complementizers and subordination in typical language acquisition and sli

Complementizers and subordination in typical language acquisitionand SLI

Maria Mastropavlou a,*, Ianthi Maria Tsimpli b

aDepartment of Linguistics, University of Ioannina, Ioannina, GreecebDepartment of Theoretical and Applied Linguistics, School of English, Aristotle University of Thessaloniki, Greece

1. Introduction

Although a great amount of research on typical and impaired language acquisition has reported findings on patterns ofemergence of a variety of morpho-syntactic features, little is known about the way Greek subordination emerges in childlanguage, especially in SLI grammars. Research findings from English and Hebrew typical acquisition data have shown thatsubordination first emerges in the use of complement clauses, while adverbial clauses follow and relative clauses are the lastto appear in child speech (Bloom et al., 1980; Dromi and Breman, 1986). This pattern has been accounted for in terms ofstructural complexity of embedding in combination with conceptual complexity and discourse requirements, wherebycomplement clauses are more accessible since they are selected by the verb. On the other hand, a number of studies haveindicated the reverse order, with relative clauses emerging before complements (e.g. Armon-Lotem, 2005). This pattern isexplained on account of the feature complexity of the embedded clause. Specifically, complement clauses require knowledgeof the subcategorization properties of the selecting verb, thus lexical information needs to be available apart from syntacticknowledge, which renders their acquisition more demanding.

Lingua 121 (2011) 442–462

A R T I C L E I N F O

Article history:

Received 1 May 2007

Received in revised form 2 November 2008

Accepted 27 September 2010

Available online 20 November 2010

Keywords:

Complementizers

Subordination

Language acquisition

Language disorders

SLI

A B S T R A C T

This study aims to report on the pattern of use of complementizers by Greek SLI children as

well as describe differences and/or similarities with patterns of emergence in typical

language development. The complementizers na (corresponding to the English infinitival

marker to), oti and pos (corresponding to the English that) and pu (introducing factive

complements and relative clauses) were investigated in spontaneous speech samples of 8

children with SLI and two control groups: 8 language-matched and 8 age-matched

children. The theoretical frameworks adopted are that of the Interpretability Hypothesis,

according towhich LF-interpretability plays a determining role in the acquisition of formal

features by SLI children, and Roussou’s account of the C domain in Greek. In linewith these

accounts, the child data was analysed with respect to feature specification, posing a

distinction between pu on the one hand, specified for the interpretable feature of

definiteness, and oti and pos on the other, while na holds a unique status, functioning as

marker of mood/modality and a clause-typing element. Additionally, the selectional

restrictions these complementizers impose on inflection were also investigated. The

results indicate that complementizers with low specification for LF- interpretability are

more sensitive in SLI, while their selectional properties are active, revealing a problem in

the morpho-phonological operation of spell-out rather than their lexical representation in

the children’s underlying grammar.

� 2010 Elsevier B.V. All rights reserved.

* Corresponding author. Tel.: +30 26510 05143.

E-mail addresses: [email protected] (M. Mastropavlou), [email protected] (I.M. Tsimpli).

Contents lists available at ScienceDirect

Lingua

journa l homepage: www.e lsev ier .com/ locate / l ingua

0024-3841/$ – see front matter � 2010 Elsevier B.V. All rights reserved.

doi:10.1016/j.lingua.2010.10.009

http://dx.doi.org/10.1016/j.lingua.2010.10.009

mailto:[email protected]

mailto:[email protected]

http://www.sciencedirect.com/science/journal/00243841

http://www.elsevier.com/locate/lingua

http://dx.doi.org/10.1016/j.lingua.2010.10.009

Greek data have revealed that adverbial (purpose) clauses appear along with some complement clauses, while relativeclauses seem to emerge earlier than they do in the studies mentioned above and precede temporal, causal and conditionalclauses (Stephany, 1995). This pattern requires a different explanation as it cannot be accounted for by reference to thedegree of structural complexity of these clause types alone. The acquisition of subordination is approached from a differentperspective in this paper: patterns of emergence and early use are associated with the morpho-syntactic properties ofsubordinators. Specifically, the use of subordination markers and especially complementizers is investigated in associationwith the formal features they encode. Use (versus omission) of complementizers is analyzed with respect to the encoding ofintrinsic LF-interpretable features they involve rather than the structural complexity of the clauses they introduce. Althoughone cannot completely disregard the factor of structural complexity, it is not seen as the sole determining factor affectingacquisition patterns in this paper.1

The study is placed within the theoretical framework which suggests that SLI development is affected by feature-interpretability at LF (Tsimpli, 2001; Tsimpli and Mastropavlou, 2007; Tsimpli and Stavrakaki, 1999). Since very little isknown about the way subordination emerges in SLI grammar, the acquisition patterns of Greek complementizers arepresented and analyzed with respect to their differences in feature-specification in terms of interpretability.

1.1. Theoretical background

1.1.1. Specific Language Impairment

Childrenwithspecific language impairment formahighlyheterogeneouspopulationwith respect to the characteristics theyexhibit in acquiring language. This heterogeneity has led to variability in the descriptions of the symptoms associatedwith thedisorder, which is why a consistent set of diagnostic markers has not been established as yet, while formal diagnosis of SLI isperformed on the basis of exclusion criteria – i.e. lack of pathological, neurological or psychosocial disorders to account for thelinguistic setbacks.What ismore, even greater controversy has been raised in researchers’ attempt to formulate an account onthe aetiology of the disorder, while the locus of the dispute lies in two fundamental questions: first, is SLI specific to language ordoes it affect cognitivedevelopment aswell?Andsecond, is languagedevelopment in SLI deviantordelayed?With respect to thefirstquestion, two major trends have been noted in the literature, namely the psychological approaches and the linguistic ones.Psychological approaches see SLI either as a result of a general perceptual deficit (e.g. Leonard, 1989; Tallal, 1976) or as theoutcome of problems related to phonological memory (e.g. Gathercole and Baddeley, 1990). However, adopting a modularapproach to language, the linguistic accounts are more relevant to the purposes of the present study.

Seven major linguistic accounts of the SLI aetiology, forming four theoretical trends, are identified in the literature. Insome of these accounts the disorder is seen as a representational deficit, causing problems to the children’s ability toformulate grammatical rules (Implicit Rule Formation hypothesis – IRF, Gopnik and Crago, 1991), establish agreementrelations (Missing Agreement account –MA, Clahsen, 1989), acquire morpho-syntactic features (Feature Blindness hypothesis– FB,Gopnik, 1990) or specifically those features that are LF-uninterpretable in Chomsky’s terms (Interpretability Hypothesis –IH, Tsimpli and Stavrakaki, 1999). Alternatively, accounts that argue for a computational deficit maintain that SLI affects thechildren’s language production system, impeding their ability to activate syntactic operations such as Movement(Representational Deficit of Dependent Relations – RDDR, van der Lely, 1994; van der Lely et al., 2011) or leading them to avoidcomputational complexity by leaving features or bundles of features optionally unpronounced (Jakubowicz, 2011; Jakubowiczand Roulet, 2007). Finally, the postulation that SLI constitutes a processing deficit has also been put forward, arguing thataffected children have limited abilities of processing morpho-syntactic input of low phonetic substance (Sparse Morphology

Hypothesis – SMH, Leonard et al., 1988). The accounts mentioned so far argue that SLI impedes on language acquisitioncausing deviant developmental patterns, contrary to accounts that view SLI as a developmental delay, rendering childrenunable to grow out of the Optional Infinitives stage, omitting thus tense and/or agreement markers (Extended Optional

Infinitives account – EOI and Agreement/Tense Omission Model – ATOM, Wexler et al., 1998).In a nutshell, three linguistic loci are proposed by the above mentioned approaches: the perceptual mechanisms

responsible for the analysis of grammatical information, the underlying representations of formal features, and thederivational component responsible for the application of grammatical operations. These accounts have failed to provide acomplete description of the problems SLI causes to language development, giving a wide range of characteristics, which,however, do not always coexist in the SLI population. This is attributed to the fact that most of these accounts have beenformulated on the basis of linguistic evidence from a single language, while cross-linguistic variation constitutes a seriouschallenge for most of them. However, some of them have been more successful in explaining contradictory findings fromdifferent languages more than others. For instance, the EOI and ATOM models have proven problematic in accounting forcross-linguistic differences due to the different status of optional infinitives in different languages, as well as because of thefact that tense is not affected by SLI in some languages as much as in others (de Villiers, 2003; Dromi et al., 2003; Clahsenet al., 1997; Clahsen and Dalalakis, 1999; Mastropavlou, 2006; Tsimpli and Stavrakaki, 1999, and others). On the other hand,Leonard’s sparse morphology hypothesis acknowledges great cross-linguistic variation, since SLI symptoms are highlydependent on the morpho-phonological characteristics of each language. This account lies at the opposite extreme as itimplies that there are no language areas that are crosslinguistically affected by SLI. Yet, despite the heterogeneity that

1 At the same time, we also acknowledge that factors other than the morphosyntactic properties of subordinators, such as factivity, can also affect

patterns of acquisition.

M. Mastropavlou, I.M. Tsimpli / Lingua 121 (2011) 442–462 443

characterizes SLI populations and the cross-linguistic variation evident in relevant research, there seem to be some areas thatare more vulnerable to the disorder in all languages.

1.1.2. The [5_TD$DIFF]language [6_TD$DIFF]system in Minimalism (Chomsky, 1995, 2001)

The approach adopted in this study is based on Chomsky’sMinimalist Framework and on the assumption that the languagesystem is comprised of twomain components, the lexicon and the computational system (Chomsky, 1995). The lexicon is wherelexical itemsalongwith their idiosyncraticproperties aswell as formaland intrinsic featuresare stored,while the computational

system is described as amechanism that generates linguistic expressions usingmaterial from the lexicon via operations such asMerge andAgree. Linguistic expressions are seen as pairs of representations at two interface levels, namely the Phonetic Form (PF)and the Logical Form (LF), the semantic interface. Each of these representation levels constitutes the interface with a differentdomain: PFwith the articulatory-perceptual system and LF with the conceptual-intentional system, part of central cognition. Thecomputational systemthus involvesprocesses that are relevant toPFonly– the PF component–or to LFonly– the LF component,while parts that are relevant to both PF and LF constitute the overt syntax of a language.

The linguistic expressions generated by the computational system encode formal features, drawn from the lexicon,whichcan be visible at PF, LF or both. In that sense, formal features can be ‘‘interpretable’’ or ‘‘uninterpretable’’ at each of the twointerface levels; PF-interpretability refers to the property of a formal feature being morpho-phonologically realised in aspecific language, while LF-interpretability refers to the relevance of that feature to the semantic/conceptual representation.It then follows that features that are relevant – or interpretable – at PF only (and are thus LF-uninterpretable) are subject tocrosslinguistic variation and are involved in syntactic operations, while LF-interpretable features are involved in building asemantic representation of a structure but do not trigger syntactic operations.

1.1.3. Interpretability Hypothesis and SLI

On the basis of a case[7_TD$DIFF] study on Greek SLI, Tsimpli and Stavrakaki (1999) first proposed that the distinction betweenfeatures that are interpretable at LF and features that are not determines the degree to which they are accessible indeveloping grammars. The Interpretability Hypothesis (IH) was initially put forward to account for language problemsobserved in SLI and provide an explanation for the crosslinguistic variation exhibited in SLI research, while later it wasrelated to findings from second language acquisition, successfully accounting for the linguistic performance of L2 learners ofGreek (Tsimpli, 2003; Tsimpli and Dimitrakopoulou, 2007). Specifically, the IH maintains that features that are LF-interpretable are privileged in learnability terms because, apart from their lexical entry in the linguistic lexicon, they are alsoassociatedwith semantic/conceptual features in themental lexicon; these features are, by definition, non-modular. Contraryto interpretable features, uninterpretable ones serve a purely morpho-syntactic function, with a possible spell-out at PF. Inthat sense they are only found within the language module, which is assumed to reduce their learnability (Tsimpli andMastropavlou, 2007). Assuming that SLI is a deficit specific to language, it is argued that children with SLI are characterizedby difficulties in analyzing morpho-syntactic input. When features in this input are purely morpho-syntactic in the sensethat they are not mapped onto conceptual features, SLI grammars face difficulties, whereas when features share propertieswith the semantic/conceptual interface, their emergence and acquisition is more target-like. The IH assigns a significant roleto the properties formal features have at the phonetic/phonological level (PF). Contrary to the SMH, the IH does not argue forprocessing or perceptual problems in SLI, but rather sees phonology as a resource where children turn to as a means ofcompensation for the underlying morpho-syntactic deficit that characterizes the disorder. Evidence of compensatorystrategies at the PF level used by childrenwith specific language impairment is provided in previous studies (e.g. Goad, 1998;Tsimpli and Stavrakaki, 1999), where children tend to alter the prosodic characteristics of their utterances in order to expressgrammatical features they are unable to express morpho-syntactically. Given the fact that PF interpretability also plays asignificant role in determining SLI performance, the phonetic realization of formal features has been argued to be responsiblefor the variation in the way children from different linguistic backgrounds acquire language (Mastropavlou, 2006; Tsimpli,2003). A good example is the feature of tense, which seems to pose serious problems to childrenwith SLI in English, whereasGreek children appear to struggle significantly less, if at all. This pattern has been attributed to the difference in the PFrealization of tense in the two languages, a fact that has also been empirically supported (see Mastropavlou, 2006).

The crosslinguistic variation observed in SLI performance does not necessarily render the disorder language-specific.Although it is true that different diagnostic markers have been proposed for different languages (i.e. tense and agreement inEnglish, clitics and determiners in Greek (Tsimpli and Stavrakaki, 1999), clitics for French (Jakubowicz et al., 1998; Hamann,2003)), the locus of the disorder could still lie on underlying linguistic representations but be masked due to thecrosslinguistic differences in the PF-realization of grammatical features.

1.2. Complementizers in Greek

1.2.1. A general description

Four major types of subordinate clauses are identified in Greek: complement clauses, adverbial clauses, conditionals andrelatives. Complement clauses are further distinguished between declarative and interrogative. Complement clauses arenormally introduced with five main markers: oti and pos (used in declarative complements, corresponding to the Englishcomplementizer that), pu (introducing factive complements of psychological verbs, equivalent to the English factive that), an(introducing interrogative complements, equivalent to English if or whether) and na (a Mood (subjunctive) marker

M. Mastropavlou, I.M. Tsimpli / Lingua 121 (2011) 442–462444

introducing verb complements frequently equivalent to infinitival or gerundive complement clauses in English) (Roussou,1994, 2006). Three of these elements, namely pu, an ‘if/whether’ and na also introduce non-complement clauses: pu alsointroduces relative clauseswhen no relative pronoun is present (corresponding to the English that used in relatives), while analso introduces conditional clauses. Finally, na seems to bear a unique status, as it can introduce adverbial clauses of purpose,cause and result (in the form of a prepositional complementizer ja na, corresponding to the English in order to), as well asmain clauses expressing modality with an imperative or hortative function.

Beginning with the complementizers oti and pos, they introduce declarative complement clauses as the examples belowillustrate:

(1) a. Ipe pos/oti tha erthi noris

said-he that will come-he early

‘He said that he would come early.’

b. Vlepo oti/pos dhulevis poli

see-I that work-you a lot

‘I can see that you work a lot.’

It should be noted, however, that oti and pos can be omittedwhen they denote opinion or belief and introduce object clauses ofverbs like nomizo ‘think’ and pistevo ‘believe’ inwhich cases thematrix verb can optionally follow the complement clause (2c):

(2) a. Nomizo oti lei psemata.

think-1sg that tell-3sg lies

‘I think that he is lying.’

b. Nomizo lei psemata.

think-1sg tell-3sg lies.

‘I think he is lying.’

c. Lei psemata, nomizo.

tell-3sg lies think-1st

‘He’s lying, I think.’

However, oti cannot be omitted if the clause it introduces complements verbs or expressions that denote knowledge orcertainty, as the following examples indicate:

(3) a. Ksero *(oti) lei psemata

know-1st (that) tell-3sg lies

‘I know (that) he is lying.’

b. Ime siguros *(oti) lei psemata

be-1sg sure (that) tell-3sg lies

‘I am sure (that) he is lying.’

Moving to the complementizer pu, it introduces factive complements of verbs of perception and emotional state, cause orpurpose, result or assumption and contrast (some indicative examples are given below). In these cases, the clause introducedby the complementizer pu expresses a true proposition encoding the feature of factivity (Christidis, 1986; Roussou, 1994;Varlokosta, 1994).

(4) a. Vlepis pu irtha stin ora mu telika?

see-2sg that came-1sg on time my after all

‘See [that] I came on time after all?’

b. Xerome pu se vlepo.

be glad-1sg that you see-1sg

‘I’m glad that I see you.’


As mentioned above pu also introduces relative clauses, which is actually its most frequent function:

(5) To pedhi pu idhes prin ine jos mu.

the child that saw-2sg before is-3sg son my

‘The child that you saw before is my son.’

The complementizer an ‘if/whether’ introduces indirect questions, following verbs of asking, wondering and knowing (see(6) below). Like English if, an also introduces conditional clauses (7):

(6) Me rotise an tha pao sto party.

me asked-3sg whether will go-1sg to the party

‘He asked me whether I would go to the party.’

(7) An dhjavasis, tha perasis tis eksetasis

if study-2sg(prf) will pass-2sg the exams

‘If you study, you will pass the exams.’

Finally, the complementizer na introduces clausal complements of verbs (8a) but also of adjectives (8b) or nouns (8c):

(8) a. Thelo na matho Agglika.

want-1sg na learn-1sg English.

‘I want to learn English.’

b. Itan etimos na fiji.

was-3sg ready-msc na leave-3sg

‘She was ready to leave.’

c. I epithimia tu na pai sto panepistimio. . .

The wish his to go-3sg to-the university

‘His wish to go to university. . .’

Na also introduces purpose adverbial (adjunct) clauses2 as in example (9) below:

(9) Agorasa manitaria ja na ftiakso pitsa.

bought-1sg mushrooms for na make-1sg pizza

‘I bought mushrooms (in order) to make pizza.’

The unique status of na lies in that it is a complementizer with a mood feature and, thus, functions as a subjunctive marker.

This inflectional property of na renders its characterization as a complementizer a controversial issue (cf. Agouraki, 1991;Philippaki-Warburton, 1992, 1994, 1998; Rivero, 1994; Roussou, 2000). As will be presented in the following section, the Cstatus of na can easily be accommodated in a Rizzi-type articulated C-structure (see Roussou, 2000).

Themood feature of na in certain contexts clearly encodesmodality,which justifies its appearance in root clauses (Roussou,2006; Tsangalidis, 2002). Na root clauses can be either declarative with imperative force (8a), interrogative (8b) or hortative(8c):

(10) a. Na fijis.

na go-2sg

‘(I want you) to go.’

b. Na fiγo?na go-1sg?

‘Can) I go?’

2 Other adverbial clauses of time and cause are introduced by temporal and causal connectives, e.g. otan, eno ‘when, while’, jati, epidhi ‘because’.We do not

investigate the use of these connectives in this paper.


c. Na efevge!

Na leave-Imp-Past-1sg

‘(I wish) he would leave.’

1.2.2. Feature specification and syntactic position of Greek complementizers

In what follows, we adopt Roussou’s (2000) analysis of Greek complementizers and elaborate on its implications in termsof the IH.

Assuming amodified version of the split-CP structure in (11) in the sense of Rizzi (1997), Roussou suggests that the Greekcomplementizers pu, oti and na can be shown to occupy different syntactic positions in the C domain:

(11) [Force [Topic/Focus [Fin(iteness) [IP. . .]]]]

According to Rizzi (1997), Force is the position where clause-typing features are found in the form of complementizers suchas the declarative that. Fin, on the other hand, is associated with inflectional features. Thus, complementizers occupying theFin position bear or are sensitive to inflectional properties of the clause they introduce.

Roussou’s analysis of the Greek complementizers is based on evidence from the different distribution of focus, topic andwh-phrases in clauses introduced by these complementizers but also from the differences in the intrinsic features borne byeach. Since the learnability claim advocated by the IH is based on differences in the feature specification of lexical items, weconcentrate on the set of features which characterize each of the complementizers studied here.

To start with, pu is the only complementizer that does not allow for nominalization, as opposed to na, an, and oti, as theexamples in (12) indicate (Roussou, 2006):

(12) a. To na dhulevis se sxolio dhen ine efkolo.

the na work-2sg in school not is-3sg easy

‘Working in a school is not easy.’

b. To an tha ertho i oxi dhe se afora.

the if will come-1sg or not not you concern-3sg

‘Whether I am coming or not doesn’t concern you.’

c. Ine apistefto to oti kerdhise to vravio.

is-3sg unbelievable the that won-3sg the prize

‘The (fact) that he won the prize is unbelievable.’

d. *Ine apistefto to pu kerdhise to vravio.

is-3sg unbelievable the that won-3sg the prize

This distinct characteristic of pu is attributed to the fact that it bears the feature [definiteness] when introducingcomplement clauses and, as such it is capable of nominalizing the clause by itself (Christidis, 1986). Oti, an and na, on theother hand, are not specified for [definiteness], which iswhy they require the definite article in order for the clause to acquireargument status. Furthermore, the complementizer an ‘if/whether’ is specified for the interpretable feature [wh], while oti

and pos are the only complementizers that bear no feature other than the categorial C (being clause-typing elements).The distinct nature of pu, which stems from its specification for [definiteness], is also shown through its use to express

factivity (Roussou, 1994; Varlokosta, 1994). In a comparison between factive complements with pu and oti, Roussou (2006)demonstrates this distinction:

(13) a. Thimame oti efije, ala bori ke na kano lathos.

‘I remember that he left, but I could be wrong.’

b. #Thimame pu efije, ala bori ke na kano lathos.

‘I remember him leaving, but I could be wrong.’

Although the semantic differentiation between the two examples is minor, the fact that (13b) constitutes a less acceptablesentence than (13a) shows that the requirement expressed by the oti clause can be cancelledmore easily than that expressedby the pu clause. This difference constitutes evidence for the claims that the complementizer pu contains the lexical featureof [definiteness]. Additionally, Roussou (2006) argues that, because oti is not lexically loaded with [definiteness], it can only


appear in factive complements when the requirement of factivity is fulfilled by the lexical value of the verb. In the case of pu,on the other hand, factivity is a feature of the verb and of pu itself. This lexical distinctiveness of pu allows its presence infactive complements that do not follow factive verbs:

(14) a. Fovame oti vrexi

‘I am afraid that it’s raining.’

b. Fovame pu vrexi

‘I am afraid because it’s raining.’

Here, the semantic difference of factivity between the two examples is quite clear. Example (14a) expresses a fear of asituation that is probably going on, while this situation is definitely going on in (14b).

Recall that pu is also the relative complementizer in Greek. Although relative clauses are not selected by the head noun,restrictive relatives share the property of rendering the modified noun specific. In this respect, it can be argued that it is notaccidental that pu is the complementizer used in Greek relatives given that it is the only C in Greek that bears the feature of[definiteness].

In her analysis of the structure of the CP domain, Roussou (2000) argues for a tripartite C-system, modifying Rizzi’s (1997)analysis and adjusting it to Greek. Specifically, Roussou proposes a C-system with three head positions: the highest headhosts the subordination feature (C) and is occupied by the complementizer pu, themiddle head hosts operator information forclause-typing (COP) and is occupied by the complementizers oti and an, while the lower C head hosts the Mood feature (CM)and is occupied by themodal particles na, as ‘let’s’ and tha ‘will/would’. However, na and as are differentiated from tha in thatthey raise to the COP position along with the complementizers oti and an, functioning thus as operators and spelling outclause-typing as well. Roussou’s proposed structure is presented below:

[TD$INLINE]

(15) [C pu[Topic/Focus [COP oti/an/na/as[Neg dhen/min[CM tha/t(na/as) [I cl+V…]]]]]]

[TD$INLINE] [Roussou, 2000]

As it is shown in this analysis, na bears a special status among Greek complementizers, since it grammaticalizes two distinctfeatures, namely [mood] – as a subjunctive marker – and [clause-typing] (Roussou, 2006). Its uniqueness also lies in that itemploys a distinct negation marker,min (in contrast to oti and pu, which make use of the negation dhen) as well as in that itcan also introduce matrix clauses.

Greek complementizers are also differentiated with respect to whether they impose any restrictions on the aspect of theembedded verb. Particularly, the complementizers oti, pu and an are always followed by a tense-independent verb form,while use of a tense-dependent form (i.e. nonpast, perfective) would be ungrammatical3:

(16) a. To pedhi ipe oti *pinasi

the child said-3sg that be hungry-3sg[nonpast, perfective]

b. Xerome pu *perasis kala

be-1sg glad that have-2sg[nonpast, perfective] fun

Na, on the other hand, is the only complementizer that introduces tense-dependent clauses, although tense-independentverb forms are also allowed (see also examples (13)–(14) above).Na therefore does not impose any selectional restrictions onthe inflection of the embedded verb, except under specific circumstances (see Holton et al., 1999, for a detailed description).In that sense, na can be specified as [+tense] in cases like (17a) below or [Øtense] in cases like (17b):

(17) a. Elpizo na efije.

hope-1sg na leave-3sg[past, perfective]

‘I hope he left.’

b. Thelo na peksume.

want-1sg to play-1pl[nonpast, perfective]

‘I want to play.’

3 The complementizer an ‘if’ allows for the verb-dependent form when it introduces a conditional clause.


Roussou (2000) provides the basic set of features that characterize and distinguish betweenGreek complementizers: pu isa subordinator, oti and an are clause-typing elements and na ismodal and clause-typing.Moreover, the [definiteness] featureof pumakes it more ‘specialized’ than the other complementizers (Roussou, 2000:80). In terms of interpretability and on thebasis of the preceding presentation, it can be proposed that oti is the least specified declarative complementizer being amorpheme introducing declarative (i.e. [�wh]) clauses. On the other hand, pu bears the interpretable feature of[definiteness] and an ‘if/whether’ is [+wh]. Finally, na has a special status, encoding the roles of a complementizer and amodality marker: it is the last property, i.e. the [mood/modality] feature which characterizes it in terms of interpretability.

1.3. Subordinators and complementizers in child language and SLI

Findings from English report that sentential complements (introduced with that) appear quite early in child speech,earlier thanwh- and if-complements (Diessel, 2004). However, these first instances are characterized as highly formulaic inthe sense that thematrix clauses used basically ‘‘function as epistemic markers, attention getters ormarkers of the illocutionary

force’’ (Diessel, 2004:90). However, research findings on the way Greek children acquire subordination are quite limited.Some evidence comes from Stephany (1995), who reports that children begin using purpose and object clauses with na ataround 1;10. At that time, frequent omissions of na occur, while by 2;4, object clauses appear with the verbs thelo ‘want’ andboro ‘can’ and purpose clauses with the verb pao ‘go’. Oti ‘that’ clauses appear later, at around 2;10, in reported speech asobject complements. Finally, adverbials are much rarer at this stage, indicating a slower emergence in the children’s speech.Similar findings are provided by Baslis (1996), who also describes the same order of emergence, which occurs between theages of 1;10 and 2;0.

A more recent analysis of the Stephany corpus4 with respect to the acquisition of complementizers revealed that puappears earlier than oti, but its first emerging instances involve relative rather than factive use (Mastropavlou, 2005). Na isthe first complementizer to appear and is first used in purpose, root and object complement clauses, while omissions at theage of 1;10 aremore frequent in object than in purpose clauses and less frequent in root na clauses. However, omissions of na– or a frequent use of the vocalic placeholder a – in root clauses have been frequently attested in early stages of languageacquisition, where instances of tense-dependent forms are encounteredwith a strongmodal reading (Hyams, 2001; Tsimpli,1996; Varlokosta et al., 1998). Finally, the same analysis of the Stephany corpus revealed absence of an complements till theage of 2;9 (Mastropavlou, 2005).

Analyses of different corpora also indicate similar results. Specifically, Katis and Nikiforidou (2003) report that pu is firstused in relative clauses and appears in complement clauses with the verbs kita ‘look!’ and thimase ‘remember-2sg’ at around2;1, while its use in pseudo-relatives with adverbial reference begins between 2;4 and 2;7. A different analysis of the samecorpus is given by Stambouliadou (2006), where pu appears earlier than oti ‘that’, although the two complementizers exhibitthe reverse pattern with respect to frequency of use. The emergence of pu is preceded by the use of a phonetically simplifiedsyllable (a) around the age of 1;9, which constitutes a precursor for the emergence of pu complements. As far as oti isconcerned, an instance of incorrect use is reported, which rather constitutes overuse in an inappropriate context (instead ofan ‘if/whether’). Finally, as far as accuracy of use is concerned, very few omissions of oti are reported, which, however, do notresult in ungrammatical sentences but rather constitute instances of acceptable drop of oti (Katis and Nikiforidou, 2005).

Summing up the above, it seems that na clauses are the easiest to acquire and emerge first, followed by pu-relative, pu-complement and finally oti-clauses. This pattern of emergence is in line with the predictions made under the IH hypothesis,as children seem to acquire first and more easily semantically richer elements than those with lower semantic import. Notethat na clauses are initially used in root contexts with a strong modal interpretation, in purposive adverbials or ascomplements of modal verbs. In this respect, early contexts of na encode the modal feature of this element rather than itscomplementizer function alone. In root contexts, modality is encoded in the use of na itself, whereas in the complements ofmodal verbs, na is mainly a complementizer with mood properties (subjunctive). Finally, in adverbial clauses the mostfrequent of which are purpose clauses, na is attached to the prepositional complementizer ja ‘for’ and acts as a marker oftense-dependency. In these cases, na lacks a strong modal reading but is phonologically cliticized onto the prepositionalcomplementizer.

Moving to SLI, evidence from acquisition patterns of complementation in SLI mainly comes from English. Specifically, SLIchildren are reported to use non-finite complement clauses (i.e. to-clauses) by age 5, while finite complements (e.g. that-clauses) emerge a bit later (Eisenberg, 2003). It has been attested, though, that SLI children tend to use the non-finite particleto later and at lower rates than their typically developing peers (e.g. Leonard et al., 1997), while omissions of thecomplementizer that are also reported in spontaneous speech samples, which are, however, at similar rates as thoseexhibited by typically developing children (Leonard, 1995).

More recent evidence indicates that SLI children tend to produce more ungrammatical complement clauses than theirunaffected peers, even in non-finite contexts (Owen and Leonard, 2006). Erroneous responses involve omitted arguments infinite-complement contexts as well as omission of agreement markers on embedded verbs, indicating difficulties with finitecomplements. Additionally, omissions of the non-finite complementizer to have been observed in English SLI, a rare findingin data from unaffected children. Finally, instances of substitution of to by a filler syllable schwa are also reported in both SLIand control data (Schuele and Dykes, 2005), while omissions of the complementizer that by SLI children significantly

4 The Stephany corpus is available through the CHILDES database (http://childes.psy.cmu.edu/).


http://childes.psy.cmu.edu/

exceeded those noted in their unaffected controls data (Owen and Leonard, 2006). What is more, evidence from Italianindicates late emergence of complement clauses, even later than relative clauses (Cipriani et al., 1998). Finally, a largenumber of studies have shown various kinds of difficulties in the acquisition of relative clauses by SLI children, eithermanifesting as a systematic omission of the relative complementizer (e.g. Hakansson and Hansson, 2000 for Swedish) orattributed to problems in assigning theta roles to moved constituents rather than a ‘‘structural deficit in embedding’’(Friedmann andNovogrodsky, 2007; Novogrodsky and Friedmann, 2006 for Hebrew) or functional category deficits (Schueleand Tolbert, 2001; Schuele and Nicholls, 2000 for English).

Coming to Greek, no evidence has been provided on the patterns of acquisition of complementation by children with SLI,which renders the present study of greater significance.

1.4. Specific purposes and predictions

The focus of this paper lies on the acquisition of Greek complementizers aiming at an analysis in terms of the features theyencode. A descriptive analysis of the subordination types encountered in the children’s speech is given, aiming to provide aninsight to the status of subordinators in typically developing and SLI grammars. Yet, it is a primary goal of the study toqualitatively compare the complement clauses used by SLI children to those used by unaffected controls, aiming to provide adescription of the way these elements are affected by the disorder and establish an account of its locus in the child languagesystem. This is attempted through analyses of the following:

Na-clauses, functioning either as root, complement or adverbial clauses. Target use as opposed to omission of the na

marker is analyzed, while omissions are investigated in root as opposed to embedded clauses;Pu-clauses, functioning as relative or factive complements. Use versus omission of the pu marker is analyzed;Oti-(that)clauses, functioning as verb complements, are analyzed. Target use as opposed to omission of oti is presented.Finally, we look into selectional restrictions of complementizers imposed on the inflection of the embedded verb.A number of predictions can be formulated in terms of the IH. First of all, it is predicted that the CP layer is present in the

SLI grammar, enabling children to handle Force,Modality and Finiteness. This is based on the assumption that the availabilityof structural positions is not affected as such, but only in relation to the interpretability of features of themorphemesmergedin these positions (Tsimpli, 2005). SLI children should be able to produce the correct finiteness markers on embedded verbswhen these are selected by the C used, indicating that C-I selection is intact in Greek SLI. Additionally, based on the IH, it ispredicted that all groups will showmore accurate use of the factive and relative complementizer pu compared to oti and pos

‘that’ complementizers, on account of their interpretability distinction, which makes the former more accessible than thelatter. Accordingly, it is expected that oti omissions will be significantly more frequent that pu omissions in obligatorycontexts in the speech of the SLI group, a difference not expected to bear the same statistical significance in the speech of thecontrol groups. Such an observation will provide support to the fact that semantically uninterpretable features pose greaterdifficulties to language impaired subjects than interpretable ones. Furthermore, although the difference in acquiringinterpretable and uninterpretable features has also been observed in typical language development, its presence in impairedacquisition has been claimed to be far more intense, a discrepancy which is also expected to characterize the results in thepresent study.

2. Method

2.1. The subjects

Three groups of children were recruited for the purposes of this study. One group of children diagnosed with specificlanguage impairment (SLI group) and two control groups, onematched to SLI based on chronological age (CA group) and onematched in language development (LD group).

2.1.1. SLI group

Eight (8) children aged between 4;2 and 5;9 comprised the SLI group. All the children had been formally diagnosed withdevelopmental language impairment by specialized staff in Speech and Language Therapy centers in Athens, Greece.Specifically, six of these children were recruited from the Centre of Community Mental Health of Kaesariani, while one ofthem was obtained through the Doxiadis Unit of Child Research in Athens. Finally, one child was recruited through afreelance clinician, who was willing to help by providing access to the child’s contact details.

All children were carefully selected so that they fulfilled the diagnostic criteria reported in the relevant literature for SLI(Leonard, 1998). Specifically, all selected children were of non-verbal intelligence that falls within the range of normal (i.e.scores >75 in standardised intelligence scales), while none of them exhibited any type of non-verbal disorders (i.e.pathological, neurological or anatomical abnormalities, hearing problems, psychological disorders [8_TD$DIFF], etc.). All of the childrenwere attending language intervention programmes, which however did not exceed 9 months and focused on phonologicalskills. A number of exclusion criteria were used to ensure the suitability and homogeneity of the group, such as severephonological problems and unintelligible speech, semantic-pragmatic disorders and therapy history longer than 10months.Finally, a screening process was used for the children of the SLI group with the Diagnostic Verbal IQ (DVIQ) test (Stavrakakiand Tsimpli, 1999), in order to obtain their language development profile.


2.1.2. Control groups

Two control groups were formed to facilitate the investigation of SLI performance; one group with eight (8) typicallydeveloping children matched to the SLI with respect to chronological age (CA group), and a group of eight (8) childrenmatched to the SLI with respect to language development (LD group). Note that pair-matching was used in the formation ofthe control groups.

The sixteen (16) children of the control groups were obtained through the private kindergarten Ta Zavolakia in Piraeus,Greece, and were selected based on the criterion that they would not present any language delay, learning disabilities,attention deficit disorder (ADD) or hyperactivity (HADD) syndromes. Their typical language development was ensuredthrough reports obtained from their teachers and the headmaster of the school andwas further established by their scores inthe DVIQ test.

The language-matched children were aged between 3;0 and 3;7, while the matching between SLI and LD children wascarried out on the basis of their scores in the morpho-syntax and sentence recall sections of the DVIQ test (production andcomprehension). The aim of this decision was to establish very strict matching, so that any differences between the twogroups’ performances would be adequate to lead to assumptions on the delay/deviance issue. [9_TD$DIFF]

Table 1 presents information on the three groups of the study.

2.2. Data collection and analyses

The data consists of speech samples collected in three individual sessions (one 40-min[10_TD$DIFF] and two 45-min[11_TD$DIFF] long each) witheach child. The first, 40-min[12_TD$DIFF] session was used for the administration of the DVIQ test, so that the children’s produced speechduring that timewas not included in the analysis. Therefore, 90 min[13_TD$DIFF] of spontaneous speechwas collected from each child foranalysis. The materials used to elicit natural speech included picture story books and toys suitable for children of preschoolage. The same materials were used in all sessions to ensure the comparability of the conditions under which each child wasinterviewed.

Data collectionwith the SLI children took place between October 2002 and February 2003, while the typically developingchildren were interviewed between October and December 2004. Conversations were tape-recorded and transcribed, whilethe children’s utterances were scrutinized during analysis in order to exclude stereotypical expressions and repetitions ofthe researcher’s utterances.

In analyzing the children’s utterances, the following elements were measured and included in the analyses: na clauses(complement, adverbial and matrix), pu clauses (both relative and factive), oti clauses (complement) and an clauses(complements and conditionals). Finally, within-groups analyses were carried out to investigate any differences in the waychildren use the complementizers studied and between-groups comparisons aimed to establish any differences between theexperimental and the two control groups with respect to frequency and accuracy of use.

3. Results

In this section, the use of embedded and especially complement clauses by the three groups of the study is presentedwithrespect to frequency of occurrence as well as accuracy of use. Comparisons within each group’s performance across thedifferent complementizers under investigation, follows in section 4.

3.1. Frequency of embedded clauses in the children’s speech

Beginning the description of the children’s use of subordination, a presentation of the numbers of the clauses analyzedfrom the children’s speech samples follows in[14_TD$DIFF] Table 2.

Table 1Matching between the experimental (SLI) and the two control groups (CA, LD).

CA SLI LD

Name Age DVIQ Name Age DVIQ Name Age DVIQ

Antonis 4;2 83 (80.6%) Leonidas 4;2 58 (58.3%) Sofianna 3;0 56 (54.4%)

Michalis 4;3 82 (79.6%) Stathis 4;4 30 (29.1%) Mando 3;2 30 (29.1%)

Aggeliki 4;6 82 (79.6%) Haris 4;6 50 (48.5%) Maria 3;0 47 (45.6%)

Labros 5;0 82 (79.6%) Stelios 5;1 44 (42.7%) Nikos 3;2 40 (38.8%)

Alexandra 5;5 83 (80.6%) Dimitris 5;2 65 (63.1%) Vasilis 3;2 62 (60.2%)

Thodora 5;7 81 (78.6%) Spiros 5;7 48 (46.6%) Ifijenia 3;2 44 (42.7%)

Thodoris 5;9 73 (70.9%) Kostas 5;9 52 (50.5%) Sotiris 3;5 51 (49.5%)

Giorgos L 6;0 78 (75.7%) Damianos 5;9 65 (63.1%) Giorgos G 3;7 66 (64.1%)

Mean 5;1 77.9% Mean 5;0 50.2% Mean 3;2 48.1%

SD 0.71 3.52 SD 0.64 11.44 SD 0.20 11.51


It is evident in Table 2 that the older control children produced the largest number of clauses (426), followed by the SLIgroup and the language-matched controlswho produced fewer relevant contexts than the other two groups. In fact, since thespeech samples collected from the children of the three groups were of exactly the same length and involved the same tasks,these differences can be indicative of the children’s skills to a considerable extent. At first glance, the above numbers couldindicate that the SLI children’s skills in producing subordinate clauses are comparable to those of their unaffected peers andseemingly more advanced than those exhibited by the younger controls, at least with respect to frequency of use ofembedded clauses. Yet, the fact that SLI scores are considerably widespread (SD = 41.24) reveals a performance of a highlyheterogeneous group, especiallywhen compared to the two control groups, whose frequency scores seemmore consistent. Itwould thus be implausible to reach any early assumptions on SLI children’s ability to produce subordinate structures, at leastat this point.

As far as the types of subordinate clauses used are concerned, all groups exhibited comparable patterns. Table 3 illustratesthe distribution of embedded clauses across the different types analyzed.

Evidently, complement clauses comprise the most frequent type of embedding used by all three groups. It is, however,important to note at this point that root na clauses have been excluded from this analysis as they could be masking thechildren’s actual skills in producing embedded clauses. Indeed, looking at the total number of clauses with the root naexcluded, it is evident that the SLI speech samples exhibited lower frequency of embedded clauses compared to the age-matched control group, while the number of subordinate clauses in the LD data is similar to that of the SLI group. Yet, furtheranalyses of the na clauses used by the three groups can be quite revealing.

3.2. Na clauses

It was shown in the previous section that complement clauses are the most frequently encountered contexts ofembedding used by all groups. Analyzing na clauses in particular also revealed that the use of na introducing verbcomplements was its most frequent use in the data.

According to [16_TD$DIFF] Table 4, all groups used more complement na clauses than root or adverbials. The preference is stronger forthe LD group, who produced significantlymore complement na clauses than root or adverbials (x2 = 9.48, df = 2, p < .01) andweaker for the SLI and CA groups, whose preference for complement clauses is not significant (SLI: x2 = 2.00, df = 2, p = n.s.;CA: x2 = 1.75, df = 2, p = n.s.).

Analysing the accuracy of the na clauses used by the three groups revealed a clear difference between SLI and controlgroups. The[17_TD$DIFF] graph (Fig. 1) illustrates the use of na clauses by the children in each group.

What is interesting in this graph is the relatively high omission rate of na in the SLI data, which reaches 32.8% of theobligatory contexts produced. Omissions are very rare in the LD data and absent from the CA results, revealing a clear

Table 3Clause types used by the three groups.

SLI LD CA

Compl 126 (56.2%) 154 (56.7%) 157 (44.6%)

Root na 124 30 81

Adv 65 (31.6%) 75 (24.2%) 108 (32.4%)

Relative 44 (11.3%) 29 (19.1%) 70 (20.4%)

Conditionals 5 (0.9%) 0 (0%) 10 (2.6%)

Total 240 258 345

Total (root na included) 364 288 426

Table 2Numbers of clauses analysed.

SLI LD CA

Child n Child n Child n

Leonidas 22 Sofianna 48 Antonis 58

Stathis 6 Mando 4 Michalis 33

Haris 33 Maria 26 Aggeliki 41

Stelios 17 Nikos 17 Labros 42

Dimitris 60 Vasilis 43 Alexandra 56

Spiros 125 Ifijenia 62 Thodora 48

Kostas 84 Sotiris 44 Thodoris 59

Damianos 17 Giorgos G 44 Giorgos L 89

Total n 364 Total n 288 Total n 426

Mean n 45.5 Mean n 36 Mean n 53.3

SD 41.24 SD 18.83 SD 17.14


difference between the experimental and control groups. Although use of the vocalic placeholder a is also observed in the SLIand LD speech – although quite rarely – but not in the CA data, na omissions seem to be themain type of error encountered inthe SLI speech. However, despite the high omission rate of the group, not all of the SLI children who were tested faced thesame difficulties, which is evident by looking at their individual rates of na omissions in obligatory contexts (Table 5).

It becomes clear that two of the eight SLI children omitted na in more than half of the obligatory contexts they used, onechild omitted na in 40% of contexts, while the omission rates of the other five children did not exceed 20%. This heterogeneitywithin the SLI group is also revealed by the high SD value, contrary to that of the LD group, where only two children omittedna in few instances. However, although SLI performance does not seem consistent enough to lead to reliable group results,between-group comparisons revealed a statistically significant difference (x2 = 19.55, df = 2, p < .000). Specifically, SLIomission rates were significantly higher than those of the two other groups (SLI-LD: U = .000, N1 = 8, N2 = 8, p < .001 two-tailed; SLI-CA: U = .000, N1 = 8, N2 = 8, p < .000 two-tailed), while the two control groups did not differ significantly fromeach other (U = 24.000,N1,N2 = 8, p = n.s.). These results shows that SLI children could be undergoing an extended stage of na-omission, a stage that unaffected children typically overcome by the age of 3. This reveals difficulties with thecomplementizer na faced by children with SLI, difficulties that persist until a much later stage than they do in typicallanguage development.

An analysis of the contexts where na omissions occurred also lead to some interesting observations.

Table 5Individual rates of na omissions.

SLI LD CA

n % n % n %

Leonidas 3/21 14.3% Sofianna 0/38 0% Antonis 0/41 0%

Stathis 4/5 80% Mando 0/1 0% Michalis 0/14 0%

Haris 24/28 85.7% Maria 0/19 0% Aggeliki 0/36 0%

Stelios 3/16 18.8% Nikos 0/17 0% Labros 0/30 0%

Dimitris 4/39 10.3% Vaggelis 0/41 0% Alexandra 0/39 0%

Spiros 6/101 5.9% Ifijenia 2/58 3.4% Thodora 0/28 0%

Kostas 6/77 7.8% Sotiris 1/35 2.9% Thodoris 0/32 0%

Damianos 6/15 40% Giorgos G 0/39 0% Giorgos L 0/65 0%

T/m 56/302 32.8% T/m 3/248 0.8% T/m 0/285 0%

SD 32.68 SD 1.47 SD 0

[()TD$FIG]

64,9

94,3 100

32,8

0,80

2,3 4,9 0

0%

20%

40%

60%

80%

100%

SLI LD CA

accu

racy

of

na

clau

ses

-%

a

OMIT

CORR

Fig. 1. Na clauses – accuracy of use.

Table 4Group means of na clause uses.

Compl Root Adv

SLI Mean % 43.4% 33.3% 21.4%

SD (%) 19.22 24.91 12.38

LD Mean % 61.5% 14% 24.5%

SD (%) 22.88 12.78 22.34

CA Mean % 38.1% 28.5% 33.4%

SD (%) 15.4 18.27 14.26


According to [18_TD$DIFF]the graph (Fig. 2), na was dropped in 37.3% of the complement clauses SLI children used, while omissionswere fewer in main clauses and even fewer in adverbials. The examples below illustrate the types of errors the childrenproduced.

(18) a. Complement clauses

Thelo *paro [thelo na to paro]

want-1sg take-1sg/prf [want na it take-1sg/prf]

‘I want to take it’ (Stathis, SLI)

b. Main clauses

*to valis eki [na to valis eki]

it put-2sg/prf there [na it put-2sg/prf there]

‘Put it there’ (Spiros, SLI)

c. Adverbial clauses

Vadhi to podhi tu *to tupitune

put-3sg the leg his it wipe-3sg/prf

‘He puts his leg (there) in order for them to wipe it’ (Harris, SLI)

[Vazi to podhi tu ja na tu to skupisune]

Although statistical analyses did not reveal significant differences among the clause types where omissions occurred (SLI:x2 = 3.31, df = 2, p = n.s.), the fact that omissions were more frequent in complements could be explained on account of thesemantic function of na. As stated in section 1.2.2, in complement clauses na serves as amoodmarker and it is underspecifiedfor [modality] contrary to its use inmain clauses, where a strongmodal reading is entailed. Thus, themore frequent omissionof na in complement vs. main clauses is expected. However, its less frequent omission in adverbials is not straightforwardlyexplained in terms of IH. This is because na is amoodmarker in purposive adverbials too, as in complement clauses. The basicdifference from complement clauses, however, is that in adverbials na is phonologically cliticized onto the prepositionalcomplementizer ja ‘for’. As such, na omissions in adverbials coincides with omission of the prepositional complementizer aswell, as all instances of na drop in adverbials also involved omission of ja. Consequently, omission of na in these contexts ismore likely to involve a phonological effect affecting both C elements.

Looking at the omission patterns exhibited by SLI and LD children, it is clear that language-matched controls omit na invery few instances and only in complement clauses, while SLI children seem to drop the least specified C in a systematicway.Finally, note that all instances of na omissions involved a tense-dependent verb form in the na clause, which is compatiblewith patterns observed in typical language acquisition, whereby tense-dependent verb forms signify the emergence ofembedding before the emergence of na. It therefore appears that in na-omission cases SLI children observe C-I selectionalproperties in the embedded clause showing appropriate aspect marking on the verb.

Since aspectual marking of the embedded verb expresses C-I selection, it is important to analyze the production of verbsin na clauses it terms of aspect. The graph (Fig. 3) illustrates the verb use in the three groups’ na clauses.

According to this graph, all groups showed preference for tense-dependent verb forms (i.e. non-past perfective), whileaspect errors are very rare in the SLI and LD data. All aspect errors involved using an imperfective verb form in clauses whichrequire a tense-dependent form, as the examples below indicate.

[()TD$FIG]

37.3SD:40.85

1,1 0

29.3SD:30.05

0 0

19.4SD:26.47

0 00

20

40

60

SLI LD CA

mea

n n

a o

mis

sio

ns

-%

Compl

Main

Adv

Fig. 2. Contexts of na omissions.


(19) Inappropriate aspect:

a. na t’ #aniγo? (ta matia mu) [na t’ anikso?]

na them open-1sg/imprf [to them open-1sg/prf]

‘Can I open my eyes?’ (Stavros, SLI)

b. na #klino pali matia? [na kliso pali matia?]

na close-1sg/imprf again eyes [to close-1sg/prf again eyes]

‘Should I close my eyes again?’ (Stavros, SLI)

c. ali fora boro na #skafalono ta ksila [. . .na skarfaloso. . .]

other time can-1sg na climb-1sg/imprf the woods [. . .to climb-1sg/prf. . .]

‘Next time I can climb the wooden stairs’ (Dimitris, SLI)

d. na sikothis ke na #pianis, ala na mi s’akusi i mama mu

na get up-2sg/prf and na get-2sg/imprf but don’t let my mom hear you

‘No, get up and get it, but don’t let my mom hear you’ (Dimitris, SLI)

e. to silo to afise na #pefti sti thalasa [. . .na pesi. . .]

the stick it leave-3sg/pst na fall-3sg/imprf in the sea [. . .to fall-3sg/prf. . .]

‘He left the stick to fall into the sea’ (Vaggelis, LD)

All the above examples involve instances where a perfective verb would be appropriate, although the use of an imperfectiveform is not ungrammatical. One instance of verb omission is given in (20) below.

(20) Verb omission

e, ato na eki [na to valume eki]

this na there [to it put-1pl/prf there]

‘Let’s put this there’ (Stavros, SLI)

Looking at the examples in (19), it seems that instances of inappropriate aspect mainly involvedmatrix na clauses (19a, b, d)rather than complement or adverbial clauses, which, however, did not prove true after further analysis. The table belowpresents group mean percentages of verb use in complement, root and adverbial na clauses.

As shown in[19_TD$DIFF] Table 6, aspect errorswere almost equally frequent in complement and root na clauses for the SLI group (3.4%and 4% respectively), butmore frequent in root na clauses for the language-matched controls. As for the children’s preferencefor tense-dependent verb forms, it seems stronger in root clauses for SLI and the age-matched controls, although language-matched children exhibit stronger preference for perfective verbs in complement clauses. However, statistical analyses didnot reveal a significant effect of clause type in the children’s preferences (SLI: x2 = 4.33, df = 2, p = n.s.; LD: x2 = 1.13, df = 2,p = n.s.; CA: x2 = 3.60, df = 2, p = n.s.). Finally, a comparison between the SLI and LD error rates revealed that the two groupsdid not differ significantly from each other (U = 27.500, N1,2 = 8, p = n.s.), suggesting that C-I selection is not affected by SLI.

[()TD$FIG]

90,597,4

93,2

4,4 1,16,83,5 1,5 01,6 0 0

0

20

40

60

80

100

SLI LD CAN

a ve

rb -

%

Dep

Indep

Indep.Inc.

V omitted

Fig. 3. Use of verbs in na- clauses.


3.3. Pu-clauses

As far aspu clauses are concerned, obligatory contextswere fewer than those ofna clauses in the speechof all groups. The24children used 171 pu clauses in total, a number that includes both relative and complement clauses. The older control groupproduced 82 clauses, more than the other two groups, while SLI children produced 51 obligatory contexts for pu complementand relative clauses. The language-matched children produced the fewest number of clauses, which did not exceed 38. Allgroups’ clauses were mostly relatives, while factive complements with pu were rarer in their speech, as [20_TD$DIFF]Table 7 presents.

Table 7 shows that relative pu clauses were preferred by all children, although the SLI group seems to have the highestmean preference (32.3%). However, this score is not informative due to the very high variability in the children’s scores, asthe SD value reveals (46.82). Indeed, five of the eight children seemed to use relative clauses in more than 70% of the pu

contexts in their speech, while only two children used one factive complement as opposed to no relative clauseswhatsoever.On the other hand, the LD and CA performances seem more consistent, which was also established through statisticalanalyses that revealed a significant preference for relative over complement clauses for these two groups (LD: z = �2.21,p < .05 two-tailed; CA: z = �2.39, p < .05 two-tailed) but not for the SLI group (SLI: z = �1.71, p = n.s. two-tailed). Looking atthe raw numbers of factive complements used by each group it is shown that the SLI group overall produced the fewest pucomplements than the two control groups (7 as opposed to 9 and 12 respectively). In fact, statistical analyses revealed a

Table 6Use of verb form in Complement, Main and Adverbial na clauses.

Dep Indep. Indep.Inc. V omit

SLI Compl m 87.7% 8.9% 3.4% 0%

SD 15.92 10.20 6.65 0

Root m 91.8% 1% 4% 3.2%

SD 16.17 1.89 8.32 8.4

Adv m 96.8% 2.5% 0.7% 0%

SD 7.06 7.07 1.96 0

LD Compl m 97.7% 1% 1.3% 0%

SD 3.65 1.96 3.54 0

Root m 94.8% 0% 5.2% 0%

SD 8.51 0 8.51 0

Adv m 98.8% 1.2% 0% 0%

SD 3.15 3.15 0 0

CA Compl m 86.6% 13.4% 0% 0%

SD 19.79 19.79 0 0

Root m 97.7% 2.3% 0% 0%

SD 4.27 4.27 0 0

Adv m 96.9% 3.1% 0% 0%

SD 6.2 6.2 0 0

Table 7Pu clauses used by the three groups.

SLI LD CA

Rel Fact Rel Fact Rel Fact

Leonidas 1

100%

0

0%

Sofianna 7

77.8%

2

22.2%

Antonis 13

86.7%

2

13.3%

Stathis – – Mando 3

100%

0

0%

Michalis 11

100%

0

0%

Haris 5

100%

0

0%

Maria 2

33.3%

4

66.7%

Aggeliki 3

100%

0

0%

Stelios 0

0%

1

100%

Nikos 2

100%

0

0%

Labros 4

80%

1

20%

Dimitris 15

78.9%

4

21.1%

Vaggelis 1

50%

1

50%

Alexandra 1

25%

3

75%

Spiros 18

94.7%

1

5.3%

Ifijenia 4

100%

0

0%

Thodora 9

81.8%

2

18.2%

Kostas 5

100%

0

0%

Sotiris 6

85.7%

1

14.3%

Thodoris 17

94.4%

1

5.6%

Damianos 0

0%

1

100%

Giorgos G 4

80%

1

20%

Giorgos L 12

80%

3

20%

Total n 44 7 Total n 29 9 Total n 70 12

Mean % 67.7% 32.3% Mean % 78.4% 21.6% Mean % 81% 19%

SD 46.82 SD 24.76 SD 24.12


significant difference between SLI and CA preference rates for factive complements (U = 11.000, N1 = 7, N2 = 8, p < .05 two-tailed), indicating that SLI children have not yet acquired this function of pu to the same extent as their unaffected age-controls. Yet, no significant difference was revealed between SLI and LD (U = 16.500, N1 = 7, N2 = 8, p = n.s.) nor between thetwo control groups (U = 30.000, N1,2 = 8, p = n.s.). The pattern seen in the control groups’ results is consistent with therelevant literature, which traces the acquisition of factivity around the age of 4;0 (Abbeduto and Rosenberg, 1985; Leger,2008; Schulz, 2003). Yet, the fact that SLI children appear to produce very few factive pu complements indicates a delay inacquiring the notion of factivity rather than difficulties in producing pu complements as such.

With regard to accuracy of pu-use, all groups exhibited no notable difficulties producing the subordinator in obligatorycontexts. Fig. 4 illustrates the average accuracy rates of the three groups.

Only two instanceswhere puwas omittedwere found in the SLI speech data, whichwere both produced by the same childand occurred in relative clauses:

(21) a. oxi ato, *ato exume meta [. . .afto pu exume mesa]

not this, this have-1pl inside [. . .this that have-1pl inside]

‘Not this one, the one that we have inside.’ (Haris, SLI)

b. edo ine *ena katikaki evale kerata [. . .ena katsikaki pu evale. . .]

here be-3sg a goat put-3sg/pst horns [. . .a goat that put-3sg/pst. . .]

‘Here is a goat that put on horns.’ (Haris, SLI)

Apart from these two instances, all children seemed to make appropriate use of pu clauses, relative and factive, while nodifferences were revealed between groups (x2 = 2.29, df = 2, p = n.s.). Finally, all pu clauses produced by the children of thethree groups included target verb forms, that is, tense-independent forms (imperfective).

3.4. Oti clauses

Moving to oti clauses, obligatory contextsweremuch fewer for the SLI group and their language-matched controls but notfor the older control children. Only 4 oti clauses in the SLI data and only 2 in LD,while 26 oti clauseswere used by the childrenof the CA group. This pattern reveals a distinction between the SLI and LD groups on the one hand and the CA group on theother. In other words, oti complements seem to develop later in typical language acquisition, while SLI children also appearto make very conservative use of this type of complementation. In fact, analyzing accuracy of using these clauses lead tosome interesting observations:

As it appears in [23_TD$DIFF]the graph (Fig. 5), SLI children exhibited low accuracy scores in producing oti clauses. The complementizerotiwas omitted in three out of the four clauses used, contrary to the two control groups, where all clauses produced reflectedtarget use. The examples below present the oti omissions that occurred in SLI data:

(22) a. Investigator: aftos lei oti lipate. Ki aftos?

‘He says he is sorry. And that one?’

SM: #aftos lei ponai [. . .lei oti ponai]

he say-3sg hurt-3sg [. . .say-3sg that hurt-3sg]

‘He says he is hurting.’

[()TD$FIG]

94,3

100100

20

40

60

80

100

SLI LD CA

Co

rrec

t -

%

Pu use

Fig. 4. Pu use.


b. *kseri tha fiji [kseri oti tha fiji]

know-3sg will go-3sg [know-3sg that will go-3sg]

‘He knows (that) he will leave.’ (Kostas, SLI)

c. edho dhe *tseri epese [. . .dhen kseri oti epese]

here not know-3sg fall-3sg/pst [. . .not know-3sg that fall-3sg/pst]

‘Here, he doesn’t know (that) he fell.’ (Kostas, SLI)

As shown in (22), two of the three omissions of oti occurred in obligatory contexts (22b and c), while (22a) does not somuchconstitute a grammatical error as an unacceptable or unnatural utterance. Although the very small number of contexts limitsthe generalizeability of this finding, the fact that oti was omitted in 3 out 4 times by SLI children as opposed to 100% targetusemade by both groups of typically developing children is indicative of a distinction between the experimental and controlgroups. In fact, the control groups’ results are consistentwith findings in the relevant literature (Katis andNikiforidou, 2005),contrary to SLI patterns that involve omissions of oti resulting in ungrammatical or unnatural sentences, patterns that havenot been noted in typical language acquisition. This distinction is likely to be attributed to the limited semantic import of thecomplementizer oti compared to na and pu, a possibility that will be discussed in section 4. Finally, it is worth noting that alloti verbs were accurately used by all children, namely in a tense-independent form.

3.5. An clauses

An clauses were the rarest type of subordinate clauses encountered in the speech samples of all children. Obligatorycontexts were few, including both conditional and Wh functions, as [24_TD$DIFF] Table 8 [25_TD$DIFF]reveals:

Oncemore, the older controls produced the highest number of obligatory contexts, while only two of the children did notproduce any an clauses. Their clauses were mostly conditionals, while five an complements were used by four of thechildren:

(23) a. theli na dhi an ine ta pedhakja tu

want-3sg to see-3sg if be-3pl the children his-gen/msc

‘He wants to see if those are his children.’ (Antonis, CA)

b. tha perimeno na dho an ine kalos mathitis

will wait-1sg to see-1sg if be-3sg good student

‘I will wait and see if he is a good student.’ (Giorgos G, CA)

[()TD$FIG]

1 2

26

3 0

0

0

5

10

15

20

25

30

SLI LD CAsu

m o

ti u

se -

n

Omit

Corr

Fig. 5. Accuracy of oti clauses.

Table 8An clauses: Conditional vs. Wh.

Cond Wh Total

SLI 5 (2a) 0 5 (2a)

LD 0 1 1

CA 10 (5a) 5 (4a) 15 (6a)

a Indicates the number of children who contributed to these numbers.


In contrast, only two of the eight children in the SLI group produced an clauses, which were all conditionals rather thancomplements:

(24) a. an paris tin kokini xanis

if pick up-2sg/prf the red lose-2sg/imprf

‘If you pick up the red one, you lose.’ (Dimitris, SLI)

b. an ixa ki ala tuvlakja. . .

if have-1sg/pst and other bricks

‘If I had more bricks. . .’ (Stavros, SLI)

Finally, only one an clause was found in the language-matched control data, which was a complement clause and is given in(25) below:

(25) dhen iksere an tha ’rthi

not know-3sg/pst if will come

‘He didn’t knowwhether she would come.’ (Sotiris, LD)

All verbs in an clauses were accurately used by all groups, either in the perfective or imperfective aspect.

4. Discussion

It is evident from Table 9, which summarizes the frequencies of use of each of the complementizers examined, that na isthe most frequently used complementizer by all three groups.

This is to be expected since in Greek na is found in more contexts than any of the other complementizers, namely main,complement and adverbial clauses. Inmore formal terms, the frequent occurrence of na is attributed to its status as a clause-typing element and a mood particle. With regard to the other complementizers, the SLI and LD group seem to show similarfrequency of use while the older children are more productive.

Accuracy of use is the aspect of children’s performance relevant to the predictions made by the IH. As shown by Table 10,pu (as a complementizer) is used 100% accurately by all groups whereas oti seems to be least accurate in the SLI group whenboth control groups exhibit target use:

Although the number of obligatory contexts for use of oti is small, it is important to note that omission rather than use isthe tendency for SLI but not LD children. As far as na omission is concerned, recall that the distribution of such cases is not thesame for adverbial, complement andmatrix clauses. Complement clauses, inwhich na is amoodmarker and does not conveya strong modal reading, are the contexts where omission is most frequent. On the other hand, adverbial clauses show thehighest accuracy rates for na, a fact which does not seem consistent with the IH, given that in these contexts too, use of theparticle is primarily inflectional, i.e. as a mood marker of a tense-dependent clause. It was argued, however, that it isthe phonological cliticization of na on the prepositional complementizer ja, which leads to lower omission rates in

Table 10Accuracy of complementizer use by the three groups.

SLI LD CA

na 98/115a 139/142a 114/114a

pu 7/7 12/12 12/12

oti 1/4 2/2 26/26

an – 1/1 15/15

a Note that these numbers involve sums of raw numbers rather than mean performances, as the percentages in the graphs in section 3 represent.

Table 9Frequency of complementizers in the three groups’ data.

SLI LD CA

na 115 142 114

pu 7 9 12

oti 4 2 26

an 0 1 5


adverbials, rather than the feature-specification of C elements. Indeed, it was shown that jawas always dropped along withna by SLI children, confirming the above claim.5

Furthermore, use of the [definite] complementizer pu is target-like in all groups’ data. The specialized function of thiscomplementizer which is encoded in its interpretable [definiteness] feature is consistent with the predictions of the IH.Finally, the pattern of use of an complements noted in the control groups’ data is in line with previous findings that haverevealed late emergence of wh and if complements both in English (Diessel, 2004) and Greek (Mastropavlou, 2005). At thesame time, the absence of these clauses from the SLI speech was predictable given their general performance; it is alsoindicative of delayed development.

Note that in the data presented, use/omission of these complementizers does not seem to affect C-I selectionalrestrictions. Thus, errors in the use of appropriate verb forms selected by the complementizer na are rather few even in theSLI group. This leads to a paradoxical situation where the complementizer may be omitted and, hence, not merged in thesyntactic position, whereas its selectional restrictions are still operative. This is particularly relevant to the omission cases ofna which, as mentioned above, is the only complementizer which can introduce tense-dependent verb forms, i.e. the non-past, perfective form. Recall that the number of errors in the choice of verb form is very small and these errors are only foundin na clauses (and not in oti clauses where C is omitted). Moreover, the dependent form is only used in na clauses and not,incorrectly, with any of the other complementizers. Wemust, therefore, conclude that even in the case of omission, childrenknow the selectional properties imposed by C and fail to access or spell-out the required complementizer. In this respect, theproblem is not in the lexicon but in whatever operation is assumed to dictate spell-out of the appropriate C in the rightclausal context.6 Thus, LF-interpretability seems to constrain morphological spell-out depending on the choice of thecomplementizer, without, at the same time, sacrificing selectional restrictions in the verb form used. If this suggestion is onthe right track, it follows that spell-out constraints in developing grammars (pathological or not) follow an economyhierarchy (partly) regulated by semantic properties as well.

5. Conclusion

The aim of this paper was to provide an account on the way Greek SLI children acquire complementizers with respect tothe realization of these elements at LF. Given the claim that SLI impedes the acquisition of LF-uninterpretable features, it washypothesized that the semantic import of Greek complementizers would be manifested in their acquisition patterns. It wasshown that all groups produced na clausesmore frequently than oti, an and pu clauses, whichwas attributed to its function asa mood/modality and clause typing marker, apart from the fact that it occurs in more contexts than the othercomplementizers studied. It was also shown that all children exhibited target use of pu, contrary to oti, which SLI childrenomitted in [28_TD$DIFF]3/4 of the obligatory contexts they produced. This difference is in line with the predictions made under the IHhypothesis, based on the distinction between these two complementizers in terms of LF-interpretability. Similarly, SLIchildren appeared to face notable difficulties with the complementizer na, dropping it in obligatory contexts significantlymore than both control groups, indicating that the least specified use of na is more vulnerable in their grammar. Finally,along with the distinction between complementizers that encode interpretable features and those that do not, anotherimportant finding was that SLI does not seem to impinge on the children’s knowledge of the selectional restrictions eachcomplementizer imposes on the embedded verb.

Based on the above findings, it follows that SLI does not cause problems to the representation of C elements in the lexicon,as their properties seem to be active in their grammar. Additionally, the fact that no overgeneralization of complementizersin inappropriate contexts was noted in the SLI data also indicates that the problem does not lie in the representation of theseelements in the children’s grammar. It is therefore more plausible to claim that those C elements that are less specified interms of LF interpretability aremore sensitive at spell-out in the sense that they are optionally used by SLI children to amuchgreater extent than is seen in typical language acquisition.

Acknowledgements

Special thanks go to the speech therapists Ms Maria Vlassopoulou, Ms Hara Karamitsou and Ms Maria Diamanti forproviding access to SLI children and assisted with the whole process. We would also like to thank all the children, theirparents and teachers who provided their consent and assisted with the data collection. Finally, we are grateful to the twoanonymous reviewers and to Dimitra Katis, for their insightful comments and suggestions.

5 An anonymous reviewer noted that itmight be the semantic import of ja that posits fewer omissions of na in adverbials. However, we do not think that it

is the semantic content of ja in adverbials that holds na in place, since ja is not obligatory in purposive clauses in the first place:

e.g. Efere to vivlio (ja) na mu to dhiksi.brought-2sg the book (ja) na me it show-3sg/prf‘He brought the book (in order) to show it tome.’

However, when ja is dropped in adverbial clauses, its semantic load seems to be transferred on the particle na, which in these instances indicates the notion

of purpose.6 This could mean that morphological representations of these elements are intact and fully active at a morphological structure level but fail to gain

substance at PF (Halle and Marantz, 1993).


References

Abbeduto, L., Rosenberg, S., 1985. Children’s knowledge of the presuppositions of know and other cognitive verbs. Journal of Child Language 12, 621–641.Agouraki, Y., 1991. A Modern Greek complementizer and its significance for UG. UCL Working Papers in Linguistics 3, 1–24.Armon-Lotem, S., 2005. The acquisition of subordination: from preconjunctionals to later use. In: Ravid, D.D., Bat-Zeev Shyldkrot, H. (Eds.), Perspectives on

Language and Language Development: Essays in Honor of Ruth A. Berman. Kluwer, Dordrecht, pp. 191–202.Baslis, I.N., 1996. H αnptyjƞ toy ypotαktikoy� loγoy stƞ γlv�ssα tvn pαiδiv�n ƞlikıαz 0-36 mƞnv�n [The development of subordinate speech in the

language of children aged 0–36 months]. Studies in Greek Linguistics 39, 24–34.Bloom, L., Lifter, K., Hafitz, J., 1980. Semantics of verbs and the development of verb inflection in child language. Language 56, 386–412.Chomsky, N., 1995. The Minimalist Program. MIT Press, Cambridge, MA.Chomsky, N., 2001. Beyond Explanatory Adequacy. MS., MIT.Christidis, A-Ph., 1986. To morwƞmα ‘poy’ sαn αnαworikoz δeıktƞz [The pu morpheme as a definite clause nominaliser]. Studies in Greek Linguistics 7,

135–148.Cipriani, P., et al., 1998. A longitudinal perspective on the study of specific language impairment: the long term follow-up of an Italian child. International

Journal of Language and Communication Disorders 33, 245–280.Clahsen, H., 1989. The grammatical characterisation of developmental dysphasia. Linguistics 27, 897–920.Clahsen, H., Bartke, S., Gollner, S., 1997. Formal features in impaired grammars: a comparison of English and German SLI children. Essex Research Reports in

Linguistics 14, 42–75.Clahsen, H., Dalalakis, J., 1999. Tense and agreement in Greek SLI: a case study. Essex Reports in Linguistics 24, 1–24.de Villiers, J.G., 2003. Defining SLI: a linguistic perspective. In: Levy, Y., Schaeffer, J. (Eds.), Language Competence across Populations: Toward a Definition of

Specific Language Impairment. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 425–447.Diessel, H., 2004. The Acquisition of Complex Sentences. Cambridge University Press, Cambridge.Dromi, E., Breman, R.A., 1986. Language-specific and language-general in developing syntax. Journal of Child Language 13, 371–387.Dromi, E., Leonard, L.B., Blass, A., 2003. Differentmethodologies yield incongruous results: a study of the spontaneous use of verb forms in Hebrew. In: Levy,

Y., Schaeffer, J. (Eds.), Language Competence across Populations: Toward a Definition of Specific Language Impairment. Lawrence Erlbaum Associates,Mahwah, NJ, pp. 273–290.

Eisenberg, S.L., 2003. Production of infinitival complement clauses in the conversational speech of 5-year-old children with language impairment. FirstLanguage 23, 327–341.

Friedmann, N., Novogrodsky, R., 2007. Is themovement deficit in syntactic SLI related to traces or to thematic role transfer? Brain and Language 101, 50–63.Gathercole, S.E., Baddeley, A.D., 1990. Phonological memory deficits in language disordered children: Is there a causal connection? Journal of Memory and

Language 29, 336–360.Goad, H., 1998. Plurals in SLI: Prosodic deficit or morphological deficit? Language Acquisition 7, 247–284.Gopnik, M., 1990. Feature blindness: a case study. Language Acquisition 1, 139–164.Gopnik, M., Crago, M.B., 1991. Familial aggregation of a developmental language disorder. Cognition 39, 1–50.Hakansson, G., Hansson, K., 2000. Comprehension and production of relative clauses: a comparison between Swedish impaired and unimpaired children.

Journal of Child Language 27, 313–333.Halle, M., Marantz, A., 1993. Distributed Morphology and the pieces of inflection. In: Hale, K., Keyser, J. (Eds.), The View from Building 20. MIT Press,

Cambridge, MA, pp. 111–176.Hamann, C., 2003. Phenomena in French normal and impaired language acquisition and their implications for hypotheses on language development. Probus

15, 91–122.Holton, D., Mackridge, P., Philippaki-Warburton, I., 1999. Greek: A Comprehensive Grammar of the Modern Language. Routledge, London.Hyams, N., 2001. Now you hear it, now you don’t: the nature of optionality in child language. In: Do, A.H.-J., Dominguez, L., Johansen, A. (Eds.), BUCLD 25:

Proceedings of the 25th annual Boston University Conference on Language Development. Cascadilla Press, Sommerville, MA, pp. 34–58.Jakubowicz, C., 2011. Measuring derivational complexity: New evidence from typically developing and SLI learners of L1 French. Lingua. 121, 339–351.Jakubowicz, C., et al., 1998. Determiners and clitic pronouns in French-speaking children with SLI. Language Acquisition 7, 113–160.Jakubowicz, C., Roulet, L., 2007. Narrow syntax or interface deficit? Gender agreement in French SLI. In: Goodluck, H., Liceras, J., Zobl, H. (Eds.), The Role of

Formal Features in Second Language Acquisition. Routledge, London, pp. 184–225.Katis, D., Nikiforidou, V., 2003. Hpolysƞmıα toy δeıktƞpoy:symβolZαnαptyjiαkv�n δeδomԑnvnstƞn ejZγƞsZ tƞz [The polysemy of the subordinator

pu: a contribution of developmental data in its explanation]. Proceedings of the 6th International Conference in Linguistics, University of Crete, E-book.Katis, D., Nikiforidou, V., 2005. The development of complement clauses in Greek. Paper presented at the X. International Conference of the Association for

Research in Child Language, Berlin.Leger, C., 2008. The acquisition of two types of factive complements. In: Gavarro, A., Freitas, M.J. (Eds.), Language Acquisition and Development:

Proceedings of GALA 2007. Cambridge Scholars Publishing, Cambridge, pp. 337–347.Leonard, L.B., 1989. Language learnability and specific language impairment in children. Applied Psycholinguistics 10, 179–202.Leonard, L.B., 1995. Functional categories in the grammars of childrenwith specific language impairment. Journal of Speech andHearing Research 38, 1270–

1283.Leonard, L.B., 1998. Children with Specific Language Impairment. MIT Press, Cambridge, MA.Leonard, L.B., et al., 1997. Three accounts of the grammatical morpheme difficulties of English-speaking children with specific language impairment,

Journal of Speech. Language, and Hearing Research 40, 741–753.Leonard, L.B., et al., 1988. Some influences on the grammar of English- and Italian-speaking children with specific language impairment. Applied

Psycholinguistics 9, 39–57.Mastropavlou, M., 2005. Greek markers of subordination and complementation in language acquisition. Unpublished manuscript.Mastropavlou, M., 2006. The Role of Phonological Salience and Feature Interpretability in the Grammar of Typically Developing and Language Impaired

Children. Unpublished Ph.D thesis, Aristotle University of Thessaloniki.Novogrodsky, R., Friedmann, N., 2006. The production of relative clauses in SLI: a window to the nature of the impairment. International Journal of Speech-

Language Pathology 8, 364–375.Owen, A.J., Leonard, L.B., 2006. The production of finite and nonfinite complement clauses by childrenwith specific language impairment and their typically

developing peers. Journal of Speech, Language, and Hearing Research 49, 548–571.Philippaki-Warburton, I., 1992. On mood and complementizers in Modern Greek. Reading University Working Papers in Linguistics 1, 5–41.Philippaki-Warburton, I., 1994. Verb movement and clitics in Modern Greek. In: Philippaki-Warburton, I., Nikolaidis, K., Sifianou, M. (Eds.), Themes in

Greek Linguistics. Papers from the First International Conference on Greek Linguistics, Reading September. John Benjamins, Amsterdam, pp. 53–60.Philippaki-Warburton, I., 1998. Functional categories and Modern Greek syntax. The Linguistic Review 15, 159–186.Rivero, M.L., 1994. Clause structure and v-movement in the languages of the Balkans. Natural Language and Linguistic Theory 12, 63–120.Rizzi, L., 1997. The fine structure of the left periphery. In: Haegeman, L. (Ed.), Elements of Grammar: Handbook of Generative Syntax. Kluwer, Dordrecht, pp.

281–337.Roussou, A., 1994. The Syntax of Complementizers. Unpublished Ph.D thesis, University College, London.Roussou, A., 2000. On the left periphery: Modal particles and complementizers. Journal of Greek Linguistics 1, 65–94.Roussou, A., 2006. Symplƞrvmαtikoı Deıktez [Complementizers]. Patakis, Athens.


Schuele, C.M., Dykes, J., 2005. Complex syntax acquisition: a longitudinal case study of a child with specific language impairment. Clinical Linguistics andPhonetics 19, 295–318.

Schuele, C.M., Nicholls, L.M., 2000. Relative clauses: evidence of continued linguistic vulnerability in children with specific language impairment. ClinicalLinguistics and Phonetics 14, 563–585.

Schuele, C.M., Tolbert, L., 2001. Omissions of obligatory relative markers in children with specific language impairment. Clinical Linguistics and Phonetics15, 257–274.

Schulz, P., 2003. Factivity: Its Nature and Acquisition. Linguistische Arbeiten 480. Max Niemeyer Verlag, Tubingen.Stambouliadou, C., 2006. H αnptyjƞ tvn symplƞrvmαtikv�n protsevn me δeıktez oti kαi poy stƞn pαiδikZ γlv�ssα [The development of

complement clauses introduced by oti and pu in child language]. MA thesis, University of Athens.Stavrakaki, S., Tsimpli, I.M., 1999. Diagnostic verbal IQ test for Greek preschool and school age children. Poster presented at the 5th European Conference on

Psychological Assessment, University of Patras.Stephany, U., 1995. The Acquisition of Greek. Arbeitspapiere des Instituts fur Sprachwissenschaft der Universitat zu Koln.Tallal, P., 1976. Rapid auditory processing in normal and disordered language development. Journal of Speech and Hearing Research 19, 561–571.Tsangalidis, A., 2002. Homonymy, polysemy, category membership: the case of Greek modal particles. Belgian Journal of Linguistics 16, 135–150.Tsimpli, I.M., 1996. The Prefunctional Stage of Language Acquisition. Garland, New York.Tsimpli, I.M., 2001. LF-Interpretability and language development: a study of verbal and nominal features in Greek normally developing and SLI children.

Brain and Language 77, 432–448.Tsimpli, I.M., 2003. H kαtktƞsƞ toy γԑnoyz stƞn EllƞnikZ vz δey�terƞ γlv�ssα [The acquisition of gender in Greek as a second language]. In:

Anastiasiadhi-Simeonidi, A., Ralli, A., Chila-Markopoulou, D. (Eds.), Qԑmαtα NeoellƞnikZz GrαmmαtikZz: To Gԑnoz. Patakis, Athens, pp. 168–189.Tsimpli, I.M., 2005. Peripheral positions in Early Greek. In: Stavrou, M., Terzi, A. (Eds.), Advances in Greek Generative Syntax. John Benjamins, Amsterdam,

pp. 178–216.Tsimpli, I.M., Dimitrakopoulou, M., 2007. The interpretability hypothesis: evidence fromwh-interrogatives in L2 acquisition. Second Language Research 23,

215–242.Tsimpli, I.M., Mastropavlou, M., 2007. Feature-interpretability in L2 acquisition and SLI: Greek clitics and determiners. In: Liceras, J., Zobl, H., Goodluck, H.

(Eds.), The Role of Formal Features in Second Language Acquisition. Routledge, London, pp. 142–183.Tsimpli, I.M., Stavrakaki, S., 1999. The effects of a morpho-syntactic deficit in the determiner system: the case of a Greek SLI child. Lingua 108, 31–85.van der Lely, H.K.J., 1994. Canonical linking rules: forward vs. reverse linking in normally developing and specifically impaired children. Cognition 51, 29–

72.van der Lely, H.K.J., Jones, M., Marshal, C.R., 2011. Who did Buzz see someone? Grammaticality judgement of wh-questions in typically developing children

and children with grammatical-SLI. Lingua. 121, 408–422.Varlokosta, S., 1994. Factive complements in Modern Greek. University of Maryland Working Papers in Linguistics 2, 238–258.Varlokosta, S., Rohrbacher, B., Vainikka, A., 1998. Functional projections, markedness, and ‘root infinitives’ in Early Child Greek. Linguistic Review 15, 187–

207.Wexler, K., Schutze, C.T., Rice, M., 1998. Subject case in children with SLI and unaffected controls: Evidence for the Agr/Tns omission model. Language

Acquisition 7, 317–344.


complementizers and subordination in typical language acquisition and sli

Documents