2.1 The monolingual children

Longitudinal studies

Two monolingual children were analyzed: Grégoire (French) and Martina (Italian). The longitudinal data of both children are taken from the CHILDES database (see MacWhinney and Snow, Reference MacWhinney and Snow1985). Table 3 gives a survey of their age in years;months,days, the (word-based) MLUw value, and the standard deviation (from MLUw).Footnote ⁴

Table 3.

The monolingual French and Italian children.

Cross-section studies

In addition to the longitudinal monolingual studies introduced in the preceding subsection, further studies on pronoun acquisition by monolingual French and Italian children are relevant for the present study and enlarge the monolingual database. These studies will be briefly presented below.

Jakubowicz et al. (Reference Jakubowicz, Müller, Ok-Kyung, Riemer, Rigaut, Springfellow, Cahana-Amitay, Hughes and Zukowski1996, Reference Jakubowicz, Müller, Riemer, Rigaut, Hughes, Hughes and Greenhill1997) conducted an elicited production task with monolingual German and French children between the ages 2 and 3. In addition to this test, each session also contained about 30 minutes of recordings of spontaneous speech on which the following analysis is exclusively based. Tiedemann (Reference Tiedemann1999) conducted the same test with monolingual Italian children.Footnote ⁵Tables 4 and 5 provide a summary of the number of children, their age in years;months, days, the (word-based) MLUw value, and the standard deviation. Two groups of children were set up: in group 1, children had a mean MLUw of ≤ 3 and (almost) lacked structures related to the adult C-system (which include subordinate clauses introduced by a lexical complementizer). Group 2 children had a mean MLUw > 3 and used the relevant structures productively.

Table 4.

Monolingual Italian children in elicited production task.

Table 5.

Monolingual French children in elicited production task.

2.2 The bilingual children

For the purpose of the present article, we analyzed two German–French and two German–Italian bilingual children, all of them recorded in the research project “Bilingualism in Early Childhood: Comparing German/Italian and German/French”.Footnote ⁶ All of these bilingual children are being raised in Hamburg; their parents follow the strategy of “one parent – one language” (see Ronjat, Reference Ronjat1913). Except for Céline, the mothers are native speakers of French or Italian and the fathers are German. Céline's mother is German and her father is French. Table 6 shows the number of chosen recordings and the age span in years;months,days as well as the range for the MLUw, base (number of utterances) and standard deviation (SD) of the four children in their Romance language.

Table 6.

The bilingual German–French and German–Italian children.

2.3 Analysis

The following analysis of the data focuses on the use of different types of pronouns in the acquisition of French and Italian. In particular, the order of acquisition of strong and clitic pronouns in subject and object position is investigated, as well as that of reflexive and non-reflexive object clitics and subject (nominative) clitics. Both the onset of acquisition and development towards the target are studied, where development is considered successful or target-like when the adult norm is reached.Footnote ⁷ We will refer to the adult data represented in Figure 1 (cf. section 2). The figures showing occurrences of subject and object pronouns are based on percentages.Footnote ⁸ Information about DP use (la fille, la ragazza “the girl”) is added, allowing the reader to compare it with pronoun use and to infer the omission rates.Footnote ⁹

Pronouns in monolingual longitudinal studies

Subject and object pronouns

This section compares the use of strong and clitic pronouns in subject and object position in the acquisition of French and Italian by the monolingual children. Figure 2 shows for the monolingual French child Grégoire that he produces both strong and clitic subject pronouns from the first recording onwards, reaching the adult rate of 80% of subject clitics at age 2;5,1.

Figure 2.

Clitic versus strong subject pronouns in Grégoire.

In the object position, Grégoire uses clitics rather intermittently (maximally to 20%, only half of the percentage of the French adults in Figure 1) and only regularly, i.e., in every recording session, from 2;3 onwards. The same holds for strong object pronouns, as Figure 3 shows. The few occurrences of strong pronouns appear one recording session later than the first clitics, i.e., within a time span of 13 days, which can be considered to be simultaneously.

Figure 3.

Clitic versus strong object pronouns in Grégoire.

Comparing both positions, both strong and clitic object pronouns appear clearly later than strong and clitic subject pronouns, and objects are mainly realized as DP.

The Italian monolingual child Martina uses strong subject pronouns from the first recording on. They increase from less than five in the first recordings up to a maximum of about 30 at the age of 2;4,13. Although it is not possible to represent these rare occurrences in a percentage-based graph, we would like to add this information since the rare use of strong pronouns is caused by the null subject property of Italian and reflects adult use.

In object position, Martina initially uses clitic pronouns at a rate of less than 20%, i.e., only at half the rate of the Italian adults in Figure 1. The number of object clitics increases during development, reaching the adult rate of 40% in the last recording as Figure 4 shows. Both strong and clitic object pronouns are employed at this time, but strong pronouns appear later in object position (not in the first two recordings) than in subject position. Strong object pronouns are used regularly (i.e., in every recording) from 1;11,20 on to a maximum rate of 20% (as the adults' rate). In the beginning, a large number of objects are realized as DP, as in Grégoire.

Figure 4.

Clitic versus strong object pronouns in Martina.

The comparison between the use of strong and clitic pronouns in subject and object position by monolingual children clearly shows that subject clitics and strong subject pronouns start to be used at the same time. Subject clitics appear before object clitics. In object position, however, pronouns are in general acquired later than in subject position. Once objects are pronominalized, strong pronouns and object clitics appear simultaneously.

While the observations above emphasize the onset of acquisition, we can also state that both monolingual children reached the adult rates of clitic use in both subject (French) and object (French and Italian) position.

Reflexives

In the following, the first appearances and total number of reflexive object clitics are presented and compared with the onset of both subject and non-reflexive object clitics. Notice, however, that reflexive verbs appear rarely in the spontaneous data. Table 7 shows that Grégoire, the French child, uses reflexive clitics very late, only in the last two recordings, unlike subject clitics, which appear already in the first recording. Also, Martina does not use reflexive clitics from the beginning of the investigation period.

Table 7.

First appearance and total use of reflexive object clitics in the monolingual longitudinal studies.

Table 8 compares the first appearances of reflexive and non-reflexive object clitics.

Table 8.

First appearances of reflexive and non-reflexive object clitics.

Summarizing, the results for the reflexive object clitics show that the monolingual French child acquires reflexive clitics later than subject clitics. With respect to non-reflexive object clitics, the comparison of the moments of acquisition indicates that both monolingual children use non-reflexive object clitics before reflexives.Footnote ¹⁰

Pronouns in the cross-section studies

Subject and object pronouns

This section compares the onset and development of strong and clitic pronouns in subject and object position in the spontaneous production part of the cross-section studies introduced in 2.1. Figure 5 shows that in French, both groups of children use subject clitics to a much higher degree (F1 50%, F2 more than 80%) than strong subject pronouns (less than 10%). The second French group (F2) thus almost reaches the adult rates for both subject clitics and strong subject pronouns. The Italian children use strong pronouns approximately 20% of the time, which corresponds to the Italian adult rate.

Figure 5.

Clitic versus strong subject pronouns in French and Italian.

In object position, a completely different picture emerges: Figure 6 shows that the first groups of French and Italian children (F1 and I1) use slightly more strong object pronouns (F1 10%, I1 22%) than clitics (F1 9%, I1 18%), while the second groups (F2 and I2) produce clearly more clitics than strong pronouns; in French, however, with ca. 20 percentage points less than in subject position and far less than adults (40%), which means that the second groups of French and Italian children had not yet reached the target norm. The Italian children use more object clitics (ca. 42%) than the French children. This rate is also higher than that of DP objects while in both French groups and the first Italian group the highest rate of realized objects consists of DP's as in the longitudinal data from Grégoire and Martina (cf. also Müller, Schmitz, Cantone and Kupisch, Reference Müller, Schmitz, Cantone, Kupisch, Torrens and Escobar2006, who find parallel developments in their cross-sectional study).

Figure 6.

Clitic versus strong object pronouns in French and Italian.

With regard to strong object pronouns we observed that French children of both groups (F1 and F2) use them rather rarely (ca. 10%), while the Italian children of the first group (I1) produce them ca. 22% of the time. In group I2, this rate decreases to ca. 10% (mostly in favor of object clitics). Both French and Italian adults, however, use strong objects around 20% of the time. This rate is not reached by either of the second groups (F2 and I2).

These results can be summarized as follows: strong subject pronouns and subject clitics are acquired at the same time as strong object pronouns. The developmental process of acquisition of object clitics clearly takes place after the acquisition of subject clitics. The adult norm for subject clitics is almost reached by F2 children, while the adult rate for object clitics is still far from having been reached. These observations show that although the acquisition process towards the target norm of strong object pronouns starts very early, it still overlaps with that of object clitics.

Reflexives

Following the same procedure as in the cross-section studies, we compare the onset of use of reflexive object clitics with that of subject and non-reflexive object clitics.

The analysis of the reflexive pronouns in French monolingual children is based on the results taken from Jakubowicz and Rigaut (Reference Jakubowicz and Rigaut2000), an expanded version of the investigation in Jakubowicz et al. (Reference Jakubowicz, Müller, Ok-Kyung, Riemer, Rigaut, Springfellow, Cahana-Amitay, Hughes and Zukowski1996, Reference Jakubowicz, Müller, Riemer, Rigaut, Hughes, Hughes and Greenhill1997) with new data from the original and several new children. The criteria for group division (MLUw, C-system and finite verb production) correspond to the criteria applied in the present study. Spontaneous interaction and elicited production are clearly separated. In contrast to Jakubowicz and Rigaut (Reference Jakubowicz and Rigaut2000), only accusative reflexive pronouns in spontaneous production are counted here. Figure 7 shows that reflexives were only used by the second French group while both subject and non-reflexive object clitics were also used by the first French group (see Figures 5 and 6).

Figure 7.

Reflexives in French monolingual children.

The analysis of reflexive pronouns in Italian is based on the study of Tiedemann (Reference Tiedemann1999). Figure 8 shows that reflexive pronouns are acquired earlier in Italian than in French, the reflexive si being used already by the group 1 children.

Figure 8.

Reflexives in Italian monolingual children.

The results for the ongoing acquisition of reflexives can be summarized as follows: the acquisition of subject clitics takes place before the acquisition of reflexive object clitics, as the French study clearly shows. The order of acquisition of non-reflexive and reflexive object clitics seems to depend on the language: French group 1 children exhibited a very low percentage of non-reflexive object clitics (ca. 10%) and they did not use reflexives at all. Italian children in group 1 used more non-reflexive object clitics (18%) than their French peers, and they also produced reflexive object clitics (ca. 10%). The children of the second French group (F2) produced reflexives at a rate of ca. 15% while the Italian children in the second group (I2) already produced reflexives at a rate of ca. 40%.

We may thus conclude that reflexive object clitics pattern like non-reflexive object clitics in both languages and that Italian children have less difficulty producing these types of pronouns as compared to French children.

Pronouns in bilingual longitudinal studies

Subject and object pronouns

This section compares the use of strong and clitic pronouns in subject and object position in French and Italian by the bilingual children. Starting with the German–French bilingual child Alexander, we can observe in Figure 9 that he uses strong subject pronouns and subject clitics simultaneously from the first recording onwards. By the age of 2;6 he quantitatively approaches the target system's use of subject clitics (a rate of more than 80%, cf. Figure 1), thus discovering the near-obligatoriness of subject clitics. Until then, clitics are used at a rate of ca. 80%. Strong subject pronouns are also used in every recording, but to a very low average degree of less than 10%, which corresponds to the adult use.

Figure 9.

Clitic vs. strong subject pronouns in Alexander.

Figure 10 shows that Alexander does not use object clitics in the first two recordings but only strong pronouns. Object clitics are produced from 2;2,27 onwards, but to varying degrees of far below the adult rate (e.g., 10% at 2;7,27) and also exceeding the adult rate (e.g., 50% at age 2;8,28). A consistent use corresponding to the adult rate of 40% cannot be observed during the investigation period. The onset of use of object clitics, however, clearly takes place after that of strong object pronouns. Overall, Alexander prefers to realize objects as DP's like the monolingual children (cf. subsections above).

Figure 10.

Clitic vs. strong object pronouns in Alexander.

Figure 11 shows that the other bilingual German–French child, Céline, behaves similarly with respect to subjects: from the first recordings onwards, she produces both strong and clitic subject pronouns (even in the period from 2;0 to 3;1 where she speaks much less French than Alexander).Footnote ¹¹ Interestingly, she starts to use subject clitics at the 80% level from 3;3,12 onwards, i.e., much later than Alexander as far as each child's respective age in concerned. Notice however that their MLUw is comparable: Alexander has reached an MLUw of 3.5, Céline an MLUw of 3.4 when they use nominative clitics at the 80% level, i.e., at almost the adult rate. Unlike Alexander, Céline does not use strong subject pronouns in every recording and to a lesser degree than the adults, but the onset of production occurs at the same time as the onset of production of subject clitics, as was the case with Alexander and the monolingual children.

Figure 11.

Clitic vs. strong subject pronouns in Céline.

By contrast, Céline uses no clitic pronouns in object position until 3;3,12. Like Alexander, she already uses strong pronouns in the object position during the very first recordings. The adult 40% level for object clitics is reached only at 3;7,17 for the first time, as Figure 12 shows. Apart from one recording (3;4,23), Céline uses strong object pronouns in every recording with rates of up to 40%, reaching the 20% adult level only from 3;8,14.

Figure 12.

Clitic vs. strong object pronouns in Céline.

We now turn to the German–Italian bilingual children. The bilingual German–Italian child Carlotta shows an increasing use of strong subject pronouns from 2;0 onwards. Although we cannot represent the overall rare use of strong subject pronouns (due to the null subject property, as with Martina) the onset of the use of strong subject pronouns is relevant since it corresponds to the onset of regular use of strong object pronouns (2;0,11) as Figure 13 shows. In the object position, Carlotta also uses only strong pronouns until 2;3,2 which is the onset of using object clitics. But a regular use, i.e., in every recording, takes place only from 2;4,21 on. Then Carlotta produces them, like the bilingual German–French child Alexander, to highly varying rates, both below the adult rate (ca. 10% at e.g., 2;11,13) and up to ca. 60% (e.g., at 2;4,21), then exceeding the adult rate of ca. 40% object clitics. An adult-like use of ca. 40% is reached in the last recording (3;0,25). Like the German–French child Céline, Carlotta does not use strong subject pronouns in every recording. If they are used, however, they mostly appear at a rate of ca. 20%, which corresponds to the adult rate.

Figure 13.

Clitic vs. strong object pronouns in Carlotta.

Lukas, the second bilingual German–Italian child, behaves similarly: in the subject position he uses strong pronouns from 2;2 onwards (to a varying rate). Compared with the object position, Figure 14 shows that Lukas first uses only strong object pronouns (from 2;4,9 onwards) and starts to use clitic pronouns slightly later (at 2;4,23). The onset of the use of strong pronouns in both subject and object position thus clearly precedes that of the object clitics. Like Carlotta (and the German–French child Alexander), Lukas uses object clitics to a varying degree (up to 50%, then exceeding the adult rate of ca. 40% object clitics). The adult rate is met only in the last recording (3;1;16). Lukas uses strong object pronouns in every recording from 2;4 on, but mostly to a rate of 10% and less. Not until the last recording does their use correspond to the adult rate of 20%.

Figure 14.

Clitic vs. strong object pronouns in Lukas.

We can summarize the results of this section as follows: strong subject pronouns and subject clitics are acquired at the same time as strong object pronouns. Subject clitics are acquired clearly before object clitics. Strong object pronouns are acquired before object clitics.

Reflexives

As in the preceding studies, we compare the onset of the use of reflexive object clitics with subject and non-reflexive object clitics. The first appearance of SE and SI and the absolute number of produced reflexive clitics in the investigation periods are shown in Table 9.

Table 9.

First appearance and total number of reflexive object clitics.

A comparison of the appearance of reflexive clitics with the one of non-reflexive clitics in Table 10 shows that non-reflexive object clitics are used somewhat earlier than reflexive object clitics by three of the four children – in the French-German child Céline both appear at the same time and quite late, her overall development in French being slowed down. Notice, however, that reflexive verbs occur rarely in the spontaneous data of all children.

Table 10.

Comparison of first occurrences of non-reflexive and reflexive object clitics.

A summary of the results for the acquisition of reflexive clitics by bilingual children demonstrates the following: subject clitics are clearly acquired before reflexive object clitics. The order of acquisition of non-reflexive and reflexive object clitics is less clear: both are acquired either at the same time or reflexive object clitics slightly later than non-reflexive object clitics or vice versa.Footnote ¹²

In comparing the monolingual and the bilingual data, we can determine the following similarities and differences:

1. • Subject clitics appear before object clitics.

2. • Subject clitics appear at the same time as strong subject pronouns.

3. • In elicited production of monolinguals and longitudinal bilingual data strong object pronouns appear at the same time as subject clitics and strong subject pronouns.

4. • In the longitudinal monolingual data object pronouns appear generally later, i.e., independently of their status as clitic or strong pronoun.

5. • Reflexive clitics appear much later that subject clitics.

6. • The appearance of reflexive clitics in relation to non-reflexive clitics is random, i.e., they can appear simultaneously or one before the other.

The general conclusion is that the order of appearance of clitic and strong pronouns shows no difference in monolingual and bilingual children. The only difference is between the monolingual children of the longitudinal studies and those of the cross-sectional studies regarding the strong object pronouns, which might be due to the investigation of only two longitudinal studies compared to a much higher (and more representative) number of children in the cross-sectional studies. The following sections will examine the internal and external structure of clitics and strong pronouns in the two Romance languages and will explain the acquisition facts on the basis of this structure.

3.1 Clitics as DPs with internal difference

The first type of approach within generative theory argues that clitics differ from other pronouns in their internal structure. Some researchers have argued for their status as heads (mostly assumed to be D(eterminer)-like elements) in syntax (with a pro – an empty pronominal – in argument position which satisfies Case- and Theta-requirements; e.g., Roberge, Reference Roberge, Wanner and Kibbee1991), others for their status as elements which are inserted into syntax as categories of some type (XPs) but which move as heads (Xs) (e.g., Belletti, Reference Belletti and van Riemsdijk1999; for a discussion, see Müller, et al., Reference Müller, Crysmann and Kaiser1996). In this approach, all pronouns are of the same categorial type (e.g., DP) but their internal structures are different. The main evidence supporting this analysis comes from the observation that Romance object clitics often have the same phonological form as determiners. Furthermore, Raposo (Reference Raposo, Schwegler, Tranel and Uribe-Etxebarria1998) shows that there is a relation concerning the possibility to omit object clitics at the sentence level and to omit determiners at the level of DP.

The predictions for acquisition on the basis of this account are clear: clitics should pattern together (as they at least move as heads) and contrast with strong pronouns, the latter being analyzed uniformly as DPs. The summary of section 2 clearly showed that the predictions are not borne out by the data, neither by the monolingual nor by the bilingual children.

3.2 Clitics as a proper class of pronouns

This section presents in more detail the approaches from Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, Reference Cardinaletti, Starke, Powers and Hamann2000) and Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002), which assign a different categorial status to the various types of pronouns, according to their degree of deficiency. Apart from using different criteria and categorial labels for their classifications, the two approaches treat clitics differently: while Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, Reference Cardinaletti, Starke, Powers and Hamann2000) establish a homogenous class of clitic pronouns, Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002) group the clitics in different classes corresponding to their distinctive properties.

The tripartition of pronouns by Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, Reference Cardinaletti, Starke, Powers and Hamann2000)

Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999) propose a tripartition into strong, weak and clitic pronouns which exhibit different degrees of syntactic, semantic, and morphophonological deficiency, expressed in the hierarchy in (1):

1. (1) Hierarchy of deficiency of pronouns

   

With respect to their syntactic derivation, strong and weak pronouns move as XPs while clitics move as heads. They also differ in the representation of projection layers which are each correlated to functional categories on the sentence level: strong pronouns are full nominal projections, with C° encoding case and referential information, parallel to CP (Complementizer Phrase) in sentence structure. Weak pronouns lack the highest functional projection and are compared to ΣPs, a functional projection between CP and IP (Inflection Phrase) encoding polarity and prosodic features. Clitics lack the two highest functional projections and are thus compared to IPs, the lowest functional category in sentence structure. Cf. the structures in (2).

1. (2)

Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, p. 197) further claim that the choice of a particular pronoun type is regulated by a strong cross-linguistic generalization according to which the most deficient form must be chosen if available. The authors (1999, p. 215) specify that this rule holds only if the deficient pronoun is associated with the same features as those which would have been contained in the strong counterpart. This generalization is then reduced to the statement that a “smaller structure” is necessarily chosen if possible, expressed as the Economy Principle “Minimise Structure” (Cardinaletti and Starke, Reference Cardinaletti, Starke and van Riemsdijk1999, p. 198) and generalized even more to “Minimalise α” (Cardinaletti and Starke, Reference Cardinaletti, Starke and van Riemsdijk1999, p. 201). This principle leads to the following rule: strong elements have priority but a deficient element has to be chosen whenever it is associated with the same features that would be encoded by the corresponding less deficient element. Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, fn. 71) claim that acquisition data show the primacy of strong pronouns in language acquisition: if confronted with a pronoun which in adult grammar can be classified as deficient or strong, a child always resolves ambiguity by postulating only one form, namely the strong one.

To sum up the statements regarding the acquisition of pronouns by Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999), we see that they predict that children react to the primacy of strong pronouns and therefore disobey the Principle Minimise α. It thus appears that two different generalizations hold, one for each period: initially a primacy of strong pronouns and later the Principle Minimise α. Since one of them is always active, the theory cannot be falsified by empirical results: if children start first with the production of strong pronouns, they obey the primacy of strong pronouns; if they start first with the production of clitics they respect Minimise α. What cannot be predicted, though, is the simultaneous acquisition of strong and clitic pronouns and the acquisition of different types of clitics at different points in time. Interestingly, both monolingual and bilingual children seem to acquire strong subject pronouns and subject clitics at the same time, as presented in section 2. Furthermore, subject clitics appeared before object clitics. We can thus conclude that this approach does not correctly explain the appearance of pronouns in child language.

The tripartition of pronouns by Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002)

In what follows we present the approach by Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002), which differs in one important aspect from the one by Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, Reference Cardinaletti, Starke, Powers and Hamann2000): while Cardinaletti and Starke compare the different pronoun types to functional categories at the sentence level, the proposal by Déchaine and Wiltschko refers to the properties of the functional category D in the nominal domain. They propose the following three pronominal types in (3):

1. (3)

Comparing pronouns in Halkomelem, Shuswap, Japanese, English and French, Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002) argue that the clitics which are of interest here, IL, LE/LO and SE/SI, are of the categorial type ΦP.Footnote ^13, Footnote ¹⁴ Strong pronouns are pro-DPs and the clitics en/ne are Pro-NPs. In other words, this approach differentiates among pronouns assigning them a different categorial status. In contrast to Cardinaletti and Starke, clitics are not generally conceived of as the most deficient elements and the class of clitics contains categories of the type pro-ΦP and pro-NP.

Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002, p. 427ff.) apply their approach to French clitics and conclude that en is a pro-NP clitic while 1st/2nd/3rd person subject and object clitics are pro-ΦPs. For the acquisition of the clitics of interest here, their approach predicts first that these clitics are acquired simultaneously and second that strong pronouns (as pro-DPs) and clitics are not acquired at the same time in development, if acquisition depends on the external syntax of pronouns. In essence, on the basis of Déchaine and Wiltschko's work, the same predictions can be made as with Cardinaletti and Starke's approach. As we have seen in section 2, neither prediction is borne out by the data.

In the next section, we will present an approach which combines the internal and external syntactic analysis of pronouns.

3.3 Combining the external and internal syntax of clitics

In what follows we will combine the two approaches from Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999) and Déchaine and Wiltschko (Reference Déchaine and Wiltschko2002) on the one hand, and Raposo (Reference Raposo, Schwegler, Tranel and Uribe-Etxebarria1998) on the other hand, following a proposal made by Gabriel and Müller (Reference Gabriel, Müller and Kaiser2005). We will argue for both an external and internal deficiency of clitics in the two Romance languages. If deficiency relates to external as well as internal syntax, we have the following possibilities, which are discussed in detail in Gabriel and Müller (Reference Gabriel, Müller and Kaiser2005), and which give rise to the typology in (4).

1. (4) A typology of clitics in relation to strong pronouns

   

IL is a pro-ΦP with the most articulated structure, LE/LO is a pro-ΦP as well, but lacking an N-layer, while SE/SI is a Φ-head. Clitics, as a class of Φ-expressions, differ from strong pronouns in that the former are of the categorial type Φ, the latter are DPs. In this respect clitics represent a homogeneous class. However, a crucial aspect of the proposal made by Gabriel and Müller (Reference Gabriel, Müller and Kaiser2005) is the fact that the class of clitics shows quantitative and qualitative differences. Therefore, the proposal is presented in two steps: the first subsection of 3.3 puts forward several reasons why clitics form a pronoun class in their own right which differs from strong pronouns, while the second subsection argues for a further subdivision of the class of clitics. The third and last subsection of 3.3 summarizes the proposal and makes predictions holding both for monolingual and bilingual first language acquisition.

Clitics as a homogeneous class of Φ

The first question is why the clitics IL, LE/LO and SE/SI represent a homogeneous class, differing in well-defined respects from strong pronouns.

It has been shown in the literature (Cardinaletti and Starke, Reference Cardinaletti, Starke and van Riemsdijk1999 among others) that Romance clitics exhibit a homogeneous syntactic, semantic and phonological behavior. On a syntactic level, the deficiency of clitics is evident under coordination. Whereas strong pronouns can be coordinated, clitic pronouns cannot, as shown in (5) for French.Footnote ¹⁵

1. (5)

   1. a.

      *Il et celui de Jean sont beaux. <+h>

   2. b.

   3. c.

If we turn to the semantics of clitics, they can have human and non-human reference. Strong pronouns are restricted to human reference (see e.g., Jakubowicz et al., Reference Jakubowicz, Nash, Rigaut and Gérard1998; Cardinaletti and Starke, Reference Cardinaletti, Starke and van Riemsdijk1999). Both semantic properties assumed hold cross-linguistically and are exemplified in (6) for Italian:

1. (6)

   1. a.

      “Here is the house of Piero. Gianni sees it with pleasure.”

   2. b. Ecco Silvia. Gianni vede lei con piacere. <+h>

      “Here is Silvia. Gianni sees her with pleasure.”

   3. c.

      “I will see Gianni tomorrow.”

   4. d.

      “I don't know if you want the wine now or later.”

Also with respect to the argument status, clitics (LE/LO) are less restrictive than strong pronouns since the former may act as arguments and as predicates. If clitics represent predicates, the unmarked (masculine) form LE/LO is used, as evidenced in (7) for French and Italian.

1. (7)

   1. a.

   2. b.

      “Mary is an advocate and Silvia will it be soon too.”

   3. b′.

      “Mary is an advocate and Silvia will be it soon too.”

Furthermore, clitics are deficient in their phonology. French liaison is possible only with deficient pronouns. In French, the feminine 3rd person strong pronoun and the feminine 3rd person clitic are homophonous. The reading as a strong pronoun is enforced if the pronoun is prosodically prominent (indicated by capitalization in (8a)) or used in the syntactic construction of complex inversion (in (8b)). Complex inversion with a clitic in (8c) is ungrammatical, while the use of a full DP or the strong pronoun is allowed; cf. Cardinaletti and Starke (Reference Cardinaletti, Starke and van Riemsdijk1999, p. 159) for French:

1. (8)

   1. a.

      “They have said the truth.”

   2. b.

      “When did they say the truth?”

   3. c.

      “Did he say the truth?”

In brief, there is ample evidence that clitics differ from strong pronouns as deficient elements displaying a similar degree of deficiency which, in the framework of Déchaine and Wiltschko, would naturally be characterizable in terms of the syntactic category Φ. Furthermore, arguing within this framework, subject and non-reflexive object clitics are typical Binding Principle B pronouns, i.e., they must be free in their governing category and can function as bound variables (cf. French examples in (9) taken from Déchaine and Wiltschko, Reference Déchaine and Wiltschko2002, p. 429). We will argue further below that reflexive clitics cannot be used as evidence. Therefore they are irrelevant for Binding Theory.

1. (9)

   1. a.

   2. b. [Chaque homme]_i pense que Marie veut [l]_i épouser.

   3. c. [Ogni uomo]_i pensa che María [lo]_i voglia sposare.

      “Every man thinks that Mary will marry him.”

However, the split between strong pronouns and clitics is not sufficient for a description of pronouns in French and Italian. There are differences among the types of clitics which can be observed both on a quantitative and a qualitative level. In what follows, we will present these differences and discuss, following Gabriel and Müller (Reference Gabriel, Müller and Kaiser2005), how these may be described in terms of the internal syntactic architecture of clitics.

Different types of clitics

The preceding section has shown that the clitics IL, LE/LO and SE/SI represent a homogeneous class when compared with strong pronouns. Within the present framework, it seems plausible to represent the homogeneity of clitics by the external syntactic category Φ, which all clitics share, since they all encode the phi-feature person, some also number and/or gender. However, there are also qualitative and quantitative differences among the clitics stemming, in our view, from their different internal syntax. As presented in section 2, spontaneous interactions by French and Italian adults show quantitative differences between the realization of the subject and the object position in the two languages which we interpret as reflecting qualitative differences.

Figure 1 in section 2 above shows that the two languages resemble each other in their realizations of the object position. This observation might suggest that the usage of object clitics is regulated by constraints on pronominalization, i.e., by discourse-pragmatic conditions. In other words, the usage of object clitics is not required by syntax, as will be argued for subject clitics in more detail below. Notice further that DPs are quite frequent in object position in both languages. Since objects often encode new information, this observation is not surprising.

If we compare subjects and objects, Figure 1 (see section 2) can be interpreted as revealing that the presence of subject clitics in French and the absence of a phonetically realized element in subject position in Italian are systematic syntactic properties of the two languages.

Figure 1 shows a very small number of subject omissions (ca. 10%) and a very high number of subject clitics (ca. 85%) in French. Interestingly, all subject omissions were omissions of the expletive pronoun il with impersonal verbs like falloir, for example, in constructions as (il) faut boire beaucoup “(one) has to drink a lot”. The high number of subject clitics points to the near-mandatory usage of subject clitics in French. We may thus conclude that the subject position in French must be phonetically realized, as a nominative clitic, as a strong pronoun or as a DP. Within syntactic theory, one may argue that the presence of subject clitics has syntactic reasons (as well as the absence of a phonetically realized element in Italian). Some researchers, like Jakubowicz et al. (Reference Jakubowicz, Nash, Rigaut and Gérard1998), have suggested that subject clitics are markers of finiteness, since they are restricted to finite clauses in French. However, not only clitics but also strong pronouns and DPs are excluded in non-finite clauses. Therefore, the near-mandatory nature of clitic pronouns in subject position seems rather to stem from the fact that French is not a pro-drop language (in contrast to Italian). The presence of a subject in SpecTP (Tense Phrase), or SpecIP in older analyses, has been attributed to the Extended Projection Principle (EPP) which claims that syntax always projects a subject position, even in the absence of an argument which may fill this position (Chomsky, Reference Chomsky1982). Within the minimalist framework the EPP has been modified to apply to every functional head, not just of T. The EPP is a parameterized property crucially related to the presence of the EPP feature on the head of a functional category. If present, it accounts for the presence of a mandatory specifier in the functional phrase since the EPP feature must be checked.Footnote ¹⁶ Quantitative analysis of the French corpus indicates that it is reasonable to ascribe the checking of the EPP feature to the clitic. We might say that subject clitics are expressions of two different functions: (i) they lexicalize the phi-features of T and check off the EPP feature which in turn forces the set of phi-features to be spelled out, and (ii) they act as arguments. Let us add that the assumption that subject clitics can satisfy the EPP like DPs does not imply that they are analyzed syntactically as DPs. In essence, we have seen in the preceding section that strong pronouns (like lexical DPs) are DPs, while subject clitics are ΦPs (like the other clitics). Subject clitics share with strong subject pronouns and lexical DPs that they are able to check off the EPP feature, a property which we would like to relate to referentiality or presence of an N-layer.

We have argued that the quantitative differences between subject and object clitics can be interpreted in terms of qualitative differences. There are also distinct qualitative differences between subject and object clitics in French.Footnote ¹⁷ Subject clitics can have wide scope over a coordinated VP, as can full DPs. Object clitics pattern differently in this respect: they have to be repeated with each VP, unlike a DP argument, which may undergo ellipsis as the examples in (10) show.Footnote ¹⁸ We may account for this difference by assuming that subject clitics resemble DP arguments in providing an N-layer, while object clitics are pure functional projections. If this is plausible, subject clitics do not differ from full DP arguments with respect to their internal structure. Object clitics, on the other hand, are purely functional categories which lack an N-layer, i.e., they lack referential properties altogether.Footnote ¹⁹

1. (10)

   1. a. Il/L'étudiant travaille à son article et pense à ses problèmes.

      “He/The student works at his article and thinks about his problems.”

   2. b.

      “Paul hates her and considers her to be very stupid.”

      (Kayne, 1975, p. 97)

   3. b′.

      “Jean will wash and eat the apple which he bought yesterday.”

The lack of specifically referential characteristics may also be the reason why object clitics in Romance can represent either arguments or predicates, as shown in (7) above. Intuitively speaking, if co-indexed with an argument they can spell out the Φ-features of this argument. Put differently, they can act as arguments due to their co-indexation with an argument. In other cases, the object clitic can represent a predicate; in this situation, it surfaces with default Φ-features. Our approach would then assume that syntactic elements with an N-layer are genuine referential arguments; however, there is another way to act as an argument in syntax, namely by co-indexation with an N-layered element. The former case is indicative of subject clitics, the latter of object clitics. According to a recent proposal by López (Reference López2003), who has analyzed Catalan dislocations, object clitics mark presupposition, which would be the only LF-interpretable feature they express. The different quantitative and qualitative behavior of object and subject clitics we have outlined may be interpreted as evidence that subject and object clitics are not expressions of the same feature matrix. Whereas subject clitics are expressions of the EPP feature, object clitics spell-out [+presupposition].

Let us finally turn to SE/SI, for which it has been argued that it is a head (Zubizarreta, Reference Zubizarreta1982; Zribi-Hertz, Reference Zribi-Hertz1987; Crysmann and Müller, Reference Crysmann, Müller, Powers and Hamann2000; Gabriel and Müller, Reference Gabriel, Müller and Kaiser2005). It manipulates the argument structure of the lexical verb with which it appears. French causatives are sensitive to the transitivity of the embedded lexical verb. If the verb has one argument, the former external/internal argument of the embedded verb is realized as a DP. If the verb has more than one argument, the former external argument must be realized as a PP (see Crysmann and Müller (Reference Crysmann, Müller, Powers and Hamann2000) for an overview of the literature). The examples in (11) show that a se+V embedded under faire (11d) behaves syntactically as a verb with one argument (11a) and precludes the possibility of assigning argument status to se.

1. (11)

   1. a.

      “Jean has Paul cry.”

   2. b.

      “Jean has the children wash the dishes.”

   3. c.

      “Jean has the children wash them.”

      (Wehrli, 1986, p. 271)

   4. d.

      “Jean has the children wash themselves.”

      (Wehrli, 1986, p. 271)

We may assume that Italian SI behaves like French SE with regard to the manipulation of the argument structure of lexical verbs in Italian (see e.g., Manzini, Reference Manzini and Borer1986). This view is supported by the fact that Italian ne-cliticization and absolute participle constructions prove that SI functions as an ergativizer, i.e., it blocks the realization of the external argument in syntax. Ne cliticizes parts of internal arguments, not of external arguments (see Burzio, Reference Burzio1986). Since all readings of SI are compatible with ne-cliticization, it is clearly shown that these constructions do not involve an external argument.

1. (12)

   1. a.

      “I-eat three apples.” / “I eat three of them.”

   2. b.

      “Three men sleep.” / “Three of them sleep.”

   3. c.

      “Many children wash themselves.” /

      “Many of them wash themselves.”

Summary of the proposal and predictions for language acquisition

To summarize, we have argued for a similar degree of deficiency of clitics when compared with strong pronouns. Of course, we will have to refine our approach in future research by implementing the observation that the different forms of clitics do not encode all Φ-features (gender, number, person), an extreme case being reflexive se/si, which only encodes person. The purpose of the present article is not to discuss these different Φ-layers.

The discussion of the different clitic types has shown that there are qualitative and quantitative differences among them which we will capture by their different internal syntax, exemplified by the typology in (4), repeated in (13).

1. (13) A typology of clitics in relation to strong pronouns

   

We believe that our typology cogently reflects what we have argued for in the preceding paragraphs: we have shown that clitics differ from strong pronouns, the difference being displayed in terms of external syntax: strong pronouns, as lexical DPs, project a D-layer where clitics are homogeneously deficient in the sense that they project only as high as Φ. The typology can also explain the difference among the clitics: subject clitics are arguments, reflected in the presence of an N-layer, while non-reflexive object and reflexive clitics are pure functional categories, the difference between the two being that reflexives are heads whereas non-reflexive object clitics are maximal projections.

The typology of clitics and strong pronouns which we propose can make two predictions for acquisition:

1. (A) If the acquisition process is guided by the external syntactic architecture of pronouns, strong (subject and object) pronouns – D-elements – should be acquired either before or after clitic pronouns, i.e., all elements of the category Φ.

2. (B) If children rely on the internal syntactic architecture of pronouns, one can predict that strong pronouns and subject clitics – elements which possess an N-layer – should be acquired together, either before or after object clitics which lack the N-layer.

In section 2, we presented acquisition data which demonstrate the following generalizations both for monolingual and for bilingual first language acquisition:

• • strong subject pronouns are acquired at the same time as strong object pronouns – same external (DP) and internal syntax (N-layer), supporting (A) and (B);

• • subject clitics are acquired at the same time as strong subject pronouns – same internal syntax (N-layer), supporting (B);

• • subject clitics are acquired at the same time as strong subject pronouns – same internal syntax (N-layer), supporting (B);

• • subject clitics are acquired before object clitics – different internal syntax, supporting (B);

• • reflexive clitics are acquired at the same time as object clitics – same internal syntax, supporting (B).

This observation can be explained by the following generalization: syntactic categories with an N-layer are acquired before categories which lack this layer (i.e., which are mere functional categories). In other words, internal syntax – not external syntax – matters for acquisition.

In the next section we will discuss our results with respect to language acquisition theory; in particular, we will address the issue of robustness of our findings for the different acquisition types.