Polarity-item anything in L3 English: Where does transfer come from when the L1 is Catalan and the L2 is Spanish?

Abstract

This study explores the source of transfer in third language (L3) English by two distinct groups of Catalan–Spanish bilinguals, simultaneous bilinguals and late bilinguals. Our study addresses two research questions: (1) Does transfer come from the first language (L1), the second language (L2), or both? and (2) Does age of acquisition of the L2 affect how transfer occurs? We compare beginner and advanced English speakers from both L3 groups with beginner and advanced L1-Spanish L2-English speakers, and find that, on an acceptablity judgment task that investigates knowledge of the distribution of polarity item anything, the two L3 groups demonstrate a different response pattern from the L2 group. The results suggest that both L3 groups transfer from Catalan, and not from their L2, Spanish. Additionally, the cross-sectional nature of the study shows that negative transfer from the initial stages of acquisition is overcome to different extents by the L3 vs. the L2 groups. We conclude that the results show strong evidence against the L2 status factor (Bardel and Falk, 2007, 2012) and the cumulative enhancement (Flynn et al., 2004) models of L3 acquisition, while they can be accounted for by the typological primacy model (Rothman, 2010, 2011, 2015), although other models that predict L1 transfer in L3 acquisition are not ruled out. Further, our findings show no effect of age of acquisition of the L2 on L3 development.

On the status of transfer in adult third language acquisition of early bilinguals

“…To the best of our knowledge, this is the first study to offer results across three languages, for the same speakers, from a wide array of unrelated, novel domains of morphosyntax, four in total (Differential Object Marking, Determiner + Proper nouns, VSO word order, causative structures). These four properties come together with data from two other domains of grammar reported on in other work, collected at the same time in the same population: definiteness effects [ 11 ] and negative polarity [ 9 , 12 ]. As a result, we rely on evidence from six domains of grammar in the same learners, tested after eight weeks of a purposefully designed English language course, to adjudicate between claims of full versus property-by-property transfer in L3/L n morphosyntax.…”

“…In order to focus on an early stage of L3 development, we designed a two-month course for true ab initio learners of English [see 9 , 12 ]. Designing and administering this course allowed us to (a) narrow down our population sampling to people without previous experience with naturalistic exposure or formal instruction in English; and (b) ensure that they learned the minimum vocabulary required to undertake all experimental tasks administered at the end of the course, while receiving virtually no exposure to the critical grammatical properties.…”

“…Each item was presented individually in the middle of the screen. The AJT was conducted within the framework of a larger study [ 12 ], being one of four tasks participants took in each experimental session.…”

“…As argued in previous work with the population in this study (Spanish/Catalan bilinguals with English as an L3), lexical similarity is likely to favor neither of the two previous languages, as the vast majority of lexical overlap will come from the presence of words of Romance (from Middle French) and Greek origin in English, which will have cognates in both Catalan and Spanish. However, the next level, phonology, is argued to offer sufficient information to adjudicate between the two, as Catalan exhibits more proximity to English in this domain than does Spanish [e.g., 9 , 11 , 22 , 23 ]. First, at the segmental level Spanish has a very reduced set of word-final legitimate consonants (/r/, /s/, /d/, /l/, /θ/, /n/); Catalan, in contrast, is closer to English in also admitting in this context palatal consonants, bilabial consonants or dental consonants.…”

“…As a result, we rely on evidence from six domains of grammar in the same learners, tested after eight weeks of a purposefully designed English language course, to adjudicate between claims of full versus property-by-property transfer in L3/L n morphosyntax. This is an improvement from the vast majority of previous studies, which have looked at single domains of grammar [ 12 , 13 for review] and are thus unable to estimate the extent of transfer (full vs. partial) with any degree of confidence.…”