Incomplete End State L2 Acquisition

Incomplete End State L2 Acquisition: L2 Spanish CLLD and English CLD constructions 1. Acquisition theory Two contending theories of adult L2 acquisition, which make predictions for the L2 end state grammar, are the Representational Deficit Hypothesis (RDH) (Hawkins 2000, 2002) and Full Transfer/Full Access (FT/FA) (Schwartz and Sprouse 1994, 1996). On the one hand, the RDH predicts that the underlying representation of the L2 end state is not native-like but that learners can have native-like performance. On the other hand, the FT/FA contends that the L2 end state can be native-like due to nativelike representations. If both theories predict native-like performance, then how does one tease the two theories apart? A solution is to look beyond surface word order and morphology to the interpretive domain of the grammar. Sorace (1993, 1999, 2000, 2003), Robertson and Sorace (1999), and Tsimpli et al. (2003) argue that the interpretive domain, where the syntax interfaces with semantics, of the L2 end state grammar is vulnerable to fossilization and base persistent non-target forms on L1 influence. When the L2 input data are insufficient and the L2 speaker cannot restructure his/her grammar, the learner is unable to ‘let go’ of one of the two forms of the construction. The presence of two forms for one interpretation results in a permanent optionality between both variations. This permanent optionality is not found in the native grammar of the target language and is therefore a form of incomplete L2 acquisition. The present paper reports on a bidirectional study which examines the status of topic constructions, a discourse level operator, in the near-native L2 Spanish of L1 English speakers and the near-native L2 English of L1 Spanish speakers. We will show that, consistent with Sorace’s theory of optionality, fossilization occurs at the interpretive level. 2. Linguistic background 2.1 Topic constructions in English and Spanish In Germanic and Romance languages, topicalization is typically expressed by setting a phrase apart from a clause in order to reintroduce it into the discourse. The dislocated phrase is the topic and is one of a set of known items (discourse antecedents). The lower clause is the comment, or what the topic is about: (1) [Context: I have a hat and gloves.] The hat, I bought in Toronto. TOPIC Topic constructions are expressed using contrastive left dislocation (CLD) in Germanic languages (shown in (2)) while clitic left dislocation (CLLD) is a typical form of topicalization in Romance languages (as in (3)). Rizzi (1997) assumes a unified structure for CLLD and CLD in which a parameter differentiates English and Romance topic-comment structures. The topic parameter depends on the availability or not of a null anaphoric operator in the two types of structures. The parametric difference between the two language groups is that in English, the preposed topic connects to the comment via a null anaphoric operator (Op), (as in (2)), whereas in Spanish, the connection between the topic and the comment is made using a clitic (as in (3)): (2) (3) These shoes, I bought Op in Madrid. Estos zapatos, los compré en Madrid These shoes CL I-bought in Madrid ‘These shoes, I bought them in Madrid.’ In Spanish, the notion of specificity is crucial for topicalization. When a topicalized element is specific, it takes a clitic (CLLD structure) however, when a topicalized element is non-specific or generic (shown in (4) – (5) respectively), it does not take a clitic (Contreras 1976; Liceras et al. 1992; Arregi 2003): (4) (5) Este libro, *(lo) he leído muchas veces. This book, *(CL) I-have read many times ‘This book, I have read it many times.’ Revistas, (*las) leo a menudo. Magazines, (*CL) I-read often ‘Magazines, I read often.’ (specific) (non-specific) Zubizarreta (2001) points out that in Spanish, and other Romance languages, the specificity of an indefinite DP can be determined using a relative clause. A specific DP will take the indicative mood in the relative while the non-specific DP will take the subjunctive mood in the relative clause. This is shown in (6a) – (6b): (6) a. b. Una chica que conozco1 A girl that I-know(INDICATIVE) Una chica que conozca A girl that I-know(SUBJUNCTIVE) (una chica = specific) (una chica = non-specific) Thus, with respect to Spanish topic constructions, the presence or absence of the clitic results in a difference in interpretation of the topicalized element. Namely, with the CLLD construction the topic is interpreted as specific whereas with the CLD construction the topic in interpreted as non-specific. Such a contrast is not available in English topicalization. English is thus restricted to the CLD construction regardless of the interpretation of the topic, as in (7) and (8): (7) (8) This book, I have (*it2) read many times. Magazines, I often read (*them). 2.2 Clitic left dislocation and contrastive left dislocation We will assume that non-clitic topicalization in Spanish of the type found with non-specific topic constructions is an instance of CLD of the English type. We will base this assumption on the fact that Spanish non-specific topic constructions pattern with English-type CLD and not with Spanish-type CLLD. There are a number of asymmetries between CLLD and CLD3 but we will examine the following two crucial differences in the present study. First, the resumptive element in CLD is a null anaphoric operator (Op) whereas in CLLD it is a clitic. Second, the environment in which the left-dislocated phrase can occur differs between CLD and CLLD constructions. Specifically, in CLLD the left-dislocated phrase can occur in either root or embedded clauses, as in (9). The left-dislocated phrase in CLD, on the other hand, can only occur at the left periphery of root clauses, as in (10): (9) a. b. (10) 1 a. Un libro, lo leí anoche CLLD A book, CL I-read last night ‘A book, I read last night’ Me pregunto que, a María, el libro, quién se lo dio CLLD Me I-ask that to María the book who CL CL gave ‘I wonder who gave the book to María’ Un libro, leí anoche CLD A book, I-read last night These examples are taken from Zubizarreta (2001). Although English does not have Romance-type clitics, pronouns represent the same overt resumptive element. 3 For a comprehensive description of the asymmetries between CLD and CLLD see Anagnostopoulou (1997) and Valenzuela (to appear). 2 b. c. d. ‘A book, I read last night’ A book, I read last night. CLD *I wonder who, the book, to María, gave *Me pregunto que, tarjetas, a amigos, quién mandará4 CLD Me I-ask that cards, to friends, who will send ‘I wonder who will send cards to friends’ The minimal pairs in (9a) and (10a) show the different constructions with a non-specific topicalized phrase. In (9a), the topic is interpreted as a specific indefinite while in (10a) the topic is interpreted as an indefinite. Here, the native speaker preference is to choose the construction in CLD (10a) for indefinite DPs, bare plurals and mass nouns. A summary of the syntactic properties of CLD and CLLD is found in Table 1 below: Table 1. CLLD versus CLD CLLD CLD ROOT clauses: ROOT clauses: Este libro, lo he leído muchas veces Libros, leo a menudo (pero no revistas) This book, I-have read it many times Books, I read often (but not magazines) EMBEDDED clauses: Te aseguro que, tu secreto, no se lo he dicho a nadie I assure you that, your secret, I have not told it to anyone EMBEDDED clauses: * Me pregunto si, secretos, puede guardar *I wonder whether, secrets, she can keep 2.3 Research questions Assuming Spanish topicalization shows a distinction with respect to the specificity of the preposed topic where a non-specific topic takes the English-type CLD construction and a specific topic takes the Romance-type CLLD construction. Moreover, assuming English topicalization does not show sensitivity to specificity and is therefore restricted to the CLD construction, the following research questions can be made: i. ii. iii. 3. Will adult L2 Spanish learners be able to acquire the syntactic properties associated with CLLD? Will adult L2 English learners be able to let go of the syntactic properties associated with CLLD? Will there be a directional difference with respect to the associated interpretive properties? That is, will it be easier to acquire or to ‘let go’ of a property? Previous L2 research on topic/focus Previous research on discourse level properties (such as topic and focus) has found that such areas of the grammar are particularly vulnerable to fossilization (Lozano submitted; Hertel 2003; Pérez-Leroux and Glass 1999 for focus in L2 Spanish). In a study on topic constructions in the adult L2 Spanish of L1 English speakers, Valenzuela (2002) found that learners overgeneralized the use of the CLLD structure thereby not displaying a sensitivity to the specificity of the preposed topic. Given these results, we expect a 4 The sentence in (10d) is rendered grammatical as a CLLD (with a clitic) but the dislocated phrases, ‘tarjetas’ cards and ‘amigos’ friends, are in this case interpreted as specific: i. Me pregunto que, tarjetas, a amigos, quién se las mandará Me I-ask that cards, to friends, who CL CL will send ‘I wonder who will send cards to friends’ similar vulnerability in the interpretation of the Spanish and English topic constructions of the present study. 4. Methodology 4.1 Participants In order to examine the status of topic constructions in the L2 end state grammar a bidirectional study (L1 English/L2 Spanish and L1 Spanish/L2 English) was conducted. Participants for study 1 (L1 English/L2 Spanish) consisted of 15 L1 English speakers of L2 Spanish who had had their first exposure to Spanish after puberty. Participants were from England, Canada and United States and were living in Spain at the time of testing. In all cases, their work was conducted in Spanish and in most cases both work and home life were conducted in Spanish. The L2 participants were end state, near-native speakers of L2 Spanish. In addition to the L2 Spanish/L1 English experimental group, 25 monolingual L1 Spanish participants were tested in Spain as a control group. Participants for study 2 (L1 Spanish/L2 English) were 17 Spanish speakers of L2 English who had had their first exposure to English after childhood. Participants were from various Spanish-speaking countries and were living in either Canada or Spain at the time of testing. In all cases, their work was conducted in English and in most cases both work and home life were conducted in English. The L2 participants were end state, near-native speakers of L2 English. In addition to the L2 English/L1 Spanish experimental group, 15 monolingual L1 English participants were tested in Canada as a control group. For both studies, inclusion in the experiment was based on the learner being both a near-native and an end state speaker of L2 Spanish. Following similar procedure to that implemented in White and Genesee (1996) and Montrul and Slabakova (2003), speech samples from all participants (controls and L2) were extracted from short oral interviews. Based on the native speakers’ score margin, L2 speakers whose average scores also fell within that margin were deemed near-native. Scores on the near-nativeness interview together with the age of first exposure and number of years living in a Spanish environment were the criteria for inclusion in the study. 4.2 Tasks Participants were given two tasks, a sentence selection (SS) task and a sentence completion (SC) task5. All tasks aimed at testing: i. Sensitivity to specific versus non-specific topics in root clauses ii. Sensitivity to specific versus non-specific topics in embedded clauses 4.3 Oral Sentence Selection Task In the oral sentence selection task, participants read and listened to a context story and were asked to select the most appropriate concluding sentence. Context stories forced either a specific or a non-specific interpretation of the target topic in the concluding sentence. A Spanish sample is given in (11) and the English counterpart sample is given in (12): (11) Lola está haciendo los deberes de la universidad pero se acaba de dar cuenta que le faltan unos apuntes importantes. Mira por todas partes en la biblioteca, en su habitación, y en la clase pero... a. b. c. d. (12) 5 Esos apuntes, no encuentra por ninguna parte. Esos apuntes, no los encuentra por ninguna parte. Ni a ni b Ambas a y b DESIRED RESPONSE Lola is doing her homework. However, she just noticed that she is missing some important class notes. Lola looks in the library, in her room, and in the classroom but… a. Those class notes, she cannot find anywhere. DESIRED RESPONSE There was a third task, an oral grammaticality judgment task, which will not be reported on in the present paper. b. c. d. Those class notes, she cannot find them anywhere. Neither (a) nor (b). Both (a) and (b). 4.4 Sentence Completion Task In this task, participants read a context story and were then presented with a sentence that was begun (the topicalized element was provided) and they were asked to complete the sentence. Context stories forced either a specific or a non-specific interpretation of the target topic in the concluding sentence. A Spanish example is given in (13) and its English counterpart is given in (14): (13) El Sr Martínez ve mucho la televisión. No ve programas de deportes sino ve programas policíacos y documentales. Le pregunto por qué no ve programas de deportes y me explica: “Deportes, ___no miro_____ porque me aburren enormemente.” (14) Mr. Martinez watches a lot of television. He does not watch sports but rather he watches crime shows and documentaries. I ask him why and he explains: “Sports, __I don’t watch___ because they bore me enormously.” Participant responses were evaluated based on the construction type used to complete the sentence (suppliance of an overt resumptive element: a clitic in the case of the Spanish tests or an overt pronoun in the case of the English tests). 4.5 Results: Study 1 The results for the Spanish oral sentence selection are shown in Figure 1. The graph shows the distribution of group responses for topic constructions in root contexts. Figure 1: Results Spanish Oral Sentence Selection Task – root contexts 5% 1% 3% 14% 20% DISTRIBUTION RESPONSES 19% 40% 100% neither 94% 53% both no clitic clitic 37% 14% NEAR NATIVE CONTROL ROOT SPECIFIC NEAR NATIVE CONTROL ROOT NON-SPECIFIC As can be seen in Figure 1, both participant groups correctly selected the specific topic with a clitic (CLLD) in contexts which forced a specific interpretation of the topicalized element in the root clause. For example, participants correctly chose sentences as in (15) in contexts forcing a specific topic: (15) A Rafael, le voy a pedir la receta To Rafael, CL I-going to ask the recipe ‘Rafael, I am going to ask him for the recipe’ As for the contexts forcing a non-specific topic, the controls correctly chose the non-specific topic without a clitic (CLD), such as in (16): (16) Agua, toma por la mañana Water, s/he-drinks in the morning ‘Water, s/he drinks in the morning’ The near-natives, on the other hand, incorrectly chose the non-specific topic with a clitic 37% of the time. That is, they chose sentences as in (17): (17) Agua, la toma todas las mañanas Water, CL s/he-drinks every the morning ‘Water, s/he drinks it every morning’ Figure 2 shows the distribution of group responses in the oral sentence selection task for topic constructions in embedded contexts. Figure 2: Results Spanish Sentence Selection Task – embedded contexts 7% DISTRIBUTION RESPONSES 31% 3% 77% 100% neither both 93% no clitic clitic 67% 1% 1% 22% NEAR NATIVE CONTROL EMBEDDED SPECIFIC NEAR NATIVE CONTROL EMBEDDED NON-SPECIFIC As illustrated in Figure 2, both groups correctly preferred the CLLD construction with tokens eliciting specific topics in embedded contexts. That is, the showed a preference for sentences as in (18): (18) Insistió en que, los libros, los devolviera ahora mismo S/he-insited that, the books, CL I-return right now For the non-specific topics in embedded contexts, the desired response is a rejection of both options since CLD, unlike CLLD, cannot appear in embedded contexts. The controls showed a strong tendency of correctly selecting the option ‘neither’ (that is, of rejecting the sentence with or without a clitic) for the non-specific topics, while the near-natives preferred the ‘clitic’ (CLLD) with the non-specific topics (as in example (19)): (19) *Me parece que, café, lo debería tomar menos Me I-seem that, coffee, CL I-should drink less We will now turn to the results for the Spanish sentence completion task. In Figure 3, group percentage of sentences completed with or without a clitic for topics in root contexts are shown. Figure 3: Results Spanish Sentence Completion Task – root contexts 0% 11% DISTRIBUTION OF RESPONSES 47% 83% 100% null operator 89% pronoun 53% 17% NEAR NATIVE CONTROL ROOT SPECIFIC NEAR NATIVE CONTROL ROOT NON-SPECIFIC Both groups correctly provided clitics with specific left-dislocated topics in root environments. In contexts forcing non-specific interpretation, however, near-natives completed them with a clitic over 50% of the time. A single-factor ANOVA showed a significant difference between the groups on ROOTns sentences (F(1,38) = 19.11303, p<.01). In Figure 4, group percentage of sentences completed with or without a clitic for topics in embedded contexts are shown. Figure 4: Results Spanish Sentence Completion Task – embedded contexts 4% 1% DISTRIBUTION OF RESPONSES 27% 83% 96% 99% null operator pronoun 73% 17% NEAR NATIVE CONTROL EMBEDDED SPECIFIC NEAR NATIVE CONTROL EMBEDDED NON-SPECIFIC As can be seen from Figure 4, both groups correctly produced sentences with a clitic in contexts where a specific clitic was provided. For contexts where a non-specific embedded topic was provided, the control group produced sentences without clitics. The near-natives produced sentences with clitics where a clearly non-specific topic in an embedded clause was provided. A single-factor ANOVA only showed a significant difference between groups for the EMBns sentences (F(1,38)= 42.1244, p<.01). In summary, the results for study 1 indicate that the Spanish near-native group does appear to have acquired the CLLD structure (producing and accepting clitic constructions). However, they appear to prefer the CLLD construction (with clitic) regardless of specificity thereby not showing the specificity distinction. This indicates that the interpretive properties are problematic for the near-natives. Having seen the results for study 1, we will now turn to the results for study 2 (L2 English/L1 Spanish). 4.6 Results: L2 English /L1 Spanish The results for the English oral sentence selection are shown in Figure 5. The graph shows the distribution of group responses for topic constructions in root contexts. Figure 5: Results English Sentence Selection Task – root contexts 10% 17% 27% 31% 13% DISTRIBUTION OF RESPONSES 13% 15% 13% 48% neither both no pronoun pronoun 63% 48% 55% 30% 8% NEAR NATIVE CONTROL ROOT SPECIFIC 7% NEAR NATIVE 4% CONTROL ROOT NON-SPECIFIC In Figure 5 we see that the controls correctly selected the ‘no pronoun’ (as opposed to the ‘pronoun’) option most often for topics in specific contexts (48%) and non-specific contexts (55%). The near-natives selected the ‘no pronoun’ option for specific topics (48%) and non-specific topics (63%). That is, they chose sentences as in (20) - (21) in specific and non-specific contexts respectively: (20) (21) Peter, she could not talk to. Water, he never drinks. SPECIFIC NON-SPECIFIC With respect to the root specific topics, while near-native speakers are choosing the ‘no pronoun’ option more often, their selection of the ‘pronoun’ option, as in (22), is high at 30%. (22) Water, he never drinks it. Results for the English sentence selection task for topics in embedded contexts are shown in Figure 6. Recall that embedding is not possible in English CLD constructions, therefore, the expected response for these sentences is the ‘neither’ option. Figure 6: Results English Sentence Selection Task – embedded contexts DISTRIBUTION OF RESPONSES 35% 46% 47% 6% 5% 37% 14% 18% 27% 31% 20% NEAR NATIVE CONTROL EMBEDDED SPECIFIC 13% 27% 25% 24% 24% NEAR NATIVE neither both no pronoun pronoun CONTROL EMBEDDED NON-SPECIFIC As can be seen from Figure 6, both groups correctly preferred the ‘neither’ option for all embedded topics in English. Turning to the English sentence completion task, Figure 7 shows the results for topics in root contexts. Recall that in this task, the topicalized element was given and the participant was asked to complete the sentence. Since CLD is restricted to root contexts, only topics in root contexts were included in the task. Figure 7: Results English Sentence Completion Task – root contexts DISTRIBUTION OF RESPONSES 46% 64% 71% 83% null operator pronoun 54% 36% 29% 17% NEAR NATIVE CONTROL ROOT SPECIFIC NEAR NATIVE CONTROL ROOT NON-SPECIFIC The controls are predictably low in their suppliance of pronouns for sentences with both specific topics in root CLD (mean 29%) and non-specific topics in root CLD (17%). That is, as expected, the control group is not showing a difference in their treatment of specific and non-specific topics since both are given with the CLD construction. The near-natives, on the other hand, are supplying pronouns with the specific topics (mean 54%) more often than with the non-specific topics (mean 36%). This sensitivity to the specificity of the topicalized element shows influence from their L1 Spanish. In summary, study 2 shows that the near-native speakers are sensitive to the syntactic properties which distinguish CLD and CLLD. That is, appear to know the constraints on the contexts in which CLD can appear (only in root contexts). However, the near-native group is both accepting and producing pronouns (overt resumptive elements like clitics) with the specific topics there by exhibiting L1 influence. 5. Discussion The aim of this paper was to examine three main research questions. First, whether adult L2 Spanish learners (L1 English) were able to acquire the syntactic properties associated with Spanish CLLD constructions. Second, whether adult L2 English learners were able to ‘let go’ of the syntactic properties associated with CLLD, which are present in their L1 Spanish but not in their L2 English. Third, whether there was a directional difference with respect to the associated interpretive properties associated with CLD and CLLD. With respect to the first research question, the L2 Spanish group did appear to have acquired the syntactic properties of CLLD. That is, they appeared to know the CLLD structure. However, they overgeneralized the CLLD structure to extend to topics in non-speciifc contexts (unlike the control group) suggesting a lack of sensitivity to the specificity distinction. This overgeneralization cannot be attributed to L1 influence since English does not have this distinction. For the second research question, the Spanish near-native group showed influence from their L1 English in that they provided an overt resumptive element with specific topics. That is, although English does not distinguish between specific and non-specific topics in the way that Spanish does, the L2 English subjects were showing a specificity distinction. Finally, with respect to the third research question, the L2 English results showed more L1 influence at the interpretive level than the L2 Spanish results. This suggests that ‘letting go’ of an L1 property is more difficult than acquiring one based on the L2 input. Moreover, consistent with Sorace (1993, 1999, 2000, 2003) and Robertson and Sorace (1999), vulnerability appears to occur at the interpretive level. In conclusion, these results provide evidence in support of Sorace’s theory of optionality whereby nontarget forms are based on L1 influence and the area of vulnerability is at the interpretive level and not at the level of syntax.