrie, B. 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry 8: 63-99.

Kidd, E., Brandt, S., Lieven, E. & Tomasello, M. 2007. Object relatives made easy: Across-linguistic comparison of the constraints influencing young childrens

processing of relative clauses. Language and Cognitive Processes 22: 860-897.

Kuno, S. 1987. Functional syntax; anaphora, discourse and empathy . Chicago: TheUniversity of Chicago Press.

Lambrecht, K. 1988. There was a farmer had a dog: syntactic amalgams revisited. Berkeley Linguistic Society 14: 319-339.

Lehman, Ch. 1988. Towards a typology of clause linkage. In Clause Combining inGrammar and Discourse, J. Haiman & S. Thompson (eds), 181-225. Amsterdam:John Benjamins.

Limber, J. 1973. The genesis of complex sentences. In Cognitive Development and the Acquisition of Language , T. Moore (Ed.). New York: Academic Press.

Lieven, E., Pine, J., & Baldwin, G. 1997. Lexically-based learning and early grammaticaldevelopment. Journal of Child Language, 24: 187-220.

Lope Blanch, J. M. 1986. La despronominalizacin de los relativos, Estudios de lingsticaespaola, 119-136. Mxico: Instituto de Investigaciones Filolgicas-Universidad

Nacional Autnoma de Mxico.

McClure, C., Pine, J. & Lieven, E. 2006. Investigating the abstractness of children's earlyknowledge of argument structure. Journal of Child Language 33(4): 693-720.

Moreno Cabrera, J. C. 1999. Las funciones informativas: Las perfrasis de relativo y otrasconstrucciones perifrsticas. In Gramtica descriptive de la lengua espaola vol.III, I. Bosque & V. Demonte (coords), 4245- 4302. Madrid: Espasa.

Newport, E. 1990. Maturation constraints on language learning. Cognitive Science 14: 11-28.

Palacios, M. 1983. Sintaxis de los relativos en el habla culta de la ciudad de Mxico. Mxico: Instituto de Investigaciones Filolgicas-Universidad Nacional Autnomade Mxico.

Prideaux, G. D. & Baker, W. 1986. Strategies and structures: The processing of relativeclauses . Amsterdam: John Benjamins.

Rojas-Nieto, C. 2003. Early acquisition of verb inflexion in Spanish. A Usage-basedaccount. Psychology of Language and Communication 7(2): 33-56.

Rojas-Nieto, C. 2007. La base de datos ETAL: Etapas tempranas en la adquisicin dellenguaje. Jornadas Filolgicas 2005 . Mxico: Instituto de InvestigacionesFilolgicas-Universidad Nacional Autnoma de Mxico.

Rojas-Nieto, C. 2008. Before grammar. Cut and paste in early complex sentences


31/35

(submitted).

Seidenberg, M. S., & MacDonald, M. C. 1999. A probabilistic constraints approach tolanguage acquisition and processing. Cognitive Science 23: 569-588.

Sicuro Corra, L. 1995. An alternative assessment of childrens comprehension of relativeclauses. Journal of Psycholinguistic Research 24: 225-262.

Tavakolian, S. 1981. The conjoined clause analysis of relative clauses. In Languageacquisition and Linguistic theory, S. Tavakolian, (ed), 167-187. Cambridge Mass.:The MIT Press.

Tomasello, M. 2000. Do young children have adult syntactic competence? Cognition 74:209-253.

Tomasello, M. 2003. Constructing a language. A usage based view of languageacquisition. Cambridge, Mass.: Harvard University Press.

Thompson, S. 1998. A discourse approach to the crosslinguistic category of adjective. In Explaining language universals, J. Hawkins, (ed), 167- 185. Londres: BasilBlackwell.

Van Valin, R. D. & LaPolla R. 1997. Syntax. Cambridge: Cambridge University Press.

Zubizarreta, M. L. 1999. Las funciones informativas: Tema y foco. In Gramticadescriptiva de la lengua espaola vol. III, I. Bosque & V. Demonte (coords),4215- 4224. Madrid: Espasa.


32/35

Table 1. Data Base

Observations Childconversational

turns

Age range(in months)

Time RCs

FLOR(Fem)

31-67 24,318 27 m 48 m 74 hrs. 188

ELIA(Fem)

20-29 7,244 40 m 50 m 20 hrs. 80

JULIO(Male)

20-30 8,835 37 m 48 m 22 hrs. 44

Total 57 40,397 27 m 50 m 116 hrs. 312

Table 2. First five RCs exemplars for every child

Predicate framing Type of head

Lex-fixed Free NPHeads

NP Pronoun Det

FLOR mira DET REL 1 4 4 0 1ELIA este es DET N REL 3 2 5 0 0JULIO mira REL 4 1 1 0 0

Table 3. Syntactic integration of [Head RCs] constructions

Integrated RCs Non-Integrated RCs Total RCs

Tokens % Tokens % ! TokensFLOR 81 .431 107 .569 188ELIA 27 .338 53 .662 80JULIO 24 .545 20 .455 44Group 132 .423 180 .577 312

Table 4. Free Head RCs conditions

Tokens %Dialogic niche 48 .267

Isolated Head-RCs 41 .223Intonation break 32 .178Lexical or Pronominal closure 59 .328Total 180 1.000


33/35

Table 5. Verb frames for [Head-RC] insertion

JULIO ELIA FLOR Frame diversity /Tokens ( 39/132) 9/24 12/27 32/81Lexical Frames: este es [ ] this is 3 14 14mira [ ] look-IMP 12 7hay /haba [ ] there is/was 3 3 5tengo [ ] I have 1 1 5[DET REL vb ] es FN be cleft 6era una vez [ ] there was once 5dame [ ] give me 1 4viste/ ves [ ] have yo seen 1 3llevas [ ] you take 4oye [ ] listen-IMP 3hace he makes 3quiero [ ] I want 1 2

sabes [ ] you know 2

Various (1 token/each) 3 8 18

Table 6. External position of [Head RCs]

Syntactic slot of Head RC in a Predicativeframe

! Embedded

RCsO % S % PN % Obl %

FLOR 43 .531 18 .222 15 .185 5 .027 81ELIA 9 .333 2 .074 14 .519 2 .074 27

JULIO 20 .833 0 0 3 .125 1 .042 24Group 72 .545 15 .113 41 .311 8 .061 132

Table 7. Internal gap, possibly association to REL*No gap Possible

gapInternal

GapUnplanned THM RES [ S ] [O] ! RCs

FLOR 9 11 13 103 59 188ELIA 4 5 4 59 9 80JULIO 3 3 2 21 15 44Group 16 19 19 183 76 312% 54 /312 = .173 = .587 = .244

* In this table, RES cases have been left out from S-REL and O-REL counting


34/35

Table 8. Internal Syntax: Verb types and Object continuity RCs

ELIA JULIO FLOR Verb types / RCs tokens 7/9 8/15 22/59Contact-holding verbTengo-tienes I/you have 5 6 12Object transfer verbsencontr/ perdI found , lost,me dieron /compraron/ regalaron / trajeronthey gave/ bought/ presented/ brought to me

5 6 12

Object manipulation verbs Poner, sacar, llevar, quitar, dejar, pescarput in, put out, take in, take out, leave, catch

1 1 10

Speech activity verbsdecir , cantar , contar , leer (a story) pedir say, sing, tell , read -a story- ask, demand

1 1 4

Various Activity verbs Hacer make 1 1 5borrar , comer , romper usar erase, eat, brake, use

1 7

Other non activity verbs 2quiero want 4ver see 3hay exist 1


35/35

Figure 1. Topic continuity in RCs and posible REL function

Documents

Construcciones lingüisticas relativas tempranas.pdf