In this study we focus on the types of subject that Italian-speaking children produced in the context of an experiment eliciting Clitic Left Dislocations (henceforth ClLD), with the left dislocated phrase corresponding to the direct object. In particular, we are interested in analyzing the nature of the subject in relation to the presence of the left peripheral object topic. The resulting sentences correspond to a DP₁ DP₂ CL V structure, as illustrated in (1a, b), with both DPs lexically restricted:1

The data we will illustrate and discuss come from Belletti & Manetti’s (2019) work which precisely aimed at eliciting ClLDs of the type illustrated in (1), with pre-school children. In fact, the elicited structures could either contain an overtly expressed left peripheral direct object topic or no overtly expressed direct object topic, depending on discourse conditions. In the quoted work, the main results of the experiments showed that pre-school children, aged from 4- to 6-year-old, were able to produce ClLDs with an overtly expressed direct object topic in the appropriate discourse conditions, thus indicating that left-peripheral topic positions are available and accessible to young children when the appropriate context is provided.

In the present work, we provide a different perspective on Belletti & Manetti’s (2019) production data in order to zoom in on the nature of the subject used by children in the discourse conditions created by the experiments mentioned. Specifically, we analyze the different types of subject (lexical subject, null subject or overt pronominal subject) within sentences with no overtly expressed direct object topic (Clitic Pronoun structures) and with an overt left peripheral direct object topic (ClLD structures).

The reason why we focus our investigation on these productions is related to the role that different types of subject play in the computation of ClLDs when the overtly expressed object DP is lexical and left dislocated. Specifically, the ClLDs under investigation may instantiate an intervention configuration when the left dislocated object is overtly expressed and the subject is preverbal and lexical, intervening between the left-peripheral object topic and its clause internal merge position. For this reason, we will analyze the subject of ClLD structures in light of the grammatical principle featural Relativized Minimality (fRM), following Friedmann et al.’s (2009) account and its further elaboration (Belletti et al. 2012), in line with Belletti & Manetti (2019).

Under the Relativized Minimality principle (Rizzi 1990; 2004), in a ClLD, the left dislocated object topic encounters the intervening lexical subject (Z) in the establishment of the dependency between its Target (X) position in the left-peripheral topic position and its Merge object position (Y) within the clause, as shown in (2):

In Italian, which is a multiple topic language (Rizzi 1997), the DP₁ DP₂ Cl V structure can correspond to two different word orders, either DP_SBJ DP_OBJ Cl V (3a) with also DP_SBJ occupying a left-peripheral topic position, or DP_OBJ DP_SBJ Cl V (3b), with DP_SBJ either in the left periphery or in the clause internal subject position (as given in 2 for illustration): both orders give rise to the locality problem:

Following from this, the role of the subject is crucial in this particular structure since it could yield different degrees of intervention. If the subject is lexical and preverbal and matches in number with the object, it leads to the intervention situation of (feature) inclusion, which is the most challenging for children as previously shown for other A-bar dependencies (e.g. Object relative clauses, Friedmann et al. 2009). In case of number mismatch between the subject and the object the structure corresponds to a different feature configuration, intersection, which is well mastered by children at age 5 (Belletti at al. 2012; Manetti et al. 2016 on ClLD) and in the acquisition of other structures (e.g., object relative clauses, see Adani et al. 2010). In case the subject is pronominal (in fact null in Italian when not focal or otherwise prominent), this yields the most accessible structure from the perspective of intervention, since no lexical DP intervenes between the clause internal merge position of the object and its left dislocated position, giving rise to feature disjunction, hence no intervention.

To summarize, in the present study we aim at providing a baseline from the behavior of monolingual Italian-speaking young children in their mastering of the proper use of overt lexical subjects, null subjects and, possibly, overt pronominal subjects, in the specific discourse conditions investigated leading to the production of ClLD, with a left dislocated direct object. The type of subject will be examined in relation to the intervention configurations arising in this specific clitic left dislocated discourse context, which may be particularly challenging for pre-school children.

The data we will illustrate and analyze come from the study carried out by Belletti & Manetti (2019), which tested two groups of typically-developing pre-school children, aged from 4 to 6 years old. The production experiments were designed with the aim of eliciting structures with left dislocated object topics.2 We will first summarize the method and materials adopted in the two experiments, in order to show the discourse contexts relevant for analyzing the different types of subject (lexical, pronominal null, pre- or post-verbal; see the quoted work for all relevant details), the central aim of the present work.

We will then focus on the analysis of the subjects actually produced by children in the context of Clitic Pronoun and clitic left-dislocated structures, ClLD. The research questions we address are the following:

In this study we analyzed the data from Belletti & Manetti (2019) in the aim of highlighting what type of subject children produced when answering to patient-oriented questions. We focused our analyses on ClLDs and Clitic Pronoun structures and examined whether the subject of children’s answers was appropriate to the discourse conditions set up through the prompting questions. As for our first research question (Q1: Did children produce the appropriate type of subject (expected to be lexical and preverbal) in their answers?), different results emerged in the two experiments.

In Experiment 1, children showed to master the use of the appropriate expected type of subject: they preferred lexical subjects which were correctly informative as to who performed the actions on the patients in Clitic Pronoun and ClLD structures and privileged the preverbal position of the subject since the referents were already given in the discourse context.9 This shows that at age 4, children already master appropriate production of (lexical vs. null referential) subjects in general confirming previous findings (see Belletti & Guasti 2015: Chapter 7 for a review; Manetti 2017; De Cat 2009 for French). In addition, the absence of pronominal overt subjects (e.g. lui/he; lei/she), which would be appropriate in different contexts (e.g. with focal interpretation) given their status of strong pronouns (see Cardinaletti & Starke 1999) highlights that at this age monolingual Italian-speaking children properly master the distinction and production of null pronominal vs. overt pronominal subjects in the elicited production contexts, in line with previous research from spontaneous production studies (e.g. Serratrice 2005). This result could constitute a baseline measure for testing other populations, such as L2 or multilingual speakers, who instead are known to find difficulty in the acquisition of the use of subject pronouns in relation to different pragmatic discourse properties (see Serratrice 2005, 2007; Sorace & Filiaci 2006; Belletti et al. 2007; Belletti & Guasti 2015: Chapter 7, for a summary on the acquisition of subjects across different populations).

Differently from Experiment 1, in Experiment 2 which should have elicited the same type of subject (namely, lexical and preverbal), children diverged from the expected pattern. Our second research question (Q2: In case children produced other types of subjects (e.g. pronominal null, pronominal overt), what did they opt for?) precisely aimed at analyzing children’s alternative choice which mainly consisted of null subjects, which were overwhelmingly in the plural form, as illustrated in (12).

The plural null option was only seldom used in Experiment 1 (footnote 8), where the referent of the subject was always a singular character in the stimuli (cf. Figure 1 and 2) and the subject was mostly overt in children’s ClLDs answers, as discussed. Notice that the sentences containing null subjects were not specific but rather remain vague about the two agent referents. This tendency to overuse subject pronouns instead of lexical DPs also emerged in previous studies investigating children’s acquisition of subjects (e.g. Karmiloff-Smith 1981; Koster, Hoeks & Hendriks 2011; among others). Notice, however, that in the two experiments the information about the agent referents could be easily recovered from the context since both the experimenter and the child could see the images on the screen. Thus this experimental factor could have favored the overuse of subject null pronouns instead of referential lexical subjects, in that children are also known to be more likely to use ambiguous subject pronouns in place of referential lexical subjects when the information is recoverable in a shared context (e.g. Serratrice 2008; De Cat 2009; among others).

Through our third research question (Q3: Did the type of subject vary depending on the presence of a lexically expressed left dislocated object (in ClLDs)?), we further analyzed the production of null subjects by investigating the possible relationship between the subject and the presence of the overtly expressed left dislocated object.

Overall, in both experiments, children relied on null subjects more often when the object topic was overtly realized in the left periphery: this suggests that in both experiments ClLDs with preverbal lexical subjects were challenging and hence avoided in different ways. This difficulty is expected under the featural Relativized Minimality principle (Friedmann et al. 2009; Belletti et al. 2012; Belletti 2017), which analyzes DP_OBJ DP_SUBJ Cl V structures as creating the hardest intervention configuration for children when both DPs are lexical, since the (preverbal) lexical subject intervenes between the left-peripheral object topic and its clausal internal merge position. Hence having a null subject could constitute a way to avoid intervention, as it eliminates the lexically restricted intervener. This is illustrated in (13a) and (13b):

In (13a) there is inclusion of [+NP] feature corresponding to the lexical restriction, which is the hardest structure for young children. In contrast, in (13b) there is no lexical restriction feature on the pronominal null subject, thus creating a feature disjunction relation that can be computed by young children (see Belletti & Manetti 2019 and references quoted there for all relevant details).

As mentioned earlier, in Experiment 1 the null subject was very seldom produced by children, who instead adopted other strategies to modulate the intervention configuration in ClLDs, for example through the a-marking of left dislocated object topics (see Belletti 2018a, b for a detailed discussion on a-Topics). It is worth highlighting that when children opted for null subjects they used both the singular and plural forms (cf. footnote 8), and singular null subjects were very rare in their production. This indicates that, in Experiment 1, children were consistent in conveying the information about the subject referents and tried to handle the grammatical pressure of the intervention configuration instantiated by the ClLDs (see 13a).

In Experiment 2, in contrast, children overwhelmingly relied on the production of plural null subjects both in Clitic Pronoun and ClLD structures, but significantly more often in ClLDs: we interpret this result as a possible way to avoid the intervention configuration, resulting in feature disjunction configuration illustrated in (13b). Building on the evidence from Experiment 1, in which children produced few (singular) null subjects and more consistently used referential lexical subjects, we suggest that the plural null subject found in Experiment 2 was not referential but rather generic. Possibly the plural nature of the null subject, rarely present in Experiment 1 where the agent referent was always singular, could be triggered by the fact that the images in Experiment 2 featured a plural subject, thus constituting a kind of priming for the use of the (generic) plural null subject. Notice that this type of answer containing a plural null subject, as in (13b), is an appropriate answering strategy to the patient-oriented question, as it conveys the information about the topic patient which the question is about and just remains vague about the agent referents. That the plural null subject is likely to be interpreted as a generic (plural) subject is supported by the fact that this interpretation is a possible option in several languages, including standard Italian, as shown in (14) (example 26 from Belletti & Manetti 2019 and related discussion), where the null pronominal subject can correspond to either a singular or a plural referent:

We further suggest that children’s reliance on the pro_PL in ClLDs, with the generic interpretation of the subject, may be a suitable alternative to the passive, which was the much preferred answer by adults. Indeed, the structures in (15a), typical children’s production, and in (15b), typical adults’ production, can be considered closely related to each other: They are judged as appropriate in the same discourse context, and crosslinguistic evidence indicates that languages which do not have (a productive use of) passive may resort to active sentences with the generic plural subject instead (Belletti & Manetti 2019: 167 and footnote 25). The fact that children largely preferred (15a) over (15b) indicates that the computation for passive is still not adequately mastered at this age.

In conclusion, our study presented an overview on the production of subjects in relation to the structural and discourse related properties of Clitic Pronoun and ClLD sentences. Importantly, we underline that children never used an overt pronominal subject in compliance with the discourse conditions provided by our experiments: this shows that they were aware of the distribution of overt vs. null pronominal subjects. This result is consistent with the another important finding of our work that indicated appropriate mastery in the production of overt lexical subjects vs. null pronominal subjects. We conclude by pointing out that children properly produced null and lexical subjects depending both on the discourse conditions and on their grammatical competence. In particular, children adopted appropriate discourse related options available in Italian such as the use of the plural null subject with the generic interpretation. This helped them cope with the grammatical pressure otherwise arising in ClLDs in which both the left dislocated object topic and the preverbal subject were lexically restricted.