Session 4G

Usage-Based Theory of Language Acquisition proposes that acquisition of both L1 and L2 is affected by frequency of input at different levels of language. Type frequency of constructions in L2 input induces formation of abstract schemas in L2 speaker's mind (Bybee 2013, Ellis 2015: 50). Through the process of generalization, abstract constructions like "verb + object" may occur in cognitive representations of L2. However, in morphologically rich languages such as Lithuanian each object is marked by a grammatical case (accusative, genitive, dative or instrumental). Consequently, a schema "verb + object" must contain morphological information: "verb + object (acc/gen/dat/instr)". Moreover, this abstract schema in the acquisition of Lithuanian as L2 is also hardly possible, as the marking of object is lexically governed. Different groups of verbs govern different grammatical cases of objects. Nevertheless, these groups of verbs are not determined by any salient semantical information. Therefore, it let us believe that less abstract constructions such as "a particular verb + objacc / gen / dat / instr" are more common in Lithuanian as L2. If semiabstract constructions containing concrete verbs and abstract objects form cognitive morphosyntactic representations of Lithuanian as L2, then the frequency of input of a particular verb would affect the acquisition of object marking in Lithuanian as L2. This paper aims to answer the question if the frequency of concrete verbs is related to the correctness of object marking in Lithuanian as L2. In order to answer the question 78 Lithuanian as L2 speakers (age = 16–18 years, L1 = Russian) were tested. A written version of Elicited Oral Imitation Test (Erlam 2006) was designed to measure the correctness of object marking in Lithuanian as L2. The test contained 28 "verb + object" constructions. The frequency of each verb varied from 1 to 1350 uses in the "Frequency Dictionary of Written Lithuanian" (Utka 2008) (the dictionary is based on 1-million-word corpus). Constructions conflicting with participants' L1 were not involved in the test. Collocational structures were also avoided. The task of the participants of the study was to listen to a statement, to mark if they agree or disagree with it, and then to write down the statement they heard. 2 184 uses of "verb + object" constructions in Lithuanian as L2 were collected. The data were analyzed using Generalized Linear Mixed Effects Models. The frequency of verbs and different grammatical cases marking an object were treated as fixed effects, while items and participants, as random effects. The analysis shows that participants of the study had a higher probability to mark an object correctly when it was governed by a more frequent verb (? = 0.001, SE = 0.0005, z = 2.34, p = 0.19). Objects marked by genitive, which is the most frequently used grammatical case in Lithuanian (Rimkut? 2006: 42), were also used significantly more correctly in the test (? = 0.099, SE = 0.39, z = 2.53, p = 0.01). The analysis showed no significant effect of marking an object by other grammatical cases. References Bybee, J., 2013. Usage-based Theory and Exemplar Representations of Constructions. In: The Oxford Handbook of Construction Grammar, (eds.) Hoffman, Th., Trousdale, G. Oxford University Press, pp. 49–69. Ellis, N. C. 2015. Cognitive and Social Aspects of Learning from Usage. In Cadierno, T & Eskildsen, S. (Eds.), Usage-based perspectives on second language learning, pp. 49-73. Berlin: DeGruyter Mouton. Erlam, R. 2006. Elicited Imitation as a Measure of L2 Implicit Knowledge: An Empirical Validation Study. Applied Linguistics, 27(3), 464–491. Rimkut?, E., 2006. Dabartin?s lietuvi? kalbos gramatini? form? vartosena morfologiškai anotuotame tekstyne. Lituanistica, 2(66), pp. 34–55.

In recent years several SLA studies have discussed the construct of morphological complexity, proposing theoretical and operational definitions and measuring its development over time or its variation across tasks (e.g Brezina & Pallotti 2019; De Clercq & Housen, 2019). Most of these works assessed morphological complexity in terms of the range and variety of morphological processes displayed in learners' texts, without clearly differentiating between regular and irregular phenomena. For many languages, including most of the Indo-European family, the regular/irregular distinction corresponds to a distinction between predictable affixation processes and relatively unpredictable stem alternations. The former may be said to pertain to grammar proper, as they are generalizable and may be productively applied to new words and even non-words; the latter belong to an intermediate level between grammar and the lexicon, as some of these patterns do show some regularity ("minor rules", Lakoff 1970; "sub-regularities", Marquis & Royle, 2019), but these apply to specific (groups of) lexemes, and hence may also be considered as properties of single words (Wunderlich, 1996). Research on L1 and L2 acquisition, or on computational models like neural networks, has confirmed that regularity / irregularity, rather than being a dichotomy, are more adequately represented as a continuum: regular forms do exhibit frequency effects, which proves that they are to some extent stored in long-term memory, while irregulars are sometimes over-extended, showing that they may be analyzed and productively processed (Baayen et al, 2018; Krause et al, 2015; Marquis & Royle, 2019; Ramscar et al, 2018). Thus, although both theory and research show that it is difficult to draw a sharp line between regular and irregular morphological processes, the distinction may still be relevant on a descriptive level, to track how these processes develop over time and differentially contribute to morphological complexity. The presentation will address these issues, firstly by providing a conceptual and theoretical overview of how the regular/irregular distinction has been treated in theoretical linguistics, with special attention to the verbal domain. This will be followed by a discussion of how the constructs of regular and irregular morphology may be operationalized in the analysis of linguistic texts produced by native and non-native speakers. In particular, three options will be examined: 1) identifying regular, stem-independent morphology; 2) treating stems as the result of minor stem-formation rules (e.g. that giving rise to the find/found, grind/ground, bind/bound alternations in English); 3) treating stems as properties of individual lexical entries (in this case bound, ground and found would be treated as three different lexical stems). These operationalizations will be empirically tested by re-analyzing the data employed by Brezina & Pallotti (2019), who treated morphological complexity as a unitary phenomenon, i.e., without distinguishing between regular and irregular processes. Morphological complexity will thus be computed on essays written by Dutch learners of L2 Italian and Italian learners of L2 English of varying proficiency levels, with their corresponding control groups of native speakers (taken from the CALC and the ICLE corpora). Different operationalizations of the construct will be proposed, seeing it as a unitary construct with no internal differentiations, or following the different operationalizations of 'irregularity' given above. Results will be discussed to determine the strengths and limitations of different approaches, in order to better qualify the notion of morphological complexity in SLA research.

The placement of clitic pronouns is known as a problematic yet interesting aspect in the acquisition of European Portuguese (EP). As has been well described in the literature (Madeira 1993; Duarte and Matos 2000, etc.), EP presents quite unique clitic placement patterns, which are not related to finiteness, unlike some other Romance languages, but are determined by different syntactic conditions. Some authors (e.g. Martins 1994) attribute these patterns to specific properties of the functional domain in EP and hence the acquisition of clitic placement is dependent on the acquisition of these functional properties. Two influential proposals have been put forward regarding the possibility of acquiring new functional properties in an L2: the Full Transfer/Full Access (FT/FA) hypothesis (Schwartz and Sprouse, 1996), which predicts that they may be successfully acquired whether they are present in the L1 or not, and the Interpretability Hypothesis (IH) (e.g. Tsimpli and Mastropavlou, 2007), which predicts that (uninterpretable) features not present in the L1 will not be acquired. Previous studies on the L1 acquisition of clitics (Varlokosta et al. 2015; Costa & Lobo 2009, 2013; Costa, Fiéis & Lobo 2015, etc.) show that, differently from what is observed in most of other Romance languages, where clitic placement is acquired early, in EP, children tend to display deviant patterns till quite late. Some of these studies suggest a fixed developmental path in the acquisition of these patterns, marked by the initial generalization of enclisis and gradual acquisition of the different contexts for proclisis. In L2 EP, similar observations have been made (see Rosário 2005; Madeira, Crispim & Xavier 2006; Madeira & Xavier 2009, etc.), although there have been few systematic studies exploring this issue. In this talk we present an empirical study on the acquisition of clitic placement in EP by adult Chinese learners which aims to answer the following questions: (1) Do Chinese learners of EP generalize enclisis? (2) Is it more difficult for them to acquire proclisis patterns than enclisis? (3) Are there asymmetries in the acquisition of the different proclitic contexts? (4) If so, what is the acquisition order? As their L1 does not have clitic pronouns, development of the syntactic properties related to clitic placement in EP will only occur if learners are able to acquire the relevant new functional properties - predicted to be possible by FT/FA but not by the IH. Two tasks were used: an untimed acceptability judgement task (72 test items and 36 distractors), as well as a centered auto-paced reading task (20 test items and 16 distractors). 58 Chinese students of Portuguese from a university in Shanghai and a control group of 26 native speakers of EP participated in the study. In both two tasks, the participants were asked to judge the sentences presented on the screen, on a scale of 1 (completely unacceptable) to 5 (perfectly acceptable). The preliminary results we obtained suggest that, in the acceptability judgement task, the Chinese participants tend to follow a certain development path in the acquisition of clitic placement, with better performance in enclisis contexts, more difficulty and asymmetries in the acquisition of the different proclitic contexts, as observed in previous works on L1 EP and FL EP with other L1s, but display signs of development of the syntactic properties related to clitic placement. The centered auto-paced reading task seems to show, instead, some instability in the “acquired” knowledge. We will continue this study with some further statistical analysis.

The video of this presentation has been removed at the request of the author.
Heritage language acquisition, normally the weaker language of early bilinguals, has been oftentimes found to be more similar to adult L2 than child L1 acquisition, especially for morphosyntax (Montrul, 2008, 2010, 2016). This heritage-L2 similarity calls into question the advantages typically associated with early age of exposure, namely the critical period, given that heritage speakers are in some cases early bilinguals, exposed to their family language at an age and in measures similar to monolingual children. Despite the reported divergence between heritage and L1 acquisition, an interesting divide exists in the literature to suggest that early exposure can lead to more monolingual-like performance: while L2 and heritage speakers are similar in their knowledge of morphological properties (Montrul, Foote, & Perpiñán, 2008; Polinsky, 2008; Silva-Corvalán, 2014), in some areas of syntax, heritage speaker knowledge resembles L1 more than L2 speakers (Montrul, 2010; Chung, 2013, Håkansson, 1995). These studies, however, have not sufficiently considered the effects of syntactic complexity on knowledge of morphology. In this talk, I report on the effects of age of exposure and syntactic complexity by comparing early heritage, late L2, and L1 speakers on knowledge of Italian accusative clitics in three verb structures, lexical, functional, and quasi-functional, where clitics are surmised to be abstractly represented in either a lexical or functional projection (Cardinaletti & Schlonky, 2004): Under the assumption that lexical projections are less syntactically complex than functional projections, it was predicted that: (1) lexical and quasi-functional structures would be represented more strongly than the functional (syntactic prediction); (2) differences in complexity between lexical and functional domains would lead to lower accuracy in use of clitics in the latter case (morphological prediction). It was also predicted that heritage speakers would perform more similarly to L2ers with respect to morphology (i.e. clitic realization) than syntax (i.e. strength of syntactic representation of the three structures). An oral structural priming task was used to prime the strength of structure representation and measure accuracy in use of clitic forms (i.e. clitic realization). In addition, a timed GJT was used to elicit explicit judgments on clitic form in each of the three structures. Results confirmed the prediction that while heritage and L1 abstract representation of clitic structures approximate each other, heritage morphological knowledge of clitics aligns with the L2ers', suggesting early exposure has tangible effects only on syntactic knowledge. In turn, syntactic complexity affected the representation of the three clitic structures in the predicted manner, with lexical and quasi-functional structures showing stronger priming than the functional. Complexity, however, was inconsequential to knowledge of morphological forms in monolingual and bilingual speakers as evidenced by both the priming and GJ task. The key difference in morphological knowledge between the bilingual and monolingual speakers amounts to clitic omission which I attribute to interface vulnerability, namely differences in the integration of syntactic and pragmatic knowledge in clitic left-dislocated structures.