3.1 Null versus overt pronominal subjects and preverbal versus postverbal subjects

Although Spanish has the grammatical option of omitting lexical and pronominal subjects, their omission is not entirely optional but constrained by discourse–pragmatic considerations.Footnote ⁶ In general terms, the overt use of a personal pronoun in a position where it could normally be omitted results from emphatic or contrastive reasons, whereas its omission indicates neutrality or lack of emphasis, as the examples below illustrate:

1. (6) ¡Él hizo trampas!

   “He cheated!” (and not me or anyone else – emphatic use)

2. (7) ¡Ø Hizo trampas! (neutral use)

Overt subject pronouns are hence used for emphasis and the disambiguation of the referent and are interpreted as focused elements which give new and contrastive information whereas null-subject pronouns represent continuity in discourse topic and hence cannot be used when the information of the subject is new in the discourse:

1. (8)

   1. a. ¿Quién viene?

      “Who comes?”

   2. b. Ella viene.

      “She comes”

   3. c. *Viene.

      “(She) comes”

1. (9)

   1. a. ¿Quién cree Juan_i que __ ganará el premio?

      “Who does Juan believe will win the prize?”

   2. b. Juan_i cree que él_i/*pro_i ganará el premio.

      “Juan believes that he will win the prize”

   3. c. ¿Qué cree Juan_i que obtendrá __ en el concurso?

      “What does Juan believe he will get in the contest?”

   4. d. Juan_i cree que *él_i/pro_i ganará el premio.

      “Juan believes that he will win the prize”

      (Pérez-Leroux and Glass, Reference Pérez-Leroux and Glass1999, p. 226)

In (8), an overt pronoun is required as the subject encodes new information. In (9), the questions set the discourse interpretation. Thus (9a) focuses the discourse on the subject of the embedded clause and hence a null pronoun is disallowed, whereas in (9c) the question is about the object of the embedded clause, and therefore the subject represents old information in the discourse (i.e., the topic) and may remain null.

Given that pronoun omission indicates lack of emphasis, it has been claimed that the expression vs. omission of pronouns in Spanish has the same function and anaphoric use as strong (emphatic use) vs. weak pronouns (neutral use) in languages such as English, which do not permit pronoun omission and in which the use of overt pronouns is non-emphatic by default (Luján, Reference Luján, Bosque and Demonte1999). Thus English (10a) would correspond to Spanish (10b) in terms of pronoun use:

1. (10)

   1. a. He/He (and not me or anyone else) cheated!

   2. b. ¡Ø/Él (and not me or anyone else) hizo trampas!

In the case of a series of sentences or complex sentences with a subordinate clause, the use of pronouns and their co-reference is also constrained by discourse–pragmatic considerations. According to Montrul (Reference Montrul2004a), if we have two sentences in a discourse with the same reference, an overt pronoun in the second sentence is pragmatically odd and an overt lexical subject is ungrammatical, as in (11). Yet pronoun omission is odd when there is a switch in reference, as in (12):

1. (11) Pepe no vino hoy a trabajar. *Pepe/?Él/Ø estará enfermo.

   Pepe no came today to work. Pepe/He/Ø will be sick.

   “Pepe did not come to work today. He must be sick.”

2. (12) Hoy no fui a trabajar. Pepe/Él/*Ø pensó que estaba enferma.

   Today no I went to work. Pepe/He/Ø thought that I was sick.

   “Today I did not go to work. Pepe/He though I was sick.” (2004a, p. 176)

Within sentences with a main and a subordinate clause, the presence vs. the omission of a subject pronoun also gives rise to significant differences in co-reference and interpretation. The following sentences (Luján, Reference Luján, Bosque and Demonte1999) illustrate this:

1. (13)

   1. a. Cuando él/Ø trabaja, Juan no bebe. (él ≠ Juan; Ø = Juan)

   2. a′. “When he/he works, Juan doesn't drink.” (he ≠ Juan; he = Juan)

   3. b. Cuando Juan trabaja, él/Ø no bebe. (él ≠ Juan; Ø = Juan)

   4. b′. “When Juan works, he/he doesn't drink.” (he ≠ Juan; he = Juan)

   5. c. Juan no bebe cuando él/Ø trabaja. (él = Juan; Ø = Juan)

   6. c′. “Juan doesn't drink when he/he works.” (he = Juan; he = Juan)

   7. d. Él/Ø no bebe cuando Juan trabaja. (él/Ø ≠ Juan)

   8. d′. “He/he doesn't drink when Juan works.” (He/he ≠ Juan)

As the examples indicate, the differences in co-reference between the referential element Juan and the pronominal element él or the null element in Spanish correspond to the differences in co-reference between the referential expression Juan and the strong emphatic pronoun he or the weak neutral pronoun he in English. In (13a) and (13b), the overt pronoun él gives rise to “disjunctive reference” (Luján, Reference Luján, Bosque and Demonte1999, p. 1279) with respect to the antecedent Juan, whereas the null pronoun can co-refer with it. That is to say, él necessarily designates someone different from Juan, whereas the null pronoun can refer to it. In (13c), the subordinate clause, where the overt and null pronoun are, is postposed and hence part of the predicate of the main clause. According to Luján (Reference Luján, Bosque and Demonte1999), in that case both the overt pronoun and the null element can co-refer with the lexical referential element Juan. The overt pronoun is not necessarily “disjunctive” here. In (13d), both the overt and the null pronouns give rise to “disjunctive reference” in that neither él nor the null element can co-refer with Juan. A referential expression (i.e., Juan) cannot be given an antecedent.Footnote ⁷ The above-mentioned “disjunctive” effects of the overt pronoun may disappear and the pronoun can indeed co-refer with its antecedent (preceding or following the pronoun) if the discourse provides a context where the emphatic and contrastive use of the pronoun is required.

Regarding the apparently free word order in Spanish, the possibility for subjects to appear pre -and postverbally is strongly determined by lexical verb class and discourse structure factors, which follow from the minimalist syntactic accounts of Spanish preverbal subjects by which subjects are moved from spec-vP to a preverbal non-argumental position higher than TP and within the CP-domain for pragmatic reasons (Platzack, Reference Platzack2004; Rizzi, Reference Rizzi and Belletti2004, Reference Rizzi2005; Alexiadou and Anagnostopoulou, Reference Alexiadou and Anagnostopoulou1998; Ordóñez and Treviño, Reference Ordóñez and Treviño1999). In sentences with transitive verbs, the subject can appear either preverbally (SVO), postverbally (VSO) or after the object (VOS), although in neutral focus structures, native Spanish speakers tend to prefer SVO orders. Both VSO and VOS orders are possible in contrastive focus (i.e., emphatic) structures and information/presentational focus (i.e., new versus old information) although native Spanish speakers tend to prefer VOS orders and consider VSO orders unnatural and strongly marked.

As for intransitive verbs, they are classified into unergatives and unaccusatives according to the Unaccusative Hypothesis or the Split-Intransitivity Hypothesis (Burzio, Reference Burzio1986; Levin and Rappaport-Hovav, Reference Levin and Rappaport-Hovav1995) and on the basis of where the subject is base-generated. The subject of unergative verbs like gritar “to shout” is generated preverbally in spec-VP, whereas the subject of unaccusative verbs like llegar “to arrive” is base-generated postverbally in object position. The discourse neutral word order for unergative verbs is thus SV, whereas for unaccusative verbs it is VS. Neutral contexts are elicited by the “out of the blue” question “What happened?”, which requires an unfocused answer, as the whole sentence is new information. Spanish inappropriate word order, though not ungrammatical, yields pragmatic anomaly, whereas inappropriate word order in English results in ungrammaticality:

1. (14)

2. (15)

While word order is constrained in Spanish by the Unaccusative Hypothesis in unfocused contexts, it is constrained by discourse structure and more specifically by information focus at the syntax–discourse interface. Informationally focused elements (i.e., new information) are placed in sentence-final position in Spanish (Zubizarreta, Reference Zubizarreta1998) and hence syntactic word order is affected. When the subject is informationally focused, as in the answer to the question “Who called while I was gone?”, both unergative and unaccusative verbs trigger VS word orders. In contrast, informational focus is realized only phonologically, but not syntactically, in English, where the focused element is stressed:

1. (16)

2. (17)

The syntactic account of information focus to be followed in this study is based on Belletti's (Reference Belletti, Hulk and Pollock2000, Reference Belletti and Rizzi2003) TP-internal presentational focus hypothesis, by which the functional projection Focus Phrase (FocP) is internally merged in TP and above VP. The informationally focused subject is displaced to Spec-FocP, that is to say, in a postverbal position in Spanish, whose verbs typically raise to TP and whose subjects do not raise obligatorily to Spec-TP, as has been seen above. The focus head contains a [+strong] focus feature in Spanish, which triggers movement of the presentationally focused element to Spec-FocP. At the same time, the uninterpretable focus feature in the focus head attracts the interpretable focus feature in the focused subject for checking purposes. In contrast, the focus feature is [–strong] in English and hence requires the focused subject to remain in situ. Recall that English subjects obligatorily raise to Spec-TP from Spec-vP in order to check the EPP feature on T, and that the verb does not raise to TP. The result is that even if the subject is focused, word order remains SV and focus is marked by prosody. Yet interpretative facts also hold in English. The focused subject in Spec-TP carries the interpretable focus feature, which is attracted by and agrees with the uninterpretable focus feature under T.

3.2 L2A research on the discourse distribution of null and overt subjects

Although most L2A studies on the syntactic and discourse properties of L2 Spanish have been conducted with adult learners, and hence direct comparisons with the child L2A data in the present study cannot be made, some of the most relevant and recent research is briefly reviewed here, as it contributes to our understanding of the data presented.

As for the distribution of null and overt subject pronouns, Pérez-Leroux and Glass (Reference Pérez-Leroux, Glass, Pérez-Leroux and Glass1997, Reference Pérez-Leroux and Glass1999) conducted experimental research on elementary, intermediate and advanced adult English speakers learning L2 Spanish and focused on topic/focus environments and Montalbetti's (Reference Montalbetti1984) Overt Pronoun Constraint (OPC) environments.Footnote ⁸ The Topic/Focus study focused on adult English-speaking learners of Spanish in a classroom setting, who were required to read eight short stories illustrated by a picture to fix the interpretation of the embedded subject as co-referent with the subject in the main clause, and followed by a subject/object comprehension question which they had to answer. Subject questions required an answer with a focused and hence overt subject in the embedded clause, whereas an object question elicited an answer in which the focus was not placed on the embedded subject and hence this was a topic and should be null.Footnote ⁹

Results indicate that L2 learners increased their use of null pronouns with proficiency in the topic condition and kept them relatively low in the focus condition, although the elementary and intermediate groups produced very similar percentages of overt and null pronouns in this condition. However, the fact that the task involved answering an open question contributed to the large number of “other” responses, which were not relevant for the study. Results were not significantly different between groups and according to story type but, according to the authors, their findings still indicate that all groups discriminate in their use of null-pronoun subjects in topic and focus environments, as the percentages of null pronouns are higher when the pronoun is interpreted as topic than when the pronoun is interpreted as the focus of the sentence. Pérez-Leroux and Glass (Reference Pérez-Leroux, Glass, Pérez-Leroux and Glass1997) conclude that pragmatic/discourse constraints are acquired later, over time and experience, than syntactic constraints, but that results indeed show that L2 learners of Spanish can learn the syntactic and pragmatic distribution properties of null pronouns and the fact that overt and null pronouns are not in free variation.

Also related to null/overt pronominal subject distribution, Sorace and Filiaci (Reference Sorace and Filiaci2006) provided a processing account for the interpretation of intrasentential anaphora by English adult near-native speakers of L2 Italian and native Italian speakers. The authors claim that the syntax–pragmatics interface vulnerability results “not so much from representational deficits, but from inadequate processing resources or ‘shallow’ parsing strategies” (p. 341), along the lines of Clahsen and Felser (Reference Clahsen and Felser2006) and Carminati (Reference Carminati2002, Reference Carminati2005). Participants were presented with complex sentences in which the subordinate clause included an overt or null pronominal subject and either preceded (backwards anaphora) or followed (forwards anaphora) the main clause, and carried out a picture verification task interpreting the possible antecedents of the null/overt pronominal subject.

The authors predicted that the interpretation of anaphoric overt pronouns would be more problematic than that of null pronouns (Belletti, Bennati and Sorace, Reference Belletti, Bennati and Sorace2005; Filiaci, Reference Filiaci2003), allowing overt pronouns in subordinate clauses to co-refer with the subject of the main clause, and that backwards anaphora would create more processing difficulties than forwards anaphora for the near-native group. Results supported these predictions and the authors conclude that although near-native speakers had acquired the syntactic constraints of a null-subject grammar, they did not have the necessary processing strategies to coordinate syntactic and discourse knowledge when linking pronouns to their antecedents, as the optionality found in the interpretation of overt pronouns indicates.

Also focusing on adult L1 English near-native speakers of Italian, Belletti et al. (Reference Belletti, Bennati and Sorace2007) conducted several experiments on the controlled and spontaneous production and interpretation of null and overt pronominal subjects and postverbal subjects across verb types and established comparisons with a group of control native speakers of Italian in order to examine the role of syntactic and discourse factors in near-native subject use. Participants were tested on spontaneous production of null and overt subjects with a story-telling task, on their controlled production of focus new information postverbal subjects with videos and comprehension questions, on their interpretation of null and overt pronominal subjects in anaphora contexts with a picture verification task and on their production of preverbal and postverbal subjects with a headline creation task.

Results indicate that near-native speakers of Italian mastered the crucial properties of the null-subject parameter in their L2, although they showed residual optionality of some grammatical and discourse subject features of their L1 English which are not native-like but compatible with the L2 grammatical system. More specifically, near-native speakers displayed a native-like use and interpretation of null pronominal subjects but an overuse of overt pronouns in inappropriate discourse contexts and a non-native-like interpretation of overt pronominal subjects in anaphora contexts. Likewise, postverbal subjects were underused across verb classes (except with unaccusatives) and new information focused preverbal subjects were overproduced. The authors concluded that the differences found between native and near-native grammars did not only involve the presence of formal grammatical principles but “rather the interplay of formal principles and grammatical options in domains at the interface with discourse” (Belletti et al., Reference Belletti, Bennati and Sorace2007, p. 682). The authors also point out that the dissociation between the accurate use of null subjects and the non-native use of postverbal subjects suggests that the traditional formulation of the null-subject parameter in terms of pro-licensing (Rizzi, Reference Rizzi1982) may not be sufficient to account for both properties in L2 grammars and that discourse factors need to be seriously considered and incorporated in the theory.Footnote ¹⁰

Hertel (Reference Hertel2003) and Lozano (Reference Lozano2006)Footnote ¹¹ tested native English adult L2 learners of Spanish on their acquisition of subject inversion according to both syntactic and discourse structure properties. Both studies predicted that L2 learners would present more difficulties acquiring word order with informationally focused structures than in neutral contexts, where the distribution is determined by the unaccusative–unergative distinction.

Hertel's (Reference Hertel2003) study included beginner, low-intermediate, high-intermediate and advanced English-speaking adult learners of Spanish, who completed a contextualized written production task in which they were presented with short stories and had to answer a question which targeted unaccusative and unergative verbs with different information structures. The questions focused on the entire sentence or on the subject, thus triggering a discourse-neutral interpretation or an informationally focused subject, respectively.

Results on the answers to the global questions show that inversion in unaccusatives was mainly produced by the advanced learners and the control group, as the lower-proficiency learners produced little or no inversion, possibly transferring their L1 word order. As for global questions with unergative verbs, percentages of VS orders were very low, though advanced learners produced significantly more inverted sentences than the other learner groups and the native speakers. According to the author, advanced learners seem to have generalized inversion to the unergative verbs in this type of questions. All groups except for beginners produced more VS sentences with unaccusatives than with unergatives, though the difference in verb type was only significant in the case of advanced learners and the native group. Results on focused subject questions revealed that beginner and low-intermediate learners transferred their L1 SV order again, showing that they were not sensitive to Spanish discourse structure and producing practically no VS responses with either type of verb. High-intermediate and advanced learners increasingly produced more VS orders with unaccusative and unergative verbs.

Hertel concluded that the word order effects of verb class and discourse structure are acquired late in L2 Spanish, namely at the advanced level, and preceded by L1 Transfer. Thus the prediction that discourse constraints on word order would be acquired after verb class lexical constraints on word order was not met in this study, as VS percentages rise earlier for focused subject questions in the two verb types than for global questions in the unaccusative verb type. Although, according to the author, advanced learners show native-like sensitivity to verb class and discourse structure, the lexical and focus-related inversion remains optional.Footnote ¹²

Lozano (Reference Lozano2006) carried out a very similar study to test the distribution of SV and VS word order in adult advanced L2 Spanish, which included two experimental groups of Greek and English adult advanced learners of Spanish and native Spanish speakers. Learners completed a contextualized acceptability judgment test with paired target sentences. Each contextualized stimulus and question represented an unfocused or informationally focused environment with unergative and unnacusative verbs, thus favouring an SV or VS word order. Learners were provided with both orders and were required to rate their acceptability on a five-point rating scale.

Results on unfocused contexts with unergative verbs revealed that both Greek and English learners of Spanish significantly prefer SV to VS, as native speakers do and as predicted by the author's hypothesis. The acceptance of SV did not differ between groups, though the acceptance of the pragmatically odd VS order did differ between groups, with the English and Greek groups accepting a higher percentage of VS orders than the native control group. As for the unfocused contexts with unaccusative verbs, results showed that both groups of learners significantly prefer VS to SV, as the native control group did, and supporting the author's predictions. In this case, the acceptance of both VS and SV did not differ between groups, showing native-like knowledge.

Regarding informationally focused contexts, where VS is the preferred order with both unergative and unaccusative verbs, results revealed that whereas the native control group significantly preferred VS to SV order with both types of verbs, L2 learners simultaneously accepted both SV and VS orders, resulting in non-native optionality. Between-group comparisons show that although the learners’ acceptance of the preferred VS order did not differ significantly from that of the native control group, their acceptance of the pragmatically odd SV order was significantly higher that that of the control group. Overall, results clearly indicate that properties at the syntax–discourse interface were persistently more problematic for advanced speakers than syntactic properties, as was observed in previous studies.

I would like to point out that the L2 studies reviewed in this section cannot be directly compared to the child L2 data obtained and discussed in the present study, since these studies deal with L2 adult informants, whose input and type of exposure are radically different from the ones children in the present study received and since the data elicitation techniques also differ in each study. However, and as cross-sectional studies on the child L2A of subjects are practically non-existent, the present review of the literature contributes to the relevance of the data presented here.

4.1 Aims, research questions and hypothesis

This study examines the acquisition of both syntactic and stylistic subject properties of L2 Spanish in five-, ten- and seventeen-year-old British children, who were all first exposed to Spanish at the age of four in the context of an immersion Spanish school. More specifically, and in the light of previous L2 studies on the syntactic and discourse subject properties focusing only on adult learners, it aims to explore whether and to what extent British child L2 learners of Spanish can indeed achieve native-like results in their judgment and preference of overt vs. null pronominal subjects and preverbal vs. postverbal subjects in a cross-sectional experimental task.

As all the children tested in the study started their L2 acquisition of Spanish at the age of four in an immersion context, and Spanish was not their home language, we can safely assume that these children are true child L2 learners, whose acquisition will be compared to L1 Spanish children. The study of child L2A has lately become a focus of research, particularly regarding its similarities and differences in different grammatical areas with respect to child L1A and adult L2A (Schwartz, Reference Schwartz, van Kampen and Baauw2004; Unsworth, Reference Unsworth2005b; Blom, Reference Blom, Haznedar and Gavruseva2008; Meisel, Reference Meisel2007, Reference Meisel, Haznedar and Gavruseva2008). Unsworth (Reference Unsworth2005b) studied and compared the acquisition of Dutch direct object scrambling over negation by L1 English child and adult L2 learners and also compared it to L1A. In general terms, and with respect to the production of this particular property, learners were found to go through the same developmental stages in child and adult L2A which, on the assumption that child L2A is guided by UG, was taken to imply that adult L2A is also guided by UG. Both child and adult learners’ production was affected by L1 transfer, and this was the only difference found with respect to the developmental stages in L1A. Schwartz (Reference Schwartz, van Kampen and Baauw2004) compared L1A and L2A in terms of ultimate attainment and course of development and with regard to syntax and inflectional morphology. As for ultimate attainment, she found differences between L1A and adult L2A and, regarding developmental course, she claims that child L2A is like adult L2A and that both differ from L1A with respect to syntax but that child L2A is like L1A and different from adult L2A in terms of inflectional morphology. In Meisel (Reference Meisel2007, Reference Meisel, Haznedar and Gavruseva2008), results suggested that child L2A is like adult L2A and different from bilingual L1A in the morphological domain and possibly also in the syntactic domain. Further research needs to be done and, as the author also claims, child and adult L2A clearly differ in many other aspects, such as rate of acquisition.

Yet what is clear is that the L2 child shares representational and developmental properties of both L1A and adult L2A and hence is a valuable tool in the general picture of language acquisition (Lakshmanan, Reference Lakshmanan1995; Schwartz, Reference Schwartz1992, Reference Schwartz, Beachley, Brown and Conlin2003, Reference Schwartz, van Kampen and Baauw2004). The child L2 learner is cognitively more mature than the child L1 acquirer but less mature than the adult L2 learner; she is affected by the existence of the L1 grammar and she is remarkably successful in ultimate attainment studies. It is also important to distinguish child L2A from bilingual L1A. The former is defined as successive child bilingualism, whereas the latter is defined as simultaneous child bilingualism, which refers to the acquisition of two languages from birth. Child L2A implies acquiring an L1 and starting L2 exposure at least when most grammatical principles of the native language are assumed to have been acquired, around the age of four (Guasti, Reference Guasti2002). Yet, and on the basis of maturational effects on successful ultimate attainment, another age of onset needs to be established as the limit after which native-like ultimate attainment is no longer possible, namely around the age of seven (Schwartz, Reference Schwartz, van Kampen and Baauw2004; Unsworth, Reference Unsworth, Brugos, Micciulla and Smith2005a, Reference Unsworthb; DeKeyser, Reference DeKeyser2000). Non-native acquirers who are first exposed to an L2 between the ages of four and seven are thus child L2 acquirers, like the ones tested in the present research. Following Schwartz (Reference Schwartz, van Kampen and Baauw2004), reaching native-like levels of ultimate attainment implies that child L2A is indeed constrained and guided by UG, which does not necessarily mean that child L2A and L1A are alike. Rather, UG clearly plays a role in child L2A but L1 influence is also argued to be present. Therefore, I follow the assumption that child L2 development is characterized by Full Access to UG, initial presence of Full L1 Transfer and possible but not inevitable successful ultimate attainment (Schwartz and Sprouse, Reference Schwartz and Sprouse1996; Schwartz, Reference Schwartz1998, Reference Schwartz, Beachley, Brown and Conlin2003, Reference Schwartz, van Kampen and Baauw2004). It remains to be seen if and how L2 discourse–pragmatic properties are acquired.

In the context of this study, we seek to explore whether children are able to acquire the L2 feature values responsible for the target syntactic and discourse properties of Spanish subjects and how this acquisition compares to that of the control children. More specifically, we want to test whether L1 English children acquire the strong pronominal [+interpretable] Agreement features on the verb, by which Spanish checks the EPP by raising the verb to T, and therefore empty subjects and postverbal subjects are allowed (as outlined in Platzack, Reference Platzack2004). Once they have acquired empty and postverbal subjects, we also want to test whether these children acquire the syntactic and discourse factors which apply in L2 Spanish subject use, namely topic vs. focus contexts and co-reference in null/overt subjects in backwards anaphora contexts, and lexical verb class and the presence of a [+strong] uninterpretable functional focus feature (Belletti, Reference Belletti, Hulk and Pollock2000, Reference Belletti and Rizzi2003) in preverbal/postverbal subject contexts.

We hypothesize that child L2 learners of Spanish will develop their judgment and preference of both the syntactic and discursive properties of subjects towards native-like results. However, L1 transfer is predicted to be present and discourse properties are predicted to be more complex to acquire and hence delayed, if child and adult L2A pattern alike in this respect and given that most studies in the field indicate that properties at the syntax–discourse interface are vulnerable and residual optionality is observed at least in adult advanced learners and even near-native speakers (Lozano, Reference Lozano2002, Reference Lozano2006; Hertel, Reference Hertel2003; Montrul, Reference Montrul2004b; Montrul and Rodríguez Louro, Reference Montrul, Rodríguez Louro, Torrens and Escobar2006; Belletti et al., Reference Belletti, Bennati and Sorace2007; Sorace and Filiaci, Reference Sorace and Filiaci2006). This delay in acquisition is also to be expected as discourse contrasts are expressed prosodically but not syntactically in the children's L1, English, and non-native-like word order or empty subject distribution do not result in ungrammaticality or lack of comprehension and hence input data may be confusing.

5.1 A comparison of the three age groups: non-native five-, ten- and seventeen-year-olds

With the aim of comparing the percentages of correct judgments and preference judgments of the three age groups of non-native child L2 learners of Spanish, and in order to detect significant differences among the three groups, a non-parametric one-way analysis of variance Kruskal–Wallis H test was first applied. As for the correct judgments, the Kruskal–Wallis H test showed that at least one of the age groups has a significantly different median in each linguistic variable (H = 7.214, p = .027; H = 13.255, p = .001; H = 9.848, p = .007; and H = 9 .613, p = .008), except “Preverbal/Postverbal Subjects with Transitive Verbs” and “Preverbal/Postverbal Subjects with Unergative Verbs”, which display similar results along the three age groups (H = 3.089, p = .213; H = 1.533, p = .465). Regarding the preference judgments, the Kruskal–Wallis H test also showed that at least one of the age groups has a significantly different median in each linguistic variable (H = 12.649, p = .002; H = 15.788, p < .001; H = 7.349, p = .025; H = 6.120, p = .047; H = 6.247, p = .044; and H = 7.784, p = .020) except “Preference for Preverbal Subjects with Transitive Verbs”, which displays identical constant results along the three age groups (H = .000, p = 1.000).

In order to find out where in the groups the significant difference emerges and to analyze the development of correct judgments and of preference judgments through the three age groups, the Mann–Whitney U-test was applied to each variable and between the five- and the ten-year-old groups and the ten- and seventeen-year-old groups. Table 4 shows the results obtained from the acceptance/rejection variables. Shadowed results indicate where the difference between the groups was significant.

Table 4.

Mann–Whitney U-test: acceptance/rejection variables.

Note *: c stands for “constant”, that is to say, all values are identical.

As can be seen in Table 4, there is a general increase-by-age tendency in all variables, although significant differences between the age groups are different in each variable. As for “Null/Overt Subjects in Main Clauses” (see Figure 1), a significant difference is observed between the ten- and the seventeen-year-old groups (U = 18.000, p = .005), whereas the first two age groups show a rather similar percentage of correct judgements. “Null/Overt Subjects in Subordinate Clauses” (see Figure 2) display much lower results in the three age groups in comparison to the other variables, showing that child learners of L2 Spanish seem to have more problems accepting/rejecting subjects in subordinate clauses than in main clauses. A statistically significant difference is also observed between the ten- and the seventeen-year-old groups (U = 11.500, p = .002) and overall development is discrete, its median values ranging from 33.33% in the five-year-old group to 66.66% in the seventeen-year-old group. “Null/Overt Expletive Subjects” (see Figure 3) present high percentages of correct judgments, which also increase by age. A significant difference is also found between the ten- and the seventeen-year-old groups (U = 20.000, p = .011). Two significant differences emerge in the results of correct judgments of “Preverbal and Postverbal Subjects with Unaccusative Verbs” (see Figure 4), namely between the five- and ten-year-old groups and the ten- and seventeen-year-old groups (U = 23.500, p = .047 and U = 20.000, p = .007), although an increase-by-age tendency is not observed here. No significant differences are observed in the “Preverbal and Postverbal Subjects with Transitive Verbs” variable (see Figure 5), as the three groups have a median of 50.00%, although their mean values present a slight increasing tendency from 50.00% to 62.50%. “Preverbal and Postverbal Subjects with Unergative Verbs” (see Figure 6) displays a great deal of overlapped variability and no significant differences emerge among the three groups of child L2 learners either.Footnote ¹⁹

Figure 1.

Box-and-whiskers plot of “Null/Overt Subjects in Main Clauses” (median %) for the three experimental and the three control groups.

Figure 2.

Box-and-whiskers plot of “Null/Overt Subjects in Subordinate Clauses” (median %) for the three experimental and the three control groups.

Figure 3.

Box-and-whiskers plot of “Null/Overt Expletive Subjects” (median %) for the three experimental and the three control groups.

Figure 4.

Box-and-whiskers plot of “Preverbal/Postverbal Subjects with Unaccusative Verbs” (median %) for the three experimental and the three control groups.

Figure 5.

Box-and-whiskers plot of “Preverbal/Postverbal Subjects with Transitive Verbs” (median %) for the three experimental and the three control groups.

Figure 6.

Box-and-whiskers plot of “Preverbal/Postverbal Subjects with Unergative Verbs” (median %) for the three experimental and the three control groups.

The Mann–Whitney U-test was also applied to the set of preference variables, which tested the learners’ preference for the L1 option. Except for the “Preference for Preverbal Subjects with Transitive Verbs”, which is the most natural expected order, and the “Preference for Preverbal Subjects with Unergative Verbs” in contexts triggering SV orders, results are expected to follow a general falling tendency, since the L1 option is expected to gradually disappear. Results are displayed in Table 5.

Table 5.

Mann–Whitney U-test: preference variables.

As Table 5 illustrates, the “Preference for Overt Subjects in Main Clauses” (see Figure 7) shows a clear decrease-by-age tendency, as was expected. The sentences included in the task to test this variable (i.e., sentences (4) and (9) from Table 1 and sentences (4), (6) and (12) from Table 2) involved a neutral non-emphatic context and hence a null subject was expected. Significant differences can be observed between the five- and the ten-year-old groups (U = 13.500, p = .008) and the ten- and seventeen-year-old groups (U = 26.000, p = .031). In “Preference for Overt Subjects in Subordinate Clauses” (see Figure 8), similar results are obtained. Sentences testing this variable in the task (i.e., sentences (2), (7) and (12) from Table 1 and sentences (9), (15) and (13) from Table 2) triggered missing subjects as no contrastive context was provided and the pronoun in the subordinate clause could not co-refer with the referential expression in the main clause. A declining tendency is observed and both the five- and ten-year-old groups and the ten- and seventeen-year-old groups differ significantly (U = 13.500, p = .008 and U = 24.000, p = .015). “Preference for Overt Expletive Subjects” (see Figure 9) already displays lower (i.e., more target-like) results from the five-year-old group. No significant differences are found between the five- and ten-year-old groups or the ten- and seventeen-year-old groups. As was explained in section 3.1, the discourse neutral word order for unergative verbs is SV whereas for unaccusative verbs it is VS and in informationally focused structures both unergative and unaccusative verbs trigger VS word orders. The sentences in the task that tested “Preference for Preverbal Subjects with Unaccusative Verbs” (i.e., sentences (1) and (5) from Table 1 and sentences (3) and (14) from Table 2; see Figure 10) present neutral and focused environments and trigger VS word order. A significant difference is observed between the ten- and seventeen-year-old groups (U = 22.000, p = .018). Results from the five-year-old group, whose median and mean are identical to those of the ten-year-old group, show a great deal of variability, thus implying, as will be explained in the Discussion section, that these children are not yet aware of the syntactic–discourse constraints applying here or to lexical verb class. Median and mean values of 100.00% for “Preference for Preverbal Subjects with Transitive Verbs” (see Figure 11) is observed in the three experimental groups and with no variability of results. A VS order in transitive constructions is indeed a marked or unnatural order, which native speakers often reject and which non-native speakers clearly do not prefer. As for “Preference for Preverbal Subjects with Unergative Verbs” in focused contexts (i.e., sentence (10) from Table 1 and sentence (1) from Table 2; see Figure 12), where a VS order is triggered, a significant difference is observed between the ten- and the seventeen-year-old groups (U = 26.500, p = .031). “Preference for Preverbal Subjects with Unergative Verbs” in neutral contexts (i.e., sentence (13) from Table 1 and sentence (8) from Table 2; see Figure 13), where an SV order is expected, shows a clear preference for preverbal subjects in the five- and ten-year-old groups, not resulting from their awareness of verb class or syntax–discourse focus factors but probably due to their generalized preverbal preference. In the seventeen-year-old group, we find a significant difference with respect to the ten-year-old group (U = 30.000, p = .031), indicating alternation between preverbal and postverbal subjects at chance level, which shows some awareness of verb class and syntax–discourse focus factors in this context but clearly not complete mastery.

Figure 7.

Box-and-whiskers plot of “Preference for Overt Subjects in Main Clauses” (median %) for the three experimental and the three control groups.

Figure 8.

Box-and-whiskers plot of “Preference for Overt Subjects in Subordinate Clauses” (median %) for the three experimental and the three control groups.

Figure 9.

Box-and-whiskers plot of “Preference for Overt Expletive Subjects” (median %) for the three experimental and the three control groups.

Figure 10.

Box-and-whiskers plot of “Preference for Preverbal Subjects with Unaccusative Verbs” (median %) for the three experimental and the three control groups.

Figure 11.

Box-and-whiskers plot of “Preference for Preverbal Subjects with Transitive Verbs” (median %) for the three experimental and the three control groups.

Figure 12.

Box-and-whiskers plot of “Preference for Preverbal Subjects with Unergative Verbs (VS)” (median %) for the three experimental and the three control groups.

Figure 13.

Box-and-whiskers plot of “Preference for Preverbal Subjects with Unergative Verbs (SV)” (median %) for the three experimental and the three control groups.

5.2 A comparison between native and non-native control responses

L2 Spanish and native Spanish groups’ results on each age group were statistically compared for each acceptance/rejection and preference variable with the Mann–Whitney U-test. As Table 6 indicates, the non-native five-year-old group's percentages of correct judgments are rather high but still lower than those of the control group in all variables. Yet they are only significantly different in “Null/Overt Expletive Subjects” (U = 3.500, p = .003), which is the only variable expressing a strict syntactic contrast. “Null/Overt Subjects in Subordinate Clauses” display similarly low results in both the experimental and the control group (i.e., below 50%), which indicates the children's difficulty in judging the grammaticality of subjects in subordinate clauses in anaphora contexts, although we will see in the Discussion section that methodological aspects may have influenced the results. “Preverbal/Postverbal Subjects with Transitive Verbs” yield percentages of median results of only 50%, as instances of postverbal subjects with this type of verb have been considered extremely marked and have been judged as ungrammatical by both the experimental and the control groups.

Table 6.

Mann–Whitney U-test between the experimental and control five-year-old groups: acceptance variables.

Table 7 shows that the percentages of correct judgments from the ten-year-old experimental group have generally increased but are still slightly lower than those of the control group. Two variables present significant differences between the non-native and the native ten-year-old groups, namely “Null/Overt Subjects in Subordinate Clauses” (U = 24.500, p = .001), which continues to display low results in comparison to the other variables, and “Null/Overt Expletive Subjects” (U = 26.000, p < .001), which might still indicate some difficulty in fully acquiring the syntactic properties of Spanish L2 subjects. Similarly to the five-year-old groups, “Preverbal/Postverbal Subjects with Transitive Verbs” seem to present difficulties for native and non-native informants, whose medians only reach 50.00%, mostly rejecting postverbal subjects in this construction. As has been observed, postverbal subjects with this kind of verb are judged as ungrammatical or impossible by virtue of rendering an extremely marked interpretation. A remarkable drop in percentage of correct judgments should be noticed with respect to “Preverbal/Postverbal Subjects” with unaccusative and unergative verbs, which obtained higher percentages in the five-year-old groups. This will be addressed in the Discussion section.

Table 7.

Mann–Whitney U-test between the experimental and control ten-year-old groups: acceptance variables.

As for the seventeen-year-old experimental and control groups, Table 8 indicates further increase of correct judgments in all variables, although “Null/Overt Subjects in Subordinate Clauses” and “Preverbal/Postverbal Subjects with Transitive Verbs” present much lower median and mean values than in the other variables, and we can observe that not much development has taken place in such constructions. The only significant difference between native and non-native children emerges in “Null/Overt Subjects in Main Clauses” (U = 12.000, p = .015). Unexpectedly, non-native children's results are significantly higher than those of the control group, whose percentage median and mean are high but in which some variability of results is observed. Figures 1 to 6 also illustrate the graphic representation of the differences between the experimental and the control groups for each acceptance/rejection variable.

Table 8.

Mann–Whitney U-test between the experimental and control seventeen-year-old groups: acceptance variables.

With respect to the preference variables in the control and experimental five-year-old groups, Table 9 shows that results are similar to those of the acceptance variables, in that significantly different medians occur in the same corresponding variable, namely “Preference for Overt Expletive Subjects”. This L1 option variable is preferred by non-native children significantly more (U = 10.500, p = .024) than by native children, who never chose it. Both the experimental and the control groups prefer overt subjects in main and subordinate clauses, when a null subject would be required, with high median and mean values (i.e., a strong preference for the L1 option), showing the great difficulty in judging derivations at the syntax–discourse interface. Unaccusative verbs yield a preference for preverbal subjects at chance level for the experimental group and a bit lower for the control group when a postverbal subject would be required. As for “Preverbal Subjects with Transitive Verbs”, they were preferred with constant value results of 100.00% by both groups and hence postverbal subjects with transitive verbs were totally dispreferred. Likewise, no significant differences between groups were found in “Preference for Preverbal Subjects with Unergative Verbs” in focused structures triggering VS word order and in neutral contexts triggering SV word order. In contexts triggering VS word order, the non-native children have a stronger preference for preverbal subjects than native children, as their medians indicate, yet their mean values come closer to each other and, as can be observed in Figure 12, their variability of results overlaps completely and ranges from 100.00% to 0.00%, which makes the difference of the results non-significant. Finally, unergative verbs in neutral contexts, which should trigger preverbal subjects, yield an almost significantly higher preference for preverbal subjects in experimental children than in control children, who might be generalizing postverbal subjects, as Figure 13 indicates.

Table 9.

Mann–Whitney U-test between the experimental and control five-year-old groups: preference variables.

The preference for the L1 option lowers considerably as age and exposure to the L2 increase, and thus the results get closer and closer to those of the native group in the ten-year-old children. As Table 10 illustrates, only two variables display significantly different results between native and non-native speakers, namely “Preference for Overt subjects in Subordinate Clauses” and “Preference for Overt Expletive Subjects” (U = 48.000, p = .026; U = 58.500, p = .033). Yet the experimental group's median and mean values are reasonably low, which indicates clear development. The preverbal/postverbal distinction still remains problematic. Preverbal and postverbal subjects are still equally preferred (i.e., with a median value of 50.00%) with unaccusative verbs by both native and non-native children, whereas the expected word order is VS in both focused and neutral contexts in Spanish. As was the case with the five-year-old groups, preverbal subjects are preferred with transitive verbs by both groups with constant result values of 100.00%, which confirms the strongly marked status of postverbal subjects with transitive verbs. Regarding the preferred word order with neutral and focused contexts with unergative verbs, preverbal subjects are highly preferred by the two groups, even in the focused context where a postverbal subject should be preferred. In short, lexical verb class and syntactic–discourse factors related to word order remain problematic for both native and non-native speakers at this age/stage, as will be explained in the Discussion section.

Table 10.

Mann–Whitney U-test between the experimental and control ten-year-old groups: preference variables.

Almost target-like preference is achieved by the seventeen-year-old group, as Table 11 indicates. No significant differences in the results emerge between the experimental and the control group, whose preference judgments are syntactically and pragmatically native-like, accurately showing 0.00% median values in all variables favouring the L1 option. Overt subjects are no longer preferred by either group. Postverbal subjects are correctly preferred with unergative verbs in focused contexts (VS) and with unaccusative verbs. Not surprisingly, preverbal subjects are strongly preferred with transitive verbs with constant result values of 100.00%, as the transitive structures presented in the task do not involve focused contexts. Yet, and even if no significant difference is observed with respect to the control group, results in “Preference for Preverbal Subjects with Unergative Verbs (SV)” in neutral contexts do not pattern alike, with only 50% preference for preverbal subjects, which, as will be seen in the Discussion section, might indicate overgeneralization of postverbal subjects to inappropriate discourse contexts with unergative verbs. Figures 7 to 13 also illustrate the graphic representation of the differences between the experimental and the control groups for each preference variable.

Table 11.

Mann–Whitney U-test between the experimental and control seventeen-year-old groups: preference variables.