Running head: Examining Morphological Variability in L2 Learners

We examined potential sources of morphological variability in adult L1-English L2-Spanish learners, with a focus on L1-L2 similarity, morphological markedness, and knowledge type (receptive vs. expressive). Experiment 1 uses event-related potentials to examine noun-adjective number (present in L1) and gender agreement (absent in L1) in online sentence comprehension (receptive knowledge). For each feature, markedness was manipulated, such that half of the critical noun-adjective combinations were feminine (marked) and the other half, masculine; half were used in the plural (marked) and the other half in the singular. With this set-up, we examined learners’ potential overreliance on unmarked forms or “defaults” (singular/masculine). Experiment 2 examines similar dependencies in spoken sentence production (expressive knowledge). Results showed that learners (n=22) performed better with number than gender overall, but their brain responses to both features were qualitatively native-like (i.e., P600), even though gender was probed with nouns that do not provide strong distributional cues to gender. In addition, variability with gender agreement was better accounted for by lexical (as opposed to syntactic) aspects. Learners showed no advantage for comprehension over production. They also showed no systematic evidence of reliance on morphological defaults, although their online processing was sensitive to markedness in a native-like manner. Overall, these results suggest that there is facilitation for properties of the L2 that exist in the L1 and that markedness impacts L2 processing, but in a native-like manner. These results also speak against proposals arguing that adult L2ers have deficits at the level of the morphology or the syntax. Examining Morphological Variability in L2 Learners 3 Adult second language (L2) learners often exhibit variability in their use of inflectional morphology, even at high levels of proficiency (e.g., Franceschina, 2005; Gillon-Dowens, Vergara, Barber, and Carreiras, 2010; Grüter, Lew-Williams, and Fernald, 2012; Keating, 2009; Lardiere, 1998, McCarthy, 2008; Rossi, Kroll, and Dussias, 2014; Sabourin and Stowe, 2008; see White, 2007 for theoretical considerations). Morphological variability refers to a learner’s inconsistent use of obligatory inflectional morphology, as exemplified in (1), which presents elicited production data from an advanced L1-English L2-Spanish learner (McCarthy, 2008, p. 478): (1) a. está poniendo las tijeras en la mochila she’s putting the scissors in the-FEM backpack-FEM b. la mochila es negro the-FEM backpack-FEM is black-MASC In (1a-b), the learner correctly establishes gender agreement between the feminine noun mochila “backpack-FEM” and the determiner la “the-FEM”, but then shows incorrect inflection on the adjective negro “black-MASC”, which is used in the masculine (and, thus, fails to agree with its controller noun). A wealth of research has examined inflectional variability in L2 learners (e.g., Franceschina, 2005; Grüter et al., 2012; Lemhöfer, Schriefers, and Indefrey, 2014; López Prego and Gabriele, 2014; McCarthy, 2008; Montrul, Foote, and Perpiñán, 2008; Morgan-Short, Sanz, Steinhauer, and Ullman, 2010; Prévost and White, 2000; Renaud, 2012; White, Valenzuela, Kozlowska-MacGregor, and Leug, 2004), and some interesting generalizations have emerged from this literature. For example, inflectional errors tend to exhibit systematicity, with some error types occurring more frequently than others (e.g., Dewaele and Véronique, 2001; Franceschina, 2001; McCarthy, 2008; Montrul et al., Examining Morphological Variability in L2 Learners 4 2008; Sabourin, 2003; White et al., 2004). To account for this asymmetry, some authors have argued that L2ers resort to the use of morphological “defaults”, that is, underspecified forms that learners use in target-like contexts and overextend to incorrect ones (e.g., McCarthy, 2008; Montrul et al., 2008; Prévost and White, 2000; White et al., 2004). With respect to number and gender agreement in Spanish, the properties of interest herein, this would mean that learners incorrectly use singular and masculine forms in plural and feminine contexts, but the reverse pattern rarely occurs. The error in (1b), where the learner incorrectly uses masculine inflection in a feminine context constitutes a good example of potential reliance on default morphology. In addition, some morphosyntactic properties exhibit greater variability than others, even at the highest proficiency levels. For example, Franceschina (2005), López-Prego and Gabriele (2014), McCarthy (2008), Rossi et al. (2014), and White et al. (2004) all compared syntactic number and gender agreement in L2-Spanish by English-speaking learners at different proficiency levels, and found that number was relatively unproblematic across the proficiency spectrum (see also Gabriele, Fiorentino, and Alemán Bañón, 2013). In contrast, gender agreement showed more variability, among both advanced L2ers (e.g., López Prego and Gabriele, 2014; McCarthy, 2008; Rossi et al., 2014; but see White et al., 2004) and even near-native speakers (e.g., Franceschina, 2005). Since grammatical gender is not instantiated by these learners’ L1, some authors have claimed that inflectional variability is due to brain maturation effects specifically affecting novel L2 syntactic properties (e.g., Franceschina, 2005; Hawkins and Chan, 1997; Long, 2005; Sabourin, 2003; Tsimpli and Dimitrakopoulou, 2007). Recent proposals for the domain of grammatical gender (Grüter et al., 2012; Hopp, 2013), however, argue that variability with grammatical gender is more tied to aspects of lexical gender assignment (i.e., linking nouns to their appropriate gender classes at the Examining Morphological Variability in L2 Learners 5 level of mental representation). Along these lines, recent studies have shown that even L2ers whose L1 realizes gender exhibit variability with gender inflection due to weak knowledge of lexical gender (e.g., Lemhöfer et al., 2014), even at high levels of proficiency (e.g., Sabourin, 2003; Sabourin and Stowe, 2008; White et al., 2004). Finally, variability appears to emerge in some tasks more than others. Several studies have shown that learners usually perform better in tasks measuring comprehension (e.g., sentence-picture matching, written recognition task), relative to those examining oral production (e.g., Alarcón, 2011; Grüter et al., 2012; Montrul et al., 2008), and some authors have proposed that inflectional variability might be a production-specific phenomenon (Prévost and White, 2000; Rothman, 2007; White, 2011). However, as pointed out by Grüter et al. (2012), the difference between comprehension and production shows a confound with processing burden in many studies. Indeed, comprehension has often been examined via offline tasks (e.g., McCarthy, 2008; Montrul et al., 2008; White et al., 2014), while the very nature of spoken language production calls for online tasks, where the processing burden is higher, as learners must retrieve and articulate the words in real-time. Therefore, the observed performance differences between comprehension and production may well be related to task type, rather than differences between the receptive and expressive knowledge of morphology. The present paper is devoted to the study of morphological variability in adult L2 learners, with a focus on the central issues highlighted above. The properties of interest are number and gender agreement in L2-Spanish, with a novel emphasis on markedness relations, since it has been argued that underspecified features (i.e., defaults) correspond to unmarked ones (e.g., Harley and Ritter, 2002). We examine the extent to which L2 inflectional variability can be accounted for by (i) reliance on default morphology; (ii) Examining Morphological Variability in L2 Learners 6 the properties of the learners’ L1; and (iii) the type of knowledge tapped into (receptive vs. expressive) related to methodological design (comprehension vs. production). Number and Gender Agreement in Spanish Spanish nouns belong to one of two genders, masculine or feminine. Although neither gender value is associated with a unique marker (Harris, 1991), a clear regularity can be observed: 99.8% of nouns ending in –o are masculine and 96.3% of nouns ending in –a are feminine (Teschner and Russell, 1984). These transparent nouns make up approximately two thirds of the Spanish lexicon (Harris, 1991), suggesting that the – o and –a markers provide strong distributional cues to gender. However, the Spanish lexicon includes many nouns ending in vowel –e or a consonant, for which gender can be less reliably determined. These less transparent nouns are the focus of the present study. Several observations suggest that, in Spanish, feminine is marked for gender and masculine is underspecified (Battistella, 1990; Bonet, 1995; Harris, 1991). For example, when a genderless word (e.g., preposition para “for”) is modified by an agreement-bearing element (e.g., the indefinite adjective demasiado “too-many”), the latter must show masculine inflection (demasiados paras en ese párrafo “too-many-MASC fors-NO-GENDER in that paragraph”) (Harris, 1991). Likewise, when masculine and feminine nouns are conjoined, all agreement targets must also show masculine inflection. This suggests that masculine inflection is underspecified for gender, since it can appear with genderless elements and even feminine ones, but feminine forms are marked since they can only appear with feminine nouns. The Spanish number system distinguishes between singular and plural. Singular shows zero inflection, while plural is formed by suffixing –s or –es to the singular form Examining Morphological Variability in L2 Learners 7 (the root) (e.g., coche/coches “car/cars”, árbol/árboles “tree/trees”) (Saporta, 1965). This asymmetry with respect to the presence of overt inflection has been taken as evidence that plural forms are marked, relative to singular (e.g., Battistella, 1990). Additional evidence that singular and plural are asymmetrically represented is that singular has a broader syntactic distribution than plural. For example, the singular dative clitic le can be coindexed with a plural phrase (Julia lei teme [a las ratas]i “Julia CL-SG fears rat-PL”), but its plural counterpart les cannot be coindexed with a singular phrase. This suggests that singular forms are underspecified for number, since they can agree both with singular and plural phrases, but plural forms are marked, since they are restricted to plural elements. Theories on L2 Morphological Variability Different L2 theoretical models make competing claims regarding the locus and nature of L2 morphological variability. The “representational accounts” posit that L2 morphological variability stems from a representational deficit at the level of the syntax (e.g., Franceschina, 2005; Hawkins, 2001; Tsimpli and Dimitrakopoulou, 2007). Under these models, only syntactic properties of the L2 that exist in the learners’ L1 can be acquired to native-like levels, due to maturation. For novel properties, it is argued that L2ers use compensatory strategies. With respect to the acquisition of grammatical gender by speakers of gender-free languages, one potential strategy would be phonological rhyming between noun endings and inflectional forms (Hawkins, 2001; White et al., 2004). This position is well represented by the Interpretability Hypothesis (Tsimpli and Dimitrakopoulou, 2007), for which it is novel syntactic features (i.e., those which make no semantic contribution to the interpretation of a lexical item) that become Examining Morphological Variability in L2 Learners 8 inaccessible in adult L2 acquisition. For syntactic agreement, this would be the case with number and gender information on determiners and adjectives. In contrast, the “computational accounts” argue that the properties of the learner’s L1 do not constrain L2 acquisition, but rather that morphological variability is a corollary of performance limitations (e.g., Haznedar and Schwartz, 1997; Hopp, 2010; Prévost and White, 2000). This is the position adopted by the Missing Surface Inflection Hypothesis “MSIH” (Prévost and White, 2000) according to which, variability results from the difficulty associated with the retrieval of the appropriate inflectional forms and their mapping onto lexical items, particularly in oral production (White, 2011). The proponents of the MSIH offer the following analysis for the observation that L2ers often adopt defaults. They assume that features are fully specified in the syntax, but not in the morphology (Halle and Marantz, 1993; Harley and Noyer, 1999). In the morphology, singular and masculine are underspecified, whereas plural and feminine are marked (i.e., fully specified) (Bonet, 1995; Cowper, 2005; Harley, 1994; Harley and Ritter, 2002; Harris, 1991). For agreement to be successful, the features on lexical items must be compatible with those of the syntax. A perfect match is not required, but there can be no feature clash. For cases where the syntax (e.g., the Determiner Phrase) is specified as plural or feminine, the parser will select a plural or feminine form (i.e., fully specified in the morphology), as they provide a perfect match (e.g., luz roja “light-FEM red-FEM”). However, masculine or singular forms do not clash in this context, due to their lack of specification (e.g., luz rojo “light-FEM red-UNDERSPECIFIED”). For cases where the syntax is specified as singular or masculine, only underspecified forms can be inserted, since inflectional forms that are fully specified as masculine or singular are not available (e.g., coche rojo “car-MASC red-UNDERSPECIFIED”), and the insertion of plural or feminine forms would cause a feature clash (e.g., coche roja “car-MASC red-FEM”). The Examining Morphological Variability in L2 Learners 9 proponents of the MSIH argue that, although L2ers can acquire the full specification of features, they have trouble retrieving them in production, due to processing burden. In such cases, L2ers select a “good enough form” (i.e., an underspecified form or default) even if a better candidate is available. This yields the well-attested asymmetric pattern of errors in production, where learners are more likely to underspecify a feature, as in luz rojo (light-FEM red-UNDERSPECIFIED), than to produce a feature clash. This was the pattern observed by Prévost and White (2000) with respect to the acquisition of finite forms in adult L2 learners of French and German. Grüter et al. (2012) agree that inflectional variability with grammatical gender is tied to difficulty with lexical retrieval, but point to gender assignment (i.e., linking nouns to their gender classes) as the source of variability. The authors examined gender agreement in advanced L1-English L2-Spanish learners, and found that they were native-like in offline comprehension, but made errors of gender assignment in production and could not utilize gender predictively in online comprehension. Grüter et al. propose that the links between nouns and their abstract gender classes are weaker in L2ers. Consequently, L2ers have difficulty with the retrieval and use of gender information online. A subsequent study by Hopp (2013) looking at L1-English L2-German learners provides support for this proposal. Hopp found that only those L2ers who showed stable knowledge of lexical gender (i.e., those who assigned almost all nouns to their appropriate gender values) behaved like German native speakers in their ability to utilize gender information predictively. Taken together, these studies suggest that the quality of the learners’ lexical representations for gender accounts for variability with gender agreement. Following Hopp (2013), we will refer to this proposal as the Lexical Gender Learning Hypothesis. Examining Morphological Variability in L2 Learners 10 Finally, an alternative account of inflectional variability is provided by McCarthy (2008), who builds on the idea that variability is systematic and consists of the overuse of default morphology. McCarthy distinguishes between two types of errors, default/underspecification errors and feature clash errors. Default errors are cases where the syntax is fully specified as plural or feminine, but the learner uses an underspecified form on lexical items (i.e., singular, masculine). This is the case in (2a) and (2b) for number and gender, respectively: (2) a. las mochilas son *negra the backpack-FEM-PL are black-FEM-SG b. la mochila es *negro the backpack-FEM-SG is black-MASC-SG Feature clash errors show the opposite pattern; the syntax is fully specified as singular or masculine, but the lexical items are fully specified as plural or feminine. Examples are shown in (3a) and (3b) for number and gender, respectively: (3) a. el bolso es *negros the purse-MASC-SG is black-MASC-PL b. el bolso es *negra the purse-MASC-SG is black-FEM-SG The main tenet of McCarthy’s proposal (2008) is that L2ers’ errors mainly consist of default errors. Unlike the MSIH, however, McCarthy argues that variability is representational, and that overreliance on default morphology is not specific to production, but also emerges in comprehension. Her proposal also differs from other representational accounts in two ways. First, the deficit is located at the level of the morphology. That is, L2ers are assumed to be able to acquire all syntactic projections of the L2, but not the full specification of features in the morphology. Second, variability Examining Morphological Variability in L2 Learners 11 is not restricted to novel properties, but can also emerge for properties instantiated in the L1.