Noun-raising and Adjectival Interpretative Reflexes in the L2 Spanish of Germanic and Italian Learners

Noun-raising and Adjectival Interpretative Reflexes in the L2 Spanish of Germanic and Italian Learners Jason Rothman 1, Pedro Guijarro-Fuentes 2, Michael Iverson 1 and Tiffany Judy 1 University of Iowa 1, Plymouth University 2 1. Introduction Experimentally, we investigate L2 Spanish knowledge of properties related to noun-raising, focusing on the semantic reflexes of adjectival interpretive restrictions mediated by their placement relative to head nouns in Romance languages. The semantic properties we investigate fall out from obligatory nominal feature-checking movement, which obtains in Romance languages and differently from Germanic languages where nouns remain in situ in the overt syntax irrespective of the meaning of the adjective (i.e. restrictive versus nonrestrictive interpretations). Focusing on the developmental predictions that follow from the Full Transfer/Full Access model (Schwartz & Sprouse 1996), we note that if transfer occurs, Italian learners of L2 Spanish should have few (perhaps no) problems with Spanish adjectival interpretation from the earliest of L2 proficiency stages. However, German and English learners of L2 Spanish, due to the lack of noun raising in the L1, have no recourse to transfer the necessary features and are therefore predicted to lag behind Italian learners significantly at lower levels of proficiency for noun-raising related properties. If, however, adults have full accessibility to new functional categories and features, then the predicted difference between these sets of L2 Spanish learners at the beginning and intermediate levels should lessen significantly, if not disappear, by the advanced speaker levels. Conversely, the Processibility Theory (PT) (e.g. Pienemann 1998), which maintains that a then-current ability to process and parse L2 input and not L1 transfer is both the principle variable influencing adult initial state hypotheses and what drives inter-language transitions, predicts no significant differences between the Italian and Germanic groups at any level of proficiency. Furthermore, we test the predictions of so-called Representational Deficits (RD)accounts (e.g. Hawkins & Chan 1997, Hawkins & Hattori 2006) which argue that adults have no access to features not instantiated in the syntax of their L1, and therefore movement dependant on new L2 features, such as the nounraising for Germanic learners of L2 Spanish, should at best remain optional. Because these theoretical accounts make distinctive predictions either for the role of transfer in L2 interlanguage development and/or the competence of speakers at advanced levels of proficiency for the properties under investigation, this study s cross-sectional design will help to examine their relative

epistemological values. As we will see, the data support only Full Accessibility (FA) and Full Transfer (FT) accounts. 2. Syntax Although Germanic and Romance languages share similarities in terms of some DP-features (i.e. number in all cases, gender between German and Romance), they are syntactically different in significant ways, which in turn explains differences in the syntactic position of adjectives and how related semantic reflexes are calculated. The main purpose of the current investigation concerns two interrelated aspects of the morphosyntax of nominals: on the one hand, gender agreement (which exists in Spanish, Italian and German, but not in English; 1 in the Romance languages, but not in German, gender marking is realized inside the nominal DP, namely, between the head noun and adjectives) and, on the other hand, the lack of interpretive ambiguity depending on the iteration of adjectives within the Romance DP (i.e. in prenominal and postnominal positions) as compared to the inherent ambiguity of the adjectival position in Germanic languages. 2.1 German Following recent linguistic analyses of the German DP (Penner 1993; Penner & Schönenberger 1992; Eisenbeiss 2003; Parodi, Schwartz & Clahsen 2004 and works cited therein), German DPs are head-initial, adjectives are based-generated to the left of the noun and agreement is realized syntactically within the DP as illustrated in (1). (1) DP Spec D D NP AP NP A Spec N A N 1 Because space limitations and the overwhelming similarities between English and German, we will not review English language properties, but occasionally we touch on it. For explanations related to English language see Judy et al. (2008) and Guijarro-Fuentes et al. (to appear).

In (1) the adjective phrase (AP) is assumed to be adjoined to the NP explaining the fixed linear order within a German DP, as shown in (2): (2) Die schöne rote Rose the.fem. beautiful.fem. red.fem. rose The beautiful red rose. Regarding morphological agreement, German has a rich nominal agreement system in D which spreads to all elements within the DP by Spec-Head agreement. German has grammatical gender and number inflection (which are realized on determiners and adjectives) on D-elements together with case distinctions. However, due to syncretism these syntactical features are difficult to tease apart (namely, while in German there is mainly agreement between the determiner and the adjective, gender marking on the noun is hardly visible). 2.2 Spanish/Italian Spanish and Italian possess full nominal agreement systems (so-called phifeatures, see Chomsky 1981, 1995) and in the expanded nominal the different components of the Spanish/Italian DP (i.e. determiners, nouns and adjectives) concord with each other in both gender and number. Gender concord is realized between the head noun and its modifying article and between the head noun and adjectives; namely, if the head noun is feminine (niña/ragazza girl ), then the article as well as all adjectives are feminine as well, as illustrated in (3). (3) La niña bonita La ragazza bella The.FEM.SG. girl.fem.sg. pretty.fem.sg As shown in (3), while determiners always precede the noun, adjectives most often appear after the noun. Acknowledging that there are subsets of adjectives that have a strict word order (either pre- (e.g. temporal modifiers) or post-nominal (e.g. nationalities)), we focus here on the vast majority of qualitative adjectives that can appear in two syntactic positions with syntactically imposed necessary differences in interpretation (consider (4a) and (4b)). (4) a. Los hombres fuertes b. Los fuertes hombres Gli uomini forti i forti uomini The strong men The strong men

Thus, unlike in German where the prenominal position (in most cases the only available position) of these adjectives is inherently ambiguous between a restrictive (group-denoting) and a non-restrictive (kind-denoting) meaning, Spanish/Italian adjectival interpretation is unambiguous, depending on the adjective s spell-out (or surface) position. Whereas (4a) can only refer to some of all existing men (a group-denoting reading), 4(b) must be interpreted as all men, such that fuertes/forti strong is simply a defining characteristic of being a man (a kind-denoting reading). In both English and German, where only one word order is possible, both readings are available and thus, outside of context is ambiguous. In order to explain these descriptive collocational aspects of Spanish/Italian adjectives, we adhere to Bernstein s (1993, 2001) analysis that claims that intersective adjectives are essentially prenominal in Spanish/Italian, and that their postnominal position is derived by noun movement over the adjective. Adopting Bernstein s analysis, the head noun moves up (as symbolized by X in (8)) from N to the head of Word Marker Phrase (WMP) and then to Number Phrase (NumP). 2 Following this reasoning, non-intersective adjectives are merged higher in the DP, either in the Specifier position of NumP or some other functional projection above NumP, thus allowing the possibility of [Det Adj N] word order even after obligatory noun raising. 3 Leaving aside other details, the resulting distinctive possible word orders between the DP in Romance and Germanic languages are syntactically restricted and have semantic consequences. Since movement in Spanish/Italian is overt, no semantic ambiguity can obtain in the sense that the surface position of the adjective correlates faithfully to the high or low syntactic position for adjectives, from which their meaning falls out universally. Nevertheless, since movement in German and English is covert at LF semantic ambiguity is generated. Let us consider the following example: (5) [ DP ein ] [ NumP [ AP armes] [ Num +singular]] [ np [ AP armes] [ N Kind]] unfortunate not rich Taking a closer look at (5), the German phrase ein armes Kind is ambiguous because armes poor can be underlyingly linked to either the higher clausal position, that is, NumP, which can only be interpreted as non-intersective, or to the lower clausal position in n, which must be interpreted as intersective. In Spanish/Italian, on the other hand, N raises to NumP, so that any prenominal 2 We are aware that more recent analyses motivate the placement of the adjectives in terms of the existence of some extra categories, for example, [np] (e.g. Demonte 2008). 3 For feature checking reasons, adjectives must raise in Spanish even when they appear overtly to the right of the head noun. This must mean that adjectives in such a position move to a higher functional category within the DP-layer.

adjective must be non-intersective, and any postnominal adjective must be intersective. Consequently, the meanings are uniquely determined by the syntactic position of the adjective poor and, therefore, adjectival interpretation in this respect is a strong sign of the underlying mental representation of the syntax. 3. Methodology A total of 107 subjects participated in this study. The control group consisted of twelve native Spanish speakers and there were two L2 subject groups (Italian and Germanic) that were divided into two proficiency groups (Advanced and Intermediate) based on their performance on a widely used 50- point two-part cloze and multiple choice proficiency exam. There were 35 L1 Italian subjects (n=23 Advanced, n=12 Intermediate) and 59 L1 Germanic participants (n=12 English Advanced, n=15 German Advanced, n=29 English Intermediate, n=4 German Intermediate). In addition to the experimental tasks completed by all subjects, each subject completed an extensive linguistic history background questionnaire. The subject groups described above completed three tasks designed to test their acquisition of the syntactic and semantic properties of the Spanish DP outlined in the previous section. The first task, a Grammaticality Judgment/Correction Task (GJCT) was designed to test for knowledge of gender and number agreement between nouns and adjectives (n=20) and determiners and nouns (n=20), as can be seen in (10) below. Recall that rejection required correction, so we were able to know the reason for which rejection occurred. Rejections were only counted as successful if the appropriate correction was done. For space limitations we do not present the types of sentences that served as filled, however, we not that there were three filler categories with 20 tokens each for a total of 60 fillers or a 3/2 ratio of fillers to test sentences: (10) Knowledge of gender and number agreement a. Ellas prefieren comida fresca. (grammatical Noun-Adj gender/number) They prefer food FEM.SING. fresh FEM.SING. They prefer fresh food. b. El chico siempre lee aquí. (grammatical Det-Noun gender/number agreement)' The MASC.SING. boy MASC.SING. always reads here The boy always reads here. c. *Mis tíos de Los Ángeles son famosas. (ungrammatical Noun-Adj gender agreement long distance) My uncles MASC.PLRL. from L.A. are famous FEM.PLRL. My uncles from L.A. are famous. d. *Iván necesita lavar el ropa. (ungrammatical Det-Noun gender agreement)

Iván needs to wash the MASC.SING. clothes FEM.SING. Iván needs to wash the clothes. e. *Su vida académicas es muy difícil. (ungrammatical Noun-Adj number agreement) His life FEM.SING. academic FEM.PLRL.is very difficult His academic life is very difficult. f. *Felipe me compró las pulsera. (ungrammatical Det-Noun number agreement) Felipe me bought the FEM.PLRL. bracelet FEM.SING. Felipe bought me the bracelet. The second task was a Semantic Interpretation task in which the subjects read a short sentence that contained a DP with either a prenominal or a postnominal adjective. From this sentence, they were asked to choose the correct meaning based on two interpretations provided to them. Example (11) is a prenominal token (n=5) and example (12) is a postnominal token (n=5). The correct interpretation is bolded. (11) Su vecino es un pobre hombre. The man is unfortunate. The man is not wealthy. (12) Ella es la persona única de quien te hablé. She is the only person that I told you about. She is the unique person that I told you about. For the final task, the Context-based Collocation task, subjects were instructed to read a short context and write the adjective found at the end of the token in either the prenominal or postnominal blank also provided. Example (13) is a prenominal token (n=5) and example (14) is a postnominal token (n=5). (13) Mi mejor amiga se llama Magda. Ella es una persona muy amable y cariñosa. Aunque solo tenemos 22 años, hace mucho tiempo que somos amigas. Magda es una vieja amiga (viejo). My best friend is named Magda. She is a very nice and affectionate person. Even though we are only 22 years old, we have been friends for a long time. Magda is an old friend. (14) Creo que la gente que tiene mucho dinero puede ser muy arrogante. Pero la semana pasada conocimos a unos millonarios que no son así. Los

millonarios simpáticos (simpático) que conocimos me cayeron muy bien. I think that people that have a lot of money can be very arrogant. But, last week we met some millionaires that aren t like that. I really like the nice millionaires that we met! The results of these empirical experiments are presented in the following section, inclusive of a statistical analysis as well as an interpretation of them. 4. Results In this section, the results of the three experimental tasks are presented and statistically analyzed. Each of the analyses compares the performance of the five participant groups mentioned above using a mixed-model ANOVA. Any posthoc test conducted was of the Bonferroni type, and all analyses used a significance level of α = 0.05. 4.1 Task 1: Grammaticality Judgment/Correction Task Task 1 was a grammaticality judgment/correction task in which participants were asked to rate sentences with permissible and impermissible instances of agreement as grammatical or ungrammatical. Group means of acceptance (out of 10) and standard deviation are seen in Table 1 below: Table 1: Experiment 1, GJT. Group means, number accepted (Standard deviations) NS ItAdv ItInt GrAdv GrInt NA Good 10 (0) 9.57 (0.66) 9.33 (0.98) 9.89 (0.32) 9.48 (0.76) NA Bad 0.42 (0.79) 0.39 (0.50) 0.75 (0.62) 0.70 (0.78) 2.82 (2.31) DN Good 10 (0) 9.61 (0.58) 9.67 (0.49) 9.78 (0.42) 9.73 (0.63) DN Bad 1.25 (1.14) 0.30 (0.56) 0.58 (0.67) 0.67 (0.68) 1.52 (1.75) NS=Native Speaker; It=Italian; Gr=Germanic; Adv=Advanced; Int=Intermediate; NA=Noun/Adjective agreement; DN=Determiner/Noun agreement Inferential statistics showed a significant interaction between group and grammaticality (F(4,102)=10.60; p<0.001) and group and item type (noun/adjective agreement vs. determiner/noun agreement) (F(4,102)=6.40; p<0.001). So, there was an overall difference somewhere among groups in the distinction they made between grammatical and ungrammatical items and between item types. Post-hoc tests revealed that the intermediate Germanic group differed significantly from every other group with respect to

noun/adjective agreement (p<0.007 for all comparisons) and Ungrammatical items (p<0.001 for all comparisons). There were no other statistically significant differences anywhere among any of the other groups. In sum, all L2 learners except for the intermediate Germanic group demonstrated robust knowledge gender agreement morphological accord, performing in a way similar to natives with respect to gender/number agreement between nouns and adjectives and determiners and nouns. Since this is an off-line task and is not production based, there is no way to assess whether or not these learners would perform equally well with gender agreement in production. 4.2 Task 2: Semantic Interpretation Task (SIT) In this task, participants were instructed to select the meaning of a prenominal or postnominal adjective. Group means (out of 5) and standard deviations of number of items correct for a given adjectival position are given in Table 2 below: Table 2: Experiment 2, SIT. Group means, number correct (Standard deviations) NS ItAdv ItInt GrAdv GrInt Prenominal 4.5 (0.52) 4.35 (0.65) 4.17 (0.72) 4.37 (0.74) 3.30 (1.16) Postnominal 4.42 (0.51) 4.26 (0.69) 4.08 (0.67) 3.96 (1.13) 2.64 (0.90) NS = Native Speaker; It = Italian; Gr = Germanic; Adv = Advanced; Int = Intermediate; Prenominal = prenominal adjectives; Postnominal = postnominal adjectives The results of the ANOVA showed main effects for group (F(4,102)=22.43; p<0.001) and adjectival position (pre/post; F(1,102)=5.00; p = 0.003)), meaning that there was at least one group performing differently from the others and that adjectival position influenced performance. Post-hoc tests uncovered that, again, the Intermediate Germanic group performed differently than all the other groups regardless of the position of the adjective (p<0.008 when compared to all other groups). There were no other statistically significant differences between any other groups. Additionally, all groups fared slightly better with prenominal adjectives than with postnominal adjectives. Since all groups except the Intermediate Germanic group performed similarly to native speakers, we can infer that every group except the Intermediate Germanic group has a similar mental representation for the syntax of the target DP from which the semantic interpretations of pre/postnominal adjectives obtain. 4.3 Task 3: Collocation Task (CT)

Task 3 was a production task in which participants had to place an adjective in either a prenominal or postnominal position depending on a short context. Table 3 shows the average number correct (out of 5) and standard deviation for each group for both prenominal and postnominal items: Table 3: Experiment 3, CT. Group means, number correct (Standard deviations) NS ItAdv ItInt GrAdv GrInt Prenominal 3.08 (0.67) 4.09 (0.73) 4.08 (0.67) 3.70 (1.23) 1.94 (0.97) Postnominal 4.83 (0.39) 4.52 (0.59) 4.42 (0.67) 4.41 (0.93) 4.09 (1.23) NS = Native Speaker; It = Italian; Gr = Germanic; Adv = Advanced; Int = Intermediate; Prenominal = prenominal adjectives; Postnominal = postnominal adjectives A statistical analysis revealed an interaction between group and adjectival position (F(4,102)=10.89; p<0.001), suggesting that while a group (or groups) may not differ from the others on this task in a global sense, it (they) may differ from the others in judging a given adjectival position. Further investigation found that, again, the Intermediate Germanic group performed significantly lower than each group with respect to postnominal adjectives (p < 0.001 for all comparisons) and that there were no statistically significant differences between any of the other groups, as also seen in the first two tasks. So, while the Intermediate Germanic learners may fare well with postnominal adjectives (which, notably, is claimed as the default position in Spanish classroom instruction, cf. Anderson (2007) for French, Rothman (2008) for pedagogical effects), they do not do so with prenominal adjectives and therefore do not have complete command of adjectival semantics. 5. Experimental Discussion Looking globally at the results of the three tasks allows us to comment about the groups overall knowledge and their relationship to each other with respect to DP-morphological agreement and adjectival semantics. With the exception of the Intermediate Germanic group, each L2 group shows knowledge of the grammatical and ungrammatical instances of gender/number agreement and knowledge of the semantic nuances of prenominal and postnominal adjectives. Furthermore, this knowledge does not differ from native speaker knowledge of the same phenomena. The Intermediate Germanic group differed overall from every other group, and from this we can conclude that they have not yet acquired the nominal phi-features of Spanish that are part of the

morphophonological forms that they use indeterminately (task 1) and are necessary to converge on the knowledge that Spanish has obligatory noun raising, a precursor to adjectival interpretation in Spanish (tasks 2 and 3). Equally, since knowledge of these DP properties hinges upon the instantiation of nominal phi-features, we can infer that any group performing like natives with respect to these phenomena have the appropriate features, acquired either while learning the L2 (in the case of Germanic learners) or transferred from the L1 (in the case of the Italian learners). 6. Conclusion As it relates to the properties we tested, the introduction specified the predictions of three L2 acquisition theories: the developmental predictions stemming from the Full Transfer/Full Access model (Schwartz & Sprouse 1996) as compared to Representational Deficit accounts (Hawkins & Chan 1997; Hawkins & Hattori 2006) and Processibility Theory accounts (Pienemann 1998). To briefly recap, the FT/FA model claims that L2 learners initially transfer their L1 in its entirety (except for lexical items), but functional properties (categories and features) provided by UG are purported to remain available to adult L2 learners. Therefore, in earlier stages of development, FT/FA predicts that the L1 Italian subjects will perform quite accurately with respect to the properties tested herein precisely because, like Spanish, Italian has nominal phi-features that require noun-raising within the overt syntax. Similarly, FT/FA predicts that L1 Germanic subjects at early stages will perform less accurately at the very least with adjectival interpretation since German and English do not have nounraising. Since German does have grammatical gender (although differently than in Romance) it could be that German and English learners at lower levels would perform differently on the GJCT, but the results did not show this (p=0.48). In any case, since UG is argued to be accessible, FT/FA predicts that with sufficient exposure to the target language, it is possible for all L2 learners, regardless of their L1, to acquire target representations and thus perform like native speakers on all tasks at higher levels of proficiency. In contrast, RDAs claim that only those features instantiated in the L1 are available for use in the L2. Therefore, RDAs predict that only Italian speakers (and crucially not Germanic speakers) will demonstrate invariable target knowledge of the syntax and semantics of the Spanish DP. Because UG is purported to be inaccessible to adult L2 learners, RDAs predict, in this case, that L1 Italian subjects will be initially more accurate than L1 Germanic subjects as well as ad infinitum. The third hypothesis tested herein, the Processibility Theory, predicts that no differences should be found between the Italian and Germanic groups, regardless of proficiency level since L1 transfer is supposedly not what drives interlanguage development, but rather the learner s

processibility level. That is, it is predicted to be equally difficult for Italian and Germanic speakers at all developmental stages to parse L2 Spanish. When the results of the three experimental tasks are examined, it is clear that they strongly support the Full Transfer/Full Access model and in doing so provide evidence against the predictions of RDAs and Processibility Theory. Both the Advanced and Intermediate Italian subject groups performed at the native control group level across all three experimental tasks, thus suggesting that the L1 Italian learners have successfully converged upon the syntax and semantics of the Spanish DP. Since this is not the case at the intermediate level of the Germanic group, it is reasonable to deduce that L1 transfer explains this difference. And so, this comparative pattern presents a problem of explanation for Processibility accounts. Nevertheless, the Advanced Germanic group also performed like the native control group across all tasks, suggesting that in due course and via exposure to Spanish data the acquisition of new phi-features (values) can be accomplished, thus implicating a direct role for UG in adult L2 acquisition. This finding is consistent with the prediction of Full Accessibility accounts. When the proficiency levels are collapsed, the data reveal development for one group (Germanic) and no development due to ceiling performance already at the intermediate level (Italian); this finding is also consistent with L1 transfer hypotheses. In sum, since both Italian subject groups and the Advanced Germanic group performed like the control group and, crucially, because the Intermediate Germanic group was the only group to diverge, we conclude that the results support the FT/FA model only and provide evidence against RDAs and the Processibility Theory. References Anderson, Bruce. (2007). Pedagogical Rules and their Relationship to Frequency in the Input: Observational and Empirical Data from L2 French. Applied Lingusitics 28, 286-308. Bernstein, Judy. (1993). Topics in the syntax of nominal structure across Romance. Doctoral dissertation, Graduate Center. City University of New York. Bernstein, Judy. (2001). The DP Hypothesis: Identifying Clausal Properties in the Nominal Domain. In M. Baltin and C. Collins (Eds.) The Handbook of Contemporary Syntactic Theory. London: Blackwell. Chomsky, Noam. (1981). Lectures on government and binding. Foris, Dordrecht. Chomsky, Noam. (1995). The minimalist program. Cambridge, MA: MIT Press. Demonte, Violeta. (2008). Meaning-Form correlations and adjective position in Spanish. In Ch. Kennedy and L. Mcnally (Eds.), The semantics of adjectives and adverbs. Oxford University Press. Eisenbeiss, S. (2003). Feature-Driven structure building. A study on the acquisition of Noun-Phrase and inflection. Unpublished Ph.D. University of Düsseldorf. Guijarro-Fuentes, Pedro, Judy, Tiffany & Rothman, Jason. (to appear). The acquisition of

adjectives in instructed Spanish: the role of individual characteristics. In A. Benati (ed.), Issues in Second Language Proficiency. Continuum Publishing. Hawkins, Roger & Chan, Cecilia. (1997). The partial accessibility of Universal Grammar in second language acquisition: The failed functional features hypothesis. Second Language Research, 13, 187 226. Hawkins, Roger & Hattori, Hajime. (2006). Interpretation of English multiple whquestions by Japanese speakers, a missing uninterpretable feature account. Second Language Research, 22, 269 301. Judy, Tiffany, Guijarro-Fuentes, Pedro & Rothman, Jason. (2008). Adult Accessibility to L2 Representational Primitives: Evidence from the Spanish DP. In Bowles, M.; Foote, R., Perpiñán, S. and R. Bhatt (eds.) Selected Proceedings of Second Language Research Forum 2007, 1-21. Somerville, MA: Cascadilla Press. Parodi, T., Schwartz, B. & Clashen, H. (2004). On the L2 acquisition of the morphosyntax of German nominals. Linguistics 42 (3), 669-705. Penner, Z. (1993). The earliest stage in the acquisition of the nominal phrase in Bernese Swiss German: syntactic bootstrapping and the architecture of language learning. Arbeit-spapier 30, Institut für Sprachwissenschaft, University of Bern. Penner, Z. & Schönenberger, M. (1992). The distribution of nominal agreement features in Swiss-German dialects and the strong DP/CP parallelism hypothesis. In Z. Penner (ed.), Topics in Swiss Grammar Syntax, 331-346. Bern: Lang. Pienemann, Manfred. (1998). Language Processing and Second Language Development:Processability Theory, Amsterdam: Benjamins. Rothman, Jason. (2008). Aspectual Morphology Use in Adult L2 Spanish & The Competing Systems Hypothesis: When Pedagogical and Linguistic Rules Conflict. Languages in Contrast, 8, 1, 74-106. Rothman, Jason. (2009). Pragmatic Deficits with Syntactic Consequences: L2 Pronominal Subjects and the Syntax-Pragmatics Interface. Journal of Pragmatics, 41. Schwartz, Bonnie & Sprouse, Rex. (1996). L2 cognitive states and the Full Transfer/Full Access model. Second Language Research 12, 40 72. Blackwell.