Language Acquisition

This article was downloaded by:[University of Milan] On: 13 April 2008 Access Details: [subscription number 789756591] Publisher: Psychology Press Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Language Acquisition Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t775653668

Article Omission Across Child Languages

Maria Teresa Guasti a; Anna Gavarró b; Joke de Lange c; Claudia Caprin a a Università di Milano-Bicocca, b Universitat Autònoma de Barcelona, c Universiteit Utrecht, Online Publication Date: 01 April 2008 To cite this Article: Guasti, Maria Teresa, Gavarró, Anna, de Lange, Joke and Caprin, Claudia (2008) 'Article Omission Across Child Languages', Language Acquisition, 15:2, 89 - 119 To link to this article: DOI: 10.1080/10489220801937346 URL: http://dx.doi.org/10.1080/10489220801937346

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Downloaded By: [University of Milan] At: 08:26 13 April 2008

Language Acquisition, 15:89–119, 2008 Copyright © Taylor & Francis Group, LLC ISSN: 1048-9223 print/1532-7817 online DOI: 10.1080/10489220801937346

Article Omission Across Child Languages Maria Teresa Guasti Università di Milano-Bicocca

Anna Gavarró Universitat Autònoma de Barcelona

Joke de Lange Universiteit Utrecht

Claudia Caprin Università di Milano-Bicocca

Article omission is known to be a feature of early grammar, although it does not affect all child languages to the same extent. In this article we analyze the production of articles by 12 children, 4 speakers of Catalan, 4 speakers of Italian, and 4 speakers of Dutch. We consider the results in the light of (i) the adult input the children are exposed to, (ii) the prosodic properties of articles in the three languages, and (iii) the properties of the syntax-semantics mapping of nouns in the languages under consideration. We show that the proportion of bare nouns (grammatical or ungrammatical) in the adult input does not bear any systematic relation to child production/omission of articles and that the full developmental pattern observed can be explained by appealing to the role of the nominal mapping parameter (NMP) in guiding acquisition, in conjunction with prosodic properties of articles and with discourse conditions.

Correspondence should be sent to Maria Teresa Guasti, University of Milano-Bicocca, Department of Psychology, V.le dell’Innovazione 10, 20126 Milano, Italy. E-mail: mariateresa.guasti@ unimib.it

89


90

GUASTI ET AL.

1. INTRODUCTION Children speaking a wide variety of languages omit articles in their earliest productions, a phenomenon that has been documented over recent years (Hoekstra, Hyams, and Becker (1996), Chierchia, Guasti, and Gualmini (1999), Marinis (1999), Pérez-Leroux and Roeper (1999), Crisma and Tomasutti (2000), Avrutin and Brun (2001)). Less documented is the fact that children’s omission of articles varies across languages. Exceptions are the work by Lleó and Demuth (1999) (see also Lleó (2001)) and the work by Chierchia et al. (1999). Lleó and Demuth observed that there is a temporal displacement in the development of (proto)articles: children speaking Spanish produce proto-articles at an earlier age than German-speaking children. Chierchia et al. (1999) observed that Frenchand Italian-speaking children cease to omit articles at an earlier point of linguistic development than English- and Swedish-speaking children. Different explanations were offered for these cross-linguistic differences. Lleó and Demuth (1999) argued in favor of a prosodic account; according to them, prosodic differences in the early vocabulary of Spanish and German are responsible for the earlier use of articles by Spanish learners. More specifically, the presence of a prosodic model at the lexical level was held to be the critical factor that justified the earlier use of articles by Spanish-speaking children. By contrast, according to Chierchia et al. (1999), it is the mapping between syntactic categories and semantic types that is at the heart of the difference. Languages vary as to the mapping between syntactic categories and semantic types. In the Romance languages, all nouns are mapped into the semantic-type predicate and articles are needed to turn predicates into arguments. Thus, once the child has chosen this mapping, s/he knows that articles need to be used. In Germanic languages, instead, nouns can be mapped into two semantic types, depending on whether they are count or mass nouns. The latter are mapped into the type argument and thus can be used as bare nouns, for example, Water is falling on the floor, while the former are mapped into the type predicate. In this case, an article is needed, as in Romance (see the next section as to why the Germanic languages allow bare plurals). Thus, once the child exposed to a Germanic language has discovered that nouns can be mapped into two semantic types, s/he still needs to establish whether each specific noun is mass or count and consequently whether it is mapped into argument or predicate. This takes time as it is tied to the acquisition of the lexicon: the mapping has to be decided every time a noun enters the child’s lexicon. Initially, the child that adopts the Germanic setting may be unsure about this mapping and produce bare count singular nouns; with time, s/he may develop a more efficient procedure and immediately recognize whether the noun is mass or count. To disentangle the factors that determine these cross-linguistic differences, we look at two additional languages, Catalan and Dutch, and compare them to Italian. As far as syntax is concerned, Dutch is like English, German, and


ARTICLE OMISSION

91

Swedish in allowing bare nouns in all argumental positions, while Catalan, Italian, and Spanish do not. As regards the prosodics of words, Dutch is like English and Catalan, in some sense, is in between Italian and Dutch: like Italian, it has trisyllabic words, but like Dutch it also has monosyllabic words. We will show that the role of prosody in explaining the developmental course of articles is rather limited and cannot be valid for all three languages, but prosodic factors can be integrated fruitfully with a syntax-semantic interface model that accounts for differences in the cross-linguistic patterns. The article starts with a description of the system of articles and of the syntax/semantics of nouns in the three languages under scrutiny. Then it presents the investigation of the speech of 12 children (4 for each language) and of the speech of their caregivers. It ends with a discussion of the results and with a model that accounts for the properties of article development found.

2. BACKGROUND: THE SYNTAX, SEMANTICS, AND PROSODY OF ARTICLES First let us briefly consider some basic facts concerning the cross-linguistic variation one finds with respect to the syntax of nouns in the three languages that we consider: Catalan, Dutch, and Italian. Catalan and Italian have a full paradigm of articles: definite and indefinite articles varying in gender and number. In Italian, there are also some variants of the masculine articles (lo, gli); in Catalan and Italian the definite singular feminine and masculine (lo, el, la) articles can be reduced to l0 in front of vowels. Dutch, instead, has three articles, two definite articles (de used with common gender nouns and het used with neuter nouns) and an indefinite one (een). Articles are used in front of common nouns in all three languages. In most varieties of Catalan (like the one under consideration), they must also be placed in front of proper nouns which are not vocative; in some varieties of Italian, other than those spoken by the children in our corpora, proper nouns can also be accompanied by an article.1 In addition, bare plural and mass nouns are also allowed, but their distribution varies to a significant extent. Bare plurals and bare mass nouns are grammatical in Dutch in all argumental positions. In Italian and Catalan, bare plural and mass nouns are accepted only as sisters to a lexical head (the object of a transitive verb and the post-verbal 1 There is no systematic study regarding which varieties of Italian require or allow the use of articles in front of proper nouns under certain circumstances. In addition, it is not clear whether this is a feature of the standard variety spoken in a certain area or of the substandard variety. This use is more widespread in northern regions than in central and southern regions. In Lombardia, a northern Italian region, articles with proper nouns are generally used in spoken language, but as a speaker of that variety the first author would not feel that it is obligatory. Moreover, it is more common with feminine nouns than with masculine ones.


92

GUASTI ET AL.

subject of an unaccusative verb and the object of a preposition) (Longobardi (1994)). One influential account of this cross-linguistic variation is that of Chierchia (1998) who claims that the syntax of nouns is governed by the nominal mapping parameter (NMP), a parameter that constrains the mapping between the syntactic category Noun and the corresponding semantic type. According to Chierchia (1998), the category N(oun) is associated with two independent features, each with a parametric value, which determine what nouns denote. The first feature is arg(umental) and determines whether nouns are kind-denoting and have a mass syntax; the second is pred(icative) and determines whether nouns are predicatedenoting and have a count syntax. The combination of these features gives rise to the following possibilities: (1) a. b. c. d.

N N N N

[Carg pred] [Carg Cpred] [ arg Cpred] [ arg pred]

Classifier languages (Chinese) Germanic (English/Dutch) Romance (Catalan/Italian/Spanish) impossible

The setting [ arg pred] is ruled out on logical grounds, as it would prevent nouns from having any interpretation at all. The other settings are instantiated by different languages. In languages like Chinese, Nouns are mapped into [Carg]. This means that they are kind-denoting and have the syntactic properties of mass nouns in languages such as English, in particular no article is needed and bare NPs can be used in all argument positions. In Romance, nouns are mapped into [Cpred] and to turn them into arguments D must be projected and filled with an article. Thus, bare nominal arguments are generally disallowed in these languages (apparent bare nominals, found as sisters of lexical heads, are hypothesized to be preceded by a null determiner; see Longobardi (1994)).2 Finally, Germanic languages are the union of the two types of languages just discussed. Nouns are either [Cpred] or [Carg]. If they are [Cpred] (count nouns), they behave like Romance nouns and need to be accompanied by an article; if they are [Carg] (mass nouns), they may occur as bare nominals. Bare plurals in this system are derived from plural properties through a type-shifting operation that yields the maximal sets of objects. This type-shifting operation is allowed 2 We

assume that Romance-speaking children need to establish that nouns are mapped into [Cpred] and need to be accompanied by articles, but they also have to learn that there is a null article licensed in certain contexts. This picture may lead us to conjecture that children who have set the value of the NMP to [Cpred] overuse articles with NPs in object position. Unfortunately, this prediction cannot be tested with our data. Although bare mass and plural nouns are licit in object position, there are no data that inform us as to the real use of this option in spoken language; in addition, there are speakers of Italian who prefer the use of determiners even in these cases, i.e., they prefer to say (i) Voglio del latte ‘I want some milk‘ (with del being a partitive article) to (ii) Voglio latte ‘I want milk‘. So, to test this prediction, one would need to set up an experiment in which children would have to judge which of the sentences (i) and (ii) they preferred.


ARTICLE OMISSION

93

in Germanic languages because in these languages, nouns can be [Carg] and used, for example, in generic sentences, e.g., Usually, dogs bark. Type-shifting is not allowed in Romance; this means that what in the Germanic languages is expressed through type-shifting, in Romance must be expressed otherwise. In fact, in these languages, the equivalent of Usually, dogs bark requires the use of an article, as the following Italian example shows: Solitamente, i cani abbaiano (lit. ‘usually the dogs bark’). Similarly, while in a Germanic language, like English, one can say I want milk, in a Romance language, like French, one has to use the partitive article: Je veux du lait (lit. ‘I want some milk’); for some speakers of Italian, the use of the partitive article is also required and for others it is just possible: Voglio del latte ((I) want some milk) (see footnote 2). This means that the Romance paradigm of articles is different from the Germanic one; in particular, the Romance paradigm is complete in that it has definite and indefinite singular and plural articles, while the Germanic paradigm does not have a plural indefinite (the partitive article). Articles in the three languages under investigation also have different prosodic properties. In Catalan and Italian, articles are proclitic to the following word. The proclitic status of articles is not questioned for Eastern Catalan, the variety spoken by the children participating in our study (Bonet (2002)).3 For Italian, on the other hand, the proclitic status of articles has been questioned. Crisma and Tomasutti (2000) proposed that an article in object position forms a foot with the preceding verb when this is a monosyllabic item.4 In addition, a reviewer notes that in Italian rapid speech, the article il (the-masc-sg) and i (the-masc-pl) are enclitic to the preceding verb, whether this is a mono- or disyllabic word. The evidence for this assertion is that, in rapid speech, the vowel [i] of the article il is not present in (2a0 ), and in (2b) the [i] is pronounced as a glide. (2) a.

Ho (I) have b. Ho Ho (I) have

visto il seen the visto i [vistoj] seen-the

cane. ! a0 . Ho visto ‘l cane. dog (I) have seen-the dog cani. cani dogs

These examples show that the articles il and i may be syllabified with the preceding words. Resyllabification, however, does not entail enclisis. For example, 3 For Western Catalan, enclisis of articles is a possibility that may be entertained (see Mascaró and Rigau (2002)), but this possibility is irrelevant for us, as the variety under study is Eastern Catalan. 4 Through a production experiment, Crisma and Tomasutti (2000) found that, in verbal utterances, omission of the article la (the-fem-sg) was higher when the verb was disyllabic than when it was monosyllabic. Caprin and Guasti (2006) did not replicate this result based on spontaneous data, nor did Caprin and Ioghà (2007), based on an elicited imitation study. Specifically, Caprin and Guasti found that omission of la did not differ depending on whether the verb was mono- or disyllabic. The same result was found by Caprin and Ioghà in a controlled experiment.


94

GUASTI ET AL.

in (3), for some Italian speakers, the auxiliary ha (‘has’) is resyllabified with the preceding noun, but it is not enclitic to the preceding noun.5 Resyllabification occurs at the level of the syllable and may occur inside the intonational phrase, while enclisis occurs at the level of the clitic group, a smaller unit than the intonational phrase (see, for example, Nespor and Vogel (1986)). (3) Il bus ha fatto tardi. Il busa. fatto tardi The bus has made late ‘The bus was late.’ Thus, we believe that the data in (2) are not compelling evidence for the enclitic status of Italian articles, but could simply indicate that in rapid speech some Italian articles can be resyllabified with the preceding word. For this reason, throughout the article we will assume that articles are proclitic in Italian, although we think that a more careful investigation, which cannot be pursued within the scope of this article, is in order. For the sake of argument, we will also discuss the consequences of the possibility that Italian articles may be both proclitic and enclitic. Let us turn now to the prosodic properties of articles in Dutch. The articles de ‘the’ and een ‘a’ are weak forms and prosodify either as proclitic to the following word or as enclitic to the preceding word, often a verb. Articles are proclitic when they appear in utterance initial position, e.g., the first position of the sentence or with nouns in isolation. Het is unstressed and, like all weak forms, cliticizes like the other Dutch articles (see Gussenhoven (1985), Booij (1995)). Thus, the three languages under investigation differ with respect to the prosody as well as with respect to the syntax/semantics of articles.

3. DATA Our investigation is based on the speech of 4 Catalan- (Guillem, Gisela, Laura, and Pep), 4 Italian- (Diana, Martina, Raffaello, and Rosa), and 4 Dutch- (Abel, Peter, Sara, and Tom) speaking children (Catalan: from the corpus of Serra and Solé; Italian: from the Calambrone corpus, Cipriani et al. (1989); Dutch: from the Groningen corpus, Bol (1996), and the corpus of van Kampen (1994)) all data being available through CHILDES (MacWhinney and Snow (1985)). The Italian children are from Tuscany and speak the central variety of Italian. 5 Notice that this resyllabification occurs only in rapid speech and it is unlikely that adults usually and only address children with rapid speech. Although we cannot discard the hypothesis that children hear rapid speech in their environment, they certainly hear many utterances in which the article il is not resyllabified with the preceding verb.


ARTICLE OMISSION

95

TABLE 1 Data Used in the Study

Gisela Guillem Laura Pep Diana Martina Raffaello Rosa Abel Peter Sara Tom

Files

Age Range

MLU Range

N. Utterances Requiring an Article

4, 5, 6, 7, 8, 9, 10, 11, 12 15, 18, 21, 22, 26, 29, 30, 31, 32 4, 6, 9, 11, 12 13, 15, 17, 21, 24, 26, 27, 29, 31 2–10 2, 4, 6, 8, 11, 13, 16 2, 3, 5, 7, 9, 10, 11, 12, 13, 14, 16, 17 2, 5, 7, 10, 12, 13, 14, 16, 18, 20 111, 201, 203, 205, 207, 210, 211, 301 200, 202, 205, 207, 208 2, 7, 11, 17, 20, 27 110, 202, 205, 206, 209, 210, 301

1;10–2;11 1;9–2;11 2;2–3;0 1;5–3;0 1;10–2;6 1;7–2;7 1;9–2;11 1;9–2;11 1;11–3;0 1;11–2;08 1;7–2;11 1;10–3;0

1.1–2.6 1.3–2.4 1.5–2.2 1.2–2.9 2.0–4.5 1.2–2.7 1.3–3.7 1.4–2.8 1.2–3.7 1.8–3.5 1.2–3.0 1.2–3.1

326 485 485 696 457 500 445 388 487 537 428 553

Table 1 summarizes the data used for the analyses, unless otherwise specified. It includes information about the files used, the age range during the period investigated, the MLU range and the number of utterances in which an article was either used or omitted. Care was taken to have roughly the same number of utterances in each language. The MLU was calculated for all children on the first 100 utterances of each file used (unless the number of utterances was less than 100, in which case it was based on the total file). The total number of utterances was 1992 for Catalan, 1790 for Italian, and 2005 for Dutch. One reviewer raised the question of the reliability of the transcriptions, in particular of children’s renditions of articles and presyntactic devices. Although this problem is not specific to our study and concerns all research conducted using the CHILDES database, we think that it is a important concern for studies such as ours. For Catalan, transcripts were carried out by one transcriber, but in case of doubt a team of three people checked the video and decided on which transcription was to be considered correct; if the transcription was still problematic, a special symbol was introduced and, therefore, no article or protoarticle appears in the transcript. The procedure was sufficiently conservative to guarantee that our recount does not overestimate article production (we thank Miquel Serra, Universitat de Barcelona, for discussing this issue with us). For Italian, the transcriptions were carefully carried out (and then checked) with the goal of teasing apart presyntactic devices, which were coded with a special symbol and thus are easily distinguishable in the files from the full forms of the various functional words. The criteria for what counts as a presyntactic device are reported in Bottari, Cipriani, and Chilosi (1993/4). Finally, the transcriptions of all the Dutch corpora were checked after the first transcription was made by


96

GUASTI ET AL.

at least one other person and presyntactic devices were coded differently than articles. Therefore, we feel confident that, although the transcriptions might not be perfect, they are reliable (see Table 1). We eliminated unclear sentences, immediate repetitions of the same sentence, idiomatic expressions, and routine sentences. We identified routine sentences with stereotyped expressions that can possibly be learned by children as chunks (e.g., Italian fare nanna, lit. do lullaby, ‘sleep’). Then, we analyzed the relevant utterances by hand distinguishing between article use and omission in different contexts: in utterances including a verb and utterances without a verb. These latter include nouns in isolation, as in answers to questions (e.g., What is this? Horse) or nouns accompanied by adjectives or prepositions (e.g., Big pig, For horse). This distinction is motivated by the following considerations. It is sometimes difficult to decide whether, in utterances without verbs, articles are needed or not. For example, when one names an object, in lists and in answers to some questions, an article is not necessary. Thus, it is possible that, in isolation, omission of articles is overestimated. By contrast, in sentences, this problem does not arise. We counted as omissions of articles all those contexts in which a noun in the target language would necessarily have required an article. This means that, for Dutch, plural and mass nouns without an article were considered ungrammatical only in those contexts in which the use of an article was obligatory in the target language. Similarly, for Italian and Catalan, plural and mass nouns without an article were considered ungrammatical unless they occurred as direct objects or objects of prepositions. Proper nouns without articles were counted as omissions for Catalan, because in this variety of the language they obligatorily have to co-occur with a definite article. In our counts we excluded some copula sentences where the distinction between argument or predicate was impossible to draw, as for sentences like Gianni è (un) dottore in Italian or El Joan és (un) metge in Catalan ‘John is a doctor’. This decision is motivated by the fact that predicates in these constructions can be used legitimately without an article. Finally, all the files used included at least three contexts requiring an article. We analyzed sentences with all kinds of articles, definite and indefinite, singular and plural, and unless otherwise specified we did not include in our counts presyntactic devices or proto-forms of articles, which, as we said, were clearly identifiable in the transcripts.

4. RESULTS 4.1. Age Range and MLU The ages of the children during the period investigated are roughly the same in the three languages (Kruskal-Wallis, 2 D 0:47; df D 2, p D 0:78; age in months: Catalan M D 29, SD D 5:1, age range: 17 to 36; Dutch M D 29,


ARTICLE OMISSION

97

SD D 4:7, age range: 19 to 36; Italian M D 27, SD D 4:9, age range: 19 to 35).6 Thus, differences between the three languages cannot be attributed to the fact that some children are older than others. Regarding MLU, no significant difference was found among the three languages (Kruskal-Wallis, 2 D 0:2; df D 2, p D 0:36; Catalan M D 1:89, SD D 0:2; Dutch M D 2:42, SD D 0:35; Italian M D 2:42, SD D 0:52). At first sight, this may be unexpected since MLU is in part a function of article use, as pointed out by an anonymous reviewer. However, it is possible that children who do not use articles use other words in their multi-word utterances and this is what explains the same length. This hypothesis is plausible given the fact that the composition of the vocabulary of children exposed to different languages varies. Caselli, Casadio, and Bates (2001) found that when the vocabulary of Italian- and English-speaking children is between 100 and 600 words, there are more function words in Italian than in English. It is not unreasonable to think that when these children form utterances, on average, they may use the same number of words, but which words those are may be partly a function of which words are present in their respective vocabularies. Mean length of utterance is a quantitative measure; it does not inform us about the qualitative aspects of utterances. 4.2. Article Omission in the Three Early Languages First, we analyzed whether there were differences in the proportion of article omission in the three languages, collapsing together data for the whole period investigated. In this analysis, we considered only articles, that is, target-like forms; presyntactic devices were tallied separately and are considered in section 6. Since no difference was found between omission in verbal utterances and utterances without a verb, we put them together. A significant difference was found between languages (Kruskal-Wallis N D 12; 2 D 8:0; df D 2, p D :01). The Mann-Whitney test shows that there is a significant difference between Catalan and Dutch, on the one hand, and Italian and Dutch on the other (z D 2:3, p D 0:02). Proportion and standard deviations are as follows: Catalan M D 0:27, SD D 0:09; Italian M D 0:37, SD D 0:15; Dutch M D 0:63, SD D 0:03. Omission of articles is higher in Dutch than in Catalan and Italian. In order to further investigate the development of children’s omission of articles in the three languages, we had to match the children’s performance based on some measure of linguistic development. We discarded age since children at the same age may be more or less advanced and there is enormous variability among children in the age range relevant here (Bates, Dale, and Thal (1995)). 6 We used nonparametric analyses as the requirement of homogeneity necessary for the application of parametric analysis was not always met. However, we repeated all the analyses using ANOVA and we did not find any differences with respect to the nonparametric analyses. Therefore, we decided to report the results of nonparametric tests.


98

GUASTI ET AL.

TABLE 2 Proportion of Article Omission in Different Periods of Linguistic Development as Measured by Proportion of VU, in Utterances With and Without Verbs Stage 1 0.03–0.29 VU

Catalan Italian Dutch

Stage 2 0.3–0.6 VU

M

SD

M

SD

0.45 0.59 0.92

0.11 0.20 0.06

0.07 0.16 0.39

0.04 0.08 0.08

As an independent measure of linguistic development, we used the rate of verbal utterances (VU) (see Valian (1991)).7 Therefore, we calculated for each child the rate of VU in a given file and divided the data of article use/omission into two stages: the first stage includes observations obtained when the children’s proportion of VU was between 0.03 and 0.29 and the second one when it was between 0.30 and 0.60 (VU ranges between 0.06–0.60, M D 0:34, SD D 0:17 in Catalan; it ranges between 0.03–0.60, M D 0:28, SD D 0:15 in Italian; and it ranges between 0.05–0.54, M D 0:30, SD D 0:16 in Dutch). Table 2 reports the proportion of article omission and SD at the two different stages of linguistic development. As can be seen in Table 2, at stage 1 all children omit articles in half or more of the contexts where they are required according to the adult grammar; Dutchspeaking children omit many more than the other two groups. At stage 2, Catalan and Italian children have almost stopped omitting articles; Dutch-speaking children omit fewer than in stage 1, but many more than the other two groups of children. These findings are corroborated by statistical analyses. The KruskalWallis test shows that there is a significant difference among the three languages at stage 1 (2 D 7:73, df D 2, p D 0:02) and at stage 2 (2 D 8:76, df D 2, p D 0:01). The Mann-Whitney test shows that at stage 1, Dutch is different from Catalan (z D 2:3, p D 0:02) and from Italian (z D 2:03, p D 0:02). At stage 2, Dutch is still different from both Catalan and Italian (z D 2:3, p D 0:02). We can observe that by stage 2, omission has decreased to less than 20% in the two Romance languages, while it is still about 40% in Dutch. In summary, there are differences between the two Romance languages, on the one hand, and Dutch (a Germanic language), on the other. First, omission of articles 7 The

question of how to compare the linguistic development of children speaking the same language or different languages is a matter of debate. MLU is often used, but other measures have also been used, such as number of verbal utterances, and proportion of presyntactic devices (Veneziano and Sinclair (2000), Deen (2005)). In another work, we have used the number of different words in the productive vocabulary (see Guasti et al. (2004)).


ARTICLE OMISSION

99

is generally higher in Dutch than in the two Romance languages; second, at the same stage of linguistic development, children exposed to a Romance language have almost stopped omitting articles, that is, they are more advanced than their Dutch peers in their mastery of the grammar of article use. So far we have discussed the data of children as a group and found that initially (stage 1) all children optionally omit articles and then, while Romancespeaking children almost stop omitting articles, Dutch learners continue to do so at a relatively high rate, although less than initially (stage 2). To have a complete picture, we now turn to individual developmental curves in article use which are reported in Figures 1, 2, and 3. The curves in these figures confirm what we have found at the group level. All children start by omitting articles very frequently with omission being higher in Dutch than in the two Romance languages. All Dutch children and one Italian child (Raffaello) even start by omitting them 100% of the time in the first files. Then omission decreases for all children. At around 29– 32 months all Catalan- and Italian-speaking children omit less than 20% of articles in obligatory contexts. By contrast, Dutch learners omit around 20% at around 36 months. Finally, children speaking a given language behave quite uniformly and it is not the case that some children are quite good at using articles and others poor inside the group. These facts confirm our findings based on the analysis of the data from the three groups, and they add some further information. For example, they show that some subjects do not use articles at all in the first files; this suggests that there is a stage before the two we have found in which articles are completely absent, a stage that in our data

FIGURE 1 Developmental curves of article omission (over the total number of obligatory contexts of article use) for the four Catalan children.


100

GUASTI ET AL.

FIGURE 2 Developmental curves of article omission (over the total number of obligatory contexts of article use) for the four Italian children.

FIGURE 3 Developmental curves of article omission (over the total number of obligatory contexts of article use) for the four Dutch children.

is documented in the speech of only some children.8 They also show that in Catalan and Italian omission of articles decreases even more rapidly in terms of 8 In the nonparametric analysis of the group data presented in section 4.2, we computed the means concerning stage 1 by also including the data points corresponding to 100% of article omission. If we leave out these data points, a significant difference still exists between Catalan and Italian on the one hand and Dutch on the other (Mann-Whitney, z D 2:3, p D 0:02).


ARTICLE OMISSION

101

age. They further support the conclusion that there are differences among early languages in the development of article use. In the subsequent sections, we examine factors that may be advanced to explain the pattern of article omission observed.

5. INFLUENCES OF THE INPUT As a first step, we performed an analysis of the input children receive. It is possible that children omit articles because they are exposed to cases of ungrammatical omission or because they hear bare nouns. In addition, it is also possible that the proportion of bare nouns children hear differs, depending on the language, and this may be the source of the cross-linguistic variation we observed. Thus, for each child, we examined the incidence of adult use of bare nouns (whether adults were talking to the child or among themselves) at three points in the child’s development: when omission of articles is high in the child’s production, when it has started to decline, and when article omission has almost disappeared. For adults, we made two computations. First, we calculated the rate of bare nouns (be they grammatical or ungrammatical) over the total number of nouns with and without an article. For Italian and Dutch, we only included common nouns, while for Catalan we also included proper nouns, since these require an article. Our reason for including grammatical and ungrammatical bare nouns is dictated by the hypothesis that we want to test, namely, that children omit articles because they hear bare nouns in the input and they do not know which nouns can occur with articles and which cannot. In the second count, we computed the rate of bare nouns over all nouns including not only common nouns but also proper nouns (for Italian and Dutch) and nouns used in the vocative case (that do not require an article) for all three languages. In both counts, we included nouns in utterances with verbs and without verbs. Table 3 reports the results of this analysis. Table 3 shows that there are bare nouns in the caregivers’ speech to children, but there is no systematic decrease in the adults’ use of bare nouns that parallels the decrease in the rate of bare nouns in children’s speech, as might be expected if children merely imitated the input. For some adults there is a decrease, but for others there is an increase in the use of bare nouns. Thus, it seems unlikely that children omit and then cease to omit because they mimic the input. If this were the case, we should have found a systematic change in the input (adults’ speech) parallel to the change in the output (children’s speech). This is in contrast to what has been reported by Bohnacker (2005) who found a large proportion of bare nouns in the child-directed speech to one Swedish child and a parallel decrease in the child and adult speech. In fact, in some cases we also found a parallel decrease, but this was not so for all children, although all children used fewer bare nouns over time.


102

GUASTI ET AL.

TABLE 3 Percentages of Omission of Articles by Children and Their Caregivers in Interaction With Them

Child Gisela 04 Gisela 09 Gisela 12 Guillem 15 Guillem 29 Guillem 31 Laura 04 Laura 09 Laura 12 Pep 13 Pep 24 Pep 31 Diana 2 Diana 4 Diana 10 Martina 2 Martina 8 Martina 16 Raffaello 7 Raffaello 10 Raffaello 17 Rosa 2 Rosa 12 Rosa 20 Abel 20102 Abel 20729 Abel 21110 Peter 11103 Peter 20529 Peter 20822 Sara 07 Sara 20 Sara 27 Tom 11011 Tom 20614 Tom 30102

Language

Children’s Bare N

Adults’ Bare N

Adults’ Common, Proper, Vocative N

Catalan Catalan Catalan Catalan Catalan Catalan Catalan Catalan Catalan Catalan Catalan Catalan Italian Italian Italian Italian Italian Italian Italian Italian Italian Italian Italian Italian Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch

96% 61% 0% 76% 6% 0% 68% 20% 7% 86% 4% 0% 43% 9% 8% 89% 73% 11% 72% 29% 3% 80% 37% 12% 91% 68% 22% 100% 51% 29% 85% 43% 17% 100% 54% 11%

6% 16% 14% 17% 19% 19% 7% 28% 26% 23% 20% 27% 17% 29% 19% 27% 12% 13% 16% 16% 6% 8% 9% 16% 33% 25% 23% 34% 25% 18% 29% 9% 23% 30% 6% 21%

17% 16% 14% 46% 40% 31% 9% 31% 28% 38% 24% 37% 29% 37% 27% 29% 16% 17% 40% 27% 27% 30% 24% 22% 43% 37% 34% 53% 34% 30% 38% 20% 36% 44% 19% 38%


ARTICLE OMISSION

103

Although the incidence of bare nouns does not seem to be responsible for the children’s article omissions, it is possible that input factors are responsible for the different rates and patterns of omission in the three languages investigated. It is possible that Dutch children take more time to stop omitting articles because they are exposed to more adult cases of article omission and are thus more confused. Therefore, we compared the overall rate of adults’ bare nouns in the three languages, using the data displayed in Table 3. We did not find any significant difference among the three languages, either when we considered just common nouns (Kruskal-Wallis 2 D 4:7; df D 2, p D 0:09) or when we added proper nouns and vocative nouns (Kruskal-Wallis 2 D 2:0; df D 2, p D 0:36). The rate of bare nouns in the caregivers’ speech to children is by and large the same in the three languages. Of course this does not mean that the bare nouns are impossible in adult Dutch to the same extent they are in adult Catalan and Italian or that caregivers use something else instead of bare nouns. It merely says that, in the input we analyzed, there were few bare nouns, presumably because caregivers in these samples spoke about the here and now and there were few contexts favoring the use of bare nouns. This does not exclude the possibility that children have heard bare nouns in other contexts. A more extensive investigation of different inputs would certainly be in order. In addition, adult ungrammatical omissions of articles are rare in all three languages: 1% in Catalan, 2% in Italian, and 4% in Dutch. These omissions tend to be cases in which the adult repeats an utterance without an article that was previously produced by the child; some very rare cases are spontaneous uses of bare singular nouns used as proper nouns, especially for parents and relatives (Cosa fa nonna? ‘What does grandmother do?’), and for animals (Guarda orso sta mangiando ‘Look, bear is eating’), a typical feature of child-directed speech and common in some varieties of Italian, including the one spoken by the children; for Dutch as well some common nouns can be used as proper nouns in child-directed speech and children’s literature. Finally, for Dutch only we found some cases of article omission with singular nouns that are not legitimate in the standard grammar, but are acceptable in the spoken language in some special contexts (Guasti, de Lange, Gavarró, and Caprin (2004)). It is noteworthy that these uses are limited to Dutch and completely unacceptable in Catalan and Italian, even in the spoken language. We return to these cases in the discussion. The data considered suggest that children do not omit articles and then stop doing it because they find bare nouns in the input in the first stage, and fewer in the later stage. In addition, Dutch-speaking children do not omit more articles and do not continue to omit articles when Romance-speaking children have ceased because overall they hear more bare nouns in the caregivers’ input. Differences among the three early languages cannot be attributed to the different rate of bare nouns that children hear in the input.


104

GUASTI ET AL.

6. AGE OF FIRST USE OF ARTICLES It is possible that children exposed to the Romance languages start to use articles earlier than Dutch-speaking children, and that because of this earlier use they have more experience with articles, use them more frequently, and converge on the target grammar at an earlier point of linguistic development. In fact, Lleó and Demuth (1999) made this point for presyntactic devices, that is, for placeholders of articles. They found that Spanish learners use presyntactic devices a bit earlier (1;6) than their German peers (1;7–1;8). If this result is confirmed, it suggests that the different course of acquisition might be explained by the earlier use of presyntactic devices.9 Then, the question is whether the timing in the emergence of (proto)articles is significantly different between the languages under examination. To investigate this dimension, we searched for both articles and presyntactic devices in all files available for a given child, unlike in the previous analyses. Including all files was motivated by the fact that an article may be present in a file that was not included in the earlier counts. In this way, excluding this file we may underestimate the age at which the article is used. We established the age of first use of each article (full form) and of each presyntactic device (PD) so that the age of first use of a PD standing for the article la ‘the-fem-sg’ may be different from the age of first use of a PD standing for il ‘the-masc-sg’.10 Then we computed the mean age of the use of articles and of presyntactic devices. Age of first use is defined as the age at which children use an article in a novel and nonrepetitive way, following Stromswold (1990). We should emphasize that the first full form of an article and of a presyntactic device was not found in the first file available in the CHILDES database, but in one of the successive files with the exception of un ‘a’ for Martina and of a presyntactic device for one Catalan-speaking child at 14 months. This is an important point. Suppose that we found articles in the first file available for the Italian learners, but in later files for the Dutch learners. Then, it would be possible that the Italian learners were already using articles before the first transcription available. And in that case, it would not be surprising that they converge on the target system earlier than their Dutch peers. However, this was not so; articles generally started to be used after the initial transcriptions available in all three languages. Table 4 reports the mean age and standard deviations of first use of full forms of articles, the age of first use of presyntactic devices, and the age range of use of articles and presyntactic devices. As can be observed, children start 9 This raises the question of why Spanish-speaking children use presyntactic devices earlier than their German peers, as pointed out by one reviewer. Lleó and Demuth’s (1999) answer would be that prosodic properties of words are responsible for this difference. See section 7 for details. 10 For Italian, the following articles were found: il ‘the-masc-sg’, la ‘the-fem-sg’, i ‘the-masc-pl’, le ‘the-fem-pl’, un ‘a-masc-sg’, una ‘a-fem-sg’; for Catalan: la ‘the-fem-sg’, el ‘the-masc-sg’, les ‘the-fem-pl’, els ‘the-masc-pl’, una ‘a-fem-sg’, un ‘a-masc-sg’; and for Dutch: de, het ‘the’, een ‘a’.


ARTICLE OMISSION

105

TABLE 4 Mean Age in Months of First Use of Articles and of Presyntactic Devices Age of First Use of Article or Presyntactic Devices

Age of First Use of Articles


M

SD

Age Range

M

SD

Age Range

24 26 25

1,4 4,1 1,7

18–32 19–33 19–34

23 23 24

2,5 1,8 1,8

14–32 19–30 19–31

to use articles or presyntactic devices more or less at the same age. In fact, the Kruskal-Wallis test shows that there is no difference among the three languages either in the means for first use of articles or in the means for the first use of presyntactic devices or articles. In this second analysis, we used the age of first use of presyntactic devices; when no presyntactic device was used, we considered the age of first use of the corresponding article. In Catalan, some presyntactic devices start to be used at 14 months. This is due to one child, however, as reported above. From this analysis, we can conclude that, on average, children do not differ in the age of first use of articles or presyntactic devices. In spite of this similarity, Italian/Catalan and Dutch learners differ significantly in the rate of omission of articles. Thus, we can conclude that the reason why Italian and Catalan learners omit less and converge more quickly on the target system does not depend on the fact that they start to use articles or presyntactic devices at a significantly earlier age.

7. PROSODIC CONSTRAINTS: THE ROLE OF THE PROSODIC PROPERTIES OF WORDS Lleó and Demuth (1999) claim that the differences in the development of article use between Spanish and German are determined by prosodic constraints operating at the word level and by the way articles are prosodified in a particular language. First, they show that the words children try to produce approximate the shape of words in the target vocabulary, that is, Spanish learners tend to produce more di- and trisyllabic words and German learners tend to produce mono- and disyllabic words. Then, they suggest that Spanish-speaking children start to use articles earlier than German learners because they have models for the prosodic integration of articles provided at the lexical level by trisyllabic wSw (Weak Strong Weak) words, in which the first weak syllable has the status of an unfooted or extrametrical syllable directly attached at the level of the prosodic


106

GUASTI ET AL.

word. In fact, in Spanish, and more generally in Romance, this same prosodic structure arises at the utterance level by joining an article, a weak syllable, to a disyllabic Sw word, since articles procliticize to the following words in these languages. According to Lleó and Demuth, it is the availability of a prosodic model in the lexicon that boosts Spanish-speaking children’s production of articles and is responsible for the different timing in the use of articles between Spanish and German learners. A prosodic model for the integration of articles is rarely available in German. Moreover, in German, articles are prosodified either as foot, if they are full forms, or as enclitic to the preceding word, often a verb, if they are reduced. These two facts are supposed to explain the later use of articles in German. Lleó and Demuth (1999) also investigate the use of articles in English with mixed results: one child was quite advanced in the use of articles and her development was comparable to that of the Spanish-speaking children, and two children were less advanced and their use of articles was similar to that of the German-speaking children. The authors explain this dual pattern by invoking the fact that, in English, articles can be prosodified as proclitics (e.g., when they are in sentence initial position or are uttered with a noun in isolation) or as enclitics (e.g., when they follow a monosyllabic verb). If children focus on the proclitic status of articles (thus on the configuration of article C noun), they are led to early acquisition; if they focus on the enclitic status, they are led to later acquisition. As differences exist between Catalan/Italian and Dutch, with respect to the prosody of articles, we may ask whether and to what extent these differences are responsible for the pattern of article omission. Catalan and Italian articles have the same prosodic status as Spanish articles; they are proclitic, while Dutch articles, like English articles, can be either proclitic or enclitic. If the different developmental pattern of article use depends on the availability of a lexical model for the prosodic integration of articles, we expect to find a model in the Catalan and Italian lexicon, while such a model should be unavailable or less frequent in Dutch; in addition, since in Dutch articles can be either proclitic or enclitic, as in English, we expect to find different patterns of development among the Dutch learners, as Lleó and Demuth found for the English learners. Looking at the developmental curves of article omission in Dutch reported in Figure 3, it is clear that the second prediction is not fulfilled, as they are quite homogeneous and no child was more advanced than the others.11 The first prediction can be tested by examining the prosodic properties of words in the speech of the children and 11 We believe that this discrepancy ensues from Lleó and Demuth’s exclusive reliance on age. It is well known that there is variation among children in linguistic development (Bates et al. (1995)) and, therefore, age cannot be a very reliable measure of comparison. The dual pattern they found is, in our view, simply an indication that one child, Eve, was more advanced than the others, Adam and Sarah.


ARTICLE OMISSION

107

establishing whether and to what extent these children have internalized a model for the integration of articles. Guasti and Gavarró (2003) showed that the prosodic properties of the words produced by the Italian- and Catalan-speaking children in the early files are different in some respects, although the use of articles is very similar. Catalan has quite a number of monosyllabic content words, while Italian has almost none (27% vs. 3%). By contrast, multisyllabic words (trisyllabic words or words with more than three syllables) are significantly more frequent in Italian than in Catalan (38% vs. 19%). In addition, these trisyllabic words in Italian frequently have the pattern wSw and can be a model for the integration of articles with disyllabic Sw words. Trisyllabic words, being less frequent in Catalan, may not be a model for the integration of articles; however, Catalan has another model for the integration of articles that was not considered by Guasti and Gavarró (2003): disyllabic wS words could be a model for the integration of articles with monosyllabic words that are quite frequent in Catalan. In Dutch, articles can be proclitic or enclitic; thus, the prosodic models for the integration of articles could be wS(w) for proclitic articles and Sw for enclitic articles. Therefore, to test whether it is the presence of a prosodic model that influences the use of articles, we examined the di- and trisyllabic words in the speech corpora of Catalan, Dutch, and Italian children. We computed the proportions of disyllabic wS and Sw words out of the total number of disyllabic words and the proportion of trisyllabic wSw words out of the total number of trisyllabic words in the three languages under investigation in the first stage based on VU value (see section 4.2). We investigated the first stage because we wanted to test whether there is indeed a model at the earliest stage of multi-word production that could boost the use of articles, as suggested by Lleó and Demuth.12 Table 5 reports the proportion of wS, Sw, and wSw words and the standard deviation at stage 1. From this table, it emerges that wS words are very frequent in Catalan; Sw words are frequent in all three languages, but especially in Italian and Dutch; and wSw words are most frequent in Italian. In fact, the Kruskal-Wallis test on the proportion of wS words shows that there is a significant difference between the three languages (2 D 6:0, df D 2, p < 0:05). The Mann-Whitney test shows that there is a difference between Italian and Catalan and between Catalan and 12 We examined the vocabulary of children because, as Lleó and Demuth note, the prosodic shapes of words children produce approximate the adult target. In addition, the leading idea of this approach is that children use a prosodic model for the integration of articles if this model is part of their competence concerning the prosody of words. Thus, the best way to test this conjecture is by examining the prosodic properties of children’s words. In any event, there is a close match between the children’s productions and the adult target; Guasti and Gavarró looked at the first 500 most frequent words of an Italian and a Catalan frequency dictionary and found that children’s vocabularies closely match the target vocabulary in terms of types of prosodic words and of frequency of these words.


108

GUASTI ET AL.

TABLE 5 Proportion of wS, Sw, and wSw Words and Standard Deviation at Stage 1 wS


Sw

wSw

M

SD

M

SD

M

SD

0.41 0.007 0.03

0.06 0.006 0.04

0.54 0.99 0.96

0.03 0.009 0.04

0.31 0.89 0.21

0.33 0.07 0.05

Dutch (z D 2:3, p < 0:02). The Kruskal-Wallis test on the proportion of Sw words shows that there is a significant difference between the three languages (2 D 8:0, df D 2, p < 0:05). The Mann-Whitney test shows that there is a difference between Italian and Catalan and between Catalan and Dutch (z D 2:3, p < 0:02). Finally, the Kruskal-Wallis test on the proportion of wSw words shows that there is a significant difference among the three languages (2 D 8:0, df D 2, p < 0:05). The Mann-Whitney test shows that there is a difference between Italian and Catalan, and between Italian and Dutch (z D 2:3, p < 0:02). These results indicate that there are distinct models for the prosodic integration of articles in the vocabulary of each language investigated: wS disyllabic words in Catalan, wSw trisyllabic words in Italian, and Sw disyllabic words in Dutch. Interestingly, these models are quite frequent, especially in Italian and Dutch. However, the presence of a model seems to influence children’s production of articles in Italian and Catalan, but not in Dutch. One reason may be that, in the case of Dutch, the Sw model is only a model for the integration of enclitic articles, but in our analysis, reported in section 4.2, we collapsed together contexts for enclitic and proclitic articles and this may have obscured the influence of a prosodic model for the integration of articles. To be fair, the analysis of article omission should include only enclitic contexts for Dutch, as it is just in this case that there is a frequently available model for the integration of articles at the lexical level. For Catalan and Italian, on the other hand, such limitation is not needed, under the hypothesis that articles are uniformly proclitic, as we have assumed thus far. Alternatively, it could be that the presence of a model is not enough. What matters is whether there is no ambiguity in the way of prosodifying an article. That is, it may be that Dutch learners could not benefit from the prosodic model because articles can be prosodified either as enclitic or as proclitic. To disentangle these two hypotheses, we carried out a new analysis. We re-examined article omission in stage 1 using the same data used in section 4.2 (Table 2) for Catalan and Italian and using only enclitic contexts for Dutch. If the effect of a prosodic model was obscured by the kind of analysis carried out in section 4.2, in


ARTICLE OMISSION

109

which we also included contexts for which Dutch-speaking children did not have a model, then no difference would be expected among the three languages now, as only enclitic contexts are included for Dutch. Table 6 reports the proportion of article omissions in the three languages at stage 1. From this table, it clearly emerges that Dutch-speaking children still omit articles much more than the children from the other two groups. This is supported by the Kruskal-Wallis test that shows that there is a significant difference among the three languages (2 D 8:0, df D 2, p < 0:02). The Mann-Whitney test shows that there is a difference between Catalan and Dutch (z D 2:3, p < 0:02) and between Italian and Dutch (z D 2:02, p < 0:04). Thus, in spite of the fact that we limit the comparison to those Dutch contexts for which a prosodic model for the integration of articles is available at the word level, we still find a significant difference. Thus, we can discard the hypothesis that the difference in article omission between Dutch, on the one hand, and Catalan and Italian, on the other, was due to the fact that in our previous analysis we included contexts (proclitic articles) for which Dutch-speaking children do not have a frequent model. Even when the analysis is limited to enclitic contexts for which Dutchspeaking children have a model for the integration of articles, we still find a difference with respect to the Romance languages. Therefore, we can reject the claim that there was an oversight in our previous analysis. But the question still remains as to why Dutch-speaking children do not take advantage of the Sw model. It is possible that a prosodic model for the integration of articles is only useful if the way of prosodifying articles is unambiguous, as suggested by Lleó and Demuth for English. Thus, on the basis of our investigation, one could argue that Catalan and Italian learners omit less than Dutch learners because their lexicon includes a model for the integration of articles with the subsequent words and because this model is the only prosodic model that is used to accommodate articles prosodically. Dutchspeaking children cannot benefit from a prosodic model because of the dual prosodic nature of articles (proclitic or enclitic). This argument rests on the assumption that articles are proclitic in Catalan and Italian. As we discussed TABLE 6 Proportion of Article Omission During Stage 1. For Dutch, Only Contexts in Which the Article Would Be Enclitic to the Preceding Word Are Included Stage 1: Proportion of VU 0.03–0.29


M

SD

0.45 0.59 0.88

0.11 0.20 0.10


110

GUASTI ET AL.

in section 2, this assumption has been questioned for Italian. If future research provided clear arguments in favor of the double prosodic nature of articles in Italian, this would make Italian more similar to Dutch. In that case, either the presence of a prosodic model for the integration of articles has no bearing on the production of articles or it has an influence, but under some conditions that have yet to be discovered.

8. DISCUSSION Although all children omit articles in their first multi-word combinations, the developmental course of article use is subject to cross-linguistic variation. First, articles or their phonetic approximations are used more consistently at an earlier point of linguistic development in Catalan and Italian than in Dutch. We observe differences in the rate of omission already at stage 1. Second, at stage 2 there is almost no omission in Catalan and Italian, but omission is still high in Dutch. Thus, Romance learners reach a higher level of article use at a lower point of linguistic development when compared to Dutch-speaking children. These findings raise two questions: (i) why do children omit articles? and (ii) why does article omission display different patterns in early languages? We established that children start to use articles at more or less the same chronological age and that children’s omission of articles is not input-driven, at least not in a superficial sense. Adults use bare nouns in their child-directed speech, but this use is not subject to changes during the course of development. Children hear more or less the same proportion of bare nouns over the developmental course examined here. Still, their use of articles increases; so, their omission of articles does not depend on changes in the input. Second, we established that the proportion of bare nouns in the caregivers’ speech is not different in the three languages investigated. This fact eliminates the possibility that differences among the three languages are determined by the proportion of bare arguments in the adult speech. Although bare nouns (mass and plural nouns) are clearly an option in Dutch, this option is not expressed in the input we examined, presumably because the kind of interaction does not create the contexts appropriate for the use of bare nouns. Of course, it is possible that children hear bare nouns in other interactions. We then examined whether prosodic constraints can play a role in the use of articles. Specifically, we tested Lleó and Demuth’s (1999) hypothesis that the use of articles depends on the availability in the lexicon of a prosodic model for the integration of articles in combination with the way articles are prosodified in a particular language. Our results showed that there are models for the prosodic integration of articles in all three languages, disyllabic iambic words (wS words) for Catalan, disyllabic trochaic words (Sw words) for Dutch enclitic articles, and trisyllabic words with penultimate stress (wSw words) for Italian, but given our


ARTICLE OMISSION

111

results, we had to conclude that only Catalan and Italian children benefited from the model. This conclusion stands if it is the case that the way of prosodifying articles is not ambiguous in Italian, a controversial question that we had to leave open for future research. A proposal similar to the one offered by Lleó and Demuth was advanced by Giusti and Gozzi (2007) on the basis of the spontaneous production of one Italian child. They suggested that article production passes through three stages that differ in the prosodic strings that the child is able to produce. In stage one, the child produces only trochaic words (Sw) and does not produce articles. In the second stage, she produces trisyllabic words (wSw) and articles with trochaic words, that is, she uses trisyllabic words as a model for the integration of articles with disyllabic (Sw) words. Finally, in the third stage, articles are used with trisyllabic words. Although it is not easy to evaluate this proposal, as no information about, for example, the MLU at the different stages is given, nor information about whether the NPs occur in isolation or in sentences, what also emerges from this work is that prosodic factors can help children’s use of articles at least at the beginning. Although we think that the availability of a prosodic model can help some children, it is clearly the case that the grammar of articles can be acquired without such help. Dutch children do not seem to benefit from prosody, but they omit less in the second stage. Bohnacker (2005), unlike Giusti and Gozzi (2007), found no relation between the types of prosodic words and prosodic utterances, including articles, in the speech of one Swedish-speaking child. She noted that this child rarely produced wSw or wS lexical words, but frequently produced wSw utterances (resulting from the attachment of an indefinite article to a Sw word) between 1;10 and 1;11. What then can explain the course of article production in the three early languages investigated? We propose that the course of development of article production is best explained in relation to the model proposed by Chierchia (1998) and expressed in the NMP that governs the mapping between nominal syntactic categories and semantic types (following Chierchia, Guasti, and Gualmini (1999); see De Lange (2004) for a different approach). This mapping varies across languages and purports to determine how nouns are used in a given language. Above, we saw that three types of languages exist: languages in which nouns are mapped into [Carg] and are thus kind denoting; languages in which nouns are mapped into [Cpred] and thus are predicate denoting, and finally languages in which nouns can be mapped into [Carg] or [Cpred], a choice that is lexically determined (to the extent that being mass or count is lexically determined). The situation is summarized in (1), repeated below: (1) a. N [Carg pred] b. N [Carg Cpred] c. N [ arg Cpred]

Classifier languages (Chinese) Germanic (English/Dutch) Romance (Catalan/Italian)


112

GUASTI ET AL.

If we assume that UG provides the child with the NMP, the question arises as to how children set its value. Suppose that parameters are associated with a default value, which, following Chierchia (1998), we take to be [Carg pred], that is, the Chinese setting.13 Thus, initially all learners start with the Chinese setting and map nouns into the semantic type [Carg]. Nouns will be kind denoting and have the same distribution that mass nouns have in English, that is, they can be used without articles in all argumental positions. This means that there is a point in development at which children do not produce articles, although they produce multi-word combinations. Evidence for this stage in our results comes from the individual developmental curves of the Dutch children and of one Italian child, who do not produce any articles in the first files. In the literature, a phase of complete omission of articles has also been observed by Penner and Weissenborn (1996) for a Bernese-speaking child, by Bohnacker (1997) for a Swedish child, and by Chierchia et al. (1999) who report that some of their children speaking French, English, and Swedish did not use articles in their first multi-word productions. Eisenbeiss (2000) also found complete omission of articles for one of the German-speaking children she investigated, as did Marinis (2003) for a Greek learner and Küpish (2004) for a bilingual Italian-German child. It is possible that this phase is not found in the productions of all subjects either because when the transcriptions start this phase is over or because some children skip this phase altogether in their productions, but there is clear evidence that this phase is at least sometimes attested.14 We suggest that because children are especially endowed to pay attention to functional elements (e.g., Hoehle and Weissenborn (2003), Kedar, Casasola, and Lust (2004)), if the input language doesn’t match the default Chinese setting, they find this out by discovering that there are articles and morphologically plural nouns. This informs children that the Chinese setting is not valid, but this discovery is compatible both with the Romance setting and with the Germanic setting. We assume that children, regardless of the language of exposure, switch to the Germanic setting first, as they know that there are articles but they do not know if the article paradigm is complete (i.e., the language has definite and indefinite singular and plural, where the plural indefinite are the partitive articles). This switch is motivated by the subset principle. If German-speaking children were to switch to the Romance setting, they may never recover from this 13 If

children started from the Germanic or the Romance setting, they would need negative evidence to recover from the choice, if the language of exposure were Chinese. Based on either of these settings, they would have to look for articles forever, never finding one, but never knowing that none exists in the language. 14 One might expect that children in the Chinese phase use classifiers. However, this is unlikely as the use of most classifiers by Mandarin-speaking children occurs late, at around age 2;6, for reasons related to their complexity, according to Erbaugh (1992), and by this age the Chinese phase is finished.


ARTICLE OMISSION

113

choice, unless negative evidence was provided. In fact, once they have chosen the Romance setting, they are obliged to assume that there is a full paradigm of articles, but this is not the case and thus they will try to fill all the slots of the article paradigm forever. It may be possible that some children exposed to a Romance language, who have gathered evidence for a full paradigm of articles, switch directly from the Chinese to the Romance setting. In this case, we would expect such children to use articles consistently. This was not attested in our data, but it is a legitimate path. The change from the Chinese setting to the Germanic setting is exemplified in our stage 1 when Catalan, Italian, and Dutch children all optionally omit articles. The optional use of articles at stage 1 can be interpreted as a reflex of this intermediary parameter (mis)setting. In assuming this setting, when children learn a new noun, they have to decide whether it is to be mapped into [Carg] (it is kind denoting and has a mass syntax) or [Cpred] (it denotes a predicate and in order to be used in argumental position needs to be accompanied by an article). This does not mean that the parameter has to be set for each noun. The parameter is set to the value [Carg] and [Cpred]. But as children learn nouns, they have to decide for each noun whether it has the feature mass or count (or [Carg], [Cpred]). This is something like deciding whether a noun has the feature masculine or feminine or neuter gender. If a noun is mass, children will have to map it into [Carg], and if it is countable it will have to be mapped into [Cpred]. Under this approach, when children say ‘I ate apple’, they are (mis)analyzing a noun like ‘apple’ together with nouns like ‘furniture’ and take it to be grammatically mass [Carg]. We observed that at this stage, omission of articles is higher in Dutch than in Romance. Why is this so, if both groups of children have reset the parameter to the Germanic setting? We think that prosody may play a role here. As we said earlier, once children have recognized that there are articles in their language, they may rely on the availability of unambiguous prosodic models for their prosodic integration, that is, prosody may boost Romance children’s use of articles and account for the different rate of article use. Although Dutch learners know that articles can be used because of the [Cpred] setting, they lack an unambiguous bootstrapping model from prosody and are more reluctant to produce articles. It is even possible that the availability of a prosodic model helps Romance learners to get out of the Chinese phase more rapidly; this would explain why more cases of complete omission are found among Germanic learners than among Romance learners. We think that another reason for the higher omission in Dutch is the fact that Dutch allows the use of bare singular nouns in spoken language under certain conditions. In some spoken registers, one finds examples with bare singular nouns, both in subject and object position, with some also present in our corpus (although their frequency was not high, 3–4%); the example in (4) was produced by an adult speaker interacting with a child.


114

GUASTI ET AL.

(4) Dat is een meisje van twee huizen verderop. Meisje van een jaar of zes / zeven. This is a girl from two houses away. Girl from a year or six/seven This is a girl who lives nearby. Girl aged about six/seven years old. (Peter 20419) Sentences like the one above are fully acceptable in spoken Dutch despite the fact that they are unexpected given the value of the NMP. However, they are acceptable only under some specific discourse conditions: when the speaker and the listener share some knowledge, either because they are talking about someone familiar to both or because they are discussing something that has happened in the context and has been witnessed by both. Under the same discourse conditions, examples like the aforementioned are not acceptable in Catalan and Italian. It is possible that the availability of examples like (4) contributes to the still higher omission rate of articles in Dutch than in the two Romance languages. Thus, at stage 1, Dutch and Romance learners would omit articles because both have set the NMP to the Germanic setting and thus temporarily take some count nouns to be mass and use them without an article. Then Dutch learners omit more than Romance learners because of the lack of an unambiguous prosodic model and because of the availability of bare singular nouns in special registers, whose use is governed by discourse conditions that children do not yet know (see Baauw, de Roo, and Avrutin (2002) and Avrutin (2004)). Finally, we suggest that when Romance learners discover that the German setting is not valid for their language of exposure, they switch the value of NMP to the category [ arg Cpred]. Nouns are all mapped into the category predicate, and to turn them into arguments they need to be accompanied by articles. So, articles must always be used. This seems to happen at stage 2, when Italian and Catalan learners have almost ceased to omit articles. This change in the parameter value of the Catalan and Italian grammar is triggered by the discovery that the article paradigm is complete in Romance, that is, that it has definite singular and plural articles and indefinite singular and plural articles, i.e., partitive articles like dei bambini (which roughly corresponds to English some children) (see also Chierchia et al. (1999)).15 At stage 2, Dutch learners still omit articles, but much less than at stage 1. As in stage 1, this depends on the Germanic setting of the NMP and again on the availability of bare singular nouns in Dutch special registers (as we already saw for stage 1). In fact, we claim that no change in the setting of the NMP 15 One might think of the use of expletive articles, that is, articles with proper nouns (see Marinis (1999)), which in Catalan are obligatory, as triggers. However, as pointed out by an anonymous reviewer, expletive articles are also possible in some Germanic languages, e.g., German; therefore, it is unlikely that they could act as triggers for the Romance setting of the NMP.


ARTICLE OMISSION

115

occurs in Dutch at stage 2. Children improve because they have established the mapping of more nouns than at stage 1, but they still omit articles at times because the Germanic setting of the NMP requires that the child decides for each newly learned noun whether it is [Carg] or [Cpred] and this decision takes time; moreover, they may not have learned the pragmatic conditions for the use of bare singular nouns. As we said earlier, because of the setting of the NMP, Dutch-speaking children may be led to misclassify some countable nouns, such as apple, as mass, as is the case for furniture or hair. A misclassification in the opposite direction is implausible, that is, nouns of substances like water are not expected to be classified as count nouns, according to Chierchia et al. (1999), because the child’s early ontology is fairly structured, including a robust notion of a solid object that contrasts sharply with that of a liquid or paste, as shown by Soja, Carey, and Spelke (1991). Such an ontological distinction is clearly related to (but does not coincide with) the grammatical notion of mass vs. count and may help the child in classifying mass nouns, at least in the clear cases. For nouns like furniture, hair, or count nouns, the decision as to whether they are count or mass is more problematic, as there is no reliable extra-linguistic guide. We can identify the atoms in the case of count nouns (the smallest sample of the kind), but we can do so also in the case of mass nouns like furniture or hair and this may lead one to take these mass nouns as count. Thus, the child may be confused in the classification of nouns and only be able to rely on the grammatical context. Evidence for the view that English- (and presumably Germanic-) speaking children find it difficult to classify nouns like furniture comes from Gordon’s (1985) study. There, it was shown that 3;0 to 5;11-year-olds rely on syntactic contexts to define subcategories of nouns. When syntax is ambiguous, a noun is classified as mass when it first names a plurality of objects and as count when it first names a single object (see also Bloom (1994)). Finally, Barner and Snedeker (2005) tested 4-year-old children and adults with flexible nouns that could either be mass or count, such as paper/s, stone/s. They found that both populations tend to answer questions (with count syntax) like Who has more stones? by pointing at the figure with three small stones as opposed to the figure with one big stone. In contrast, when they were asked the question (with mass syntax) Who has more stone?, they make that choice less often. Interestingly, however, although a clear difference in responses was observed both in children and adults, some errors were observed in children when the question with mass syntax was used; in fact, 25% of the responses were based on numerical quantity, i.e., children chose the figure with three small stones, although the question with mass syntax was employed. On the basis of this fact, it seems fair to conclude that the ontological distinction can help Dutch- (or generally Germanic-) speaking children in the grammatical classification of a noun as [Carg] in the clear cases. In the less clear cases, when atoms are fairly easy to individuate, children must rely on the syntactic context to discriminate between those nouns that have a mass syntax and need to be mapped into [Carg] and those that have a count syntax and need


116

GUASTI ET AL.

to be mapped into [Cpred]. Nouns whose atoms are easy to identify pose a problem of classification to children, as they are compatible with a mass and a count syntax. Thus, as the same noun can be used with a mass or count syntax, children may be confused and sometimes misclassify it. A parametric approach such as the one here may seem at odds with some current thinking, since in the last 15 years children have been shown to know a lot about grammar and to make fine-grained distinctions that are not immediately apparent. A case in point is the optional omission of subjects by children speaking languages other than Spanish or Italian. The first proposal aimed at accounting for this phenomenon was Hyams’ (1986) and was a parametric approach. In a nutshell, it was proposed that children speaking English, in the period in which they were omitting subjects, omitted them because they had misset the null subject parameter, which later would be reset. Several works in the years to follow showed that subject omission by children speaking those languages that require overt subjects had different distributional properties than subject omission in the null subject languages and this was taken as evidence against a parameter setting and resetting approach. Whether a parametric approach to acquisition is viable or not is an empirical question, however. One recent proposal about subject omission by children is again a parameter setting and resetting approach. Rizzi (2002) argues that children’s omission of subjects results from the adoption of a parameter value (the root null subject parameter) valid in some adult languages, a parametric choice due to the limited resources of the children’s computational system. Thus, a parametric approach cannot be ruled out a priori. That there are parameters set at a very early stage (see Wexler (1998)) does not imply that all parameters are; in this respect, our approach is in line with current research in language acquisition. In summary, the different developmental patterns of children’s omission of articles cannot be explained by input frequency or prosodic scaffolding. An account that would fit the current data well and that can be integrated with other factors is one in terms of the NMP, plus a local uncertainty about the classification of nouns as [Carg] (mass) or [Cpred] (count) that has to be resolved lexical item by lexical item. The setting of the NMP presumably interacts with prosodic constraints in that, once children discover that articles are to be used (stage 1), production is facilitated if they are helped by prosody either in the way proposed by Lleó and Demuth (1999) or by Giusti and Gozzi (2007). This, coupled with another peculiar property of Dutch special registers, namely, the availability of bare singular nouns, is deemed to explain why at stage 1 Romance learners omit fewer articles than Dutch learners, although both groups of learners are assuming the Germanic setting of the NMP. At stage 2, once Romance children figure out that nouns must be mapped into [Cpred], they stop omitting articles, while Dutch children continue to omit them because of the lexical component of the Germanic setting of the NMP and of the need to figure out the pragmatic conditions operating in special registers and allowing omission of articles with bare singular nouns.


ARTICLE OMISSION

117

ACKNOWLEDGMENTS Maria Teresa Guasti acknowledges the support from the PRIN National project “Hierarchical structures and recursivity in natural languages.” Anna Gavarró wishes to acknowledge the financial support of the Ministerio de Educación y Ciencia/FEDER through project HUM2006–13295–CO2–01 and the Generalitat de Catalunya through SGR (2005SGR–00753).

REFERENCES Avrutin, S. (2004) “Beyond Narrow Syntax,” in L. Jenkins, ed., Genetics of Language, Elsevier, London. Avrutin, S. and D. Brun (2001) “The Expression of Specificity in a Language Without Determiners: Evidence from Child Russian,” in A. H.-J. Do, L. Domínguez, and A. Johansen, eds., Proceedings of the 25th Annual Boston University Conference on Language Development, Cascadilla Press, Somerville, MA, 70–81. Baauw, S., E. de Roo, and S. Avrutin (2002) “Determiner Omission in Language Acquisition and Language Impairment: Syntactic and Discourse Factors,” in B. Skarabela, S. Fish, and A. H.-J. Do, eds., Proceedings of the 26th Annual Boston University Conference on Language Development, Cascadilla Press, Somerville, MA, 24–35. Barner, D. and J. Snedeker (2005) “Quantity Judgments and Individuation: Evidence That Mass Nouns Counts,” Cognition 97, 41–66. Bates, E., P. S. Dale, and D. Thal (1995) “Individual Differences and Implication for Theories of Language Development,” in P. Fletcher and B. MacWhinney, eds., Handbook of Child Language, Blackwell, Oxford. Bloom, P. (1994) “Semantic Competence as an Explanation for Some Transitions in Language Development,” in Y. Levy, ed., Other Children, Other Languages: Theoretical Issues in Language Development, Lawrence Erlbaum, Hillsdale, NJ. Bohnacker, U. (1997) “Determiner Phrases and the Debate on Functional Categories in Early Child Language,” Language Acquisition 6, 49–90. Bohnacker, U. (2005) “Frequency Pattern in Child Swedish and Parental Speech,” paper presented at the International Association for the Study of Child Language (IASCL), Freie Universität, Berlin. Bol, G. W. (1996) “Optional Subjects in Dutch Child Language,” in C. Koster and F. Wijnen, eds., Proceedings of the Groningen Assembly on Language Acquisition, 125–135. Bonet, E. (2002) “Cliticització,” in J. Solà, M. R. Lloret, J. Mascaró, and M. Pérez-Saldanya, eds., Gramàtica del català contemporani, Editorial Empúries, Barcelona. Booij, G. E. (1995) The Phonology of Dutch, Clarendon Press, Oxford. Bottari, P., P. Cipriani, and A. M. Chilosi (1993/4) “Protosyntactic Devices in the Acquisition of Italian Free Morphology,” Language Acquisition 3, 327–369. Caprin, C. and M. T. Guasti (2006) “The Acquisition of Morpho-Syntax in Italian: A Cross-sectional Study,” manuscript, University of Milano-Bicocca. Caprin, C. and C. Ioghà (2007) “Article Omission and the Role of the Root,” in A. Belletti, E. Bennati, C. Chesi, E. di Domenico, and I. Ferrari, eds., Language Acquisition and Development— Proceedings of GALA 2005, Cambridge Scholars Press, Newcastle, 96–104. Caselli, M. C., P. Casadio, and E. Bates (2001) “Lexical Development in English and Italian,” in M. Tomasello, and E. Bates, eds., Language Development. The Essential Readings, Blackwell, Oxford. Chierchia, G. (1998) “Reference to Kinds Across Languages,” Natural Language Semantics 6, 339– 405.


118

GUASTI ET AL.

Chierchia, G., M. T. Guasti, and A. Gualmini (1999) Nouns and Articles in Child Grammar and the Syntax/Semantics Map, manuscript, University of Milano-Bicocca and Maryland. Paper presented at Generative Approaches to Language Acquisition (GALA), Potsdam. Cipriani, P., P. Pfanner, A. Chilosi, L. Cittadoni, A. Ciuti, A. Maccari, N. Pantano, L. Pfanner, P. Poli, S. Sarno, P. Bottari, G. Cappelli, C. Colombo, and E. Veneziano (1989) Protocolli diagnostici e terapeutici nello sviluppo e nella patologia del linguaggio (1/84 Italian Ministry of Health), Stella Maris Foundation. Crisma, P. and E. Tomasutti (2000) “Phonological Effects on Article Omission in the Acquisition of Italian,” in S. C. Howell, S. A. Fish, and T. Keith-Lucas, eds., Proceedings of the 24th Boston University Conference on Language Development, Cascadilla Press, Somerville, 220–231. Deen, K. (2005) The Acquisition of Swahili, John Benjamins, Amsterdam/Philadelphia. De Lange, J. (2004) “Article Omission in Child Language and Headlines,” in A. Kerkhoff, J. De Lange, and O. Sadeh Leicht, eds., Yearbook 2004, Utrecht Institute of Linguistics, Utrecht, The Netherlands. Eisenbeiss, S. (2000) “The Acquisition of the Determiner Phrase in German Child Language,” in M. A. Friedemann and L. Rizzi, eds., The Acquisition of Syntax, Longman, Harlow, UK. Erbaugh, M. S. (1992) “The Acquisition of Mandarin,” in D. I. Slobin, ed., The Cross-Linguistic Study of Language Acquisition, vol. 3. Lawrence Erlbaum, Hillsdale, NY. Giusti, G. and R. Gozzi (2007) “The Acquisition of Determiners: Evidence for the Full Competence Hypothesis,” in A. Belletti, E. Bennati, C. Chesi, E. di Domenico, and I. Ferrari (eds.), Language Acquisition and Development—Proceedings of GALA 2005, Cambridge Scholars Press, Newcastle, 232–238. Gordon, P. (1985) “Evaluating the Semantic Categories Hypothesis: The Case of the Count/Mass Distinction,” Cognition 20, 209–242. Guasti, M. T., J. de Lange, A. Gavarró, and C. Caprin (2004) “Article Omission: Across Child Languages and Across Special Registers,” in J. van Kampen and S. Baauw, eds., Proceedings of GALA 2003, LOT Occasional Series, Vol. 1, Utrecht, 199–210. Guasti, M. T. and A. Gavarró (2003) “Catalan as a Testing Hypothesis Concerning Article Omission,” in B. Beachley, A. Brown, and F. Conlin, eds., Proceedings of the 27th Annual Boston University Conference on Language Development, Cascadilla Press, Somerville, MA, 288–298. Gussenhoven, C. (1985) “Over de Fonologie van Nederlandse Clitica,” Spektator 15, 185–200. Hyams, N. (1986) Language Acquisition and the Theory of Parameters, Reidel Publishing Company, Dordrecht, H1, The Netherlands. Hoehle, B. and J. Weissenborn (2003) “German-Learning Infants’ Ability to Detect Unstressed Closed-Class Elements in Continuous Speech,” Developmental Science 6, 122–127. Hoekstra, T., N. Hyams, and M. Becker (1996) “The Underspecification of Number and the Licensing of Root Infinitives,” in E. Hughes, M. Hughes, and A. Greenhill, eds., in Proceedings of the 21st Annual Boston University Conference on Language Development, Cascadilla Press, Somerville, MA, 293–306. Kedar, Y., M. Casasola, and B. Lust (2004) “24-Month-Old Toddlers’ Sensitivity to the Syntactic Role of Function Words in English Sentences: Noun Phrase Determiners,” Child Language Research Forum, Stanford University, Palo Alto, CA. Küpish, T. (2004) The Acquisition of Determiners in Bilingual German-Italian and German French Children, Doctoral dissertation, University of Hamburg. Lleó, C. (2001) “The Emergence of the Article in the Early Acquisition of Spanish and German,” in J. Weissenborn and B. Hoehle, eds., Approaches to Bootstrapping, John Benjamins, Amsterdam/ Phila. Lleó, C. and K. Demuth (1999) “Prosodic Constraints on the Emergence of Grammatical Morphemes: Crosslinguistic Evidence from Germanic and Romance Languages,” in A. Greenhill, H. Littlefield, and C. Tano, eds., Proceedings of the 23rd Annual Boston University Conference on Language Development, Cascadilla Press, Somerville, MA, 407–418. Longobardi, G. (1994) “Reference and Proper Names: A Theory of N-movement in Syntax and Logical Form,” Linguistic Inquiry 25, 609–665.


ARTICLE OMISSION

119

MacWhinney, B. and C. Snow (1985) “The Child Language Data Exchange System,” Journal of Child Language 12, 271–296. Marinis, T. (1999) “Minimal Inquiries and Acquisition of the Definite Article in Modern Greek,” in Proceedings of the 34th Colloquium of Linguistics, Peter Lang Verlag, Frankfurt, 73–82. Marinis, T. (2003) The Acquisition of the DP in Modern Greek, John Benjamins, Amsterdam/ Philadelphia. Mascaró, J. & G. Rigau (2002) “Introduction. The Grammar of Clitics,” Catalan Journal of Linguistics 1, 9–15. Nespor, M. and I. Vogel (1986) Prosodic Phonology, Foris, Dordrecht. Penner, Z. and J. Weissenborn (1996) “Strong Continuity, Parameter Setting, and Trigger Hierarchy,” in H. Clahsen, ed., Generative Perspectives on Language Acquisition, John Benjamins, Amsterdam/Philadelphia, 161–200. Pérez-Leroux, A.-T. and T. Roeper (1999) “Scope and the Structure of Bare Nominals: Evidence from Child Language,” Linguistics 37, 927–960. Rizzi, L. (2002) “On the Grammatical Basis of Language Development: A Case Study,” manuscript, University of Siena, Italy. Soja, N., S. Carey, and E. Spelke (1991) “Ontological Categories Guide Young Children’s Indications of Word Meaning: Object Terms and Substance Terms,” Cognition 38, 179–211. Stromswold, K. (1990) Learnability and the Acquisition of Auxiliaries, Doctoral dissertation, MIT, Cambridge, MA. Valian, V. (1991) “Syntactic Subjects in the Early Speech of American and Italian Children,” Cognition 40, 21–81. van Kampen, N. J. (1994) “The Learnability of the Left Branch Condition,” in R. Bok-Bennema and C. Cremers, eds., Linguistics in the Netherlands 1994, John Benjamins, Amsterdam. Veneziano, E. and H. Sinclair (2000) “The Changing Status of ‘Filler Syllables’ on the Way to Grammatical Morphemes,” Journal of Child Language 27, 461–500. Wexler, K. (1998) “Very Early Parameter Setting and the Unique Checking Constraint: A New Explanation of the Optional Infinitive Stage,” Lingua 106, 23–79.

Submitted 24 February 2006 Final version accepted 27 September 2007

Language Acquisition

Language Acquisition

Suggest Documents

Language Acquisition

Second Language Acquisition - Mathematics

Second Language Vocabulary Acquisition

Second Language Acquisition

READING 1 Language Acquisition

Second Language Acquisition

Second Language Acquisition

Sign Language Acquisition

Language Teaching Methodology and Second Language Acquisition

Second Language Acquisition and Language Teaching - Eric

Language Teaching Methodology and Second Language Acquisition

second language acquisition and the language curriculum

Oral Language Development in Second Language Acquisition

Child Language Acquisition Is Language Behavior?

Second Language Acquisition and Second Language Learning ...

Second Language Acquisition and First Language Phonological ...

Emergentism and Second Language Acquisition*

First Language Acquisition Theories - Clientes

Second language English morphosyntax acquisition

Philosophy of Second Language Acquisition

On Automated Language Acquisition

Episodic Organization and Language Acquisition

FEEDBACK IN SECOND LANGUAGE ACQUISITION

Language Acquisition in Computers